{ "best_metric": null, "best_model_checkpoint": null, "epoch": 9.458594019409823, "eval_steps": 500, "global_step": 24000, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.0, "learning_rate": 5.249343832020997e-10, "loss": 0.328, "step": 1 }, { "epoch": 0.0, "learning_rate": 1.0498687664041995e-09, "loss": 0.3429, "step": 2 }, { "epoch": 0.0, "learning_rate": 1.5748031496062992e-09, "loss": 0.3305, "step": 3 }, { "epoch": 0.0, "learning_rate": 2.099737532808399e-09, "loss": 0.3301, "step": 4 }, { "epoch": 0.0, "learning_rate": 2.6246719160104985e-09, "loss": 0.3391, "step": 5 }, { "epoch": 0.0, "learning_rate": 3.1496062992125984e-09, "loss": 0.3493, "step": 6 }, { "epoch": 0.0, "learning_rate": 3.674540682414698e-09, "loss": 0.3361, "step": 7 }, { "epoch": 0.0, "learning_rate": 4.199475065616798e-09, "loss": 0.328, "step": 8 }, { "epoch": 0.0, "learning_rate": 4.724409448818897e-09, "loss": 0.3159, "step": 9 }, { "epoch": 0.0, "learning_rate": 5.249343832020997e-09, "loss": 0.3343, "step": 10 }, { "epoch": 0.0, "learning_rate": 5.7742782152230965e-09, "loss": 0.3375, "step": 11 }, { "epoch": 0.0, "learning_rate": 6.299212598425197e-09, "loss": 0.3329, "step": 12 }, { "epoch": 0.01, "learning_rate": 6.824146981627296e-09, "loss": 0.3358, "step": 13 }, { "epoch": 0.01, "learning_rate": 7.349081364829396e-09, "loss": 0.33, "step": 14 }, { "epoch": 0.01, "learning_rate": 7.874015748031496e-09, "loss": 0.3282, "step": 15 }, { "epoch": 0.01, "learning_rate": 8.398950131233596e-09, "loss": 0.3306, "step": 16 }, { "epoch": 0.01, "learning_rate": 8.923884514435695e-09, "loss": 0.343, "step": 17 }, { "epoch": 0.01, "learning_rate": 9.448818897637795e-09, "loss": 0.3231, "step": 18 }, { "epoch": 0.01, "learning_rate": 9.973753280839894e-09, "loss": 0.3389, "step": 19 }, { "epoch": 0.01, "learning_rate": 1.0498687664041994e-08, "loss": 0.3125, "step": 20 }, { "epoch": 0.01, "learning_rate": 1.1023622047244093e-08, "loss": 0.3273, "step": 21 }, { "epoch": 0.01, "learning_rate": 1.1548556430446193e-08, "loss": 0.3398, "step": 22 }, { "epoch": 0.01, "learning_rate": 1.2073490813648292e-08, "loss": 0.3283, "step": 23 }, { "epoch": 0.01, "learning_rate": 1.2598425196850394e-08, "loss": 0.3346, "step": 24 }, { "epoch": 0.01, "learning_rate": 1.3123359580052493e-08, "loss": 0.3442, "step": 25 }, { "epoch": 0.01, "learning_rate": 1.3648293963254593e-08, "loss": 0.3279, "step": 26 }, { "epoch": 0.01, "learning_rate": 1.4173228346456692e-08, "loss": 0.3264, "step": 27 }, { "epoch": 0.01, "learning_rate": 1.4698162729658792e-08, "loss": 0.3399, "step": 28 }, { "epoch": 0.01, "learning_rate": 1.522309711286089e-08, "loss": 0.3372, "step": 29 }, { "epoch": 0.01, "learning_rate": 1.5748031496062992e-08, "loss": 0.3356, "step": 30 }, { "epoch": 0.01, "learning_rate": 1.627296587926509e-08, "loss": 0.3506, "step": 31 }, { "epoch": 0.01, "learning_rate": 1.679790026246719e-08, "loss": 0.3345, "step": 32 }, { "epoch": 0.01, "learning_rate": 1.732283464566929e-08, "loss": 0.3237, "step": 33 }, { "epoch": 0.01, "learning_rate": 1.784776902887139e-08, "loss": 0.3319, "step": 34 }, { "epoch": 0.01, "learning_rate": 1.837270341207349e-08, "loss": 0.338, "step": 35 }, { "epoch": 0.01, "learning_rate": 1.889763779527559e-08, "loss": 0.3546, "step": 36 }, { "epoch": 0.01, "learning_rate": 1.942257217847769e-08, "loss": 0.341, "step": 37 }, { "epoch": 0.01, "learning_rate": 1.994750656167979e-08, "loss": 0.3478, "step": 38 }, { "epoch": 0.02, "learning_rate": 2.047244094488189e-08, "loss": 0.3402, "step": 39 }, { "epoch": 0.02, "learning_rate": 2.0997375328083988e-08, "loss": 0.3289, "step": 40 }, { "epoch": 0.02, "learning_rate": 2.152230971128609e-08, "loss": 0.3173, "step": 41 }, { "epoch": 0.02, "learning_rate": 2.2047244094488187e-08, "loss": 0.3461, "step": 42 }, { "epoch": 0.02, "learning_rate": 2.2572178477690288e-08, "loss": 0.3509, "step": 43 }, { "epoch": 0.02, "learning_rate": 2.3097112860892386e-08, "loss": 0.3352, "step": 44 }, { "epoch": 0.02, "learning_rate": 2.3622047244094487e-08, "loss": 0.3327, "step": 45 }, { "epoch": 0.02, "learning_rate": 2.4146981627296585e-08, "loss": 0.3473, "step": 46 }, { "epoch": 0.02, "learning_rate": 2.4671916010498686e-08, "loss": 0.3199, "step": 47 }, { "epoch": 0.02, "learning_rate": 2.5196850393700787e-08, "loss": 0.3244, "step": 48 }, { "epoch": 0.02, "learning_rate": 2.572178477690289e-08, "loss": 0.343, "step": 49 }, { "epoch": 0.02, "learning_rate": 2.6246719160104986e-08, "loss": 0.336, "step": 50 }, { "epoch": 0.02, "learning_rate": 2.6771653543307087e-08, "loss": 0.3172, "step": 51 }, { "epoch": 0.02, "learning_rate": 2.7296587926509185e-08, "loss": 0.3213, "step": 52 }, { "epoch": 0.02, "learning_rate": 2.7821522309711287e-08, "loss": 0.3418, "step": 53 }, { "epoch": 0.02, "learning_rate": 2.8346456692913384e-08, "loss": 0.3433, "step": 54 }, { "epoch": 0.02, "learning_rate": 2.8871391076115486e-08, "loss": 0.3368, "step": 55 }, { "epoch": 0.02, "learning_rate": 2.9396325459317583e-08, "loss": 0.3347, "step": 56 }, { "epoch": 0.02, "learning_rate": 2.9921259842519685e-08, "loss": 0.3352, "step": 57 }, { "epoch": 0.02, "learning_rate": 3.044619422572178e-08, "loss": 0.3394, "step": 58 }, { "epoch": 0.02, "learning_rate": 3.097112860892389e-08, "loss": 0.3351, "step": 59 }, { "epoch": 0.02, "learning_rate": 3.1496062992125985e-08, "loss": 0.3381, "step": 60 }, { "epoch": 0.02, "learning_rate": 3.202099737532808e-08, "loss": 0.3157, "step": 61 }, { "epoch": 0.02, "learning_rate": 3.254593175853018e-08, "loss": 0.3357, "step": 62 }, { "epoch": 0.02, "learning_rate": 3.3070866141732285e-08, "loss": 0.3172, "step": 63 }, { "epoch": 0.03, "learning_rate": 3.359580052493438e-08, "loss": 0.3312, "step": 64 }, { "epoch": 0.03, "learning_rate": 3.412073490813648e-08, "loss": 0.3136, "step": 65 }, { "epoch": 0.03, "learning_rate": 3.464566929133858e-08, "loss": 0.3084, "step": 66 }, { "epoch": 0.03, "learning_rate": 3.517060367454068e-08, "loss": 0.3138, "step": 67 }, { "epoch": 0.03, "learning_rate": 3.569553805774278e-08, "loss": 0.3176, "step": 68 }, { "epoch": 0.03, "learning_rate": 3.6220472440944886e-08, "loss": 0.3313, "step": 69 }, { "epoch": 0.03, "learning_rate": 3.674540682414698e-08, "loss": 0.3269, "step": 70 }, { "epoch": 0.03, "learning_rate": 3.727034120734908e-08, "loss": 0.3097, "step": 71 }, { "epoch": 0.03, "learning_rate": 3.779527559055118e-08, "loss": 0.3217, "step": 72 }, { "epoch": 0.03, "learning_rate": 3.8320209973753284e-08, "loss": 0.3432, "step": 73 }, { "epoch": 0.03, "learning_rate": 3.884514435695538e-08, "loss": 0.3239, "step": 74 }, { "epoch": 0.03, "learning_rate": 3.937007874015748e-08, "loss": 0.3063, "step": 75 }, { "epoch": 0.03, "learning_rate": 3.989501312335958e-08, "loss": 0.3029, "step": 76 }, { "epoch": 0.03, "learning_rate": 4.041994750656168e-08, "loss": 0.3036, "step": 77 }, { "epoch": 0.03, "learning_rate": 4.094488188976378e-08, "loss": 0.3019, "step": 78 }, { "epoch": 0.03, "learning_rate": 4.146981627296588e-08, "loss": 0.3133, "step": 79 }, { "epoch": 0.03, "learning_rate": 4.1994750656167975e-08, "loss": 0.317, "step": 80 }, { "epoch": 0.03, "learning_rate": 4.251968503937008e-08, "loss": 0.3152, "step": 81 }, { "epoch": 0.03, "learning_rate": 4.304461942257218e-08, "loss": 0.3137, "step": 82 }, { "epoch": 0.03, "learning_rate": 4.3569553805774276e-08, "loss": 0.3325, "step": 83 }, { "epoch": 0.03, "learning_rate": 4.4094488188976374e-08, "loss": 0.3223, "step": 84 }, { "epoch": 0.03, "learning_rate": 4.461942257217848e-08, "loss": 0.313, "step": 85 }, { "epoch": 0.03, "learning_rate": 4.5144356955380576e-08, "loss": 0.313, "step": 86 }, { "epoch": 0.03, "learning_rate": 4.566929133858268e-08, "loss": 0.2931, "step": 87 }, { "epoch": 0.03, "learning_rate": 4.619422572178477e-08, "loss": 0.2855, "step": 88 }, { "epoch": 0.04, "learning_rate": 4.6719160104986876e-08, "loss": 0.2789, "step": 89 }, { "epoch": 0.04, "learning_rate": 4.7244094488188974e-08, "loss": 0.2728, "step": 90 }, { "epoch": 0.04, "learning_rate": 4.776902887139108e-08, "loss": 0.2838, "step": 91 }, { "epoch": 0.04, "learning_rate": 4.829396325459317e-08, "loss": 0.2965, "step": 92 }, { "epoch": 0.04, "learning_rate": 4.8818897637795274e-08, "loss": 0.2773, "step": 93 }, { "epoch": 0.04, "learning_rate": 4.934383202099737e-08, "loss": 0.2754, "step": 94 }, { "epoch": 0.04, "learning_rate": 4.9868766404199477e-08, "loss": 0.2672, "step": 95 }, { "epoch": 0.04, "learning_rate": 5.0393700787401574e-08, "loss": 0.2882, "step": 96 }, { "epoch": 0.04, "learning_rate": 5.091863517060367e-08, "loss": 0.2816, "step": 97 }, { "epoch": 0.04, "learning_rate": 5.144356955380578e-08, "loss": 0.3082, "step": 98 }, { "epoch": 0.04, "learning_rate": 5.196850393700787e-08, "loss": 0.2691, "step": 99 }, { "epoch": 0.04, "learning_rate": 5.249343832020997e-08, "loss": 0.2824, "step": 100 }, { "epoch": 0.04, "learning_rate": 5.301837270341207e-08, "loss": 0.2753, "step": 101 }, { "epoch": 0.04, "learning_rate": 5.3543307086614175e-08, "loss": 0.2902, "step": 102 }, { "epoch": 0.04, "learning_rate": 5.4068241469816266e-08, "loss": 0.2828, "step": 103 }, { "epoch": 0.04, "learning_rate": 5.459317585301837e-08, "loss": 0.2657, "step": 104 }, { "epoch": 0.04, "learning_rate": 5.5118110236220475e-08, "loss": 0.2778, "step": 105 }, { "epoch": 0.04, "learning_rate": 5.564304461942257e-08, "loss": 0.2619, "step": 106 }, { "epoch": 0.04, "learning_rate": 5.6167979002624664e-08, "loss": 0.2782, "step": 107 }, { "epoch": 0.04, "learning_rate": 5.669291338582677e-08, "loss": 0.2843, "step": 108 }, { "epoch": 0.04, "learning_rate": 5.721784776902887e-08, "loss": 0.2549, "step": 109 }, { "epoch": 0.04, "learning_rate": 5.774278215223097e-08, "loss": 0.2712, "step": 110 }, { "epoch": 0.04, "learning_rate": 5.826771653543306e-08, "loss": 0.265, "step": 111 }, { "epoch": 0.04, "learning_rate": 5.879265091863517e-08, "loss": 0.2654, "step": 112 }, { "epoch": 0.04, "learning_rate": 5.931758530183727e-08, "loss": 0.2566, "step": 113 }, { "epoch": 0.04, "learning_rate": 5.984251968503937e-08, "loss": 0.2702, "step": 114 }, { "epoch": 0.05, "learning_rate": 6.036745406824147e-08, "loss": 0.2693, "step": 115 }, { "epoch": 0.05, "learning_rate": 6.089238845144356e-08, "loss": 0.2521, "step": 116 }, { "epoch": 0.05, "learning_rate": 6.141732283464566e-08, "loss": 0.2444, "step": 117 }, { "epoch": 0.05, "learning_rate": 6.194225721784777e-08, "loss": 0.2563, "step": 118 }, { "epoch": 0.05, "learning_rate": 6.246719160104986e-08, "loss": 0.2491, "step": 119 }, { "epoch": 0.05, "learning_rate": 6.299212598425197e-08, "loss": 0.2166, "step": 120 }, { "epoch": 0.05, "learning_rate": 6.351706036745407e-08, "loss": 0.2287, "step": 121 }, { "epoch": 0.05, "learning_rate": 6.404199475065617e-08, "loss": 0.2266, "step": 122 }, { "epoch": 0.05, "learning_rate": 6.456692913385826e-08, "loss": 0.2141, "step": 123 }, { "epoch": 0.05, "learning_rate": 6.509186351706036e-08, "loss": 0.2066, "step": 124 }, { "epoch": 0.05, "learning_rate": 6.561679790026247e-08, "loss": 0.2076, "step": 125 }, { "epoch": 0.05, "learning_rate": 6.614173228346457e-08, "loss": 0.1998, "step": 126 }, { "epoch": 0.05, "learning_rate": 6.666666666666665e-08, "loss": 0.2045, "step": 127 }, { "epoch": 0.05, "learning_rate": 6.719160104986877e-08, "loss": 0.1984, "step": 128 }, { "epoch": 0.05, "learning_rate": 6.771653543307086e-08, "loss": 0.2141, "step": 129 }, { "epoch": 0.05, "learning_rate": 6.824146981627296e-08, "loss": 0.2061, "step": 130 }, { "epoch": 0.05, "learning_rate": 6.876640419947506e-08, "loss": 0.19, "step": 131 }, { "epoch": 0.05, "learning_rate": 6.929133858267716e-08, "loss": 0.1986, "step": 132 }, { "epoch": 0.05, "learning_rate": 6.981627296587927e-08, "loss": 0.198, "step": 133 }, { "epoch": 0.05, "learning_rate": 7.034120734908137e-08, "loss": 0.2019, "step": 134 }, { "epoch": 0.05, "learning_rate": 7.086614173228345e-08, "loss": 0.2052, "step": 135 }, { "epoch": 0.05, "learning_rate": 7.139107611548556e-08, "loss": 0.1938, "step": 136 }, { "epoch": 0.05, "learning_rate": 7.191601049868766e-08, "loss": 0.2062, "step": 137 }, { "epoch": 0.05, "learning_rate": 7.244094488188977e-08, "loss": 0.1985, "step": 138 }, { "epoch": 0.05, "learning_rate": 7.296587926509186e-08, "loss": 0.1907, "step": 139 }, { "epoch": 0.06, "learning_rate": 7.349081364829395e-08, "loss": 0.1858, "step": 140 }, { "epoch": 0.06, "learning_rate": 7.401574803149606e-08, "loss": 0.1865, "step": 141 }, { "epoch": 0.06, "learning_rate": 7.454068241469816e-08, "loss": 0.1825, "step": 142 }, { "epoch": 0.06, "learning_rate": 7.506561679790026e-08, "loss": 0.1781, "step": 143 }, { "epoch": 0.06, "learning_rate": 7.559055118110236e-08, "loss": 0.1761, "step": 144 }, { "epoch": 0.06, "learning_rate": 7.611548556430446e-08, "loss": 0.1812, "step": 145 }, { "epoch": 0.06, "learning_rate": 7.664041994750657e-08, "loss": 0.1795, "step": 146 }, { "epoch": 0.06, "learning_rate": 7.716535433070865e-08, "loss": 0.1687, "step": 147 }, { "epoch": 0.06, "learning_rate": 7.769028871391076e-08, "loss": 0.1778, "step": 148 }, { "epoch": 0.06, "learning_rate": 7.821522309711286e-08, "loss": 0.1889, "step": 149 }, { "epoch": 0.06, "learning_rate": 7.874015748031496e-08, "loss": 0.1604, "step": 150 }, { "epoch": 0.06, "learning_rate": 7.926509186351706e-08, "loss": 0.1848, "step": 151 }, { "epoch": 0.06, "learning_rate": 7.979002624671915e-08, "loss": 0.1763, "step": 152 }, { "epoch": 0.06, "learning_rate": 8.031496062992125e-08, "loss": 0.1813, "step": 153 }, { "epoch": 0.06, "learning_rate": 8.083989501312336e-08, "loss": 0.1692, "step": 154 }, { "epoch": 0.06, "learning_rate": 8.136482939632545e-08, "loss": 0.169, "step": 155 }, { "epoch": 0.06, "learning_rate": 8.188976377952756e-08, "loss": 0.169, "step": 156 }, { "epoch": 0.06, "learning_rate": 8.241469816272966e-08, "loss": 0.1663, "step": 157 }, { "epoch": 0.06, "learning_rate": 8.293963254593176e-08, "loss": 0.1594, "step": 158 }, { "epoch": 0.06, "learning_rate": 8.346456692913385e-08, "loss": 0.1661, "step": 159 }, { "epoch": 0.06, "learning_rate": 8.398950131233595e-08, "loss": 0.1641, "step": 160 }, { "epoch": 0.06, "learning_rate": 8.451443569553806e-08, "loss": 0.1566, "step": 161 }, { "epoch": 0.06, "learning_rate": 8.503937007874016e-08, "loss": 0.161, "step": 162 }, { "epoch": 0.06, "learning_rate": 8.556430446194224e-08, "loss": 0.1569, "step": 163 }, { "epoch": 0.06, "learning_rate": 8.608923884514436e-08, "loss": 0.1543, "step": 164 }, { "epoch": 0.07, "learning_rate": 8.661417322834645e-08, "loss": 0.1518, "step": 165 }, { "epoch": 0.07, "learning_rate": 8.713910761154855e-08, "loss": 0.1532, "step": 166 }, { "epoch": 0.07, "learning_rate": 8.766404199475065e-08, "loss": 0.1462, "step": 167 }, { "epoch": 0.07, "learning_rate": 8.818897637795275e-08, "loss": 0.1608, "step": 168 }, { "epoch": 0.07, "learning_rate": 8.871391076115486e-08, "loss": 0.1382, "step": 169 }, { "epoch": 0.07, "learning_rate": 8.923884514435696e-08, "loss": 0.1449, "step": 170 }, { "epoch": 0.07, "learning_rate": 8.976377952755904e-08, "loss": 0.1516, "step": 171 }, { "epoch": 0.07, "learning_rate": 9.028871391076115e-08, "loss": 0.1413, "step": 172 }, { "epoch": 0.07, "learning_rate": 9.081364829396325e-08, "loss": 0.1516, "step": 173 }, { "epoch": 0.07, "learning_rate": 9.133858267716536e-08, "loss": 0.1422, "step": 174 }, { "epoch": 0.07, "learning_rate": 9.186351706036745e-08, "loss": 0.149, "step": 175 }, { "epoch": 0.07, "learning_rate": 9.238845144356954e-08, "loss": 0.1541, "step": 176 }, { "epoch": 0.07, "learning_rate": 9.291338582677165e-08, "loss": 0.1391, "step": 177 }, { "epoch": 0.07, "learning_rate": 9.343832020997375e-08, "loss": 0.149, "step": 178 }, { "epoch": 0.07, "learning_rate": 9.396325459317585e-08, "loss": 0.1357, "step": 179 }, { "epoch": 0.07, "learning_rate": 9.448818897637795e-08, "loss": 0.1456, "step": 180 }, { "epoch": 0.07, "learning_rate": 9.501312335958005e-08, "loss": 0.151, "step": 181 }, { "epoch": 0.07, "learning_rate": 9.553805774278216e-08, "loss": 0.1333, "step": 182 }, { "epoch": 0.07, "learning_rate": 9.606299212598424e-08, "loss": 0.1369, "step": 183 }, { "epoch": 0.07, "learning_rate": 9.658792650918634e-08, "loss": 0.1456, "step": 184 }, { "epoch": 0.07, "learning_rate": 9.711286089238845e-08, "loss": 0.1351, "step": 185 }, { "epoch": 0.07, "learning_rate": 9.763779527559055e-08, "loss": 0.1393, "step": 186 }, { "epoch": 0.07, "learning_rate": 9.816272965879265e-08, "loss": 0.1356, "step": 187 }, { "epoch": 0.07, "learning_rate": 9.868766404199474e-08, "loss": 0.126, "step": 188 }, { "epoch": 0.07, "learning_rate": 9.921259842519684e-08, "loss": 0.1299, "step": 189 }, { "epoch": 0.07, "learning_rate": 9.973753280839895e-08, "loss": 0.1314, "step": 190 }, { "epoch": 0.08, "learning_rate": 1.0026246719160105e-07, "loss": 0.1258, "step": 191 }, { "epoch": 0.08, "learning_rate": 1.0078740157480315e-07, "loss": 0.1188, "step": 192 }, { "epoch": 0.08, "learning_rate": 1.0131233595800523e-07, "loss": 0.1417, "step": 193 }, { "epoch": 0.08, "learning_rate": 1.0183727034120734e-07, "loss": 0.1243, "step": 194 }, { "epoch": 0.08, "learning_rate": 1.0236220472440944e-07, "loss": 0.1257, "step": 195 }, { "epoch": 0.08, "learning_rate": 1.0288713910761155e-07, "loss": 0.1201, "step": 196 }, { "epoch": 0.08, "learning_rate": 1.0341207349081365e-07, "loss": 0.1242, "step": 197 }, { "epoch": 0.08, "learning_rate": 1.0393700787401574e-07, "loss": 0.1182, "step": 198 }, { "epoch": 0.08, "learning_rate": 1.0446194225721785e-07, "loss": 0.1231, "step": 199 }, { "epoch": 0.08, "learning_rate": 1.0498687664041995e-07, "loss": 0.1279, "step": 200 }, { "epoch": 0.08, "learning_rate": 1.0551181102362203e-07, "loss": 0.1226, "step": 201 }, { "epoch": 0.08, "learning_rate": 1.0603674540682414e-07, "loss": 0.1157, "step": 202 }, { "epoch": 0.08, "learning_rate": 1.0656167979002624e-07, "loss": 0.1186, "step": 203 }, { "epoch": 0.08, "learning_rate": 1.0708661417322835e-07, "loss": 0.1125, "step": 204 }, { "epoch": 0.08, "learning_rate": 1.0761154855643045e-07, "loss": 0.1275, "step": 205 }, { "epoch": 0.08, "learning_rate": 1.0813648293963253e-07, "loss": 0.1152, "step": 206 }, { "epoch": 0.08, "learning_rate": 1.0866141732283464e-07, "loss": 0.1013, "step": 207 }, { "epoch": 0.08, "learning_rate": 1.0918635170603674e-07, "loss": 0.1125, "step": 208 }, { "epoch": 0.08, "learning_rate": 1.0971128608923883e-07, "loss": 0.1044, "step": 209 }, { "epoch": 0.08, "learning_rate": 1.1023622047244095e-07, "loss": 0.1114, "step": 210 }, { "epoch": 0.08, "learning_rate": 1.1076115485564303e-07, "loss": 0.1215, "step": 211 }, { "epoch": 0.08, "learning_rate": 1.1128608923884515e-07, "loss": 0.1051, "step": 212 }, { "epoch": 0.08, "learning_rate": 1.1181102362204724e-07, "loss": 0.1188, "step": 213 }, { "epoch": 0.08, "learning_rate": 1.1233595800524933e-07, "loss": 0.1031, "step": 214 }, { "epoch": 0.08, "learning_rate": 1.1286089238845144e-07, "loss": 0.1132, "step": 215 }, { "epoch": 0.09, "learning_rate": 1.1338582677165354e-07, "loss": 0.1062, "step": 216 }, { "epoch": 0.09, "learning_rate": 1.1391076115485562e-07, "loss": 0.1009, "step": 217 }, { "epoch": 0.09, "learning_rate": 1.1443569553805775e-07, "loss": 0.1051, "step": 218 }, { "epoch": 0.09, "learning_rate": 1.1496062992125983e-07, "loss": 0.109, "step": 219 }, { "epoch": 0.09, "learning_rate": 1.1548556430446194e-07, "loss": 0.1023, "step": 220 }, { "epoch": 0.09, "learning_rate": 1.1601049868766404e-07, "loss": 0.1037, "step": 221 }, { "epoch": 0.09, "learning_rate": 1.1653543307086612e-07, "loss": 0.0935, "step": 222 }, { "epoch": 0.09, "learning_rate": 1.1706036745406825e-07, "loss": 0.0985, "step": 223 }, { "epoch": 0.09, "learning_rate": 1.1758530183727033e-07, "loss": 0.0973, "step": 224 }, { "epoch": 0.09, "learning_rate": 1.1811023622047243e-07, "loss": 0.1045, "step": 225 }, { "epoch": 0.09, "learning_rate": 1.1863517060367454e-07, "loss": 0.1013, "step": 226 }, { "epoch": 0.09, "learning_rate": 1.1916010498687663e-07, "loss": 0.0932, "step": 227 }, { "epoch": 0.09, "learning_rate": 1.1968503937007874e-07, "loss": 0.0973, "step": 228 }, { "epoch": 0.09, "learning_rate": 1.2020997375328082e-07, "loss": 0.0986, "step": 229 }, { "epoch": 0.09, "learning_rate": 1.2073490813648293e-07, "loss": 0.1017, "step": 230 }, { "epoch": 0.09, "learning_rate": 1.2125984251968505e-07, "loss": 0.101, "step": 231 }, { "epoch": 0.09, "learning_rate": 1.2178477690288713e-07, "loss": 0.0977, "step": 232 }, { "epoch": 0.09, "learning_rate": 1.2230971128608921e-07, "loss": 0.0993, "step": 233 }, { "epoch": 0.09, "learning_rate": 1.2283464566929133e-07, "loss": 0.0992, "step": 234 }, { "epoch": 0.09, "learning_rate": 1.2335958005249344e-07, "loss": 0.0987, "step": 235 }, { "epoch": 0.09, "learning_rate": 1.2388451443569555e-07, "loss": 0.0983, "step": 236 }, { "epoch": 0.09, "learning_rate": 1.2440944881889763e-07, "loss": 0.1005, "step": 237 }, { "epoch": 0.09, "learning_rate": 1.2493438320209972e-07, "loss": 0.0897, "step": 238 }, { "epoch": 0.09, "learning_rate": 1.2545931758530183e-07, "loss": 0.0885, "step": 239 }, { "epoch": 0.09, "learning_rate": 1.2598425196850394e-07, "loss": 0.0959, "step": 240 }, { "epoch": 0.09, "learning_rate": 1.2650918635170602e-07, "loss": 0.0974, "step": 241 }, { "epoch": 0.1, "learning_rate": 1.2703412073490814e-07, "loss": 0.0894, "step": 242 }, { "epoch": 0.1, "learning_rate": 1.2755905511811022e-07, "loss": 0.1003, "step": 243 }, { "epoch": 0.1, "learning_rate": 1.2808398950131233e-07, "loss": 0.0885, "step": 244 }, { "epoch": 0.1, "learning_rate": 1.2860892388451444e-07, "loss": 0.0886, "step": 245 }, { "epoch": 0.1, "learning_rate": 1.2913385826771653e-07, "loss": 0.0894, "step": 246 }, { "epoch": 0.1, "learning_rate": 1.2965879265091864e-07, "loss": 0.0928, "step": 247 }, { "epoch": 0.1, "learning_rate": 1.3018372703412072e-07, "loss": 0.0911, "step": 248 }, { "epoch": 0.1, "learning_rate": 1.307086614173228e-07, "loss": 0.0859, "step": 249 }, { "epoch": 0.1, "learning_rate": 1.3123359580052494e-07, "loss": 0.0962, "step": 250 }, { "epoch": 0.1, "learning_rate": 1.3175853018372703e-07, "loss": 0.1094, "step": 251 }, { "epoch": 0.1, "learning_rate": 1.3228346456692914e-07, "loss": 0.1012, "step": 252 }, { "epoch": 0.1, "learning_rate": 1.3280839895013123e-07, "loss": 0.0892, "step": 253 }, { "epoch": 0.1, "learning_rate": 1.333333333333333e-07, "loss": 0.0871, "step": 254 }, { "epoch": 0.1, "learning_rate": 1.3385826771653542e-07, "loss": 0.0892, "step": 255 }, { "epoch": 0.1, "learning_rate": 1.3438320209973753e-07, "loss": 0.094, "step": 256 }, { "epoch": 0.1, "learning_rate": 1.3490813648293962e-07, "loss": 0.0848, "step": 257 }, { "epoch": 0.1, "learning_rate": 1.3543307086614173e-07, "loss": 0.094, "step": 258 }, { "epoch": 0.1, "learning_rate": 1.359580052493438e-07, "loss": 0.0934, "step": 259 }, { "epoch": 0.1, "learning_rate": 1.3648293963254592e-07, "loss": 0.0912, "step": 260 }, { "epoch": 0.1, "learning_rate": 1.3700787401574803e-07, "loss": 0.0851, "step": 261 }, { "epoch": 0.1, "learning_rate": 1.3753280839895012e-07, "loss": 0.0847, "step": 262 }, { "epoch": 0.1, "learning_rate": 1.3805774278215223e-07, "loss": 0.0932, "step": 263 }, { "epoch": 0.1, "learning_rate": 1.3858267716535431e-07, "loss": 0.0747, "step": 264 }, { "epoch": 0.1, "learning_rate": 1.3910761154855643e-07, "loss": 0.0848, "step": 265 }, { "epoch": 0.1, "learning_rate": 1.3963254593175854e-07, "loss": 0.0898, "step": 266 }, { "epoch": 0.11, "learning_rate": 1.4015748031496062e-07, "loss": 0.0834, "step": 267 }, { "epoch": 0.11, "learning_rate": 1.4068241469816273e-07, "loss": 0.0893, "step": 268 }, { "epoch": 0.11, "learning_rate": 1.4120734908136482e-07, "loss": 0.0889, "step": 269 }, { "epoch": 0.11, "learning_rate": 1.417322834645669e-07, "loss": 0.0878, "step": 270 }, { "epoch": 0.11, "learning_rate": 1.4225721784776904e-07, "loss": 0.0852, "step": 271 }, { "epoch": 0.11, "learning_rate": 1.4278215223097112e-07, "loss": 0.0886, "step": 272 }, { "epoch": 0.11, "learning_rate": 1.433070866141732e-07, "loss": 0.0805, "step": 273 }, { "epoch": 0.11, "learning_rate": 1.4383202099737532e-07, "loss": 0.0829, "step": 274 }, { "epoch": 0.11, "learning_rate": 1.443569553805774e-07, "loss": 0.0783, "step": 275 }, { "epoch": 0.11, "learning_rate": 1.4488188976377954e-07, "loss": 0.0787, "step": 276 }, { "epoch": 0.11, "learning_rate": 1.4540682414698163e-07, "loss": 0.0895, "step": 277 }, { "epoch": 0.11, "learning_rate": 1.459317585301837e-07, "loss": 0.0839, "step": 278 }, { "epoch": 0.11, "learning_rate": 1.4645669291338582e-07, "loss": 0.0809, "step": 279 }, { "epoch": 0.11, "learning_rate": 1.469816272965879e-07, "loss": 0.0896, "step": 280 }, { "epoch": 0.11, "learning_rate": 1.4750656167979002e-07, "loss": 0.0844, "step": 281 }, { "epoch": 0.11, "learning_rate": 1.4803149606299213e-07, "loss": 0.0833, "step": 282 }, { "epoch": 0.11, "learning_rate": 1.4855643044619421e-07, "loss": 0.0915, "step": 283 }, { "epoch": 0.11, "learning_rate": 1.4908136482939633e-07, "loss": 0.0862, "step": 284 }, { "epoch": 0.11, "learning_rate": 1.496062992125984e-07, "loss": 0.0812, "step": 285 }, { "epoch": 0.11, "learning_rate": 1.5013123359580052e-07, "loss": 0.0845, "step": 286 }, { "epoch": 0.11, "learning_rate": 1.5065616797900263e-07, "loss": 0.0831, "step": 287 }, { "epoch": 0.11, "learning_rate": 1.5118110236220472e-07, "loss": 0.0818, "step": 288 }, { "epoch": 0.11, "learning_rate": 1.517060367454068e-07, "loss": 0.0823, "step": 289 }, { "epoch": 0.11, "learning_rate": 1.522309711286089e-07, "loss": 0.0781, "step": 290 }, { "epoch": 0.11, "learning_rate": 1.5275590551181102e-07, "loss": 0.0841, "step": 291 }, { "epoch": 0.12, "learning_rate": 1.5328083989501313e-07, "loss": 0.0804, "step": 292 }, { "epoch": 0.12, "learning_rate": 1.5380577427821522e-07, "loss": 0.0753, "step": 293 }, { "epoch": 0.12, "learning_rate": 1.543307086614173e-07, "loss": 0.0745, "step": 294 }, { "epoch": 0.12, "learning_rate": 1.5485564304461942e-07, "loss": 0.0765, "step": 295 }, { "epoch": 0.12, "learning_rate": 1.5538057742782153e-07, "loss": 0.086, "step": 296 }, { "epoch": 0.12, "learning_rate": 1.559055118110236e-07, "loss": 0.079, "step": 297 }, { "epoch": 0.12, "learning_rate": 1.5643044619422572e-07, "loss": 0.0776, "step": 298 }, { "epoch": 0.12, "learning_rate": 1.569553805774278e-07, "loss": 0.0819, "step": 299 }, { "epoch": 0.12, "learning_rate": 1.5748031496062992e-07, "loss": 0.0803, "step": 300 }, { "epoch": 0.12, "learning_rate": 1.58005249343832e-07, "loss": 0.0865, "step": 301 }, { "epoch": 0.12, "learning_rate": 1.5853018372703411e-07, "loss": 0.0875, "step": 302 }, { "epoch": 0.12, "learning_rate": 1.5905511811023622e-07, "loss": 0.08, "step": 303 }, { "epoch": 0.12, "learning_rate": 1.595800524934383e-07, "loss": 0.0866, "step": 304 }, { "epoch": 0.12, "learning_rate": 1.601049868766404e-07, "loss": 0.087, "step": 305 }, { "epoch": 0.12, "learning_rate": 1.606299212598425e-07, "loss": 0.0821, "step": 306 }, { "epoch": 0.12, "learning_rate": 1.6115485564304462e-07, "loss": 0.0804, "step": 307 }, { "epoch": 0.12, "learning_rate": 1.6167979002624673e-07, "loss": 0.0823, "step": 308 }, { "epoch": 0.12, "learning_rate": 1.622047244094488e-07, "loss": 0.0818, "step": 309 }, { "epoch": 0.12, "learning_rate": 1.627296587926509e-07, "loss": 0.0807, "step": 310 }, { "epoch": 0.12, "learning_rate": 1.63254593175853e-07, "loss": 0.0835, "step": 311 }, { "epoch": 0.12, "learning_rate": 1.6377952755905512e-07, "loss": 0.0798, "step": 312 }, { "epoch": 0.12, "learning_rate": 1.643044619422572e-07, "loss": 0.0798, "step": 313 }, { "epoch": 0.12, "learning_rate": 1.6482939632545931e-07, "loss": 0.0716, "step": 314 }, { "epoch": 0.12, "learning_rate": 1.653543307086614e-07, "loss": 0.0752, "step": 315 }, { "epoch": 0.12, "learning_rate": 1.658792650918635e-07, "loss": 0.0811, "step": 316 }, { "epoch": 0.12, "learning_rate": 1.6640419947506562e-07, "loss": 0.0796, "step": 317 }, { "epoch": 0.13, "learning_rate": 1.669291338582677e-07, "loss": 0.0757, "step": 318 }, { "epoch": 0.13, "learning_rate": 1.6745406824146982e-07, "loss": 0.0783, "step": 319 }, { "epoch": 0.13, "learning_rate": 1.679790026246719e-07, "loss": 0.08, "step": 320 }, { "epoch": 0.13, "learning_rate": 1.6850393700787399e-07, "loss": 0.0746, "step": 321 }, { "epoch": 0.13, "learning_rate": 1.6902887139107612e-07, "loss": 0.0989, "step": 322 }, { "epoch": 0.13, "learning_rate": 1.695538057742782e-07, "loss": 0.0705, "step": 323 }, { "epoch": 0.13, "learning_rate": 1.7007874015748032e-07, "loss": 0.0726, "step": 324 }, { "epoch": 0.13, "learning_rate": 1.706036745406824e-07, "loss": 0.0783, "step": 325 }, { "epoch": 0.13, "learning_rate": 1.711286089238845e-07, "loss": 0.0748, "step": 326 }, { "epoch": 0.13, "learning_rate": 1.716535433070866e-07, "loss": 0.071, "step": 327 }, { "epoch": 0.13, "learning_rate": 1.721784776902887e-07, "loss": 0.0811, "step": 328 }, { "epoch": 0.13, "learning_rate": 1.727034120734908e-07, "loss": 0.0759, "step": 329 }, { "epoch": 0.13, "learning_rate": 1.732283464566929e-07, "loss": 0.0791, "step": 330 }, { "epoch": 0.13, "learning_rate": 1.73753280839895e-07, "loss": 0.0753, "step": 331 }, { "epoch": 0.13, "learning_rate": 1.742782152230971e-07, "loss": 0.0792, "step": 332 }, { "epoch": 0.13, "learning_rate": 1.7480314960629921e-07, "loss": 0.0794, "step": 333 }, { "epoch": 0.13, "learning_rate": 1.753280839895013e-07, "loss": 0.0772, "step": 334 }, { "epoch": 0.13, "learning_rate": 1.758530183727034e-07, "loss": 0.0796, "step": 335 }, { "epoch": 0.13, "learning_rate": 1.763779527559055e-07, "loss": 0.0778, "step": 336 }, { "epoch": 0.13, "learning_rate": 1.769028871391076e-07, "loss": 0.0775, "step": 337 }, { "epoch": 0.13, "learning_rate": 1.7742782152230972e-07, "loss": 0.0828, "step": 338 }, { "epoch": 0.13, "learning_rate": 1.779527559055118e-07, "loss": 0.0771, "step": 339 }, { "epoch": 0.13, "learning_rate": 1.784776902887139e-07, "loss": 0.0783, "step": 340 }, { "epoch": 0.13, "learning_rate": 1.79002624671916e-07, "loss": 0.0786, "step": 341 }, { "epoch": 0.13, "learning_rate": 1.7952755905511808e-07, "loss": 0.0763, "step": 342 }, { "epoch": 0.14, "learning_rate": 1.8005249343832022e-07, "loss": 0.076, "step": 343 }, { "epoch": 0.14, "learning_rate": 1.805774278215223e-07, "loss": 0.0759, "step": 344 }, { "epoch": 0.14, "learning_rate": 1.811023622047244e-07, "loss": 0.0769, "step": 345 }, { "epoch": 0.14, "learning_rate": 1.816272965879265e-07, "loss": 0.0794, "step": 346 }, { "epoch": 0.14, "learning_rate": 1.8215223097112858e-07, "loss": 0.0768, "step": 347 }, { "epoch": 0.14, "learning_rate": 1.8267716535433072e-07, "loss": 0.0692, "step": 348 }, { "epoch": 0.14, "learning_rate": 1.832020997375328e-07, "loss": 0.0747, "step": 349 }, { "epoch": 0.14, "learning_rate": 1.837270341207349e-07, "loss": 0.0778, "step": 350 }, { "epoch": 0.14, "learning_rate": 1.84251968503937e-07, "loss": 0.0796, "step": 351 }, { "epoch": 0.14, "learning_rate": 1.8477690288713909e-07, "loss": 0.0815, "step": 352 }, { "epoch": 0.14, "learning_rate": 1.853018372703412e-07, "loss": 0.0743, "step": 353 }, { "epoch": 0.14, "learning_rate": 1.858267716535433e-07, "loss": 0.0714, "step": 354 }, { "epoch": 0.14, "learning_rate": 1.863517060367454e-07, "loss": 0.0657, "step": 355 }, { "epoch": 0.14, "learning_rate": 1.868766404199475e-07, "loss": 0.07, "step": 356 }, { "epoch": 0.14, "learning_rate": 1.874015748031496e-07, "loss": 0.0746, "step": 357 }, { "epoch": 0.14, "learning_rate": 1.879265091863517e-07, "loss": 0.0773, "step": 358 }, { "epoch": 0.14, "learning_rate": 1.884514435695538e-07, "loss": 0.0712, "step": 359 }, { "epoch": 0.14, "learning_rate": 1.889763779527559e-07, "loss": 0.0733, "step": 360 }, { "epoch": 0.14, "learning_rate": 1.8950131233595798e-07, "loss": 0.0842, "step": 361 }, { "epoch": 0.14, "learning_rate": 1.900262467191601e-07, "loss": 0.073, "step": 362 }, { "epoch": 0.14, "learning_rate": 1.905511811023622e-07, "loss": 0.08, "step": 363 }, { "epoch": 0.14, "learning_rate": 1.9107611548556431e-07, "loss": 0.0828, "step": 364 }, { "epoch": 0.14, "learning_rate": 1.916010498687664e-07, "loss": 0.0814, "step": 365 }, { "epoch": 0.14, "learning_rate": 1.9212598425196848e-07, "loss": 0.0699, "step": 366 }, { "epoch": 0.14, "learning_rate": 1.926509186351706e-07, "loss": 0.0653, "step": 367 }, { "epoch": 0.15, "learning_rate": 1.9317585301837268e-07, "loss": 0.0688, "step": 368 }, { "epoch": 0.15, "learning_rate": 1.937007874015748e-07, "loss": 0.0741, "step": 369 }, { "epoch": 0.15, "learning_rate": 1.942257217847769e-07, "loss": 0.0749, "step": 370 }, { "epoch": 0.15, "learning_rate": 1.9475065616797899e-07, "loss": 0.0724, "step": 371 }, { "epoch": 0.15, "learning_rate": 1.952755905511811e-07, "loss": 0.0771, "step": 372 }, { "epoch": 0.15, "learning_rate": 1.9580052493438318e-07, "loss": 0.0729, "step": 373 }, { "epoch": 0.15, "learning_rate": 1.963254593175853e-07, "loss": 0.0743, "step": 374 }, { "epoch": 0.15, "learning_rate": 1.968503937007874e-07, "loss": 0.0747, "step": 375 }, { "epoch": 0.15, "learning_rate": 1.973753280839895e-07, "loss": 0.0776, "step": 376 }, { "epoch": 0.15, "learning_rate": 1.9790026246719157e-07, "loss": 0.0725, "step": 377 }, { "epoch": 0.15, "learning_rate": 1.9842519685039368e-07, "loss": 0.0789, "step": 378 }, { "epoch": 0.15, "learning_rate": 1.989501312335958e-07, "loss": 0.0695, "step": 379 }, { "epoch": 0.15, "learning_rate": 1.994750656167979e-07, "loss": 0.0771, "step": 380 }, { "epoch": 0.15, "learning_rate": 2e-07, "loss": 0.0765, "step": 381 }, { "epoch": 0.15, "learning_rate": 2.005249343832021e-07, "loss": 0.0653, "step": 382 }, { "epoch": 0.15, "learning_rate": 2.0104986876640419e-07, "loss": 0.0694, "step": 383 }, { "epoch": 0.15, "learning_rate": 2.015748031496063e-07, "loss": 0.0826, "step": 384 }, { "epoch": 0.15, "learning_rate": 2.020997375328084e-07, "loss": 0.0741, "step": 385 }, { "epoch": 0.15, "learning_rate": 2.0262467191601047e-07, "loss": 0.0701, "step": 386 }, { "epoch": 0.15, "learning_rate": 2.0314960629921258e-07, "loss": 0.083, "step": 387 }, { "epoch": 0.15, "learning_rate": 2.036745406824147e-07, "loss": 0.0695, "step": 388 }, { "epoch": 0.15, "learning_rate": 2.041994750656168e-07, "loss": 0.0679, "step": 389 }, { "epoch": 0.15, "learning_rate": 2.0472440944881889e-07, "loss": 0.0722, "step": 390 }, { "epoch": 0.15, "learning_rate": 2.05249343832021e-07, "loss": 0.067, "step": 391 }, { "epoch": 0.15, "learning_rate": 2.057742782152231e-07, "loss": 0.0708, "step": 392 }, { "epoch": 0.15, "learning_rate": 2.0629921259842517e-07, "loss": 0.0791, "step": 393 }, { "epoch": 0.16, "learning_rate": 2.068241469816273e-07, "loss": 0.0657, "step": 394 }, { "epoch": 0.16, "learning_rate": 2.0734908136482941e-07, "loss": 0.0765, "step": 395 }, { "epoch": 0.16, "learning_rate": 2.0787401574803147e-07, "loss": 0.0688, "step": 396 }, { "epoch": 0.16, "learning_rate": 2.0839895013123358e-07, "loss": 0.0701, "step": 397 }, { "epoch": 0.16, "learning_rate": 2.089238845144357e-07, "loss": 0.0814, "step": 398 }, { "epoch": 0.16, "learning_rate": 2.0944881889763778e-07, "loss": 0.0708, "step": 399 }, { "epoch": 0.16, "learning_rate": 2.099737532808399e-07, "loss": 0.0779, "step": 400 }, { "epoch": 0.16, "learning_rate": 2.10498687664042e-07, "loss": 0.0706, "step": 401 }, { "epoch": 0.16, "learning_rate": 2.1102362204724406e-07, "loss": 0.0757, "step": 402 }, { "epoch": 0.16, "learning_rate": 2.1154855643044617e-07, "loss": 0.0691, "step": 403 }, { "epoch": 0.16, "learning_rate": 2.1207349081364828e-07, "loss": 0.0663, "step": 404 }, { "epoch": 0.16, "learning_rate": 2.1259842519685042e-07, "loss": 0.0695, "step": 405 }, { "epoch": 0.16, "learning_rate": 2.1312335958005248e-07, "loss": 0.0698, "step": 406 }, { "epoch": 0.16, "learning_rate": 2.136482939632546e-07, "loss": 0.074, "step": 407 }, { "epoch": 0.16, "learning_rate": 2.141732283464567e-07, "loss": 0.0688, "step": 408 }, { "epoch": 0.16, "learning_rate": 2.1469816272965878e-07, "loss": 0.0644, "step": 409 }, { "epoch": 0.16, "learning_rate": 2.152230971128609e-07, "loss": 0.0719, "step": 410 }, { "epoch": 0.16, "learning_rate": 2.15748031496063e-07, "loss": 0.0692, "step": 411 }, { "epoch": 0.16, "learning_rate": 2.1627296587926506e-07, "loss": 0.07, "step": 412 }, { "epoch": 0.16, "learning_rate": 2.1679790026246718e-07, "loss": 0.0734, "step": 413 }, { "epoch": 0.16, "learning_rate": 2.173228346456693e-07, "loss": 0.0639, "step": 414 }, { "epoch": 0.16, "learning_rate": 2.1784776902887137e-07, "loss": 0.0634, "step": 415 }, { "epoch": 0.16, "learning_rate": 2.1837270341207348e-07, "loss": 0.0737, "step": 416 }, { "epoch": 0.16, "learning_rate": 2.188976377952756e-07, "loss": 0.0733, "step": 417 }, { "epoch": 0.16, "learning_rate": 2.1942257217847765e-07, "loss": 0.0692, "step": 418 }, { "epoch": 0.17, "learning_rate": 2.1994750656167976e-07, "loss": 0.0751, "step": 419 }, { "epoch": 0.17, "learning_rate": 2.204724409448819e-07, "loss": 0.0747, "step": 420 }, { "epoch": 0.17, "learning_rate": 2.20997375328084e-07, "loss": 0.075, "step": 421 }, { "epoch": 0.17, "learning_rate": 2.2152230971128607e-07, "loss": 0.0719, "step": 422 }, { "epoch": 0.17, "learning_rate": 2.2204724409448818e-07, "loss": 0.0665, "step": 423 }, { "epoch": 0.17, "learning_rate": 2.225721784776903e-07, "loss": 0.0784, "step": 424 }, { "epoch": 0.17, "learning_rate": 2.2309711286089238e-07, "loss": 0.0715, "step": 425 }, { "epoch": 0.17, "learning_rate": 2.236220472440945e-07, "loss": 0.0658, "step": 426 }, { "epoch": 0.17, "learning_rate": 2.241469816272966e-07, "loss": 0.0683, "step": 427 }, { "epoch": 0.17, "learning_rate": 2.2467191601049866e-07, "loss": 0.0735, "step": 428 }, { "epoch": 0.17, "learning_rate": 2.2519685039370077e-07, "loss": 0.081, "step": 429 }, { "epoch": 0.17, "learning_rate": 2.2572178477690288e-07, "loss": 0.0688, "step": 430 }, { "epoch": 0.17, "learning_rate": 2.2624671916010496e-07, "loss": 0.0661, "step": 431 }, { "epoch": 0.17, "learning_rate": 2.2677165354330708e-07, "loss": 0.0624, "step": 432 }, { "epoch": 0.17, "learning_rate": 2.2729658792650919e-07, "loss": 0.0664, "step": 433 }, { "epoch": 0.17, "learning_rate": 2.2782152230971124e-07, "loss": 0.0655, "step": 434 }, { "epoch": 0.17, "learning_rate": 2.2834645669291338e-07, "loss": 0.0683, "step": 435 }, { "epoch": 0.17, "learning_rate": 2.288713910761155e-07, "loss": 0.0639, "step": 436 }, { "epoch": 0.17, "learning_rate": 2.293963254593176e-07, "loss": 0.0643, "step": 437 }, { "epoch": 0.17, "learning_rate": 2.2992125984251966e-07, "loss": 0.0787, "step": 438 }, { "epoch": 0.17, "learning_rate": 2.3044619422572177e-07, "loss": 0.0722, "step": 439 }, { "epoch": 0.17, "learning_rate": 2.3097112860892388e-07, "loss": 0.0776, "step": 440 }, { "epoch": 0.17, "learning_rate": 2.3149606299212597e-07, "loss": 0.0718, "step": 441 }, { "epoch": 0.17, "learning_rate": 2.3202099737532808e-07, "loss": 0.0673, "step": 442 }, { "epoch": 0.17, "learning_rate": 2.325459317585302e-07, "loss": 0.0688, "step": 443 }, { "epoch": 0.17, "learning_rate": 2.3307086614173225e-07, "loss": 0.0791, "step": 444 }, { "epoch": 0.18, "learning_rate": 2.3359580052493436e-07, "loss": 0.0697, "step": 445 }, { "epoch": 0.18, "learning_rate": 2.341207349081365e-07, "loss": 0.0722, "step": 446 }, { "epoch": 0.18, "learning_rate": 2.3464566929133856e-07, "loss": 0.0661, "step": 447 }, { "epoch": 0.18, "learning_rate": 2.3517060367454067e-07, "loss": 0.0708, "step": 448 }, { "epoch": 0.18, "learning_rate": 2.3569553805774278e-07, "loss": 0.0702, "step": 449 }, { "epoch": 0.18, "learning_rate": 2.3622047244094486e-07, "loss": 0.0614, "step": 450 }, { "epoch": 0.18, "learning_rate": 2.3674540682414697e-07, "loss": 0.0729, "step": 451 }, { "epoch": 0.18, "learning_rate": 2.3727034120734909e-07, "loss": 0.0743, "step": 452 }, { "epoch": 0.18, "learning_rate": 2.377952755905512e-07, "loss": 0.0652, "step": 453 }, { "epoch": 0.18, "learning_rate": 2.3832020997375325e-07, "loss": 0.0767, "step": 454 }, { "epoch": 0.18, "learning_rate": 2.3884514435695537e-07, "loss": 0.0676, "step": 455 }, { "epoch": 0.18, "learning_rate": 2.393700787401575e-07, "loss": 0.0722, "step": 456 }, { "epoch": 0.18, "learning_rate": 2.3989501312335954e-07, "loss": 0.0683, "step": 457 }, { "epoch": 0.18, "learning_rate": 2.4041994750656165e-07, "loss": 0.0735, "step": 458 }, { "epoch": 0.18, "learning_rate": 2.409448818897638e-07, "loss": 0.0582, "step": 459 }, { "epoch": 0.18, "learning_rate": 2.4146981627296587e-07, "loss": 0.0633, "step": 460 }, { "epoch": 0.18, "learning_rate": 2.41994750656168e-07, "loss": 0.0678, "step": 461 }, { "epoch": 0.18, "learning_rate": 2.425196850393701e-07, "loss": 0.0695, "step": 462 }, { "epoch": 0.18, "learning_rate": 2.4304461942257215e-07, "loss": 0.0711, "step": 463 }, { "epoch": 0.18, "learning_rate": 2.4356955380577426e-07, "loss": 0.0711, "step": 464 }, { "epoch": 0.18, "learning_rate": 2.4409448818897637e-07, "loss": 0.0635, "step": 465 }, { "epoch": 0.18, "learning_rate": 2.4461942257217843e-07, "loss": 0.0619, "step": 466 }, { "epoch": 0.18, "learning_rate": 2.4514435695538054e-07, "loss": 0.0642, "step": 467 }, { "epoch": 0.18, "learning_rate": 2.4566929133858265e-07, "loss": 0.0697, "step": 468 }, { "epoch": 0.18, "learning_rate": 2.4619422572178476e-07, "loss": 0.075, "step": 469 }, { "epoch": 0.19, "learning_rate": 2.467191601049869e-07, "loss": 0.0662, "step": 470 }, { "epoch": 0.19, "learning_rate": 2.47244094488189e-07, "loss": 0.0734, "step": 471 }, { "epoch": 0.19, "learning_rate": 2.477690288713911e-07, "loss": 0.0678, "step": 472 }, { "epoch": 0.19, "learning_rate": 2.4829396325459315e-07, "loss": 0.0712, "step": 473 }, { "epoch": 0.19, "learning_rate": 2.4881889763779527e-07, "loss": 0.0686, "step": 474 }, { "epoch": 0.19, "learning_rate": 2.493438320209974e-07, "loss": 0.0643, "step": 475 }, { "epoch": 0.19, "learning_rate": 2.4986876640419943e-07, "loss": 0.0651, "step": 476 }, { "epoch": 0.19, "learning_rate": 2.5039370078740155e-07, "loss": 0.0666, "step": 477 }, { "epoch": 0.19, "learning_rate": 2.5091863517060366e-07, "loss": 0.0693, "step": 478 }, { "epoch": 0.19, "learning_rate": 2.5144356955380577e-07, "loss": 0.062, "step": 479 }, { "epoch": 0.19, "learning_rate": 2.519685039370079e-07, "loss": 0.0693, "step": 480 }, { "epoch": 0.19, "learning_rate": 2.5249343832021e-07, "loss": 0.0697, "step": 481 }, { "epoch": 0.19, "learning_rate": 2.5301837270341205e-07, "loss": 0.0658, "step": 482 }, { "epoch": 0.19, "learning_rate": 2.5354330708661416e-07, "loss": 0.0727, "step": 483 }, { "epoch": 0.19, "learning_rate": 2.5406824146981627e-07, "loss": 0.0708, "step": 484 }, { "epoch": 0.19, "learning_rate": 2.545931758530184e-07, "loss": 0.0758, "step": 485 }, { "epoch": 0.19, "learning_rate": 2.5511811023622044e-07, "loss": 0.0652, "step": 486 }, { "epoch": 0.19, "learning_rate": 2.5564304461942255e-07, "loss": 0.0665, "step": 487 }, { "epoch": 0.19, "learning_rate": 2.5616797900262466e-07, "loss": 0.0675, "step": 488 }, { "epoch": 0.19, "learning_rate": 2.5669291338582677e-07, "loss": 0.0693, "step": 489 }, { "epoch": 0.19, "learning_rate": 2.572178477690289e-07, "loss": 0.0608, "step": 490 }, { "epoch": 0.19, "learning_rate": 2.57742782152231e-07, "loss": 0.0628, "step": 491 }, { "epoch": 0.19, "learning_rate": 2.5826771653543305e-07, "loss": 0.0718, "step": 492 }, { "epoch": 0.19, "learning_rate": 2.5879265091863516e-07, "loss": 0.0694, "step": 493 }, { "epoch": 0.19, "learning_rate": 2.593175853018373e-07, "loss": 0.0669, "step": 494 }, { "epoch": 0.2, "learning_rate": 2.5984251968503933e-07, "loss": 0.0653, "step": 495 }, { "epoch": 0.2, "learning_rate": 2.6036745406824144e-07, "loss": 0.0644, "step": 496 }, { "epoch": 0.2, "learning_rate": 2.6089238845144356e-07, "loss": 0.0607, "step": 497 }, { "epoch": 0.2, "learning_rate": 2.614173228346456e-07, "loss": 0.0762, "step": 498 }, { "epoch": 0.2, "learning_rate": 2.619422572178477e-07, "loss": 0.0642, "step": 499 }, { "epoch": 0.2, "learning_rate": 2.624671916010499e-07, "loss": 0.0625, "step": 500 }, { "epoch": 0.2, "learning_rate": 2.62992125984252e-07, "loss": 0.0662, "step": 501 }, { "epoch": 0.2, "learning_rate": 2.6351706036745406e-07, "loss": 0.0665, "step": 502 }, { "epoch": 0.2, "learning_rate": 2.6404199475065617e-07, "loss": 0.066, "step": 503 }, { "epoch": 0.2, "learning_rate": 2.645669291338583e-07, "loss": 0.0648, "step": 504 }, { "epoch": 0.2, "learning_rate": 2.6509186351706034e-07, "loss": 0.0622, "step": 505 }, { "epoch": 0.2, "learning_rate": 2.6561679790026245e-07, "loss": 0.0682, "step": 506 }, { "epoch": 0.2, "learning_rate": 2.6614173228346456e-07, "loss": 0.0632, "step": 507 }, { "epoch": 0.2, "learning_rate": 2.666666666666666e-07, "loss": 0.0669, "step": 508 }, { "epoch": 0.2, "learning_rate": 2.6719160104986873e-07, "loss": 0.0614, "step": 509 }, { "epoch": 0.2, "learning_rate": 2.6771653543307084e-07, "loss": 0.0685, "step": 510 }, { "epoch": 0.2, "learning_rate": 2.6824146981627295e-07, "loss": 0.056, "step": 511 }, { "epoch": 0.2, "learning_rate": 2.6876640419947506e-07, "loss": 0.0659, "step": 512 }, { "epoch": 0.2, "learning_rate": 2.692913385826772e-07, "loss": 0.0599, "step": 513 }, { "epoch": 0.2, "learning_rate": 2.6981627296587923e-07, "loss": 0.069, "step": 514 }, { "epoch": 0.2, "learning_rate": 2.7034120734908134e-07, "loss": 0.0638, "step": 515 }, { "epoch": 0.2, "learning_rate": 2.7086614173228346e-07, "loss": 0.0638, "step": 516 }, { "epoch": 0.2, "learning_rate": 2.7139107611548557e-07, "loss": 0.0596, "step": 517 }, { "epoch": 0.2, "learning_rate": 2.719160104986876e-07, "loss": 0.059, "step": 518 }, { "epoch": 0.2, "learning_rate": 2.7244094488188974e-07, "loss": 0.0633, "step": 519 }, { "epoch": 0.2, "learning_rate": 2.7296587926509185e-07, "loss": 0.0685, "step": 520 }, { "epoch": 0.21, "learning_rate": 2.7349081364829396e-07, "loss": 0.061, "step": 521 }, { "epoch": 0.21, "learning_rate": 2.7401574803149607e-07, "loss": 0.0721, "step": 522 }, { "epoch": 0.21, "learning_rate": 2.745406824146982e-07, "loss": 0.0696, "step": 523 }, { "epoch": 0.21, "learning_rate": 2.7506561679790024e-07, "loss": 0.0602, "step": 524 }, { "epoch": 0.21, "learning_rate": 2.7559055118110235e-07, "loss": 0.0605, "step": 525 }, { "epoch": 0.21, "learning_rate": 2.7611548556430446e-07, "loss": 0.0667, "step": 526 }, { "epoch": 0.21, "learning_rate": 2.766404199475065e-07, "loss": 0.0683, "step": 527 }, { "epoch": 0.21, "learning_rate": 2.7716535433070863e-07, "loss": 0.0684, "step": 528 }, { "epoch": 0.21, "learning_rate": 2.7769028871391074e-07, "loss": 0.0617, "step": 529 }, { "epoch": 0.21, "learning_rate": 2.7821522309711285e-07, "loss": 0.0625, "step": 530 }, { "epoch": 0.21, "learning_rate": 2.7874015748031496e-07, "loss": 0.0667, "step": 531 }, { "epoch": 0.21, "learning_rate": 2.792650918635171e-07, "loss": 0.0685, "step": 532 }, { "epoch": 0.21, "learning_rate": 2.797900262467192e-07, "loss": 0.0655, "step": 533 }, { "epoch": 0.21, "learning_rate": 2.8031496062992124e-07, "loss": 0.0696, "step": 534 }, { "epoch": 0.21, "learning_rate": 2.8083989501312335e-07, "loss": 0.0634, "step": 535 }, { "epoch": 0.21, "learning_rate": 2.8136482939632547e-07, "loss": 0.0618, "step": 536 }, { "epoch": 0.21, "learning_rate": 2.818897637795275e-07, "loss": 0.0644, "step": 537 }, { "epoch": 0.21, "learning_rate": 2.8241469816272963e-07, "loss": 0.0716, "step": 538 }, { "epoch": 0.21, "learning_rate": 2.8293963254593175e-07, "loss": 0.0627, "step": 539 }, { "epoch": 0.21, "learning_rate": 2.834645669291338e-07, "loss": 0.0664, "step": 540 }, { "epoch": 0.21, "learning_rate": 2.8398950131233597e-07, "loss": 0.0642, "step": 541 }, { "epoch": 0.21, "learning_rate": 2.845144356955381e-07, "loss": 0.0657, "step": 542 }, { "epoch": 0.21, "learning_rate": 2.8503937007874014e-07, "loss": 0.069, "step": 543 }, { "epoch": 0.21, "learning_rate": 2.8556430446194225e-07, "loss": 0.0638, "step": 544 }, { "epoch": 0.21, "learning_rate": 2.8608923884514436e-07, "loss": 0.0654, "step": 545 }, { "epoch": 0.22, "learning_rate": 2.866141732283464e-07, "loss": 0.0555, "step": 546 }, { "epoch": 0.22, "learning_rate": 2.8713910761154853e-07, "loss": 0.0601, "step": 547 }, { "epoch": 0.22, "learning_rate": 2.8766404199475064e-07, "loss": 0.0662, "step": 548 }, { "epoch": 0.22, "learning_rate": 2.8818897637795275e-07, "loss": 0.0632, "step": 549 }, { "epoch": 0.22, "learning_rate": 2.887139107611548e-07, "loss": 0.0672, "step": 550 }, { "epoch": 0.22, "learning_rate": 2.892388451443569e-07, "loss": 0.0662, "step": 551 }, { "epoch": 0.22, "learning_rate": 2.897637795275591e-07, "loss": 0.0677, "step": 552 }, { "epoch": 0.22, "learning_rate": 2.9028871391076114e-07, "loss": 0.0656, "step": 553 }, { "epoch": 0.22, "learning_rate": 2.9081364829396325e-07, "loss": 0.0676, "step": 554 }, { "epoch": 0.22, "learning_rate": 2.9133858267716536e-07, "loss": 0.0634, "step": 555 }, { "epoch": 0.22, "learning_rate": 2.918635170603674e-07, "loss": 0.0596, "step": 556 }, { "epoch": 0.22, "learning_rate": 2.9238845144356953e-07, "loss": 0.0671, "step": 557 }, { "epoch": 0.22, "learning_rate": 2.9291338582677165e-07, "loss": 0.0621, "step": 558 }, { "epoch": 0.22, "learning_rate": 2.934383202099737e-07, "loss": 0.0631, "step": 559 }, { "epoch": 0.22, "learning_rate": 2.939632545931758e-07, "loss": 0.0597, "step": 560 }, { "epoch": 0.22, "learning_rate": 2.944881889763779e-07, "loss": 0.0717, "step": 561 }, { "epoch": 0.22, "learning_rate": 2.9501312335958004e-07, "loss": 0.0621, "step": 562 }, { "epoch": 0.22, "learning_rate": 2.9553805774278215e-07, "loss": 0.0604, "step": 563 }, { "epoch": 0.22, "learning_rate": 2.9606299212598426e-07, "loss": 0.0632, "step": 564 }, { "epoch": 0.22, "learning_rate": 2.9658792650918637e-07, "loss": 0.0675, "step": 565 }, { "epoch": 0.22, "learning_rate": 2.9711286089238843e-07, "loss": 0.063, "step": 566 }, { "epoch": 0.22, "learning_rate": 2.9763779527559054e-07, "loss": 0.0657, "step": 567 }, { "epoch": 0.22, "learning_rate": 2.9816272965879265e-07, "loss": 0.0714, "step": 568 }, { "epoch": 0.22, "learning_rate": 2.986876640419947e-07, "loss": 0.0591, "step": 569 }, { "epoch": 0.22, "learning_rate": 2.992125984251968e-07, "loss": 0.0603, "step": 570 }, { "epoch": 0.23, "learning_rate": 2.9973753280839893e-07, "loss": 0.0624, "step": 571 }, { "epoch": 0.23, "learning_rate": 3.0026246719160104e-07, "loss": 0.0577, "step": 572 }, { "epoch": 0.23, "learning_rate": 3.0078740157480315e-07, "loss": 0.0571, "step": 573 }, { "epoch": 0.23, "learning_rate": 3.0131233595800526e-07, "loss": 0.0594, "step": 574 }, { "epoch": 0.23, "learning_rate": 3.018372703412073e-07, "loss": 0.063, "step": 575 }, { "epoch": 0.23, "learning_rate": 3.0236220472440943e-07, "loss": 0.0633, "step": 576 }, { "epoch": 0.23, "learning_rate": 3.0288713910761154e-07, "loss": 0.0645, "step": 577 }, { "epoch": 0.23, "learning_rate": 3.034120734908136e-07, "loss": 0.0621, "step": 578 }, { "epoch": 0.23, "learning_rate": 3.039370078740157e-07, "loss": 0.0662, "step": 579 }, { "epoch": 0.23, "learning_rate": 3.044619422572178e-07, "loss": 0.0546, "step": 580 }, { "epoch": 0.23, "learning_rate": 3.0498687664041994e-07, "loss": 0.0578, "step": 581 }, { "epoch": 0.23, "learning_rate": 3.0551181102362205e-07, "loss": 0.0602, "step": 582 }, { "epoch": 0.23, "learning_rate": 3.0603674540682416e-07, "loss": 0.0633, "step": 583 }, { "epoch": 0.23, "learning_rate": 3.0656167979002627e-07, "loss": 0.0608, "step": 584 }, { "epoch": 0.23, "learning_rate": 3.0708661417322833e-07, "loss": 0.0624, "step": 585 }, { "epoch": 0.23, "learning_rate": 3.0761154855643044e-07, "loss": 0.06, "step": 586 }, { "epoch": 0.23, "learning_rate": 3.0813648293963255e-07, "loss": 0.0671, "step": 587 }, { "epoch": 0.23, "learning_rate": 3.086614173228346e-07, "loss": 0.068, "step": 588 }, { "epoch": 0.23, "learning_rate": 3.091863517060367e-07, "loss": 0.0555, "step": 589 }, { "epoch": 0.23, "learning_rate": 3.0971128608923883e-07, "loss": 0.0631, "step": 590 }, { "epoch": 0.23, "learning_rate": 3.102362204724409e-07, "loss": 0.0663, "step": 591 }, { "epoch": 0.23, "learning_rate": 3.1076115485564305e-07, "loss": 0.0608, "step": 592 }, { "epoch": 0.23, "learning_rate": 3.1128608923884516e-07, "loss": 0.0683, "step": 593 }, { "epoch": 0.23, "learning_rate": 3.118110236220472e-07, "loss": 0.0628, "step": 594 }, { "epoch": 0.23, "learning_rate": 3.1233595800524933e-07, "loss": 0.0571, "step": 595 }, { "epoch": 0.23, "learning_rate": 3.1286089238845144e-07, "loss": 0.0676, "step": 596 }, { "epoch": 0.24, "learning_rate": 3.1338582677165355e-07, "loss": 0.0609, "step": 597 }, { "epoch": 0.24, "learning_rate": 3.139107611548556e-07, "loss": 0.0653, "step": 598 }, { "epoch": 0.24, "learning_rate": 3.144356955380577e-07, "loss": 0.0629, "step": 599 }, { "epoch": 0.24, "learning_rate": 3.1496062992125984e-07, "loss": 0.0686, "step": 600 }, { "epoch": 0.24, "learning_rate": 3.154855643044619e-07, "loss": 0.0615, "step": 601 }, { "epoch": 0.24, "learning_rate": 3.16010498687664e-07, "loss": 0.0655, "step": 602 }, { "epoch": 0.24, "learning_rate": 3.1653543307086617e-07, "loss": 0.0622, "step": 603 }, { "epoch": 0.24, "learning_rate": 3.1706036745406823e-07, "loss": 0.0646, "step": 604 }, { "epoch": 0.24, "learning_rate": 3.1758530183727034e-07, "loss": 0.0594, "step": 605 }, { "epoch": 0.24, "learning_rate": 3.1811023622047245e-07, "loss": 0.0611, "step": 606 }, { "epoch": 0.24, "learning_rate": 3.186351706036745e-07, "loss": 0.0597, "step": 607 }, { "epoch": 0.24, "learning_rate": 3.191601049868766e-07, "loss": 0.0574, "step": 608 }, { "epoch": 0.24, "learning_rate": 3.1968503937007873e-07, "loss": 0.0627, "step": 609 }, { "epoch": 0.24, "learning_rate": 3.202099737532808e-07, "loss": 0.0556, "step": 610 }, { "epoch": 0.24, "learning_rate": 3.207349081364829e-07, "loss": 0.0581, "step": 611 }, { "epoch": 0.24, "learning_rate": 3.21259842519685e-07, "loss": 0.0668, "step": 612 }, { "epoch": 0.24, "learning_rate": 3.217847769028871e-07, "loss": 0.0566, "step": 613 }, { "epoch": 0.24, "learning_rate": 3.2230971128608923e-07, "loss": 0.0619, "step": 614 }, { "epoch": 0.24, "learning_rate": 3.2283464566929134e-07, "loss": 0.0671, "step": 615 }, { "epoch": 0.24, "learning_rate": 3.2335958005249345e-07, "loss": 0.0644, "step": 616 }, { "epoch": 0.24, "learning_rate": 3.238845144356955e-07, "loss": 0.0726, "step": 617 }, { "epoch": 0.24, "learning_rate": 3.244094488188976e-07, "loss": 0.0615, "step": 618 }, { "epoch": 0.24, "learning_rate": 3.2493438320209973e-07, "loss": 0.0659, "step": 619 }, { "epoch": 0.24, "learning_rate": 3.254593175853018e-07, "loss": 0.0647, "step": 620 }, { "epoch": 0.24, "learning_rate": 3.259842519685039e-07, "loss": 0.0632, "step": 621 }, { "epoch": 0.25, "learning_rate": 3.26509186351706e-07, "loss": 0.0636, "step": 622 }, { "epoch": 0.25, "learning_rate": 3.270341207349081e-07, "loss": 0.0654, "step": 623 }, { "epoch": 0.25, "learning_rate": 3.2755905511811024e-07, "loss": 0.057, "step": 624 }, { "epoch": 0.25, "learning_rate": 3.2808398950131235e-07, "loss": 0.0532, "step": 625 }, { "epoch": 0.25, "learning_rate": 3.286089238845144e-07, "loss": 0.0547, "step": 626 }, { "epoch": 0.25, "learning_rate": 3.291338582677165e-07, "loss": 0.0652, "step": 627 }, { "epoch": 0.25, "learning_rate": 3.2965879265091863e-07, "loss": 0.0644, "step": 628 }, { "epoch": 0.25, "learning_rate": 3.3018372703412074e-07, "loss": 0.0583, "step": 629 }, { "epoch": 0.25, "learning_rate": 3.307086614173228e-07, "loss": 0.0606, "step": 630 }, { "epoch": 0.25, "learning_rate": 3.312335958005249e-07, "loss": 0.0585, "step": 631 }, { "epoch": 0.25, "learning_rate": 3.31758530183727e-07, "loss": 0.0566, "step": 632 }, { "epoch": 0.25, "learning_rate": 3.3228346456692913e-07, "loss": 0.0617, "step": 633 }, { "epoch": 0.25, "learning_rate": 3.3280839895013124e-07, "loss": 0.0627, "step": 634 }, { "epoch": 0.25, "learning_rate": 3.3333333333333335e-07, "loss": 0.0581, "step": 635 }, { "epoch": 0.25, "learning_rate": 3.338582677165354e-07, "loss": 0.0612, "step": 636 }, { "epoch": 0.25, "learning_rate": 3.343832020997375e-07, "loss": 0.0549, "step": 637 }, { "epoch": 0.25, "learning_rate": 3.3490813648293963e-07, "loss": 0.0544, "step": 638 }, { "epoch": 0.25, "learning_rate": 3.354330708661417e-07, "loss": 0.0623, "step": 639 }, { "epoch": 0.25, "learning_rate": 3.359580052493438e-07, "loss": 0.0568, "step": 640 }, { "epoch": 0.25, "learning_rate": 3.364829396325459e-07, "loss": 0.0557, "step": 641 }, { "epoch": 0.25, "learning_rate": 3.3700787401574797e-07, "loss": 0.0599, "step": 642 }, { "epoch": 0.25, "learning_rate": 3.375328083989501e-07, "loss": 0.0607, "step": 643 }, { "epoch": 0.25, "learning_rate": 3.3805774278215225e-07, "loss": 0.0561, "step": 644 }, { "epoch": 0.25, "learning_rate": 3.3858267716535436e-07, "loss": 0.0643, "step": 645 }, { "epoch": 0.25, "learning_rate": 3.391076115485564e-07, "loss": 0.0564, "step": 646 }, { "epoch": 0.25, "learning_rate": 3.3963254593175853e-07, "loss": 0.0613, "step": 647 }, { "epoch": 0.26, "learning_rate": 3.4015748031496064e-07, "loss": 0.063, "step": 648 }, { "epoch": 0.26, "learning_rate": 3.406824146981627e-07, "loss": 0.0529, "step": 649 }, { "epoch": 0.26, "learning_rate": 3.412073490813648e-07, "loss": 0.0582, "step": 650 }, { "epoch": 0.26, "learning_rate": 3.417322834645669e-07, "loss": 0.0534, "step": 651 }, { "epoch": 0.26, "learning_rate": 3.42257217847769e-07, "loss": 0.0566, "step": 652 }, { "epoch": 0.26, "learning_rate": 3.427821522309711e-07, "loss": 0.0725, "step": 653 }, { "epoch": 0.26, "learning_rate": 3.433070866141732e-07, "loss": 0.0584, "step": 654 }, { "epoch": 0.26, "learning_rate": 3.438320209973753e-07, "loss": 0.0575, "step": 655 }, { "epoch": 0.26, "learning_rate": 3.443569553805774e-07, "loss": 0.0636, "step": 656 }, { "epoch": 0.26, "learning_rate": 3.4488188976377953e-07, "loss": 0.062, "step": 657 }, { "epoch": 0.26, "learning_rate": 3.454068241469816e-07, "loss": 0.0583, "step": 658 }, { "epoch": 0.26, "learning_rate": 3.459317585301837e-07, "loss": 0.0588, "step": 659 }, { "epoch": 0.26, "learning_rate": 3.464566929133858e-07, "loss": 0.0608, "step": 660 }, { "epoch": 0.26, "learning_rate": 3.469816272965879e-07, "loss": 0.0605, "step": 661 }, { "epoch": 0.26, "learning_rate": 3.4750656167979e-07, "loss": 0.0607, "step": 662 }, { "epoch": 0.26, "learning_rate": 3.480314960629921e-07, "loss": 0.0626, "step": 663 }, { "epoch": 0.26, "learning_rate": 3.485564304461942e-07, "loss": 0.0515, "step": 664 }, { "epoch": 0.26, "learning_rate": 3.490813648293963e-07, "loss": 0.0559, "step": 665 }, { "epoch": 0.26, "learning_rate": 3.4960629921259843e-07, "loss": 0.0617, "step": 666 }, { "epoch": 0.26, "learning_rate": 3.5013123359580054e-07, "loss": 0.0571, "step": 667 }, { "epoch": 0.26, "learning_rate": 3.506561679790026e-07, "loss": 0.0533, "step": 668 }, { "epoch": 0.26, "learning_rate": 3.511811023622047e-07, "loss": 0.0611, "step": 669 }, { "epoch": 0.26, "learning_rate": 3.517060367454068e-07, "loss": 0.0567, "step": 670 }, { "epoch": 0.26, "learning_rate": 3.522309711286089e-07, "loss": 0.0618, "step": 671 }, { "epoch": 0.26, "learning_rate": 3.52755905511811e-07, "loss": 0.0561, "step": 672 }, { "epoch": 0.27, "learning_rate": 3.532808398950131e-07, "loss": 0.0626, "step": 673 }, { "epoch": 0.27, "learning_rate": 3.538057742782152e-07, "loss": 0.0584, "step": 674 }, { "epoch": 0.27, "learning_rate": 3.543307086614173e-07, "loss": 0.064, "step": 675 }, { "epoch": 0.27, "learning_rate": 3.5485564304461943e-07, "loss": 0.0593, "step": 676 }, { "epoch": 0.27, "learning_rate": 3.5538057742782154e-07, "loss": 0.0554, "step": 677 }, { "epoch": 0.27, "learning_rate": 3.559055118110236e-07, "loss": 0.0665, "step": 678 }, { "epoch": 0.27, "learning_rate": 3.564304461942257e-07, "loss": 0.0574, "step": 679 }, { "epoch": 0.27, "learning_rate": 3.569553805774278e-07, "loss": 0.0593, "step": 680 }, { "epoch": 0.27, "learning_rate": 3.574803149606299e-07, "loss": 0.0572, "step": 681 }, { "epoch": 0.27, "learning_rate": 3.58005249343832e-07, "loss": 0.0616, "step": 682 }, { "epoch": 0.27, "learning_rate": 3.585301837270341e-07, "loss": 0.0614, "step": 683 }, { "epoch": 0.27, "learning_rate": 3.5905511811023616e-07, "loss": 0.0544, "step": 684 }, { "epoch": 0.27, "learning_rate": 3.5958005249343833e-07, "loss": 0.0567, "step": 685 }, { "epoch": 0.27, "learning_rate": 3.6010498687664044e-07, "loss": 0.062, "step": 686 }, { "epoch": 0.27, "learning_rate": 3.606299212598425e-07, "loss": 0.0571, "step": 687 }, { "epoch": 0.27, "learning_rate": 3.611548556430446e-07, "loss": 0.0592, "step": 688 }, { "epoch": 0.27, "learning_rate": 3.616797900262467e-07, "loss": 0.0581, "step": 689 }, { "epoch": 0.27, "learning_rate": 3.622047244094488e-07, "loss": 0.0539, "step": 690 }, { "epoch": 0.27, "learning_rate": 3.627296587926509e-07, "loss": 0.0602, "step": 691 }, { "epoch": 0.27, "learning_rate": 3.63254593175853e-07, "loss": 0.0616, "step": 692 }, { "epoch": 0.27, "learning_rate": 3.637795275590551e-07, "loss": 0.0657, "step": 693 }, { "epoch": 0.27, "learning_rate": 3.6430446194225717e-07, "loss": 0.0588, "step": 694 }, { "epoch": 0.27, "learning_rate": 3.648293963254593e-07, "loss": 0.0582, "step": 695 }, { "epoch": 0.27, "learning_rate": 3.6535433070866144e-07, "loss": 0.0613, "step": 696 }, { "epoch": 0.27, "learning_rate": 3.658792650918635e-07, "loss": 0.0546, "step": 697 }, { "epoch": 0.28, "learning_rate": 3.664041994750656e-07, "loss": 0.0584, "step": 698 }, { "epoch": 0.28, "learning_rate": 3.669291338582677e-07, "loss": 0.0595, "step": 699 }, { "epoch": 0.28, "learning_rate": 3.674540682414698e-07, "loss": 0.0561, "step": 700 }, { "epoch": 0.28, "learning_rate": 3.679790026246719e-07, "loss": 0.0539, "step": 701 }, { "epoch": 0.28, "learning_rate": 3.68503937007874e-07, "loss": 0.0575, "step": 702 }, { "epoch": 0.28, "learning_rate": 3.6902887139107606e-07, "loss": 0.0587, "step": 703 }, { "epoch": 0.28, "learning_rate": 3.6955380577427817e-07, "loss": 0.0627, "step": 704 }, { "epoch": 0.28, "learning_rate": 3.700787401574803e-07, "loss": 0.0566, "step": 705 }, { "epoch": 0.28, "learning_rate": 3.706036745406824e-07, "loss": 0.0582, "step": 706 }, { "epoch": 0.28, "learning_rate": 3.711286089238845e-07, "loss": 0.0576, "step": 707 }, { "epoch": 0.28, "learning_rate": 3.716535433070866e-07, "loss": 0.0606, "step": 708 }, { "epoch": 0.28, "learning_rate": 3.7217847769028873e-07, "loss": 0.0533, "step": 709 }, { "epoch": 0.28, "learning_rate": 3.727034120734908e-07, "loss": 0.0556, "step": 710 }, { "epoch": 0.28, "learning_rate": 3.732283464566929e-07, "loss": 0.0565, "step": 711 }, { "epoch": 0.28, "learning_rate": 3.73753280839895e-07, "loss": 0.059, "step": 712 }, { "epoch": 0.28, "learning_rate": 3.7427821522309707e-07, "loss": 0.0542, "step": 713 }, { "epoch": 0.28, "learning_rate": 3.748031496062992e-07, "loss": 0.0585, "step": 714 }, { "epoch": 0.28, "learning_rate": 3.753280839895013e-07, "loss": 0.0649, "step": 715 }, { "epoch": 0.28, "learning_rate": 3.758530183727034e-07, "loss": 0.071, "step": 716 }, { "epoch": 0.28, "learning_rate": 3.763779527559055e-07, "loss": 0.063, "step": 717 }, { "epoch": 0.28, "learning_rate": 3.769028871391076e-07, "loss": 0.0612, "step": 718 }, { "epoch": 0.28, "learning_rate": 3.774278215223097e-07, "loss": 0.0579, "step": 719 }, { "epoch": 0.28, "learning_rate": 3.779527559055118e-07, "loss": 0.0561, "step": 720 }, { "epoch": 0.28, "learning_rate": 3.784776902887139e-07, "loss": 0.0568, "step": 721 }, { "epoch": 0.28, "learning_rate": 3.7900262467191596e-07, "loss": 0.0557, "step": 722 }, { "epoch": 0.28, "learning_rate": 3.7952755905511807e-07, "loss": 0.06, "step": 723 }, { "epoch": 0.29, "learning_rate": 3.800524934383202e-07, "loss": 0.0566, "step": 724 }, { "epoch": 0.29, "learning_rate": 3.805774278215223e-07, "loss": 0.0591, "step": 725 }, { "epoch": 0.29, "learning_rate": 3.811023622047244e-07, "loss": 0.0594, "step": 726 }, { "epoch": 0.29, "learning_rate": 3.816272965879265e-07, "loss": 0.0592, "step": 727 }, { "epoch": 0.29, "learning_rate": 3.8215223097112863e-07, "loss": 0.0596, "step": 728 }, { "epoch": 0.29, "learning_rate": 3.826771653543307e-07, "loss": 0.0587, "step": 729 }, { "epoch": 0.29, "learning_rate": 3.832020997375328e-07, "loss": 0.055, "step": 730 }, { "epoch": 0.29, "learning_rate": 3.837270341207349e-07, "loss": 0.0597, "step": 731 }, { "epoch": 0.29, "learning_rate": 3.8425196850393697e-07, "loss": 0.0595, "step": 732 }, { "epoch": 0.29, "learning_rate": 3.847769028871391e-07, "loss": 0.0544, "step": 733 }, { "epoch": 0.29, "learning_rate": 3.853018372703412e-07, "loss": 0.0538, "step": 734 }, { "epoch": 0.29, "learning_rate": 3.8582677165354325e-07, "loss": 0.0544, "step": 735 }, { "epoch": 0.29, "learning_rate": 3.8635170603674536e-07, "loss": 0.0593, "step": 736 }, { "epoch": 0.29, "learning_rate": 3.868766404199475e-07, "loss": 0.0504, "step": 737 }, { "epoch": 0.29, "learning_rate": 3.874015748031496e-07, "loss": 0.066, "step": 738 }, { "epoch": 0.29, "learning_rate": 3.879265091863517e-07, "loss": 0.0541, "step": 739 }, { "epoch": 0.29, "learning_rate": 3.884514435695538e-07, "loss": 0.0586, "step": 740 }, { "epoch": 0.29, "learning_rate": 3.889763779527559e-07, "loss": 0.0599, "step": 741 }, { "epoch": 0.29, "learning_rate": 3.8950131233595797e-07, "loss": 0.0587, "step": 742 }, { "epoch": 0.29, "learning_rate": 3.900262467191601e-07, "loss": 0.0549, "step": 743 }, { "epoch": 0.29, "learning_rate": 3.905511811023622e-07, "loss": 0.0581, "step": 744 }, { "epoch": 0.29, "learning_rate": 3.9107611548556425e-07, "loss": 0.0607, "step": 745 }, { "epoch": 0.29, "learning_rate": 3.9160104986876636e-07, "loss": 0.0602, "step": 746 }, { "epoch": 0.29, "learning_rate": 3.921259842519685e-07, "loss": 0.0611, "step": 747 }, { "epoch": 0.29, "learning_rate": 3.926509186351706e-07, "loss": 0.0522, "step": 748 }, { "epoch": 0.3, "learning_rate": 3.931758530183727e-07, "loss": 0.0615, "step": 749 }, { "epoch": 0.3, "learning_rate": 3.937007874015748e-07, "loss": 0.0586, "step": 750 }, { "epoch": 0.3, "learning_rate": 3.9422572178477687e-07, "loss": 0.0543, "step": 751 }, { "epoch": 0.3, "learning_rate": 3.94750656167979e-07, "loss": 0.0521, "step": 752 }, { "epoch": 0.3, "learning_rate": 3.952755905511811e-07, "loss": 0.0582, "step": 753 }, { "epoch": 0.3, "learning_rate": 3.9580052493438315e-07, "loss": 0.0581, "step": 754 }, { "epoch": 0.3, "learning_rate": 3.9632545931758526e-07, "loss": 0.0532, "step": 755 }, { "epoch": 0.3, "learning_rate": 3.9685039370078737e-07, "loss": 0.0588, "step": 756 }, { "epoch": 0.3, "learning_rate": 3.973753280839895e-07, "loss": 0.0653, "step": 757 }, { "epoch": 0.3, "learning_rate": 3.979002624671916e-07, "loss": 0.0619, "step": 758 }, { "epoch": 0.3, "learning_rate": 3.984251968503937e-07, "loss": 0.0591, "step": 759 }, { "epoch": 0.3, "learning_rate": 3.989501312335958e-07, "loss": 0.0595, "step": 760 }, { "epoch": 0.3, "learning_rate": 3.9947506561679787e-07, "loss": 0.0552, "step": 761 }, { "epoch": 0.3, "learning_rate": 4e-07, "loss": 0.06, "step": 762 }, { "epoch": 0.3, "learning_rate": 3.99999998370152e-07, "loss": 0.0595, "step": 763 }, { "epoch": 0.3, "learning_rate": 3.9999999348060793e-07, "loss": 0.0541, "step": 764 }, { "epoch": 0.3, "learning_rate": 3.99999985331368e-07, "loss": 0.055, "step": 765 }, { "epoch": 0.3, "learning_rate": 3.999999739224322e-07, "loss": 0.0577, "step": 766 }, { "epoch": 0.3, "learning_rate": 3.9999995925380087e-07, "loss": 0.0586, "step": 767 }, { "epoch": 0.3, "learning_rate": 3.9999994132547414e-07, "loss": 0.0652, "step": 768 }, { "epoch": 0.3, "learning_rate": 3.9999992013745233e-07, "loss": 0.0568, "step": 769 }, { "epoch": 0.3, "learning_rate": 3.999998956897358e-07, "loss": 0.0504, "step": 770 }, { "epoch": 0.3, "learning_rate": 3.999998679823249e-07, "loss": 0.0564, "step": 771 }, { "epoch": 0.3, "learning_rate": 3.999998370152202e-07, "loss": 0.0636, "step": 772 }, { "epoch": 0.3, "learning_rate": 3.99999802788422e-07, "loss": 0.0578, "step": 773 }, { "epoch": 0.31, "learning_rate": 3.9999976530193106e-07, "loss": 0.0512, "step": 774 }, { "epoch": 0.31, "learning_rate": 3.999997245557479e-07, "loss": 0.059, "step": 775 }, { "epoch": 0.31, "learning_rate": 3.999996805498732e-07, "loss": 0.0521, "step": 776 }, { "epoch": 0.31, "learning_rate": 3.999996332843076e-07, "loss": 0.06, "step": 777 }, { "epoch": 0.31, "learning_rate": 3.99999582759052e-07, "loss": 0.0561, "step": 778 }, { "epoch": 0.31, "learning_rate": 3.999995289741072e-07, "loss": 0.0557, "step": 779 }, { "epoch": 0.31, "learning_rate": 3.9999947192947394e-07, "loss": 0.0559, "step": 780 }, { "epoch": 0.31, "learning_rate": 3.9999941162515337e-07, "loss": 0.0582, "step": 781 }, { "epoch": 0.31, "learning_rate": 3.999993480611463e-07, "loss": 0.0541, "step": 782 }, { "epoch": 0.31, "learning_rate": 3.999992812374538e-07, "loss": 0.0529, "step": 783 }, { "epoch": 0.31, "learning_rate": 3.99999211154077e-07, "loss": 0.0548, "step": 784 }, { "epoch": 0.31, "learning_rate": 3.9999913781101703e-07, "loss": 0.0589, "step": 785 }, { "epoch": 0.31, "learning_rate": 3.999990612082751e-07, "loss": 0.0583, "step": 786 }, { "epoch": 0.31, "learning_rate": 3.999989813458524e-07, "loss": 0.0579, "step": 787 }, { "epoch": 0.31, "learning_rate": 3.999988982237502e-07, "loss": 0.0576, "step": 788 }, { "epoch": 0.31, "learning_rate": 3.999988118419701e-07, "loss": 0.057, "step": 789 }, { "epoch": 0.31, "learning_rate": 3.9999872220051314e-07, "loss": 0.0557, "step": 790 }, { "epoch": 0.31, "learning_rate": 3.9999862929938113e-07, "loss": 0.0539, "step": 791 }, { "epoch": 0.31, "learning_rate": 3.999985331385753e-07, "loss": 0.0572, "step": 792 }, { "epoch": 0.31, "learning_rate": 3.999984337180974e-07, "loss": 0.0531, "step": 793 }, { "epoch": 0.31, "learning_rate": 3.99998331037949e-07, "loss": 0.0526, "step": 794 }, { "epoch": 0.31, "learning_rate": 3.9999822509813173e-07, "loss": 0.0604, "step": 795 }, { "epoch": 0.31, "learning_rate": 3.9999811589864735e-07, "loss": 0.0519, "step": 796 }, { "epoch": 0.31, "learning_rate": 3.999980034394977e-07, "loss": 0.0549, "step": 797 }, { "epoch": 0.31, "learning_rate": 3.9999788772068453e-07, "loss": 0.0563, "step": 798 }, { "epoch": 0.31, "learning_rate": 3.999977687422097e-07, "loss": 0.0522, "step": 799 }, { "epoch": 0.32, "learning_rate": 3.999976465040753e-07, "loss": 0.0522, "step": 800 }, { "epoch": 0.32, "learning_rate": 3.999975210062832e-07, "loss": 0.0504, "step": 801 }, { "epoch": 0.32, "learning_rate": 3.999973922488354e-07, "loss": 0.0563, "step": 802 }, { "epoch": 0.32, "learning_rate": 3.9999726023173413e-07, "loss": 0.0537, "step": 803 }, { "epoch": 0.32, "learning_rate": 3.9999712495498145e-07, "loss": 0.058, "step": 804 }, { "epoch": 0.32, "learning_rate": 3.9999698641857955e-07, "loss": 0.0622, "step": 805 }, { "epoch": 0.32, "learning_rate": 3.999968446225308e-07, "loss": 0.0538, "step": 806 }, { "epoch": 0.32, "learning_rate": 3.999966995668375e-07, "loss": 0.0647, "step": 807 }, { "epoch": 0.32, "learning_rate": 3.9999655125150183e-07, "loss": 0.0598, "step": 808 }, { "epoch": 0.32, "learning_rate": 3.9999639967652643e-07, "loss": 0.0594, "step": 809 }, { "epoch": 0.32, "learning_rate": 3.9999624484191365e-07, "loss": 0.0567, "step": 810 }, { "epoch": 0.32, "learning_rate": 3.99996086747666e-07, "loss": 0.0562, "step": 811 }, { "epoch": 0.32, "learning_rate": 3.9999592539378606e-07, "loss": 0.0615, "step": 812 }, { "epoch": 0.32, "learning_rate": 3.9999576078027654e-07, "loss": 0.0592, "step": 813 }, { "epoch": 0.32, "learning_rate": 3.9999559290714013e-07, "loss": 0.0526, "step": 814 }, { "epoch": 0.32, "learning_rate": 3.999954217743795e-07, "loss": 0.0503, "step": 815 }, { "epoch": 0.32, "learning_rate": 3.9999524738199735e-07, "loss": 0.0575, "step": 816 }, { "epoch": 0.32, "learning_rate": 3.9999506972999675e-07, "loss": 0.0624, "step": 817 }, { "epoch": 0.32, "learning_rate": 3.9999488881838044e-07, "loss": 0.0528, "step": 818 }, { "epoch": 0.32, "learning_rate": 3.999947046471513e-07, "loss": 0.0535, "step": 819 }, { "epoch": 0.32, "learning_rate": 3.9999451721631257e-07, "loss": 0.0595, "step": 820 }, { "epoch": 0.32, "learning_rate": 3.999943265258671e-07, "loss": 0.0563, "step": 821 }, { "epoch": 0.32, "learning_rate": 3.999941325758181e-07, "loss": 0.0492, "step": 822 }, { "epoch": 0.32, "learning_rate": 3.999939353661687e-07, "loss": 0.0585, "step": 823 }, { "epoch": 0.32, "learning_rate": 3.9999373489692204e-07, "loss": 0.0645, "step": 824 }, { "epoch": 0.33, "learning_rate": 3.999935311680815e-07, "loss": 0.0559, "step": 825 }, { "epoch": 0.33, "learning_rate": 3.999933241796503e-07, "loss": 0.0593, "step": 826 }, { "epoch": 0.33, "learning_rate": 3.9999311393163194e-07, "loss": 0.051, "step": 827 }, { "epoch": 0.33, "learning_rate": 3.9999290042402973e-07, "loss": 0.0586, "step": 828 }, { "epoch": 0.33, "learning_rate": 3.999926836568472e-07, "loss": 0.0577, "step": 829 }, { "epoch": 0.33, "learning_rate": 3.999924636300879e-07, "loss": 0.0539, "step": 830 }, { "epoch": 0.33, "learning_rate": 3.999922403437555e-07, "loss": 0.0619, "step": 831 }, { "epoch": 0.33, "learning_rate": 3.999920137978534e-07, "loss": 0.053, "step": 832 }, { "epoch": 0.33, "learning_rate": 3.999917839923855e-07, "loss": 0.0513, "step": 833 }, { "epoch": 0.33, "learning_rate": 3.999915509273554e-07, "loss": 0.052, "step": 834 }, { "epoch": 0.33, "learning_rate": 3.99991314602767e-07, "loss": 0.0503, "step": 835 }, { "epoch": 0.33, "learning_rate": 3.9999107501862414e-07, "loss": 0.0598, "step": 836 }, { "epoch": 0.33, "learning_rate": 3.9999083217493075e-07, "loss": 0.0559, "step": 837 }, { "epoch": 0.33, "learning_rate": 3.999905860716906e-07, "loss": 0.0562, "step": 838 }, { "epoch": 0.33, "learning_rate": 3.9999033670890805e-07, "loss": 0.0581, "step": 839 }, { "epoch": 0.33, "learning_rate": 3.9999008408658684e-07, "loss": 0.0596, "step": 840 }, { "epoch": 0.33, "learning_rate": 3.9998982820473123e-07, "loss": 0.057, "step": 841 }, { "epoch": 0.33, "learning_rate": 3.9998956906334535e-07, "loss": 0.0573, "step": 842 }, { "epoch": 0.33, "learning_rate": 3.9998930666243344e-07, "loss": 0.0497, "step": 843 }, { "epoch": 0.33, "learning_rate": 3.9998904100199983e-07, "loss": 0.0473, "step": 844 }, { "epoch": 0.33, "learning_rate": 3.9998877208204876e-07, "loss": 0.0601, "step": 845 }, { "epoch": 0.33, "learning_rate": 3.9998849990258463e-07, "loss": 0.0588, "step": 846 }, { "epoch": 0.33, "learning_rate": 3.99988224463612e-07, "loss": 0.059, "step": 847 }, { "epoch": 0.33, "learning_rate": 3.9998794576513517e-07, "loss": 0.0531, "step": 848 }, { "epoch": 0.33, "learning_rate": 3.999876638071588e-07, "loss": 0.0571, "step": 849 }, { "epoch": 0.33, "learning_rate": 3.9998737858968744e-07, "loss": 0.0506, "step": 850 }, { "epoch": 0.34, "learning_rate": 3.9998709011272574e-07, "loss": 0.0518, "step": 851 }, { "epoch": 0.34, "learning_rate": 3.999867983762785e-07, "loss": 0.0533, "step": 852 }, { "epoch": 0.34, "learning_rate": 3.999865033803503e-07, "loss": 0.0517, "step": 853 }, { "epoch": 0.34, "learning_rate": 3.9998620512494606e-07, "loss": 0.0502, "step": 854 }, { "epoch": 0.34, "learning_rate": 3.9998590361007065e-07, "loss": 0.0598, "step": 855 }, { "epoch": 0.34, "learning_rate": 3.999855988357289e-07, "loss": 0.0535, "step": 856 }, { "epoch": 0.34, "learning_rate": 3.9998529080192594e-07, "loss": 0.0549, "step": 857 }, { "epoch": 0.34, "learning_rate": 3.999849795086666e-07, "loss": 0.0573, "step": 858 }, { "epoch": 0.34, "learning_rate": 3.9998466495595603e-07, "loss": 0.055, "step": 859 }, { "epoch": 0.34, "learning_rate": 3.9998434714379936e-07, "loss": 0.0511, "step": 860 }, { "epoch": 0.34, "learning_rate": 3.999840260722018e-07, "loss": 0.0525, "step": 861 }, { "epoch": 0.34, "learning_rate": 3.9998370174116854e-07, "loss": 0.0512, "step": 862 }, { "epoch": 0.34, "learning_rate": 3.999833741507049e-07, "loss": 0.0569, "step": 863 }, { "epoch": 0.34, "learning_rate": 3.9998304330081617e-07, "loss": 0.0554, "step": 864 }, { "epoch": 0.34, "learning_rate": 3.9998270919150784e-07, "loss": 0.056, "step": 865 }, { "epoch": 0.34, "learning_rate": 3.9998237182278524e-07, "loss": 0.053, "step": 866 }, { "epoch": 0.34, "learning_rate": 3.999820311946539e-07, "loss": 0.0564, "step": 867 }, { "epoch": 0.34, "learning_rate": 3.9998168730711937e-07, "loss": 0.0503, "step": 868 }, { "epoch": 0.34, "learning_rate": 3.999813401601874e-07, "loss": 0.0476, "step": 869 }, { "epoch": 0.34, "learning_rate": 3.9998098975386345e-07, "loss": 0.0511, "step": 870 }, { "epoch": 0.34, "learning_rate": 3.9998063608815325e-07, "loss": 0.0543, "step": 871 }, { "epoch": 0.34, "learning_rate": 3.9998027916306265e-07, "loss": 0.053, "step": 872 }, { "epoch": 0.34, "learning_rate": 3.9997991897859747e-07, "loss": 0.0497, "step": 873 }, { "epoch": 0.34, "learning_rate": 3.999795555347635e-07, "loss": 0.0563, "step": 874 }, { "epoch": 0.34, "learning_rate": 3.9997918883156674e-07, "loss": 0.0521, "step": 875 }, { "epoch": 0.35, "learning_rate": 3.999788188690131e-07, "loss": 0.0522, "step": 876 }, { "epoch": 0.35, "learning_rate": 3.9997844564710864e-07, "loss": 0.0514, "step": 877 }, { "epoch": 0.35, "learning_rate": 3.999780691658595e-07, "loss": 0.05, "step": 878 }, { "epoch": 0.35, "learning_rate": 3.9997768942527173e-07, "loss": 0.0521, "step": 879 }, { "epoch": 0.35, "learning_rate": 3.9997730642535155e-07, "loss": 0.0494, "step": 880 }, { "epoch": 0.35, "learning_rate": 3.9997692016610524e-07, "loss": 0.0497, "step": 881 }, { "epoch": 0.35, "learning_rate": 3.9997653064753903e-07, "loss": 0.0542, "step": 882 }, { "epoch": 0.35, "learning_rate": 3.9997613786965934e-07, "loss": 0.0507, "step": 883 }, { "epoch": 0.35, "learning_rate": 3.9997574183247247e-07, "loss": 0.0531, "step": 884 }, { "epoch": 0.35, "learning_rate": 3.9997534253598493e-07, "loss": 0.0577, "step": 885 }, { "epoch": 0.35, "learning_rate": 3.9997493998020333e-07, "loss": 0.0547, "step": 886 }, { "epoch": 0.35, "learning_rate": 3.999745341651341e-07, "loss": 0.0565, "step": 887 }, { "epoch": 0.35, "learning_rate": 3.999741250907838e-07, "loss": 0.0546, "step": 888 }, { "epoch": 0.35, "learning_rate": 3.999737127571593e-07, "loss": 0.0528, "step": 889 }, { "epoch": 0.35, "learning_rate": 3.999732971642671e-07, "loss": 0.0563, "step": 890 }, { "epoch": 0.35, "learning_rate": 3.999728783121142e-07, "loss": 0.0522, "step": 891 }, { "epoch": 0.35, "learning_rate": 3.999724562007072e-07, "loss": 0.0519, "step": 892 }, { "epoch": 0.35, "learning_rate": 3.9997203083005317e-07, "loss": 0.0511, "step": 893 }, { "epoch": 0.35, "learning_rate": 3.999716022001589e-07, "loss": 0.0566, "step": 894 }, { "epoch": 0.35, "learning_rate": 3.9997117031103146e-07, "loss": 0.0557, "step": 895 }, { "epoch": 0.35, "learning_rate": 3.9997073516267783e-07, "loss": 0.06, "step": 896 }, { "epoch": 0.35, "learning_rate": 3.999702967551052e-07, "loss": 0.0549, "step": 897 }, { "epoch": 0.35, "learning_rate": 3.9996985508832066e-07, "loss": 0.0578, "step": 898 }, { "epoch": 0.35, "learning_rate": 3.9996941016233135e-07, "loss": 0.0499, "step": 899 }, { "epoch": 0.35, "learning_rate": 3.9996896197714456e-07, "loss": 0.0553, "step": 900 }, { "epoch": 0.36, "learning_rate": 3.999685105327677e-07, "loss": 0.055, "step": 901 }, { "epoch": 0.36, "learning_rate": 3.999680558292079e-07, "loss": 0.0517, "step": 902 }, { "epoch": 0.36, "learning_rate": 3.999675978664728e-07, "loss": 0.0539, "step": 903 }, { "epoch": 0.36, "learning_rate": 3.9996713664456977e-07, "loss": 0.0568, "step": 904 }, { "epoch": 0.36, "learning_rate": 3.999666721635063e-07, "loss": 0.062, "step": 905 }, { "epoch": 0.36, "learning_rate": 3.9996620442328994e-07, "loss": 0.0531, "step": 906 }, { "epoch": 0.36, "learning_rate": 3.9996573342392843e-07, "loss": 0.0521, "step": 907 }, { "epoch": 0.36, "learning_rate": 3.9996525916542935e-07, "loss": 0.051, "step": 908 }, { "epoch": 0.36, "learning_rate": 3.9996478164780055e-07, "loss": 0.0434, "step": 909 }, { "epoch": 0.36, "learning_rate": 3.999643008710496e-07, "loss": 0.0567, "step": 910 }, { "epoch": 0.36, "learning_rate": 3.999638168351845e-07, "loss": 0.0496, "step": 911 }, { "epoch": 0.36, "learning_rate": 3.999633295402131e-07, "loss": 0.0506, "step": 912 }, { "epoch": 0.36, "learning_rate": 3.999628389861433e-07, "loss": 0.0523, "step": 913 }, { "epoch": 0.36, "learning_rate": 3.9996234517298316e-07, "loss": 0.0514, "step": 914 }, { "epoch": 0.36, "learning_rate": 3.999618481007407e-07, "loss": 0.0586, "step": 915 }, { "epoch": 0.36, "learning_rate": 3.9996134776942404e-07, "loss": 0.0567, "step": 916 }, { "epoch": 0.36, "learning_rate": 3.999608441790413e-07, "loss": 0.0534, "step": 917 }, { "epoch": 0.36, "learning_rate": 3.999603373296008e-07, "loss": 0.0573, "step": 918 }, { "epoch": 0.36, "learning_rate": 3.999598272211106e-07, "loss": 0.0484, "step": 919 }, { "epoch": 0.36, "learning_rate": 3.999593138535791e-07, "loss": 0.0526, "step": 920 }, { "epoch": 0.36, "learning_rate": 3.9995879722701475e-07, "loss": 0.0537, "step": 921 }, { "epoch": 0.36, "learning_rate": 3.999582773414259e-07, "loss": 0.0591, "step": 922 }, { "epoch": 0.36, "learning_rate": 3.99957754196821e-07, "loss": 0.0565, "step": 923 }, { "epoch": 0.36, "learning_rate": 3.999572277932086e-07, "loss": 0.0594, "step": 924 }, { "epoch": 0.36, "learning_rate": 3.999566981305973e-07, "loss": 0.0518, "step": 925 }, { "epoch": 0.36, "learning_rate": 3.999561652089958e-07, "loss": 0.0531, "step": 926 }, { "epoch": 0.37, "learning_rate": 3.9995562902841257e-07, "loss": 0.051, "step": 927 }, { "epoch": 0.37, "learning_rate": 3.999550895888566e-07, "loss": 0.0524, "step": 928 }, { "epoch": 0.37, "learning_rate": 3.9995454689033645e-07, "loss": 0.0471, "step": 929 }, { "epoch": 0.37, "learning_rate": 3.999540009328612e-07, "loss": 0.0522, "step": 930 }, { "epoch": 0.37, "learning_rate": 3.9995345171643953e-07, "loss": 0.0507, "step": 931 }, { "epoch": 0.37, "learning_rate": 3.9995289924108057e-07, "loss": 0.0543, "step": 932 }, { "epoch": 0.37, "learning_rate": 3.9995234350679316e-07, "loss": 0.0551, "step": 933 }, { "epoch": 0.37, "learning_rate": 3.999517845135865e-07, "loss": 0.0599, "step": 934 }, { "epoch": 0.37, "learning_rate": 3.9995122226146966e-07, "loss": 0.0482, "step": 935 }, { "epoch": 0.37, "learning_rate": 3.999506567504518e-07, "loss": 0.0586, "step": 936 }, { "epoch": 0.37, "learning_rate": 3.9995008798054204e-07, "loss": 0.0568, "step": 937 }, { "epoch": 0.37, "learning_rate": 3.999495159517498e-07, "loss": 0.056, "step": 938 }, { "epoch": 0.37, "learning_rate": 3.9994894066408436e-07, "loss": 0.0549, "step": 939 }, { "epoch": 0.37, "learning_rate": 3.9994836211755506e-07, "loss": 0.0537, "step": 940 }, { "epoch": 0.37, "learning_rate": 3.9994778031217126e-07, "loss": 0.0594, "step": 941 }, { "epoch": 0.37, "learning_rate": 3.9994719524794266e-07, "loss": 0.0518, "step": 942 }, { "epoch": 0.37, "learning_rate": 3.9994660692487856e-07, "loss": 0.0489, "step": 943 }, { "epoch": 0.37, "learning_rate": 3.999460153429887e-07, "loss": 0.0624, "step": 944 }, { "epoch": 0.37, "learning_rate": 3.9994542050228264e-07, "loss": 0.0525, "step": 945 }, { "epoch": 0.37, "learning_rate": 3.9994482240277014e-07, "loss": 0.0472, "step": 946 }, { "epoch": 0.37, "learning_rate": 3.9994422104446095e-07, "loss": 0.0531, "step": 947 }, { "epoch": 0.37, "learning_rate": 3.999436164273647e-07, "loss": 0.06, "step": 948 }, { "epoch": 0.37, "learning_rate": 3.9994300855149153e-07, "loss": 0.0552, "step": 949 }, { "epoch": 0.37, "learning_rate": 3.999423974168511e-07, "loss": 0.0512, "step": 950 }, { "epoch": 0.37, "learning_rate": 3.999417830234535e-07, "loss": 0.0534, "step": 951 }, { "epoch": 0.38, "learning_rate": 3.999411653713087e-07, "loss": 0.0553, "step": 952 }, { "epoch": 0.38, "learning_rate": 3.999405444604268e-07, "loss": 0.0536, "step": 953 }, { "epoch": 0.38, "learning_rate": 3.9993992029081796e-07, "loss": 0.0541, "step": 954 }, { "epoch": 0.38, "learning_rate": 3.9993929286249224e-07, "loss": 0.0559, "step": 955 }, { "epoch": 0.38, "learning_rate": 3.999386621754599e-07, "loss": 0.0554, "step": 956 }, { "epoch": 0.38, "learning_rate": 3.999380282297313e-07, "loss": 0.0541, "step": 957 }, { "epoch": 0.38, "learning_rate": 3.9993739102531663e-07, "loss": 0.0543, "step": 958 }, { "epoch": 0.38, "learning_rate": 3.9993675056222636e-07, "loss": 0.0505, "step": 959 }, { "epoch": 0.38, "learning_rate": 3.9993610684047096e-07, "loss": 0.0472, "step": 960 }, { "epoch": 0.38, "learning_rate": 3.999354598600609e-07, "loss": 0.0532, "step": 961 }, { "epoch": 0.38, "learning_rate": 3.999348096210067e-07, "loss": 0.0477, "step": 962 }, { "epoch": 0.38, "learning_rate": 3.9993415612331894e-07, "loss": 0.0544, "step": 963 }, { "epoch": 0.38, "learning_rate": 3.999334993670083e-07, "loss": 0.055, "step": 964 }, { "epoch": 0.38, "learning_rate": 3.999328393520855e-07, "loss": 0.0531, "step": 965 }, { "epoch": 0.38, "learning_rate": 3.999321760785613e-07, "loss": 0.0544, "step": 966 }, { "epoch": 0.38, "learning_rate": 3.999315095464465e-07, "loss": 0.0568, "step": 967 }, { "epoch": 0.38, "learning_rate": 3.9993083975575195e-07, "loss": 0.0504, "step": 968 }, { "epoch": 0.38, "learning_rate": 3.9993016670648856e-07, "loss": 0.0563, "step": 969 }, { "epoch": 0.38, "learning_rate": 3.999294903986673e-07, "loss": 0.0535, "step": 970 }, { "epoch": 0.38, "learning_rate": 3.999288108322992e-07, "loss": 0.0496, "step": 971 }, { "epoch": 0.38, "learning_rate": 3.9992812800739533e-07, "loss": 0.0448, "step": 972 }, { "epoch": 0.38, "learning_rate": 3.999274419239668e-07, "loss": 0.0572, "step": 973 }, { "epoch": 0.38, "learning_rate": 3.999267525820249e-07, "loss": 0.0516, "step": 974 }, { "epoch": 0.38, "learning_rate": 3.999260599815808e-07, "loss": 0.0597, "step": 975 }, { "epoch": 0.38, "learning_rate": 3.999253641226457e-07, "loss": 0.0521, "step": 976 }, { "epoch": 0.39, "learning_rate": 3.999246650052311e-07, "loss": 0.0523, "step": 977 }, { "epoch": 0.39, "learning_rate": 3.9992396262934826e-07, "loss": 0.0497, "step": 978 }, { "epoch": 0.39, "learning_rate": 3.9992325699500865e-07, "loss": 0.0483, "step": 979 }, { "epoch": 0.39, "learning_rate": 3.9992254810222383e-07, "loss": 0.0496, "step": 980 }, { "epoch": 0.39, "learning_rate": 3.999218359510053e-07, "loss": 0.0529, "step": 981 }, { "epoch": 0.39, "learning_rate": 3.9992112054136473e-07, "loss": 0.0456, "step": 982 }, { "epoch": 0.39, "learning_rate": 3.9992040187331375e-07, "loss": 0.0586, "step": 983 }, { "epoch": 0.39, "learning_rate": 3.99919679946864e-07, "loss": 0.0512, "step": 984 }, { "epoch": 0.39, "learning_rate": 3.999189547620274e-07, "loss": 0.0542, "step": 985 }, { "epoch": 0.39, "learning_rate": 3.999182263188156e-07, "loss": 0.0532, "step": 986 }, { "epoch": 0.39, "learning_rate": 3.9991749461724054e-07, "loss": 0.0605, "step": 987 }, { "epoch": 0.39, "learning_rate": 3.9991675965731424e-07, "loss": 0.0508, "step": 988 }, { "epoch": 0.39, "learning_rate": 3.9991602143904853e-07, "loss": 0.061, "step": 989 }, { "epoch": 0.39, "learning_rate": 3.9991527996245556e-07, "loss": 0.0531, "step": 990 }, { "epoch": 0.39, "learning_rate": 3.9991453522754733e-07, "loss": 0.0526, "step": 991 }, { "epoch": 0.39, "learning_rate": 3.9991378723433596e-07, "loss": 0.0526, "step": 992 }, { "epoch": 0.39, "learning_rate": 3.9991303598283375e-07, "loss": 0.0493, "step": 993 }, { "epoch": 0.39, "learning_rate": 3.9991228147305287e-07, "loss": 0.0478, "step": 994 }, { "epoch": 0.39, "learning_rate": 3.9991152370500564e-07, "loss": 0.0562, "step": 995 }, { "epoch": 0.39, "learning_rate": 3.999107626787044e-07, "loss": 0.0509, "step": 996 }, { "epoch": 0.39, "learning_rate": 3.999099983941616e-07, "loss": 0.0487, "step": 997 }, { "epoch": 0.39, "learning_rate": 3.999092308513896e-07, "loss": 0.0561, "step": 998 }, { "epoch": 0.39, "learning_rate": 3.9990846005040096e-07, "loss": 0.0537, "step": 999 }, { "epoch": 0.39, "learning_rate": 3.9990768599120834e-07, "loss": 0.055, "step": 1000 }, { "epoch": 0.39, "learning_rate": 3.999069086738242e-07, "loss": 0.0494, "step": 1001 }, { "epoch": 0.39, "learning_rate": 3.9990612809826124e-07, "loss": 0.051, "step": 1002 }, { "epoch": 0.4, "learning_rate": 3.999053442645322e-07, "loss": 0.051, "step": 1003 }, { "epoch": 0.4, "learning_rate": 3.9990455717264997e-07, "loss": 0.0539, "step": 1004 }, { "epoch": 0.4, "learning_rate": 3.999037668226272e-07, "loss": 0.0496, "step": 1005 }, { "epoch": 0.4, "learning_rate": 3.9990297321447685e-07, "loss": 0.0545, "step": 1006 }, { "epoch": 0.4, "learning_rate": 3.9990217634821184e-07, "loss": 0.0589, "step": 1007 }, { "epoch": 0.4, "learning_rate": 3.9990137622384523e-07, "loss": 0.051, "step": 1008 }, { "epoch": 0.4, "learning_rate": 3.9990057284138995e-07, "loss": 0.0516, "step": 1009 }, { "epoch": 0.4, "learning_rate": 3.9989976620085917e-07, "loss": 0.0504, "step": 1010 }, { "epoch": 0.4, "learning_rate": 3.9989895630226604e-07, "loss": 0.053, "step": 1011 }, { "epoch": 0.4, "learning_rate": 3.998981431456237e-07, "loss": 0.0565, "step": 1012 }, { "epoch": 0.4, "learning_rate": 3.9989732673094546e-07, "loss": 0.0448, "step": 1013 }, { "epoch": 0.4, "learning_rate": 3.9989650705824454e-07, "loss": 0.0496, "step": 1014 }, { "epoch": 0.4, "learning_rate": 3.9989568412753446e-07, "loss": 0.0527, "step": 1015 }, { "epoch": 0.4, "learning_rate": 3.998948579388285e-07, "loss": 0.0507, "step": 1016 }, { "epoch": 0.4, "learning_rate": 3.9989402849214013e-07, "loss": 0.0509, "step": 1017 }, { "epoch": 0.4, "learning_rate": 3.998931957874829e-07, "loss": 0.0528, "step": 1018 }, { "epoch": 0.4, "learning_rate": 3.9989235982487043e-07, "loss": 0.048, "step": 1019 }, { "epoch": 0.4, "learning_rate": 3.9989152060431626e-07, "loss": 0.0514, "step": 1020 }, { "epoch": 0.4, "learning_rate": 3.9989067812583414e-07, "loss": 0.053, "step": 1021 }, { "epoch": 0.4, "learning_rate": 3.998898323894377e-07, "loss": 0.0565, "step": 1022 }, { "epoch": 0.4, "learning_rate": 3.9988898339514085e-07, "loss": 0.0492, "step": 1023 }, { "epoch": 0.4, "learning_rate": 3.9988813114295736e-07, "loss": 0.0512, "step": 1024 }, { "epoch": 0.4, "learning_rate": 3.9988727563290104e-07, "loss": 0.0463, "step": 1025 }, { "epoch": 0.4, "learning_rate": 3.99886416864986e-07, "loss": 0.0489, "step": 1026 }, { "epoch": 0.4, "learning_rate": 3.9988555483922614e-07, "loss": 0.0482, "step": 1027 }, { "epoch": 0.41, "learning_rate": 3.9988468955563553e-07, "loss": 0.0494, "step": 1028 }, { "epoch": 0.41, "learning_rate": 3.998838210142283e-07, "loss": 0.0501, "step": 1029 }, { "epoch": 0.41, "learning_rate": 3.9988294921501853e-07, "loss": 0.049, "step": 1030 }, { "epoch": 0.41, "learning_rate": 3.998820741580205e-07, "loss": 0.0514, "step": 1031 }, { "epoch": 0.41, "learning_rate": 3.998811958432484e-07, "loss": 0.0498, "step": 1032 }, { "epoch": 0.41, "learning_rate": 3.998803142707167e-07, "loss": 0.0576, "step": 1033 }, { "epoch": 0.41, "learning_rate": 3.998794294404396e-07, "loss": 0.0461, "step": 1034 }, { "epoch": 0.41, "learning_rate": 3.9987854135243157e-07, "loss": 0.0498, "step": 1035 }, { "epoch": 0.41, "learning_rate": 3.998776500067071e-07, "loss": 0.0497, "step": 1036 }, { "epoch": 0.41, "learning_rate": 3.998767554032807e-07, "loss": 0.0484, "step": 1037 }, { "epoch": 0.41, "learning_rate": 3.9987585754216695e-07, "loss": 0.0572, "step": 1038 }, { "epoch": 0.41, "learning_rate": 3.9987495642338055e-07, "loss": 0.0478, "step": 1039 }, { "epoch": 0.41, "learning_rate": 3.9987405204693613e-07, "loss": 0.0468, "step": 1040 }, { "epoch": 0.41, "learning_rate": 3.9987314441284843e-07, "loss": 0.0544, "step": 1041 }, { "epoch": 0.41, "learning_rate": 3.9987223352113225e-07, "loss": 0.0539, "step": 1042 }, { "epoch": 0.41, "learning_rate": 3.998713193718024e-07, "loss": 0.0505, "step": 1043 }, { "epoch": 0.41, "learning_rate": 3.9987040196487383e-07, "loss": 0.0516, "step": 1044 }, { "epoch": 0.41, "learning_rate": 3.998694813003615e-07, "loss": 0.049, "step": 1045 }, { "epoch": 0.41, "learning_rate": 3.9986855737828036e-07, "loss": 0.0549, "step": 1046 }, { "epoch": 0.41, "learning_rate": 3.998676301986455e-07, "loss": 0.0513, "step": 1047 }, { "epoch": 0.41, "learning_rate": 3.9986669976147207e-07, "loss": 0.0504, "step": 1048 }, { "epoch": 0.41, "learning_rate": 3.9986576606677514e-07, "loss": 0.0561, "step": 1049 }, { "epoch": 0.41, "learning_rate": 3.9986482911457e-07, "loss": 0.0518, "step": 1050 }, { "epoch": 0.41, "learning_rate": 3.9986388890487196e-07, "loss": 0.0532, "step": 1051 }, { "epoch": 0.41, "learning_rate": 3.9986294543769625e-07, "loss": 0.0628, "step": 1052 }, { "epoch": 0.41, "learning_rate": 3.998619987130583e-07, "loss": 0.0474, "step": 1053 }, { "epoch": 0.42, "learning_rate": 3.9986104873097347e-07, "loss": 0.0538, "step": 1054 }, { "epoch": 0.42, "learning_rate": 3.9986009549145737e-07, "loss": 0.0559, "step": 1055 }, { "epoch": 0.42, "learning_rate": 3.998591389945254e-07, "loss": 0.0515, "step": 1056 }, { "epoch": 0.42, "learning_rate": 3.998581792401933e-07, "loss": 0.0542, "step": 1057 }, { "epoch": 0.42, "learning_rate": 3.9985721622847654e-07, "loss": 0.0517, "step": 1058 }, { "epoch": 0.42, "learning_rate": 3.99856249959391e-07, "loss": 0.0542, "step": 1059 }, { "epoch": 0.42, "learning_rate": 3.998552804329522e-07, "loss": 0.0539, "step": 1060 }, { "epoch": 0.42, "learning_rate": 3.998543076491762e-07, "loss": 0.0551, "step": 1061 }, { "epoch": 0.42, "learning_rate": 3.9985333160807864e-07, "loss": 0.0497, "step": 1062 }, { "epoch": 0.42, "learning_rate": 3.9985235230967554e-07, "loss": 0.0473, "step": 1063 }, { "epoch": 0.42, "learning_rate": 3.998513697539829e-07, "loss": 0.0501, "step": 1064 }, { "epoch": 0.42, "learning_rate": 3.998503839410166e-07, "loss": 0.0546, "step": 1065 }, { "epoch": 0.42, "learning_rate": 3.998493948707928e-07, "loss": 0.0481, "step": 1066 }, { "epoch": 0.42, "learning_rate": 3.998484025433276e-07, "loss": 0.046, "step": 1067 }, { "epoch": 0.42, "learning_rate": 3.998474069586372e-07, "loss": 0.0487, "step": 1068 }, { "epoch": 0.42, "learning_rate": 3.998464081167378e-07, "loss": 0.0516, "step": 1069 }, { "epoch": 0.42, "learning_rate": 3.9984540601764564e-07, "loss": 0.0483, "step": 1070 }, { "epoch": 0.42, "learning_rate": 3.998444006613771e-07, "loss": 0.0517, "step": 1071 }, { "epoch": 0.42, "learning_rate": 3.9984339204794856e-07, "loss": 0.0497, "step": 1072 }, { "epoch": 0.42, "learning_rate": 3.9984238017737644e-07, "loss": 0.0485, "step": 1073 }, { "epoch": 0.42, "learning_rate": 3.998413650496773e-07, "loss": 0.0494, "step": 1074 }, { "epoch": 0.42, "learning_rate": 3.9984034666486755e-07, "loss": 0.051, "step": 1075 }, { "epoch": 0.42, "learning_rate": 3.9983932502296396e-07, "loss": 0.0479, "step": 1076 }, { "epoch": 0.42, "learning_rate": 3.9983830012398306e-07, "loss": 0.0518, "step": 1077 }, { "epoch": 0.42, "learning_rate": 3.998372719679416e-07, "loss": 0.048, "step": 1078 }, { "epoch": 0.43, "learning_rate": 3.998362405548563e-07, "loss": 0.0554, "step": 1079 }, { "epoch": 0.43, "learning_rate": 3.99835205884744e-07, "loss": 0.0512, "step": 1080 }, { "epoch": 0.43, "learning_rate": 3.9983416795762154e-07, "loss": 0.0517, "step": 1081 }, { "epoch": 0.43, "learning_rate": 3.998331267735059e-07, "loss": 0.0454, "step": 1082 }, { "epoch": 0.43, "learning_rate": 3.99832082332414e-07, "loss": 0.0562, "step": 1083 }, { "epoch": 0.43, "learning_rate": 3.998310346343628e-07, "loss": 0.054, "step": 1084 }, { "epoch": 0.43, "learning_rate": 3.998299836793696e-07, "loss": 0.0491, "step": 1085 }, { "epoch": 0.43, "learning_rate": 3.998289294674512e-07, "loss": 0.0576, "step": 1086 }, { "epoch": 0.43, "learning_rate": 3.9982787199862506e-07, "loss": 0.0423, "step": 1087 }, { "epoch": 0.43, "learning_rate": 3.998268112729082e-07, "loss": 0.0524, "step": 1088 }, { "epoch": 0.43, "learning_rate": 3.9982574729031814e-07, "loss": 0.0456, "step": 1089 }, { "epoch": 0.43, "learning_rate": 3.9982468005087203e-07, "loss": 0.0445, "step": 1090 }, { "epoch": 0.43, "learning_rate": 3.9982360955458736e-07, "loss": 0.0588, "step": 1091 }, { "epoch": 0.43, "learning_rate": 3.9982253580148155e-07, "loss": 0.0583, "step": 1092 }, { "epoch": 0.43, "learning_rate": 3.998214587915721e-07, "loss": 0.0528, "step": 1093 }, { "epoch": 0.43, "learning_rate": 3.9982037852487655e-07, "loss": 0.0511, "step": 1094 }, { "epoch": 0.43, "learning_rate": 3.9981929500141256e-07, "loss": 0.0483, "step": 1095 }, { "epoch": 0.43, "learning_rate": 3.9981820822119773e-07, "loss": 0.0539, "step": 1096 }, { "epoch": 0.43, "learning_rate": 3.998171181842498e-07, "loss": 0.0535, "step": 1097 }, { "epoch": 0.43, "learning_rate": 3.998160248905865e-07, "loss": 0.0482, "step": 1098 }, { "epoch": 0.43, "learning_rate": 3.9981492834022567e-07, "loss": 0.0598, "step": 1099 }, { "epoch": 0.43, "learning_rate": 3.9981382853318526e-07, "loss": 0.0486, "step": 1100 }, { "epoch": 0.43, "learning_rate": 3.9981272546948306e-07, "loss": 0.0491, "step": 1101 }, { "epoch": 0.43, "learning_rate": 3.9981161914913717e-07, "loss": 0.0477, "step": 1102 }, { "epoch": 0.43, "learning_rate": 3.9981050957216554e-07, "loss": 0.0455, "step": 1103 }, { "epoch": 0.44, "learning_rate": 3.9980939673858627e-07, "loss": 0.0525, "step": 1104 }, { "epoch": 0.44, "learning_rate": 3.998082806484175e-07, "loss": 0.0455, "step": 1105 }, { "epoch": 0.44, "learning_rate": 3.998071613016775e-07, "loss": 0.0475, "step": 1106 }, { "epoch": 0.44, "learning_rate": 3.998060386983844e-07, "loss": 0.0474, "step": 1107 }, { "epoch": 0.44, "learning_rate": 3.9980491283855653e-07, "loss": 0.0501, "step": 1108 }, { "epoch": 0.44, "learning_rate": 3.9980378372221224e-07, "loss": 0.0459, "step": 1109 }, { "epoch": 0.44, "learning_rate": 3.998026513493699e-07, "loss": 0.053, "step": 1110 }, { "epoch": 0.44, "learning_rate": 3.998015157200481e-07, "loss": 0.0531, "step": 1111 }, { "epoch": 0.44, "learning_rate": 3.998003768342652e-07, "loss": 0.0507, "step": 1112 }, { "epoch": 0.44, "learning_rate": 3.997992346920399e-07, "loss": 0.0426, "step": 1113 }, { "epoch": 0.44, "learning_rate": 3.9979808929339064e-07, "loss": 0.0572, "step": 1114 }, { "epoch": 0.44, "learning_rate": 3.997969406383362e-07, "loss": 0.0541, "step": 1115 }, { "epoch": 0.44, "learning_rate": 3.997957887268953e-07, "loss": 0.054, "step": 1116 }, { "epoch": 0.44, "learning_rate": 3.997946335590867e-07, "loss": 0.049, "step": 1117 }, { "epoch": 0.44, "learning_rate": 3.9979347513492924e-07, "loss": 0.0466, "step": 1118 }, { "epoch": 0.44, "learning_rate": 3.9979231345444173e-07, "loss": 0.0556, "step": 1119 }, { "epoch": 0.44, "learning_rate": 3.9979114851764327e-07, "loss": 0.0495, "step": 1120 }, { "epoch": 0.44, "learning_rate": 3.9978998032455263e-07, "loss": 0.0527, "step": 1121 }, { "epoch": 0.44, "learning_rate": 3.99788808875189e-07, "loss": 0.0491, "step": 1122 }, { "epoch": 0.44, "learning_rate": 3.997876341695714e-07, "loss": 0.0523, "step": 1123 }, { "epoch": 0.44, "learning_rate": 3.99786456207719e-07, "loss": 0.0577, "step": 1124 }, { "epoch": 0.44, "learning_rate": 3.997852749896511e-07, "loss": 0.0438, "step": 1125 }, { "epoch": 0.44, "learning_rate": 3.9978409051538674e-07, "loss": 0.0511, "step": 1126 }, { "epoch": 0.44, "learning_rate": 3.997829027849454e-07, "loss": 0.05, "step": 1127 }, { "epoch": 0.44, "learning_rate": 3.9978171179834635e-07, "loss": 0.0533, "step": 1128 }, { "epoch": 0.44, "learning_rate": 3.997805175556091e-07, "loss": 0.0523, "step": 1129 }, { "epoch": 0.45, "learning_rate": 3.9977932005675297e-07, "loss": 0.0489, "step": 1130 }, { "epoch": 0.45, "learning_rate": 3.997781193017975e-07, "loss": 0.0498, "step": 1131 }, { "epoch": 0.45, "learning_rate": 3.9977691529076246e-07, "loss": 0.0535, "step": 1132 }, { "epoch": 0.45, "learning_rate": 3.997757080236672e-07, "loss": 0.046, "step": 1133 }, { "epoch": 0.45, "learning_rate": 3.997744975005316e-07, "loss": 0.056, "step": 1134 }, { "epoch": 0.45, "learning_rate": 3.9977328372137533e-07, "loss": 0.0544, "step": 1135 }, { "epoch": 0.45, "learning_rate": 3.9977206668621815e-07, "loss": 0.049, "step": 1136 }, { "epoch": 0.45, "learning_rate": 3.997708463950798e-07, "loss": 0.0506, "step": 1137 }, { "epoch": 0.45, "learning_rate": 3.9976962284798034e-07, "loss": 0.0488, "step": 1138 }, { "epoch": 0.45, "learning_rate": 3.9976839604493964e-07, "loss": 0.0548, "step": 1139 }, { "epoch": 0.45, "learning_rate": 3.9976716598597777e-07, "loss": 0.0498, "step": 1140 }, { "epoch": 0.45, "learning_rate": 3.997659326711146e-07, "loss": 0.0498, "step": 1141 }, { "epoch": 0.45, "learning_rate": 3.9976469610037035e-07, "loss": 0.0525, "step": 1142 }, { "epoch": 0.45, "learning_rate": 3.9976345627376525e-07, "loss": 0.0502, "step": 1143 }, { "epoch": 0.45, "learning_rate": 3.997622131913193e-07, "loss": 0.0474, "step": 1144 }, { "epoch": 0.45, "learning_rate": 3.997609668530529e-07, "loss": 0.0523, "step": 1145 }, { "epoch": 0.45, "learning_rate": 3.997597172589864e-07, "loss": 0.0526, "step": 1146 }, { "epoch": 0.45, "learning_rate": 3.997584644091401e-07, "loss": 0.0534, "step": 1147 }, { "epoch": 0.45, "learning_rate": 3.997572083035344e-07, "loss": 0.0542, "step": 1148 }, { "epoch": 0.45, "learning_rate": 3.9975594894218984e-07, "loss": 0.0494, "step": 1149 }, { "epoch": 0.45, "learning_rate": 3.9975468632512684e-07, "loss": 0.0525, "step": 1150 }, { "epoch": 0.45, "learning_rate": 3.997534204523661e-07, "loss": 0.0504, "step": 1151 }, { "epoch": 0.45, "learning_rate": 3.997521513239282e-07, "loss": 0.0472, "step": 1152 }, { "epoch": 0.45, "learning_rate": 3.9975087893983374e-07, "loss": 0.0497, "step": 1153 }, { "epoch": 0.45, "learning_rate": 3.997496033001036e-07, "loss": 0.0499, "step": 1154 }, { "epoch": 0.46, "learning_rate": 3.9974832440475853e-07, "loss": 0.0539, "step": 1155 }, { "epoch": 0.46, "learning_rate": 3.9974704225381933e-07, "loss": 0.052, "step": 1156 }, { "epoch": 0.46, "learning_rate": 3.9974575684730694e-07, "loss": 0.0495, "step": 1157 }, { "epoch": 0.46, "learning_rate": 3.997444681852423e-07, "loss": 0.0529, "step": 1158 }, { "epoch": 0.46, "learning_rate": 3.9974317626764636e-07, "loss": 0.0454, "step": 1159 }, { "epoch": 0.46, "learning_rate": 3.9974188109454026e-07, "loss": 0.0548, "step": 1160 }, { "epoch": 0.46, "learning_rate": 3.997405826659451e-07, "loss": 0.0548, "step": 1161 }, { "epoch": 0.46, "learning_rate": 3.9973928098188196e-07, "loss": 0.0507, "step": 1162 }, { "epoch": 0.46, "learning_rate": 3.9973797604237205e-07, "loss": 0.0495, "step": 1163 }, { "epoch": 0.46, "learning_rate": 3.9973666784743675e-07, "loss": 0.0475, "step": 1164 }, { "epoch": 0.46, "learning_rate": 3.997353563970974e-07, "loss": 0.0509, "step": 1165 }, { "epoch": 0.46, "learning_rate": 3.997340416913753e-07, "loss": 0.0528, "step": 1166 }, { "epoch": 0.46, "learning_rate": 3.9973272373029177e-07, "loss": 0.0501, "step": 1167 }, { "epoch": 0.46, "learning_rate": 3.9973140251386847e-07, "loss": 0.0493, "step": 1168 }, { "epoch": 0.46, "learning_rate": 3.997300780421268e-07, "loss": 0.0491, "step": 1169 }, { "epoch": 0.46, "learning_rate": 3.9972875031508854e-07, "loss": 0.0504, "step": 1170 }, { "epoch": 0.46, "learning_rate": 3.997274193327751e-07, "loss": 0.0489, "step": 1171 }, { "epoch": 0.46, "learning_rate": 3.9972608509520834e-07, "loss": 0.0497, "step": 1172 }, { "epoch": 0.46, "learning_rate": 3.997247476024099e-07, "loss": 0.051, "step": 1173 }, { "epoch": 0.46, "learning_rate": 3.997234068544016e-07, "loss": 0.0478, "step": 1174 }, { "epoch": 0.46, "learning_rate": 3.997220628512053e-07, "loss": 0.0471, "step": 1175 }, { "epoch": 0.46, "learning_rate": 3.9972071559284305e-07, "loss": 0.0511, "step": 1176 }, { "epoch": 0.46, "learning_rate": 3.9971936507933655e-07, "loss": 0.0483, "step": 1177 }, { "epoch": 0.46, "learning_rate": 3.9971801131070796e-07, "loss": 0.0512, "step": 1178 }, { "epoch": 0.46, "learning_rate": 3.9971665428697934e-07, "loss": 0.0535, "step": 1179 }, { "epoch": 0.47, "learning_rate": 3.9971529400817277e-07, "loss": 0.0551, "step": 1180 }, { "epoch": 0.47, "learning_rate": 3.997139304743105e-07, "loss": 0.0509, "step": 1181 }, { "epoch": 0.47, "learning_rate": 3.997125636854146e-07, "loss": 0.0533, "step": 1182 }, { "epoch": 0.47, "learning_rate": 3.9971119364150756e-07, "loss": 0.0502, "step": 1183 }, { "epoch": 0.47, "learning_rate": 3.997098203426115e-07, "loss": 0.0508, "step": 1184 }, { "epoch": 0.47, "learning_rate": 3.9970844378874886e-07, "loss": 0.0623, "step": 1185 }, { "epoch": 0.47, "learning_rate": 3.997070639799422e-07, "loss": 0.0577, "step": 1186 }, { "epoch": 0.47, "learning_rate": 3.9970568091621386e-07, "loss": 0.0551, "step": 1187 }, { "epoch": 0.47, "learning_rate": 3.997042945975865e-07, "loss": 0.0503, "step": 1188 }, { "epoch": 0.47, "learning_rate": 3.997029050240826e-07, "loss": 0.0493, "step": 1189 }, { "epoch": 0.47, "learning_rate": 3.9970151219572485e-07, "loss": 0.0469, "step": 1190 }, { "epoch": 0.47, "learning_rate": 3.99700116112536e-07, "loss": 0.0523, "step": 1191 }, { "epoch": 0.47, "learning_rate": 3.996987167745387e-07, "loss": 0.0554, "step": 1192 }, { "epoch": 0.47, "learning_rate": 3.996973141817559e-07, "loss": 0.0504, "step": 1193 }, { "epoch": 0.47, "learning_rate": 3.9969590833421034e-07, "loss": 0.046, "step": 1194 }, { "epoch": 0.47, "learning_rate": 3.99694499231925e-07, "loss": 0.0522, "step": 1195 }, { "epoch": 0.47, "learning_rate": 3.996930868749228e-07, "loss": 0.0484, "step": 1196 }, { "epoch": 0.47, "learning_rate": 3.9969167126322684e-07, "loss": 0.0508, "step": 1197 }, { "epoch": 0.47, "learning_rate": 3.996902523968601e-07, "loss": 0.0491, "step": 1198 }, { "epoch": 0.47, "learning_rate": 3.996888302758457e-07, "loss": 0.0472, "step": 1199 }, { "epoch": 0.47, "learning_rate": 3.996874049002069e-07, "loss": 0.0468, "step": 1200 }, { "epoch": 0.47, "learning_rate": 3.996859762699669e-07, "loss": 0.0515, "step": 1201 }, { "epoch": 0.47, "learning_rate": 3.996845443851489e-07, "loss": 0.0496, "step": 1202 }, { "epoch": 0.47, "learning_rate": 3.996831092457764e-07, "loss": 0.0534, "step": 1203 }, { "epoch": 0.47, "learning_rate": 3.996816708518727e-07, "loss": 0.0472, "step": 1204 }, { "epoch": 0.47, "learning_rate": 3.996802292034612e-07, "loss": 0.0524, "step": 1205 }, { "epoch": 0.48, "learning_rate": 3.996787843005655e-07, "loss": 0.0495, "step": 1206 }, { "epoch": 0.48, "learning_rate": 3.9967733614320904e-07, "loss": 0.0578, "step": 1207 }, { "epoch": 0.48, "learning_rate": 3.996758847314154e-07, "loss": 0.0476, "step": 1208 }, { "epoch": 0.48, "learning_rate": 3.9967443006520847e-07, "loss": 0.0544, "step": 1209 }, { "epoch": 0.48, "learning_rate": 3.996729721446117e-07, "loss": 0.0453, "step": 1210 }, { "epoch": 0.48, "learning_rate": 3.99671510969649e-07, "loss": 0.0434, "step": 1211 }, { "epoch": 0.48, "learning_rate": 3.9967004654034417e-07, "loss": 0.0554, "step": 1212 }, { "epoch": 0.48, "learning_rate": 3.99668578856721e-07, "loss": 0.0488, "step": 1213 }, { "epoch": 0.48, "learning_rate": 3.996671079188034e-07, "loss": 0.0458, "step": 1214 }, { "epoch": 0.48, "learning_rate": 3.996656337266155e-07, "loss": 0.0512, "step": 1215 }, { "epoch": 0.48, "learning_rate": 3.9966415628018114e-07, "loss": 0.0502, "step": 1216 }, { "epoch": 0.48, "learning_rate": 3.9966267557952447e-07, "loss": 0.0492, "step": 1217 }, { "epoch": 0.48, "learning_rate": 3.9966119162466975e-07, "loss": 0.0477, "step": 1218 }, { "epoch": 0.48, "learning_rate": 3.99659704415641e-07, "loss": 0.0523, "step": 1219 }, { "epoch": 0.48, "learning_rate": 3.996582139524625e-07, "loss": 0.0489, "step": 1220 }, { "epoch": 0.48, "learning_rate": 3.9965672023515856e-07, "loss": 0.0479, "step": 1221 }, { "epoch": 0.48, "learning_rate": 3.9965522326375357e-07, "loss": 0.0526, "step": 1222 }, { "epoch": 0.48, "learning_rate": 3.996537230382719e-07, "loss": 0.0493, "step": 1223 }, { "epoch": 0.48, "learning_rate": 3.996522195587379e-07, "loss": 0.0528, "step": 1224 }, { "epoch": 0.48, "learning_rate": 3.9965071282517616e-07, "loss": 0.045, "step": 1225 }, { "epoch": 0.48, "learning_rate": 3.9964920283761127e-07, "loss": 0.0487, "step": 1226 }, { "epoch": 0.48, "learning_rate": 3.996476895960678e-07, "loss": 0.0522, "step": 1227 }, { "epoch": 0.48, "learning_rate": 3.9964617310057044e-07, "loss": 0.0474, "step": 1228 }, { "epoch": 0.48, "learning_rate": 3.9964465335114385e-07, "loss": 0.0456, "step": 1229 }, { "epoch": 0.48, "learning_rate": 3.9964313034781287e-07, "loss": 0.0514, "step": 1230 }, { "epoch": 0.49, "learning_rate": 3.9964160409060224e-07, "loss": 0.0586, "step": 1231 }, { "epoch": 0.49, "learning_rate": 3.99640074579537e-07, "loss": 0.0489, "step": 1232 }, { "epoch": 0.49, "learning_rate": 3.9963854181464177e-07, "loss": 0.0473, "step": 1233 }, { "epoch": 0.49, "learning_rate": 3.996370057959419e-07, "loss": 0.057, "step": 1234 }, { "epoch": 0.49, "learning_rate": 3.9963546652346214e-07, "loss": 0.0509, "step": 1235 }, { "epoch": 0.49, "learning_rate": 3.9963392399722775e-07, "loss": 0.0489, "step": 1236 }, { "epoch": 0.49, "learning_rate": 3.996323782172638e-07, "loss": 0.053, "step": 1237 }, { "epoch": 0.49, "learning_rate": 3.996308291835954e-07, "loss": 0.0474, "step": 1238 }, { "epoch": 0.49, "learning_rate": 3.99629276896248e-07, "loss": 0.0549, "step": 1239 }, { "epoch": 0.49, "learning_rate": 3.9962772135524676e-07, "loss": 0.0496, "step": 1240 }, { "epoch": 0.49, "learning_rate": 3.9962616256061705e-07, "loss": 0.0469, "step": 1241 }, { "epoch": 0.49, "learning_rate": 3.9962460051238424e-07, "loss": 0.0493, "step": 1242 }, { "epoch": 0.49, "learning_rate": 3.9962303521057395e-07, "loss": 0.0514, "step": 1243 }, { "epoch": 0.49, "learning_rate": 3.996214666552115e-07, "loss": 0.0475, "step": 1244 }, { "epoch": 0.49, "learning_rate": 3.9961989484632256e-07, "loss": 0.0567, "step": 1245 }, { "epoch": 0.49, "learning_rate": 3.996183197839327e-07, "loss": 0.0533, "step": 1246 }, { "epoch": 0.49, "learning_rate": 3.996167414680676e-07, "loss": 0.0519, "step": 1247 }, { "epoch": 0.49, "learning_rate": 3.9961515989875303e-07, "loss": 0.0508, "step": 1248 }, { "epoch": 0.49, "learning_rate": 3.996135750760147e-07, "loss": 0.0487, "step": 1249 }, { "epoch": 0.49, "learning_rate": 3.9961198699987853e-07, "loss": 0.044, "step": 1250 }, { "epoch": 0.49, "learning_rate": 3.9961039567037027e-07, "loss": 0.0469, "step": 1251 }, { "epoch": 0.49, "learning_rate": 3.99608801087516e-07, "loss": 0.052, "step": 1252 }, { "epoch": 0.49, "learning_rate": 3.996072032513416e-07, "loss": 0.0445, "step": 1253 }, { "epoch": 0.49, "learning_rate": 3.996056021618732e-07, "loss": 0.0542, "step": 1254 }, { "epoch": 0.49, "learning_rate": 3.996039978191368e-07, "loss": 0.0474, "step": 1255 }, { "epoch": 0.49, "learning_rate": 3.996023902231586e-07, "loss": 0.048, "step": 1256 }, { "epoch": 0.5, "learning_rate": 3.9960077937396486e-07, "loss": 0.0457, "step": 1257 }, { "epoch": 0.5, "learning_rate": 3.9959916527158173e-07, "loss": 0.0464, "step": 1258 }, { "epoch": 0.5, "learning_rate": 3.995975479160356e-07, "loss": 0.0506, "step": 1259 }, { "epoch": 0.5, "learning_rate": 3.9959592730735273e-07, "loss": 0.0502, "step": 1260 }, { "epoch": 0.5, "learning_rate": 3.995943034455596e-07, "loss": 0.0464, "step": 1261 }, { "epoch": 0.5, "learning_rate": 3.9959267633068273e-07, "loss": 0.0473, "step": 1262 }, { "epoch": 0.5, "learning_rate": 3.995910459627485e-07, "loss": 0.0555, "step": 1263 }, { "epoch": 0.5, "learning_rate": 3.9958941234178365e-07, "loss": 0.0466, "step": 1264 }, { "epoch": 0.5, "learning_rate": 3.995877754678147e-07, "loss": 0.0483, "step": 1265 }, { "epoch": 0.5, "learning_rate": 3.995861353408683e-07, "loss": 0.0465, "step": 1266 }, { "epoch": 0.5, "learning_rate": 3.9958449196097123e-07, "loss": 0.0483, "step": 1267 }, { "epoch": 0.5, "learning_rate": 3.995828453281504e-07, "loss": 0.0477, "step": 1268 }, { "epoch": 0.5, "learning_rate": 3.995811954424324e-07, "loss": 0.0463, "step": 1269 }, { "epoch": 0.5, "learning_rate": 3.9957954230384423e-07, "loss": 0.0475, "step": 1270 }, { "epoch": 0.5, "learning_rate": 3.995778859124129e-07, "loss": 0.0465, "step": 1271 }, { "epoch": 0.5, "learning_rate": 3.9957622626816534e-07, "loss": 0.0517, "step": 1272 }, { "epoch": 0.5, "learning_rate": 3.995745633711286e-07, "loss": 0.0474, "step": 1273 }, { "epoch": 0.5, "learning_rate": 3.995728972213298e-07, "loss": 0.0433, "step": 1274 }, { "epoch": 0.5, "learning_rate": 3.995712278187962e-07, "loss": 0.0473, "step": 1275 }, { "epoch": 0.5, "learning_rate": 3.9956955516355475e-07, "loss": 0.0423, "step": 1276 }, { "epoch": 0.5, "learning_rate": 3.9956787925563286e-07, "loss": 0.05, "step": 1277 }, { "epoch": 0.5, "learning_rate": 3.995662000950579e-07, "loss": 0.0434, "step": 1278 }, { "epoch": 0.5, "learning_rate": 3.995645176818572e-07, "loss": 0.0512, "step": 1279 }, { "epoch": 0.5, "learning_rate": 3.995628320160581e-07, "loss": 0.0515, "step": 1280 }, { "epoch": 0.5, "learning_rate": 3.995611430976883e-07, "loss": 0.0487, "step": 1281 }, { "epoch": 0.51, "learning_rate": 3.99559450926775e-07, "loss": 0.046, "step": 1282 }, { "epoch": 0.51, "learning_rate": 3.9955775550334596e-07, "loss": 0.0487, "step": 1283 }, { "epoch": 0.51, "learning_rate": 3.995560568274289e-07, "loss": 0.049, "step": 1284 }, { "epoch": 0.51, "learning_rate": 3.9955435489905127e-07, "loss": 0.046, "step": 1285 }, { "epoch": 0.51, "learning_rate": 3.9955264971824103e-07, "loss": 0.05, "step": 1286 }, { "epoch": 0.51, "learning_rate": 3.9955094128502587e-07, "loss": 0.048, "step": 1287 }, { "epoch": 0.51, "learning_rate": 3.995492295994337e-07, "loss": 0.0477, "step": 1288 }, { "epoch": 0.51, "learning_rate": 3.9954751466149227e-07, "loss": 0.0475, "step": 1289 }, { "epoch": 0.51, "learning_rate": 3.995457964712297e-07, "loss": 0.04, "step": 1290 }, { "epoch": 0.51, "learning_rate": 3.9954407502867387e-07, "loss": 0.049, "step": 1291 }, { "epoch": 0.51, "learning_rate": 3.995423503338529e-07, "loss": 0.0475, "step": 1292 }, { "epoch": 0.51, "learning_rate": 3.9954062238679487e-07, "loss": 0.0409, "step": 1293 }, { "epoch": 0.51, "learning_rate": 3.99538891187528e-07, "loss": 0.0498, "step": 1294 }, { "epoch": 0.51, "learning_rate": 3.9953715673608047e-07, "loss": 0.0512, "step": 1295 }, { "epoch": 0.51, "learning_rate": 3.995354190324806e-07, "loss": 0.0458, "step": 1296 }, { "epoch": 0.51, "learning_rate": 3.995336780767566e-07, "loss": 0.0448, "step": 1297 }, { "epoch": 0.51, "learning_rate": 3.9953193386893685e-07, "loss": 0.0461, "step": 1298 }, { "epoch": 0.51, "learning_rate": 3.995301864090499e-07, "loss": 0.0523, "step": 1299 }, { "epoch": 0.51, "learning_rate": 3.995284356971242e-07, "loss": 0.0437, "step": 1300 }, { "epoch": 0.51, "learning_rate": 3.9952668173318813e-07, "loss": 0.0489, "step": 1301 }, { "epoch": 0.51, "learning_rate": 3.9952492451727046e-07, "loss": 0.052, "step": 1302 }, { "epoch": 0.51, "learning_rate": 3.9952316404939977e-07, "loss": 0.0521, "step": 1303 }, { "epoch": 0.51, "learning_rate": 3.9952140032960473e-07, "loss": 0.0467, "step": 1304 }, { "epoch": 0.51, "learning_rate": 3.995196333579141e-07, "loss": 0.0481, "step": 1305 }, { "epoch": 0.51, "learning_rate": 3.995178631343567e-07, "loss": 0.0528, "step": 1306 }, { "epoch": 0.52, "learning_rate": 3.995160896589613e-07, "loss": 0.0508, "step": 1307 }, { "epoch": 0.52, "learning_rate": 3.995143129317569e-07, "loss": 0.0495, "step": 1308 }, { "epoch": 0.52, "learning_rate": 3.995125329527725e-07, "loss": 0.0503, "step": 1309 }, { "epoch": 0.52, "learning_rate": 3.9951074972203696e-07, "loss": 0.051, "step": 1310 }, { "epoch": 0.52, "learning_rate": 3.995089632395794e-07, "loss": 0.0472, "step": 1311 }, { "epoch": 0.52, "learning_rate": 3.9950717350542903e-07, "loss": 0.0474, "step": 1312 }, { "epoch": 0.52, "learning_rate": 3.9950538051961493e-07, "loss": 0.0451, "step": 1313 }, { "epoch": 0.52, "learning_rate": 3.9950358428216634e-07, "loss": 0.0493, "step": 1314 }, { "epoch": 0.52, "learning_rate": 3.9950178479311253e-07, "loss": 0.0454, "step": 1315 }, { "epoch": 0.52, "learning_rate": 3.9949998205248283e-07, "loss": 0.0499, "step": 1316 }, { "epoch": 0.52, "learning_rate": 3.994981760603066e-07, "loss": 0.0527, "step": 1317 }, { "epoch": 0.52, "learning_rate": 3.9949636681661334e-07, "loss": 0.0443, "step": 1318 }, { "epoch": 0.52, "learning_rate": 3.9949455432143253e-07, "loss": 0.0546, "step": 1319 }, { "epoch": 0.52, "learning_rate": 3.994927385747936e-07, "loss": 0.0446, "step": 1320 }, { "epoch": 0.52, "learning_rate": 3.994909195767263e-07, "loss": 0.0465, "step": 1321 }, { "epoch": 0.52, "learning_rate": 3.9948909732726017e-07, "loss": 0.0505, "step": 1322 }, { "epoch": 0.52, "learning_rate": 3.9948727182642494e-07, "loss": 0.0496, "step": 1323 }, { "epoch": 0.52, "learning_rate": 3.994854430742504e-07, "loss": 0.0488, "step": 1324 }, { "epoch": 0.52, "learning_rate": 3.9948361107076626e-07, "loss": 0.0413, "step": 1325 }, { "epoch": 0.52, "learning_rate": 3.994817758160025e-07, "loss": 0.0469, "step": 1326 }, { "epoch": 0.52, "learning_rate": 3.994799373099889e-07, "loss": 0.0517, "step": 1327 }, { "epoch": 0.52, "learning_rate": 3.994780955527555e-07, "loss": 0.0472, "step": 1328 }, { "epoch": 0.52, "learning_rate": 3.994762505443324e-07, "loss": 0.0512, "step": 1329 }, { "epoch": 0.52, "learning_rate": 3.994744022847495e-07, "loss": 0.048, "step": 1330 }, { "epoch": 0.52, "learning_rate": 3.99472550774037e-07, "loss": 0.0419, "step": 1331 }, { "epoch": 0.52, "learning_rate": 3.9947069601222517e-07, "loss": 0.0461, "step": 1332 }, { "epoch": 0.53, "learning_rate": 3.9946883799934417e-07, "loss": 0.042, "step": 1333 }, { "epoch": 0.53, "learning_rate": 3.994669767354242e-07, "loss": 0.0589, "step": 1334 }, { "epoch": 0.53, "learning_rate": 3.994651122204956e-07, "loss": 0.0482, "step": 1335 }, { "epoch": 0.53, "learning_rate": 3.99463244454589e-07, "loss": 0.047, "step": 1336 }, { "epoch": 0.53, "learning_rate": 3.994613734377345e-07, "loss": 0.0437, "step": 1337 }, { "epoch": 0.53, "learning_rate": 3.994594991699628e-07, "loss": 0.0506, "step": 1338 }, { "epoch": 0.53, "learning_rate": 3.994576216513045e-07, "loss": 0.046, "step": 1339 }, { "epoch": 0.53, "learning_rate": 3.9945574088179e-07, "loss": 0.0505, "step": 1340 }, { "epoch": 0.53, "learning_rate": 3.9945385686145014e-07, "loss": 0.048, "step": 1341 }, { "epoch": 0.53, "learning_rate": 3.9945196959031546e-07, "loss": 0.0459, "step": 1342 }, { "epoch": 0.53, "learning_rate": 3.994500790684169e-07, "loss": 0.0486, "step": 1343 }, { "epoch": 0.53, "learning_rate": 3.994481852957851e-07, "loss": 0.0462, "step": 1344 }, { "epoch": 0.53, "learning_rate": 3.9944628827245103e-07, "loss": 0.0516, "step": 1345 }, { "epoch": 0.53, "learning_rate": 3.994443879984456e-07, "loss": 0.0449, "step": 1346 }, { "epoch": 0.53, "learning_rate": 3.994424844737998e-07, "loss": 0.0461, "step": 1347 }, { "epoch": 0.53, "learning_rate": 3.994405776985446e-07, "loss": 0.0423, "step": 1348 }, { "epoch": 0.53, "learning_rate": 3.9943866767271104e-07, "loss": 0.0538, "step": 1349 }, { "epoch": 0.53, "learning_rate": 3.994367543963304e-07, "loss": 0.052, "step": 1350 }, { "epoch": 0.53, "learning_rate": 3.994348378694337e-07, "loss": 0.045, "step": 1351 }, { "epoch": 0.53, "learning_rate": 3.994329180920522e-07, "loss": 0.0459, "step": 1352 }, { "epoch": 0.53, "learning_rate": 3.994309950642174e-07, "loss": 0.0535, "step": 1353 }, { "epoch": 0.53, "learning_rate": 3.9942906878596037e-07, "loss": 0.0461, "step": 1354 }, { "epoch": 0.53, "learning_rate": 3.9942713925731264e-07, "loss": 0.0492, "step": 1355 }, { "epoch": 0.53, "learning_rate": 3.9942520647830563e-07, "loss": 0.0506, "step": 1356 }, { "epoch": 0.53, "learning_rate": 3.9942327044897084e-07, "loss": 0.0436, "step": 1357 }, { "epoch": 0.54, "learning_rate": 3.994213311693398e-07, "loss": 0.043, "step": 1358 }, { "epoch": 0.54, "learning_rate": 3.9941938863944425e-07, "loss": 0.0456, "step": 1359 }, { "epoch": 0.54, "learning_rate": 3.9941744285931566e-07, "loss": 0.0483, "step": 1360 }, { "epoch": 0.54, "learning_rate": 3.9941549382898587e-07, "loss": 0.0431, "step": 1361 }, { "epoch": 0.54, "learning_rate": 3.994135415484866e-07, "loss": 0.0433, "step": 1362 }, { "epoch": 0.54, "learning_rate": 3.994115860178496e-07, "loss": 0.0455, "step": 1363 }, { "epoch": 0.54, "learning_rate": 3.9940962723710694e-07, "loss": 0.0513, "step": 1364 }, { "epoch": 0.54, "learning_rate": 3.994076652062904e-07, "loss": 0.048, "step": 1365 }, { "epoch": 0.54, "learning_rate": 3.994056999254319e-07, "loss": 0.0515, "step": 1366 }, { "epoch": 0.54, "learning_rate": 3.994037313945636e-07, "loss": 0.0444, "step": 1367 }, { "epoch": 0.54, "learning_rate": 3.9940175961371753e-07, "loss": 0.0446, "step": 1368 }, { "epoch": 0.54, "learning_rate": 3.9939978458292585e-07, "loss": 0.0491, "step": 1369 }, { "epoch": 0.54, "learning_rate": 3.9939780630222075e-07, "loss": 0.0497, "step": 1370 }, { "epoch": 0.54, "learning_rate": 3.9939582477163445e-07, "loss": 0.051, "step": 1371 }, { "epoch": 0.54, "learning_rate": 3.993938399911992e-07, "loss": 0.0441, "step": 1372 }, { "epoch": 0.54, "learning_rate": 3.9939185196094747e-07, "loss": 0.0455, "step": 1373 }, { "epoch": 0.54, "learning_rate": 3.993898606809115e-07, "loss": 0.0433, "step": 1374 }, { "epoch": 0.54, "learning_rate": 3.99387866151124e-07, "loss": 0.0455, "step": 1375 }, { "epoch": 0.54, "learning_rate": 3.993858683716172e-07, "loss": 0.0478, "step": 1376 }, { "epoch": 0.54, "learning_rate": 3.993838673424238e-07, "loss": 0.052, "step": 1377 }, { "epoch": 0.54, "learning_rate": 3.993818630635764e-07, "loss": 0.0515, "step": 1378 }, { "epoch": 0.54, "learning_rate": 3.993798555351076e-07, "loss": 0.0508, "step": 1379 }, { "epoch": 0.54, "learning_rate": 3.993778447570502e-07, "loss": 0.046, "step": 1380 }, { "epoch": 0.54, "learning_rate": 3.99375830729437e-07, "loss": 0.0422, "step": 1381 }, { "epoch": 0.54, "learning_rate": 3.9937381345230074e-07, "loss": 0.0517, "step": 1382 }, { "epoch": 0.55, "learning_rate": 3.9937179292567434e-07, "loss": 0.0507, "step": 1383 }, { "epoch": 0.55, "learning_rate": 3.9936976914959075e-07, "loss": 0.0498, "step": 1384 }, { "epoch": 0.55, "learning_rate": 3.993677421240829e-07, "loss": 0.0459, "step": 1385 }, { "epoch": 0.55, "learning_rate": 3.993657118491839e-07, "loss": 0.0534, "step": 1386 }, { "epoch": 0.55, "learning_rate": 3.9936367832492676e-07, "loss": 0.0482, "step": 1387 }, { "epoch": 0.55, "learning_rate": 3.993616415513446e-07, "loss": 0.0481, "step": 1388 }, { "epoch": 0.55, "learning_rate": 3.9935960152847077e-07, "loss": 0.0455, "step": 1389 }, { "epoch": 0.55, "learning_rate": 3.993575582563384e-07, "loss": 0.0448, "step": 1390 }, { "epoch": 0.55, "learning_rate": 3.9935551173498083e-07, "loss": 0.051, "step": 1391 }, { "epoch": 0.55, "learning_rate": 3.993534619644314e-07, "loss": 0.0569, "step": 1392 }, { "epoch": 0.55, "learning_rate": 3.993514089447235e-07, "loss": 0.0456, "step": 1393 }, { "epoch": 0.55, "learning_rate": 3.9934935267589065e-07, "loss": 0.0502, "step": 1394 }, { "epoch": 0.55, "learning_rate": 3.9934729315796626e-07, "loss": 0.0464, "step": 1395 }, { "epoch": 0.55, "learning_rate": 3.9934523039098405e-07, "loss": 0.0431, "step": 1396 }, { "epoch": 0.55, "learning_rate": 3.9934316437497753e-07, "loss": 0.0402, "step": 1397 }, { "epoch": 0.55, "learning_rate": 3.993410951099804e-07, "loss": 0.0467, "step": 1398 }, { "epoch": 0.55, "learning_rate": 3.9933902259602635e-07, "loss": 0.051, "step": 1399 }, { "epoch": 0.55, "learning_rate": 3.993369468331492e-07, "loss": 0.0494, "step": 1400 }, { "epoch": 0.55, "learning_rate": 3.993348678213827e-07, "loss": 0.0437, "step": 1401 }, { "epoch": 0.55, "learning_rate": 3.993327855607609e-07, "loss": 0.0548, "step": 1402 }, { "epoch": 0.55, "learning_rate": 3.9933070005131766e-07, "loss": 0.0456, "step": 1403 }, { "epoch": 0.55, "learning_rate": 3.993286112930869e-07, "loss": 0.0475, "step": 1404 }, { "epoch": 0.55, "learning_rate": 3.993265192861028e-07, "loss": 0.0517, "step": 1405 }, { "epoch": 0.55, "learning_rate": 3.9932442403039934e-07, "loss": 0.0512, "step": 1406 }, { "epoch": 0.55, "learning_rate": 3.9932232552601066e-07, "loss": 0.0529, "step": 1407 }, { "epoch": 0.55, "learning_rate": 3.9932022377297105e-07, "loss": 0.0494, "step": 1408 }, { "epoch": 0.56, "learning_rate": 3.993181187713147e-07, "loss": 0.047, "step": 1409 }, { "epoch": 0.56, "learning_rate": 3.99316010521076e-07, "loss": 0.051, "step": 1410 }, { "epoch": 0.56, "learning_rate": 3.993138990222892e-07, "loss": 0.0498, "step": 1411 }, { "epoch": 0.56, "learning_rate": 3.9931178427498883e-07, "loss": 0.0488, "step": 1412 }, { "epoch": 0.56, "learning_rate": 3.9930966627920927e-07, "loss": 0.0494, "step": 1413 }, { "epoch": 0.56, "learning_rate": 3.99307545034985e-07, "loss": 0.0413, "step": 1414 }, { "epoch": 0.56, "learning_rate": 3.993054205423507e-07, "loss": 0.0499, "step": 1415 }, { "epoch": 0.56, "learning_rate": 3.9930329280134093e-07, "loss": 0.0439, "step": 1416 }, { "epoch": 0.56, "learning_rate": 3.9930116181199046e-07, "loss": 0.0447, "step": 1417 }, { "epoch": 0.56, "learning_rate": 3.9929902757433387e-07, "loss": 0.0462, "step": 1418 }, { "epoch": 0.56, "learning_rate": 3.9929689008840613e-07, "loss": 0.0467, "step": 1419 }, { "epoch": 0.56, "learning_rate": 3.9929474935424193e-07, "loss": 0.0466, "step": 1420 }, { "epoch": 0.56, "learning_rate": 3.992926053718762e-07, "loss": 0.0428, "step": 1421 }, { "epoch": 0.56, "learning_rate": 3.992904581413439e-07, "loss": 0.0476, "step": 1422 }, { "epoch": 0.56, "learning_rate": 3.9928830766267996e-07, "loss": 0.0452, "step": 1423 }, { "epoch": 0.56, "learning_rate": 3.9928615393591957e-07, "loss": 0.0463, "step": 1424 }, { "epoch": 0.56, "learning_rate": 3.992839969610977e-07, "loss": 0.0451, "step": 1425 }, { "epoch": 0.56, "learning_rate": 3.992818367382496e-07, "loss": 0.0473, "step": 1426 }, { "epoch": 0.56, "learning_rate": 3.9927967326741046e-07, "loss": 0.0471, "step": 1427 }, { "epoch": 0.56, "learning_rate": 3.992775065486155e-07, "loss": 0.0483, "step": 1428 }, { "epoch": 0.56, "learning_rate": 3.992753365818999e-07, "loss": 0.0488, "step": 1429 }, { "epoch": 0.56, "learning_rate": 3.992731633672994e-07, "loss": 0.0453, "step": 1430 }, { "epoch": 0.56, "learning_rate": 3.9927098690484906e-07, "loss": 0.0465, "step": 1431 }, { "epoch": 0.56, "learning_rate": 3.992688071945845e-07, "loss": 0.0562, "step": 1432 }, { "epoch": 0.56, "learning_rate": 3.992666242365413e-07, "loss": 0.0465, "step": 1433 }, { "epoch": 0.57, "learning_rate": 3.992644380307549e-07, "loss": 0.0462, "step": 1434 }, { "epoch": 0.57, "learning_rate": 3.992622485772611e-07, "loss": 0.0494, "step": 1435 }, { "epoch": 0.57, "learning_rate": 3.992600558760954e-07, "loss": 0.0489, "step": 1436 }, { "epoch": 0.57, "learning_rate": 3.9925785992729367e-07, "loss": 0.0484, "step": 1437 }, { "epoch": 0.57, "learning_rate": 3.992556607308916e-07, "loss": 0.0481, "step": 1438 }, { "epoch": 0.57, "learning_rate": 3.9925345828692524e-07, "loss": 0.0432, "step": 1439 }, { "epoch": 0.57, "learning_rate": 3.992512525954302e-07, "loss": 0.0495, "step": 1440 }, { "epoch": 0.57, "learning_rate": 3.9924904365644265e-07, "loss": 0.0469, "step": 1441 }, { "epoch": 0.57, "learning_rate": 3.992468314699985e-07, "loss": 0.0416, "step": 1442 }, { "epoch": 0.57, "learning_rate": 3.9924461603613377e-07, "loss": 0.0442, "step": 1443 }, { "epoch": 0.57, "learning_rate": 3.9924239735488466e-07, "loss": 0.0463, "step": 1444 }, { "epoch": 0.57, "learning_rate": 3.9924017542628724e-07, "loss": 0.0487, "step": 1445 }, { "epoch": 0.57, "learning_rate": 3.9923795025037785e-07, "loss": 0.0433, "step": 1446 }, { "epoch": 0.57, "learning_rate": 3.992357218271926e-07, "loss": 0.048, "step": 1447 }, { "epoch": 0.57, "learning_rate": 3.99233490156768e-07, "loss": 0.0478, "step": 1448 }, { "epoch": 0.57, "learning_rate": 3.9923125523914017e-07, "loss": 0.0434, "step": 1449 }, { "epoch": 0.57, "learning_rate": 3.992290170743457e-07, "loss": 0.041, "step": 1450 }, { "epoch": 0.57, "learning_rate": 3.992267756624211e-07, "loss": 0.0449, "step": 1451 }, { "epoch": 0.57, "learning_rate": 3.992245310034028e-07, "loss": 0.0492, "step": 1452 }, { "epoch": 0.57, "learning_rate": 3.992222830973275e-07, "loss": 0.0456, "step": 1453 }, { "epoch": 0.57, "learning_rate": 3.9922003194423173e-07, "loss": 0.0471, "step": 1454 }, { "epoch": 0.57, "learning_rate": 3.992177775441522e-07, "loss": 0.0438, "step": 1455 }, { "epoch": 0.57, "learning_rate": 3.992155198971257e-07, "loss": 0.0504, "step": 1456 }, { "epoch": 0.57, "learning_rate": 3.99213259003189e-07, "loss": 0.047, "step": 1457 }, { "epoch": 0.57, "learning_rate": 3.992109948623789e-07, "loss": 0.0486, "step": 1458 }, { "epoch": 0.58, "learning_rate": 3.992087274747324e-07, "loss": 0.0446, "step": 1459 }, { "epoch": 0.58, "learning_rate": 3.992064568402864e-07, "loss": 0.0496, "step": 1460 }, { "epoch": 0.58, "learning_rate": 3.9920418295907783e-07, "loss": 0.0463, "step": 1461 }, { "epoch": 0.58, "learning_rate": 3.9920190583114394e-07, "loss": 0.0527, "step": 1462 }, { "epoch": 0.58, "learning_rate": 3.991996254565217e-07, "loss": 0.0437, "step": 1463 }, { "epoch": 0.58, "learning_rate": 3.991973418352483e-07, "loss": 0.0514, "step": 1464 }, { "epoch": 0.58, "learning_rate": 3.99195054967361e-07, "loss": 0.0525, "step": 1465 }, { "epoch": 0.58, "learning_rate": 3.9919276485289705e-07, "loss": 0.0446, "step": 1466 }, { "epoch": 0.58, "learning_rate": 3.991904714918937e-07, "loss": 0.0445, "step": 1467 }, { "epoch": 0.58, "learning_rate": 3.991881748843885e-07, "loss": 0.0477, "step": 1468 }, { "epoch": 0.58, "learning_rate": 3.9918587503041873e-07, "loss": 0.0461, "step": 1469 }, { "epoch": 0.58, "learning_rate": 3.9918357193002194e-07, "loss": 0.0438, "step": 1470 }, { "epoch": 0.58, "learning_rate": 3.9918126558323564e-07, "loss": 0.0434, "step": 1471 }, { "epoch": 0.58, "learning_rate": 3.991789559900974e-07, "loss": 0.0471, "step": 1472 }, { "epoch": 0.58, "learning_rate": 3.991766431506449e-07, "loss": 0.047, "step": 1473 }, { "epoch": 0.58, "learning_rate": 3.991743270649159e-07, "loss": 0.0481, "step": 1474 }, { "epoch": 0.58, "learning_rate": 3.9917200773294806e-07, "loss": 0.0482, "step": 1475 }, { "epoch": 0.58, "learning_rate": 3.9916968515477914e-07, "loss": 0.0462, "step": 1476 }, { "epoch": 0.58, "learning_rate": 3.991673593304471e-07, "loss": 0.0468, "step": 1477 }, { "epoch": 0.58, "learning_rate": 3.991650302599898e-07, "loss": 0.0483, "step": 1478 }, { "epoch": 0.58, "learning_rate": 3.991626979434452e-07, "loss": 0.0518, "step": 1479 }, { "epoch": 0.58, "learning_rate": 3.991603623808513e-07, "loss": 0.0423, "step": 1480 }, { "epoch": 0.58, "learning_rate": 3.9915802357224615e-07, "loss": 0.0511, "step": 1481 }, { "epoch": 0.58, "learning_rate": 3.99155681517668e-07, "loss": 0.048, "step": 1482 }, { "epoch": 0.58, "learning_rate": 3.991533362171548e-07, "loss": 0.0441, "step": 1483 }, { "epoch": 0.58, "learning_rate": 3.99150987670745e-07, "loss": 0.0458, "step": 1484 }, { "epoch": 0.59, "learning_rate": 3.9914863587847676e-07, "loss": 0.0454, "step": 1485 }, { "epoch": 0.59, "learning_rate": 3.991462808403884e-07, "loss": 0.0504, "step": 1486 }, { "epoch": 0.59, "learning_rate": 3.9914392255651833e-07, "loss": 0.0486, "step": 1487 }, { "epoch": 0.59, "learning_rate": 3.99141561026905e-07, "loss": 0.049, "step": 1488 }, { "epoch": 0.59, "learning_rate": 3.991391962515869e-07, "loss": 0.049, "step": 1489 }, { "epoch": 0.59, "learning_rate": 3.991368282306025e-07, "loss": 0.051, "step": 1490 }, { "epoch": 0.59, "learning_rate": 3.991344569639905e-07, "loss": 0.0448, "step": 1491 }, { "epoch": 0.59, "learning_rate": 3.9913208245178946e-07, "loss": 0.0441, "step": 1492 }, { "epoch": 0.59, "learning_rate": 3.991297046940382e-07, "loss": 0.0447, "step": 1493 }, { "epoch": 0.59, "learning_rate": 3.991273236907753e-07, "loss": 0.0419, "step": 1494 }, { "epoch": 0.59, "learning_rate": 3.991249394420397e-07, "loss": 0.0434, "step": 1495 }, { "epoch": 0.59, "learning_rate": 3.9912255194787024e-07, "loss": 0.044, "step": 1496 }, { "epoch": 0.59, "learning_rate": 3.9912016120830583e-07, "loss": 0.0506, "step": 1497 }, { "epoch": 0.59, "learning_rate": 3.991177672233854e-07, "loss": 0.0464, "step": 1498 }, { "epoch": 0.59, "learning_rate": 3.99115369993148e-07, "loss": 0.0471, "step": 1499 }, { "epoch": 0.59, "learning_rate": 3.991129695176327e-07, "loss": 0.0506, "step": 1500 }, { "epoch": 0.59, "learning_rate": 3.9911056579687855e-07, "loss": 0.0444, "step": 1501 }, { "epoch": 0.59, "learning_rate": 3.9910815883092484e-07, "loss": 0.0396, "step": 1502 }, { "epoch": 0.59, "learning_rate": 3.9910574861981076e-07, "loss": 0.0491, "step": 1503 }, { "epoch": 0.59, "learning_rate": 3.9910333516357556e-07, "loss": 0.0492, "step": 1504 }, { "epoch": 0.59, "learning_rate": 3.991009184622586e-07, "loss": 0.0492, "step": 1505 }, { "epoch": 0.59, "learning_rate": 3.9909849851589926e-07, "loss": 0.0429, "step": 1506 }, { "epoch": 0.59, "learning_rate": 3.9909607532453704e-07, "loss": 0.0478, "step": 1507 }, { "epoch": 0.59, "learning_rate": 3.990936488882113e-07, "loss": 0.0483, "step": 1508 }, { "epoch": 0.59, "learning_rate": 3.9909121920696174e-07, "loss": 0.042, "step": 1509 }, { "epoch": 0.6, "learning_rate": 3.9908878628082784e-07, "loss": 0.0465, "step": 1510 }, { "epoch": 0.6, "learning_rate": 3.990863501098493e-07, "loss": 0.042, "step": 1511 }, { "epoch": 0.6, "learning_rate": 3.990839106940659e-07, "loss": 0.0429, "step": 1512 }, { "epoch": 0.6, "learning_rate": 3.990814680335173e-07, "loss": 0.045, "step": 1513 }, { "epoch": 0.6, "learning_rate": 3.9907902212824327e-07, "loss": 0.0467, "step": 1514 }, { "epoch": 0.6, "learning_rate": 3.9907657297828383e-07, "loss": 0.049, "step": 1515 }, { "epoch": 0.6, "learning_rate": 3.9907412058367876e-07, "loss": 0.0488, "step": 1516 }, { "epoch": 0.6, "learning_rate": 3.990716649444681e-07, "loss": 0.0438, "step": 1517 }, { "epoch": 0.6, "learning_rate": 3.9906920606069183e-07, "loss": 0.0479, "step": 1518 }, { "epoch": 0.6, "learning_rate": 3.9906674393239013e-07, "loss": 0.0464, "step": 1519 }, { "epoch": 0.6, "learning_rate": 3.99064278559603e-07, "loss": 0.045, "step": 1520 }, { "epoch": 0.6, "learning_rate": 3.9906180994237066e-07, "loss": 0.0487, "step": 1521 }, { "epoch": 0.6, "learning_rate": 3.990593380807334e-07, "loss": 0.0485, "step": 1522 }, { "epoch": 0.6, "learning_rate": 3.9905686297473145e-07, "loss": 0.0458, "step": 1523 }, { "epoch": 0.6, "learning_rate": 3.9905438462440514e-07, "loss": 0.0476, "step": 1524 }, { "epoch": 0.6, "learning_rate": 3.990519030297949e-07, "loss": 0.0445, "step": 1525 }, { "epoch": 0.6, "learning_rate": 3.9904941819094115e-07, "loss": 0.0489, "step": 1526 }, { "epoch": 0.6, "learning_rate": 3.990469301078844e-07, "loss": 0.0512, "step": 1527 }, { "epoch": 0.6, "learning_rate": 3.990444387806652e-07, "loss": 0.0472, "step": 1528 }, { "epoch": 0.6, "learning_rate": 3.990419442093242e-07, "loss": 0.05, "step": 1529 }, { "epoch": 0.6, "learning_rate": 3.9903944639390204e-07, "loss": 0.0457, "step": 1530 }, { "epoch": 0.6, "learning_rate": 3.9903694533443935e-07, "loss": 0.0505, "step": 1531 }, { "epoch": 0.6, "learning_rate": 3.99034441030977e-07, "loss": 0.0485, "step": 1532 }, { "epoch": 0.6, "learning_rate": 3.9903193348355567e-07, "loss": 0.0458, "step": 1533 }, { "epoch": 0.6, "learning_rate": 3.990294226922164e-07, "loss": 0.0444, "step": 1534 }, { "epoch": 0.6, "learning_rate": 3.9902690865700005e-07, "loss": 0.0422, "step": 1535 }, { "epoch": 0.61, "learning_rate": 3.9902439137794756e-07, "loss": 0.0451, "step": 1536 }, { "epoch": 0.61, "learning_rate": 3.990218708550999e-07, "loss": 0.0461, "step": 1537 }, { "epoch": 0.61, "learning_rate": 3.990193470884983e-07, "loss": 0.0405, "step": 1538 }, { "epoch": 0.61, "learning_rate": 3.990168200781838e-07, "loss": 0.0473, "step": 1539 }, { "epoch": 0.61, "learning_rate": 3.9901428982419754e-07, "loss": 0.0501, "step": 1540 }, { "epoch": 0.61, "learning_rate": 3.990117563265809e-07, "loss": 0.0467, "step": 1541 }, { "epoch": 0.61, "learning_rate": 3.9900921958537504e-07, "loss": 0.0502, "step": 1542 }, { "epoch": 0.61, "learning_rate": 3.990066796006214e-07, "loss": 0.0421, "step": 1543 }, { "epoch": 0.61, "learning_rate": 3.9900413637236133e-07, "loss": 0.0474, "step": 1544 }, { "epoch": 0.61, "learning_rate": 3.990015899006363e-07, "loss": 0.0518, "step": 1545 }, { "epoch": 0.61, "learning_rate": 3.989990401854878e-07, "loss": 0.0458, "step": 1546 }, { "epoch": 0.61, "learning_rate": 3.989964872269573e-07, "loss": 0.0468, "step": 1547 }, { "epoch": 0.61, "learning_rate": 3.9899393102508665e-07, "loss": 0.0462, "step": 1548 }, { "epoch": 0.61, "learning_rate": 3.989913715799173e-07, "loss": 0.0472, "step": 1549 }, { "epoch": 0.61, "learning_rate": 3.9898880889149097e-07, "loss": 0.0407, "step": 1550 }, { "epoch": 0.61, "learning_rate": 3.989862429598495e-07, "loss": 0.0448, "step": 1551 }, { "epoch": 0.61, "learning_rate": 3.989836737850347e-07, "loss": 0.044, "step": 1552 }, { "epoch": 0.61, "learning_rate": 3.9898110136708853e-07, "loss": 0.0433, "step": 1553 }, { "epoch": 0.61, "learning_rate": 3.989785257060527e-07, "loss": 0.0425, "step": 1554 }, { "epoch": 0.61, "learning_rate": 3.989759468019694e-07, "loss": 0.045, "step": 1555 }, { "epoch": 0.61, "learning_rate": 3.989733646548805e-07, "loss": 0.0412, "step": 1556 }, { "epoch": 0.61, "learning_rate": 3.989707792648282e-07, "loss": 0.0504, "step": 1557 }, { "epoch": 0.61, "learning_rate": 3.9896819063185463e-07, "loss": 0.0453, "step": 1558 }, { "epoch": 0.61, "learning_rate": 3.9896559875600194e-07, "loss": 0.0461, "step": 1559 }, { "epoch": 0.61, "learning_rate": 3.989630036373123e-07, "loss": 0.0491, "step": 1560 }, { "epoch": 0.62, "learning_rate": 3.989604052758282e-07, "loss": 0.0452, "step": 1561 }, { "epoch": 0.62, "learning_rate": 3.989578036715918e-07, "loss": 0.0428, "step": 1562 }, { "epoch": 0.62, "learning_rate": 3.9895519882464567e-07, "loss": 0.047, "step": 1563 }, { "epoch": 0.62, "learning_rate": 3.9895259073503204e-07, "loss": 0.0426, "step": 1564 }, { "epoch": 0.62, "learning_rate": 3.989499794027937e-07, "loss": 0.0474, "step": 1565 }, { "epoch": 0.62, "learning_rate": 3.9894736482797293e-07, "loss": 0.0479, "step": 1566 }, { "epoch": 0.62, "learning_rate": 3.9894474701061254e-07, "loss": 0.0446, "step": 1567 }, { "epoch": 0.62, "learning_rate": 3.989421259507551e-07, "loss": 0.0438, "step": 1568 }, { "epoch": 0.62, "learning_rate": 3.989395016484435e-07, "loss": 0.0478, "step": 1569 }, { "epoch": 0.62, "learning_rate": 3.989368741037203e-07, "loss": 0.0505, "step": 1570 }, { "epoch": 0.62, "learning_rate": 3.9893424331662835e-07, "loss": 0.0508, "step": 1571 }, { "epoch": 0.62, "learning_rate": 3.989316092872106e-07, "loss": 0.044, "step": 1572 }, { "epoch": 0.62, "learning_rate": 3.9892897201551e-07, "loss": 0.0445, "step": 1573 }, { "epoch": 0.62, "learning_rate": 3.9892633150156947e-07, "loss": 0.0466, "step": 1574 }, { "epoch": 0.62, "learning_rate": 3.98923687745432e-07, "loss": 0.0508, "step": 1575 }, { "epoch": 0.62, "learning_rate": 3.9892104074714086e-07, "loss": 0.044, "step": 1576 }, { "epoch": 0.62, "learning_rate": 3.98918390506739e-07, "loss": 0.0475, "step": 1577 }, { "epoch": 0.62, "learning_rate": 3.9891573702426974e-07, "loss": 0.0493, "step": 1578 }, { "epoch": 0.62, "learning_rate": 3.9891308029977626e-07, "loss": 0.0442, "step": 1579 }, { "epoch": 0.62, "learning_rate": 3.989104203333019e-07, "loss": 0.046, "step": 1580 }, { "epoch": 0.62, "learning_rate": 3.9890775712489e-07, "loss": 0.0485, "step": 1581 }, { "epoch": 0.62, "learning_rate": 3.9890509067458397e-07, "loss": 0.0422, "step": 1582 }, { "epoch": 0.62, "learning_rate": 3.989024209824272e-07, "loss": 0.0422, "step": 1583 }, { "epoch": 0.62, "learning_rate": 3.9889974804846335e-07, "loss": 0.0465, "step": 1584 }, { "epoch": 0.62, "learning_rate": 3.9889707187273587e-07, "loss": 0.0439, "step": 1585 }, { "epoch": 0.63, "learning_rate": 3.9889439245528835e-07, "loss": 0.0436, "step": 1586 }, { "epoch": 0.63, "learning_rate": 3.9889170979616465e-07, "loss": 0.0498, "step": 1587 }, { "epoch": 0.63, "learning_rate": 3.9888902389540827e-07, "loss": 0.045, "step": 1588 }, { "epoch": 0.63, "learning_rate": 3.9888633475306304e-07, "loss": 0.0467, "step": 1589 }, { "epoch": 0.63, "learning_rate": 3.9888364236917286e-07, "loss": 0.0453, "step": 1590 }, { "epoch": 0.63, "learning_rate": 3.988809467437816e-07, "loss": 0.0472, "step": 1591 }, { "epoch": 0.63, "learning_rate": 3.988782478769332e-07, "loss": 0.047, "step": 1592 }, { "epoch": 0.63, "learning_rate": 3.9887554576867157e-07, "loss": 0.049, "step": 1593 }, { "epoch": 0.63, "learning_rate": 3.988728404190408e-07, "loss": 0.0538, "step": 1594 }, { "epoch": 0.63, "learning_rate": 3.98870131828085e-07, "loss": 0.0475, "step": 1595 }, { "epoch": 0.63, "learning_rate": 3.988674199958483e-07, "loss": 0.0464, "step": 1596 }, { "epoch": 0.63, "learning_rate": 3.9886470492237486e-07, "loss": 0.0484, "step": 1597 }, { "epoch": 0.63, "learning_rate": 3.9886198660770905e-07, "loss": 0.0462, "step": 1598 }, { "epoch": 0.63, "learning_rate": 3.98859265051895e-07, "loss": 0.0483, "step": 1599 }, { "epoch": 0.63, "learning_rate": 3.9885654025497725e-07, "loss": 0.0428, "step": 1600 }, { "epoch": 0.63, "learning_rate": 3.9885381221700006e-07, "loss": 0.0418, "step": 1601 }, { "epoch": 0.63, "learning_rate": 3.98851080938008e-07, "loss": 0.0463, "step": 1602 }, { "epoch": 0.63, "learning_rate": 3.988483464180455e-07, "loss": 0.0454, "step": 1603 }, { "epoch": 0.63, "learning_rate": 3.988456086571572e-07, "loss": 0.0486, "step": 1604 }, { "epoch": 0.63, "learning_rate": 3.988428676553877e-07, "loss": 0.048, "step": 1605 }, { "epoch": 0.63, "learning_rate": 3.9884012341278156e-07, "loss": 0.049, "step": 1606 }, { "epoch": 0.63, "learning_rate": 3.9883737592938366e-07, "loss": 0.0445, "step": 1607 }, { "epoch": 0.63, "learning_rate": 3.9883462520523876e-07, "loss": 0.0456, "step": 1608 }, { "epoch": 0.63, "learning_rate": 3.9883187124039165e-07, "loss": 0.0475, "step": 1609 }, { "epoch": 0.63, "learning_rate": 3.9882911403488727e-07, "loss": 0.0431, "step": 1610 }, { "epoch": 0.63, "learning_rate": 3.9882635358877047e-07, "loss": 0.0477, "step": 1611 }, { "epoch": 0.64, "learning_rate": 3.9882358990208635e-07, "loss": 0.0419, "step": 1612 }, { "epoch": 0.64, "learning_rate": 3.988208229748798e-07, "loss": 0.0454, "step": 1613 }, { "epoch": 0.64, "learning_rate": 3.9881805280719603e-07, "loss": 0.0477, "step": 1614 }, { "epoch": 0.64, "learning_rate": 3.988152793990802e-07, "loss": 0.0403, "step": 1615 }, { "epoch": 0.64, "learning_rate": 3.9881250275057747e-07, "loss": 0.0471, "step": 1616 }, { "epoch": 0.64, "learning_rate": 3.988097228617331e-07, "loss": 0.0469, "step": 1617 }, { "epoch": 0.64, "learning_rate": 3.988069397325924e-07, "loss": 0.0508, "step": 1618 }, { "epoch": 0.64, "learning_rate": 3.9880415336320077e-07, "loss": 0.048, "step": 1619 }, { "epoch": 0.64, "learning_rate": 3.9880136375360354e-07, "loss": 0.0445, "step": 1620 }, { "epoch": 0.64, "learning_rate": 3.9879857090384625e-07, "loss": 0.0448, "step": 1621 }, { "epoch": 0.64, "learning_rate": 3.9879577481397433e-07, "loss": 0.0441, "step": 1622 }, { "epoch": 0.64, "learning_rate": 3.987929754840335e-07, "loss": 0.042, "step": 1623 }, { "epoch": 0.64, "learning_rate": 3.987901729140693e-07, "loss": 0.0451, "step": 1624 }, { "epoch": 0.64, "learning_rate": 3.9878736710412737e-07, "loss": 0.0434, "step": 1625 }, { "epoch": 0.64, "learning_rate": 3.9878455805425346e-07, "loss": 0.0391, "step": 1626 }, { "epoch": 0.64, "learning_rate": 3.987817457644934e-07, "loss": 0.0478, "step": 1627 }, { "epoch": 0.64, "learning_rate": 3.98778930234893e-07, "loss": 0.0441, "step": 1628 }, { "epoch": 0.64, "learning_rate": 3.9877611146549813e-07, "loss": 0.0502, "step": 1629 }, { "epoch": 0.64, "learning_rate": 3.9877328945635476e-07, "loss": 0.0454, "step": 1630 }, { "epoch": 0.64, "learning_rate": 3.987704642075089e-07, "loss": 0.0426, "step": 1631 }, { "epoch": 0.64, "learning_rate": 3.987676357190065e-07, "loss": 0.0438, "step": 1632 }, { "epoch": 0.64, "learning_rate": 3.9876480399089383e-07, "loss": 0.043, "step": 1633 }, { "epoch": 0.64, "learning_rate": 3.9876196902321686e-07, "loss": 0.0437, "step": 1634 }, { "epoch": 0.64, "learning_rate": 3.987591308160219e-07, "loss": 0.0423, "step": 1635 }, { "epoch": 0.64, "learning_rate": 3.987562893693552e-07, "loss": 0.0456, "step": 1636 }, { "epoch": 0.65, "learning_rate": 3.9875344468326304e-07, "loss": 0.0505, "step": 1637 }, { "epoch": 0.65, "learning_rate": 3.987505967577918e-07, "loss": 0.0484, "step": 1638 }, { "epoch": 0.65, "learning_rate": 3.9874774559298793e-07, "loss": 0.0449, "step": 1639 }, { "epoch": 0.65, "learning_rate": 3.987448911888978e-07, "loss": 0.0428, "step": 1640 }, { "epoch": 0.65, "learning_rate": 3.987420335455681e-07, "loss": 0.0493, "step": 1641 }, { "epoch": 0.65, "learning_rate": 3.987391726630452e-07, "loss": 0.0474, "step": 1642 }, { "epoch": 0.65, "learning_rate": 3.987363085413759e-07, "loss": 0.0409, "step": 1643 }, { "epoch": 0.65, "learning_rate": 3.987334411806068e-07, "loss": 0.0442, "step": 1644 }, { "epoch": 0.65, "learning_rate": 3.987305705807847e-07, "loss": 0.0434, "step": 1645 }, { "epoch": 0.65, "learning_rate": 3.9872769674195626e-07, "loss": 0.046, "step": 1646 }, { "epoch": 0.65, "learning_rate": 3.987248196641684e-07, "loss": 0.0431, "step": 1647 }, { "epoch": 0.65, "learning_rate": 3.9872193934746806e-07, "loss": 0.0404, "step": 1648 }, { "epoch": 0.65, "learning_rate": 3.987190557919021e-07, "loss": 0.0445, "step": 1649 }, { "epoch": 0.65, "learning_rate": 3.9871616899751754e-07, "loss": 0.0425, "step": 1650 }, { "epoch": 0.65, "learning_rate": 3.9871327896436144e-07, "loss": 0.0411, "step": 1651 }, { "epoch": 0.65, "learning_rate": 3.9871038569248096e-07, "loss": 0.0499, "step": 1652 }, { "epoch": 0.65, "learning_rate": 3.987074891819231e-07, "loss": 0.0427, "step": 1653 }, { "epoch": 0.65, "learning_rate": 3.987045894327352e-07, "loss": 0.0425, "step": 1654 }, { "epoch": 0.65, "learning_rate": 3.9870168644496453e-07, "loss": 0.0471, "step": 1655 }, { "epoch": 0.65, "learning_rate": 3.9869878021865834e-07, "loss": 0.0438, "step": 1656 }, { "epoch": 0.65, "learning_rate": 3.9869587075386397e-07, "loss": 0.0458, "step": 1657 }, { "epoch": 0.65, "learning_rate": 3.98692958050629e-07, "loss": 0.0471, "step": 1658 }, { "epoch": 0.65, "learning_rate": 3.986900421090007e-07, "loss": 0.0497, "step": 1659 }, { "epoch": 0.65, "learning_rate": 3.986871229290267e-07, "loss": 0.051, "step": 1660 }, { "epoch": 0.65, "learning_rate": 3.986842005107546e-07, "loss": 0.046, "step": 1661 }, { "epoch": 0.66, "learning_rate": 3.98681274854232e-07, "loss": 0.0467, "step": 1662 }, { "epoch": 0.66, "learning_rate": 3.986783459595065e-07, "loss": 0.045, "step": 1663 }, { "epoch": 0.66, "learning_rate": 3.9867541382662597e-07, "loss": 0.0413, "step": 1664 }, { "epoch": 0.66, "learning_rate": 3.986724784556382e-07, "loss": 0.0416, "step": 1665 }, { "epoch": 0.66, "learning_rate": 3.9866953984659093e-07, "loss": 0.0419, "step": 1666 }, { "epoch": 0.66, "learning_rate": 3.9866659799953207e-07, "loss": 0.0421, "step": 1667 }, { "epoch": 0.66, "learning_rate": 3.986636529145096e-07, "loss": 0.0455, "step": 1668 }, { "epoch": 0.66, "learning_rate": 3.986607045915716e-07, "loss": 0.046, "step": 1669 }, { "epoch": 0.66, "learning_rate": 3.9865775303076604e-07, "loss": 0.0418, "step": 1670 }, { "epoch": 0.66, "learning_rate": 3.9865479823214095e-07, "loss": 0.0433, "step": 1671 }, { "epoch": 0.66, "learning_rate": 3.986518401957446e-07, "loss": 0.0393, "step": 1672 }, { "epoch": 0.66, "learning_rate": 3.9864887892162515e-07, "loss": 0.0454, "step": 1673 }, { "epoch": 0.66, "learning_rate": 3.98645914409831e-07, "loss": 0.0451, "step": 1674 }, { "epoch": 0.66, "learning_rate": 3.986429466604102e-07, "loss": 0.0465, "step": 1675 }, { "epoch": 0.66, "learning_rate": 3.986399756734114e-07, "loss": 0.0485, "step": 1676 }, { "epoch": 0.66, "learning_rate": 3.9863700144888284e-07, "loss": 0.0406, "step": 1677 }, { "epoch": 0.66, "learning_rate": 3.98634023986873e-07, "loss": 0.0475, "step": 1678 }, { "epoch": 0.66, "learning_rate": 3.986310432874305e-07, "loss": 0.0475, "step": 1679 }, { "epoch": 0.66, "learning_rate": 3.986280593506039e-07, "loss": 0.0463, "step": 1680 }, { "epoch": 0.66, "learning_rate": 3.986250721764418e-07, "loss": 0.0496, "step": 1681 }, { "epoch": 0.66, "learning_rate": 3.986220817649929e-07, "loss": 0.0404, "step": 1682 }, { "epoch": 0.66, "learning_rate": 3.986190881163059e-07, "loss": 0.0456, "step": 1683 }, { "epoch": 0.66, "learning_rate": 3.9861609123042966e-07, "loss": 0.0422, "step": 1684 }, { "epoch": 0.66, "learning_rate": 3.9861309110741297e-07, "loss": 0.0448, "step": 1685 }, { "epoch": 0.66, "learning_rate": 3.986100877473048e-07, "loss": 0.0479, "step": 1686 }, { "epoch": 0.66, "learning_rate": 3.9860708115015405e-07, "loss": 0.0465, "step": 1687 }, { "epoch": 0.67, "learning_rate": 3.986040713160097e-07, "loss": 0.0437, "step": 1688 }, { "epoch": 0.67, "learning_rate": 3.9860105824492086e-07, "loss": 0.0415, "step": 1689 }, { "epoch": 0.67, "learning_rate": 3.9859804193693657e-07, "loss": 0.0479, "step": 1690 }, { "epoch": 0.67, "learning_rate": 3.9859502239210603e-07, "loss": 0.049, "step": 1691 }, { "epoch": 0.67, "learning_rate": 3.9859199961047853e-07, "loss": 0.047, "step": 1692 }, { "epoch": 0.67, "learning_rate": 3.985889735921032e-07, "loss": 0.0477, "step": 1693 }, { "epoch": 0.67, "learning_rate": 3.9858594433702936e-07, "loss": 0.0448, "step": 1694 }, { "epoch": 0.67, "learning_rate": 3.9858291184530657e-07, "loss": 0.0474, "step": 1695 }, { "epoch": 0.67, "learning_rate": 3.98579876116984e-07, "loss": 0.0449, "step": 1696 }, { "epoch": 0.67, "learning_rate": 3.9857683715211137e-07, "loss": 0.0473, "step": 1697 }, { "epoch": 0.67, "learning_rate": 3.9857379495073805e-07, "loss": 0.0453, "step": 1698 }, { "epoch": 0.67, "learning_rate": 3.985707495129136e-07, "loss": 0.0428, "step": 1699 }, { "epoch": 0.67, "learning_rate": 3.9856770083868784e-07, "loss": 0.0433, "step": 1700 }, { "epoch": 0.67, "learning_rate": 3.985646489281103e-07, "loss": 0.045, "step": 1701 }, { "epoch": 0.67, "learning_rate": 3.9856159378123077e-07, "loss": 0.0473, "step": 1702 }, { "epoch": 0.67, "learning_rate": 3.9855853539809905e-07, "loss": 0.0462, "step": 1703 }, { "epoch": 0.67, "learning_rate": 3.9855547377876493e-07, "loss": 0.049, "step": 1704 }, { "epoch": 0.67, "learning_rate": 3.985524089232784e-07, "loss": 0.0476, "step": 1705 }, { "epoch": 0.67, "learning_rate": 3.9854934083168934e-07, "loss": 0.0427, "step": 1706 }, { "epoch": 0.67, "learning_rate": 3.9854626950404786e-07, "loss": 0.0426, "step": 1707 }, { "epoch": 0.67, "learning_rate": 3.985431949404038e-07, "loss": 0.0458, "step": 1708 }, { "epoch": 0.67, "learning_rate": 3.9854011714080755e-07, "loss": 0.0421, "step": 1709 }, { "epoch": 0.67, "learning_rate": 3.985370361053091e-07, "loss": 0.0443, "step": 1710 }, { "epoch": 0.67, "learning_rate": 3.9853395183395867e-07, "loss": 0.0501, "step": 1711 }, { "epoch": 0.67, "learning_rate": 3.9853086432680656e-07, "loss": 0.0419, "step": 1712 }, { "epoch": 0.68, "learning_rate": 3.9852777358390315e-07, "loss": 0.0471, "step": 1713 }, { "epoch": 0.68, "learning_rate": 3.985246796052987e-07, "loss": 0.0393, "step": 1714 }, { "epoch": 0.68, "learning_rate": 3.985215823910437e-07, "loss": 0.0497, "step": 1715 }, { "epoch": 0.68, "learning_rate": 3.9851848194118865e-07, "loss": 0.0471, "step": 1716 }, { "epoch": 0.68, "learning_rate": 3.98515378255784e-07, "loss": 0.0453, "step": 1717 }, { "epoch": 0.68, "learning_rate": 3.985122713348805e-07, "loss": 0.0459, "step": 1718 }, { "epoch": 0.68, "learning_rate": 3.985091611785286e-07, "loss": 0.0472, "step": 1719 }, { "epoch": 0.68, "learning_rate": 3.9850604778677915e-07, "loss": 0.0478, "step": 1720 }, { "epoch": 0.68, "learning_rate": 3.985029311596827e-07, "loss": 0.046, "step": 1721 }, { "epoch": 0.68, "learning_rate": 3.984998112972902e-07, "loss": 0.0425, "step": 1722 }, { "epoch": 0.68, "learning_rate": 3.984966881996525e-07, "loss": 0.0471, "step": 1723 }, { "epoch": 0.68, "learning_rate": 3.9849356186682045e-07, "loss": 0.04, "step": 1724 }, { "epoch": 0.68, "learning_rate": 3.98490432298845e-07, "loss": 0.0446, "step": 1725 }, { "epoch": 0.68, "learning_rate": 3.9848729949577724e-07, "loss": 0.0488, "step": 1726 }, { "epoch": 0.68, "learning_rate": 3.984841634576681e-07, "loss": 0.0521, "step": 1727 }, { "epoch": 0.68, "learning_rate": 3.9848102418456876e-07, "loss": 0.0421, "step": 1728 }, { "epoch": 0.68, "learning_rate": 3.9847788167653035e-07, "loss": 0.0457, "step": 1729 }, { "epoch": 0.68, "learning_rate": 3.9847473593360423e-07, "loss": 0.044, "step": 1730 }, { "epoch": 0.68, "learning_rate": 3.9847158695584145e-07, "loss": 0.0428, "step": 1731 }, { "epoch": 0.68, "learning_rate": 3.9846843474329346e-07, "loss": 0.0448, "step": 1732 }, { "epoch": 0.68, "learning_rate": 3.9846527929601167e-07, "loss": 0.0399, "step": 1733 }, { "epoch": 0.68, "learning_rate": 3.984621206140474e-07, "loss": 0.0424, "step": 1734 }, { "epoch": 0.68, "learning_rate": 3.984589586974523e-07, "loss": 0.0408, "step": 1735 }, { "epoch": 0.68, "learning_rate": 3.984557935462777e-07, "loss": 0.0448, "step": 1736 }, { "epoch": 0.68, "learning_rate": 3.984526251605753e-07, "loss": 0.04, "step": 1737 }, { "epoch": 0.68, "learning_rate": 3.9844945354039675e-07, "loss": 0.0487, "step": 1738 }, { "epoch": 0.69, "learning_rate": 3.984462786857937e-07, "loss": 0.0448, "step": 1739 }, { "epoch": 0.69, "learning_rate": 3.984431005968179e-07, "loss": 0.0419, "step": 1740 }, { "epoch": 0.69, "learning_rate": 3.9843991927352116e-07, "loss": 0.0435, "step": 1741 }, { "epoch": 0.69, "learning_rate": 3.984367347159553e-07, "loss": 0.0409, "step": 1742 }, { "epoch": 0.69, "learning_rate": 3.9843354692417234e-07, "loss": 0.0496, "step": 1743 }, { "epoch": 0.69, "learning_rate": 3.984303558982241e-07, "loss": 0.0445, "step": 1744 }, { "epoch": 0.69, "learning_rate": 3.984271616381626e-07, "loss": 0.0451, "step": 1745 }, { "epoch": 0.69, "learning_rate": 3.9842396414404e-07, "loss": 0.0461, "step": 1746 }, { "epoch": 0.69, "learning_rate": 3.984207634159083e-07, "loss": 0.0409, "step": 1747 }, { "epoch": 0.69, "learning_rate": 3.9841755945381974e-07, "loss": 0.0485, "step": 1748 }, { "epoch": 0.69, "learning_rate": 3.984143522578266e-07, "loss": 0.0435, "step": 1749 }, { "epoch": 0.69, "learning_rate": 3.9841114182798094e-07, "loss": 0.0525, "step": 1750 }, { "epoch": 0.69, "learning_rate": 3.9840792816433533e-07, "loss": 0.0465, "step": 1751 }, { "epoch": 0.69, "learning_rate": 3.9840471126694197e-07, "loss": 0.0452, "step": 1752 }, { "epoch": 0.69, "learning_rate": 3.984014911358534e-07, "loss": 0.0501, "step": 1753 }, { "epoch": 0.69, "learning_rate": 3.9839826777112205e-07, "loss": 0.0414, "step": 1754 }, { "epoch": 0.69, "learning_rate": 3.9839504117280047e-07, "loss": 0.0464, "step": 1755 }, { "epoch": 0.69, "learning_rate": 3.983918113409412e-07, "loss": 0.052, "step": 1756 }, { "epoch": 0.69, "learning_rate": 3.98388578275597e-07, "loss": 0.0414, "step": 1757 }, { "epoch": 0.69, "learning_rate": 3.983853419768204e-07, "loss": 0.0472, "step": 1758 }, { "epoch": 0.69, "learning_rate": 3.9838210244466424e-07, "loss": 0.0397, "step": 1759 }, { "epoch": 0.69, "learning_rate": 3.983788596791814e-07, "loss": 0.0392, "step": 1760 }, { "epoch": 0.69, "learning_rate": 3.983756136804246e-07, "loss": 0.041, "step": 1761 }, { "epoch": 0.69, "learning_rate": 3.9837236444844676e-07, "loss": 0.047, "step": 1762 }, { "epoch": 0.69, "learning_rate": 3.9836911198330095e-07, "loss": 0.0459, "step": 1763 }, { "epoch": 0.7, "learning_rate": 3.9836585628504e-07, "loss": 0.045, "step": 1764 }, { "epoch": 0.7, "learning_rate": 3.983625973537172e-07, "loss": 0.048, "step": 1765 }, { "epoch": 0.7, "learning_rate": 3.9835933518938545e-07, "loss": 0.0466, "step": 1766 }, { "epoch": 0.7, "learning_rate": 3.98356069792098e-07, "loss": 0.0454, "step": 1767 }, { "epoch": 0.7, "learning_rate": 3.9835280116190814e-07, "loss": 0.0444, "step": 1768 }, { "epoch": 0.7, "learning_rate": 3.9834952929886907e-07, "loss": 0.0438, "step": 1769 }, { "epoch": 0.7, "learning_rate": 3.983462542030341e-07, "loss": 0.0463, "step": 1770 }, { "epoch": 0.7, "learning_rate": 3.9834297587445664e-07, "loss": 0.0477, "step": 1771 }, { "epoch": 0.7, "learning_rate": 3.983396943131901e-07, "loss": 0.0405, "step": 1772 }, { "epoch": 0.7, "learning_rate": 3.983364095192881e-07, "loss": 0.0452, "step": 1773 }, { "epoch": 0.7, "learning_rate": 3.983331214928039e-07, "loss": 0.0458, "step": 1774 }, { "epoch": 0.7, "learning_rate": 3.983298302337914e-07, "loss": 0.0403, "step": 1775 }, { "epoch": 0.7, "learning_rate": 3.9832653574230397e-07, "loss": 0.0413, "step": 1776 }, { "epoch": 0.7, "learning_rate": 3.983232380183955e-07, "loss": 0.0446, "step": 1777 }, { "epoch": 0.7, "learning_rate": 3.9831993706211967e-07, "loss": 0.041, "step": 1778 }, { "epoch": 0.7, "learning_rate": 3.9831663287353025e-07, "loss": 0.0413, "step": 1779 }, { "epoch": 0.7, "learning_rate": 3.983133254526811e-07, "loss": 0.0457, "step": 1780 }, { "epoch": 0.7, "learning_rate": 3.983100147996262e-07, "loss": 0.0454, "step": 1781 }, { "epoch": 0.7, "learning_rate": 3.9830670091441936e-07, "loss": 0.0445, "step": 1782 }, { "epoch": 0.7, "learning_rate": 3.983033837971148e-07, "loss": 0.0424, "step": 1783 }, { "epoch": 0.7, "learning_rate": 3.983000634477665e-07, "loss": 0.0399, "step": 1784 }, { "epoch": 0.7, "learning_rate": 3.982967398664284e-07, "loss": 0.0449, "step": 1785 }, { "epoch": 0.7, "learning_rate": 3.982934130531549e-07, "loss": 0.0448, "step": 1786 }, { "epoch": 0.7, "learning_rate": 3.982900830080001e-07, "loss": 0.0434, "step": 1787 }, { "epoch": 0.7, "learning_rate": 3.9828674973101835e-07, "loss": 0.042, "step": 1788 }, { "epoch": 0.71, "learning_rate": 3.982834132222639e-07, "loss": 0.0435, "step": 1789 }, { "epoch": 0.71, "learning_rate": 3.982800734817912e-07, "loss": 0.0419, "step": 1790 }, { "epoch": 0.71, "learning_rate": 3.982767305096547e-07, "loss": 0.038, "step": 1791 }, { "epoch": 0.71, "learning_rate": 3.982733843059088e-07, "loss": 0.0401, "step": 1792 }, { "epoch": 0.71, "learning_rate": 3.9827003487060804e-07, "loss": 0.0409, "step": 1793 }, { "epoch": 0.71, "learning_rate": 3.982666822038071e-07, "loss": 0.0408, "step": 1794 }, { "epoch": 0.71, "learning_rate": 3.9826332630556053e-07, "loss": 0.0422, "step": 1795 }, { "epoch": 0.71, "learning_rate": 3.982599671759231e-07, "loss": 0.0417, "step": 1796 }, { "epoch": 0.71, "learning_rate": 3.982566048149495e-07, "loss": 0.0438, "step": 1797 }, { "epoch": 0.71, "learning_rate": 3.982532392226946e-07, "loss": 0.0489, "step": 1798 }, { "epoch": 0.71, "learning_rate": 3.9824987039921315e-07, "loss": 0.0464, "step": 1799 }, { "epoch": 0.71, "learning_rate": 3.9824649834456015e-07, "loss": 0.0451, "step": 1800 }, { "epoch": 0.71, "learning_rate": 3.982431230587905e-07, "loss": 0.0431, "step": 1801 }, { "epoch": 0.71, "learning_rate": 3.9823974454195923e-07, "loss": 0.0478, "step": 1802 }, { "epoch": 0.71, "learning_rate": 3.982363627941215e-07, "loss": 0.0439, "step": 1803 }, { "epoch": 0.71, "learning_rate": 3.982329778153323e-07, "loss": 0.0468, "step": 1804 }, { "epoch": 0.71, "learning_rate": 3.9822958960564675e-07, "loss": 0.0456, "step": 1805 }, { "epoch": 0.71, "learning_rate": 3.982261981651203e-07, "loss": 0.0468, "step": 1806 }, { "epoch": 0.71, "learning_rate": 3.98222803493808e-07, "loss": 0.0418, "step": 1807 }, { "epoch": 0.71, "learning_rate": 3.982194055917653e-07, "loss": 0.0513, "step": 1808 }, { "epoch": 0.71, "learning_rate": 3.9821600445904753e-07, "loss": 0.0435, "step": 1809 }, { "epoch": 0.71, "learning_rate": 3.982126000957102e-07, "loss": 0.0442, "step": 1810 }, { "epoch": 0.71, "learning_rate": 3.982091925018086e-07, "loss": 0.04, "step": 1811 }, { "epoch": 0.71, "learning_rate": 3.982057816773985e-07, "loss": 0.04, "step": 1812 }, { "epoch": 0.71, "learning_rate": 3.9820236762253546e-07, "loss": 0.0416, "step": 1813 }, { "epoch": 0.71, "learning_rate": 3.98198950337275e-07, "loss": 0.0478, "step": 1814 }, { "epoch": 0.72, "learning_rate": 3.9819552982167287e-07, "loss": 0.0471, "step": 1815 }, { "epoch": 0.72, "learning_rate": 3.9819210607578484e-07, "loss": 0.0437, "step": 1816 }, { "epoch": 0.72, "learning_rate": 3.9818867909966663e-07, "loss": 0.0355, "step": 1817 }, { "epoch": 0.72, "learning_rate": 3.9818524889337424e-07, "loss": 0.0488, "step": 1818 }, { "epoch": 0.72, "learning_rate": 3.981818154569635e-07, "loss": 0.0515, "step": 1819 }, { "epoch": 0.72, "learning_rate": 3.9817837879049035e-07, "loss": 0.0503, "step": 1820 }, { "epoch": 0.72, "learning_rate": 3.981749388940108e-07, "loss": 0.0446, "step": 1821 }, { "epoch": 0.72, "learning_rate": 3.98171495767581e-07, "loss": 0.0442, "step": 1822 }, { "epoch": 0.72, "learning_rate": 3.9816804941125693e-07, "loss": 0.0447, "step": 1823 }, { "epoch": 0.72, "learning_rate": 3.9816459982509487e-07, "loss": 0.0461, "step": 1824 }, { "epoch": 0.72, "learning_rate": 3.98161147009151e-07, "loss": 0.0473, "step": 1825 }, { "epoch": 0.72, "learning_rate": 3.981576909634816e-07, "loss": 0.042, "step": 1826 }, { "epoch": 0.72, "learning_rate": 3.9815423168814306e-07, "loss": 0.0427, "step": 1827 }, { "epoch": 0.72, "learning_rate": 3.981507691831916e-07, "loss": 0.045, "step": 1828 }, { "epoch": 0.72, "learning_rate": 3.9814730344868385e-07, "loss": 0.0463, "step": 1829 }, { "epoch": 0.72, "learning_rate": 3.9814383448467615e-07, "loss": 0.045, "step": 1830 }, { "epoch": 0.72, "learning_rate": 3.981403622912251e-07, "loss": 0.0511, "step": 1831 }, { "epoch": 0.72, "learning_rate": 3.981368868683873e-07, "loss": 0.0415, "step": 1832 }, { "epoch": 0.72, "learning_rate": 3.9813340821621935e-07, "loss": 0.0487, "step": 1833 }, { "epoch": 0.72, "learning_rate": 3.98129926334778e-07, "loss": 0.0419, "step": 1834 }, { "epoch": 0.72, "learning_rate": 3.9812644122412e-07, "loss": 0.0464, "step": 1835 }, { "epoch": 0.72, "learning_rate": 3.9812295288430206e-07, "loss": 0.0407, "step": 1836 }, { "epoch": 0.72, "learning_rate": 3.9811946131538115e-07, "loss": 0.0415, "step": 1837 }, { "epoch": 0.72, "learning_rate": 3.9811596651741415e-07, "loss": 0.044, "step": 1838 }, { "epoch": 0.72, "learning_rate": 3.9811246849045787e-07, "loss": 0.0362, "step": 1839 }, { "epoch": 0.73, "learning_rate": 3.981089672345696e-07, "loss": 0.0439, "step": 1840 }, { "epoch": 0.73, "learning_rate": 3.981054627498062e-07, "loss": 0.0393, "step": 1841 }, { "epoch": 0.73, "learning_rate": 3.981019550362248e-07, "loss": 0.0405, "step": 1842 }, { "epoch": 0.73, "learning_rate": 3.9809844409388266e-07, "loss": 0.0424, "step": 1843 }, { "epoch": 0.73, "learning_rate": 3.980949299228369e-07, "loss": 0.043, "step": 1844 }, { "epoch": 0.73, "learning_rate": 3.9809141252314494e-07, "loss": 0.0492, "step": 1845 }, { "epoch": 0.73, "learning_rate": 3.9808789189486396e-07, "loss": 0.0513, "step": 1846 }, { "epoch": 0.73, "learning_rate": 3.980843680380514e-07, "loss": 0.0419, "step": 1847 }, { "epoch": 0.73, "learning_rate": 3.9808084095276474e-07, "loss": 0.047, "step": 1848 }, { "epoch": 0.73, "learning_rate": 3.980773106390614e-07, "loss": 0.0441, "step": 1849 }, { "epoch": 0.73, "learning_rate": 3.980737770969989e-07, "loss": 0.0416, "step": 1850 }, { "epoch": 0.73, "learning_rate": 3.980702403266349e-07, "loss": 0.0478, "step": 1851 }, { "epoch": 0.73, "learning_rate": 3.9806670032802707e-07, "loss": 0.0399, "step": 1852 }, { "epoch": 0.73, "learning_rate": 3.9806315710123295e-07, "loss": 0.0437, "step": 1853 }, { "epoch": 0.73, "learning_rate": 3.980596106463104e-07, "loss": 0.043, "step": 1854 }, { "epoch": 0.73, "learning_rate": 3.980560609633173e-07, "loss": 0.0417, "step": 1855 }, { "epoch": 0.73, "learning_rate": 3.9805250805231136e-07, "loss": 0.0451, "step": 1856 }, { "epoch": 0.73, "learning_rate": 3.9804895191335055e-07, "loss": 0.0429, "step": 1857 }, { "epoch": 0.73, "learning_rate": 3.9804539254649285e-07, "loss": 0.0449, "step": 1858 }, { "epoch": 0.73, "learning_rate": 3.980418299517962e-07, "loss": 0.0415, "step": 1859 }, { "epoch": 0.73, "learning_rate": 3.9803826412931873e-07, "loss": 0.0441, "step": 1860 }, { "epoch": 0.73, "learning_rate": 3.980346950791185e-07, "loss": 0.0451, "step": 1861 }, { "epoch": 0.73, "learning_rate": 3.980311228012538e-07, "loss": 0.0402, "step": 1862 }, { "epoch": 0.73, "learning_rate": 3.980275472957827e-07, "loss": 0.0416, "step": 1863 }, { "epoch": 0.73, "learning_rate": 3.980239685627636e-07, "loss": 0.0416, "step": 1864 }, { "epoch": 0.74, "learning_rate": 3.980203866022547e-07, "loss": 0.0447, "step": 1865 }, { "epoch": 0.74, "learning_rate": 3.980168014143146e-07, "loss": 0.0427, "step": 1866 }, { "epoch": 0.74, "learning_rate": 3.980132129990014e-07, "loss": 0.0425, "step": 1867 }, { "epoch": 0.74, "learning_rate": 3.980096213563739e-07, "loss": 0.0431, "step": 1868 }, { "epoch": 0.74, "learning_rate": 3.980060264864905e-07, "loss": 0.0431, "step": 1869 }, { "epoch": 0.74, "learning_rate": 3.9800242838940975e-07, "loss": 0.0457, "step": 1870 }, { "epoch": 0.74, "learning_rate": 3.979988270651904e-07, "loss": 0.0403, "step": 1871 }, { "epoch": 0.74, "learning_rate": 3.979952225138911e-07, "loss": 0.0406, "step": 1872 }, { "epoch": 0.74, "learning_rate": 3.9799161473557057e-07, "loss": 0.0493, "step": 1873 }, { "epoch": 0.74, "learning_rate": 3.979880037302876e-07, "loss": 0.0454, "step": 1874 }, { "epoch": 0.74, "learning_rate": 3.9798438949810107e-07, "loss": 0.0459, "step": 1875 }, { "epoch": 0.74, "learning_rate": 3.979807720390699e-07, "loss": 0.0376, "step": 1876 }, { "epoch": 0.74, "learning_rate": 3.979771513532531e-07, "loss": 0.0461, "step": 1877 }, { "epoch": 0.74, "learning_rate": 3.979735274407096e-07, "loss": 0.0517, "step": 1878 }, { "epoch": 0.74, "learning_rate": 3.979699003014985e-07, "loss": 0.0422, "step": 1879 }, { "epoch": 0.74, "learning_rate": 3.9796626993567884e-07, "loss": 0.0449, "step": 1880 }, { "epoch": 0.74, "learning_rate": 3.979626363433099e-07, "loss": 0.049, "step": 1881 }, { "epoch": 0.74, "learning_rate": 3.9795899952445085e-07, "loss": 0.0381, "step": 1882 }, { "epoch": 0.74, "learning_rate": 3.97955359479161e-07, "loss": 0.0364, "step": 1883 }, { "epoch": 0.74, "learning_rate": 3.9795171620749953e-07, "loss": 0.0417, "step": 1884 }, { "epoch": 0.74, "learning_rate": 3.9794806970952607e-07, "loss": 0.0414, "step": 1885 }, { "epoch": 0.74, "learning_rate": 3.9794441998529987e-07, "loss": 0.0478, "step": 1886 }, { "epoch": 0.74, "learning_rate": 3.979407670348805e-07, "loss": 0.041, "step": 1887 }, { "epoch": 0.74, "learning_rate": 3.979371108583274e-07, "loss": 0.0447, "step": 1888 }, { "epoch": 0.74, "learning_rate": 3.9793345145570027e-07, "loss": 0.0496, "step": 1889 }, { "epoch": 0.74, "learning_rate": 3.9792978882705863e-07, "loss": 0.0438, "step": 1890 }, { "epoch": 0.75, "learning_rate": 3.9792612297246236e-07, "loss": 0.0401, "step": 1891 }, { "epoch": 0.75, "learning_rate": 3.9792245389197107e-07, "loss": 0.0435, "step": 1892 }, { "epoch": 0.75, "learning_rate": 3.979187815856446e-07, "loss": 0.0484, "step": 1893 }, { "epoch": 0.75, "learning_rate": 3.979151060535427e-07, "loss": 0.0427, "step": 1894 }, { "epoch": 0.75, "learning_rate": 3.9791142729572544e-07, "loss": 0.0446, "step": 1895 }, { "epoch": 0.75, "learning_rate": 3.979077453122527e-07, "loss": 0.0475, "step": 1896 }, { "epoch": 0.75, "learning_rate": 3.979040601031845e-07, "loss": 0.0432, "step": 1897 }, { "epoch": 0.75, "learning_rate": 3.979003716685809e-07, "loss": 0.0435, "step": 1898 }, { "epoch": 0.75, "learning_rate": 3.97896680008502e-07, "loss": 0.0413, "step": 1899 }, { "epoch": 0.75, "learning_rate": 3.97892985123008e-07, "loss": 0.0444, "step": 1900 }, { "epoch": 0.75, "learning_rate": 3.978892870121591e-07, "loss": 0.0413, "step": 1901 }, { "epoch": 0.75, "learning_rate": 3.978855856760155e-07, "loss": 0.0424, "step": 1902 }, { "epoch": 0.75, "learning_rate": 3.9788188111463776e-07, "loss": 0.0392, "step": 1903 }, { "epoch": 0.75, "learning_rate": 3.97878173328086e-07, "loss": 0.0418, "step": 1904 }, { "epoch": 0.75, "learning_rate": 3.978744623164208e-07, "loss": 0.0454, "step": 1905 }, { "epoch": 0.75, "learning_rate": 3.9787074807970253e-07, "loss": 0.0424, "step": 1906 }, { "epoch": 0.75, "learning_rate": 3.978670306179919e-07, "loss": 0.0443, "step": 1907 }, { "epoch": 0.75, "learning_rate": 3.9786330993134934e-07, "loss": 0.0496, "step": 1908 }, { "epoch": 0.75, "learning_rate": 3.978595860198355e-07, "loss": 0.0481, "step": 1909 }, { "epoch": 0.75, "learning_rate": 3.9785585888351116e-07, "loss": 0.0431, "step": 1910 }, { "epoch": 0.75, "learning_rate": 3.978521285224371e-07, "loss": 0.0449, "step": 1911 }, { "epoch": 0.75, "learning_rate": 3.9784839493667394e-07, "loss": 0.0419, "step": 1912 }, { "epoch": 0.75, "learning_rate": 3.9784465812628264e-07, "loss": 0.035, "step": 1913 }, { "epoch": 0.75, "learning_rate": 3.9784091809132414e-07, "loss": 0.0407, "step": 1914 }, { "epoch": 0.75, "learning_rate": 3.9783717483185934e-07, "loss": 0.0411, "step": 1915 }, { "epoch": 0.76, "learning_rate": 3.9783342834794933e-07, "loss": 0.0504, "step": 1916 }, { "epoch": 0.76, "learning_rate": 3.9782967863965504e-07, "loss": 0.0482, "step": 1917 }, { "epoch": 0.76, "learning_rate": 3.9782592570703767e-07, "loss": 0.0392, "step": 1918 }, { "epoch": 0.76, "learning_rate": 3.9782216955015836e-07, "loss": 0.0413, "step": 1919 }, { "epoch": 0.76, "learning_rate": 3.9781841016907836e-07, "loss": 0.0456, "step": 1920 }, { "epoch": 0.76, "learning_rate": 3.9781464756385894e-07, "loss": 0.0428, "step": 1921 }, { "epoch": 0.76, "learning_rate": 3.978108817345614e-07, "loss": 0.0437, "step": 1922 }, { "epoch": 0.76, "learning_rate": 3.9780711268124715e-07, "loss": 0.047, "step": 1923 }, { "epoch": 0.76, "learning_rate": 3.978033404039775e-07, "loss": 0.0384, "step": 1924 }, { "epoch": 0.76, "learning_rate": 3.977995649028141e-07, "loss": 0.0485, "step": 1925 }, { "epoch": 0.76, "learning_rate": 3.977957861778184e-07, "loss": 0.0399, "step": 1926 }, { "epoch": 0.76, "learning_rate": 3.9779200422905204e-07, "loss": 0.0354, "step": 1927 }, { "epoch": 0.76, "learning_rate": 3.977882190565766e-07, "loss": 0.0401, "step": 1928 }, { "epoch": 0.76, "learning_rate": 3.9778443066045375e-07, "loss": 0.0423, "step": 1929 }, { "epoch": 0.76, "learning_rate": 3.977806390407453e-07, "loss": 0.0412, "step": 1930 }, { "epoch": 0.76, "learning_rate": 3.9777684419751304e-07, "loss": 0.0461, "step": 1931 }, { "epoch": 0.76, "learning_rate": 3.9777304613081877e-07, "loss": 0.0428, "step": 1932 }, { "epoch": 0.76, "learning_rate": 3.9776924484072446e-07, "loss": 0.0386, "step": 1933 }, { "epoch": 0.76, "learning_rate": 3.97765440327292e-07, "loss": 0.0414, "step": 1934 }, { "epoch": 0.76, "learning_rate": 3.977616325905835e-07, "loss": 0.0404, "step": 1935 }, { "epoch": 0.76, "learning_rate": 3.9775782163066094e-07, "loss": 0.0417, "step": 1936 }, { "epoch": 0.76, "learning_rate": 3.977540074475864e-07, "loss": 0.0427, "step": 1937 }, { "epoch": 0.76, "learning_rate": 3.9775019004142205e-07, "loss": 0.0403, "step": 1938 }, { "epoch": 0.76, "learning_rate": 3.977463694122302e-07, "loss": 0.044, "step": 1939 }, { "epoch": 0.76, "learning_rate": 3.977425455600731e-07, "loss": 0.0441, "step": 1940 }, { "epoch": 0.76, "learning_rate": 3.97738718485013e-07, "loss": 0.0377, "step": 1941 }, { "epoch": 0.77, "learning_rate": 3.977348881871124e-07, "loss": 0.0444, "step": 1942 }, { "epoch": 0.77, "learning_rate": 3.977310546664335e-07, "loss": 0.0491, "step": 1943 }, { "epoch": 0.77, "learning_rate": 3.97727217923039e-07, "loss": 0.0461, "step": 1944 }, { "epoch": 0.77, "learning_rate": 3.977233779569914e-07, "loss": 0.047, "step": 1945 }, { "epoch": 0.77, "learning_rate": 3.9771953476835316e-07, "loss": 0.0466, "step": 1946 }, { "epoch": 0.77, "learning_rate": 3.9771568835718704e-07, "loss": 0.0481, "step": 1947 }, { "epoch": 0.77, "learning_rate": 3.977118387235557e-07, "loss": 0.0418, "step": 1948 }, { "epoch": 0.77, "learning_rate": 3.977079858675219e-07, "loss": 0.0454, "step": 1949 }, { "epoch": 0.77, "learning_rate": 3.977041297891484e-07, "loss": 0.0402, "step": 1950 }, { "epoch": 0.77, "learning_rate": 3.9770027048849803e-07, "loss": 0.0409, "step": 1951 }, { "epoch": 0.77, "learning_rate": 3.976964079656337e-07, "loss": 0.044, "step": 1952 }, { "epoch": 0.77, "learning_rate": 3.9769254222061843e-07, "loss": 0.0457, "step": 1953 }, { "epoch": 0.77, "learning_rate": 3.976886732535152e-07, "loss": 0.0449, "step": 1954 }, { "epoch": 0.77, "learning_rate": 3.97684801064387e-07, "loss": 0.0426, "step": 1955 }, { "epoch": 0.77, "learning_rate": 3.97680925653297e-07, "loss": 0.0416, "step": 1956 }, { "epoch": 0.77, "learning_rate": 3.976770470203084e-07, "loss": 0.0484, "step": 1957 }, { "epoch": 0.77, "learning_rate": 3.9767316516548433e-07, "loss": 0.0412, "step": 1958 }, { "epoch": 0.77, "learning_rate": 3.976692800888881e-07, "loss": 0.0485, "step": 1959 }, { "epoch": 0.77, "learning_rate": 3.97665391790583e-07, "loss": 0.0435, "step": 1960 }, { "epoch": 0.77, "learning_rate": 3.9766150027063244e-07, "loss": 0.0418, "step": 1961 }, { "epoch": 0.77, "learning_rate": 3.976576055290999e-07, "loss": 0.0388, "step": 1962 }, { "epoch": 0.77, "learning_rate": 3.976537075660487e-07, "loss": 0.0473, "step": 1963 }, { "epoch": 0.77, "learning_rate": 3.976498063815425e-07, "loss": 0.0445, "step": 1964 }, { "epoch": 0.77, "learning_rate": 3.976459019756449e-07, "loss": 0.0417, "step": 1965 }, { "epoch": 0.77, "learning_rate": 3.9764199434841947e-07, "loss": 0.0477, "step": 1966 }, { "epoch": 0.78, "learning_rate": 3.9763808349992984e-07, "loss": 0.0388, "step": 1967 }, { "epoch": 0.78, "learning_rate": 3.976341694302399e-07, "loss": 0.0438, "step": 1968 }, { "epoch": 0.78, "learning_rate": 3.976302521394133e-07, "loss": 0.0435, "step": 1969 }, { "epoch": 0.78, "learning_rate": 3.97626331627514e-07, "loss": 0.0445, "step": 1970 }, { "epoch": 0.78, "learning_rate": 3.9762240789460584e-07, "loss": 0.0422, "step": 1971 }, { "epoch": 0.78, "learning_rate": 3.976184809407528e-07, "loss": 0.0467, "step": 1972 }, { "epoch": 0.78, "learning_rate": 3.976145507660188e-07, "loss": 0.0422, "step": 1973 }, { "epoch": 0.78, "learning_rate": 3.9761061737046806e-07, "loss": 0.0438, "step": 1974 }, { "epoch": 0.78, "learning_rate": 3.976066807541645e-07, "loss": 0.043, "step": 1975 }, { "epoch": 0.78, "learning_rate": 3.976027409171725e-07, "loss": 0.0478, "step": 1976 }, { "epoch": 0.78, "learning_rate": 3.9759879785955606e-07, "loss": 0.043, "step": 1977 }, { "epoch": 0.78, "learning_rate": 3.975948515813795e-07, "loss": 0.0419, "step": 1978 }, { "epoch": 0.78, "learning_rate": 3.975909020827072e-07, "loss": 0.0386, "step": 1979 }, { "epoch": 0.78, "learning_rate": 3.9758694936360356e-07, "loss": 0.0414, "step": 1980 }, { "epoch": 0.78, "learning_rate": 3.9758299342413284e-07, "loss": 0.0434, "step": 1981 }, { "epoch": 0.78, "learning_rate": 3.975790342643597e-07, "loss": 0.0434, "step": 1982 }, { "epoch": 0.78, "learning_rate": 3.9757507188434846e-07, "loss": 0.0488, "step": 1983 }, { "epoch": 0.78, "learning_rate": 3.9757110628416396e-07, "loss": 0.045, "step": 1984 }, { "epoch": 0.78, "learning_rate": 3.975671374638707e-07, "loss": 0.0418, "step": 1985 }, { "epoch": 0.78, "learning_rate": 3.9756316542353327e-07, "loss": 0.0438, "step": 1986 }, { "epoch": 0.78, "learning_rate": 3.9755919016321657e-07, "loss": 0.0386, "step": 1987 }, { "epoch": 0.78, "learning_rate": 3.9755521168298534e-07, "loss": 0.0452, "step": 1988 }, { "epoch": 0.78, "learning_rate": 3.9755122998290436e-07, "loss": 0.0455, "step": 1989 }, { "epoch": 0.78, "learning_rate": 3.975472450630386e-07, "loss": 0.0498, "step": 1990 }, { "epoch": 0.78, "learning_rate": 3.97543256923453e-07, "loss": 0.0497, "step": 1991 }, { "epoch": 0.79, "learning_rate": 3.975392655642125e-07, "loss": 0.0424, "step": 1992 }, { "epoch": 0.79, "learning_rate": 3.975352709853822e-07, "loss": 0.0458, "step": 1993 }, { "epoch": 0.79, "learning_rate": 3.9753127318702726e-07, "loss": 0.048, "step": 1994 }, { "epoch": 0.79, "learning_rate": 3.9752727216921277e-07, "loss": 0.0399, "step": 1995 }, { "epoch": 0.79, "learning_rate": 3.9752326793200387e-07, "loss": 0.0403, "step": 1996 }, { "epoch": 0.79, "learning_rate": 3.97519260475466e-07, "loss": 0.0433, "step": 1997 }, { "epoch": 0.79, "learning_rate": 3.975152497996643e-07, "loss": 0.0392, "step": 1998 }, { "epoch": 0.79, "learning_rate": 3.9751123590466427e-07, "loss": 0.0403, "step": 1999 }, { "epoch": 0.79, "learning_rate": 3.9750721879053123e-07, "loss": 0.0443, "step": 2000 }, { "epoch": 0.79, "learning_rate": 3.975031984573308e-07, "loss": 0.0389, "step": 2001 }, { "epoch": 0.79, "learning_rate": 3.974991749051283e-07, "loss": 0.04, "step": 2002 }, { "epoch": 0.79, "learning_rate": 3.974951481339894e-07, "loss": 0.0463, "step": 2003 }, { "epoch": 0.79, "learning_rate": 3.9749111814397985e-07, "loss": 0.0461, "step": 2004 }, { "epoch": 0.79, "learning_rate": 3.9748708493516514e-07, "loss": 0.0355, "step": 2005 }, { "epoch": 0.79, "learning_rate": 3.9748304850761115e-07, "loss": 0.0419, "step": 2006 }, { "epoch": 0.79, "learning_rate": 3.974790088613836e-07, "loss": 0.0388, "step": 2007 }, { "epoch": 0.79, "learning_rate": 3.9747496599654833e-07, "loss": 0.0444, "step": 2008 }, { "epoch": 0.79, "learning_rate": 3.9747091991317116e-07, "loss": 0.0408, "step": 2009 }, { "epoch": 0.79, "learning_rate": 3.9746687061131824e-07, "loss": 0.0418, "step": 2010 }, { "epoch": 0.79, "learning_rate": 3.9746281809105537e-07, "loss": 0.0458, "step": 2011 }, { "epoch": 0.79, "learning_rate": 3.974587623524487e-07, "loss": 0.0453, "step": 2012 }, { "epoch": 0.79, "learning_rate": 3.974547033955643e-07, "loss": 0.0396, "step": 2013 }, { "epoch": 0.79, "learning_rate": 3.9745064122046833e-07, "loss": 0.0443, "step": 2014 }, { "epoch": 0.79, "learning_rate": 3.97446575827227e-07, "loss": 0.0438, "step": 2015 }, { "epoch": 0.79, "learning_rate": 3.974425072159065e-07, "loss": 0.0471, "step": 2016 }, { "epoch": 0.79, "learning_rate": 3.974384353865734e-07, "loss": 0.0406, "step": 2017 }, { "epoch": 0.8, "learning_rate": 3.974343603392937e-07, "loss": 0.038, "step": 2018 }, { "epoch": 0.8, "learning_rate": 3.97430282074134e-07, "loss": 0.0456, "step": 2019 }, { "epoch": 0.8, "learning_rate": 3.974262005911608e-07, "loss": 0.0447, "step": 2020 }, { "epoch": 0.8, "learning_rate": 3.9742211589044065e-07, "loss": 0.0415, "step": 2021 }, { "epoch": 0.8, "learning_rate": 3.9741802797204e-07, "loss": 0.0425, "step": 2022 }, { "epoch": 0.8, "learning_rate": 3.974139368360255e-07, "loss": 0.0438, "step": 2023 }, { "epoch": 0.8, "learning_rate": 3.974098424824639e-07, "loss": 0.0439, "step": 2024 }, { "epoch": 0.8, "learning_rate": 3.9740574491142193e-07, "loss": 0.0401, "step": 2025 }, { "epoch": 0.8, "learning_rate": 3.974016441229663e-07, "loss": 0.0429, "step": 2026 }, { "epoch": 0.8, "learning_rate": 3.973975401171639e-07, "loss": 0.0458, "step": 2027 }, { "epoch": 0.8, "learning_rate": 3.9739343289408155e-07, "loss": 0.0399, "step": 2028 }, { "epoch": 0.8, "learning_rate": 3.973893224537863e-07, "loss": 0.0423, "step": 2029 }, { "epoch": 0.8, "learning_rate": 3.9738520879634514e-07, "loss": 0.0408, "step": 2030 }, { "epoch": 0.8, "learning_rate": 3.97381091921825e-07, "loss": 0.0396, "step": 2031 }, { "epoch": 0.8, "learning_rate": 3.973769718302931e-07, "loss": 0.0405, "step": 2032 }, { "epoch": 0.8, "learning_rate": 3.9737284852181647e-07, "loss": 0.042, "step": 2033 }, { "epoch": 0.8, "learning_rate": 3.973687219964624e-07, "loss": 0.0422, "step": 2034 }, { "epoch": 0.8, "learning_rate": 3.9736459225429807e-07, "loss": 0.0421, "step": 2035 }, { "epoch": 0.8, "learning_rate": 3.9736045929539096e-07, "loss": 0.0415, "step": 2036 }, { "epoch": 0.8, "learning_rate": 3.9735632311980827e-07, "loss": 0.0464, "step": 2037 }, { "epoch": 0.8, "learning_rate": 3.973521837276174e-07, "loss": 0.0419, "step": 2038 }, { "epoch": 0.8, "learning_rate": 3.97348041118886e-07, "loss": 0.0355, "step": 2039 }, { "epoch": 0.8, "learning_rate": 3.973438952936814e-07, "loss": 0.04, "step": 2040 }, { "epoch": 0.8, "learning_rate": 3.9733974625207125e-07, "loss": 0.0423, "step": 2041 }, { "epoch": 0.8, "learning_rate": 3.973355939941231e-07, "loss": 0.0366, "step": 2042 }, { "epoch": 0.81, "learning_rate": 3.9733143851990475e-07, "loss": 0.0436, "step": 2043 }, { "epoch": 0.81, "learning_rate": 3.973272798294838e-07, "loss": 0.041, "step": 2044 }, { "epoch": 0.81, "learning_rate": 3.973231179229281e-07, "loss": 0.0419, "step": 2045 }, { "epoch": 0.81, "learning_rate": 3.973189528003056e-07, "loss": 0.0414, "step": 2046 }, { "epoch": 0.81, "learning_rate": 3.9731478446168393e-07, "loss": 0.0433, "step": 2047 }, { "epoch": 0.81, "learning_rate": 3.9731061290713115e-07, "loss": 0.042, "step": 2048 }, { "epoch": 0.81, "learning_rate": 3.973064381367153e-07, "loss": 0.0418, "step": 2049 }, { "epoch": 0.81, "learning_rate": 3.973022601505044e-07, "loss": 0.051, "step": 2050 }, { "epoch": 0.81, "learning_rate": 3.972980789485665e-07, "loss": 0.0446, "step": 2051 }, { "epoch": 0.81, "learning_rate": 3.9729389453096977e-07, "loss": 0.0394, "step": 2052 }, { "epoch": 0.81, "learning_rate": 3.972897068977824e-07, "loss": 0.0404, "step": 2053 }, { "epoch": 0.81, "learning_rate": 3.9728551604907266e-07, "loss": 0.0477, "step": 2054 }, { "epoch": 0.81, "learning_rate": 3.9728132198490884e-07, "loss": 0.0413, "step": 2055 }, { "epoch": 0.81, "learning_rate": 3.9727712470535935e-07, "loss": 0.047, "step": 2056 }, { "epoch": 0.81, "learning_rate": 3.9727292421049253e-07, "loss": 0.0389, "step": 2057 }, { "epoch": 0.81, "learning_rate": 3.9726872050037685e-07, "loss": 0.0403, "step": 2058 }, { "epoch": 0.81, "learning_rate": 3.9726451357508085e-07, "loss": 0.0468, "step": 2059 }, { "epoch": 0.81, "learning_rate": 3.972603034346731e-07, "loss": 0.0431, "step": 2060 }, { "epoch": 0.81, "learning_rate": 3.9725609007922215e-07, "loss": 0.0415, "step": 2061 }, { "epoch": 0.81, "learning_rate": 3.972518735087968e-07, "loss": 0.0431, "step": 2062 }, { "epoch": 0.81, "learning_rate": 3.972476537234657e-07, "loss": 0.0437, "step": 2063 }, { "epoch": 0.81, "learning_rate": 3.9724343072329756e-07, "loss": 0.0405, "step": 2064 }, { "epoch": 0.81, "learning_rate": 3.9723920450836133e-07, "loss": 0.0409, "step": 2065 }, { "epoch": 0.81, "learning_rate": 3.972349750787258e-07, "loss": 0.037, "step": 2066 }, { "epoch": 0.81, "learning_rate": 3.9723074243445994e-07, "loss": 0.0443, "step": 2067 }, { "epoch": 0.82, "learning_rate": 3.9722650657563273e-07, "loss": 0.0414, "step": 2068 }, { "epoch": 0.82, "learning_rate": 3.9722226750231327e-07, "loss": 0.0399, "step": 2069 }, { "epoch": 0.82, "learning_rate": 3.9721802521457053e-07, "loss": 0.0448, "step": 2070 }, { "epoch": 0.82, "learning_rate": 3.9721377971247376e-07, "loss": 0.037, "step": 2071 }, { "epoch": 0.82, "learning_rate": 3.9720953099609204e-07, "loss": 0.0415, "step": 2072 }, { "epoch": 0.82, "learning_rate": 3.972052790654948e-07, "loss": 0.0426, "step": 2073 }, { "epoch": 0.82, "learning_rate": 3.9720102392075117e-07, "loss": 0.0411, "step": 2074 }, { "epoch": 0.82, "learning_rate": 3.9719676556193055e-07, "loss": 0.0419, "step": 2075 }, { "epoch": 0.82, "learning_rate": 3.9719250398910234e-07, "loss": 0.0398, "step": 2076 }, { "epoch": 0.82, "learning_rate": 3.9718823920233604e-07, "loss": 0.0403, "step": 2077 }, { "epoch": 0.82, "learning_rate": 3.9718397120170117e-07, "loss": 0.0427, "step": 2078 }, { "epoch": 0.82, "learning_rate": 3.9717969998726717e-07, "loss": 0.0479, "step": 2079 }, { "epoch": 0.82, "learning_rate": 3.971754255591038e-07, "loss": 0.0448, "step": 2080 }, { "epoch": 0.82, "learning_rate": 3.971711479172806e-07, "loss": 0.0453, "step": 2081 }, { "epoch": 0.82, "learning_rate": 3.9716686706186746e-07, "loss": 0.0406, "step": 2082 }, { "epoch": 0.82, "learning_rate": 3.97162582992934e-07, "loss": 0.0485, "step": 2083 }, { "epoch": 0.82, "learning_rate": 3.9715829571055e-07, "loss": 0.0524, "step": 2084 }, { "epoch": 0.82, "learning_rate": 3.971540052147856e-07, "loss": 0.0429, "step": 2085 }, { "epoch": 0.82, "learning_rate": 3.9714971150571037e-07, "loss": 0.0456, "step": 2086 }, { "epoch": 0.82, "learning_rate": 3.971454145833946e-07, "loss": 0.0399, "step": 2087 }, { "epoch": 0.82, "learning_rate": 3.9714111444790815e-07, "loss": 0.0406, "step": 2088 }, { "epoch": 0.82, "learning_rate": 3.9713681109932117e-07, "loss": 0.0415, "step": 2089 }, { "epoch": 0.82, "learning_rate": 3.971325045377038e-07, "loss": 0.0459, "step": 2090 }, { "epoch": 0.82, "learning_rate": 3.9712819476312615e-07, "loss": 0.0448, "step": 2091 }, { "epoch": 0.82, "learning_rate": 3.971238817756586e-07, "loss": 0.0404, "step": 2092 }, { "epoch": 0.82, "learning_rate": 3.9711956557537133e-07, "loss": 0.042, "step": 2093 }, { "epoch": 0.83, "learning_rate": 3.971152461623347e-07, "loss": 0.0406, "step": 2094 }, { "epoch": 0.83, "learning_rate": 3.9711092353661925e-07, "loss": 0.0433, "step": 2095 }, { "epoch": 0.83, "learning_rate": 3.9710659769829526e-07, "loss": 0.0392, "step": 2096 }, { "epoch": 0.83, "learning_rate": 3.9710226864743323e-07, "loss": 0.0436, "step": 2097 }, { "epoch": 0.83, "learning_rate": 3.970979363841039e-07, "loss": 0.0406, "step": 2098 }, { "epoch": 0.83, "learning_rate": 3.970936009083776e-07, "loss": 0.0415, "step": 2099 }, { "epoch": 0.83, "learning_rate": 3.970892622203253e-07, "loss": 0.045, "step": 2100 }, { "epoch": 0.83, "learning_rate": 3.970849203200175e-07, "loss": 0.0425, "step": 2101 }, { "epoch": 0.83, "learning_rate": 3.97080575207525e-07, "loss": 0.0406, "step": 2102 }, { "epoch": 0.83, "learning_rate": 3.970762268829188e-07, "loss": 0.0424, "step": 2103 }, { "epoch": 0.83, "learning_rate": 3.9707187534626947e-07, "loss": 0.0466, "step": 2104 }, { "epoch": 0.83, "learning_rate": 3.9706752059764813e-07, "loss": 0.0416, "step": 2105 }, { "epoch": 0.83, "learning_rate": 3.9706316263712576e-07, "loss": 0.0456, "step": 2106 }, { "epoch": 0.83, "learning_rate": 3.970588014647733e-07, "loss": 0.0404, "step": 2107 }, { "epoch": 0.83, "learning_rate": 3.9705443708066186e-07, "loss": 0.0416, "step": 2108 }, { "epoch": 0.83, "learning_rate": 3.970500694848626e-07, "loss": 0.0414, "step": 2109 }, { "epoch": 0.83, "learning_rate": 3.9704569867744666e-07, "loss": 0.0482, "step": 2110 }, { "epoch": 0.83, "learning_rate": 3.970413246584853e-07, "loss": 0.0411, "step": 2111 }, { "epoch": 0.83, "learning_rate": 3.970369474280499e-07, "loss": 0.0408, "step": 2112 }, { "epoch": 0.83, "learning_rate": 3.970325669862117e-07, "loss": 0.0393, "step": 2113 }, { "epoch": 0.83, "learning_rate": 3.9702818333304205e-07, "loss": 0.0387, "step": 2114 }, { "epoch": 0.83, "learning_rate": 3.970237964686125e-07, "loss": 0.0412, "step": 2115 }, { "epoch": 0.83, "learning_rate": 3.970194063929945e-07, "loss": 0.0364, "step": 2116 }, { "epoch": 0.83, "learning_rate": 3.9701501310625965e-07, "loss": 0.0388, "step": 2117 }, { "epoch": 0.83, "learning_rate": 3.970106166084794e-07, "loss": 0.0417, "step": 2118 }, { "epoch": 0.84, "learning_rate": 3.9700621689972567e-07, "loss": 0.0363, "step": 2119 }, { "epoch": 0.84, "learning_rate": 3.9700181398006993e-07, "loss": 0.0404, "step": 2120 }, { "epoch": 0.84, "learning_rate": 3.9699740784958404e-07, "loss": 0.0418, "step": 2121 }, { "epoch": 0.84, "learning_rate": 3.969929985083398e-07, "loss": 0.0391, "step": 2122 }, { "epoch": 0.84, "learning_rate": 3.969885859564091e-07, "loss": 0.039, "step": 2123 }, { "epoch": 0.84, "learning_rate": 3.9698417019386386e-07, "loss": 0.04, "step": 2124 }, { "epoch": 0.84, "learning_rate": 3.9697975122077596e-07, "loss": 0.0422, "step": 2125 }, { "epoch": 0.84, "learning_rate": 3.969753290372175e-07, "loss": 0.0383, "step": 2126 }, { "epoch": 0.84, "learning_rate": 3.969709036432606e-07, "loss": 0.0426, "step": 2127 }, { "epoch": 0.84, "learning_rate": 3.969664750389773e-07, "loss": 0.0444, "step": 2128 }, { "epoch": 0.84, "learning_rate": 3.9696204322443983e-07, "loss": 0.0423, "step": 2129 }, { "epoch": 0.84, "learning_rate": 3.9695760819972034e-07, "loss": 0.0371, "step": 2130 }, { "epoch": 0.84, "learning_rate": 3.9695316996489125e-07, "loss": 0.0426, "step": 2131 }, { "epoch": 0.84, "learning_rate": 3.969487285200248e-07, "loss": 0.0431, "step": 2132 }, { "epoch": 0.84, "learning_rate": 3.9694428386519337e-07, "loss": 0.0419, "step": 2133 }, { "epoch": 0.84, "learning_rate": 3.9693983600046946e-07, "loss": 0.0366, "step": 2134 }, { "epoch": 0.84, "learning_rate": 3.9693538492592553e-07, "loss": 0.0465, "step": 2135 }, { "epoch": 0.84, "learning_rate": 3.9693093064163416e-07, "loss": 0.0433, "step": 2136 }, { "epoch": 0.84, "learning_rate": 3.9692647314766787e-07, "loss": 0.0371, "step": 2137 }, { "epoch": 0.84, "learning_rate": 3.969220124440994e-07, "loss": 0.041, "step": 2138 }, { "epoch": 0.84, "learning_rate": 3.969175485310014e-07, "loss": 0.0404, "step": 2139 }, { "epoch": 0.84, "learning_rate": 3.9691308140844673e-07, "loss": 0.0341, "step": 2140 }, { "epoch": 0.84, "learning_rate": 3.96908611076508e-07, "loss": 0.0417, "step": 2141 }, { "epoch": 0.84, "learning_rate": 3.969041375352582e-07, "loss": 0.0416, "step": 2142 }, { "epoch": 0.84, "learning_rate": 3.9689966078477025e-07, "loss": 0.0399, "step": 2143 }, { "epoch": 0.84, "learning_rate": 3.9689518082511703e-07, "loss": 0.0425, "step": 2144 }, { "epoch": 0.85, "learning_rate": 3.9689069765637164e-07, "loss": 0.0386, "step": 2145 }, { "epoch": 0.85, "learning_rate": 3.9688621127860705e-07, "loss": 0.0363, "step": 2146 }, { "epoch": 0.85, "learning_rate": 3.968817216918965e-07, "loss": 0.0425, "step": 2147 }, { "epoch": 0.85, "learning_rate": 3.968772288963132e-07, "loss": 0.0358, "step": 2148 }, { "epoch": 0.85, "learning_rate": 3.968727328919302e-07, "loss": 0.0377, "step": 2149 }, { "epoch": 0.85, "learning_rate": 3.968682336788208e-07, "loss": 0.043, "step": 2150 }, { "epoch": 0.85, "learning_rate": 3.968637312570585e-07, "loss": 0.043, "step": 2151 }, { "epoch": 0.85, "learning_rate": 3.9685922562671657e-07, "loss": 0.0383, "step": 2152 }, { "epoch": 0.85, "learning_rate": 3.968547167878684e-07, "loss": 0.0422, "step": 2153 }, { "epoch": 0.85, "learning_rate": 3.9685020474058754e-07, "loss": 0.0456, "step": 2154 }, { "epoch": 0.85, "learning_rate": 3.968456894849476e-07, "loss": 0.045, "step": 2155 }, { "epoch": 0.85, "learning_rate": 3.96841171021022e-07, "loss": 0.0409, "step": 2156 }, { "epoch": 0.85, "learning_rate": 3.968366493488845e-07, "loss": 0.0419, "step": 2157 }, { "epoch": 0.85, "learning_rate": 3.968321244686088e-07, "loss": 0.0385, "step": 2158 }, { "epoch": 0.85, "learning_rate": 3.968275963802686e-07, "loss": 0.0407, "step": 2159 }, { "epoch": 0.85, "learning_rate": 3.9682306508393777e-07, "loss": 0.0434, "step": 2160 }, { "epoch": 0.85, "learning_rate": 3.9681853057969007e-07, "loss": 0.0478, "step": 2161 }, { "epoch": 0.85, "learning_rate": 3.9681399286759943e-07, "loss": 0.0377, "step": 2162 }, { "epoch": 0.85, "learning_rate": 3.9680945194773986e-07, "loss": 0.0401, "step": 2163 }, { "epoch": 0.85, "learning_rate": 3.9680490782018536e-07, "loss": 0.0432, "step": 2164 }, { "epoch": 0.85, "learning_rate": 3.968003604850099e-07, "loss": 0.0387, "step": 2165 }, { "epoch": 0.85, "learning_rate": 3.9679580994228773e-07, "loss": 0.0383, "step": 2166 }, { "epoch": 0.85, "learning_rate": 3.9679125619209294e-07, "loss": 0.0428, "step": 2167 }, { "epoch": 0.85, "learning_rate": 3.967866992344998e-07, "loss": 0.0415, "step": 2168 }, { "epoch": 0.85, "learning_rate": 3.9678213906958245e-07, "loss": 0.0443, "step": 2169 }, { "epoch": 0.86, "learning_rate": 3.967775756974154e-07, "loss": 0.0388, "step": 2170 }, { "epoch": 0.86, "learning_rate": 3.967730091180729e-07, "loss": 0.0425, "step": 2171 }, { "epoch": 0.86, "learning_rate": 3.9676843933162936e-07, "loss": 0.0375, "step": 2172 }, { "epoch": 0.86, "learning_rate": 3.967638663381594e-07, "loss": 0.0445, "step": 2173 }, { "epoch": 0.86, "learning_rate": 3.9675929013773744e-07, "loss": 0.0399, "step": 2174 }, { "epoch": 0.86, "learning_rate": 3.96754710730438e-07, "loss": 0.0425, "step": 2175 }, { "epoch": 0.86, "learning_rate": 3.967501281163359e-07, "loss": 0.0471, "step": 2176 }, { "epoch": 0.86, "learning_rate": 3.9674554229550574e-07, "loss": 0.0376, "step": 2177 }, { "epoch": 0.86, "learning_rate": 3.9674095326802225e-07, "loss": 0.0407, "step": 2178 }, { "epoch": 0.86, "learning_rate": 3.9673636103396017e-07, "loss": 0.0451, "step": 2179 }, { "epoch": 0.86, "learning_rate": 3.9673176559339447e-07, "loss": 0.0362, "step": 2180 }, { "epoch": 0.86, "learning_rate": 3.967271669464e-07, "loss": 0.0406, "step": 2181 }, { "epoch": 0.86, "learning_rate": 3.9672256509305166e-07, "loss": 0.0471, "step": 2182 }, { "epoch": 0.86, "learning_rate": 3.967179600334245e-07, "loss": 0.0403, "step": 2183 }, { "epoch": 0.86, "learning_rate": 3.9671335176759363e-07, "loss": 0.042, "step": 2184 }, { "epoch": 0.86, "learning_rate": 3.96708740295634e-07, "loss": 0.0434, "step": 2185 }, { "epoch": 0.86, "learning_rate": 3.967041256176209e-07, "loss": 0.0398, "step": 2186 }, { "epoch": 0.86, "learning_rate": 3.9669950773362954e-07, "loss": 0.0411, "step": 2187 }, { "epoch": 0.86, "learning_rate": 3.966948866437351e-07, "loss": 0.0463, "step": 2188 }, { "epoch": 0.86, "learning_rate": 3.96690262348013e-07, "loss": 0.0383, "step": 2189 }, { "epoch": 0.86, "learning_rate": 3.966856348465385e-07, "loss": 0.0407, "step": 2190 }, { "epoch": 0.86, "learning_rate": 3.966810041393872e-07, "loss": 0.0459, "step": 2191 }, { "epoch": 0.86, "learning_rate": 3.9667637022663435e-07, "loss": 0.0401, "step": 2192 }, { "epoch": 0.86, "learning_rate": 3.966717331083556e-07, "loss": 0.0431, "step": 2193 }, { "epoch": 0.86, "learning_rate": 3.9666709278462653e-07, "loss": 0.0377, "step": 2194 }, { "epoch": 0.87, "learning_rate": 3.966624492555227e-07, "loss": 0.0415, "step": 2195 }, { "epoch": 0.87, "learning_rate": 3.9665780252111984e-07, "loss": 0.0392, "step": 2196 }, { "epoch": 0.87, "learning_rate": 3.966531525814937e-07, "loss": 0.0382, "step": 2197 }, { "epoch": 0.87, "learning_rate": 3.9664849943672013e-07, "loss": 0.0392, "step": 2198 }, { "epoch": 0.87, "learning_rate": 3.9664384308687477e-07, "loss": 0.039, "step": 2199 }, { "epoch": 0.87, "learning_rate": 3.966391835320337e-07, "loss": 0.0401, "step": 2200 }, { "epoch": 0.87, "learning_rate": 3.966345207722728e-07, "loss": 0.0389, "step": 2201 }, { "epoch": 0.87, "learning_rate": 3.96629854807668e-07, "loss": 0.0419, "step": 2202 }, { "epoch": 0.87, "learning_rate": 3.9662518563829547e-07, "loss": 0.0383, "step": 2203 }, { "epoch": 0.87, "learning_rate": 3.9662051326423126e-07, "loss": 0.0399, "step": 2204 }, { "epoch": 0.87, "learning_rate": 3.9661583768555146e-07, "loss": 0.0436, "step": 2205 }, { "epoch": 0.87, "learning_rate": 3.966111589023324e-07, "loss": 0.0423, "step": 2206 }, { "epoch": 0.87, "learning_rate": 3.966064769146502e-07, "loss": 0.0472, "step": 2207 }, { "epoch": 0.87, "learning_rate": 3.9660179172258123e-07, "loss": 0.0349, "step": 2208 }, { "epoch": 0.87, "learning_rate": 3.9659710332620183e-07, "loss": 0.0414, "step": 2209 }, { "epoch": 0.87, "learning_rate": 3.9659241172558853e-07, "loss": 0.0405, "step": 2210 }, { "epoch": 0.87, "learning_rate": 3.965877169208177e-07, "loss": 0.045, "step": 2211 }, { "epoch": 0.87, "learning_rate": 3.965830189119658e-07, "loss": 0.0364, "step": 2212 }, { "epoch": 0.87, "learning_rate": 3.9657831769910945e-07, "loss": 0.0388, "step": 2213 }, { "epoch": 0.87, "learning_rate": 3.9657361328232533e-07, "loss": 0.0394, "step": 2214 }, { "epoch": 0.87, "learning_rate": 3.965689056616901e-07, "loss": 0.0417, "step": 2215 }, { "epoch": 0.87, "learning_rate": 3.965641948372804e-07, "loss": 0.0393, "step": 2216 }, { "epoch": 0.87, "learning_rate": 3.9655948080917303e-07, "loss": 0.0374, "step": 2217 }, { "epoch": 0.87, "learning_rate": 3.9655476357744497e-07, "loss": 0.0396, "step": 2218 }, { "epoch": 0.87, "learning_rate": 3.965500431421729e-07, "loss": 0.0399, "step": 2219 }, { "epoch": 0.87, "learning_rate": 3.965453195034339e-07, "loss": 0.0457, "step": 2220 }, { "epoch": 0.88, "learning_rate": 3.965405926613049e-07, "loss": 0.0408, "step": 2221 }, { "epoch": 0.88, "learning_rate": 3.9653586261586294e-07, "loss": 0.0426, "step": 2222 }, { "epoch": 0.88, "learning_rate": 3.965311293671851e-07, "loss": 0.0425, "step": 2223 }, { "epoch": 0.88, "learning_rate": 3.965263929153486e-07, "loss": 0.0415, "step": 2224 }, { "epoch": 0.88, "learning_rate": 3.965216532604305e-07, "loss": 0.0425, "step": 2225 }, { "epoch": 0.88, "learning_rate": 3.965169104025082e-07, "loss": 0.0375, "step": 2226 }, { "epoch": 0.88, "learning_rate": 3.965121643416589e-07, "loss": 0.038, "step": 2227 }, { "epoch": 0.88, "learning_rate": 3.9650741507795995e-07, "loss": 0.0414, "step": 2228 }, { "epoch": 0.88, "learning_rate": 3.9650266261148887e-07, "loss": 0.045, "step": 2229 }, { "epoch": 0.88, "learning_rate": 3.96497906942323e-07, "loss": 0.0393, "step": 2230 }, { "epoch": 0.88, "learning_rate": 3.964931480705399e-07, "loss": 0.0359, "step": 2231 }, { "epoch": 0.88, "learning_rate": 3.964883859962171e-07, "loss": 0.0438, "step": 2232 }, { "epoch": 0.88, "learning_rate": 3.9648362071943227e-07, "loss": 0.0466, "step": 2233 }, { "epoch": 0.88, "learning_rate": 3.9647885224026304e-07, "loss": 0.0388, "step": 2234 }, { "epoch": 0.88, "learning_rate": 3.964740805587871e-07, "loss": 0.039, "step": 2235 }, { "epoch": 0.88, "learning_rate": 3.9646930567508226e-07, "loss": 0.0414, "step": 2236 }, { "epoch": 0.88, "learning_rate": 3.9646452758922635e-07, "loss": 0.0442, "step": 2237 }, { "epoch": 0.88, "learning_rate": 3.964597463012972e-07, "loss": 0.0441, "step": 2238 }, { "epoch": 0.88, "learning_rate": 3.964549618113728e-07, "loss": 0.0366, "step": 2239 }, { "epoch": 0.88, "learning_rate": 3.9645017411953115e-07, "loss": 0.0405, "step": 2240 }, { "epoch": 0.88, "learning_rate": 3.964453832258502e-07, "loss": 0.0378, "step": 2241 }, { "epoch": 0.88, "learning_rate": 3.9644058913040796e-07, "loss": 0.0441, "step": 2242 }, { "epoch": 0.88, "learning_rate": 3.964357918332828e-07, "loss": 0.0441, "step": 2243 }, { "epoch": 0.88, "learning_rate": 3.964309913345527e-07, "loss": 0.0443, "step": 2244 }, { "epoch": 0.88, "learning_rate": 3.96426187634296e-07, "loss": 0.0397, "step": 2245 }, { "epoch": 0.89, "learning_rate": 3.96421380732591e-07, "loss": 0.0435, "step": 2246 }, { "epoch": 0.89, "learning_rate": 3.9641657062951605e-07, "loss": 0.0389, "step": 2247 }, { "epoch": 0.89, "learning_rate": 3.964117573251495e-07, "loss": 0.0419, "step": 2248 }, { "epoch": 0.89, "learning_rate": 3.9640694081956976e-07, "loss": 0.0432, "step": 2249 }, { "epoch": 0.89, "learning_rate": 3.9640212111285543e-07, "loss": 0.0388, "step": 2250 }, { "epoch": 0.89, "learning_rate": 3.96397298205085e-07, "loss": 0.0409, "step": 2251 }, { "epoch": 0.89, "learning_rate": 3.9639247209633717e-07, "loss": 0.0425, "step": 2252 }, { "epoch": 0.89, "learning_rate": 3.963876427866904e-07, "loss": 0.0395, "step": 2253 }, { "epoch": 0.89, "learning_rate": 3.963828102762236e-07, "loss": 0.0395, "step": 2254 }, { "epoch": 0.89, "learning_rate": 3.963779745650155e-07, "loss": 0.0377, "step": 2255 }, { "epoch": 0.89, "learning_rate": 3.9637313565314487e-07, "loss": 0.0426, "step": 2256 }, { "epoch": 0.89, "learning_rate": 3.963682935406905e-07, "loss": 0.0353, "step": 2257 }, { "epoch": 0.89, "learning_rate": 3.963634482277315e-07, "loss": 0.0417, "step": 2258 }, { "epoch": 0.89, "learning_rate": 3.9635859971434665e-07, "loss": 0.0432, "step": 2259 }, { "epoch": 0.89, "learning_rate": 3.9635374800061507e-07, "loss": 0.0428, "step": 2260 }, { "epoch": 0.89, "learning_rate": 3.963488930866158e-07, "loss": 0.0385, "step": 2261 }, { "epoch": 0.89, "learning_rate": 3.9634403497242806e-07, "loss": 0.0379, "step": 2262 }, { "epoch": 0.89, "learning_rate": 3.963391736581309e-07, "loss": 0.0375, "step": 2263 }, { "epoch": 0.89, "learning_rate": 3.9633430914380366e-07, "loss": 0.037, "step": 2264 }, { "epoch": 0.89, "learning_rate": 3.963294414295255e-07, "loss": 0.0397, "step": 2265 }, { "epoch": 0.89, "learning_rate": 3.9632457051537587e-07, "loss": 0.0392, "step": 2266 }, { "epoch": 0.89, "learning_rate": 3.963196964014341e-07, "loss": 0.0412, "step": 2267 }, { "epoch": 0.89, "learning_rate": 3.963148190877797e-07, "loss": 0.0435, "step": 2268 }, { "epoch": 0.89, "learning_rate": 3.9630993857449207e-07, "loss": 0.0402, "step": 2269 }, { "epoch": 0.89, "learning_rate": 3.9630505486165087e-07, "loss": 0.0431, "step": 2270 }, { "epoch": 0.9, "learning_rate": 3.963001679493355e-07, "loss": 0.0408, "step": 2271 }, { "epoch": 0.9, "learning_rate": 3.962952778376258e-07, "loss": 0.0473, "step": 2272 }, { "epoch": 0.9, "learning_rate": 3.962903845266014e-07, "loss": 0.0407, "step": 2273 }, { "epoch": 0.9, "learning_rate": 3.962854880163421e-07, "loss": 0.0358, "step": 2274 }, { "epoch": 0.9, "learning_rate": 3.9628058830692764e-07, "loss": 0.044, "step": 2275 }, { "epoch": 0.9, "learning_rate": 3.9627568539843786e-07, "loss": 0.0376, "step": 2276 }, { "epoch": 0.9, "learning_rate": 3.9627077929095275e-07, "loss": 0.0455, "step": 2277 }, { "epoch": 0.9, "learning_rate": 3.962658699845522e-07, "loss": 0.0375, "step": 2278 }, { "epoch": 0.9, "learning_rate": 3.9626095747931636e-07, "loss": 0.0427, "step": 2279 }, { "epoch": 0.9, "learning_rate": 3.962560417753251e-07, "loss": 0.0376, "step": 2280 }, { "epoch": 0.9, "learning_rate": 3.9625112287265867e-07, "loss": 0.0446, "step": 2281 }, { "epoch": 0.9, "learning_rate": 3.962462007713972e-07, "loss": 0.0388, "step": 2282 }, { "epoch": 0.9, "learning_rate": 3.962412754716209e-07, "loss": 0.0407, "step": 2283 }, { "epoch": 0.9, "learning_rate": 3.962363469734101e-07, "loss": 0.0453, "step": 2284 }, { "epoch": 0.9, "learning_rate": 3.9623141527684504e-07, "loss": 0.0453, "step": 2285 }, { "epoch": 0.9, "learning_rate": 3.962264803820062e-07, "loss": 0.0368, "step": 2286 }, { "epoch": 0.9, "learning_rate": 3.962215422889739e-07, "loss": 0.0372, "step": 2287 }, { "epoch": 0.9, "learning_rate": 3.962166009978287e-07, "loss": 0.0416, "step": 2288 }, { "epoch": 0.9, "learning_rate": 3.9621165650865117e-07, "loss": 0.0409, "step": 2289 }, { "epoch": 0.9, "learning_rate": 3.962067088215218e-07, "loss": 0.0344, "step": 2290 }, { "epoch": 0.9, "learning_rate": 3.9620175793652127e-07, "loss": 0.0358, "step": 2291 }, { "epoch": 0.9, "learning_rate": 3.961968038537303e-07, "loss": 0.0417, "step": 2292 }, { "epoch": 0.9, "learning_rate": 3.961918465732296e-07, "loss": 0.0369, "step": 2293 }, { "epoch": 0.9, "learning_rate": 3.961868860951e-07, "loss": 0.0402, "step": 2294 }, { "epoch": 0.9, "learning_rate": 3.961819224194223e-07, "loss": 0.0411, "step": 2295 }, { "epoch": 0.9, "learning_rate": 3.9617695554627745e-07, "loss": 0.0391, "step": 2296 }, { "epoch": 0.91, "learning_rate": 3.961719854757463e-07, "loss": 0.045, "step": 2297 }, { "epoch": 0.91, "learning_rate": 3.9616701220791e-07, "loss": 0.0411, "step": 2298 }, { "epoch": 0.91, "learning_rate": 3.961620357428496e-07, "loss": 0.0395, "step": 2299 }, { "epoch": 0.91, "learning_rate": 3.9615705608064604e-07, "loss": 0.038, "step": 2300 }, { "epoch": 0.91, "learning_rate": 3.961520732213807e-07, "loss": 0.041, "step": 2301 }, { "epoch": 0.91, "learning_rate": 3.961470871651346e-07, "loss": 0.0376, "step": 2302 }, { "epoch": 0.91, "learning_rate": 3.9614209791198913e-07, "loss": 0.0368, "step": 2303 }, { "epoch": 0.91, "learning_rate": 3.9613710546202554e-07, "loss": 0.0394, "step": 2304 }, { "epoch": 0.91, "learning_rate": 3.9613210981532524e-07, "loss": 0.0418, "step": 2305 }, { "epoch": 0.91, "learning_rate": 3.961271109719696e-07, "loss": 0.0349, "step": 2306 }, { "epoch": 0.91, "learning_rate": 3.9612210893204017e-07, "loss": 0.036, "step": 2307 }, { "epoch": 0.91, "learning_rate": 3.961171036956184e-07, "loss": 0.0439, "step": 2308 }, { "epoch": 0.91, "learning_rate": 3.9611209526278593e-07, "loss": 0.0406, "step": 2309 }, { "epoch": 0.91, "learning_rate": 3.961070836336244e-07, "loss": 0.0384, "step": 2310 }, { "epoch": 0.91, "learning_rate": 3.9610206880821543e-07, "loss": 0.0403, "step": 2311 }, { "epoch": 0.91, "learning_rate": 3.960970507866407e-07, "loss": 0.0401, "step": 2312 }, { "epoch": 0.91, "learning_rate": 3.9609202956898215e-07, "loss": 0.0392, "step": 2313 }, { "epoch": 0.91, "learning_rate": 3.960870051553216e-07, "loss": 0.036, "step": 2314 }, { "epoch": 0.91, "learning_rate": 3.9608197754574074e-07, "loss": 0.0439, "step": 2315 }, { "epoch": 0.91, "learning_rate": 3.9607694674032175e-07, "loss": 0.0318, "step": 2316 }, { "epoch": 0.91, "learning_rate": 3.9607191273914655e-07, "loss": 0.0402, "step": 2317 }, { "epoch": 0.91, "learning_rate": 3.9606687554229714e-07, "loss": 0.0448, "step": 2318 }, { "epoch": 0.91, "learning_rate": 3.960618351498556e-07, "loss": 0.0395, "step": 2319 }, { "epoch": 0.91, "learning_rate": 3.960567915619042e-07, "loss": 0.0386, "step": 2320 }, { "epoch": 0.91, "learning_rate": 3.96051744778525e-07, "loss": 0.0375, "step": 2321 }, { "epoch": 0.92, "learning_rate": 3.9604669479980035e-07, "loss": 0.042, "step": 2322 }, { "epoch": 0.92, "learning_rate": 3.9604164162581257e-07, "loss": 0.0399, "step": 2323 }, { "epoch": 0.92, "learning_rate": 3.960365852566439e-07, "loss": 0.0395, "step": 2324 }, { "epoch": 0.92, "learning_rate": 3.9603152569237686e-07, "loss": 0.0383, "step": 2325 }, { "epoch": 0.92, "learning_rate": 3.9602646293309394e-07, "loss": 0.0413, "step": 2326 }, { "epoch": 0.92, "learning_rate": 3.960213969788776e-07, "loss": 0.0438, "step": 2327 }, { "epoch": 0.92, "learning_rate": 3.9601632782981037e-07, "loss": 0.0352, "step": 2328 }, { "epoch": 0.92, "learning_rate": 3.960112554859749e-07, "loss": 0.0422, "step": 2329 }, { "epoch": 0.92, "learning_rate": 3.960061799474539e-07, "loss": 0.0378, "step": 2330 }, { "epoch": 0.92, "learning_rate": 3.9600110121433e-07, "loss": 0.0393, "step": 2331 }, { "epoch": 0.92, "learning_rate": 3.959960192866861e-07, "loss": 0.0356, "step": 2332 }, { "epoch": 0.92, "learning_rate": 3.9599093416460496e-07, "loss": 0.0409, "step": 2333 }, { "epoch": 0.92, "learning_rate": 3.9598584584816946e-07, "loss": 0.0417, "step": 2334 }, { "epoch": 0.92, "learning_rate": 3.959807543374625e-07, "loss": 0.0417, "step": 2335 }, { "epoch": 0.92, "learning_rate": 3.9597565963256707e-07, "loss": 0.0426, "step": 2336 }, { "epoch": 0.92, "learning_rate": 3.9597056173356636e-07, "loss": 0.04, "step": 2337 }, { "epoch": 0.92, "learning_rate": 3.9596546064054327e-07, "loss": 0.044, "step": 2338 }, { "epoch": 0.92, "learning_rate": 3.9596035635358096e-07, "loss": 0.0418, "step": 2339 }, { "epoch": 0.92, "learning_rate": 3.9595524887276276e-07, "loss": 0.0395, "step": 2340 }, { "epoch": 0.92, "learning_rate": 3.9595013819817174e-07, "loss": 0.0342, "step": 2341 }, { "epoch": 0.92, "learning_rate": 3.959450243298913e-07, "loss": 0.0392, "step": 2342 }, { "epoch": 0.92, "learning_rate": 3.959399072680048e-07, "loss": 0.0462, "step": 2343 }, { "epoch": 0.92, "learning_rate": 3.9593478701259556e-07, "loss": 0.0385, "step": 2344 }, { "epoch": 0.92, "learning_rate": 3.959296635637471e-07, "loss": 0.0409, "step": 2345 }, { "epoch": 0.92, "learning_rate": 3.959245369215429e-07, "loss": 0.0389, "step": 2346 }, { "epoch": 0.92, "learning_rate": 3.959194070860665e-07, "loss": 0.0422, "step": 2347 }, { "epoch": 0.93, "learning_rate": 3.959142740574016e-07, "loss": 0.0397, "step": 2348 }, { "epoch": 0.93, "learning_rate": 3.959091378356317e-07, "loss": 0.0451, "step": 2349 }, { "epoch": 0.93, "learning_rate": 3.959039984208407e-07, "loss": 0.042, "step": 2350 }, { "epoch": 0.93, "learning_rate": 3.9589885581311216e-07, "loss": 0.0397, "step": 2351 }, { "epoch": 0.93, "learning_rate": 3.9589371001253004e-07, "loss": 0.039, "step": 2352 }, { "epoch": 0.93, "learning_rate": 3.958885610191782e-07, "loss": 0.0393, "step": 2353 }, { "epoch": 0.93, "learning_rate": 3.9588340883314044e-07, "loss": 0.0392, "step": 2354 }, { "epoch": 0.93, "learning_rate": 3.958782534545009e-07, "loss": 0.0371, "step": 2355 }, { "epoch": 0.93, "learning_rate": 3.9587309488334357e-07, "loss": 0.0321, "step": 2356 }, { "epoch": 0.93, "learning_rate": 3.9586793311975244e-07, "loss": 0.0383, "step": 2357 }, { "epoch": 0.93, "learning_rate": 3.9586276816381165e-07, "loss": 0.0426, "step": 2358 }, { "epoch": 0.93, "learning_rate": 3.9585760001560545e-07, "loss": 0.0425, "step": 2359 }, { "epoch": 0.93, "learning_rate": 3.9585242867521804e-07, "loss": 0.0389, "step": 2360 }, { "epoch": 0.93, "learning_rate": 3.958472541427337e-07, "loss": 0.0367, "step": 2361 }, { "epoch": 0.93, "learning_rate": 3.9584207641823675e-07, "loss": 0.0393, "step": 2362 }, { "epoch": 0.93, "learning_rate": 3.9583689550181164e-07, "loss": 0.0395, "step": 2363 }, { "epoch": 0.93, "learning_rate": 3.9583171139354276e-07, "loss": 0.0324, "step": 2364 }, { "epoch": 0.93, "learning_rate": 3.958265240935146e-07, "loss": 0.0388, "step": 2365 }, { "epoch": 0.93, "learning_rate": 3.9582133360181175e-07, "loss": 0.0383, "step": 2366 }, { "epoch": 0.93, "learning_rate": 3.9581613991851875e-07, "loss": 0.0425, "step": 2367 }, { "epoch": 0.93, "learning_rate": 3.9581094304372027e-07, "loss": 0.0458, "step": 2368 }, { "epoch": 0.93, "learning_rate": 3.95805742977501e-07, "loss": 0.0434, "step": 2369 }, { "epoch": 0.93, "learning_rate": 3.9580053971994575e-07, "loss": 0.039, "step": 2370 }, { "epoch": 0.93, "learning_rate": 3.9579533327113926e-07, "loss": 0.038, "step": 2371 }, { "epoch": 0.93, "learning_rate": 3.9579012363116644e-07, "loss": 0.0355, "step": 2372 }, { "epoch": 0.94, "learning_rate": 3.957849108001122e-07, "loss": 0.0431, "step": 2373 }, { "epoch": 0.94, "learning_rate": 3.957796947780614e-07, "loss": 0.0443, "step": 2374 }, { "epoch": 0.94, "learning_rate": 3.957744755650991e-07, "loss": 0.0356, "step": 2375 }, { "epoch": 0.94, "learning_rate": 3.957692531613105e-07, "loss": 0.0393, "step": 2376 }, { "epoch": 0.94, "learning_rate": 3.9576402756678054e-07, "loss": 0.0373, "step": 2377 }, { "epoch": 0.94, "learning_rate": 3.9575879878159444e-07, "loss": 0.0379, "step": 2378 }, { "epoch": 0.94, "learning_rate": 3.9575356680583745e-07, "loss": 0.0432, "step": 2379 }, { "epoch": 0.94, "learning_rate": 3.957483316395949e-07, "loss": 0.035, "step": 2380 }, { "epoch": 0.94, "learning_rate": 3.9574309328295196e-07, "loss": 0.0419, "step": 2381 }, { "epoch": 0.94, "learning_rate": 3.957378517359942e-07, "loss": 0.0371, "step": 2382 }, { "epoch": 0.94, "learning_rate": 3.9573260699880684e-07, "loss": 0.0386, "step": 2383 }, { "epoch": 0.94, "learning_rate": 3.957273590714755e-07, "loss": 0.0416, "step": 2384 }, { "epoch": 0.94, "learning_rate": 3.957221079540857e-07, "loss": 0.0402, "step": 2385 }, { "epoch": 0.94, "learning_rate": 3.9571685364672296e-07, "loss": 0.046, "step": 2386 }, { "epoch": 0.94, "learning_rate": 3.95711596149473e-07, "loss": 0.0403, "step": 2387 }, { "epoch": 0.94, "learning_rate": 3.9570633546242147e-07, "loss": 0.0425, "step": 2388 }, { "epoch": 0.94, "learning_rate": 3.9570107158565404e-07, "loss": 0.0376, "step": 2389 }, { "epoch": 0.94, "learning_rate": 3.9569580451925666e-07, "loss": 0.0383, "step": 2390 }, { "epoch": 0.94, "learning_rate": 3.9569053426331506e-07, "loss": 0.0349, "step": 2391 }, { "epoch": 0.94, "learning_rate": 3.9568526081791516e-07, "loss": 0.0397, "step": 2392 }, { "epoch": 0.94, "learning_rate": 3.956799841831429e-07, "loss": 0.0402, "step": 2393 }, { "epoch": 0.94, "learning_rate": 3.956747043590844e-07, "loss": 0.0398, "step": 2394 }, { "epoch": 0.94, "learning_rate": 3.956694213458255e-07, "loss": 0.0356, "step": 2395 }, { "epoch": 0.94, "learning_rate": 3.9566413514345244e-07, "loss": 0.0381, "step": 2396 }, { "epoch": 0.94, "learning_rate": 3.9565884575205136e-07, "loss": 0.0368, "step": 2397 }, { "epoch": 0.95, "learning_rate": 3.9565355317170845e-07, "loss": 0.04, "step": 2398 }, { "epoch": 0.95, "learning_rate": 3.9564825740251e-07, "loss": 0.0397, "step": 2399 }, { "epoch": 0.95, "learning_rate": 3.956429584445423e-07, "loss": 0.0431, "step": 2400 }, { "epoch": 0.95, "learning_rate": 3.9563765629789173e-07, "loss": 0.0469, "step": 2401 }, { "epoch": 0.95, "learning_rate": 3.956323509626447e-07, "loss": 0.0394, "step": 2402 }, { "epoch": 0.95, "learning_rate": 3.9562704243888765e-07, "loss": 0.0359, "step": 2403 }, { "epoch": 0.95, "learning_rate": 3.956217307267071e-07, "loss": 0.0412, "step": 2404 }, { "epoch": 0.95, "learning_rate": 3.956164158261897e-07, "loss": 0.0455, "step": 2405 }, { "epoch": 0.95, "learning_rate": 3.95611097737422e-07, "loss": 0.0407, "step": 2406 }, { "epoch": 0.95, "learning_rate": 3.9560577646049074e-07, "loss": 0.0385, "step": 2407 }, { "epoch": 0.95, "learning_rate": 3.956004519954826e-07, "loss": 0.0402, "step": 2408 }, { "epoch": 0.95, "learning_rate": 3.9559512434248434e-07, "loss": 0.0377, "step": 2409 }, { "epoch": 0.95, "learning_rate": 3.9558979350158284e-07, "loss": 0.0433, "step": 2410 }, { "epoch": 0.95, "learning_rate": 3.95584459472865e-07, "loss": 0.041, "step": 2411 }, { "epoch": 0.95, "learning_rate": 3.9557912225641767e-07, "loss": 0.0414, "step": 2412 }, { "epoch": 0.95, "learning_rate": 3.955737818523279e-07, "loss": 0.0381, "step": 2413 }, { "epoch": 0.95, "learning_rate": 3.9556843826068274e-07, "loss": 0.0438, "step": 2414 }, { "epoch": 0.95, "learning_rate": 3.9556309148156927e-07, "loss": 0.0354, "step": 2415 }, { "epoch": 0.95, "learning_rate": 3.9555774151507463e-07, "loss": 0.0385, "step": 2416 }, { "epoch": 0.95, "learning_rate": 3.95552388361286e-07, "loss": 0.0414, "step": 2417 }, { "epoch": 0.95, "learning_rate": 3.9554703202029067e-07, "loss": 0.0408, "step": 2418 }, { "epoch": 0.95, "learning_rate": 3.9554167249217594e-07, "loss": 0.0383, "step": 2419 }, { "epoch": 0.95, "learning_rate": 3.955363097770291e-07, "loss": 0.0387, "step": 2420 }, { "epoch": 0.95, "learning_rate": 3.955309438749376e-07, "loss": 0.0365, "step": 2421 }, { "epoch": 0.95, "learning_rate": 3.9552557478598885e-07, "loss": 0.0417, "step": 2422 }, { "epoch": 0.95, "learning_rate": 3.9552020251027046e-07, "loss": 0.0425, "step": 2423 }, { "epoch": 0.96, "learning_rate": 3.9551482704786987e-07, "loss": 0.0411, "step": 2424 }, { "epoch": 0.96, "learning_rate": 3.9550944839887473e-07, "loss": 0.0382, "step": 2425 }, { "epoch": 0.96, "learning_rate": 3.955040665633728e-07, "loss": 0.036, "step": 2426 }, { "epoch": 0.96, "learning_rate": 3.9549868154145167e-07, "loss": 0.0374, "step": 2427 }, { "epoch": 0.96, "learning_rate": 3.9549329333319913e-07, "loss": 0.0423, "step": 2428 }, { "epoch": 0.96, "learning_rate": 3.954879019387031e-07, "loss": 0.0379, "step": 2429 }, { "epoch": 0.96, "learning_rate": 3.954825073580513e-07, "loss": 0.0406, "step": 2430 }, { "epoch": 0.96, "learning_rate": 3.954771095913318e-07, "loss": 0.0376, "step": 2431 }, { "epoch": 0.96, "learning_rate": 3.9547170863863244e-07, "loss": 0.0381, "step": 2432 }, { "epoch": 0.96, "learning_rate": 3.9546630450004137e-07, "loss": 0.0368, "step": 2433 }, { "epoch": 0.96, "learning_rate": 3.9546089717564657e-07, "loss": 0.0424, "step": 2434 }, { "epoch": 0.96, "learning_rate": 3.9545548666553624e-07, "loss": 0.0445, "step": 2435 }, { "epoch": 0.96, "learning_rate": 3.9545007296979847e-07, "loss": 0.0409, "step": 2436 }, { "epoch": 0.96, "learning_rate": 3.954446560885216e-07, "loss": 0.0368, "step": 2437 }, { "epoch": 0.96, "learning_rate": 3.9543923602179393e-07, "loss": 0.0387, "step": 2438 }, { "epoch": 0.96, "learning_rate": 3.9543381276970366e-07, "loss": 0.0331, "step": 2439 }, { "epoch": 0.96, "learning_rate": 3.954283863323393e-07, "loss": 0.0388, "step": 2440 }, { "epoch": 0.96, "learning_rate": 3.9542295670978926e-07, "loss": 0.0413, "step": 2441 }, { "epoch": 0.96, "learning_rate": 3.9541752390214207e-07, "loss": 0.042, "step": 2442 }, { "epoch": 0.96, "learning_rate": 3.9541208790948624e-07, "loss": 0.0416, "step": 2443 }, { "epoch": 0.96, "learning_rate": 3.954066487319103e-07, "loss": 0.0453, "step": 2444 }, { "epoch": 0.96, "learning_rate": 3.95401206369503e-07, "loss": 0.0375, "step": 2445 }, { "epoch": 0.96, "learning_rate": 3.9539576082235303e-07, "loss": 0.0419, "step": 2446 }, { "epoch": 0.96, "learning_rate": 3.953903120905491e-07, "loss": 0.0417, "step": 2447 }, { "epoch": 0.96, "learning_rate": 3.953848601741801e-07, "loss": 0.0336, "step": 2448 }, { "epoch": 0.97, "learning_rate": 3.953794050733348e-07, "loss": 0.0427, "step": 2449 }, { "epoch": 0.97, "learning_rate": 3.9537394678810216e-07, "loss": 0.0429, "step": 2450 }, { "epoch": 0.97, "learning_rate": 3.9536848531857103e-07, "loss": 0.0433, "step": 2451 }, { "epoch": 0.97, "learning_rate": 3.9536302066483056e-07, "loss": 0.0457, "step": 2452 }, { "epoch": 0.97, "learning_rate": 3.9535755282696984e-07, "loss": 0.0383, "step": 2453 }, { "epoch": 0.97, "learning_rate": 3.9535208180507785e-07, "loss": 0.04, "step": 2454 }, { "epoch": 0.97, "learning_rate": 3.9534660759924386e-07, "loss": 0.034, "step": 2455 }, { "epoch": 0.97, "learning_rate": 3.953411302095571e-07, "loss": 0.0386, "step": 2456 }, { "epoch": 0.97, "learning_rate": 3.953356496361067e-07, "loss": 0.0399, "step": 2457 }, { "epoch": 0.97, "learning_rate": 3.9533016587898213e-07, "loss": 0.0389, "step": 2458 }, { "epoch": 0.97, "learning_rate": 3.953246789382727e-07, "loss": 0.0408, "step": 2459 }, { "epoch": 0.97, "learning_rate": 3.953191888140679e-07, "loss": 0.0334, "step": 2460 }, { "epoch": 0.97, "learning_rate": 3.9531369550645717e-07, "loss": 0.0326, "step": 2461 }, { "epoch": 0.97, "learning_rate": 3.9530819901553e-07, "loss": 0.0406, "step": 2462 }, { "epoch": 0.97, "learning_rate": 3.953026993413761e-07, "loss": 0.0434, "step": 2463 }, { "epoch": 0.97, "learning_rate": 3.9529719648408493e-07, "loss": 0.0392, "step": 2464 }, { "epoch": 0.97, "learning_rate": 3.9529169044374635e-07, "loss": 0.0411, "step": 2465 }, { "epoch": 0.97, "learning_rate": 3.9528618122045e-07, "loss": 0.0352, "step": 2466 }, { "epoch": 0.97, "learning_rate": 3.952806688142857e-07, "loss": 0.0435, "step": 2467 }, { "epoch": 0.97, "learning_rate": 3.9527515322534334e-07, "loss": 0.0356, "step": 2468 }, { "epoch": 0.97, "learning_rate": 3.952696344537127e-07, "loss": 0.0388, "step": 2469 }, { "epoch": 0.97, "learning_rate": 3.9526411249948384e-07, "loss": 0.0376, "step": 2470 }, { "epoch": 0.97, "learning_rate": 3.9525858736274664e-07, "loss": 0.0419, "step": 2471 }, { "epoch": 0.97, "learning_rate": 3.9525305904359125e-07, "loss": 0.0402, "step": 2472 }, { "epoch": 0.97, "learning_rate": 3.952475275421078e-07, "loss": 0.0367, "step": 2473 }, { "epoch": 0.98, "learning_rate": 3.952419928583864e-07, "loss": 0.0391, "step": 2474 }, { "epoch": 0.98, "learning_rate": 3.9523645499251714e-07, "loss": 0.0376, "step": 2475 }, { "epoch": 0.98, "learning_rate": 3.952309139445905e-07, "loss": 0.0396, "step": 2476 }, { "epoch": 0.98, "learning_rate": 3.9522536971469666e-07, "loss": 0.0389, "step": 2477 }, { "epoch": 0.98, "learning_rate": 3.9521982230292593e-07, "loss": 0.0374, "step": 2478 }, { "epoch": 0.98, "learning_rate": 3.9521427170936885e-07, "loss": 0.04, "step": 2479 }, { "epoch": 0.98, "learning_rate": 3.9520871793411583e-07, "loss": 0.0365, "step": 2480 }, { "epoch": 0.98, "learning_rate": 3.9520316097725745e-07, "loss": 0.0461, "step": 2481 }, { "epoch": 0.98, "learning_rate": 3.9519760083888414e-07, "loss": 0.034, "step": 2482 }, { "epoch": 0.98, "learning_rate": 3.9519203751908663e-07, "loss": 0.0414, "step": 2483 }, { "epoch": 0.98, "learning_rate": 3.9518647101795555e-07, "loss": 0.0325, "step": 2484 }, { "epoch": 0.98, "learning_rate": 3.9518090133558165e-07, "loss": 0.0435, "step": 2485 }, { "epoch": 0.98, "learning_rate": 3.951753284720557e-07, "loss": 0.0416, "step": 2486 }, { "epoch": 0.98, "learning_rate": 3.9516975242746853e-07, "loss": 0.0381, "step": 2487 }, { "epoch": 0.98, "learning_rate": 3.9516417320191105e-07, "loss": 0.0389, "step": 2488 }, { "epoch": 0.98, "learning_rate": 3.951585907954741e-07, "loss": 0.0415, "step": 2489 }, { "epoch": 0.98, "learning_rate": 3.9515300520824875e-07, "loss": 0.0321, "step": 2490 }, { "epoch": 0.98, "learning_rate": 3.9514741644032604e-07, "loss": 0.0413, "step": 2491 }, { "epoch": 0.98, "learning_rate": 3.95141824491797e-07, "loss": 0.0377, "step": 2492 }, { "epoch": 0.98, "learning_rate": 3.951362293627528e-07, "loss": 0.0354, "step": 2493 }, { "epoch": 0.98, "learning_rate": 3.951306310532847e-07, "loss": 0.0416, "step": 2494 }, { "epoch": 0.98, "learning_rate": 3.9512502956348384e-07, "loss": 0.0367, "step": 2495 }, { "epoch": 0.98, "learning_rate": 3.9511942489344155e-07, "loss": 0.0414, "step": 2496 }, { "epoch": 0.98, "learning_rate": 3.9511381704324916e-07, "loss": 0.0361, "step": 2497 }, { "epoch": 0.98, "learning_rate": 3.951082060129981e-07, "loss": 0.0362, "step": 2498 }, { "epoch": 0.98, "learning_rate": 3.9510259180277983e-07, "loss": 0.0361, "step": 2499 }, { "epoch": 0.99, "learning_rate": 3.950969744126858e-07, "loss": 0.0373, "step": 2500 }, { "epoch": 0.99, "learning_rate": 3.950913538428076e-07, "loss": 0.0376, "step": 2501 }, { "epoch": 0.99, "learning_rate": 3.950857300932369e-07, "loss": 0.0369, "step": 2502 }, { "epoch": 0.99, "learning_rate": 3.950801031640652e-07, "loss": 0.0444, "step": 2503 }, { "epoch": 0.99, "learning_rate": 3.950744730553844e-07, "loss": 0.0393, "step": 2504 }, { "epoch": 0.99, "learning_rate": 3.950688397672861e-07, "loss": 0.0384, "step": 2505 }, { "epoch": 0.99, "learning_rate": 3.9506320329986223e-07, "loss": 0.0434, "step": 2506 }, { "epoch": 0.99, "learning_rate": 3.950575636532046e-07, "loss": 0.0432, "step": 2507 }, { "epoch": 0.99, "learning_rate": 3.950519208274051e-07, "loss": 0.0366, "step": 2508 }, { "epoch": 0.99, "learning_rate": 3.950462748225558e-07, "loss": 0.0448, "step": 2509 }, { "epoch": 0.99, "learning_rate": 3.9504062563874856e-07, "loss": 0.0336, "step": 2510 }, { "epoch": 0.99, "learning_rate": 3.9503497327607567e-07, "loss": 0.0414, "step": 2511 }, { "epoch": 0.99, "learning_rate": 3.9502931773462905e-07, "loss": 0.0426, "step": 2512 }, { "epoch": 0.99, "learning_rate": 3.950236590145009e-07, "loss": 0.041, "step": 2513 }, { "epoch": 0.99, "learning_rate": 3.9501799711578367e-07, "loss": 0.0413, "step": 2514 }, { "epoch": 0.99, "learning_rate": 3.950123320385694e-07, "loss": 0.0352, "step": 2515 }, { "epoch": 0.99, "learning_rate": 3.950066637829505e-07, "loss": 0.0358, "step": 2516 }, { "epoch": 0.99, "learning_rate": 3.9500099234901937e-07, "loss": 0.0456, "step": 2517 }, { "epoch": 0.99, "learning_rate": 3.949953177368684e-07, "loss": 0.0315, "step": 2518 }, { "epoch": 0.99, "learning_rate": 3.9498963994659016e-07, "loss": 0.0404, "step": 2519 }, { "epoch": 0.99, "learning_rate": 3.949839589782771e-07, "loss": 0.0376, "step": 2520 }, { "epoch": 0.99, "learning_rate": 3.949782748320219e-07, "loss": 0.0435, "step": 2521 }, { "epoch": 0.99, "learning_rate": 3.9497258750791707e-07, "loss": 0.0333, "step": 2522 }, { "epoch": 0.99, "learning_rate": 3.9496689700605545e-07, "loss": 0.0346, "step": 2523 }, { "epoch": 0.99, "learning_rate": 3.949612033265297e-07, "loss": 0.0427, "step": 2524 }, { "epoch": 1.0, "learning_rate": 3.9495550646943264e-07, "loss": 0.0428, "step": 2525 }, { "epoch": 1.0, "learning_rate": 3.949498064348572e-07, "loss": 0.0416, "step": 2526 }, { "epoch": 1.0, "learning_rate": 3.949441032228961e-07, "loss": 0.0357, "step": 2527 }, { "epoch": 1.0, "learning_rate": 3.9493839683364243e-07, "loss": 0.0357, "step": 2528 }, { "epoch": 1.0, "learning_rate": 3.949326872671892e-07, "loss": 0.0384, "step": 2529 }, { "epoch": 1.0, "learning_rate": 3.949269745236294e-07, "loss": 0.0367, "step": 2530 }, { "epoch": 1.0, "learning_rate": 3.949212586030561e-07, "loss": 0.0332, "step": 2531 }, { "epoch": 1.0, "learning_rate": 3.9491553950556264e-07, "loss": 0.0408, "step": 2532 }, { "epoch": 1.0, "learning_rate": 3.949098172312421e-07, "loss": 0.039, "step": 2533 }, { "epoch": 1.0, "learning_rate": 3.9490409178018774e-07, "loss": 0.0352, "step": 2534 }, { "epoch": 1.0, "learning_rate": 3.9489836315249294e-07, "loss": 0.0459, "step": 2535 }, { "epoch": 1.0, "learning_rate": 3.94892631348251e-07, "loss": 0.0417, "step": 2536 }, { "epoch": 1.0, "learning_rate": 3.9488689636755536e-07, "loss": 0.0353, "step": 2537 }, { "epoch": 1.0, "learning_rate": 3.948811582104995e-07, "loss": 0.0394, "step": 2538 }, { "epoch": 1.0, "learning_rate": 3.9487541687717704e-07, "loss": 0.0334, "step": 2539 }, { "epoch": 1.0, "learning_rate": 3.948696723676814e-07, "loss": 0.0354, "step": 2540 }, { "epoch": 1.0, "learning_rate": 3.9486392468210623e-07, "loss": 0.0385, "step": 2541 }, { "epoch": 1.0, "learning_rate": 3.948581738205453e-07, "loss": 0.034, "step": 2542 }, { "epoch": 1.0, "learning_rate": 3.9485241978309224e-07, "loss": 0.0364, "step": 2543 }, { "epoch": 1.0, "learning_rate": 3.948466625698409e-07, "loss": 0.0354, "step": 2544 }, { "epoch": 1.0, "learning_rate": 3.9484090218088515e-07, "loss": 0.036, "step": 2545 }, { "epoch": 1.0, "learning_rate": 3.948351386163187e-07, "loss": 0.0378, "step": 2546 }, { "epoch": 1.0, "learning_rate": 3.9482937187623567e-07, "loss": 0.0362, "step": 2547 }, { "epoch": 1.0, "learning_rate": 3.9482360196073e-07, "loss": 0.0406, "step": 2548 }, { "epoch": 1.0, "learning_rate": 3.9481782886989574e-07, "loss": 0.0366, "step": 2549 }, { "epoch": 1.0, "learning_rate": 3.948120526038269e-07, "loss": 0.0402, "step": 2550 }, { "epoch": 1.01, "learning_rate": 3.948062731626177e-07, "loss": 0.0356, "step": 2551 }, { "epoch": 1.01, "learning_rate": 3.948004905463623e-07, "loss": 0.0373, "step": 2552 }, { "epoch": 1.01, "learning_rate": 3.9479470475515504e-07, "loss": 0.0374, "step": 2553 }, { "epoch": 1.01, "learning_rate": 3.9478891578909006e-07, "loss": 0.0317, "step": 2554 }, { "epoch": 1.01, "learning_rate": 3.947831236482618e-07, "loss": 0.0372, "step": 2555 }, { "epoch": 1.01, "learning_rate": 3.947773283327647e-07, "loss": 0.0342, "step": 2556 }, { "epoch": 1.01, "learning_rate": 3.947715298426932e-07, "loss": 0.0282, "step": 2557 }, { "epoch": 1.01, "learning_rate": 3.9476572817814177e-07, "loss": 0.0349, "step": 2558 }, { "epoch": 1.01, "learning_rate": 3.9475992333920494e-07, "loss": 0.0331, "step": 2559 }, { "epoch": 1.01, "learning_rate": 3.9475411532597736e-07, "loss": 0.0368, "step": 2560 }, { "epoch": 1.01, "learning_rate": 3.9474830413855373e-07, "loss": 0.0345, "step": 2561 }, { "epoch": 1.01, "learning_rate": 3.947424897770287e-07, "loss": 0.0368, "step": 2562 }, { "epoch": 1.01, "learning_rate": 3.947366722414971e-07, "loss": 0.0421, "step": 2563 }, { "epoch": 1.01, "learning_rate": 3.947308515320537e-07, "loss": 0.0395, "step": 2564 }, { "epoch": 1.01, "learning_rate": 3.947250276487933e-07, "loss": 0.0372, "step": 2565 }, { "epoch": 1.01, "learning_rate": 3.9471920059181095e-07, "loss": 0.0382, "step": 2566 }, { "epoch": 1.01, "learning_rate": 3.947133703612016e-07, "loss": 0.0379, "step": 2567 }, { "epoch": 1.01, "learning_rate": 3.947075369570602e-07, "loss": 0.0321, "step": 2568 }, { "epoch": 1.01, "learning_rate": 3.947017003794819e-07, "loss": 0.0382, "step": 2569 }, { "epoch": 1.01, "learning_rate": 3.946958606285618e-07, "loss": 0.0377, "step": 2570 }, { "epoch": 1.01, "learning_rate": 3.9469001770439504e-07, "loss": 0.0373, "step": 2571 }, { "epoch": 1.01, "learning_rate": 3.946841716070769e-07, "loss": 0.0359, "step": 2572 }, { "epoch": 1.01, "learning_rate": 3.946783223367027e-07, "loss": 0.0394, "step": 2573 }, { "epoch": 1.01, "learning_rate": 3.9467246989336764e-07, "loss": 0.0337, "step": 2574 }, { "epoch": 1.01, "learning_rate": 3.946666142771672e-07, "loss": 0.0401, "step": 2575 }, { "epoch": 1.02, "learning_rate": 3.9466075548819686e-07, "loss": 0.0354, "step": 2576 }, { "epoch": 1.02, "learning_rate": 3.94654893526552e-07, "loss": 0.0406, "step": 2577 }, { "epoch": 1.02, "learning_rate": 3.9464902839232827e-07, "loss": 0.041, "step": 2578 }, { "epoch": 1.02, "learning_rate": 3.946431600856212e-07, "loss": 0.0356, "step": 2579 }, { "epoch": 1.02, "learning_rate": 3.9463728860652636e-07, "loss": 0.0357, "step": 2580 }, { "epoch": 1.02, "learning_rate": 3.9463141395513963e-07, "loss": 0.0355, "step": 2581 }, { "epoch": 1.02, "learning_rate": 3.9462553613155664e-07, "loss": 0.037, "step": 2582 }, { "epoch": 1.02, "learning_rate": 3.946196551358732e-07, "loss": 0.0331, "step": 2583 }, { "epoch": 1.02, "learning_rate": 3.9461377096818516e-07, "loss": 0.0395, "step": 2584 }, { "epoch": 1.02, "learning_rate": 3.946078836285884e-07, "loss": 0.0341, "step": 2585 }, { "epoch": 1.02, "learning_rate": 3.94601993117179e-07, "loss": 0.0365, "step": 2586 }, { "epoch": 1.02, "learning_rate": 3.9459609943405275e-07, "loss": 0.0353, "step": 2587 }, { "epoch": 1.02, "learning_rate": 3.9459020257930593e-07, "loss": 0.0361, "step": 2588 }, { "epoch": 1.02, "learning_rate": 3.945843025530345e-07, "loss": 0.0351, "step": 2589 }, { "epoch": 1.02, "learning_rate": 3.945783993553347e-07, "loss": 0.0397, "step": 2590 }, { "epoch": 1.02, "learning_rate": 3.945724929863027e-07, "loss": 0.0343, "step": 2591 }, { "epoch": 1.02, "learning_rate": 3.9456658344603475e-07, "loss": 0.0347, "step": 2592 }, { "epoch": 1.02, "learning_rate": 3.945606707346273e-07, "loss": 0.0416, "step": 2593 }, { "epoch": 1.02, "learning_rate": 3.945547548521765e-07, "loss": 0.034, "step": 2594 }, { "epoch": 1.02, "learning_rate": 3.9454883579877897e-07, "loss": 0.0318, "step": 2595 }, { "epoch": 1.02, "learning_rate": 3.9454291357453103e-07, "loss": 0.0349, "step": 2596 }, { "epoch": 1.02, "learning_rate": 3.945369881795293e-07, "loss": 0.0344, "step": 2597 }, { "epoch": 1.02, "learning_rate": 3.945310596138703e-07, "loss": 0.0378, "step": 2598 }, { "epoch": 1.02, "learning_rate": 3.9452512787765073e-07, "loss": 0.0334, "step": 2599 }, { "epoch": 1.02, "learning_rate": 3.945191929709672e-07, "loss": 0.0395, "step": 2600 }, { "epoch": 1.03, "learning_rate": 3.9451325489391643e-07, "loss": 0.0341, "step": 2601 }, { "epoch": 1.03, "learning_rate": 3.945073136465953e-07, "loss": 0.0359, "step": 2602 }, { "epoch": 1.03, "learning_rate": 3.9450136922910053e-07, "loss": 0.0369, "step": 2603 }, { "epoch": 1.03, "learning_rate": 3.94495421641529e-07, "loss": 0.0344, "step": 2604 }, { "epoch": 1.03, "learning_rate": 3.9448947088397774e-07, "loss": 0.038, "step": 2605 }, { "epoch": 1.03, "learning_rate": 3.9448351695654374e-07, "loss": 0.0392, "step": 2606 }, { "epoch": 1.03, "learning_rate": 3.94477559859324e-07, "loss": 0.0441, "step": 2607 }, { "epoch": 1.03, "learning_rate": 3.944715995924155e-07, "loss": 0.0334, "step": 2608 }, { "epoch": 1.03, "learning_rate": 3.9446563615591553e-07, "loss": 0.0417, "step": 2609 }, { "epoch": 1.03, "learning_rate": 3.944596695499213e-07, "loss": 0.0341, "step": 2610 }, { "epoch": 1.03, "learning_rate": 3.9445369977452996e-07, "loss": 0.0368, "step": 2611 }, { "epoch": 1.03, "learning_rate": 3.9444772682983886e-07, "loss": 0.0336, "step": 2612 }, { "epoch": 1.03, "learning_rate": 3.9444175071594536e-07, "loss": 0.0389, "step": 2613 }, { "epoch": 1.03, "learning_rate": 3.9443577143294675e-07, "loss": 0.034, "step": 2614 }, { "epoch": 1.03, "learning_rate": 3.9442978898094064e-07, "loss": 0.0339, "step": 2615 }, { "epoch": 1.03, "learning_rate": 3.944238033600245e-07, "loss": 0.0353, "step": 2616 }, { "epoch": 1.03, "learning_rate": 3.944178145702958e-07, "loss": 0.0367, "step": 2617 }, { "epoch": 1.03, "learning_rate": 3.944118226118522e-07, "loss": 0.0374, "step": 2618 }, { "epoch": 1.03, "learning_rate": 3.944058274847914e-07, "loss": 0.0416, "step": 2619 }, { "epoch": 1.03, "learning_rate": 3.9439982918921107e-07, "loss": 0.0368, "step": 2620 }, { "epoch": 1.03, "learning_rate": 3.9439382772520895e-07, "loss": 0.0379, "step": 2621 }, { "epoch": 1.03, "learning_rate": 3.9438782309288295e-07, "loss": 0.0354, "step": 2622 }, { "epoch": 1.03, "learning_rate": 3.9438181529233085e-07, "loss": 0.0341, "step": 2623 }, { "epoch": 1.03, "learning_rate": 3.943758043236506e-07, "loss": 0.0364, "step": 2624 }, { "epoch": 1.03, "learning_rate": 3.9436979018694005e-07, "loss": 0.0381, "step": 2625 }, { "epoch": 1.03, "learning_rate": 3.9436377288229743e-07, "loss": 0.0331, "step": 2626 }, { "epoch": 1.04, "learning_rate": 3.943577524098207e-07, "loss": 0.0376, "step": 2627 }, { "epoch": 1.04, "learning_rate": 3.943517287696079e-07, "loss": 0.0338, "step": 2628 }, { "epoch": 1.04, "learning_rate": 3.943457019617574e-07, "loss": 0.0366, "step": 2629 }, { "epoch": 1.04, "learning_rate": 3.943396719863673e-07, "loss": 0.0397, "step": 2630 }, { "epoch": 1.04, "learning_rate": 3.9433363884353594e-07, "loss": 0.037, "step": 2631 }, { "epoch": 1.04, "learning_rate": 3.943276025333616e-07, "loss": 0.0329, "step": 2632 }, { "epoch": 1.04, "learning_rate": 3.943215630559426e-07, "loss": 0.0358, "step": 2633 }, { "epoch": 1.04, "learning_rate": 3.9431552041137754e-07, "loss": 0.0358, "step": 2634 }, { "epoch": 1.04, "learning_rate": 3.9430947459976485e-07, "loss": 0.0379, "step": 2635 }, { "epoch": 1.04, "learning_rate": 3.94303425621203e-07, "loss": 0.0362, "step": 2636 }, { "epoch": 1.04, "learning_rate": 3.942973734757906e-07, "loss": 0.041, "step": 2637 }, { "epoch": 1.04, "learning_rate": 3.942913181636264e-07, "loss": 0.0391, "step": 2638 }, { "epoch": 1.04, "learning_rate": 3.942852596848089e-07, "loss": 0.0331, "step": 2639 }, { "epoch": 1.04, "learning_rate": 3.9427919803943696e-07, "loss": 0.038, "step": 2640 }, { "epoch": 1.04, "learning_rate": 3.9427313322760943e-07, "loss": 0.036, "step": 2641 }, { "epoch": 1.04, "learning_rate": 3.94267065249425e-07, "loss": 0.036, "step": 2642 }, { "epoch": 1.04, "learning_rate": 3.9426099410498274e-07, "loss": 0.0377, "step": 2643 }, { "epoch": 1.04, "learning_rate": 3.9425491979438146e-07, "loss": 0.0348, "step": 2644 }, { "epoch": 1.04, "learning_rate": 3.9424884231772027e-07, "loss": 0.034, "step": 2645 }, { "epoch": 1.04, "learning_rate": 3.942427616750981e-07, "loss": 0.0379, "step": 2646 }, { "epoch": 1.04, "learning_rate": 3.942366778666142e-07, "loss": 0.0334, "step": 2647 }, { "epoch": 1.04, "learning_rate": 3.9423059089236766e-07, "loss": 0.0416, "step": 2648 }, { "epoch": 1.04, "learning_rate": 3.9422450075245765e-07, "loss": 0.0357, "step": 2649 }, { "epoch": 1.04, "learning_rate": 3.9421840744698345e-07, "loss": 0.045, "step": 2650 }, { "epoch": 1.04, "learning_rate": 3.942123109760445e-07, "loss": 0.0307, "step": 2651 }, { "epoch": 1.05, "learning_rate": 3.942062113397399e-07, "loss": 0.0372, "step": 2652 }, { "epoch": 1.05, "learning_rate": 3.942001085381693e-07, "loss": 0.0354, "step": 2653 }, { "epoch": 1.05, "learning_rate": 3.94194002571432e-07, "loss": 0.0351, "step": 2654 }, { "epoch": 1.05, "learning_rate": 3.9418789343962767e-07, "loss": 0.0443, "step": 2655 }, { "epoch": 1.05, "learning_rate": 3.941817811428558e-07, "loss": 0.0368, "step": 2656 }, { "epoch": 1.05, "learning_rate": 3.9417566568121603e-07, "loss": 0.0381, "step": 2657 }, { "epoch": 1.05, "learning_rate": 3.94169547054808e-07, "loss": 0.036, "step": 2658 }, { "epoch": 1.05, "learning_rate": 3.941634252637315e-07, "loss": 0.0352, "step": 2659 }, { "epoch": 1.05, "learning_rate": 3.9415730030808614e-07, "loss": 0.0411, "step": 2660 }, { "epoch": 1.05, "learning_rate": 3.94151172187972e-07, "loss": 0.0382, "step": 2661 }, { "epoch": 1.05, "learning_rate": 3.9414504090348883e-07, "loss": 0.0391, "step": 2662 }, { "epoch": 1.05, "learning_rate": 3.941389064547365e-07, "loss": 0.0393, "step": 2663 }, { "epoch": 1.05, "learning_rate": 3.94132768841815e-07, "loss": 0.0356, "step": 2664 }, { "epoch": 1.05, "learning_rate": 3.9412662806482454e-07, "loss": 0.0351, "step": 2665 }, { "epoch": 1.05, "learning_rate": 3.94120484123865e-07, "loss": 0.0406, "step": 2666 }, { "epoch": 1.05, "learning_rate": 3.941143370190366e-07, "loss": 0.0353, "step": 2667 }, { "epoch": 1.05, "learning_rate": 3.941081867504396e-07, "loss": 0.0434, "step": 2668 }, { "epoch": 1.05, "learning_rate": 3.941020333181741e-07, "loss": 0.0321, "step": 2669 }, { "epoch": 1.05, "learning_rate": 3.9409587672234047e-07, "loss": 0.0381, "step": 2670 }, { "epoch": 1.05, "learning_rate": 3.94089716963039e-07, "loss": 0.0354, "step": 2671 }, { "epoch": 1.05, "learning_rate": 3.940835540403702e-07, "loss": 0.0361, "step": 2672 }, { "epoch": 1.05, "learning_rate": 3.940773879544344e-07, "loss": 0.0335, "step": 2673 }, { "epoch": 1.05, "learning_rate": 3.940712187053322e-07, "loss": 0.0382, "step": 2674 }, { "epoch": 1.05, "learning_rate": 3.9406504629316404e-07, "loss": 0.0354, "step": 2675 }, { "epoch": 1.05, "learning_rate": 3.940588707180305e-07, "loss": 0.0382, "step": 2676 }, { "epoch": 1.06, "learning_rate": 3.940526919800324e-07, "loss": 0.036, "step": 2677 }, { "epoch": 1.06, "learning_rate": 3.940465100792703e-07, "loss": 0.0375, "step": 2678 }, { "epoch": 1.06, "learning_rate": 3.9404032501584504e-07, "loss": 0.033, "step": 2679 }, { "epoch": 1.06, "learning_rate": 3.940341367898574e-07, "loss": 0.036, "step": 2680 }, { "epoch": 1.06, "learning_rate": 3.9402794540140817e-07, "loss": 0.0382, "step": 2681 }, { "epoch": 1.06, "learning_rate": 3.9402175085059835e-07, "loss": 0.0359, "step": 2682 }, { "epoch": 1.06, "learning_rate": 3.9401555313752887e-07, "loss": 0.0371, "step": 2683 }, { "epoch": 1.06, "learning_rate": 3.940093522623008e-07, "loss": 0.0368, "step": 2684 }, { "epoch": 1.06, "learning_rate": 3.9400314822501504e-07, "loss": 0.0343, "step": 2685 }, { "epoch": 1.06, "learning_rate": 3.9399694102577286e-07, "loss": 0.0387, "step": 2686 }, { "epoch": 1.06, "learning_rate": 3.939907306646754e-07, "loss": 0.0357, "step": 2687 }, { "epoch": 1.06, "learning_rate": 3.9398451714182387e-07, "loss": 0.0321, "step": 2688 }, { "epoch": 1.06, "learning_rate": 3.939783004573195e-07, "loss": 0.0348, "step": 2689 }, { "epoch": 1.06, "learning_rate": 3.9397208061126366e-07, "loss": 0.0329, "step": 2690 }, { "epoch": 1.06, "learning_rate": 3.939658576037577e-07, "loss": 0.0306, "step": 2691 }, { "epoch": 1.06, "learning_rate": 3.939596314349031e-07, "loss": 0.0319, "step": 2692 }, { "epoch": 1.06, "learning_rate": 3.939534021048012e-07, "loss": 0.0352, "step": 2693 }, { "epoch": 1.06, "learning_rate": 3.9394716961355367e-07, "loss": 0.0362, "step": 2694 }, { "epoch": 1.06, "learning_rate": 3.939409339612621e-07, "loss": 0.0381, "step": 2695 }, { "epoch": 1.06, "learning_rate": 3.9393469514802797e-07, "loss": 0.0369, "step": 2696 }, { "epoch": 1.06, "learning_rate": 3.939284531739531e-07, "loss": 0.0381, "step": 2697 }, { "epoch": 1.06, "learning_rate": 3.939222080391392e-07, "loss": 0.0358, "step": 2698 }, { "epoch": 1.06, "learning_rate": 3.93915959743688e-07, "loss": 0.0426, "step": 2699 }, { "epoch": 1.06, "learning_rate": 3.9390970828770137e-07, "loss": 0.0439, "step": 2700 }, { "epoch": 1.06, "learning_rate": 3.939034536712812e-07, "loss": 0.0392, "step": 2701 }, { "epoch": 1.06, "learning_rate": 3.938971958945295e-07, "loss": 0.0357, "step": 2702 }, { "epoch": 1.07, "learning_rate": 3.938909349575482e-07, "loss": 0.035, "step": 2703 }, { "epoch": 1.07, "learning_rate": 3.938846708604393e-07, "loss": 0.0361, "step": 2704 }, { "epoch": 1.07, "learning_rate": 3.938784036033049e-07, "loss": 0.0428, "step": 2705 }, { "epoch": 1.07, "learning_rate": 3.9387213318624717e-07, "loss": 0.0352, "step": 2706 }, { "epoch": 1.07, "learning_rate": 3.938658596093684e-07, "loss": 0.0334, "step": 2707 }, { "epoch": 1.07, "learning_rate": 3.938595828727707e-07, "loss": 0.0359, "step": 2708 }, { "epoch": 1.07, "learning_rate": 3.9385330297655647e-07, "loss": 0.0355, "step": 2709 }, { "epoch": 1.07, "learning_rate": 3.9384701992082803e-07, "loss": 0.037, "step": 2710 }, { "epoch": 1.07, "learning_rate": 3.938407337056877e-07, "loss": 0.04, "step": 2711 }, { "epoch": 1.07, "learning_rate": 3.938344443312381e-07, "loss": 0.0385, "step": 2712 }, { "epoch": 1.07, "learning_rate": 3.9382815179758167e-07, "loss": 0.0341, "step": 2713 }, { "epoch": 1.07, "learning_rate": 3.938218561048209e-07, "loss": 0.0326, "step": 2714 }, { "epoch": 1.07, "learning_rate": 3.938155572530584e-07, "loss": 0.0402, "step": 2715 }, { "epoch": 1.07, "learning_rate": 3.9380925524239696e-07, "loss": 0.0351, "step": 2716 }, { "epoch": 1.07, "learning_rate": 3.938029500729392e-07, "loss": 0.0364, "step": 2717 }, { "epoch": 1.07, "learning_rate": 3.937966417447879e-07, "loss": 0.04, "step": 2718 }, { "epoch": 1.07, "learning_rate": 3.9379033025804585e-07, "loss": 0.0382, "step": 2719 }, { "epoch": 1.07, "learning_rate": 3.93784015612816e-07, "loss": 0.0399, "step": 2720 }, { "epoch": 1.07, "learning_rate": 3.9377769780920115e-07, "loss": 0.0362, "step": 2721 }, { "epoch": 1.07, "learning_rate": 3.937713768473044e-07, "loss": 0.0348, "step": 2722 }, { "epoch": 1.07, "learning_rate": 3.9376505272722867e-07, "loss": 0.0419, "step": 2723 }, { "epoch": 1.07, "learning_rate": 3.93758725449077e-07, "loss": 0.0379, "step": 2724 }, { "epoch": 1.07, "learning_rate": 3.937523950129527e-07, "loss": 0.035, "step": 2725 }, { "epoch": 1.07, "learning_rate": 3.9374606141895874e-07, "loss": 0.0364, "step": 2726 }, { "epoch": 1.07, "learning_rate": 3.937397246671985e-07, "loss": 0.0343, "step": 2727 }, { "epoch": 1.08, "learning_rate": 3.937333847577752e-07, "loss": 0.0372, "step": 2728 }, { "epoch": 1.08, "learning_rate": 3.9372704169079214e-07, "loss": 0.0354, "step": 2729 }, { "epoch": 1.08, "learning_rate": 3.937206954663527e-07, "loss": 0.0427, "step": 2730 }, { "epoch": 1.08, "learning_rate": 3.9371434608456045e-07, "loss": 0.0362, "step": 2731 }, { "epoch": 1.08, "learning_rate": 3.937079935455187e-07, "loss": 0.0365, "step": 2732 }, { "epoch": 1.08, "learning_rate": 3.9370163784933105e-07, "loss": 0.0409, "step": 2733 }, { "epoch": 1.08, "learning_rate": 3.9369527899610107e-07, "loss": 0.037, "step": 2734 }, { "epoch": 1.08, "learning_rate": 3.9368891698593255e-07, "loss": 0.0385, "step": 2735 }, { "epoch": 1.08, "learning_rate": 3.9368255181892893e-07, "loss": 0.0336, "step": 2736 }, { "epoch": 1.08, "learning_rate": 3.9367618349519414e-07, "loss": 0.0367, "step": 2737 }, { "epoch": 1.08, "learning_rate": 3.936698120148319e-07, "loss": 0.0363, "step": 2738 }, { "epoch": 1.08, "learning_rate": 3.93663437377946e-07, "loss": 0.0374, "step": 2739 }, { "epoch": 1.08, "learning_rate": 3.9365705958464053e-07, "loss": 0.0358, "step": 2740 }, { "epoch": 1.08, "learning_rate": 3.9365067863501923e-07, "loss": 0.0367, "step": 2741 }, { "epoch": 1.08, "learning_rate": 3.936442945291862e-07, "loss": 0.0348, "step": 2742 }, { "epoch": 1.08, "learning_rate": 3.936379072672455e-07, "loss": 0.0349, "step": 2743 }, { "epoch": 1.08, "learning_rate": 3.936315168493012e-07, "loss": 0.0361, "step": 2744 }, { "epoch": 1.08, "learning_rate": 3.936251232754574e-07, "loss": 0.0341, "step": 2745 }, { "epoch": 1.08, "learning_rate": 3.9361872654581843e-07, "loss": 0.0363, "step": 2746 }, { "epoch": 1.08, "learning_rate": 3.936123266604885e-07, "loss": 0.0354, "step": 2747 }, { "epoch": 1.08, "learning_rate": 3.9360592361957185e-07, "loss": 0.0368, "step": 2748 }, { "epoch": 1.08, "learning_rate": 3.935995174231729e-07, "loss": 0.0374, "step": 2749 }, { "epoch": 1.08, "learning_rate": 3.9359310807139605e-07, "loss": 0.0365, "step": 2750 }, { "epoch": 1.08, "learning_rate": 3.935866955643458e-07, "loss": 0.0337, "step": 2751 }, { "epoch": 1.08, "learning_rate": 3.9358027990212663e-07, "loss": 0.0357, "step": 2752 }, { "epoch": 1.08, "learning_rate": 3.935738610848431e-07, "loss": 0.0306, "step": 2753 }, { "epoch": 1.09, "learning_rate": 3.9356743911259977e-07, "loss": 0.0377, "step": 2754 }, { "epoch": 1.09, "learning_rate": 3.9356101398550145e-07, "loss": 0.0356, "step": 2755 }, { "epoch": 1.09, "learning_rate": 3.9355458570365275e-07, "loss": 0.0405, "step": 2756 }, { "epoch": 1.09, "learning_rate": 3.935481542671585e-07, "loss": 0.0344, "step": 2757 }, { "epoch": 1.09, "learning_rate": 3.9354171967612346e-07, "loss": 0.0391, "step": 2758 }, { "epoch": 1.09, "learning_rate": 3.9353528193065257e-07, "loss": 0.0345, "step": 2759 }, { "epoch": 1.09, "learning_rate": 3.9352884103085067e-07, "loss": 0.0402, "step": 2760 }, { "epoch": 1.09, "learning_rate": 3.9352239697682285e-07, "loss": 0.0335, "step": 2761 }, { "epoch": 1.09, "learning_rate": 3.93515949768674e-07, "loss": 0.0398, "step": 2762 }, { "epoch": 1.09, "learning_rate": 3.9350949940650934e-07, "loss": 0.0397, "step": 2763 }, { "epoch": 1.09, "learning_rate": 3.9350304589043397e-07, "loss": 0.037, "step": 2764 }, { "epoch": 1.09, "learning_rate": 3.9349658922055296e-07, "loss": 0.0416, "step": 2765 }, { "epoch": 1.09, "learning_rate": 3.934901293969717e-07, "loss": 0.0347, "step": 2766 }, { "epoch": 1.09, "learning_rate": 3.934836664197954e-07, "loss": 0.0344, "step": 2767 }, { "epoch": 1.09, "learning_rate": 3.934772002891294e-07, "loss": 0.037, "step": 2768 }, { "epoch": 1.09, "learning_rate": 3.93470731005079e-07, "loss": 0.0336, "step": 2769 }, { "epoch": 1.09, "learning_rate": 3.934642585677498e-07, "loss": 0.0375, "step": 2770 }, { "epoch": 1.09, "learning_rate": 3.934577829772472e-07, "loss": 0.0433, "step": 2771 }, { "epoch": 1.09, "learning_rate": 3.934513042336768e-07, "loss": 0.0388, "step": 2772 }, { "epoch": 1.09, "learning_rate": 3.934448223371441e-07, "loss": 0.0297, "step": 2773 }, { "epoch": 1.09, "learning_rate": 3.934383372877548e-07, "loss": 0.0376, "step": 2774 }, { "epoch": 1.09, "learning_rate": 3.934318490856146e-07, "loss": 0.0313, "step": 2775 }, { "epoch": 1.09, "learning_rate": 3.934253577308293e-07, "loss": 0.0397, "step": 2776 }, { "epoch": 1.09, "learning_rate": 3.934188632235045e-07, "loss": 0.0337, "step": 2777 }, { "epoch": 1.09, "learning_rate": 3.9341236556374635e-07, "loss": 0.0335, "step": 2778 }, { "epoch": 1.1, "learning_rate": 3.934058647516605e-07, "loss": 0.0354, "step": 2779 }, { "epoch": 1.1, "learning_rate": 3.93399360787353e-07, "loss": 0.0372, "step": 2780 }, { "epoch": 1.1, "learning_rate": 3.933928536709299e-07, "loss": 0.036, "step": 2781 }, { "epoch": 1.1, "learning_rate": 3.9338634340249717e-07, "loss": 0.0329, "step": 2782 }, { "epoch": 1.1, "learning_rate": 3.9337982998216097e-07, "loss": 0.0324, "step": 2783 }, { "epoch": 1.1, "learning_rate": 3.933733134100275e-07, "loss": 0.0322, "step": 2784 }, { "epoch": 1.1, "learning_rate": 3.933667936862028e-07, "loss": 0.0356, "step": 2785 }, { "epoch": 1.1, "learning_rate": 3.9336027081079333e-07, "loss": 0.0377, "step": 2786 }, { "epoch": 1.1, "learning_rate": 3.933537447839053e-07, "loss": 0.0344, "step": 2787 }, { "epoch": 1.1, "learning_rate": 3.9334721560564515e-07, "loss": 0.0417, "step": 2788 }, { "epoch": 1.1, "learning_rate": 3.933406832761192e-07, "loss": 0.0388, "step": 2789 }, { "epoch": 1.1, "learning_rate": 3.9333414779543387e-07, "loss": 0.0324, "step": 2790 }, { "epoch": 1.1, "learning_rate": 3.9332760916369584e-07, "loss": 0.033, "step": 2791 }, { "epoch": 1.1, "learning_rate": 3.933210673810116e-07, "loss": 0.032, "step": 2792 }, { "epoch": 1.1, "learning_rate": 3.933145224474878e-07, "loss": 0.0349, "step": 2793 }, { "epoch": 1.1, "learning_rate": 3.9330797436323105e-07, "loss": 0.0357, "step": 2794 }, { "epoch": 1.1, "learning_rate": 3.9330142312834813e-07, "loss": 0.0359, "step": 2795 }, { "epoch": 1.1, "learning_rate": 3.9329486874294575e-07, "loss": 0.0363, "step": 2796 }, { "epoch": 1.1, "learning_rate": 3.932883112071309e-07, "loss": 0.0343, "step": 2797 }, { "epoch": 1.1, "learning_rate": 3.932817505210102e-07, "loss": 0.0354, "step": 2798 }, { "epoch": 1.1, "learning_rate": 3.932751866846908e-07, "loss": 0.0373, "step": 2799 }, { "epoch": 1.1, "learning_rate": 3.932686196982796e-07, "loss": 0.035, "step": 2800 }, { "epoch": 1.1, "learning_rate": 3.9326204956188363e-07, "loss": 0.036, "step": 2801 }, { "epoch": 1.1, "learning_rate": 3.9325547627561e-07, "loss": 0.0358, "step": 2802 }, { "epoch": 1.1, "learning_rate": 3.9324889983956575e-07, "loss": 0.0413, "step": 2803 }, { "epoch": 1.11, "learning_rate": 3.932423202538582e-07, "loss": 0.035, "step": 2804 }, { "epoch": 1.11, "learning_rate": 3.932357375185945e-07, "loss": 0.0354, "step": 2805 }, { "epoch": 1.11, "learning_rate": 3.9322915163388195e-07, "loss": 0.0371, "step": 2806 }, { "epoch": 1.11, "learning_rate": 3.9322256259982797e-07, "loss": 0.0335, "step": 2807 }, { "epoch": 1.11, "learning_rate": 3.932159704165398e-07, "loss": 0.0356, "step": 2808 }, { "epoch": 1.11, "learning_rate": 3.93209375084125e-07, "loss": 0.0448, "step": 2809 }, { "epoch": 1.11, "learning_rate": 3.93202776602691e-07, "loss": 0.0383, "step": 2810 }, { "epoch": 1.11, "learning_rate": 3.9319617497234543e-07, "loss": 0.0372, "step": 2811 }, { "epoch": 1.11, "learning_rate": 3.931895701931958e-07, "loss": 0.0396, "step": 2812 }, { "epoch": 1.11, "learning_rate": 3.931829622653498e-07, "loss": 0.0398, "step": 2813 }, { "epoch": 1.11, "learning_rate": 3.9317635118891516e-07, "loss": 0.0362, "step": 2814 }, { "epoch": 1.11, "learning_rate": 3.9316973696399953e-07, "loss": 0.0359, "step": 2815 }, { "epoch": 1.11, "learning_rate": 3.9316311959071076e-07, "loss": 0.0372, "step": 2816 }, { "epoch": 1.11, "learning_rate": 3.9315649906915676e-07, "loss": 0.0375, "step": 2817 }, { "epoch": 1.11, "learning_rate": 3.9314987539944535e-07, "loss": 0.0428, "step": 2818 }, { "epoch": 1.11, "learning_rate": 3.9314324858168454e-07, "loss": 0.03, "step": 2819 }, { "epoch": 1.11, "learning_rate": 3.9313661861598237e-07, "loss": 0.0342, "step": 2820 }, { "epoch": 1.11, "learning_rate": 3.931299855024468e-07, "loss": 0.0346, "step": 2821 }, { "epoch": 1.11, "learning_rate": 3.93123349241186e-07, "loss": 0.0339, "step": 2822 }, { "epoch": 1.11, "learning_rate": 3.931167098323081e-07, "loss": 0.0346, "step": 2823 }, { "epoch": 1.11, "learning_rate": 3.931100672759214e-07, "loss": 0.0372, "step": 2824 }, { "epoch": 1.11, "learning_rate": 3.9310342157213403e-07, "loss": 0.0328, "step": 2825 }, { "epoch": 1.11, "learning_rate": 3.930967727210544e-07, "loss": 0.0344, "step": 2826 }, { "epoch": 1.11, "learning_rate": 3.930901207227908e-07, "loss": 0.0396, "step": 2827 }, { "epoch": 1.11, "learning_rate": 3.9308346557745173e-07, "loss": 0.0347, "step": 2828 }, { "epoch": 1.11, "learning_rate": 3.930768072851457e-07, "loss": 0.0398, "step": 2829 }, { "epoch": 1.12, "learning_rate": 3.9307014584598105e-07, "loss": 0.0319, "step": 2830 }, { "epoch": 1.12, "learning_rate": 3.9306348126006646e-07, "loss": 0.0359, "step": 2831 }, { "epoch": 1.12, "learning_rate": 3.930568135275106e-07, "loss": 0.0354, "step": 2832 }, { "epoch": 1.12, "learning_rate": 3.9305014264842207e-07, "loss": 0.034, "step": 2833 }, { "epoch": 1.12, "learning_rate": 3.9304346862290963e-07, "loss": 0.0386, "step": 2834 }, { "epoch": 1.12, "learning_rate": 3.9303679145108203e-07, "loss": 0.0396, "step": 2835 }, { "epoch": 1.12, "learning_rate": 3.930301111330481e-07, "loss": 0.0408, "step": 2836 }, { "epoch": 1.12, "learning_rate": 3.9302342766891674e-07, "loss": 0.0339, "step": 2837 }, { "epoch": 1.12, "learning_rate": 3.930167410587969e-07, "loss": 0.0357, "step": 2838 }, { "epoch": 1.12, "learning_rate": 3.930100513027975e-07, "loss": 0.0348, "step": 2839 }, { "epoch": 1.12, "learning_rate": 3.9300335840102763e-07, "loss": 0.0303, "step": 2840 }, { "epoch": 1.12, "learning_rate": 3.929966623535963e-07, "loss": 0.0373, "step": 2841 }, { "epoch": 1.12, "learning_rate": 3.929899631606127e-07, "loss": 0.029, "step": 2842 }, { "epoch": 1.12, "learning_rate": 3.9298326082218607e-07, "loss": 0.0357, "step": 2843 }, { "epoch": 1.12, "learning_rate": 3.929765553384255e-07, "loss": 0.0378, "step": 2844 }, { "epoch": 1.12, "learning_rate": 3.9296984670944047e-07, "loss": 0.0426, "step": 2845 }, { "epoch": 1.12, "learning_rate": 3.929631349353402e-07, "loss": 0.0373, "step": 2846 }, { "epoch": 1.12, "learning_rate": 3.929564200162341e-07, "loss": 0.0353, "step": 2847 }, { "epoch": 1.12, "learning_rate": 3.9294970195223154e-07, "loss": 0.0376, "step": 2848 }, { "epoch": 1.12, "learning_rate": 3.929429807434421e-07, "loss": 0.0353, "step": 2849 }, { "epoch": 1.12, "learning_rate": 3.929362563899754e-07, "loss": 0.0384, "step": 2850 }, { "epoch": 1.12, "learning_rate": 3.929295288919409e-07, "loss": 0.0373, "step": 2851 }, { "epoch": 1.12, "learning_rate": 3.929227982494483e-07, "loss": 0.0421, "step": 2852 }, { "epoch": 1.12, "learning_rate": 3.9291606446260727e-07, "loss": 0.0367, "step": 2853 }, { "epoch": 1.12, "learning_rate": 3.929093275315276e-07, "loss": 0.0347, "step": 2854 }, { "epoch": 1.13, "learning_rate": 3.9290258745631914e-07, "loss": 0.0392, "step": 2855 }, { "epoch": 1.13, "learning_rate": 3.9289584423709166e-07, "loss": 0.0341, "step": 2856 }, { "epoch": 1.13, "learning_rate": 3.9288909787395507e-07, "loss": 0.0343, "step": 2857 }, { "epoch": 1.13, "learning_rate": 3.928823483670193e-07, "loss": 0.0343, "step": 2858 }, { "epoch": 1.13, "learning_rate": 3.9287559571639444e-07, "loss": 0.0361, "step": 2859 }, { "epoch": 1.13, "learning_rate": 3.9286883992219047e-07, "loss": 0.0416, "step": 2860 }, { "epoch": 1.13, "learning_rate": 3.9286208098451763e-07, "loss": 0.035, "step": 2861 }, { "epoch": 1.13, "learning_rate": 3.928553189034859e-07, "loss": 0.0337, "step": 2862 }, { "epoch": 1.13, "learning_rate": 3.9284855367920556e-07, "loss": 0.0337, "step": 2863 }, { "epoch": 1.13, "learning_rate": 3.9284178531178693e-07, "loss": 0.0372, "step": 2864 }, { "epoch": 1.13, "learning_rate": 3.928350138013403e-07, "loss": 0.0346, "step": 2865 }, { "epoch": 1.13, "learning_rate": 3.9282823914797604e-07, "loss": 0.0318, "step": 2866 }, { "epoch": 1.13, "learning_rate": 3.928214613518045e-07, "loss": 0.0325, "step": 2867 }, { "epoch": 1.13, "learning_rate": 3.928146804129362e-07, "loss": 0.0378, "step": 2868 }, { "epoch": 1.13, "learning_rate": 3.9280789633148166e-07, "loss": 0.0391, "step": 2869 }, { "epoch": 1.13, "learning_rate": 3.9280110910755145e-07, "loss": 0.0423, "step": 2870 }, { "epoch": 1.13, "learning_rate": 3.9279431874125617e-07, "loss": 0.035, "step": 2871 }, { "epoch": 1.13, "learning_rate": 3.9278752523270656e-07, "loss": 0.0335, "step": 2872 }, { "epoch": 1.13, "learning_rate": 3.927807285820132e-07, "loss": 0.034, "step": 2873 }, { "epoch": 1.13, "learning_rate": 3.9277392878928706e-07, "loss": 0.0386, "step": 2874 }, { "epoch": 1.13, "learning_rate": 3.9276712585463876e-07, "loss": 0.0312, "step": 2875 }, { "epoch": 1.13, "learning_rate": 3.927603197781794e-07, "loss": 0.0369, "step": 2876 }, { "epoch": 1.13, "learning_rate": 3.927535105600197e-07, "loss": 0.0329, "step": 2877 }, { "epoch": 1.13, "learning_rate": 3.9274669820027073e-07, "loss": 0.0345, "step": 2878 }, { "epoch": 1.13, "learning_rate": 3.927398826990436e-07, "loss": 0.0407, "step": 2879 }, { "epoch": 1.14, "learning_rate": 3.9273306405644923e-07, "loss": 0.0339, "step": 2880 }, { "epoch": 1.14, "learning_rate": 3.927262422725989e-07, "loss": 0.0394, "step": 2881 }, { "epoch": 1.14, "learning_rate": 3.927194173476037e-07, "loss": 0.0353, "step": 2882 }, { "epoch": 1.14, "learning_rate": 3.9271258928157487e-07, "loss": 0.038, "step": 2883 }, { "epoch": 1.14, "learning_rate": 3.927057580746237e-07, "loss": 0.0365, "step": 2884 }, { "epoch": 1.14, "learning_rate": 3.926989237268616e-07, "loss": 0.0359, "step": 2885 }, { "epoch": 1.14, "learning_rate": 3.9269208623839997e-07, "loss": 0.0345, "step": 2886 }, { "epoch": 1.14, "learning_rate": 3.9268524560935006e-07, "loss": 0.0379, "step": 2887 }, { "epoch": 1.14, "learning_rate": 3.926784018398236e-07, "loss": 0.0363, "step": 2888 }, { "epoch": 1.14, "learning_rate": 3.92671554929932e-07, "loss": 0.0397, "step": 2889 }, { "epoch": 1.14, "learning_rate": 3.9266470487978685e-07, "loss": 0.0379, "step": 2890 }, { "epoch": 1.14, "learning_rate": 3.9265785168949986e-07, "loss": 0.0367, "step": 2891 }, { "epoch": 1.14, "learning_rate": 3.926509953591827e-07, "loss": 0.0391, "step": 2892 }, { "epoch": 1.14, "learning_rate": 3.926441358889471e-07, "loss": 0.0346, "step": 2893 }, { "epoch": 1.14, "learning_rate": 3.926372732789049e-07, "loss": 0.0343, "step": 2894 }, { "epoch": 1.14, "learning_rate": 3.9263040752916783e-07, "loss": 0.0325, "step": 2895 }, { "epoch": 1.14, "learning_rate": 3.92623538639848e-07, "loss": 0.0371, "step": 2896 }, { "epoch": 1.14, "learning_rate": 3.9261666661105715e-07, "loss": 0.0322, "step": 2897 }, { "epoch": 1.14, "learning_rate": 3.926097914429074e-07, "loss": 0.0355, "step": 2898 }, { "epoch": 1.14, "learning_rate": 3.9260291313551086e-07, "loss": 0.0376, "step": 2899 }, { "epoch": 1.14, "learning_rate": 3.9259603168897947e-07, "loss": 0.0354, "step": 2900 }, { "epoch": 1.14, "learning_rate": 3.925891471034255e-07, "loss": 0.0343, "step": 2901 }, { "epoch": 1.14, "learning_rate": 3.9258225937896114e-07, "loss": 0.033, "step": 2902 }, { "epoch": 1.14, "learning_rate": 3.925753685156986e-07, "loss": 0.0379, "step": 2903 }, { "epoch": 1.14, "learning_rate": 3.925684745137503e-07, "loss": 0.0341, "step": 2904 }, { "epoch": 1.14, "learning_rate": 3.9256157737322846e-07, "loss": 0.0375, "step": 2905 }, { "epoch": 1.15, "learning_rate": 3.925546770942457e-07, "loss": 0.039, "step": 2906 }, { "epoch": 1.15, "learning_rate": 3.9254777367691424e-07, "loss": 0.0304, "step": 2907 }, { "epoch": 1.15, "learning_rate": 3.9254086712134675e-07, "loss": 0.0363, "step": 2908 }, { "epoch": 1.15, "learning_rate": 3.925339574276557e-07, "loss": 0.0373, "step": 2909 }, { "epoch": 1.15, "learning_rate": 3.925270445959538e-07, "loss": 0.0334, "step": 2910 }, { "epoch": 1.15, "learning_rate": 3.925201286263537e-07, "loss": 0.0364, "step": 2911 }, { "epoch": 1.15, "learning_rate": 3.925132095189681e-07, "loss": 0.0331, "step": 2912 }, { "epoch": 1.15, "learning_rate": 3.9250628727390975e-07, "loss": 0.0406, "step": 2913 }, { "epoch": 1.15, "learning_rate": 3.924993618912915e-07, "loss": 0.0343, "step": 2914 }, { "epoch": 1.15, "learning_rate": 3.924924333712262e-07, "loss": 0.0326, "step": 2915 }, { "epoch": 1.15, "learning_rate": 3.9248550171382683e-07, "loss": 0.0335, "step": 2916 }, { "epoch": 1.15, "learning_rate": 3.924785669192063e-07, "loss": 0.0322, "step": 2917 }, { "epoch": 1.15, "learning_rate": 3.924716289874777e-07, "loss": 0.0359, "step": 2918 }, { "epoch": 1.15, "learning_rate": 3.9246468791875394e-07, "loss": 0.0358, "step": 2919 }, { "epoch": 1.15, "learning_rate": 3.924577437131484e-07, "loss": 0.0397, "step": 2920 }, { "epoch": 1.15, "learning_rate": 3.9245079637077414e-07, "loss": 0.0331, "step": 2921 }, { "epoch": 1.15, "learning_rate": 3.924438458917444e-07, "loss": 0.0355, "step": 2922 }, { "epoch": 1.15, "learning_rate": 3.924368922761724e-07, "loss": 0.0361, "step": 2923 }, { "epoch": 1.15, "learning_rate": 3.924299355241715e-07, "loss": 0.0353, "step": 2924 }, { "epoch": 1.15, "learning_rate": 3.924229756358552e-07, "loss": 0.0353, "step": 2925 }, { "epoch": 1.15, "learning_rate": 3.9241601261133677e-07, "loss": 0.0349, "step": 2926 }, { "epoch": 1.15, "learning_rate": 3.924090464507298e-07, "loss": 0.0329, "step": 2927 }, { "epoch": 1.15, "learning_rate": 3.9240207715414783e-07, "loss": 0.0322, "step": 2928 }, { "epoch": 1.15, "learning_rate": 3.923951047217044e-07, "loss": 0.0349, "step": 2929 }, { "epoch": 1.15, "learning_rate": 3.9238812915351315e-07, "loss": 0.0316, "step": 2930 }, { "epoch": 1.16, "learning_rate": 3.923811504496878e-07, "loss": 0.0358, "step": 2931 }, { "epoch": 1.16, "learning_rate": 3.923741686103421e-07, "loss": 0.0296, "step": 2932 }, { "epoch": 1.16, "learning_rate": 3.9236718363558976e-07, "loss": 0.0332, "step": 2933 }, { "epoch": 1.16, "learning_rate": 3.923601955255448e-07, "loss": 0.0278, "step": 2934 }, { "epoch": 1.16, "learning_rate": 3.9235320428032097e-07, "loss": 0.0361, "step": 2935 }, { "epoch": 1.16, "learning_rate": 3.923462099000322e-07, "loss": 0.0343, "step": 2936 }, { "epoch": 1.16, "learning_rate": 3.923392123847926e-07, "loss": 0.0317, "step": 2937 }, { "epoch": 1.16, "learning_rate": 3.9233221173471616e-07, "loss": 0.039, "step": 2938 }, { "epoch": 1.16, "learning_rate": 3.9232520794991695e-07, "loss": 0.0334, "step": 2939 }, { "epoch": 1.16, "learning_rate": 3.9231820103050917e-07, "loss": 0.0358, "step": 2940 }, { "epoch": 1.16, "learning_rate": 3.92311190976607e-07, "loss": 0.0397, "step": 2941 }, { "epoch": 1.16, "learning_rate": 3.9230417778832467e-07, "loss": 0.0338, "step": 2942 }, { "epoch": 1.16, "learning_rate": 3.922971614657765e-07, "loss": 0.0344, "step": 2943 }, { "epoch": 1.16, "learning_rate": 3.922901420090769e-07, "loss": 0.0345, "step": 2944 }, { "epoch": 1.16, "learning_rate": 3.922831194183402e-07, "loss": 0.0398, "step": 2945 }, { "epoch": 1.16, "learning_rate": 3.922760936936809e-07, "loss": 0.0367, "step": 2946 }, { "epoch": 1.16, "learning_rate": 3.9226906483521355e-07, "loss": 0.0351, "step": 2947 }, { "epoch": 1.16, "learning_rate": 3.9226203284305264e-07, "loss": 0.0332, "step": 2948 }, { "epoch": 1.16, "learning_rate": 3.922549977173127e-07, "loss": 0.0364, "step": 2949 }, { "epoch": 1.16, "learning_rate": 3.922479594581086e-07, "loss": 0.0306, "step": 2950 }, { "epoch": 1.16, "learning_rate": 3.922409180655549e-07, "loss": 0.0347, "step": 2951 }, { "epoch": 1.16, "learning_rate": 3.9223387353976643e-07, "loss": 0.0328, "step": 2952 }, { "epoch": 1.16, "learning_rate": 3.9222682588085797e-07, "loss": 0.0355, "step": 2953 }, { "epoch": 1.16, "learning_rate": 3.922197750889444e-07, "loss": 0.0332, "step": 2954 }, { "epoch": 1.16, "learning_rate": 3.9221272116414065e-07, "loss": 0.0305, "step": 2955 }, { "epoch": 1.16, "learning_rate": 3.922056641065617e-07, "loss": 0.0415, "step": 2956 }, { "epoch": 1.17, "learning_rate": 3.921986039163225e-07, "loss": 0.0327, "step": 2957 }, { "epoch": 1.17, "learning_rate": 3.9219154059353816e-07, "loss": 0.0343, "step": 2958 }, { "epoch": 1.17, "learning_rate": 3.9218447413832385e-07, "loss": 0.0321, "step": 2959 }, { "epoch": 1.17, "learning_rate": 3.921774045507946e-07, "loss": 0.0364, "step": 2960 }, { "epoch": 1.17, "learning_rate": 3.9217033183106583e-07, "loss": 0.0355, "step": 2961 }, { "epoch": 1.17, "learning_rate": 3.9216325597925266e-07, "loss": 0.0379, "step": 2962 }, { "epoch": 1.17, "learning_rate": 3.921561769954705e-07, "loss": 0.0386, "step": 2963 }, { "epoch": 1.17, "learning_rate": 3.921490948798347e-07, "loss": 0.035, "step": 2964 }, { "epoch": 1.17, "learning_rate": 3.9214200963246065e-07, "loss": 0.0361, "step": 2965 }, { "epoch": 1.17, "learning_rate": 3.9213492125346393e-07, "loss": 0.0359, "step": 2966 }, { "epoch": 1.17, "learning_rate": 3.9212782974296e-07, "loss": 0.0356, "step": 2967 }, { "epoch": 1.17, "learning_rate": 3.921207351010644e-07, "loss": 0.0353, "step": 2968 }, { "epoch": 1.17, "learning_rate": 3.921136373278928e-07, "loss": 0.035, "step": 2969 }, { "epoch": 1.17, "learning_rate": 3.921065364235609e-07, "loss": 0.0363, "step": 2970 }, { "epoch": 1.17, "learning_rate": 3.920994323881845e-07, "loss": 0.0312, "step": 2971 }, { "epoch": 1.17, "learning_rate": 3.920923252218792e-07, "loss": 0.0344, "step": 2972 }, { "epoch": 1.17, "learning_rate": 3.9208521492476094e-07, "loss": 0.033, "step": 2973 }, { "epoch": 1.17, "learning_rate": 3.9207810149694575e-07, "loss": 0.0392, "step": 2974 }, { "epoch": 1.17, "learning_rate": 3.920709849385493e-07, "loss": 0.0354, "step": 2975 }, { "epoch": 1.17, "learning_rate": 3.920638652496878e-07, "loss": 0.0385, "step": 2976 }, { "epoch": 1.17, "learning_rate": 3.9205674243047714e-07, "loss": 0.0334, "step": 2977 }, { "epoch": 1.17, "learning_rate": 3.920496164810335e-07, "loss": 0.0309, "step": 2978 }, { "epoch": 1.17, "learning_rate": 3.9204248740147295e-07, "loss": 0.036, "step": 2979 }, { "epoch": 1.17, "learning_rate": 3.9203535519191177e-07, "loss": 0.0351, "step": 2980 }, { "epoch": 1.17, "learning_rate": 3.920282198524662e-07, "loss": 0.0339, "step": 2981 }, { "epoch": 1.18, "learning_rate": 3.920210813832524e-07, "loss": 0.0344, "step": 2982 }, { "epoch": 1.18, "learning_rate": 3.9201393978438684e-07, "loss": 0.0402, "step": 2983 }, { "epoch": 1.18, "learning_rate": 3.920067950559859e-07, "loss": 0.0345, "step": 2984 }, { "epoch": 1.18, "learning_rate": 3.9199964719816606e-07, "loss": 0.036, "step": 2985 }, { "epoch": 1.18, "learning_rate": 3.919924962110437e-07, "loss": 0.0372, "step": 2986 }, { "epoch": 1.18, "learning_rate": 3.919853420947355e-07, "loss": 0.0373, "step": 2987 }, { "epoch": 1.18, "learning_rate": 3.91978184849358e-07, "loss": 0.0394, "step": 2988 }, { "epoch": 1.18, "learning_rate": 3.9197102447502787e-07, "loss": 0.0343, "step": 2989 }, { "epoch": 1.18, "learning_rate": 3.919638609718618e-07, "loss": 0.0378, "step": 2990 }, { "epoch": 1.18, "learning_rate": 3.9195669433997647e-07, "loss": 0.0382, "step": 2991 }, { "epoch": 1.18, "learning_rate": 3.919495245794889e-07, "loss": 0.0394, "step": 2992 }, { "epoch": 1.18, "learning_rate": 3.9194235169051566e-07, "loss": 0.0396, "step": 2993 }, { "epoch": 1.18, "learning_rate": 3.919351756731739e-07, "loss": 0.0313, "step": 2994 }, { "epoch": 1.18, "learning_rate": 3.919279965275805e-07, "loss": 0.0315, "step": 2995 }, { "epoch": 1.18, "learning_rate": 3.919208142538524e-07, "loss": 0.0314, "step": 2996 }, { "epoch": 1.18, "learning_rate": 3.919136288521067e-07, "loss": 0.0369, "step": 2997 }, { "epoch": 1.18, "learning_rate": 3.919064403224606e-07, "loss": 0.0365, "step": 2998 }, { "epoch": 1.18, "learning_rate": 3.918992486650311e-07, "loss": 0.0309, "step": 2999 }, { "epoch": 1.18, "learning_rate": 3.918920538799356e-07, "loss": 0.033, "step": 3000 }, { "epoch": 1.18, "learning_rate": 3.918848559672912e-07, "loss": 0.0375, "step": 3001 }, { "epoch": 1.18, "learning_rate": 3.918776549272153e-07, "loss": 0.0332, "step": 3002 }, { "epoch": 1.18, "learning_rate": 3.918704507598252e-07, "loss": 0.0311, "step": 3003 }, { "epoch": 1.18, "learning_rate": 3.9186324346523843e-07, "loss": 0.0348, "step": 3004 }, { "epoch": 1.18, "learning_rate": 3.9185603304357235e-07, "loss": 0.0377, "step": 3005 }, { "epoch": 1.18, "learning_rate": 3.9184881949494455e-07, "loss": 0.0324, "step": 3006 }, { "epoch": 1.19, "learning_rate": 3.9184160281947254e-07, "loss": 0.0329, "step": 3007 }, { "epoch": 1.19, "learning_rate": 3.918343830172739e-07, "loss": 0.0397, "step": 3008 }, { "epoch": 1.19, "learning_rate": 3.918271600884665e-07, "loss": 0.0354, "step": 3009 }, { "epoch": 1.19, "learning_rate": 3.9181993403316785e-07, "loss": 0.0359, "step": 3010 }, { "epoch": 1.19, "learning_rate": 3.918127048514959e-07, "loss": 0.0338, "step": 3011 }, { "epoch": 1.19, "learning_rate": 3.9180547254356825e-07, "loss": 0.0388, "step": 3012 }, { "epoch": 1.19, "learning_rate": 3.9179823710950305e-07, "loss": 0.0319, "step": 3013 }, { "epoch": 1.19, "learning_rate": 3.91790998549418e-07, "loss": 0.0392, "step": 3014 }, { "epoch": 1.19, "learning_rate": 3.9178375686343113e-07, "loss": 0.0386, "step": 3015 }, { "epoch": 1.19, "learning_rate": 3.9177651205166057e-07, "loss": 0.0323, "step": 3016 }, { "epoch": 1.19, "learning_rate": 3.917692641142243e-07, "loss": 0.0353, "step": 3017 }, { "epoch": 1.19, "learning_rate": 3.917620130512405e-07, "loss": 0.0322, "step": 3018 }, { "epoch": 1.19, "learning_rate": 3.917547588628273e-07, "loss": 0.0411, "step": 3019 }, { "epoch": 1.19, "learning_rate": 3.91747501549103e-07, "loss": 0.0319, "step": 3020 }, { "epoch": 1.19, "learning_rate": 3.917402411101858e-07, "loss": 0.0356, "step": 3021 }, { "epoch": 1.19, "learning_rate": 3.9173297754619414e-07, "loss": 0.0356, "step": 3022 }, { "epoch": 1.19, "learning_rate": 3.917257108572463e-07, "loss": 0.0322, "step": 3023 }, { "epoch": 1.19, "learning_rate": 3.9171844104346076e-07, "loss": 0.0362, "step": 3024 }, { "epoch": 1.19, "learning_rate": 3.9171116810495607e-07, "loss": 0.0357, "step": 3025 }, { "epoch": 1.19, "learning_rate": 3.9170389204185066e-07, "loss": 0.0334, "step": 3026 }, { "epoch": 1.19, "learning_rate": 3.916966128542632e-07, "loss": 0.0325, "step": 3027 }, { "epoch": 1.19, "learning_rate": 3.9168933054231225e-07, "loss": 0.0347, "step": 3028 }, { "epoch": 1.19, "learning_rate": 3.9168204510611653e-07, "loss": 0.0362, "step": 3029 }, { "epoch": 1.19, "learning_rate": 3.9167475654579484e-07, "loss": 0.0355, "step": 3030 }, { "epoch": 1.19, "learning_rate": 3.9166746486146594e-07, "loss": 0.0281, "step": 3031 }, { "epoch": 1.19, "learning_rate": 3.916601700532487e-07, "loss": 0.0297, "step": 3032 }, { "epoch": 1.2, "learning_rate": 3.916528721212619e-07, "loss": 0.0346, "step": 3033 }, { "epoch": 1.2, "learning_rate": 3.916455710656246e-07, "loss": 0.0346, "step": 3034 }, { "epoch": 1.2, "learning_rate": 3.9163826688645575e-07, "loss": 0.0359, "step": 3035 }, { "epoch": 1.2, "learning_rate": 3.916309595838744e-07, "loss": 0.0337, "step": 3036 }, { "epoch": 1.2, "learning_rate": 3.9162364915799966e-07, "loss": 0.0347, "step": 3037 }, { "epoch": 1.2, "learning_rate": 3.916163356089507e-07, "loss": 0.0371, "step": 3038 }, { "epoch": 1.2, "learning_rate": 3.9160901893684664e-07, "loss": 0.0338, "step": 3039 }, { "epoch": 1.2, "learning_rate": 3.9160169914180677e-07, "loss": 0.0314, "step": 3040 }, { "epoch": 1.2, "learning_rate": 3.9159437622395045e-07, "loss": 0.0371, "step": 3041 }, { "epoch": 1.2, "learning_rate": 3.9158705018339695e-07, "loss": 0.0342, "step": 3042 }, { "epoch": 1.2, "learning_rate": 3.915797210202658e-07, "loss": 0.0404, "step": 3043 }, { "epoch": 1.2, "learning_rate": 3.915723887346763e-07, "loss": 0.0336, "step": 3044 }, { "epoch": 1.2, "learning_rate": 3.91565053326748e-07, "loss": 0.0363, "step": 3045 }, { "epoch": 1.2, "learning_rate": 3.915577147966005e-07, "loss": 0.038, "step": 3046 }, { "epoch": 1.2, "learning_rate": 3.9155037314435343e-07, "loss": 0.0307, "step": 3047 }, { "epoch": 1.2, "learning_rate": 3.9154302837012634e-07, "loss": 0.036, "step": 3048 }, { "epoch": 1.2, "learning_rate": 3.91535680474039e-07, "loss": 0.0317, "step": 3049 }, { "epoch": 1.2, "learning_rate": 3.915283294562112e-07, "loss": 0.0391, "step": 3050 }, { "epoch": 1.2, "learning_rate": 3.9152097531676273e-07, "loss": 0.0367, "step": 3051 }, { "epoch": 1.2, "learning_rate": 3.915136180558134e-07, "loss": 0.0327, "step": 3052 }, { "epoch": 1.2, "learning_rate": 3.915062576734832e-07, "loss": 0.0296, "step": 3053 }, { "epoch": 1.2, "learning_rate": 3.9149889416989204e-07, "loss": 0.0375, "step": 3054 }, { "epoch": 1.2, "learning_rate": 3.914915275451599e-07, "loss": 0.0325, "step": 3055 }, { "epoch": 1.2, "learning_rate": 3.9148415779940693e-07, "loss": 0.0355, "step": 3056 }, { "epoch": 1.2, "learning_rate": 3.9147678493275325e-07, "loss": 0.032, "step": 3057 }, { "epoch": 1.21, "learning_rate": 3.9146940894531897e-07, "loss": 0.0314, "step": 3058 }, { "epoch": 1.21, "learning_rate": 3.914620298372243e-07, "loss": 0.0384, "step": 3059 }, { "epoch": 1.21, "learning_rate": 3.9145464760858957e-07, "loss": 0.0338, "step": 3060 }, { "epoch": 1.21, "learning_rate": 3.91447262259535e-07, "loss": 0.0354, "step": 3061 }, { "epoch": 1.21, "learning_rate": 3.9143987379018103e-07, "loss": 0.0314, "step": 3062 }, { "epoch": 1.21, "learning_rate": 3.914324822006481e-07, "loss": 0.03, "step": 3063 }, { "epoch": 1.21, "learning_rate": 3.914250874910567e-07, "loss": 0.0296, "step": 3064 }, { "epoch": 1.21, "learning_rate": 3.914176896615272e-07, "loss": 0.0295, "step": 3065 }, { "epoch": 1.21, "learning_rate": 3.9141028871218035e-07, "loss": 0.035, "step": 3066 }, { "epoch": 1.21, "learning_rate": 3.914028846431367e-07, "loss": 0.0326, "step": 3067 }, { "epoch": 1.21, "learning_rate": 3.9139547745451695e-07, "loss": 0.039, "step": 3068 }, { "epoch": 1.21, "learning_rate": 3.9138806714644177e-07, "loss": 0.0337, "step": 3069 }, { "epoch": 1.21, "learning_rate": 3.9138065371903196e-07, "loss": 0.032, "step": 3070 }, { "epoch": 1.21, "learning_rate": 3.913732371724084e-07, "loss": 0.0339, "step": 3071 }, { "epoch": 1.21, "learning_rate": 3.913658175066919e-07, "loss": 0.0312, "step": 3072 }, { "epoch": 1.21, "learning_rate": 3.913583947220034e-07, "loss": 0.0358, "step": 3073 }, { "epoch": 1.21, "learning_rate": 3.9135096881846394e-07, "loss": 0.0322, "step": 3074 }, { "epoch": 1.21, "learning_rate": 3.913435397961945e-07, "loss": 0.0371, "step": 3075 }, { "epoch": 1.21, "learning_rate": 3.913361076553161e-07, "loss": 0.0336, "step": 3076 }, { "epoch": 1.21, "learning_rate": 3.9132867239595005e-07, "loss": 0.0319, "step": 3077 }, { "epoch": 1.21, "learning_rate": 3.9132123401821736e-07, "loss": 0.0329, "step": 3078 }, { "epoch": 1.21, "learning_rate": 3.9131379252223935e-07, "loss": 0.0361, "step": 3079 }, { "epoch": 1.21, "learning_rate": 3.913063479081373e-07, "loss": 0.0359, "step": 3080 }, { "epoch": 1.21, "learning_rate": 3.9129890017603245e-07, "loss": 0.0316, "step": 3081 }, { "epoch": 1.21, "learning_rate": 3.912914493260464e-07, "loss": 0.0336, "step": 3082 }, { "epoch": 1.22, "learning_rate": 3.9128399535830037e-07, "loss": 0.0316, "step": 3083 }, { "epoch": 1.22, "learning_rate": 3.9127653827291596e-07, "loss": 0.0295, "step": 3084 }, { "epoch": 1.22, "learning_rate": 3.9126907807001466e-07, "loss": 0.0331, "step": 3085 }, { "epoch": 1.22, "learning_rate": 3.9126161474971813e-07, "loss": 0.0333, "step": 3086 }, { "epoch": 1.22, "learning_rate": 3.9125414831214787e-07, "loss": 0.0366, "step": 3087 }, { "epoch": 1.22, "learning_rate": 3.9124667875742573e-07, "loss": 0.0332, "step": 3088 }, { "epoch": 1.22, "learning_rate": 3.9123920608567344e-07, "loss": 0.038, "step": 3089 }, { "epoch": 1.22, "learning_rate": 3.912317302970127e-07, "loss": 0.035, "step": 3090 }, { "epoch": 1.22, "learning_rate": 3.912242513915653e-07, "loss": 0.0358, "step": 3091 }, { "epoch": 1.22, "learning_rate": 3.9121676936945334e-07, "loss": 0.0362, "step": 3092 }, { "epoch": 1.22, "learning_rate": 3.912092842307986e-07, "loss": 0.0371, "step": 3093 }, { "epoch": 1.22, "learning_rate": 3.912017959757232e-07, "loss": 0.0359, "step": 3094 }, { "epoch": 1.22, "learning_rate": 3.9119430460434905e-07, "loss": 0.035, "step": 3095 }, { "epoch": 1.22, "learning_rate": 3.9118681011679835e-07, "loss": 0.0325, "step": 3096 }, { "epoch": 1.22, "learning_rate": 3.911793125131932e-07, "loss": 0.0344, "step": 3097 }, { "epoch": 1.22, "learning_rate": 3.911718117936559e-07, "loss": 0.034, "step": 3098 }, { "epoch": 1.22, "learning_rate": 3.911643079583085e-07, "loss": 0.0319, "step": 3099 }, { "epoch": 1.22, "learning_rate": 3.911568010072734e-07, "loss": 0.0348, "step": 3100 }, { "epoch": 1.22, "learning_rate": 3.91149290940673e-07, "loss": 0.0309, "step": 3101 }, { "epoch": 1.22, "learning_rate": 3.9114177775862974e-07, "loss": 0.0356, "step": 3102 }, { "epoch": 1.22, "learning_rate": 3.9113426146126596e-07, "loss": 0.0363, "step": 3103 }, { "epoch": 1.22, "learning_rate": 3.911267420487042e-07, "loss": 0.0353, "step": 3104 }, { "epoch": 1.22, "learning_rate": 3.9111921952106703e-07, "loss": 0.0348, "step": 3105 }, { "epoch": 1.22, "learning_rate": 3.9111169387847705e-07, "loss": 0.0367, "step": 3106 }, { "epoch": 1.22, "learning_rate": 3.911041651210569e-07, "loss": 0.0302, "step": 3107 }, { "epoch": 1.22, "learning_rate": 3.9109663324892934e-07, "loss": 0.0394, "step": 3108 }, { "epoch": 1.23, "learning_rate": 3.910890982622171e-07, "loss": 0.029, "step": 3109 }, { "epoch": 1.23, "learning_rate": 3.910815601610429e-07, "loss": 0.0349, "step": 3110 }, { "epoch": 1.23, "learning_rate": 3.9107401894552973e-07, "loss": 0.035, "step": 3111 }, { "epoch": 1.23, "learning_rate": 3.910664746158005e-07, "loss": 0.0354, "step": 3112 }, { "epoch": 1.23, "learning_rate": 3.9105892717197806e-07, "loss": 0.034, "step": 3113 }, { "epoch": 1.23, "learning_rate": 3.910513766141855e-07, "loss": 0.0341, "step": 3114 }, { "epoch": 1.23, "learning_rate": 3.9104382294254585e-07, "loss": 0.036, "step": 3115 }, { "epoch": 1.23, "learning_rate": 3.9103626615718226e-07, "loss": 0.0376, "step": 3116 }, { "epoch": 1.23, "learning_rate": 3.910287062582179e-07, "loss": 0.0377, "step": 3117 }, { "epoch": 1.23, "learning_rate": 3.9102114324577585e-07, "loss": 0.0359, "step": 3118 }, { "epoch": 1.23, "learning_rate": 3.910135771199796e-07, "loss": 0.0359, "step": 3119 }, { "epoch": 1.23, "learning_rate": 3.910060078809523e-07, "loss": 0.0381, "step": 3120 }, { "epoch": 1.23, "learning_rate": 3.9099843552881737e-07, "loss": 0.0325, "step": 3121 }, { "epoch": 1.23, "learning_rate": 3.9099086006369823e-07, "loss": 0.0368, "step": 3122 }, { "epoch": 1.23, "learning_rate": 3.9098328148571835e-07, "loss": 0.0327, "step": 3123 }, { "epoch": 1.23, "learning_rate": 3.9097569979500124e-07, "loss": 0.0354, "step": 3124 }, { "epoch": 1.23, "learning_rate": 3.9096811499167044e-07, "loss": 0.0342, "step": 3125 }, { "epoch": 1.23, "learning_rate": 3.9096052707584964e-07, "loss": 0.0295, "step": 3126 }, { "epoch": 1.23, "learning_rate": 3.909529360476625e-07, "loss": 0.0333, "step": 3127 }, { "epoch": 1.23, "learning_rate": 3.909453419072327e-07, "loss": 0.0324, "step": 3128 }, { "epoch": 1.23, "learning_rate": 3.90937744654684e-07, "loss": 0.0314, "step": 3129 }, { "epoch": 1.23, "learning_rate": 3.9093014429014024e-07, "loss": 0.0329, "step": 3130 }, { "epoch": 1.23, "learning_rate": 3.909225408137254e-07, "loss": 0.0353, "step": 3131 }, { "epoch": 1.23, "learning_rate": 3.909149342255632e-07, "loss": 0.03, "step": 3132 }, { "epoch": 1.23, "learning_rate": 3.909073245257778e-07, "loss": 0.0303, "step": 3133 }, { "epoch": 1.24, "learning_rate": 3.908997117144931e-07, "loss": 0.0367, "step": 3134 }, { "epoch": 1.24, "learning_rate": 3.908920957918333e-07, "loss": 0.0339, "step": 3135 }, { "epoch": 1.24, "learning_rate": 3.9088447675792244e-07, "loss": 0.0308, "step": 3136 }, { "epoch": 1.24, "learning_rate": 3.908768546128847e-07, "loss": 0.0296, "step": 3137 }, { "epoch": 1.24, "learning_rate": 3.9086922935684435e-07, "loss": 0.0333, "step": 3138 }, { "epoch": 1.24, "learning_rate": 3.9086160098992563e-07, "loss": 0.0332, "step": 3139 }, { "epoch": 1.24, "learning_rate": 3.908539695122529e-07, "loss": 0.0301, "step": 3140 }, { "epoch": 1.24, "learning_rate": 3.908463349239505e-07, "loss": 0.029, "step": 3141 }, { "epoch": 1.24, "learning_rate": 3.9083869722514295e-07, "loss": 0.0323, "step": 3142 }, { "epoch": 1.24, "learning_rate": 3.9083105641595467e-07, "loss": 0.0324, "step": 3143 }, { "epoch": 1.24, "learning_rate": 3.9082341249651016e-07, "loss": 0.0325, "step": 3144 }, { "epoch": 1.24, "learning_rate": 3.908157654669341e-07, "loss": 0.0371, "step": 3145 }, { "epoch": 1.24, "learning_rate": 3.90808115327351e-07, "loss": 0.0361, "step": 3146 }, { "epoch": 1.24, "learning_rate": 3.9080046207788565e-07, "loss": 0.0325, "step": 3147 }, { "epoch": 1.24, "learning_rate": 3.9079280571866275e-07, "loss": 0.0367, "step": 3148 }, { "epoch": 1.24, "learning_rate": 3.9078514624980705e-07, "loss": 0.034, "step": 3149 }, { "epoch": 1.24, "learning_rate": 3.9077748367144353e-07, "loss": 0.0316, "step": 3150 }, { "epoch": 1.24, "learning_rate": 3.907698179836969e-07, "loss": 0.0329, "step": 3151 }, { "epoch": 1.24, "learning_rate": 3.907621491866922e-07, "loss": 0.0343, "step": 3152 }, { "epoch": 1.24, "learning_rate": 3.907544772805544e-07, "loss": 0.0334, "step": 3153 }, { "epoch": 1.24, "learning_rate": 3.907468022654085e-07, "loss": 0.0348, "step": 3154 }, { "epoch": 1.24, "learning_rate": 3.907391241413796e-07, "loss": 0.034, "step": 3155 }, { "epoch": 1.24, "learning_rate": 3.907314429085929e-07, "loss": 0.0286, "step": 3156 }, { "epoch": 1.24, "learning_rate": 3.907237585671736e-07, "loss": 0.0374, "step": 3157 }, { "epoch": 1.24, "learning_rate": 3.9071607111724687e-07, "loss": 0.0334, "step": 3158 }, { "epoch": 1.24, "learning_rate": 3.90708380558938e-07, "loss": 0.0351, "step": 3159 }, { "epoch": 1.25, "learning_rate": 3.9070068689237237e-07, "loss": 0.034, "step": 3160 }, { "epoch": 1.25, "learning_rate": 3.9069299011767544e-07, "loss": 0.0335, "step": 3161 }, { "epoch": 1.25, "learning_rate": 3.906852902349726e-07, "loss": 0.0327, "step": 3162 }, { "epoch": 1.25, "learning_rate": 3.9067758724438926e-07, "loss": 0.0341, "step": 3163 }, { "epoch": 1.25, "learning_rate": 3.9066988114605107e-07, "loss": 0.0341, "step": 3164 }, { "epoch": 1.25, "learning_rate": 3.906621719400836e-07, "loss": 0.034, "step": 3165 }, { "epoch": 1.25, "learning_rate": 3.906544596266125e-07, "loss": 0.0375, "step": 3166 }, { "epoch": 1.25, "learning_rate": 3.906467442057635e-07, "loss": 0.0343, "step": 3167 }, { "epoch": 1.25, "learning_rate": 3.9063902567766227e-07, "loss": 0.0353, "step": 3168 }, { "epoch": 1.25, "learning_rate": 3.906313040424347e-07, "loss": 0.038, "step": 3169 }, { "epoch": 1.25, "learning_rate": 3.9062357930020663e-07, "loss": 0.0338, "step": 3170 }, { "epoch": 1.25, "learning_rate": 3.9061585145110383e-07, "loss": 0.0339, "step": 3171 }, { "epoch": 1.25, "learning_rate": 3.9060812049525247e-07, "loss": 0.0291, "step": 3172 }, { "epoch": 1.25, "learning_rate": 3.9060038643277834e-07, "loss": 0.0304, "step": 3173 }, { "epoch": 1.25, "learning_rate": 3.905926492638077e-07, "loss": 0.0384, "step": 3174 }, { "epoch": 1.25, "learning_rate": 3.9058490898846644e-07, "loss": 0.0319, "step": 3175 }, { "epoch": 1.25, "learning_rate": 3.905771656068809e-07, "loss": 0.0388, "step": 3176 }, { "epoch": 1.25, "learning_rate": 3.905694191191771e-07, "loss": 0.0288, "step": 3177 }, { "epoch": 1.25, "learning_rate": 3.9056166952548144e-07, "loss": 0.0334, "step": 3178 }, { "epoch": 1.25, "learning_rate": 3.9055391682592026e-07, "loss": 0.035, "step": 3179 }, { "epoch": 1.25, "learning_rate": 3.9054616102061976e-07, "loss": 0.0351, "step": 3180 }, { "epoch": 1.25, "learning_rate": 3.905384021097065e-07, "loss": 0.0349, "step": 3181 }, { "epoch": 1.25, "learning_rate": 3.905306400933068e-07, "loss": 0.0333, "step": 3182 }, { "epoch": 1.25, "learning_rate": 3.905228749715473e-07, "loss": 0.0303, "step": 3183 }, { "epoch": 1.25, "learning_rate": 3.905151067445545e-07, "loss": 0.0309, "step": 3184 }, { "epoch": 1.26, "learning_rate": 3.9050733541245495e-07, "loss": 0.032, "step": 3185 }, { "epoch": 1.26, "learning_rate": 3.904995609753754e-07, "loss": 0.038, "step": 3186 }, { "epoch": 1.26, "learning_rate": 3.9049178343344254e-07, "loss": 0.0395, "step": 3187 }, { "epoch": 1.26, "learning_rate": 3.904840027867832e-07, "loss": 0.038, "step": 3188 }, { "epoch": 1.26, "learning_rate": 3.9047621903552396e-07, "loss": 0.0359, "step": 3189 }, { "epoch": 1.26, "learning_rate": 3.904684321797919e-07, "loss": 0.0347, "step": 3190 }, { "epoch": 1.26, "learning_rate": 3.904606422197139e-07, "loss": 0.0366, "step": 3191 }, { "epoch": 1.26, "learning_rate": 3.9045284915541685e-07, "loss": 0.0384, "step": 3192 }, { "epoch": 1.26, "learning_rate": 3.9044505298702786e-07, "loss": 0.0337, "step": 3193 }, { "epoch": 1.26, "learning_rate": 3.9043725371467397e-07, "loss": 0.0296, "step": 3194 }, { "epoch": 1.26, "learning_rate": 3.9042945133848226e-07, "loss": 0.0353, "step": 3195 }, { "epoch": 1.26, "learning_rate": 3.904216458585799e-07, "loss": 0.0352, "step": 3196 }, { "epoch": 1.26, "learning_rate": 3.904138372750942e-07, "loss": 0.029, "step": 3197 }, { "epoch": 1.26, "learning_rate": 3.9040602558815225e-07, "loss": 0.0312, "step": 3198 }, { "epoch": 1.26, "learning_rate": 3.9039821079788145e-07, "loss": 0.0318, "step": 3199 }, { "epoch": 1.26, "learning_rate": 3.9039039290440925e-07, "loss": 0.03, "step": 3200 }, { "epoch": 1.26, "learning_rate": 3.90382571907863e-07, "loss": 0.0384, "step": 3201 }, { "epoch": 1.26, "learning_rate": 3.903747478083701e-07, "loss": 0.0355, "step": 3202 }, { "epoch": 1.26, "learning_rate": 3.903669206060582e-07, "loss": 0.0367, "step": 3203 }, { "epoch": 1.26, "learning_rate": 3.903590903010548e-07, "loss": 0.0323, "step": 3204 }, { "epoch": 1.26, "learning_rate": 3.903512568934876e-07, "loss": 0.036, "step": 3205 }, { "epoch": 1.26, "learning_rate": 3.9034342038348417e-07, "loss": 0.0321, "step": 3206 }, { "epoch": 1.26, "learning_rate": 3.9033558077117225e-07, "loss": 0.0388, "step": 3207 }, { "epoch": 1.26, "learning_rate": 3.9032773805667966e-07, "loss": 0.0341, "step": 3208 }, { "epoch": 1.26, "learning_rate": 3.9031989224013426e-07, "loss": 0.0302, "step": 3209 }, { "epoch": 1.27, "learning_rate": 3.903120433216638e-07, "loss": 0.0305, "step": 3210 }, { "epoch": 1.27, "learning_rate": 3.903041913013963e-07, "loss": 0.0322, "step": 3211 }, { "epoch": 1.27, "learning_rate": 3.9029633617945975e-07, "loss": 0.0318, "step": 3212 }, { "epoch": 1.27, "learning_rate": 3.902884779559821e-07, "loss": 0.0355, "step": 3213 }, { "epoch": 1.27, "learning_rate": 3.902806166310914e-07, "loss": 0.0376, "step": 3214 }, { "epoch": 1.27, "learning_rate": 3.9027275220491595e-07, "loss": 0.0316, "step": 3215 }, { "epoch": 1.27, "learning_rate": 3.902648846775838e-07, "loss": 0.0314, "step": 3216 }, { "epoch": 1.27, "learning_rate": 3.902570140492231e-07, "loss": 0.0313, "step": 3217 }, { "epoch": 1.27, "learning_rate": 3.902491403199623e-07, "loss": 0.031, "step": 3218 }, { "epoch": 1.27, "learning_rate": 3.902412634899297e-07, "loss": 0.0317, "step": 3219 }, { "epoch": 1.27, "learning_rate": 3.9023338355925357e-07, "loss": 0.0329, "step": 3220 }, { "epoch": 1.27, "learning_rate": 3.9022550052806247e-07, "loss": 0.0348, "step": 3221 }, { "epoch": 1.27, "learning_rate": 3.902176143964848e-07, "loss": 0.0345, "step": 3222 }, { "epoch": 1.27, "learning_rate": 3.902097251646491e-07, "loss": 0.035, "step": 3223 }, { "epoch": 1.27, "learning_rate": 3.902018328326839e-07, "loss": 0.04, "step": 3224 }, { "epoch": 1.27, "learning_rate": 3.90193937400718e-07, "loss": 0.033, "step": 3225 }, { "epoch": 1.27, "learning_rate": 3.9018603886888e-07, "loss": 0.0365, "step": 3226 }, { "epoch": 1.27, "learning_rate": 3.9017813723729854e-07, "loss": 0.0324, "step": 3227 }, { "epoch": 1.27, "learning_rate": 3.901702325061025e-07, "loss": 0.0353, "step": 3228 }, { "epoch": 1.27, "learning_rate": 3.901623246754207e-07, "loss": 0.0381, "step": 3229 }, { "epoch": 1.27, "learning_rate": 3.90154413745382e-07, "loss": 0.0386, "step": 3230 }, { "epoch": 1.27, "learning_rate": 3.901464997161154e-07, "loss": 0.0311, "step": 3231 }, { "epoch": 1.27, "learning_rate": 3.901385825877498e-07, "loss": 0.0338, "step": 3232 }, { "epoch": 1.27, "learning_rate": 3.901306623604143e-07, "loss": 0.0366, "step": 3233 }, { "epoch": 1.27, "learning_rate": 3.90122739034238e-07, "loss": 0.0389, "step": 3234 }, { "epoch": 1.27, "learning_rate": 3.9011481260934997e-07, "loss": 0.0321, "step": 3235 }, { "epoch": 1.28, "learning_rate": 3.9010688308587947e-07, "loss": 0.0335, "step": 3236 }, { "epoch": 1.28, "learning_rate": 3.9009895046395566e-07, "loss": 0.0325, "step": 3237 }, { "epoch": 1.28, "learning_rate": 3.900910147437079e-07, "loss": 0.0376, "step": 3238 }, { "epoch": 1.28, "learning_rate": 3.900830759252655e-07, "loss": 0.0354, "step": 3239 }, { "epoch": 1.28, "learning_rate": 3.9007513400875783e-07, "loss": 0.0319, "step": 3240 }, { "epoch": 1.28, "learning_rate": 3.9006718899431436e-07, "loss": 0.0355, "step": 3241 }, { "epoch": 1.28, "learning_rate": 3.900592408820646e-07, "loss": 0.0341, "step": 3242 }, { "epoch": 1.28, "learning_rate": 3.900512896721381e-07, "loss": 0.034, "step": 3243 }, { "epoch": 1.28, "learning_rate": 3.900433353646644e-07, "loss": 0.0326, "step": 3244 }, { "epoch": 1.28, "learning_rate": 3.9003537795977317e-07, "loss": 0.0329, "step": 3245 }, { "epoch": 1.28, "learning_rate": 3.9002741745759404e-07, "loss": 0.037, "step": 3246 }, { "epoch": 1.28, "learning_rate": 3.900194538582569e-07, "loss": 0.0353, "step": 3247 }, { "epoch": 1.28, "learning_rate": 3.900114871618914e-07, "loss": 0.0318, "step": 3248 }, { "epoch": 1.28, "learning_rate": 3.900035173686275e-07, "loss": 0.0317, "step": 3249 }, { "epoch": 1.28, "learning_rate": 3.8999554447859496e-07, "loss": 0.0341, "step": 3250 }, { "epoch": 1.28, "learning_rate": 3.899875684919239e-07, "loss": 0.0304, "step": 3251 }, { "epoch": 1.28, "learning_rate": 3.8997958940874416e-07, "loss": 0.0319, "step": 3252 }, { "epoch": 1.28, "learning_rate": 3.8997160722918583e-07, "loss": 0.0328, "step": 3253 }, { "epoch": 1.28, "learning_rate": 3.899636219533791e-07, "loss": 0.034, "step": 3254 }, { "epoch": 1.28, "learning_rate": 3.8995563358145394e-07, "loss": 0.0339, "step": 3255 }, { "epoch": 1.28, "learning_rate": 3.8994764211354073e-07, "loss": 0.031, "step": 3256 }, { "epoch": 1.28, "learning_rate": 3.899396475497696e-07, "loss": 0.0403, "step": 3257 }, { "epoch": 1.28, "learning_rate": 3.899316498902709e-07, "loss": 0.0312, "step": 3258 }, { "epoch": 1.28, "learning_rate": 3.89923649135175e-07, "loss": 0.0346, "step": 3259 }, { "epoch": 1.28, "learning_rate": 3.8991564528461225e-07, "loss": 0.0333, "step": 3260 }, { "epoch": 1.29, "learning_rate": 3.899076383387131e-07, "loss": 0.0348, "step": 3261 }, { "epoch": 1.29, "learning_rate": 3.898996282976081e-07, "loss": 0.0355, "step": 3262 }, { "epoch": 1.29, "learning_rate": 3.898916151614278e-07, "loss": 0.0306, "step": 3263 }, { "epoch": 1.29, "learning_rate": 3.898835989303027e-07, "loss": 0.0318, "step": 3264 }, { "epoch": 1.29, "learning_rate": 3.8987557960436356e-07, "loss": 0.0334, "step": 3265 }, { "epoch": 1.29, "learning_rate": 3.898675571837411e-07, "loss": 0.0351, "step": 3266 }, { "epoch": 1.29, "learning_rate": 3.89859531668566e-07, "loss": 0.0319, "step": 3267 }, { "epoch": 1.29, "learning_rate": 3.89851503058969e-07, "loss": 0.0365, "step": 3268 }, { "epoch": 1.29, "learning_rate": 3.8984347135508113e-07, "loss": 0.0297, "step": 3269 }, { "epoch": 1.29, "learning_rate": 3.8983543655703313e-07, "loss": 0.0363, "step": 3270 }, { "epoch": 1.29, "learning_rate": 3.8982739866495615e-07, "loss": 0.0394, "step": 3271 }, { "epoch": 1.29, "learning_rate": 3.89819357678981e-07, "loss": 0.0362, "step": 3272 }, { "epoch": 1.29, "learning_rate": 3.898113135992388e-07, "loss": 0.0375, "step": 3273 }, { "epoch": 1.29, "learning_rate": 3.898032664258607e-07, "loss": 0.0335, "step": 3274 }, { "epoch": 1.29, "learning_rate": 3.897952161589778e-07, "loss": 0.0375, "step": 3275 }, { "epoch": 1.29, "learning_rate": 3.8978716279872134e-07, "loss": 0.0374, "step": 3276 }, { "epoch": 1.29, "learning_rate": 3.897791063452226e-07, "loss": 0.0334, "step": 3277 }, { "epoch": 1.29, "learning_rate": 3.8977104679861285e-07, "loss": 0.0308, "step": 3278 }, { "epoch": 1.29, "learning_rate": 3.897629841590234e-07, "loss": 0.0316, "step": 3279 }, { "epoch": 1.29, "learning_rate": 3.897549184265858e-07, "loss": 0.0335, "step": 3280 }, { "epoch": 1.29, "learning_rate": 3.8974684960143136e-07, "loss": 0.0357, "step": 3281 }, { "epoch": 1.29, "learning_rate": 3.8973877768369165e-07, "loss": 0.0317, "step": 3282 }, { "epoch": 1.29, "learning_rate": 3.897307026734983e-07, "loss": 0.0342, "step": 3283 }, { "epoch": 1.29, "learning_rate": 3.897226245709828e-07, "loss": 0.0322, "step": 3284 }, { "epoch": 1.29, "learning_rate": 3.897145433762769e-07, "loss": 0.0296, "step": 3285 }, { "epoch": 1.3, "learning_rate": 3.8970645908951225e-07, "loss": 0.0327, "step": 3286 }, { "epoch": 1.3, "learning_rate": 3.896983717108206e-07, "loss": 0.0374, "step": 3287 }, { "epoch": 1.3, "learning_rate": 3.896902812403338e-07, "loss": 0.0307, "step": 3288 }, { "epoch": 1.3, "learning_rate": 3.8968218767818384e-07, "loss": 0.0346, "step": 3289 }, { "epoch": 1.3, "learning_rate": 3.8967409102450237e-07, "loss": 0.0341, "step": 3290 }, { "epoch": 1.3, "learning_rate": 3.8966599127942154e-07, "loss": 0.0323, "step": 3291 }, { "epoch": 1.3, "learning_rate": 3.8965788844307327e-07, "loss": 0.0348, "step": 3292 }, { "epoch": 1.3, "learning_rate": 3.8964978251558975e-07, "loss": 0.0361, "step": 3293 }, { "epoch": 1.3, "learning_rate": 3.896416734971029e-07, "loss": 0.0301, "step": 3294 }, { "epoch": 1.3, "learning_rate": 3.896335613877451e-07, "loss": 0.0415, "step": 3295 }, { "epoch": 1.3, "learning_rate": 3.8962544618764837e-07, "loss": 0.0308, "step": 3296 }, { "epoch": 1.3, "learning_rate": 3.896173278969451e-07, "loss": 0.0368, "step": 3297 }, { "epoch": 1.3, "learning_rate": 3.896092065157676e-07, "loss": 0.0374, "step": 3298 }, { "epoch": 1.3, "learning_rate": 3.8960108204424816e-07, "loss": 0.0313, "step": 3299 }, { "epoch": 1.3, "learning_rate": 3.8959295448251926e-07, "loss": 0.0295, "step": 3300 }, { "epoch": 1.3, "learning_rate": 3.8958482383071337e-07, "loss": 0.0352, "step": 3301 }, { "epoch": 1.3, "learning_rate": 3.89576690088963e-07, "loss": 0.0344, "step": 3302 }, { "epoch": 1.3, "learning_rate": 3.895685532574006e-07, "loss": 0.0314, "step": 3303 }, { "epoch": 1.3, "learning_rate": 3.8956041333615907e-07, "loss": 0.0348, "step": 3304 }, { "epoch": 1.3, "learning_rate": 3.895522703253708e-07, "loss": 0.0313, "step": 3305 }, { "epoch": 1.3, "learning_rate": 3.895441242251686e-07, "loss": 0.0322, "step": 3306 }, { "epoch": 1.3, "learning_rate": 3.895359750356852e-07, "loss": 0.0332, "step": 3307 }, { "epoch": 1.3, "learning_rate": 3.895278227570536e-07, "loss": 0.0318, "step": 3308 }, { "epoch": 1.3, "learning_rate": 3.895196673894064e-07, "loss": 0.0344, "step": 3309 }, { "epoch": 1.3, "learning_rate": 3.895115089328768e-07, "loss": 0.0348, "step": 3310 }, { "epoch": 1.3, "learning_rate": 3.895033473875975e-07, "loss": 0.0311, "step": 3311 }, { "epoch": 1.31, "learning_rate": 3.894951827537017e-07, "loss": 0.0341, "step": 3312 }, { "epoch": 1.31, "learning_rate": 3.8948701503132243e-07, "loss": 0.0332, "step": 3313 }, { "epoch": 1.31, "learning_rate": 3.894788442205928e-07, "loss": 0.0338, "step": 3314 }, { "epoch": 1.31, "learning_rate": 3.89470670321646e-07, "loss": 0.0345, "step": 3315 }, { "epoch": 1.31, "learning_rate": 3.894624933346152e-07, "loss": 0.0346, "step": 3316 }, { "epoch": 1.31, "learning_rate": 3.894543132596337e-07, "loss": 0.0334, "step": 3317 }, { "epoch": 1.31, "learning_rate": 3.8944613009683485e-07, "loss": 0.037, "step": 3318 }, { "epoch": 1.31, "learning_rate": 3.89437943846352e-07, "loss": 0.0334, "step": 3319 }, { "epoch": 1.31, "learning_rate": 3.8942975450831863e-07, "loss": 0.0313, "step": 3320 }, { "epoch": 1.31, "learning_rate": 3.894215620828681e-07, "loss": 0.0321, "step": 3321 }, { "epoch": 1.31, "learning_rate": 3.8941336657013404e-07, "loss": 0.0336, "step": 3322 }, { "epoch": 1.31, "learning_rate": 3.8940516797024997e-07, "loss": 0.0329, "step": 3323 }, { "epoch": 1.31, "learning_rate": 3.893969662833495e-07, "loss": 0.032, "step": 3324 }, { "epoch": 1.31, "learning_rate": 3.8938876150956637e-07, "loss": 0.0323, "step": 3325 }, { "epoch": 1.31, "learning_rate": 3.8938055364903424e-07, "loss": 0.0307, "step": 3326 }, { "epoch": 1.31, "learning_rate": 3.8937234270188683e-07, "loss": 0.0314, "step": 3327 }, { "epoch": 1.31, "learning_rate": 3.893641286682581e-07, "loss": 0.0346, "step": 3328 }, { "epoch": 1.31, "learning_rate": 3.893559115482819e-07, "loss": 0.0339, "step": 3329 }, { "epoch": 1.31, "learning_rate": 3.8934769134209217e-07, "loss": 0.034, "step": 3330 }, { "epoch": 1.31, "learning_rate": 3.893394680498228e-07, "loss": 0.0369, "step": 3331 }, { "epoch": 1.31, "learning_rate": 3.893312416716079e-07, "loss": 0.0314, "step": 3332 }, { "epoch": 1.31, "learning_rate": 3.8932301220758143e-07, "loss": 0.0333, "step": 3333 }, { "epoch": 1.31, "learning_rate": 3.8931477965787763e-07, "loss": 0.037, "step": 3334 }, { "epoch": 1.31, "learning_rate": 3.893065440226307e-07, "loss": 0.0349, "step": 3335 }, { "epoch": 1.31, "learning_rate": 3.892983053019748e-07, "loss": 0.0363, "step": 3336 }, { "epoch": 1.32, "learning_rate": 3.892900634960441e-07, "loss": 0.036, "step": 3337 }, { "epoch": 1.32, "learning_rate": 3.892818186049732e-07, "loss": 0.0363, "step": 3338 }, { "epoch": 1.32, "learning_rate": 3.892735706288963e-07, "loss": 0.0261, "step": 3339 }, { "epoch": 1.32, "learning_rate": 3.892653195679479e-07, "loss": 0.0347, "step": 3340 }, { "epoch": 1.32, "learning_rate": 3.8925706542226236e-07, "loss": 0.039, "step": 3341 }, { "epoch": 1.32, "learning_rate": 3.892488081919743e-07, "loss": 0.033, "step": 3342 }, { "epoch": 1.32, "learning_rate": 3.892405478772183e-07, "loss": 0.034, "step": 3343 }, { "epoch": 1.32, "learning_rate": 3.89232284478129e-07, "loss": 0.0327, "step": 3344 }, { "epoch": 1.32, "learning_rate": 3.892240179948411e-07, "loss": 0.0348, "step": 3345 }, { "epoch": 1.32, "learning_rate": 3.892157484274892e-07, "loss": 0.0322, "step": 3346 }, { "epoch": 1.32, "learning_rate": 3.892074757762082e-07, "loss": 0.0344, "step": 3347 }, { "epoch": 1.32, "learning_rate": 3.8919920004113295e-07, "loss": 0.0317, "step": 3348 }, { "epoch": 1.32, "learning_rate": 3.891909212223983e-07, "loss": 0.029, "step": 3349 }, { "epoch": 1.32, "learning_rate": 3.891826393201391e-07, "loss": 0.0295, "step": 3350 }, { "epoch": 1.32, "learning_rate": 3.891743543344905e-07, "loss": 0.0312, "step": 3351 }, { "epoch": 1.32, "learning_rate": 3.891660662655874e-07, "loss": 0.0355, "step": 3352 }, { "epoch": 1.32, "learning_rate": 3.8915777511356483e-07, "loss": 0.0361, "step": 3353 }, { "epoch": 1.32, "learning_rate": 3.891494808785581e-07, "loss": 0.0357, "step": 3354 }, { "epoch": 1.32, "learning_rate": 3.891411835607023e-07, "loss": 0.0254, "step": 3355 }, { "epoch": 1.32, "learning_rate": 3.891328831601327e-07, "loss": 0.0347, "step": 3356 }, { "epoch": 1.32, "learning_rate": 3.891245796769845e-07, "loss": 0.0345, "step": 3357 }, { "epoch": 1.32, "learning_rate": 3.8911627311139305e-07, "loss": 0.0328, "step": 3358 }, { "epoch": 1.32, "learning_rate": 3.891079634634939e-07, "loss": 0.0348, "step": 3359 }, { "epoch": 1.32, "learning_rate": 3.8909965073342223e-07, "loss": 0.0316, "step": 3360 }, { "epoch": 1.32, "learning_rate": 3.890913349213137e-07, "loss": 0.0335, "step": 3361 }, { "epoch": 1.32, "learning_rate": 3.890830160273038e-07, "loss": 0.0341, "step": 3362 }, { "epoch": 1.33, "learning_rate": 3.890746940515281e-07, "loss": 0.0306, "step": 3363 }, { "epoch": 1.33, "learning_rate": 3.8906636899412227e-07, "loss": 0.0294, "step": 3364 }, { "epoch": 1.33, "learning_rate": 3.8905804085522196e-07, "loss": 0.0337, "step": 3365 }, { "epoch": 1.33, "learning_rate": 3.890497096349629e-07, "loss": 0.0297, "step": 3366 }, { "epoch": 1.33, "learning_rate": 3.890413753334809e-07, "loss": 0.0318, "step": 3367 }, { "epoch": 1.33, "learning_rate": 3.8903303795091184e-07, "loss": 0.0386, "step": 3368 }, { "epoch": 1.33, "learning_rate": 3.890246974873915e-07, "loss": 0.0293, "step": 3369 }, { "epoch": 1.33, "learning_rate": 3.890163539430559e-07, "loss": 0.0362, "step": 3370 }, { "epoch": 1.33, "learning_rate": 3.89008007318041e-07, "loss": 0.0319, "step": 3371 }, { "epoch": 1.33, "learning_rate": 3.889996576124829e-07, "loss": 0.0299, "step": 3372 }, { "epoch": 1.33, "learning_rate": 3.889913048265175e-07, "loss": 0.0321, "step": 3373 }, { "epoch": 1.33, "learning_rate": 3.889829489602811e-07, "loss": 0.0323, "step": 3374 }, { "epoch": 1.33, "learning_rate": 3.8897459001390997e-07, "loss": 0.0322, "step": 3375 }, { "epoch": 1.33, "learning_rate": 3.8896622798754014e-07, "loss": 0.0368, "step": 3376 }, { "epoch": 1.33, "learning_rate": 3.8895786288130796e-07, "loss": 0.0413, "step": 3377 }, { "epoch": 1.33, "learning_rate": 3.8894949469534985e-07, "loss": 0.0324, "step": 3378 }, { "epoch": 1.33, "learning_rate": 3.8894112342980213e-07, "loss": 0.0381, "step": 3379 }, { "epoch": 1.33, "learning_rate": 3.8893274908480123e-07, "loss": 0.0346, "step": 3380 }, { "epoch": 1.33, "learning_rate": 3.889243716604837e-07, "loss": 0.0337, "step": 3381 }, { "epoch": 1.33, "learning_rate": 3.8891599115698605e-07, "loss": 0.0358, "step": 3382 }, { "epoch": 1.33, "learning_rate": 3.8890760757444484e-07, "loss": 0.0305, "step": 3383 }, { "epoch": 1.33, "learning_rate": 3.888992209129967e-07, "loss": 0.0286, "step": 3384 }, { "epoch": 1.33, "learning_rate": 3.8889083117277845e-07, "loss": 0.036, "step": 3385 }, { "epoch": 1.33, "learning_rate": 3.888824383539267e-07, "loss": 0.0361, "step": 3386 }, { "epoch": 1.33, "learning_rate": 3.8887404245657825e-07, "loss": 0.0328, "step": 3387 }, { "epoch": 1.34, "learning_rate": 3.888656434808699e-07, "loss": 0.0345, "step": 3388 }, { "epoch": 1.34, "learning_rate": 3.8885724142693875e-07, "loss": 0.0326, "step": 3389 }, { "epoch": 1.34, "learning_rate": 3.8884883629492147e-07, "loss": 0.0379, "step": 3390 }, { "epoch": 1.34, "learning_rate": 3.8884042808495523e-07, "loss": 0.0334, "step": 3391 }, { "epoch": 1.34, "learning_rate": 3.88832016797177e-07, "loss": 0.0318, "step": 3392 }, { "epoch": 1.34, "learning_rate": 3.8882360243172385e-07, "loss": 0.0316, "step": 3393 }, { "epoch": 1.34, "learning_rate": 3.8881518498873305e-07, "loss": 0.0331, "step": 3394 }, { "epoch": 1.34, "learning_rate": 3.8880676446834163e-07, "loss": 0.0362, "step": 3395 }, { "epoch": 1.34, "learning_rate": 3.887983408706869e-07, "loss": 0.0336, "step": 3396 }, { "epoch": 1.34, "learning_rate": 3.887899141959062e-07, "loss": 0.0343, "step": 3397 }, { "epoch": 1.34, "learning_rate": 3.8878148444413686e-07, "loss": 0.0321, "step": 3398 }, { "epoch": 1.34, "learning_rate": 3.8877305161551614e-07, "loss": 0.0323, "step": 3399 }, { "epoch": 1.34, "learning_rate": 3.8876461571018165e-07, "loss": 0.0347, "step": 3400 }, { "epoch": 1.34, "learning_rate": 3.8875617672827077e-07, "loss": 0.0329, "step": 3401 }, { "epoch": 1.34, "learning_rate": 3.887477346699211e-07, "loss": 0.0291, "step": 3402 }, { "epoch": 1.34, "learning_rate": 3.8873928953527023e-07, "loss": 0.0308, "step": 3403 }, { "epoch": 1.34, "learning_rate": 3.8873084132445575e-07, "loss": 0.0336, "step": 3404 }, { "epoch": 1.34, "learning_rate": 3.8872239003761545e-07, "loss": 0.0312, "step": 3405 }, { "epoch": 1.34, "learning_rate": 3.8871393567488697e-07, "loss": 0.0329, "step": 3406 }, { "epoch": 1.34, "learning_rate": 3.8870547823640817e-07, "loss": 0.0297, "step": 3407 }, { "epoch": 1.34, "learning_rate": 3.886970177223169e-07, "loss": 0.0333, "step": 3408 }, { "epoch": 1.34, "learning_rate": 3.88688554132751e-07, "loss": 0.0357, "step": 3409 }, { "epoch": 1.34, "learning_rate": 3.8868008746784844e-07, "loss": 0.0334, "step": 3410 }, { "epoch": 1.34, "learning_rate": 3.8867161772774724e-07, "loss": 0.032, "step": 3411 }, { "epoch": 1.34, "learning_rate": 3.886631449125854e-07, "loss": 0.0335, "step": 3412 }, { "epoch": 1.35, "learning_rate": 3.8865466902250103e-07, "loss": 0.0353, "step": 3413 }, { "epoch": 1.35, "learning_rate": 3.886461900576323e-07, "loss": 0.0319, "step": 3414 }, { "epoch": 1.35, "learning_rate": 3.8863770801811736e-07, "loss": 0.0327, "step": 3415 }, { "epoch": 1.35, "learning_rate": 3.886292229040945e-07, "loss": 0.0345, "step": 3416 }, { "epoch": 1.35, "learning_rate": 3.8862073471570193e-07, "loss": 0.0347, "step": 3417 }, { "epoch": 1.35, "learning_rate": 3.886122434530781e-07, "loss": 0.0343, "step": 3418 }, { "epoch": 1.35, "learning_rate": 3.886037491163614e-07, "loss": 0.0334, "step": 3419 }, { "epoch": 1.35, "learning_rate": 3.8859525170569015e-07, "loss": 0.031, "step": 3420 }, { "epoch": 1.35, "learning_rate": 3.8858675122120295e-07, "loss": 0.0324, "step": 3421 }, { "epoch": 1.35, "learning_rate": 3.885782476630384e-07, "loss": 0.0297, "step": 3422 }, { "epoch": 1.35, "learning_rate": 3.885697410313349e-07, "loss": 0.0376, "step": 3423 }, { "epoch": 1.35, "learning_rate": 3.885612313262313e-07, "loss": 0.0312, "step": 3424 }, { "epoch": 1.35, "learning_rate": 3.885527185478662e-07, "loss": 0.0351, "step": 3425 }, { "epoch": 1.35, "learning_rate": 3.8854420269637835e-07, "loss": 0.0355, "step": 3426 }, { "epoch": 1.35, "learning_rate": 3.885356837719065e-07, "loss": 0.0302, "step": 3427 }, { "epoch": 1.35, "learning_rate": 3.885271617745895e-07, "loss": 0.037, "step": 3428 }, { "epoch": 1.35, "learning_rate": 3.885186367045664e-07, "loss": 0.0338, "step": 3429 }, { "epoch": 1.35, "learning_rate": 3.885101085619761e-07, "loss": 0.0309, "step": 3430 }, { "epoch": 1.35, "learning_rate": 3.8850157734695737e-07, "loss": 0.03, "step": 3431 }, { "epoch": 1.35, "learning_rate": 3.8849304305964947e-07, "loss": 0.03, "step": 3432 }, { "epoch": 1.35, "learning_rate": 3.8848450570019156e-07, "loss": 0.0313, "step": 3433 }, { "epoch": 1.35, "learning_rate": 3.8847596526872255e-07, "loss": 0.0295, "step": 3434 }, { "epoch": 1.35, "learning_rate": 3.884674217653818e-07, "loss": 0.0285, "step": 3435 }, { "epoch": 1.35, "learning_rate": 3.884588751903084e-07, "loss": 0.0347, "step": 3436 }, { "epoch": 1.35, "learning_rate": 3.884503255436419e-07, "loss": 0.0353, "step": 3437 }, { "epoch": 1.35, "learning_rate": 3.884417728255215e-07, "loss": 0.0305, "step": 3438 }, { "epoch": 1.36, "learning_rate": 3.8843321703608653e-07, "loss": 0.0299, "step": 3439 }, { "epoch": 1.36, "learning_rate": 3.884246581754766e-07, "loss": 0.0279, "step": 3440 }, { "epoch": 1.36, "learning_rate": 3.88416096243831e-07, "loss": 0.0292, "step": 3441 }, { "epoch": 1.36, "learning_rate": 3.8840753124128947e-07, "loss": 0.031, "step": 3442 }, { "epoch": 1.36, "learning_rate": 3.883989631679915e-07, "loss": 0.0346, "step": 3443 }, { "epoch": 1.36, "learning_rate": 3.8839039202407674e-07, "loss": 0.0335, "step": 3444 }, { "epoch": 1.36, "learning_rate": 3.8838181780968495e-07, "loss": 0.0337, "step": 3445 }, { "epoch": 1.36, "learning_rate": 3.8837324052495585e-07, "loss": 0.0316, "step": 3446 }, { "epoch": 1.36, "learning_rate": 3.883646601700292e-07, "loss": 0.0294, "step": 3447 }, { "epoch": 1.36, "learning_rate": 3.883560767450448e-07, "loss": 0.0296, "step": 3448 }, { "epoch": 1.36, "learning_rate": 3.883474902501427e-07, "loss": 0.0316, "step": 3449 }, { "epoch": 1.36, "learning_rate": 3.8833890068546276e-07, "loss": 0.0394, "step": 3450 }, { "epoch": 1.36, "learning_rate": 3.8833030805114495e-07, "loss": 0.0314, "step": 3451 }, { "epoch": 1.36, "learning_rate": 3.883217123473294e-07, "loss": 0.0303, "step": 3452 }, { "epoch": 1.36, "learning_rate": 3.883131135741561e-07, "loss": 0.0313, "step": 3453 }, { "epoch": 1.36, "learning_rate": 3.8830451173176527e-07, "loss": 0.0318, "step": 3454 }, { "epoch": 1.36, "learning_rate": 3.882959068202971e-07, "loss": 0.032, "step": 3455 }, { "epoch": 1.36, "learning_rate": 3.8828729883989186e-07, "loss": 0.0366, "step": 3456 }, { "epoch": 1.36, "learning_rate": 3.8827868779068973e-07, "loss": 0.0308, "step": 3457 }, { "epoch": 1.36, "learning_rate": 3.8827007367283117e-07, "loss": 0.035, "step": 3458 }, { "epoch": 1.36, "learning_rate": 3.882614564864566e-07, "loss": 0.0307, "step": 3459 }, { "epoch": 1.36, "learning_rate": 3.8825283623170636e-07, "loss": 0.0333, "step": 3460 }, { "epoch": 1.36, "learning_rate": 3.88244212908721e-07, "loss": 0.0351, "step": 3461 }, { "epoch": 1.36, "learning_rate": 3.8823558651764107e-07, "loss": 0.0347, "step": 3462 }, { "epoch": 1.36, "learning_rate": 3.882269570586072e-07, "loss": 0.0296, "step": 3463 }, { "epoch": 1.37, "learning_rate": 3.8821832453175995e-07, "loss": 0.0305, "step": 3464 }, { "epoch": 1.37, "learning_rate": 3.882096889372401e-07, "loss": 0.0341, "step": 3465 }, { "epoch": 1.37, "learning_rate": 3.8820105027518834e-07, "loss": 0.0328, "step": 3466 }, { "epoch": 1.37, "learning_rate": 3.881924085457456e-07, "loss": 0.0291, "step": 3467 }, { "epoch": 1.37, "learning_rate": 3.881837637490525e-07, "loss": 0.0323, "step": 3468 }, { "epoch": 1.37, "learning_rate": 3.8817511588525006e-07, "loss": 0.0358, "step": 3469 }, { "epoch": 1.37, "learning_rate": 3.881664649544793e-07, "loss": 0.0308, "step": 3470 }, { "epoch": 1.37, "learning_rate": 3.881578109568811e-07, "loss": 0.0348, "step": 3471 }, { "epoch": 1.37, "learning_rate": 3.8814915389259653e-07, "loss": 0.034, "step": 3472 }, { "epoch": 1.37, "learning_rate": 3.881404937617667e-07, "loss": 0.032, "step": 3473 }, { "epoch": 1.37, "learning_rate": 3.881318305645328e-07, "loss": 0.0309, "step": 3474 }, { "epoch": 1.37, "learning_rate": 3.88123164301036e-07, "loss": 0.0296, "step": 3475 }, { "epoch": 1.37, "learning_rate": 3.8811449497141745e-07, "loss": 0.0272, "step": 3476 }, { "epoch": 1.37, "learning_rate": 3.881058225758186e-07, "loss": 0.0304, "step": 3477 }, { "epoch": 1.37, "learning_rate": 3.8809714711438075e-07, "loss": 0.0331, "step": 3478 }, { "epoch": 1.37, "learning_rate": 3.880884685872452e-07, "loss": 0.028, "step": 3479 }, { "epoch": 1.37, "learning_rate": 3.8807978699455356e-07, "loss": 0.0369, "step": 3480 }, { "epoch": 1.37, "learning_rate": 3.880711023364472e-07, "loss": 0.0324, "step": 3481 }, { "epoch": 1.37, "learning_rate": 3.8806241461306775e-07, "loss": 0.0293, "step": 3482 }, { "epoch": 1.37, "learning_rate": 3.8805372382455665e-07, "loss": 0.0381, "step": 3483 }, { "epoch": 1.37, "learning_rate": 3.880450299710558e-07, "loss": 0.0375, "step": 3484 }, { "epoch": 1.37, "learning_rate": 3.880363330527067e-07, "loss": 0.0316, "step": 3485 }, { "epoch": 1.37, "learning_rate": 3.880276330696511e-07, "loss": 0.0363, "step": 3486 }, { "epoch": 1.37, "learning_rate": 3.880189300220309e-07, "loss": 0.0362, "step": 3487 }, { "epoch": 1.37, "learning_rate": 3.8801022390998794e-07, "loss": 0.0358, "step": 3488 }, { "epoch": 1.38, "learning_rate": 3.8800151473366396e-07, "loss": 0.0323, "step": 3489 }, { "epoch": 1.38, "learning_rate": 3.879928024932011e-07, "loss": 0.0311, "step": 3490 }, { "epoch": 1.38, "learning_rate": 3.8798408718874124e-07, "loss": 0.0336, "step": 3491 }, { "epoch": 1.38, "learning_rate": 3.879753688204265e-07, "loss": 0.0391, "step": 3492 }, { "epoch": 1.38, "learning_rate": 3.879666473883989e-07, "loss": 0.0294, "step": 3493 }, { "epoch": 1.38, "learning_rate": 3.8795792289280065e-07, "loss": 0.0338, "step": 3494 }, { "epoch": 1.38, "learning_rate": 3.879491953337739e-07, "loss": 0.0374, "step": 3495 }, { "epoch": 1.38, "learning_rate": 3.87940464711461e-07, "loss": 0.0303, "step": 3496 }, { "epoch": 1.38, "learning_rate": 3.8793173102600406e-07, "loss": 0.032, "step": 3497 }, { "epoch": 1.38, "learning_rate": 3.8792299427754554e-07, "loss": 0.029, "step": 3498 }, { "epoch": 1.38, "learning_rate": 3.8791425446622786e-07, "loss": 0.0344, "step": 3499 }, { "epoch": 1.38, "learning_rate": 3.879055115921934e-07, "loss": 0.0317, "step": 3500 }, { "epoch": 1.38, "learning_rate": 3.878967656555847e-07, "loss": 0.0369, "step": 3501 }, { "epoch": 1.38, "learning_rate": 3.8788801665654425e-07, "loss": 0.0307, "step": 3502 }, { "epoch": 1.38, "learning_rate": 3.8787926459521474e-07, "loss": 0.0352, "step": 3503 }, { "epoch": 1.38, "learning_rate": 3.878705094717388e-07, "loss": 0.0382, "step": 3504 }, { "epoch": 1.38, "learning_rate": 3.87861751286259e-07, "loss": 0.0316, "step": 3505 }, { "epoch": 1.38, "learning_rate": 3.8785299003891816e-07, "loss": 0.0333, "step": 3506 }, { "epoch": 1.38, "learning_rate": 3.878442257298591e-07, "loss": 0.0355, "step": 3507 }, { "epoch": 1.38, "learning_rate": 3.878354583592247e-07, "loss": 0.0333, "step": 3508 }, { "epoch": 1.38, "learning_rate": 3.878266879271578e-07, "loss": 0.0387, "step": 3509 }, { "epoch": 1.38, "learning_rate": 3.878179144338013e-07, "loss": 0.0339, "step": 3510 }, { "epoch": 1.38, "learning_rate": 3.878091378792983e-07, "loss": 0.034, "step": 3511 }, { "epoch": 1.38, "learning_rate": 3.878003582637918e-07, "loss": 0.036, "step": 3512 }, { "epoch": 1.38, "learning_rate": 3.877915755874248e-07, "loss": 0.0341, "step": 3513 }, { "epoch": 1.38, "learning_rate": 3.8778278985034056e-07, "loss": 0.0335, "step": 3514 }, { "epoch": 1.39, "learning_rate": 3.877740010526823e-07, "loss": 0.0332, "step": 3515 }, { "epoch": 1.39, "learning_rate": 3.8776520919459314e-07, "loss": 0.0359, "step": 3516 }, { "epoch": 1.39, "learning_rate": 3.877564142762165e-07, "loss": 0.0344, "step": 3517 }, { "epoch": 1.39, "learning_rate": 3.877476162976956e-07, "loss": 0.0388, "step": 3518 }, { "epoch": 1.39, "learning_rate": 3.877388152591739e-07, "loss": 0.0325, "step": 3519 }, { "epoch": 1.39, "learning_rate": 3.877300111607949e-07, "loss": 0.0375, "step": 3520 }, { "epoch": 1.39, "learning_rate": 3.87721204002702e-07, "loss": 0.0286, "step": 3521 }, { "epoch": 1.39, "learning_rate": 3.877123937850388e-07, "loss": 0.0315, "step": 3522 }, { "epoch": 1.39, "learning_rate": 3.8770358050794885e-07, "loss": 0.0342, "step": 3523 }, { "epoch": 1.39, "learning_rate": 3.876947641715758e-07, "loss": 0.0304, "step": 3524 }, { "epoch": 1.39, "learning_rate": 3.876859447760634e-07, "loss": 0.0325, "step": 3525 }, { "epoch": 1.39, "learning_rate": 3.8767712232155534e-07, "loss": 0.0321, "step": 3526 }, { "epoch": 1.39, "learning_rate": 3.8766829680819536e-07, "loss": 0.0336, "step": 3527 }, { "epoch": 1.39, "learning_rate": 3.876594682361274e-07, "loss": 0.0298, "step": 3528 }, { "epoch": 1.39, "learning_rate": 3.8765063660549535e-07, "loss": 0.0309, "step": 3529 }, { "epoch": 1.39, "learning_rate": 3.876418019164431e-07, "loss": 0.029, "step": 3530 }, { "epoch": 1.39, "learning_rate": 3.876329641691146e-07, "loss": 0.0323, "step": 3531 }, { "epoch": 1.39, "learning_rate": 3.8762412336365406e-07, "loss": 0.0339, "step": 3532 }, { "epoch": 1.39, "learning_rate": 3.8761527950020546e-07, "loss": 0.035, "step": 3533 }, { "epoch": 1.39, "learning_rate": 3.8760643257891286e-07, "loss": 0.0325, "step": 3534 }, { "epoch": 1.39, "learning_rate": 3.8759758259992063e-07, "loss": 0.0276, "step": 3535 }, { "epoch": 1.39, "learning_rate": 3.875887295633729e-07, "loss": 0.0351, "step": 3536 }, { "epoch": 1.39, "learning_rate": 3.8757987346941385e-07, "loss": 0.0361, "step": 3537 }, { "epoch": 1.39, "learning_rate": 3.875710143181881e-07, "loss": 0.0339, "step": 3538 }, { "epoch": 1.39, "learning_rate": 3.875621521098399e-07, "loss": 0.03, "step": 3539 }, { "epoch": 1.4, "learning_rate": 3.8755328684451356e-07, "loss": 0.0334, "step": 3540 }, { "epoch": 1.4, "learning_rate": 3.8754441852235376e-07, "loss": 0.0285, "step": 3541 }, { "epoch": 1.4, "learning_rate": 3.87535547143505e-07, "loss": 0.0344, "step": 3542 }, { "epoch": 1.4, "learning_rate": 3.875266727081118e-07, "loss": 0.03, "step": 3543 }, { "epoch": 1.4, "learning_rate": 3.875177952163189e-07, "loss": 0.0305, "step": 3544 }, { "epoch": 1.4, "learning_rate": 3.8750891466827086e-07, "loss": 0.027, "step": 3545 }, { "epoch": 1.4, "learning_rate": 3.875000310641125e-07, "loss": 0.0394, "step": 3546 }, { "epoch": 1.4, "learning_rate": 3.8749114440398866e-07, "loss": 0.0293, "step": 3547 }, { "epoch": 1.4, "learning_rate": 3.8748225468804407e-07, "loss": 0.0313, "step": 3548 }, { "epoch": 1.4, "learning_rate": 3.874733619164237e-07, "loss": 0.0331, "step": 3549 }, { "epoch": 1.4, "learning_rate": 3.8746446608927246e-07, "loss": 0.0352, "step": 3550 }, { "epoch": 1.4, "learning_rate": 3.8745556720673524e-07, "loss": 0.0359, "step": 3551 }, { "epoch": 1.4, "learning_rate": 3.874466652689573e-07, "loss": 0.0297, "step": 3552 }, { "epoch": 1.4, "learning_rate": 3.8743776027608357e-07, "loss": 0.0313, "step": 3553 }, { "epoch": 1.4, "learning_rate": 3.8742885222825917e-07, "loss": 0.0338, "step": 3554 }, { "epoch": 1.4, "learning_rate": 3.874199411256294e-07, "loss": 0.0325, "step": 3555 }, { "epoch": 1.4, "learning_rate": 3.8741102696833936e-07, "loss": 0.0382, "step": 3556 }, { "epoch": 1.4, "learning_rate": 3.874021097565345e-07, "loss": 0.0337, "step": 3557 }, { "epoch": 1.4, "learning_rate": 3.8739318949036e-07, "loss": 0.0342, "step": 3558 }, { "epoch": 1.4, "learning_rate": 3.873842661699613e-07, "loss": 0.0306, "step": 3559 }, { "epoch": 1.4, "learning_rate": 3.8737533979548395e-07, "loss": 0.0317, "step": 3560 }, { "epoch": 1.4, "learning_rate": 3.873664103670733e-07, "loss": 0.0368, "step": 3561 }, { "epoch": 1.4, "learning_rate": 3.8735747788487483e-07, "loss": 0.0348, "step": 3562 }, { "epoch": 1.4, "learning_rate": 3.8734854234903437e-07, "loss": 0.035, "step": 3563 }, { "epoch": 1.4, "learning_rate": 3.873396037596973e-07, "loss": 0.0332, "step": 3564 }, { "epoch": 1.4, "learning_rate": 3.873306621170095e-07, "loss": 0.0291, "step": 3565 }, { "epoch": 1.41, "learning_rate": 3.873217174211165e-07, "loss": 0.0329, "step": 3566 }, { "epoch": 1.41, "learning_rate": 3.873127696721643e-07, "loss": 0.0291, "step": 3567 }, { "epoch": 1.41, "learning_rate": 3.873038188702986e-07, "loss": 0.0317, "step": 3568 }, { "epoch": 1.41, "learning_rate": 3.8729486501566533e-07, "loss": 0.031, "step": 3569 }, { "epoch": 1.41, "learning_rate": 3.872859081084104e-07, "loss": 0.0328, "step": 3570 }, { "epoch": 1.41, "learning_rate": 3.8727694814867986e-07, "loss": 0.0306, "step": 3571 }, { "epoch": 1.41, "learning_rate": 3.8726798513661963e-07, "loss": 0.0334, "step": 3572 }, { "epoch": 1.41, "learning_rate": 3.872590190723759e-07, "loss": 0.033, "step": 3573 }, { "epoch": 1.41, "learning_rate": 3.8725004995609473e-07, "loss": 0.0331, "step": 3574 }, { "epoch": 1.41, "learning_rate": 3.8724107778792234e-07, "loss": 0.0345, "step": 3575 }, { "epoch": 1.41, "learning_rate": 3.8723210256800495e-07, "loss": 0.0316, "step": 3576 }, { "epoch": 1.41, "learning_rate": 3.8722312429648885e-07, "loss": 0.0319, "step": 3577 }, { "epoch": 1.41, "learning_rate": 3.8721414297352035e-07, "loss": 0.0377, "step": 3578 }, { "epoch": 1.41, "learning_rate": 3.8720515859924595e-07, "loss": 0.037, "step": 3579 }, { "epoch": 1.41, "learning_rate": 3.8719617117381186e-07, "loss": 0.0323, "step": 3580 }, { "epoch": 1.41, "learning_rate": 3.8718718069736477e-07, "loss": 0.0317, "step": 3581 }, { "epoch": 1.41, "learning_rate": 3.871781871700511e-07, "loss": 0.0376, "step": 3582 }, { "epoch": 1.41, "learning_rate": 3.8716919059201744e-07, "loss": 0.0337, "step": 3583 }, { "epoch": 1.41, "learning_rate": 3.871601909634105e-07, "loss": 0.0376, "step": 3584 }, { "epoch": 1.41, "learning_rate": 3.8715118828437685e-07, "loss": 0.0331, "step": 3585 }, { "epoch": 1.41, "learning_rate": 3.871421825550632e-07, "loss": 0.0313, "step": 3586 }, { "epoch": 1.41, "learning_rate": 3.871331737756165e-07, "loss": 0.0327, "step": 3587 }, { "epoch": 1.41, "learning_rate": 3.8712416194618344e-07, "loss": 0.0355, "step": 3588 }, { "epoch": 1.41, "learning_rate": 3.8711514706691096e-07, "loss": 0.0329, "step": 3589 }, { "epoch": 1.41, "learning_rate": 3.871061291379459e-07, "loss": 0.0318, "step": 3590 }, { "epoch": 1.42, "learning_rate": 3.8709710815943543e-07, "loss": 0.0299, "step": 3591 }, { "epoch": 1.42, "learning_rate": 3.870880841315263e-07, "loss": 0.0351, "step": 3592 }, { "epoch": 1.42, "learning_rate": 3.870790570543658e-07, "loss": 0.0336, "step": 3593 }, { "epoch": 1.42, "learning_rate": 3.8707002692810107e-07, "loss": 0.0302, "step": 3594 }, { "epoch": 1.42, "learning_rate": 3.870609937528791e-07, "loss": 0.0317, "step": 3595 }, { "epoch": 1.42, "learning_rate": 3.870519575288473e-07, "loss": 0.0395, "step": 3596 }, { "epoch": 1.42, "learning_rate": 3.870429182561529e-07, "loss": 0.0283, "step": 3597 }, { "epoch": 1.42, "learning_rate": 3.8703387593494313e-07, "loss": 0.0348, "step": 3598 }, { "epoch": 1.42, "learning_rate": 3.8702483056536546e-07, "loss": 0.0355, "step": 3599 }, { "epoch": 1.42, "learning_rate": 3.8701578214756736e-07, "loss": 0.0309, "step": 3600 }, { "epoch": 1.42, "learning_rate": 3.8700673068169616e-07, "loss": 0.0316, "step": 3601 }, { "epoch": 1.42, "learning_rate": 3.869976761678995e-07, "loss": 0.0293, "step": 3602 }, { "epoch": 1.42, "learning_rate": 3.8698861860632494e-07, "loss": 0.0297, "step": 3603 }, { "epoch": 1.42, "learning_rate": 3.8697955799712006e-07, "loss": 0.0351, "step": 3604 }, { "epoch": 1.42, "learning_rate": 3.8697049434043257e-07, "loss": 0.0343, "step": 3605 }, { "epoch": 1.42, "learning_rate": 3.869614276364102e-07, "loss": 0.0313, "step": 3606 }, { "epoch": 1.42, "learning_rate": 3.869523578852007e-07, "loss": 0.0299, "step": 3607 }, { "epoch": 1.42, "learning_rate": 3.8694328508695196e-07, "loss": 0.0357, "step": 3608 }, { "epoch": 1.42, "learning_rate": 3.869342092418117e-07, "loss": 0.0304, "step": 3609 }, { "epoch": 1.42, "learning_rate": 3.8692513034992795e-07, "loss": 0.0322, "step": 3610 }, { "epoch": 1.42, "learning_rate": 3.8691604841144867e-07, "loss": 0.0348, "step": 3611 }, { "epoch": 1.42, "learning_rate": 3.869069634265219e-07, "loss": 0.0279, "step": 3612 }, { "epoch": 1.42, "learning_rate": 3.868978753952957e-07, "loss": 0.033, "step": 3613 }, { "epoch": 1.42, "learning_rate": 3.868887843179182e-07, "loss": 0.0353, "step": 3614 }, { "epoch": 1.42, "learning_rate": 3.868796901945375e-07, "loss": 0.0363, "step": 3615 }, { "epoch": 1.43, "learning_rate": 3.868705930253019e-07, "loss": 0.0284, "step": 3616 }, { "epoch": 1.43, "learning_rate": 3.868614928103596e-07, "loss": 0.0305, "step": 3617 }, { "epoch": 1.43, "learning_rate": 3.8685238954985904e-07, "loss": 0.037, "step": 3618 }, { "epoch": 1.43, "learning_rate": 3.868432832439485e-07, "loss": 0.0317, "step": 3619 }, { "epoch": 1.43, "learning_rate": 3.868341738927764e-07, "loss": 0.027, "step": 3620 }, { "epoch": 1.43, "learning_rate": 3.8682506149649123e-07, "loss": 0.0312, "step": 3621 }, { "epoch": 1.43, "learning_rate": 3.8681594605524145e-07, "loss": 0.0337, "step": 3622 }, { "epoch": 1.43, "learning_rate": 3.8680682756917574e-07, "loss": 0.0301, "step": 3623 }, { "epoch": 1.43, "learning_rate": 3.867977060384426e-07, "loss": 0.0329, "step": 3624 }, { "epoch": 1.43, "learning_rate": 3.867885814631908e-07, "loss": 0.0394, "step": 3625 }, { "epoch": 1.43, "learning_rate": 3.8677945384356895e-07, "loss": 0.0323, "step": 3626 }, { "epoch": 1.43, "learning_rate": 3.867703231797259e-07, "loss": 0.0326, "step": 3627 }, { "epoch": 1.43, "learning_rate": 3.867611894718105e-07, "loss": 0.0288, "step": 3628 }, { "epoch": 1.43, "learning_rate": 3.8675205271997147e-07, "loss": 0.0311, "step": 3629 }, { "epoch": 1.43, "learning_rate": 3.8674291292435785e-07, "loss": 0.03, "step": 3630 }, { "epoch": 1.43, "learning_rate": 3.867337700851186e-07, "loss": 0.0316, "step": 3631 }, { "epoch": 1.43, "learning_rate": 3.867246242024026e-07, "loss": 0.0347, "step": 3632 }, { "epoch": 1.43, "learning_rate": 3.867154752763591e-07, "loss": 0.0351, "step": 3633 }, { "epoch": 1.43, "learning_rate": 3.867063233071371e-07, "loss": 0.0338, "step": 3634 }, { "epoch": 1.43, "learning_rate": 3.866971682948858e-07, "loss": 0.0344, "step": 3635 }, { "epoch": 1.43, "learning_rate": 3.8668801023975443e-07, "loss": 0.033, "step": 3636 }, { "epoch": 1.43, "learning_rate": 3.866788491418922e-07, "loss": 0.0325, "step": 3637 }, { "epoch": 1.43, "learning_rate": 3.8666968500144846e-07, "loss": 0.0334, "step": 3638 }, { "epoch": 1.43, "learning_rate": 3.8666051781857255e-07, "loss": 0.0356, "step": 3639 }, { "epoch": 1.43, "learning_rate": 3.8665134759341387e-07, "loss": 0.0352, "step": 3640 }, { "epoch": 1.43, "learning_rate": 3.866421743261219e-07, "loss": 0.033, "step": 3641 }, { "epoch": 1.44, "learning_rate": 3.8663299801684615e-07, "loss": 0.0309, "step": 3642 }, { "epoch": 1.44, "learning_rate": 3.866238186657362e-07, "loss": 0.0324, "step": 3643 }, { "epoch": 1.44, "learning_rate": 3.866146362729417e-07, "loss": 0.0331, "step": 3644 }, { "epoch": 1.44, "learning_rate": 3.866054508386122e-07, "loss": 0.0286, "step": 3645 }, { "epoch": 1.44, "learning_rate": 3.8659626236289737e-07, "loss": 0.0325, "step": 3646 }, { "epoch": 1.44, "learning_rate": 3.8658707084594714e-07, "loss": 0.0344, "step": 3647 }, { "epoch": 1.44, "learning_rate": 3.8657787628791125e-07, "loss": 0.0324, "step": 3648 }, { "epoch": 1.44, "learning_rate": 3.8656867868893956e-07, "loss": 0.0331, "step": 3649 }, { "epoch": 1.44, "learning_rate": 3.865594780491819e-07, "loss": 0.0344, "step": 3650 }, { "epoch": 1.44, "learning_rate": 3.8655027436878826e-07, "loss": 0.0334, "step": 3651 }, { "epoch": 1.44, "learning_rate": 3.8654106764790876e-07, "loss": 0.0283, "step": 3652 }, { "epoch": 1.44, "learning_rate": 3.8653185788669333e-07, "loss": 0.0353, "step": 3653 }, { "epoch": 1.44, "learning_rate": 3.865226450852921e-07, "loss": 0.032, "step": 3654 }, { "epoch": 1.44, "learning_rate": 3.8651342924385526e-07, "loss": 0.0334, "step": 3655 }, { "epoch": 1.44, "learning_rate": 3.8650421036253296e-07, "loss": 0.0289, "step": 3656 }, { "epoch": 1.44, "learning_rate": 3.8649498844147553e-07, "loss": 0.0357, "step": 3657 }, { "epoch": 1.44, "learning_rate": 3.864857634808332e-07, "loss": 0.0327, "step": 3658 }, { "epoch": 1.44, "learning_rate": 3.8647653548075635e-07, "loss": 0.0345, "step": 3659 }, { "epoch": 1.44, "learning_rate": 3.864673044413954e-07, "loss": 0.0334, "step": 3660 }, { "epoch": 1.44, "learning_rate": 3.8645807036290075e-07, "loss": 0.0304, "step": 3661 }, { "epoch": 1.44, "learning_rate": 3.86448833245423e-07, "loss": 0.0318, "step": 3662 }, { "epoch": 1.44, "learning_rate": 3.864395930891126e-07, "loss": 0.036, "step": 3663 }, { "epoch": 1.44, "learning_rate": 3.8643034989412014e-07, "loss": 0.03, "step": 3664 }, { "epoch": 1.44, "learning_rate": 3.8642110366059636e-07, "loss": 0.0254, "step": 3665 }, { "epoch": 1.44, "learning_rate": 3.864118543886919e-07, "loss": 0.0359, "step": 3666 }, { "epoch": 1.45, "learning_rate": 3.864026020785576e-07, "loss": 0.0323, "step": 3667 }, { "epoch": 1.45, "learning_rate": 3.8639334673034407e-07, "loss": 0.033, "step": 3668 }, { "epoch": 1.45, "learning_rate": 3.863840883442024e-07, "loss": 0.0332, "step": 3669 }, { "epoch": 1.45, "learning_rate": 3.8637482692028327e-07, "loss": 0.0315, "step": 3670 }, { "epoch": 1.45, "learning_rate": 3.863655624587377e-07, "loss": 0.0347, "step": 3671 }, { "epoch": 1.45, "learning_rate": 3.863562949597168e-07, "loss": 0.0323, "step": 3672 }, { "epoch": 1.45, "learning_rate": 3.8634702442337144e-07, "loss": 0.0272, "step": 3673 }, { "epoch": 1.45, "learning_rate": 3.863377508498528e-07, "loss": 0.027, "step": 3674 }, { "epoch": 1.45, "learning_rate": 3.863284742393121e-07, "loss": 0.0362, "step": 3675 }, { "epoch": 1.45, "learning_rate": 3.863191945919004e-07, "loss": 0.0302, "step": 3676 }, { "epoch": 1.45, "learning_rate": 3.86309911907769e-07, "loss": 0.0305, "step": 3677 }, { "epoch": 1.45, "learning_rate": 3.863006261870692e-07, "loss": 0.0318, "step": 3678 }, { "epoch": 1.45, "learning_rate": 3.8629133742995237e-07, "loss": 0.0299, "step": 3679 }, { "epoch": 1.45, "learning_rate": 3.862820456365698e-07, "loss": 0.0408, "step": 3680 }, { "epoch": 1.45, "learning_rate": 3.8627275080707306e-07, "loss": 0.0316, "step": 3681 }, { "epoch": 1.45, "learning_rate": 3.8626345294161354e-07, "loss": 0.0364, "step": 3682 }, { "epoch": 1.45, "learning_rate": 3.862541520403429e-07, "loss": 0.0308, "step": 3683 }, { "epoch": 1.45, "learning_rate": 3.8624484810341256e-07, "loss": 0.0318, "step": 3684 }, { "epoch": 1.45, "learning_rate": 3.862355411309743e-07, "loss": 0.0334, "step": 3685 }, { "epoch": 1.45, "learning_rate": 3.862262311231798e-07, "loss": 0.0348, "step": 3686 }, { "epoch": 1.45, "learning_rate": 3.862169180801807e-07, "loss": 0.03, "step": 3687 }, { "epoch": 1.45, "learning_rate": 3.8620760200212886e-07, "loss": 0.0342, "step": 3688 }, { "epoch": 1.45, "learning_rate": 3.8619828288917614e-07, "loss": 0.0294, "step": 3689 }, { "epoch": 1.45, "learning_rate": 3.8618896074147436e-07, "loss": 0.0368, "step": 3690 }, { "epoch": 1.45, "learning_rate": 3.861796355591756e-07, "loss": 0.0301, "step": 3691 }, { "epoch": 1.46, "learning_rate": 3.861703073424317e-07, "loss": 0.0314, "step": 3692 }, { "epoch": 1.46, "learning_rate": 3.8616097609139465e-07, "loss": 0.0319, "step": 3693 }, { "epoch": 1.46, "learning_rate": 3.861516418062166e-07, "loss": 0.0304, "step": 3694 }, { "epoch": 1.46, "learning_rate": 3.8614230448704986e-07, "loss": 0.0347, "step": 3695 }, { "epoch": 1.46, "learning_rate": 3.861329641340464e-07, "loss": 0.0315, "step": 3696 }, { "epoch": 1.46, "learning_rate": 3.8612362074735847e-07, "loss": 0.0285, "step": 3697 }, { "epoch": 1.46, "learning_rate": 3.8611427432713837e-07, "loss": 0.0292, "step": 3698 }, { "epoch": 1.46, "learning_rate": 3.8610492487353853e-07, "loss": 0.0343, "step": 3699 }, { "epoch": 1.46, "learning_rate": 3.860955723867112e-07, "loss": 0.0292, "step": 3700 }, { "epoch": 1.46, "learning_rate": 3.86086216866809e-07, "loss": 0.0276, "step": 3701 }, { "epoch": 1.46, "learning_rate": 3.860768583139841e-07, "loss": 0.0345, "step": 3702 }, { "epoch": 1.46, "learning_rate": 3.860674967283894e-07, "loss": 0.0321, "step": 3703 }, { "epoch": 1.46, "learning_rate": 3.8605813211017713e-07, "loss": 0.0365, "step": 3704 }, { "epoch": 1.46, "learning_rate": 3.860487644595002e-07, "loss": 0.0294, "step": 3705 }, { "epoch": 1.46, "learning_rate": 3.8603939377651116e-07, "loss": 0.03, "step": 3706 }, { "epoch": 1.46, "learning_rate": 3.860300200613627e-07, "loss": 0.0261, "step": 3707 }, { "epoch": 1.46, "learning_rate": 3.8602064331420763e-07, "loss": 0.0405, "step": 3708 }, { "epoch": 1.46, "learning_rate": 3.8601126353519886e-07, "loss": 0.0338, "step": 3709 }, { "epoch": 1.46, "learning_rate": 3.860018807244891e-07, "loss": 0.0352, "step": 3710 }, { "epoch": 1.46, "learning_rate": 3.8599249488223153e-07, "loss": 0.0329, "step": 3711 }, { "epoch": 1.46, "learning_rate": 3.8598310600857887e-07, "loss": 0.0351, "step": 3712 }, { "epoch": 1.46, "learning_rate": 3.859737141036842e-07, "loss": 0.0292, "step": 3713 }, { "epoch": 1.46, "learning_rate": 3.8596431916770076e-07, "loss": 0.0265, "step": 3714 }, { "epoch": 1.46, "learning_rate": 3.859549212007814e-07, "loss": 0.0283, "step": 3715 }, { "epoch": 1.46, "learning_rate": 3.859455202030796e-07, "loss": 0.0309, "step": 3716 }, { "epoch": 1.46, "learning_rate": 3.8593611617474837e-07, "loss": 0.0298, "step": 3717 }, { "epoch": 1.47, "learning_rate": 3.8592670911594096e-07, "loss": 0.0299, "step": 3718 }, { "epoch": 1.47, "learning_rate": 3.859172990268108e-07, "loss": 0.0319, "step": 3719 }, { "epoch": 1.47, "learning_rate": 3.859078859075113e-07, "loss": 0.0365, "step": 3720 }, { "epoch": 1.47, "learning_rate": 3.858984697581958e-07, "loss": 0.0335, "step": 3721 }, { "epoch": 1.47, "learning_rate": 3.8588905057901766e-07, "loss": 0.0292, "step": 3722 }, { "epoch": 1.47, "learning_rate": 3.858796283701306e-07, "loss": 0.03, "step": 3723 }, { "epoch": 1.47, "learning_rate": 3.858702031316881e-07, "loss": 0.0303, "step": 3724 }, { "epoch": 1.47, "learning_rate": 3.8586077486384374e-07, "loss": 0.0292, "step": 3725 }, { "epoch": 1.47, "learning_rate": 3.858513435667513e-07, "loss": 0.0311, "step": 3726 }, { "epoch": 1.47, "learning_rate": 3.8584190924056435e-07, "loss": 0.0302, "step": 3727 }, { "epoch": 1.47, "learning_rate": 3.8583247188543676e-07, "loss": 0.0325, "step": 3728 }, { "epoch": 1.47, "learning_rate": 3.858230315015223e-07, "loss": 0.0312, "step": 3729 }, { "epoch": 1.47, "learning_rate": 3.8581358808897486e-07, "loss": 0.0314, "step": 3730 }, { "epoch": 1.47, "learning_rate": 3.8580414164794835e-07, "loss": 0.0341, "step": 3731 }, { "epoch": 1.47, "learning_rate": 3.857946921785967e-07, "loss": 0.0343, "step": 3732 }, { "epoch": 1.47, "learning_rate": 3.857852396810739e-07, "loss": 0.0354, "step": 3733 }, { "epoch": 1.47, "learning_rate": 3.857757841555341e-07, "loss": 0.0343, "step": 3734 }, { "epoch": 1.47, "learning_rate": 3.8576632560213137e-07, "loss": 0.0345, "step": 3735 }, { "epoch": 1.47, "learning_rate": 3.857568640210199e-07, "loss": 0.0366, "step": 3736 }, { "epoch": 1.47, "learning_rate": 3.8574739941235376e-07, "loss": 0.0335, "step": 3737 }, { "epoch": 1.47, "learning_rate": 3.857379317762874e-07, "loss": 0.0323, "step": 3738 }, { "epoch": 1.47, "learning_rate": 3.85728461112975e-07, "loss": 0.0331, "step": 3739 }, { "epoch": 1.47, "learning_rate": 3.85718987422571e-07, "loss": 0.0289, "step": 3740 }, { "epoch": 1.47, "learning_rate": 3.8570951070522973e-07, "loss": 0.0325, "step": 3741 }, { "epoch": 1.47, "learning_rate": 3.857000309611057e-07, "loss": 0.0312, "step": 3742 }, { "epoch": 1.48, "learning_rate": 3.856905481903534e-07, "loss": 0.0295, "step": 3743 }, { "epoch": 1.48, "learning_rate": 3.8568106239312735e-07, "loss": 0.0335, "step": 3744 }, { "epoch": 1.48, "learning_rate": 3.856715735695822e-07, "loss": 0.0259, "step": 3745 }, { "epoch": 1.48, "learning_rate": 3.8566208171987255e-07, "loss": 0.0353, "step": 3746 }, { "epoch": 1.48, "learning_rate": 3.856525868441532e-07, "loss": 0.0328, "step": 3747 }, { "epoch": 1.48, "learning_rate": 3.8564308894257883e-07, "loss": 0.0343, "step": 3748 }, { "epoch": 1.48, "learning_rate": 3.856335880153043e-07, "loss": 0.0298, "step": 3749 }, { "epoch": 1.48, "learning_rate": 3.856240840624843e-07, "loss": 0.0337, "step": 3750 }, { "epoch": 1.48, "learning_rate": 3.8561457708427396e-07, "loss": 0.0299, "step": 3751 }, { "epoch": 1.48, "learning_rate": 3.8560506708082804e-07, "loss": 0.0333, "step": 3752 }, { "epoch": 1.48, "learning_rate": 3.855955540523017e-07, "loss": 0.0274, "step": 3753 }, { "epoch": 1.48, "learning_rate": 3.855860379988498e-07, "loss": 0.0364, "step": 3754 }, { "epoch": 1.48, "learning_rate": 3.8557651892062763e-07, "loss": 0.0318, "step": 3755 }, { "epoch": 1.48, "learning_rate": 3.855669968177902e-07, "loss": 0.0327, "step": 3756 }, { "epoch": 1.48, "learning_rate": 3.855574716904927e-07, "loss": 0.0306, "step": 3757 }, { "epoch": 1.48, "learning_rate": 3.855479435388905e-07, "loss": 0.0349, "step": 3758 }, { "epoch": 1.48, "learning_rate": 3.8553841236313886e-07, "loss": 0.0325, "step": 3759 }, { "epoch": 1.48, "learning_rate": 3.8552887816339296e-07, "loss": 0.0335, "step": 3760 }, { "epoch": 1.48, "learning_rate": 3.8551934093980843e-07, "loss": 0.0315, "step": 3761 }, { "epoch": 1.48, "learning_rate": 3.855098006925405e-07, "loss": 0.0326, "step": 3762 }, { "epoch": 1.48, "learning_rate": 3.8550025742174483e-07, "loss": 0.0303, "step": 3763 }, { "epoch": 1.48, "learning_rate": 3.8549071112757685e-07, "loss": 0.0329, "step": 3764 }, { "epoch": 1.48, "learning_rate": 3.854811618101922e-07, "loss": 0.0335, "step": 3765 }, { "epoch": 1.48, "learning_rate": 3.8547160946974656e-07, "loss": 0.0296, "step": 3766 }, { "epoch": 1.48, "learning_rate": 3.8546205410639554e-07, "loss": 0.031, "step": 3767 }, { "epoch": 1.48, "learning_rate": 3.854524957202949e-07, "loss": 0.0314, "step": 3768 }, { "epoch": 1.49, "learning_rate": 3.8544293431160044e-07, "loss": 0.0325, "step": 3769 }, { "epoch": 1.49, "learning_rate": 3.85433369880468e-07, "loss": 0.033, "step": 3770 }, { "epoch": 1.49, "learning_rate": 3.854238024270534e-07, "loss": 0.0342, "step": 3771 }, { "epoch": 1.49, "learning_rate": 3.8541423195151277e-07, "loss": 0.0346, "step": 3772 }, { "epoch": 1.49, "learning_rate": 3.8540465845400183e-07, "loss": 0.0304, "step": 3773 }, { "epoch": 1.49, "learning_rate": 3.8539508193467676e-07, "loss": 0.0316, "step": 3774 }, { "epoch": 1.49, "learning_rate": 3.8538550239369364e-07, "loss": 0.0302, "step": 3775 }, { "epoch": 1.49, "learning_rate": 3.8537591983120857e-07, "loss": 0.0338, "step": 3776 }, { "epoch": 1.49, "learning_rate": 3.8536633424737783e-07, "loss": 0.0327, "step": 3777 }, { "epoch": 1.49, "learning_rate": 3.8535674564235744e-07, "loss": 0.0278, "step": 3778 }, { "epoch": 1.49, "learning_rate": 3.853471540163039e-07, "loss": 0.0288, "step": 3779 }, { "epoch": 1.49, "learning_rate": 3.8533755936937335e-07, "loss": 0.0335, "step": 3780 }, { "epoch": 1.49, "learning_rate": 3.8532796170172237e-07, "loss": 0.031, "step": 3781 }, { "epoch": 1.49, "learning_rate": 3.853183610135072e-07, "loss": 0.0319, "step": 3782 }, { "epoch": 1.49, "learning_rate": 3.853087573048844e-07, "loss": 0.032, "step": 3783 }, { "epoch": 1.49, "learning_rate": 3.852991505760106e-07, "loss": 0.0306, "step": 3784 }, { "epoch": 1.49, "learning_rate": 3.852895408270421e-07, "loss": 0.0344, "step": 3785 }, { "epoch": 1.49, "learning_rate": 3.852799280581358e-07, "loss": 0.0339, "step": 3786 }, { "epoch": 1.49, "learning_rate": 3.852703122694483e-07, "loss": 0.0283, "step": 3787 }, { "epoch": 1.49, "learning_rate": 3.852606934611362e-07, "loss": 0.026, "step": 3788 }, { "epoch": 1.49, "learning_rate": 3.852510716333564e-07, "loss": 0.0371, "step": 3789 }, { "epoch": 1.49, "learning_rate": 3.8524144678626567e-07, "loss": 0.0309, "step": 3790 }, { "epoch": 1.49, "learning_rate": 3.852318189200209e-07, "loss": 0.0314, "step": 3791 }, { "epoch": 1.49, "learning_rate": 3.85222188034779e-07, "loss": 0.0334, "step": 3792 }, { "epoch": 1.49, "learning_rate": 3.852125541306969e-07, "loss": 0.0309, "step": 3793 }, { "epoch": 1.5, "learning_rate": 3.8520291720793174e-07, "loss": 0.0324, "step": 3794 }, { "epoch": 1.5, "learning_rate": 3.8519327726664047e-07, "loss": 0.0331, "step": 3795 }, { "epoch": 1.5, "learning_rate": 3.851836343069802e-07, "loss": 0.0305, "step": 3796 }, { "epoch": 1.5, "learning_rate": 3.851739883291082e-07, "loss": 0.0317, "step": 3797 }, { "epoch": 1.5, "learning_rate": 3.851643393331816e-07, "loss": 0.028, "step": 3798 }, { "epoch": 1.5, "learning_rate": 3.8515468731935767e-07, "loss": 0.03, "step": 3799 }, { "epoch": 1.5, "learning_rate": 3.8514503228779375e-07, "loss": 0.0313, "step": 3800 }, { "epoch": 1.5, "learning_rate": 3.8513537423864716e-07, "loss": 0.0336, "step": 3801 }, { "epoch": 1.5, "learning_rate": 3.851257131720754e-07, "loss": 0.0349, "step": 3802 }, { "epoch": 1.5, "learning_rate": 3.8511604908823586e-07, "loss": 0.0311, "step": 3803 }, { "epoch": 1.5, "learning_rate": 3.85106381987286e-07, "loss": 0.033, "step": 3804 }, { "epoch": 1.5, "learning_rate": 3.8509671186938356e-07, "loss": 0.0326, "step": 3805 }, { "epoch": 1.5, "learning_rate": 3.8508703873468596e-07, "loss": 0.0364, "step": 3806 }, { "epoch": 1.5, "learning_rate": 3.8507736258335097e-07, "loss": 0.0309, "step": 3807 }, { "epoch": 1.5, "learning_rate": 3.850676834155362e-07, "loss": 0.0315, "step": 3808 }, { "epoch": 1.5, "learning_rate": 3.850580012313995e-07, "loss": 0.0297, "step": 3809 }, { "epoch": 1.5, "learning_rate": 3.8504831603109855e-07, "loss": 0.0278, "step": 3810 }, { "epoch": 1.5, "learning_rate": 3.850386278147914e-07, "loss": 0.033, "step": 3811 }, { "epoch": 1.5, "learning_rate": 3.850289365826358e-07, "loss": 0.0303, "step": 3812 }, { "epoch": 1.5, "learning_rate": 3.850192423347897e-07, "loss": 0.031, "step": 3813 }, { "epoch": 1.5, "learning_rate": 3.8500954507141125e-07, "loss": 0.0298, "step": 3814 }, { "epoch": 1.5, "learning_rate": 3.849998447926583e-07, "loss": 0.0289, "step": 3815 }, { "epoch": 1.5, "learning_rate": 3.849901414986891e-07, "loss": 0.0343, "step": 3816 }, { "epoch": 1.5, "learning_rate": 3.8498043518966173e-07, "loss": 0.0309, "step": 3817 }, { "epoch": 1.5, "learning_rate": 3.849707258657344e-07, "loss": 0.0311, "step": 3818 }, { "epoch": 1.51, "learning_rate": 3.849610135270654e-07, "loss": 0.0337, "step": 3819 }, { "epoch": 1.51, "learning_rate": 3.8495129817381297e-07, "loss": 0.032, "step": 3820 }, { "epoch": 1.51, "learning_rate": 3.849415798061354e-07, "loss": 0.0326, "step": 3821 }, { "epoch": 1.51, "learning_rate": 3.8493185842419125e-07, "loss": 0.035, "step": 3822 }, { "epoch": 1.51, "learning_rate": 3.849221340281388e-07, "loss": 0.0334, "step": 3823 }, { "epoch": 1.51, "learning_rate": 3.8491240661813666e-07, "loss": 0.0358, "step": 3824 }, { "epoch": 1.51, "learning_rate": 3.849026761943434e-07, "loss": 0.0305, "step": 3825 }, { "epoch": 1.51, "learning_rate": 3.848929427569174e-07, "loss": 0.0347, "step": 3826 }, { "epoch": 1.51, "learning_rate": 3.848832063060175e-07, "loss": 0.0323, "step": 3827 }, { "epoch": 1.51, "learning_rate": 3.848734668418023e-07, "loss": 0.0359, "step": 3828 }, { "epoch": 1.51, "learning_rate": 3.8486372436443066e-07, "loss": 0.029, "step": 3829 }, { "epoch": 1.51, "learning_rate": 3.8485397887406117e-07, "loss": 0.034, "step": 3830 }, { "epoch": 1.51, "learning_rate": 3.8484423037085277e-07, "loss": 0.0308, "step": 3831 }, { "epoch": 1.51, "learning_rate": 3.848344788549644e-07, "loss": 0.0288, "step": 3832 }, { "epoch": 1.51, "learning_rate": 3.8482472432655493e-07, "loss": 0.0347, "step": 3833 }, { "epoch": 1.51, "learning_rate": 3.8481496678578334e-07, "loss": 0.0374, "step": 3834 }, { "epoch": 1.51, "learning_rate": 3.8480520623280864e-07, "loss": 0.0277, "step": 3835 }, { "epoch": 1.51, "learning_rate": 3.8479544266779007e-07, "loss": 0.0298, "step": 3836 }, { "epoch": 1.51, "learning_rate": 3.847856760908865e-07, "loss": 0.0286, "step": 3837 }, { "epoch": 1.51, "learning_rate": 3.8477590650225735e-07, "loss": 0.0295, "step": 3838 }, { "epoch": 1.51, "learning_rate": 3.8476613390206174e-07, "loss": 0.0301, "step": 3839 }, { "epoch": 1.51, "learning_rate": 3.847563582904589e-07, "loss": 0.0312, "step": 3840 }, { "epoch": 1.51, "learning_rate": 3.847465796676082e-07, "loss": 0.0279, "step": 3841 }, { "epoch": 1.51, "learning_rate": 3.8473679803366916e-07, "loss": 0.0301, "step": 3842 }, { "epoch": 1.51, "learning_rate": 3.84727013388801e-07, "loss": 0.0326, "step": 3843 }, { "epoch": 1.51, "learning_rate": 3.8471722573316325e-07, "loss": 0.0381, "step": 3844 }, { "epoch": 1.52, "learning_rate": 3.847074350669155e-07, "loss": 0.037, "step": 3845 }, { "epoch": 1.52, "learning_rate": 3.8469764139021724e-07, "loss": 0.0329, "step": 3846 }, { "epoch": 1.52, "learning_rate": 3.846878447032282e-07, "loss": 0.0373, "step": 3847 }, { "epoch": 1.52, "learning_rate": 3.84678045006108e-07, "loss": 0.0316, "step": 3848 }, { "epoch": 1.52, "learning_rate": 3.8466824229901623e-07, "loss": 0.0299, "step": 3849 }, { "epoch": 1.52, "learning_rate": 3.8465843658211285e-07, "loss": 0.0301, "step": 3850 }, { "epoch": 1.52, "learning_rate": 3.846486278555576e-07, "loss": 0.0264, "step": 3851 }, { "epoch": 1.52, "learning_rate": 3.8463881611951034e-07, "loss": 0.0275, "step": 3852 }, { "epoch": 1.52, "learning_rate": 3.84629001374131e-07, "loss": 0.0267, "step": 3853 }, { "epoch": 1.52, "learning_rate": 3.846191836195795e-07, "loss": 0.0309, "step": 3854 }, { "epoch": 1.52, "learning_rate": 3.846093628560159e-07, "loss": 0.0311, "step": 3855 }, { "epoch": 1.52, "learning_rate": 3.845995390836003e-07, "loss": 0.031, "step": 3856 }, { "epoch": 1.52, "learning_rate": 3.8458971230249277e-07, "loss": 0.0268, "step": 3857 }, { "epoch": 1.52, "learning_rate": 3.845798825128535e-07, "loss": 0.0334, "step": 3858 }, { "epoch": 1.52, "learning_rate": 3.8457004971484263e-07, "loss": 0.0319, "step": 3859 }, { "epoch": 1.52, "learning_rate": 3.845602139086205e-07, "loss": 0.029, "step": 3860 }, { "epoch": 1.52, "learning_rate": 3.845503750943473e-07, "loss": 0.032, "step": 3861 }, { "epoch": 1.52, "learning_rate": 3.845405332721835e-07, "loss": 0.0299, "step": 3862 }, { "epoch": 1.52, "learning_rate": 3.8453068844228955e-07, "loss": 0.0297, "step": 3863 }, { "epoch": 1.52, "learning_rate": 3.845208406048258e-07, "loss": 0.0328, "step": 3864 }, { "epoch": 1.52, "learning_rate": 3.8451098975995275e-07, "loss": 0.0316, "step": 3865 }, { "epoch": 1.52, "learning_rate": 3.8450113590783104e-07, "loss": 0.0315, "step": 3866 }, { "epoch": 1.52, "learning_rate": 3.8449127904862123e-07, "loss": 0.0344, "step": 3867 }, { "epoch": 1.52, "learning_rate": 3.844814191824839e-07, "loss": 0.0301, "step": 3868 }, { "epoch": 1.52, "learning_rate": 3.844715563095799e-07, "loss": 0.0341, "step": 3869 }, { "epoch": 1.53, "learning_rate": 3.844616904300698e-07, "loss": 0.0295, "step": 3870 }, { "epoch": 1.53, "learning_rate": 3.8445182154411456e-07, "loss": 0.032, "step": 3871 }, { "epoch": 1.53, "learning_rate": 3.844419496518749e-07, "loss": 0.032, "step": 3872 }, { "epoch": 1.53, "learning_rate": 3.844320747535119e-07, "loss": 0.0367, "step": 3873 }, { "epoch": 1.53, "learning_rate": 3.844221968491863e-07, "loss": 0.0348, "step": 3874 }, { "epoch": 1.53, "learning_rate": 3.844123159390592e-07, "loss": 0.0286, "step": 3875 }, { "epoch": 1.53, "learning_rate": 3.8440243202329156e-07, "loss": 0.0273, "step": 3876 }, { "epoch": 1.53, "learning_rate": 3.8439254510204464e-07, "loss": 0.0333, "step": 3877 }, { "epoch": 1.53, "learning_rate": 3.843826551754794e-07, "loss": 0.0266, "step": 3878 }, { "epoch": 1.53, "learning_rate": 3.843727622437572e-07, "loss": 0.0339, "step": 3879 }, { "epoch": 1.53, "learning_rate": 3.843628663070391e-07, "loss": 0.0289, "step": 3880 }, { "epoch": 1.53, "learning_rate": 3.8435296736548655e-07, "loss": 0.0307, "step": 3881 }, { "epoch": 1.53, "learning_rate": 3.8434306541926073e-07, "loss": 0.0323, "step": 3882 }, { "epoch": 1.53, "learning_rate": 3.843331604685232e-07, "loss": 0.0281, "step": 3883 }, { "epoch": 1.53, "learning_rate": 3.8432325251343525e-07, "loss": 0.031, "step": 3884 }, { "epoch": 1.53, "learning_rate": 3.8431334155415847e-07, "loss": 0.0325, "step": 3885 }, { "epoch": 1.53, "learning_rate": 3.843034275908543e-07, "loss": 0.0284, "step": 3886 }, { "epoch": 1.53, "learning_rate": 3.8429351062368437e-07, "loss": 0.031, "step": 3887 }, { "epoch": 1.53, "learning_rate": 3.842835906528103e-07, "loss": 0.03, "step": 3888 }, { "epoch": 1.53, "learning_rate": 3.842736676783938e-07, "loss": 0.034, "step": 3889 }, { "epoch": 1.53, "learning_rate": 3.842637417005966e-07, "loss": 0.0293, "step": 3890 }, { "epoch": 1.53, "learning_rate": 3.842538127195805e-07, "loss": 0.0329, "step": 3891 }, { "epoch": 1.53, "learning_rate": 3.842438807355072e-07, "loss": 0.0298, "step": 3892 }, { "epoch": 1.53, "learning_rate": 3.842339457485387e-07, "loss": 0.0326, "step": 3893 }, { "epoch": 1.53, "learning_rate": 3.842240077588369e-07, "loss": 0.034, "step": 3894 }, { "epoch": 1.54, "learning_rate": 3.8421406676656376e-07, "loss": 0.0278, "step": 3895 }, { "epoch": 1.54, "learning_rate": 3.842041227718813e-07, "loss": 0.0325, "step": 3896 }, { "epoch": 1.54, "learning_rate": 3.8419417577495156e-07, "loss": 0.034, "step": 3897 }, { "epoch": 1.54, "learning_rate": 3.841842257759368e-07, "loss": 0.033, "step": 3898 }, { "epoch": 1.54, "learning_rate": 3.84174272774999e-07, "loss": 0.0275, "step": 3899 }, { "epoch": 1.54, "learning_rate": 3.8416431677230045e-07, "loss": 0.0287, "step": 3900 }, { "epoch": 1.54, "learning_rate": 3.841543577680035e-07, "loss": 0.0327, "step": 3901 }, { "epoch": 1.54, "learning_rate": 3.8414439576227037e-07, "loss": 0.0275, "step": 3902 }, { "epoch": 1.54, "learning_rate": 3.841344307552634e-07, "loss": 0.0274, "step": 3903 }, { "epoch": 1.54, "learning_rate": 3.8412446274714517e-07, "loss": 0.032, "step": 3904 }, { "epoch": 1.54, "learning_rate": 3.8411449173807794e-07, "loss": 0.0294, "step": 3905 }, { "epoch": 1.54, "learning_rate": 3.841045177282244e-07, "loss": 0.0334, "step": 3906 }, { "epoch": 1.54, "learning_rate": 3.8409454071774697e-07, "loss": 0.0317, "step": 3907 }, { "epoch": 1.54, "learning_rate": 3.8408456070680834e-07, "loss": 0.0331, "step": 3908 }, { "epoch": 1.54, "learning_rate": 3.840745776955711e-07, "loss": 0.0277, "step": 3909 }, { "epoch": 1.54, "learning_rate": 3.84064591684198e-07, "loss": 0.0345, "step": 3910 }, { "epoch": 1.54, "learning_rate": 3.840546026728519e-07, "loss": 0.031, "step": 3911 }, { "epoch": 1.54, "learning_rate": 3.840446106616954e-07, "loss": 0.0316, "step": 3912 }, { "epoch": 1.54, "learning_rate": 3.8403461565089155e-07, "loss": 0.0322, "step": 3913 }, { "epoch": 1.54, "learning_rate": 3.8402461764060317e-07, "loss": 0.0329, "step": 3914 }, { "epoch": 1.54, "learning_rate": 3.840146166309932e-07, "loss": 0.031, "step": 3915 }, { "epoch": 1.54, "learning_rate": 3.840046126222245e-07, "loss": 0.0294, "step": 3916 }, { "epoch": 1.54, "learning_rate": 3.839946056144605e-07, "loss": 0.0332, "step": 3917 }, { "epoch": 1.54, "learning_rate": 3.8398459560786384e-07, "loss": 0.0321, "step": 3918 }, { "epoch": 1.54, "learning_rate": 3.8397458260259805e-07, "loss": 0.0276, "step": 3919 }, { "epoch": 1.54, "learning_rate": 3.839645665988262e-07, "loss": 0.0267, "step": 3920 }, { "epoch": 1.55, "learning_rate": 3.839545475967114e-07, "loss": 0.0307, "step": 3921 }, { "epoch": 1.55, "learning_rate": 3.8394452559641706e-07, "loss": 0.0277, "step": 3922 }, { "epoch": 1.55, "learning_rate": 3.839345005981065e-07, "loss": 0.0369, "step": 3923 }, { "epoch": 1.55, "learning_rate": 3.839244726019432e-07, "loss": 0.0309, "step": 3924 }, { "epoch": 1.55, "learning_rate": 3.839144416080906e-07, "loss": 0.0318, "step": 3925 }, { "epoch": 1.55, "learning_rate": 3.8390440761671197e-07, "loss": 0.0314, "step": 3926 }, { "epoch": 1.55, "learning_rate": 3.8389437062797106e-07, "loss": 0.0347, "step": 3927 }, { "epoch": 1.55, "learning_rate": 3.8388433064203134e-07, "loss": 0.0321, "step": 3928 }, { "epoch": 1.55, "learning_rate": 3.838742876590566e-07, "loss": 0.0299, "step": 3929 }, { "epoch": 1.55, "learning_rate": 3.838642416792104e-07, "loss": 0.0353, "step": 3930 }, { "epoch": 1.55, "learning_rate": 3.8385419270265645e-07, "loss": 0.0337, "step": 3931 }, { "epoch": 1.55, "learning_rate": 3.838441407295586e-07, "loss": 0.0302, "step": 3932 }, { "epoch": 1.55, "learning_rate": 3.838340857600807e-07, "loss": 0.0314, "step": 3933 }, { "epoch": 1.55, "learning_rate": 3.8382402779438657e-07, "loss": 0.0309, "step": 3934 }, { "epoch": 1.55, "learning_rate": 3.838139668326402e-07, "loss": 0.0287, "step": 3935 }, { "epoch": 1.55, "learning_rate": 3.8380390287500547e-07, "loss": 0.0281, "step": 3936 }, { "epoch": 1.55, "learning_rate": 3.8379383592164656e-07, "loss": 0.034, "step": 3937 }, { "epoch": 1.55, "learning_rate": 3.837837659727274e-07, "loss": 0.0303, "step": 3938 }, { "epoch": 1.55, "learning_rate": 3.837736930284122e-07, "loss": 0.0336, "step": 3939 }, { "epoch": 1.55, "learning_rate": 3.837636170888651e-07, "loss": 0.0294, "step": 3940 }, { "epoch": 1.55, "learning_rate": 3.837535381542503e-07, "loss": 0.0309, "step": 3941 }, { "epoch": 1.55, "learning_rate": 3.8374345622473214e-07, "loss": 0.0322, "step": 3942 }, { "epoch": 1.55, "learning_rate": 3.8373337130047494e-07, "loss": 0.0353, "step": 3943 }, { "epoch": 1.55, "learning_rate": 3.837232833816429e-07, "loss": 0.0333, "step": 3944 }, { "epoch": 1.55, "learning_rate": 3.8371319246840067e-07, "loss": 0.0333, "step": 3945 }, { "epoch": 1.56, "learning_rate": 3.837030985609126e-07, "loss": 0.03, "step": 3946 }, { "epoch": 1.56, "learning_rate": 3.8369300165934325e-07, "loss": 0.0356, "step": 3947 }, { "epoch": 1.56, "learning_rate": 3.836829017638571e-07, "loss": 0.0314, "step": 3948 }, { "epoch": 1.56, "learning_rate": 3.8367279887461883e-07, "loss": 0.0286, "step": 3949 }, { "epoch": 1.56, "learning_rate": 3.8366269299179306e-07, "loss": 0.0313, "step": 3950 }, { "epoch": 1.56, "learning_rate": 3.836525841155446e-07, "loss": 0.0357, "step": 3951 }, { "epoch": 1.56, "learning_rate": 3.836424722460381e-07, "loss": 0.0303, "step": 3952 }, { "epoch": 1.56, "learning_rate": 3.8363235738343845e-07, "loss": 0.0317, "step": 3953 }, { "epoch": 1.56, "learning_rate": 3.836222395279104e-07, "loss": 0.0291, "step": 3954 }, { "epoch": 1.56, "learning_rate": 3.8361211867961894e-07, "loss": 0.0313, "step": 3955 }, { "epoch": 1.56, "learning_rate": 3.8360199483872895e-07, "loss": 0.0371, "step": 3956 }, { "epoch": 1.56, "learning_rate": 3.835918680054056e-07, "loss": 0.032, "step": 3957 }, { "epoch": 1.56, "learning_rate": 3.835817381798137e-07, "loss": 0.031, "step": 3958 }, { "epoch": 1.56, "learning_rate": 3.835716053621186e-07, "loss": 0.0325, "step": 3959 }, { "epoch": 1.56, "learning_rate": 3.835614695524853e-07, "loss": 0.0357, "step": 3960 }, { "epoch": 1.56, "learning_rate": 3.8355133075107896e-07, "loss": 0.0282, "step": 3961 }, { "epoch": 1.56, "learning_rate": 3.8354118895806496e-07, "loss": 0.0336, "step": 3962 }, { "epoch": 1.56, "learning_rate": 3.8353104417360853e-07, "loss": 0.0321, "step": 3963 }, { "epoch": 1.56, "learning_rate": 3.83520896397875e-07, "loss": 0.0297, "step": 3964 }, { "epoch": 1.56, "learning_rate": 3.835107456310298e-07, "loss": 0.0275, "step": 3965 }, { "epoch": 1.56, "learning_rate": 3.8350059187323833e-07, "loss": 0.0313, "step": 3966 }, { "epoch": 1.56, "learning_rate": 3.834904351246661e-07, "loss": 0.0318, "step": 3967 }, { "epoch": 1.56, "learning_rate": 3.8348027538547864e-07, "loss": 0.0338, "step": 3968 }, { "epoch": 1.56, "learning_rate": 3.8347011265584153e-07, "loss": 0.0357, "step": 3969 }, { "epoch": 1.56, "learning_rate": 3.8345994693592046e-07, "loss": 0.033, "step": 3970 }, { "epoch": 1.57, "learning_rate": 3.8344977822588113e-07, "loss": 0.0324, "step": 3971 }, { "epoch": 1.57, "learning_rate": 3.8343960652588914e-07, "loss": 0.0323, "step": 3972 }, { "epoch": 1.57, "learning_rate": 3.834294318361104e-07, "loss": 0.0284, "step": 3973 }, { "epoch": 1.57, "learning_rate": 3.834192541567107e-07, "loss": 0.0332, "step": 3974 }, { "epoch": 1.57, "learning_rate": 3.8340907348785587e-07, "loss": 0.034, "step": 3975 }, { "epoch": 1.57, "learning_rate": 3.8339888982971197e-07, "loss": 0.0299, "step": 3976 }, { "epoch": 1.57, "learning_rate": 3.8338870318244486e-07, "loss": 0.0325, "step": 3977 }, { "epoch": 1.57, "learning_rate": 3.833785135462206e-07, "loss": 0.0285, "step": 3978 }, { "epoch": 1.57, "learning_rate": 3.833683209212053e-07, "loss": 0.0303, "step": 3979 }, { "epoch": 1.57, "learning_rate": 3.8335812530756506e-07, "loss": 0.0311, "step": 3980 }, { "epoch": 1.57, "learning_rate": 3.8334792670546605e-07, "loss": 0.0341, "step": 3981 }, { "epoch": 1.57, "learning_rate": 3.8333772511507445e-07, "loss": 0.0312, "step": 3982 }, { "epoch": 1.57, "learning_rate": 3.8332752053655655e-07, "loss": 0.0367, "step": 3983 }, { "epoch": 1.57, "learning_rate": 3.833173129700788e-07, "loss": 0.0218, "step": 3984 }, { "epoch": 1.57, "learning_rate": 3.833071024158074e-07, "loss": 0.0295, "step": 3985 }, { "epoch": 1.57, "learning_rate": 3.832968888739088e-07, "loss": 0.0277, "step": 3986 }, { "epoch": 1.57, "learning_rate": 3.8328667234454954e-07, "loss": 0.0258, "step": 3987 }, { "epoch": 1.57, "learning_rate": 3.83276452827896e-07, "loss": 0.0294, "step": 3988 }, { "epoch": 1.57, "learning_rate": 3.8326623032411493e-07, "loss": 0.0335, "step": 3989 }, { "epoch": 1.57, "learning_rate": 3.832560048333728e-07, "loss": 0.034, "step": 3990 }, { "epoch": 1.57, "learning_rate": 3.832457763558363e-07, "loss": 0.0295, "step": 3991 }, { "epoch": 1.57, "learning_rate": 3.8323554489167213e-07, "loss": 0.0338, "step": 3992 }, { "epoch": 1.57, "learning_rate": 3.832253104410471e-07, "loss": 0.0248, "step": 3993 }, { "epoch": 1.57, "learning_rate": 3.8321507300412794e-07, "loss": 0.0261, "step": 3994 }, { "epoch": 1.57, "learning_rate": 3.832048325810816e-07, "loss": 0.0345, "step": 3995 }, { "epoch": 1.57, "learning_rate": 3.8319458917207485e-07, "loss": 0.0335, "step": 3996 }, { "epoch": 1.58, "learning_rate": 3.831843427772748e-07, "loss": 0.0311, "step": 3997 }, { "epoch": 1.58, "learning_rate": 3.831740933968483e-07, "loss": 0.0332, "step": 3998 }, { "epoch": 1.58, "learning_rate": 3.831638410309625e-07, "loss": 0.0328, "step": 3999 }, { "epoch": 1.58, "learning_rate": 3.8315358567978446e-07, "loss": 0.0301, "step": 4000 }, { "epoch": 1.58, "learning_rate": 3.8314332734348136e-07, "loss": 0.0324, "step": 4001 }, { "epoch": 1.58, "learning_rate": 3.831330660222204e-07, "loss": 0.0292, "step": 4002 }, { "epoch": 1.58, "learning_rate": 3.8312280171616874e-07, "loss": 0.0289, "step": 4003 }, { "epoch": 1.58, "learning_rate": 3.8311253442549374e-07, "loss": 0.0298, "step": 4004 }, { "epoch": 1.58, "learning_rate": 3.8310226415036274e-07, "loss": 0.0271, "step": 4005 }, { "epoch": 1.58, "learning_rate": 3.8309199089094307e-07, "loss": 0.0283, "step": 4006 }, { "epoch": 1.58, "learning_rate": 3.8308171464740225e-07, "loss": 0.0328, "step": 4007 }, { "epoch": 1.58, "learning_rate": 3.8307143541990776e-07, "loss": 0.0292, "step": 4008 }, { "epoch": 1.58, "learning_rate": 3.830611532086271e-07, "loss": 0.0279, "step": 4009 }, { "epoch": 1.58, "learning_rate": 3.830508680137278e-07, "loss": 0.0298, "step": 4010 }, { "epoch": 1.58, "learning_rate": 3.8304057983537767e-07, "loss": 0.0284, "step": 4011 }, { "epoch": 1.58, "learning_rate": 3.830302886737442e-07, "loss": 0.0269, "step": 4012 }, { "epoch": 1.58, "learning_rate": 3.830199945289952e-07, "loss": 0.0253, "step": 4013 }, { "epoch": 1.58, "learning_rate": 3.8300969740129843e-07, "loss": 0.0293, "step": 4014 }, { "epoch": 1.58, "learning_rate": 3.829993972908218e-07, "loss": 0.0322, "step": 4015 }, { "epoch": 1.58, "learning_rate": 3.82989094197733e-07, "loss": 0.0283, "step": 4016 }, { "epoch": 1.58, "learning_rate": 3.8297878812220023e-07, "loss": 0.0259, "step": 4017 }, { "epoch": 1.58, "learning_rate": 3.829684790643912e-07, "loss": 0.0303, "step": 4018 }, { "epoch": 1.58, "learning_rate": 3.829581670244741e-07, "loss": 0.0321, "step": 4019 }, { "epoch": 1.58, "learning_rate": 3.829478520026169e-07, "loss": 0.0298, "step": 4020 }, { "epoch": 1.58, "learning_rate": 3.8293753399898786e-07, "loss": 0.0345, "step": 4021 }, { "epoch": 1.59, "learning_rate": 3.8292721301375497e-07, "loss": 0.0276, "step": 4022 }, { "epoch": 1.59, "learning_rate": 3.829168890470865e-07, "loss": 0.0329, "step": 4023 }, { "epoch": 1.59, "learning_rate": 3.8290656209915077e-07, "loss": 0.0305, "step": 4024 }, { "epoch": 1.59, "learning_rate": 3.8289623217011604e-07, "loss": 0.0331, "step": 4025 }, { "epoch": 1.59, "learning_rate": 3.8288589926015077e-07, "loss": 0.0325, "step": 4026 }, { "epoch": 1.59, "learning_rate": 3.828755633694232e-07, "loss": 0.0282, "step": 4027 }, { "epoch": 1.59, "learning_rate": 3.8286522449810193e-07, "loss": 0.0289, "step": 4028 }, { "epoch": 1.59, "learning_rate": 3.8285488264635543e-07, "loss": 0.0294, "step": 4029 }, { "epoch": 1.59, "learning_rate": 3.828445378143523e-07, "loss": 0.0305, "step": 4030 }, { "epoch": 1.59, "learning_rate": 3.82834190002261e-07, "loss": 0.0288, "step": 4031 }, { "epoch": 1.59, "learning_rate": 3.828238392102503e-07, "loss": 0.0292, "step": 4032 }, { "epoch": 1.59, "learning_rate": 3.828134854384889e-07, "loss": 0.0341, "step": 4033 }, { "epoch": 1.59, "learning_rate": 3.8280312868714556e-07, "loss": 0.0328, "step": 4034 }, { "epoch": 1.59, "learning_rate": 3.82792768956389e-07, "loss": 0.0296, "step": 4035 }, { "epoch": 1.59, "learning_rate": 3.8278240624638816e-07, "loss": 0.0294, "step": 4036 }, { "epoch": 1.59, "learning_rate": 3.8277204055731186e-07, "loss": 0.0315, "step": 4037 }, { "epoch": 1.59, "learning_rate": 3.827616718893291e-07, "loss": 0.0322, "step": 4038 }, { "epoch": 1.59, "learning_rate": 3.827513002426089e-07, "loss": 0.0285, "step": 4039 }, { "epoch": 1.59, "learning_rate": 3.8274092561732017e-07, "loss": 0.036, "step": 4040 }, { "epoch": 1.59, "learning_rate": 3.8273054801363216e-07, "loss": 0.0372, "step": 4041 }, { "epoch": 1.59, "learning_rate": 3.8272016743171385e-07, "loss": 0.0347, "step": 4042 }, { "epoch": 1.59, "learning_rate": 3.8270978387173457e-07, "loss": 0.0343, "step": 4043 }, { "epoch": 1.59, "learning_rate": 3.8269939733386353e-07, "loss": 0.0303, "step": 4044 }, { "epoch": 1.59, "learning_rate": 3.826890078182699e-07, "loss": 0.0322, "step": 4045 }, { "epoch": 1.59, "learning_rate": 3.8267861532512316e-07, "loss": 0.031, "step": 4046 }, { "epoch": 1.59, "learning_rate": 3.8266821985459263e-07, "loss": 0.0356, "step": 4047 }, { "epoch": 1.6, "learning_rate": 3.826578214068477e-07, "loss": 0.0283, "step": 4048 }, { "epoch": 1.6, "learning_rate": 3.8264741998205795e-07, "loss": 0.0372, "step": 4049 }, { "epoch": 1.6, "learning_rate": 3.826370155803928e-07, "loss": 0.0303, "step": 4050 }, { "epoch": 1.6, "learning_rate": 3.8262660820202187e-07, "loss": 0.0294, "step": 4051 }, { "epoch": 1.6, "learning_rate": 3.8261619784711483e-07, "loss": 0.0274, "step": 4052 }, { "epoch": 1.6, "learning_rate": 3.826057845158413e-07, "loss": 0.0299, "step": 4053 }, { "epoch": 1.6, "learning_rate": 3.8259536820837093e-07, "loss": 0.029, "step": 4054 }, { "epoch": 1.6, "learning_rate": 3.8258494892487366e-07, "loss": 0.0323, "step": 4055 }, { "epoch": 1.6, "learning_rate": 3.825745266655192e-07, "loss": 0.0303, "step": 4056 }, { "epoch": 1.6, "learning_rate": 3.825641014304774e-07, "loss": 0.0254, "step": 4057 }, { "epoch": 1.6, "learning_rate": 3.8255367321991826e-07, "loss": 0.0331, "step": 4058 }, { "epoch": 1.6, "learning_rate": 3.8254324203401165e-07, "loss": 0.0275, "step": 4059 }, { "epoch": 1.6, "learning_rate": 3.825328078729277e-07, "loss": 0.0298, "step": 4060 }, { "epoch": 1.6, "learning_rate": 3.825223707368363e-07, "loss": 0.0303, "step": 4061 }, { "epoch": 1.6, "learning_rate": 3.8251193062590776e-07, "loss": 0.028, "step": 4062 }, { "epoch": 1.6, "learning_rate": 3.825014875403121e-07, "loss": 0.0323, "step": 4063 }, { "epoch": 1.6, "learning_rate": 3.8249104148021956e-07, "loss": 0.024, "step": 4064 }, { "epoch": 1.6, "learning_rate": 3.8248059244580037e-07, "loss": 0.0312, "step": 4065 }, { "epoch": 1.6, "learning_rate": 3.824701404372249e-07, "loss": 0.0286, "step": 4066 }, { "epoch": 1.6, "learning_rate": 3.824596854546635e-07, "loss": 0.0335, "step": 4067 }, { "epoch": 1.6, "learning_rate": 3.824492274982864e-07, "loss": 0.0333, "step": 4068 }, { "epoch": 1.6, "learning_rate": 3.824387665682643e-07, "loss": 0.0322, "step": 4069 }, { "epoch": 1.6, "learning_rate": 3.8242830266476755e-07, "loss": 0.027, "step": 4070 }, { "epoch": 1.6, "learning_rate": 3.824178357879667e-07, "loss": 0.0289, "step": 4071 }, { "epoch": 1.6, "learning_rate": 3.8240736593803236e-07, "loss": 0.0294, "step": 4072 }, { "epoch": 1.61, "learning_rate": 3.8239689311513527e-07, "loss": 0.0295, "step": 4073 }, { "epoch": 1.61, "learning_rate": 3.8238641731944593e-07, "loss": 0.0285, "step": 4074 }, { "epoch": 1.61, "learning_rate": 3.823759385511352e-07, "loss": 0.0315, "step": 4075 }, { "epoch": 1.61, "learning_rate": 3.8236545681037393e-07, "loss": 0.0325, "step": 4076 }, { "epoch": 1.61, "learning_rate": 3.8235497209733284e-07, "loss": 0.0345, "step": 4077 }, { "epoch": 1.61, "learning_rate": 3.8234448441218284e-07, "loss": 0.0312, "step": 4078 }, { "epoch": 1.61, "learning_rate": 3.8233399375509487e-07, "loss": 0.0276, "step": 4079 }, { "epoch": 1.61, "learning_rate": 3.8232350012623993e-07, "loss": 0.0252, "step": 4080 }, { "epoch": 1.61, "learning_rate": 3.82313003525789e-07, "loss": 0.027, "step": 4081 }, { "epoch": 1.61, "learning_rate": 3.823025039539132e-07, "loss": 0.0337, "step": 4082 }, { "epoch": 1.61, "learning_rate": 3.8229200141078374e-07, "loss": 0.0295, "step": 4083 }, { "epoch": 1.61, "learning_rate": 3.8228149589657165e-07, "loss": 0.0266, "step": 4084 }, { "epoch": 1.61, "learning_rate": 3.822709874114482e-07, "loss": 0.0266, "step": 4085 }, { "epoch": 1.61, "learning_rate": 3.822604759555847e-07, "loss": 0.0309, "step": 4086 }, { "epoch": 1.61, "learning_rate": 3.8224996152915243e-07, "loss": 0.0289, "step": 4087 }, { "epoch": 1.61, "learning_rate": 3.8223944413232275e-07, "loss": 0.0321, "step": 4088 }, { "epoch": 1.61, "learning_rate": 3.822289237652672e-07, "loss": 0.0313, "step": 4089 }, { "epoch": 1.61, "learning_rate": 3.8221840042815706e-07, "loss": 0.0336, "step": 4090 }, { "epoch": 1.61, "learning_rate": 3.8220787412116393e-07, "loss": 0.0336, "step": 4091 }, { "epoch": 1.61, "learning_rate": 3.821973448444594e-07, "loss": 0.03, "step": 4092 }, { "epoch": 1.61, "learning_rate": 3.8218681259821505e-07, "loss": 0.0274, "step": 4093 }, { "epoch": 1.61, "learning_rate": 3.821762773826026e-07, "loss": 0.0278, "step": 4094 }, { "epoch": 1.61, "learning_rate": 3.821657391977936e-07, "loss": 0.0303, "step": 4095 }, { "epoch": 1.61, "learning_rate": 3.8215519804396e-07, "loss": 0.0339, "step": 4096 }, { "epoch": 1.61, "learning_rate": 3.821446539212735e-07, "loss": 0.0316, "step": 4097 }, { "epoch": 1.62, "learning_rate": 3.821341068299059e-07, "loss": 0.0299, "step": 4098 }, { "epoch": 1.62, "learning_rate": 3.8212355677002925e-07, "loss": 0.0258, "step": 4099 }, { "epoch": 1.62, "learning_rate": 3.821130037418154e-07, "loss": 0.0313, "step": 4100 }, { "epoch": 1.62, "learning_rate": 3.821024477454363e-07, "loss": 0.0285, "step": 4101 }, { "epoch": 1.62, "learning_rate": 3.820918887810642e-07, "loss": 0.0293, "step": 4102 }, { "epoch": 1.62, "learning_rate": 3.8208132684887094e-07, "loss": 0.03, "step": 4103 }, { "epoch": 1.62, "learning_rate": 3.820707619490289e-07, "loss": 0.0283, "step": 4104 }, { "epoch": 1.62, "learning_rate": 3.8206019408171e-07, "loss": 0.0295, "step": 4105 }, { "epoch": 1.62, "learning_rate": 3.820496232470868e-07, "loss": 0.0339, "step": 4106 }, { "epoch": 1.62, "learning_rate": 3.820390494453313e-07, "loss": 0.0325, "step": 4107 }, { "epoch": 1.62, "learning_rate": 3.8202847267661606e-07, "loss": 0.0319, "step": 4108 }, { "epoch": 1.62, "learning_rate": 3.820178929411133e-07, "loss": 0.0341, "step": 4109 }, { "epoch": 1.62, "learning_rate": 3.820073102389956e-07, "loss": 0.03, "step": 4110 }, { "epoch": 1.62, "learning_rate": 3.8199672457043527e-07, "loss": 0.0276, "step": 4111 }, { "epoch": 1.62, "learning_rate": 3.8198613593560496e-07, "loss": 0.0279, "step": 4112 }, { "epoch": 1.62, "learning_rate": 3.8197554433467724e-07, "loss": 0.0285, "step": 4113 }, { "epoch": 1.62, "learning_rate": 3.819649497678247e-07, "loss": 0.0299, "step": 4114 }, { "epoch": 1.62, "learning_rate": 3.8195435223522e-07, "loss": 0.029, "step": 4115 }, { "epoch": 1.62, "learning_rate": 3.8194375173703597e-07, "loss": 0.0279, "step": 4116 }, { "epoch": 1.62, "learning_rate": 3.8193314827344523e-07, "loss": 0.0333, "step": 4117 }, { "epoch": 1.62, "learning_rate": 3.819225418446208e-07, "loss": 0.0274, "step": 4118 }, { "epoch": 1.62, "learning_rate": 3.8191193245073527e-07, "loss": 0.0338, "step": 4119 }, { "epoch": 1.62, "learning_rate": 3.8190132009196186e-07, "loss": 0.0326, "step": 4120 }, { "epoch": 1.62, "learning_rate": 3.818907047684733e-07, "loss": 0.0276, "step": 4121 }, { "epoch": 1.62, "learning_rate": 3.818800864804427e-07, "loss": 0.0284, "step": 4122 }, { "epoch": 1.62, "learning_rate": 3.818694652280432e-07, "loss": 0.0284, "step": 4123 }, { "epoch": 1.63, "learning_rate": 3.8185884101144774e-07, "loss": 0.0321, "step": 4124 }, { "epoch": 1.63, "learning_rate": 3.8184821383082955e-07, "loss": 0.0318, "step": 4125 }, { "epoch": 1.63, "learning_rate": 3.818375836863619e-07, "loss": 0.0265, "step": 4126 }, { "epoch": 1.63, "learning_rate": 3.81826950578218e-07, "loss": 0.0281, "step": 4127 }, { "epoch": 1.63, "learning_rate": 3.8181631450657116e-07, "loss": 0.0308, "step": 4128 }, { "epoch": 1.63, "learning_rate": 3.818056754715947e-07, "loss": 0.0306, "step": 4129 }, { "epoch": 1.63, "learning_rate": 3.8179503347346206e-07, "loss": 0.0296, "step": 4130 }, { "epoch": 1.63, "learning_rate": 3.817843885123467e-07, "loss": 0.0314, "step": 4131 }, { "epoch": 1.63, "learning_rate": 3.8177374058842195e-07, "loss": 0.0304, "step": 4132 }, { "epoch": 1.63, "learning_rate": 3.817630897018616e-07, "loss": 0.0303, "step": 4133 }, { "epoch": 1.63, "learning_rate": 3.817524358528391e-07, "loss": 0.0303, "step": 4134 }, { "epoch": 1.63, "learning_rate": 3.817417790415282e-07, "loss": 0.0336, "step": 4135 }, { "epoch": 1.63, "learning_rate": 3.817311192681025e-07, "loss": 0.0297, "step": 4136 }, { "epoch": 1.63, "learning_rate": 3.817204565327357e-07, "loss": 0.0268, "step": 4137 }, { "epoch": 1.63, "learning_rate": 3.8170979083560163e-07, "loss": 0.0339, "step": 4138 }, { "epoch": 1.63, "learning_rate": 3.8169912217687417e-07, "loss": 0.0292, "step": 4139 }, { "epoch": 1.63, "learning_rate": 3.816884505567272e-07, "loss": 0.0293, "step": 4140 }, { "epoch": 1.63, "learning_rate": 3.8167777597533455e-07, "loss": 0.0324, "step": 4141 }, { "epoch": 1.63, "learning_rate": 3.816670984328704e-07, "loss": 0.0329, "step": 4142 }, { "epoch": 1.63, "learning_rate": 3.816564179295085e-07, "loss": 0.0297, "step": 4143 }, { "epoch": 1.63, "learning_rate": 3.8164573446542313e-07, "loss": 0.0334, "step": 4144 }, { "epoch": 1.63, "learning_rate": 3.816350480407884e-07, "loss": 0.0302, "step": 4145 }, { "epoch": 1.63, "learning_rate": 3.8162435865577834e-07, "loss": 0.0304, "step": 4146 }, { "epoch": 1.63, "learning_rate": 3.8161366631056733e-07, "loss": 0.0288, "step": 4147 }, { "epoch": 1.63, "learning_rate": 3.8160297100532964e-07, "loss": 0.0341, "step": 4148 }, { "epoch": 1.64, "learning_rate": 3.8159227274023946e-07, "loss": 0.0279, "step": 4149 }, { "epoch": 1.64, "learning_rate": 3.815815715154712e-07, "loss": 0.0359, "step": 4150 }, { "epoch": 1.64, "learning_rate": 3.815708673311993e-07, "loss": 0.0283, "step": 4151 }, { "epoch": 1.64, "learning_rate": 3.8156016018759824e-07, "loss": 0.0299, "step": 4152 }, { "epoch": 1.64, "learning_rate": 3.8154945008484257e-07, "loss": 0.0351, "step": 4153 }, { "epoch": 1.64, "learning_rate": 3.815387370231067e-07, "loss": 0.031, "step": 4154 }, { "epoch": 1.64, "learning_rate": 3.815280210025654e-07, "loss": 0.0327, "step": 4155 }, { "epoch": 1.64, "learning_rate": 3.815173020233932e-07, "loss": 0.0322, "step": 4156 }, { "epoch": 1.64, "learning_rate": 3.815065800857648e-07, "loss": 0.0311, "step": 4157 }, { "epoch": 1.64, "learning_rate": 3.814958551898551e-07, "loss": 0.0327, "step": 4158 }, { "epoch": 1.64, "learning_rate": 3.8148512733583873e-07, "loss": 0.0314, "step": 4159 }, { "epoch": 1.64, "learning_rate": 3.8147439652389067e-07, "loss": 0.0258, "step": 4160 }, { "epoch": 1.64, "learning_rate": 3.8146366275418577e-07, "loss": 0.033, "step": 4161 }, { "epoch": 1.64, "learning_rate": 3.8145292602689885e-07, "loss": 0.0296, "step": 4162 }, { "epoch": 1.64, "learning_rate": 3.814421863422051e-07, "loss": 0.0258, "step": 4163 }, { "epoch": 1.64, "learning_rate": 3.8143144370027945e-07, "loss": 0.0274, "step": 4164 }, { "epoch": 1.64, "learning_rate": 3.81420698101297e-07, "loss": 0.0331, "step": 4165 }, { "epoch": 1.64, "learning_rate": 3.8140994954543294e-07, "loss": 0.0258, "step": 4166 }, { "epoch": 1.64, "learning_rate": 3.813991980328624e-07, "loss": 0.0303, "step": 4167 }, { "epoch": 1.64, "learning_rate": 3.813884435637606e-07, "loss": 0.0351, "step": 4168 }, { "epoch": 1.64, "learning_rate": 3.8137768613830287e-07, "loss": 0.0312, "step": 4169 }, { "epoch": 1.64, "learning_rate": 3.8136692575666453e-07, "loss": 0.0248, "step": 4170 }, { "epoch": 1.64, "learning_rate": 3.813561624190209e-07, "loss": 0.0334, "step": 4171 }, { "epoch": 1.64, "learning_rate": 3.813453961255475e-07, "loss": 0.0263, "step": 4172 }, { "epoch": 1.64, "learning_rate": 3.8133462687641974e-07, "loss": 0.0295, "step": 4173 }, { "epoch": 1.65, "learning_rate": 3.813238546718131e-07, "loss": 0.0279, "step": 4174 }, { "epoch": 1.65, "learning_rate": 3.8131307951190327e-07, "loss": 0.0293, "step": 4175 }, { "epoch": 1.65, "learning_rate": 3.8130230139686584e-07, "loss": 0.0282, "step": 4176 }, { "epoch": 1.65, "learning_rate": 3.8129152032687637e-07, "loss": 0.029, "step": 4177 }, { "epoch": 1.65, "learning_rate": 3.812807363021107e-07, "loss": 0.0335, "step": 4178 }, { "epoch": 1.65, "learning_rate": 3.812699493227446e-07, "loss": 0.0275, "step": 4179 }, { "epoch": 1.65, "learning_rate": 3.812591593889537e-07, "loss": 0.0303, "step": 4180 }, { "epoch": 1.65, "learning_rate": 3.81248366500914e-07, "loss": 0.0341, "step": 4181 }, { "epoch": 1.65, "learning_rate": 3.812375706588015e-07, "loss": 0.0309, "step": 4182 }, { "epoch": 1.65, "learning_rate": 3.8122677186279196e-07, "loss": 0.0277, "step": 4183 }, { "epoch": 1.65, "learning_rate": 3.812159701130615e-07, "loss": 0.0275, "step": 4184 }, { "epoch": 1.65, "learning_rate": 3.812051654097861e-07, "loss": 0.0294, "step": 4185 }, { "epoch": 1.65, "learning_rate": 3.811943577531419e-07, "loss": 0.0275, "step": 4186 }, { "epoch": 1.65, "learning_rate": 3.8118354714330514e-07, "loss": 0.0279, "step": 4187 }, { "epoch": 1.65, "learning_rate": 3.8117273358045186e-07, "loss": 0.0303, "step": 4188 }, { "epoch": 1.65, "learning_rate": 3.811619170647584e-07, "loss": 0.0356, "step": 4189 }, { "epoch": 1.65, "learning_rate": 3.81151097596401e-07, "loss": 0.0288, "step": 4190 }, { "epoch": 1.65, "learning_rate": 3.811402751755561e-07, "loss": 0.0372, "step": 4191 }, { "epoch": 1.65, "learning_rate": 3.8112944980239996e-07, "loss": 0.0354, "step": 4192 }, { "epoch": 1.65, "learning_rate": 3.811186214771091e-07, "loss": 0.0303, "step": 4193 }, { "epoch": 1.65, "learning_rate": 3.8110779019985994e-07, "loss": 0.0303, "step": 4194 }, { "epoch": 1.65, "learning_rate": 3.8109695597082905e-07, "loss": 0.0348, "step": 4195 }, { "epoch": 1.65, "learning_rate": 3.8108611879019305e-07, "loss": 0.034, "step": 4196 }, { "epoch": 1.65, "learning_rate": 3.810752786581285e-07, "loss": 0.0307, "step": 4197 }, { "epoch": 1.65, "learning_rate": 3.810644355748121e-07, "loss": 0.033, "step": 4198 }, { "epoch": 1.65, "learning_rate": 3.810535895404206e-07, "loss": 0.0287, "step": 4199 }, { "epoch": 1.66, "learning_rate": 3.8104274055513075e-07, "loss": 0.029, "step": 4200 }, { "epoch": 1.66, "learning_rate": 3.810318886191194e-07, "loss": 0.0308, "step": 4201 }, { "epoch": 1.66, "learning_rate": 3.8102103373256343e-07, "loss": 0.0309, "step": 4202 }, { "epoch": 1.66, "learning_rate": 3.810101758956397e-07, "loss": 0.0299, "step": 4203 }, { "epoch": 1.66, "learning_rate": 3.809993151085252e-07, "loss": 0.0281, "step": 4204 }, { "epoch": 1.66, "learning_rate": 3.8098845137139697e-07, "loss": 0.0305, "step": 4205 }, { "epoch": 1.66, "learning_rate": 3.8097758468443204e-07, "loss": 0.027, "step": 4206 }, { "epoch": 1.66, "learning_rate": 3.8096671504780754e-07, "loss": 0.0253, "step": 4207 }, { "epoch": 1.66, "learning_rate": 3.809558424617006e-07, "loss": 0.0304, "step": 4208 }, { "epoch": 1.66, "learning_rate": 3.809449669262885e-07, "loss": 0.033, "step": 4209 }, { "epoch": 1.66, "learning_rate": 3.809340884417484e-07, "loss": 0.0304, "step": 4210 }, { "epoch": 1.66, "learning_rate": 3.8092320700825766e-07, "loss": 0.0292, "step": 4211 }, { "epoch": 1.66, "learning_rate": 3.8091232262599366e-07, "loss": 0.0303, "step": 4212 }, { "epoch": 1.66, "learning_rate": 3.809014352951337e-07, "loss": 0.028, "step": 4213 }, { "epoch": 1.66, "learning_rate": 3.808905450158553e-07, "loss": 0.0333, "step": 4214 }, { "epoch": 1.66, "learning_rate": 3.808796517883359e-07, "loss": 0.0322, "step": 4215 }, { "epoch": 1.66, "learning_rate": 3.8086875561275314e-07, "loss": 0.0306, "step": 4216 }, { "epoch": 1.66, "learning_rate": 3.808578564892845e-07, "loss": 0.0372, "step": 4217 }, { "epoch": 1.66, "learning_rate": 3.8084695441810773e-07, "loss": 0.0304, "step": 4218 }, { "epoch": 1.66, "learning_rate": 3.808360493994004e-07, "loss": 0.0285, "step": 4219 }, { "epoch": 1.66, "learning_rate": 3.808251414333403e-07, "loss": 0.0308, "step": 4220 }, { "epoch": 1.66, "learning_rate": 3.808142305201053e-07, "loss": 0.0324, "step": 4221 }, { "epoch": 1.66, "learning_rate": 3.808033166598731e-07, "loss": 0.0276, "step": 4222 }, { "epoch": 1.66, "learning_rate": 3.807923998528216e-07, "loss": 0.026, "step": 4223 }, { "epoch": 1.66, "learning_rate": 3.8078148009912876e-07, "loss": 0.0247, "step": 4224 }, { "epoch": 1.67, "learning_rate": 3.8077055739897263e-07, "loss": 0.0352, "step": 4225 }, { "epoch": 1.67, "learning_rate": 3.807596317525311e-07, "loss": 0.0371, "step": 4226 }, { "epoch": 1.67, "learning_rate": 3.807487031599823e-07, "loss": 0.0271, "step": 4227 }, { "epoch": 1.67, "learning_rate": 3.8073777162150436e-07, "loss": 0.0252, "step": 4228 }, { "epoch": 1.67, "learning_rate": 3.8072683713727536e-07, "loss": 0.0276, "step": 4229 }, { "epoch": 1.67, "learning_rate": 3.8071589970747366e-07, "loss": 0.029, "step": 4230 }, { "epoch": 1.67, "learning_rate": 3.807049593322774e-07, "loss": 0.0287, "step": 4231 }, { "epoch": 1.67, "learning_rate": 3.80694016011865e-07, "loss": 0.0309, "step": 4232 }, { "epoch": 1.67, "learning_rate": 3.8068306974641476e-07, "loss": 0.0276, "step": 4233 }, { "epoch": 1.67, "learning_rate": 3.8067212053610504e-07, "loss": 0.0278, "step": 4234 }, { "epoch": 1.67, "learning_rate": 3.806611683811144e-07, "loss": 0.0279, "step": 4235 }, { "epoch": 1.67, "learning_rate": 3.8065021328162124e-07, "loss": 0.0335, "step": 4236 }, { "epoch": 1.67, "learning_rate": 3.806392552378042e-07, "loss": 0.0338, "step": 4237 }, { "epoch": 1.67, "learning_rate": 3.8062829424984173e-07, "loss": 0.0314, "step": 4238 }, { "epoch": 1.67, "learning_rate": 3.806173303179127e-07, "loss": 0.0302, "step": 4239 }, { "epoch": 1.67, "learning_rate": 3.806063634421957e-07, "loss": 0.0322, "step": 4240 }, { "epoch": 1.67, "learning_rate": 3.805953936228694e-07, "loss": 0.0302, "step": 4241 }, { "epoch": 1.67, "learning_rate": 3.805844208601127e-07, "loss": 0.0275, "step": 4242 }, { "epoch": 1.67, "learning_rate": 3.805734451541044e-07, "loss": 0.0269, "step": 4243 }, { "epoch": 1.67, "learning_rate": 3.8056246650502334e-07, "loss": 0.0327, "step": 4244 }, { "epoch": 1.67, "learning_rate": 3.805514849130485e-07, "loss": 0.0306, "step": 4245 }, { "epoch": 1.67, "learning_rate": 3.8054050037835893e-07, "loss": 0.0258, "step": 4246 }, { "epoch": 1.67, "learning_rate": 3.805295129011336e-07, "loss": 0.0262, "step": 4247 }, { "epoch": 1.67, "learning_rate": 3.805185224815515e-07, "loss": 0.029, "step": 4248 }, { "epoch": 1.67, "learning_rate": 3.8050752911979184e-07, "loss": 0.0297, "step": 4249 }, { "epoch": 1.67, "learning_rate": 3.8049653281603384e-07, "loss": 0.0335, "step": 4250 }, { "epoch": 1.68, "learning_rate": 3.804855335704567e-07, "loss": 0.0309, "step": 4251 }, { "epoch": 1.68, "learning_rate": 3.804745313832396e-07, "loss": 0.0275, "step": 4252 }, { "epoch": 1.68, "learning_rate": 3.80463526254562e-07, "loss": 0.0272, "step": 4253 }, { "epoch": 1.68, "learning_rate": 3.804525181846032e-07, "loss": 0.0292, "step": 4254 }, { "epoch": 1.68, "learning_rate": 3.804415071735425e-07, "loss": 0.0277, "step": 4255 }, { "epoch": 1.68, "learning_rate": 3.804304932215596e-07, "loss": 0.0259, "step": 4256 }, { "epoch": 1.68, "learning_rate": 3.804194763288338e-07, "loss": 0.0299, "step": 4257 }, { "epoch": 1.68, "learning_rate": 3.804084564955448e-07, "loss": 0.0312, "step": 4258 }, { "epoch": 1.68, "learning_rate": 3.8039743372187213e-07, "loss": 0.03, "step": 4259 }, { "epoch": 1.68, "learning_rate": 3.803864080079954e-07, "loss": 0.0307, "step": 4260 }, { "epoch": 1.68, "learning_rate": 3.8037537935409444e-07, "loss": 0.0299, "step": 4261 }, { "epoch": 1.68, "learning_rate": 3.803643477603489e-07, "loss": 0.0311, "step": 4262 }, { "epoch": 1.68, "learning_rate": 3.803533132269387e-07, "loss": 0.0281, "step": 4263 }, { "epoch": 1.68, "learning_rate": 3.8034227575404347e-07, "loss": 0.0287, "step": 4264 }, { "epoch": 1.68, "learning_rate": 3.8033123534184333e-07, "loss": 0.0316, "step": 4265 }, { "epoch": 1.68, "learning_rate": 3.803201919905181e-07, "loss": 0.0257, "step": 4266 }, { "epoch": 1.68, "learning_rate": 3.803091457002478e-07, "loss": 0.0309, "step": 4267 }, { "epoch": 1.68, "learning_rate": 3.802980964712125e-07, "loss": 0.0366, "step": 4268 }, { "epoch": 1.68, "learning_rate": 3.802870443035922e-07, "loss": 0.0339, "step": 4269 }, { "epoch": 1.68, "learning_rate": 3.8027598919756707e-07, "loss": 0.0296, "step": 4270 }, { "epoch": 1.68, "learning_rate": 3.802649311533173e-07, "loss": 0.0319, "step": 4271 }, { "epoch": 1.68, "learning_rate": 3.8025387017102314e-07, "loss": 0.0302, "step": 4272 }, { "epoch": 1.68, "learning_rate": 3.802428062508649e-07, "loss": 0.0237, "step": 4273 }, { "epoch": 1.68, "learning_rate": 3.802317393930228e-07, "loss": 0.0307, "step": 4274 }, { "epoch": 1.68, "learning_rate": 3.802206695976773e-07, "loss": 0.0286, "step": 4275 }, { "epoch": 1.69, "learning_rate": 3.8020959686500874e-07, "loss": 0.0324, "step": 4276 }, { "epoch": 1.69, "learning_rate": 3.801985211951977e-07, "loss": 0.0308, "step": 4277 }, { "epoch": 1.69, "learning_rate": 3.801874425884246e-07, "loss": 0.0271, "step": 4278 }, { "epoch": 1.69, "learning_rate": 3.8017636104487e-07, "loss": 0.0282, "step": 4279 }, { "epoch": 1.69, "learning_rate": 3.801652765647146e-07, "loss": 0.03, "step": 4280 }, { "epoch": 1.69, "learning_rate": 3.80154189148139e-07, "loss": 0.0283, "step": 4281 }, { "epoch": 1.69, "learning_rate": 3.801430987953239e-07, "loss": 0.026, "step": 4282 }, { "epoch": 1.69, "learning_rate": 3.8013200550645013e-07, "loss": 0.028, "step": 4283 }, { "epoch": 1.69, "learning_rate": 3.8012090928169844e-07, "loss": 0.0267, "step": 4284 }, { "epoch": 1.69, "learning_rate": 3.801098101212496e-07, "loss": 0.0325, "step": 4285 }, { "epoch": 1.69, "learning_rate": 3.8009870802528467e-07, "loss": 0.0304, "step": 4286 }, { "epoch": 1.69, "learning_rate": 3.800876029939845e-07, "loss": 0.029, "step": 4287 }, { "epoch": 1.69, "learning_rate": 3.800764950275301e-07, "loss": 0.0327, "step": 4288 }, { "epoch": 1.69, "learning_rate": 3.800653841261025e-07, "loss": 0.034, "step": 4289 }, { "epoch": 1.69, "learning_rate": 3.800542702898828e-07, "loss": 0.0274, "step": 4290 }, { "epoch": 1.69, "learning_rate": 3.8004315351905216e-07, "loss": 0.0268, "step": 4291 }, { "epoch": 1.69, "learning_rate": 3.800320338137918e-07, "loss": 0.0258, "step": 4292 }, { "epoch": 1.69, "learning_rate": 3.800209111742828e-07, "loss": 0.0268, "step": 4293 }, { "epoch": 1.69, "learning_rate": 3.800097856007066e-07, "loss": 0.0271, "step": 4294 }, { "epoch": 1.69, "learning_rate": 3.7999865709324447e-07, "loss": 0.0274, "step": 4295 }, { "epoch": 1.69, "learning_rate": 3.799875256520778e-07, "loss": 0.0313, "step": 4296 }, { "epoch": 1.69, "learning_rate": 3.79976391277388e-07, "loss": 0.0279, "step": 4297 }, { "epoch": 1.69, "learning_rate": 3.799652539693566e-07, "loss": 0.031, "step": 4298 }, { "epoch": 1.69, "learning_rate": 3.7995411372816506e-07, "loss": 0.0291, "step": 4299 }, { "epoch": 1.69, "learning_rate": 3.79942970553995e-07, "loss": 0.0319, "step": 4300 }, { "epoch": 1.7, "learning_rate": 3.7993182444702787e-07, "loss": 0.0308, "step": 4301 }, { "epoch": 1.7, "learning_rate": 3.799206754074456e-07, "loss": 0.0256, "step": 4302 }, { "epoch": 1.7, "learning_rate": 3.799095234354297e-07, "loss": 0.0308, "step": 4303 }, { "epoch": 1.7, "learning_rate": 3.7989836853116204e-07, "loss": 0.0277, "step": 4304 }, { "epoch": 1.7, "learning_rate": 3.7988721069482434e-07, "loss": 0.0277, "step": 4305 }, { "epoch": 1.7, "learning_rate": 3.7987604992659856e-07, "loss": 0.0318, "step": 4306 }, { "epoch": 1.7, "learning_rate": 3.798648862266665e-07, "loss": 0.03, "step": 4307 }, { "epoch": 1.7, "learning_rate": 3.798537195952102e-07, "loss": 0.0296, "step": 4308 }, { "epoch": 1.7, "learning_rate": 3.798425500324116e-07, "loss": 0.0311, "step": 4309 }, { "epoch": 1.7, "learning_rate": 3.7983137753845276e-07, "loss": 0.0306, "step": 4310 }, { "epoch": 1.7, "learning_rate": 3.798202021135157e-07, "loss": 0.0232, "step": 4311 }, { "epoch": 1.7, "learning_rate": 3.7980902375778274e-07, "loss": 0.0271, "step": 4312 }, { "epoch": 1.7, "learning_rate": 3.79797842471436e-07, "loss": 0.0278, "step": 4313 }, { "epoch": 1.7, "learning_rate": 3.797866582546576e-07, "loss": 0.0254, "step": 4314 }, { "epoch": 1.7, "learning_rate": 3.7977547110762995e-07, "loss": 0.0341, "step": 4315 }, { "epoch": 1.7, "learning_rate": 3.7976428103053535e-07, "loss": 0.0309, "step": 4316 }, { "epoch": 1.7, "learning_rate": 3.7975308802355617e-07, "loss": 0.0279, "step": 4317 }, { "epoch": 1.7, "learning_rate": 3.7974189208687484e-07, "loss": 0.0318, "step": 4318 }, { "epoch": 1.7, "learning_rate": 3.797306932206739e-07, "loss": 0.0314, "step": 4319 }, { "epoch": 1.7, "learning_rate": 3.797194914251357e-07, "loss": 0.0249, "step": 4320 }, { "epoch": 1.7, "learning_rate": 3.797082867004431e-07, "loss": 0.0339, "step": 4321 }, { "epoch": 1.7, "learning_rate": 3.7969707904677844e-07, "loss": 0.0297, "step": 4322 }, { "epoch": 1.7, "learning_rate": 3.7968586846432454e-07, "loss": 0.0283, "step": 4323 }, { "epoch": 1.7, "learning_rate": 3.79674654953264e-07, "loss": 0.0256, "step": 4324 }, { "epoch": 1.7, "learning_rate": 3.796634385137798e-07, "loss": 0.0298, "step": 4325 }, { "epoch": 1.7, "learning_rate": 3.7965221914605446e-07, "loss": 0.0322, "step": 4326 }, { "epoch": 1.71, "learning_rate": 3.7964099685027104e-07, "loss": 0.0286, "step": 4327 }, { "epoch": 1.71, "learning_rate": 3.7962977162661246e-07, "loss": 0.0243, "step": 4328 }, { "epoch": 1.71, "learning_rate": 3.796185434752615e-07, "loss": 0.0293, "step": 4329 }, { "epoch": 1.71, "learning_rate": 3.796073123964014e-07, "loss": 0.0322, "step": 4330 }, { "epoch": 1.71, "learning_rate": 3.7959607839021495e-07, "loss": 0.0279, "step": 4331 }, { "epoch": 1.71, "learning_rate": 3.7958484145688544e-07, "loss": 0.0306, "step": 4332 }, { "epoch": 1.71, "learning_rate": 3.7957360159659595e-07, "loss": 0.0321, "step": 4333 }, { "epoch": 1.71, "learning_rate": 3.795623588095297e-07, "loss": 0.0285, "step": 4334 }, { "epoch": 1.71, "learning_rate": 3.795511130958699e-07, "loss": 0.0315, "step": 4335 }, { "epoch": 1.71, "learning_rate": 3.7953986445579976e-07, "loss": 0.0312, "step": 4336 }, { "epoch": 1.71, "learning_rate": 3.7952861288950276e-07, "loss": 0.0293, "step": 4337 }, { "epoch": 1.71, "learning_rate": 3.795173583971622e-07, "loss": 0.0291, "step": 4338 }, { "epoch": 1.71, "learning_rate": 3.795061009789616e-07, "loss": 0.0295, "step": 4339 }, { "epoch": 1.71, "learning_rate": 3.7949484063508434e-07, "loss": 0.026, "step": 4340 }, { "epoch": 1.71, "learning_rate": 3.794835773657139e-07, "loss": 0.0311, "step": 4341 }, { "epoch": 1.71, "learning_rate": 3.7947231117103397e-07, "loss": 0.0308, "step": 4342 }, { "epoch": 1.71, "learning_rate": 3.7946104205122824e-07, "loss": 0.0302, "step": 4343 }, { "epoch": 1.71, "learning_rate": 3.794497700064801e-07, "loss": 0.0291, "step": 4344 }, { "epoch": 1.71, "learning_rate": 3.794384950369736e-07, "loss": 0.0298, "step": 4345 }, { "epoch": 1.71, "learning_rate": 3.7942721714289225e-07, "loss": 0.0299, "step": 4346 }, { "epoch": 1.71, "learning_rate": 3.7941593632442e-07, "loss": 0.0272, "step": 4347 }, { "epoch": 1.71, "learning_rate": 3.794046525817406e-07, "loss": 0.0345, "step": 4348 }, { "epoch": 1.71, "learning_rate": 3.7939336591503814e-07, "loss": 0.0257, "step": 4349 }, { "epoch": 1.71, "learning_rate": 3.7938207632449643e-07, "loss": 0.0293, "step": 4350 }, { "epoch": 1.71, "learning_rate": 3.7937078381029943e-07, "loss": 0.0315, "step": 4351 }, { "epoch": 1.72, "learning_rate": 3.793594883726313e-07, "loss": 0.0294, "step": 4352 }, { "epoch": 1.72, "learning_rate": 3.793481900116761e-07, "loss": 0.0297, "step": 4353 }, { "epoch": 1.72, "learning_rate": 3.79336888727618e-07, "loss": 0.0246, "step": 4354 }, { "epoch": 1.72, "learning_rate": 3.7932558452064126e-07, "loss": 0.0289, "step": 4355 }, { "epoch": 1.72, "learning_rate": 3.7931427739092993e-07, "loss": 0.029, "step": 4356 }, { "epoch": 1.72, "learning_rate": 3.7930296733866837e-07, "loss": 0.0294, "step": 4357 }, { "epoch": 1.72, "learning_rate": 3.7929165436404107e-07, "loss": 0.0279, "step": 4358 }, { "epoch": 1.72, "learning_rate": 3.792803384672322e-07, "loss": 0.0258, "step": 4359 }, { "epoch": 1.72, "learning_rate": 3.7926901964842635e-07, "loss": 0.0285, "step": 4360 }, { "epoch": 1.72, "learning_rate": 3.7925769790780786e-07, "loss": 0.0302, "step": 4361 }, { "epoch": 1.72, "learning_rate": 3.792463732455614e-07, "loss": 0.0229, "step": 4362 }, { "epoch": 1.72, "learning_rate": 3.792350456618715e-07, "loss": 0.0311, "step": 4363 }, { "epoch": 1.72, "learning_rate": 3.792237151569227e-07, "loss": 0.0342, "step": 4364 }, { "epoch": 1.72, "learning_rate": 3.7921238173089973e-07, "loss": 0.0282, "step": 4365 }, { "epoch": 1.72, "learning_rate": 3.7920104538398734e-07, "loss": 0.0249, "step": 4366 }, { "epoch": 1.72, "learning_rate": 3.791897061163703e-07, "loss": 0.0264, "step": 4367 }, { "epoch": 1.72, "learning_rate": 3.791783639282333e-07, "loss": 0.0259, "step": 4368 }, { "epoch": 1.72, "learning_rate": 3.7916701881976135e-07, "loss": 0.0325, "step": 4369 }, { "epoch": 1.72, "learning_rate": 3.791556707911393e-07, "loss": 0.0262, "step": 4370 }, { "epoch": 1.72, "learning_rate": 3.79144319842552e-07, "loss": 0.0283, "step": 4371 }, { "epoch": 1.72, "learning_rate": 3.791329659741847e-07, "loss": 0.0389, "step": 4372 }, { "epoch": 1.72, "learning_rate": 3.7912160918622224e-07, "loss": 0.0304, "step": 4373 }, { "epoch": 1.72, "learning_rate": 3.7911024947884976e-07, "loss": 0.0278, "step": 4374 }, { "epoch": 1.72, "learning_rate": 3.7909888685225245e-07, "loss": 0.0301, "step": 4375 }, { "epoch": 1.72, "learning_rate": 3.790875213066155e-07, "loss": 0.0282, "step": 4376 }, { "epoch": 1.73, "learning_rate": 3.790761528421242e-07, "loss": 0.0279, "step": 4377 }, { "epoch": 1.73, "learning_rate": 3.790647814589636e-07, "loss": 0.0315, "step": 4378 }, { "epoch": 1.73, "learning_rate": 3.7905340715731934e-07, "loss": 0.0316, "step": 4379 }, { "epoch": 1.73, "learning_rate": 3.790420299373766e-07, "loss": 0.0303, "step": 4380 }, { "epoch": 1.73, "learning_rate": 3.7903064979932096e-07, "loss": 0.0306, "step": 4381 }, { "epoch": 1.73, "learning_rate": 3.790192667433378e-07, "loss": 0.0282, "step": 4382 }, { "epoch": 1.73, "learning_rate": 3.790078807696126e-07, "loss": 0.0294, "step": 4383 }, { "epoch": 1.73, "learning_rate": 3.789964918783311e-07, "loss": 0.0273, "step": 4384 }, { "epoch": 1.73, "learning_rate": 3.789851000696788e-07, "loss": 0.0322, "step": 4385 }, { "epoch": 1.73, "learning_rate": 3.7897370534384133e-07, "loss": 0.0279, "step": 4386 }, { "epoch": 1.73, "learning_rate": 3.789623077010045e-07, "loss": 0.0293, "step": 4387 }, { "epoch": 1.73, "learning_rate": 3.789509071413541e-07, "loss": 0.0249, "step": 4388 }, { "epoch": 1.73, "learning_rate": 3.7893950366507584e-07, "loss": 0.0285, "step": 4389 }, { "epoch": 1.73, "learning_rate": 3.7892809727235566e-07, "loss": 0.0267, "step": 4390 }, { "epoch": 1.73, "learning_rate": 3.789166879633794e-07, "loss": 0.0278, "step": 4391 }, { "epoch": 1.73, "learning_rate": 3.789052757383331e-07, "loss": 0.0285, "step": 4392 }, { "epoch": 1.73, "learning_rate": 3.7889386059740265e-07, "loss": 0.0302, "step": 4393 }, { "epoch": 1.73, "learning_rate": 3.788824425407742e-07, "loss": 0.0286, "step": 4394 }, { "epoch": 1.73, "learning_rate": 3.788710215686338e-07, "loss": 0.032, "step": 4395 }, { "epoch": 1.73, "learning_rate": 3.7885959768116755e-07, "loss": 0.0323, "step": 4396 }, { "epoch": 1.73, "learning_rate": 3.7884817087856175e-07, "loss": 0.0328, "step": 4397 }, { "epoch": 1.73, "learning_rate": 3.788367411610026e-07, "loss": 0.0296, "step": 4398 }, { "epoch": 1.73, "learning_rate": 3.7882530852867634e-07, "loss": 0.0286, "step": 4399 }, { "epoch": 1.73, "learning_rate": 3.788138729817693e-07, "loss": 0.0249, "step": 4400 }, { "epoch": 1.73, "learning_rate": 3.788024345204679e-07, "loss": 0.0279, "step": 4401 }, { "epoch": 1.73, "learning_rate": 3.787909931449586e-07, "loss": 0.0352, "step": 4402 }, { "epoch": 1.74, "learning_rate": 3.7877954885542785e-07, "loss": 0.0305, "step": 4403 }, { "epoch": 1.74, "learning_rate": 3.787681016520621e-07, "loss": 0.0255, "step": 4404 }, { "epoch": 1.74, "learning_rate": 3.787566515350481e-07, "loss": 0.0284, "step": 4405 }, { "epoch": 1.74, "learning_rate": 3.787451985045723e-07, "loss": 0.0306, "step": 4406 }, { "epoch": 1.74, "learning_rate": 3.7873374256082144e-07, "loss": 0.0268, "step": 4407 }, { "epoch": 1.74, "learning_rate": 3.7872228370398223e-07, "loss": 0.0295, "step": 4408 }, { "epoch": 1.74, "learning_rate": 3.7871082193424134e-07, "loss": 0.0296, "step": 4409 }, { "epoch": 1.74, "learning_rate": 3.786993572517857e-07, "loss": 0.0321, "step": 4410 }, { "epoch": 1.74, "learning_rate": 3.786878896568022e-07, "loss": 0.0294, "step": 4411 }, { "epoch": 1.74, "learning_rate": 3.7867641914947763e-07, "loss": 0.0291, "step": 4412 }, { "epoch": 1.74, "learning_rate": 3.7866494572999896e-07, "loss": 0.0308, "step": 4413 }, { "epoch": 1.74, "learning_rate": 3.7865346939855325e-07, "loss": 0.0293, "step": 4414 }, { "epoch": 1.74, "learning_rate": 3.7864199015532745e-07, "loss": 0.0296, "step": 4415 }, { "epoch": 1.74, "learning_rate": 3.786305080005088e-07, "loss": 0.0302, "step": 4416 }, { "epoch": 1.74, "learning_rate": 3.7861902293428426e-07, "loss": 0.0282, "step": 4417 }, { "epoch": 1.74, "learning_rate": 3.7860753495684116e-07, "loss": 0.03, "step": 4418 }, { "epoch": 1.74, "learning_rate": 3.7859604406836674e-07, "loss": 0.0295, "step": 4419 }, { "epoch": 1.74, "learning_rate": 3.785845502690482e-07, "loss": 0.0317, "step": 4420 }, { "epoch": 1.74, "learning_rate": 3.785730535590729e-07, "loss": 0.0274, "step": 4421 }, { "epoch": 1.74, "learning_rate": 3.785615539386282e-07, "loss": 0.032, "step": 4422 }, { "epoch": 1.74, "learning_rate": 3.785500514079016e-07, "loss": 0.0316, "step": 4423 }, { "epoch": 1.74, "learning_rate": 3.785385459670805e-07, "loss": 0.0292, "step": 4424 }, { "epoch": 1.74, "learning_rate": 3.7852703761635246e-07, "loss": 0.034, "step": 4425 }, { "epoch": 1.74, "learning_rate": 3.7851552635590503e-07, "loss": 0.0334, "step": 4426 }, { "epoch": 1.74, "learning_rate": 3.785040121859259e-07, "loss": 0.0296, "step": 4427 }, { "epoch": 1.75, "learning_rate": 3.784924951066026e-07, "loss": 0.0303, "step": 4428 }, { "epoch": 1.75, "learning_rate": 3.784809751181229e-07, "loss": 0.0282, "step": 4429 }, { "epoch": 1.75, "learning_rate": 3.784694522206746e-07, "loss": 0.0271, "step": 4430 }, { "epoch": 1.75, "learning_rate": 3.784579264144454e-07, "loss": 0.0268, "step": 4431 }, { "epoch": 1.75, "learning_rate": 3.784463976996233e-07, "loss": 0.0267, "step": 4432 }, { "epoch": 1.75, "learning_rate": 3.7843486607639613e-07, "loss": 0.0315, "step": 4433 }, { "epoch": 1.75, "learning_rate": 3.7842333154495176e-07, "loss": 0.0265, "step": 4434 }, { "epoch": 1.75, "learning_rate": 3.784117941054783e-07, "loss": 0.0301, "step": 4435 }, { "epoch": 1.75, "learning_rate": 3.7840025375816375e-07, "loss": 0.0271, "step": 4436 }, { "epoch": 1.75, "learning_rate": 3.783887105031962e-07, "loss": 0.0317, "step": 4437 }, { "epoch": 1.75, "learning_rate": 3.783771643407638e-07, "loss": 0.0272, "step": 4438 }, { "epoch": 1.75, "learning_rate": 3.7836561527105467e-07, "loss": 0.0287, "step": 4439 }, { "epoch": 1.75, "learning_rate": 3.7835406329425715e-07, "loss": 0.0312, "step": 4440 }, { "epoch": 1.75, "learning_rate": 3.7834250841055944e-07, "loss": 0.0253, "step": 4441 }, { "epoch": 1.75, "learning_rate": 3.783309506201499e-07, "loss": 0.03, "step": 4442 }, { "epoch": 1.75, "learning_rate": 3.783193899232169e-07, "loss": 0.0267, "step": 4443 }, { "epoch": 1.75, "learning_rate": 3.783078263199489e-07, "loss": 0.0251, "step": 4444 }, { "epoch": 1.75, "learning_rate": 3.7829625981053427e-07, "loss": 0.0287, "step": 4445 }, { "epoch": 1.75, "learning_rate": 3.782846903951616e-07, "loss": 0.0305, "step": 4446 }, { "epoch": 1.75, "learning_rate": 3.7827311807401945e-07, "loss": 0.0285, "step": 4447 }, { "epoch": 1.75, "learning_rate": 3.782615428472964e-07, "loss": 0.0295, "step": 4448 }, { "epoch": 1.75, "learning_rate": 3.7824996471518117e-07, "loss": 0.0321, "step": 4449 }, { "epoch": 1.75, "learning_rate": 3.782383836778624e-07, "loss": 0.0311, "step": 4450 }, { "epoch": 1.75, "learning_rate": 3.782267997355288e-07, "loss": 0.0271, "step": 4451 }, { "epoch": 1.75, "learning_rate": 3.782152128883693e-07, "loss": 0.0304, "step": 4452 }, { "epoch": 1.75, "learning_rate": 3.782036231365727e-07, "loss": 0.029, "step": 4453 }, { "epoch": 1.76, "learning_rate": 3.781920304803279e-07, "loss": 0.0291, "step": 4454 }, { "epoch": 1.76, "learning_rate": 3.7818043491982373e-07, "loss": 0.0272, "step": 4455 }, { "epoch": 1.76, "learning_rate": 3.7816883645524936e-07, "loss": 0.0298, "step": 4456 }, { "epoch": 1.76, "learning_rate": 3.781572350867937e-07, "loss": 0.0297, "step": 4457 }, { "epoch": 1.76, "learning_rate": 3.7814563081464586e-07, "loss": 0.0275, "step": 4458 }, { "epoch": 1.76, "learning_rate": 3.781340236389951e-07, "loss": 0.028, "step": 4459 }, { "epoch": 1.76, "learning_rate": 3.7812241356003036e-07, "loss": 0.031, "step": 4460 }, { "epoch": 1.76, "learning_rate": 3.7811080057794104e-07, "loss": 0.0267, "step": 4461 }, { "epoch": 1.76, "learning_rate": 3.7809918469291634e-07, "loss": 0.0257, "step": 4462 }, { "epoch": 1.76, "learning_rate": 3.780875659051457e-07, "loss": 0.0285, "step": 4463 }, { "epoch": 1.76, "learning_rate": 3.780759442148183e-07, "loss": 0.0278, "step": 4464 }, { "epoch": 1.76, "learning_rate": 3.7806431962212366e-07, "loss": 0.0355, "step": 4465 }, { "epoch": 1.76, "learning_rate": 3.780526921272513e-07, "loss": 0.0239, "step": 4466 }, { "epoch": 1.76, "learning_rate": 3.7804106173039063e-07, "loss": 0.0299, "step": 4467 }, { "epoch": 1.76, "learning_rate": 3.780294284317312e-07, "loss": 0.0257, "step": 4468 }, { "epoch": 1.76, "learning_rate": 3.7801779223146274e-07, "loss": 0.0315, "step": 4469 }, { "epoch": 1.76, "learning_rate": 3.7800615312977477e-07, "loss": 0.029, "step": 4470 }, { "epoch": 1.76, "learning_rate": 3.7799451112685707e-07, "loss": 0.0317, "step": 4471 }, { "epoch": 1.76, "learning_rate": 3.7798286622289933e-07, "loss": 0.0249, "step": 4472 }, { "epoch": 1.76, "learning_rate": 3.779712184180914e-07, "loss": 0.0275, "step": 4473 }, { "epoch": 1.76, "learning_rate": 3.779595677126231e-07, "loss": 0.0267, "step": 4474 }, { "epoch": 1.76, "learning_rate": 3.779479141066843e-07, "loss": 0.0299, "step": 4475 }, { "epoch": 1.76, "learning_rate": 3.779362576004649e-07, "loss": 0.0352, "step": 4476 }, { "epoch": 1.76, "learning_rate": 3.7792459819415503e-07, "loss": 0.024, "step": 4477 }, { "epoch": 1.76, "learning_rate": 3.7791293588794457e-07, "loss": 0.0354, "step": 4478 }, { "epoch": 1.77, "learning_rate": 3.7790127068202364e-07, "loss": 0.0288, "step": 4479 }, { "epoch": 1.77, "learning_rate": 3.7788960257658246e-07, "loss": 0.0308, "step": 4480 }, { "epoch": 1.77, "learning_rate": 3.7787793157181104e-07, "loss": 0.0284, "step": 4481 }, { "epoch": 1.77, "learning_rate": 3.778662576678997e-07, "loss": 0.0313, "step": 4482 }, { "epoch": 1.77, "learning_rate": 3.778545808650387e-07, "loss": 0.0259, "step": 4483 }, { "epoch": 1.77, "learning_rate": 3.778429011634182e-07, "loss": 0.0308, "step": 4484 }, { "epoch": 1.77, "learning_rate": 3.7783121856322885e-07, "loss": 0.0346, "step": 4485 }, { "epoch": 1.77, "learning_rate": 3.778195330646609e-07, "loss": 0.0311, "step": 4486 }, { "epoch": 1.77, "learning_rate": 3.778078446679048e-07, "loss": 0.0322, "step": 4487 }, { "epoch": 1.77, "learning_rate": 3.777961533731511e-07, "loss": 0.0311, "step": 4488 }, { "epoch": 1.77, "learning_rate": 3.7778445918059024e-07, "loss": 0.0298, "step": 4489 }, { "epoch": 1.77, "learning_rate": 3.777727620904129e-07, "loss": 0.0325, "step": 4490 }, { "epoch": 1.77, "learning_rate": 3.7776106210280985e-07, "loss": 0.0287, "step": 4491 }, { "epoch": 1.77, "learning_rate": 3.777493592179715e-07, "loss": 0.0303, "step": 4492 }, { "epoch": 1.77, "learning_rate": 3.7773765343608883e-07, "loss": 0.031, "step": 4493 }, { "epoch": 1.77, "learning_rate": 3.777259447573525e-07, "loss": 0.024, "step": 4494 }, { "epoch": 1.77, "learning_rate": 3.777142331819534e-07, "loss": 0.0293, "step": 4495 }, { "epoch": 1.77, "learning_rate": 3.7770251871008236e-07, "loss": 0.0274, "step": 4496 }, { "epoch": 1.77, "learning_rate": 3.7769080134193046e-07, "loss": 0.0267, "step": 4497 }, { "epoch": 1.77, "learning_rate": 3.776790810776884e-07, "loss": 0.0303, "step": 4498 }, { "epoch": 1.77, "learning_rate": 3.7766735791754745e-07, "loss": 0.023, "step": 4499 }, { "epoch": 1.77, "learning_rate": 3.776556318616985e-07, "loss": 0.0293, "step": 4500 }, { "epoch": 1.77, "learning_rate": 3.7764390291033294e-07, "loss": 0.0267, "step": 4501 }, { "epoch": 1.77, "learning_rate": 3.7763217106364157e-07, "loss": 0.0274, "step": 4502 }, { "epoch": 1.77, "learning_rate": 3.7762043632181583e-07, "loss": 0.0293, "step": 4503 }, { "epoch": 1.78, "learning_rate": 3.77608698685047e-07, "loss": 0.0275, "step": 4504 }, { "epoch": 1.78, "learning_rate": 3.7759695815352623e-07, "loss": 0.031, "step": 4505 }, { "epoch": 1.78, "learning_rate": 3.7758521472744496e-07, "loss": 0.0286, "step": 4506 }, { "epoch": 1.78, "learning_rate": 3.7757346840699467e-07, "loss": 0.029, "step": 4507 }, { "epoch": 1.78, "learning_rate": 3.775617191923667e-07, "loss": 0.0287, "step": 4508 }, { "epoch": 1.78, "learning_rate": 3.775499670837525e-07, "loss": 0.0305, "step": 4509 }, { "epoch": 1.78, "learning_rate": 3.7753821208134375e-07, "loss": 0.0328, "step": 4510 }, { "epoch": 1.78, "learning_rate": 3.7752645418533194e-07, "loss": 0.0303, "step": 4511 }, { "epoch": 1.78, "learning_rate": 3.7751469339590873e-07, "loss": 0.0292, "step": 4512 }, { "epoch": 1.78, "learning_rate": 3.775029297132658e-07, "loss": 0.0272, "step": 4513 }, { "epoch": 1.78, "learning_rate": 3.7749116313759495e-07, "loss": 0.0261, "step": 4514 }, { "epoch": 1.78, "learning_rate": 3.7747939366908783e-07, "loss": 0.0324, "step": 4515 }, { "epoch": 1.78, "learning_rate": 3.7746762130793634e-07, "loss": 0.0269, "step": 4516 }, { "epoch": 1.78, "learning_rate": 3.774558460543324e-07, "loss": 0.03, "step": 4517 }, { "epoch": 1.78, "learning_rate": 3.7744406790846783e-07, "loss": 0.0287, "step": 4518 }, { "epoch": 1.78, "learning_rate": 3.774322868705346e-07, "loss": 0.0283, "step": 4519 }, { "epoch": 1.78, "learning_rate": 3.774205029407248e-07, "loss": 0.0281, "step": 4520 }, { "epoch": 1.78, "learning_rate": 3.7740871611923047e-07, "loss": 0.0302, "step": 4521 }, { "epoch": 1.78, "learning_rate": 3.7739692640624367e-07, "loss": 0.0266, "step": 4522 }, { "epoch": 1.78, "learning_rate": 3.773851338019566e-07, "loss": 0.0268, "step": 4523 }, { "epoch": 1.78, "learning_rate": 3.7737333830656143e-07, "loss": 0.0228, "step": 4524 }, { "epoch": 1.78, "learning_rate": 3.7736153992025043e-07, "loss": 0.0311, "step": 4525 }, { "epoch": 1.78, "learning_rate": 3.773497386432159e-07, "loss": 0.0321, "step": 4526 }, { "epoch": 1.78, "learning_rate": 3.773379344756501e-07, "loss": 0.0272, "step": 4527 }, { "epoch": 1.78, "learning_rate": 3.7732612741774557e-07, "loss": 0.0248, "step": 4528 }, { "epoch": 1.78, "learning_rate": 3.773143174696946e-07, "loss": 0.0295, "step": 4529 }, { "epoch": 1.79, "learning_rate": 3.773025046316898e-07, "loss": 0.0266, "step": 4530 }, { "epoch": 1.79, "learning_rate": 3.772906889039236e-07, "loss": 0.03, "step": 4531 }, { "epoch": 1.79, "learning_rate": 3.7727887028658865e-07, "loss": 0.0337, "step": 4532 }, { "epoch": 1.79, "learning_rate": 3.7726704877987756e-07, "loss": 0.0283, "step": 4533 }, { "epoch": 1.79, "learning_rate": 3.7725522438398303e-07, "loss": 0.0318, "step": 4534 }, { "epoch": 1.79, "learning_rate": 3.772433970990976e-07, "loss": 0.0296, "step": 4535 }, { "epoch": 1.79, "learning_rate": 3.7723156692541427e-07, "loss": 0.0266, "step": 4536 }, { "epoch": 1.79, "learning_rate": 3.772197338631258e-07, "loss": 0.0239, "step": 4537 }, { "epoch": 1.79, "learning_rate": 3.7720789791242497e-07, "loss": 0.0253, "step": 4538 }, { "epoch": 1.79, "learning_rate": 3.7719605907350463e-07, "loss": 0.0284, "step": 4539 }, { "epoch": 1.79, "learning_rate": 3.7718421734655796e-07, "loss": 0.0264, "step": 4540 }, { "epoch": 1.79, "learning_rate": 3.771723727317778e-07, "loss": 0.0306, "step": 4541 }, { "epoch": 1.79, "learning_rate": 3.771605252293573e-07, "loss": 0.0273, "step": 4542 }, { "epoch": 1.79, "learning_rate": 3.771486748394894e-07, "loss": 0.03, "step": 4543 }, { "epoch": 1.79, "learning_rate": 3.771368215623673e-07, "loss": 0.0268, "step": 4544 }, { "epoch": 1.79, "learning_rate": 3.771249653981843e-07, "loss": 0.0313, "step": 4545 }, { "epoch": 1.79, "learning_rate": 3.7711310634713354e-07, "loss": 0.0292, "step": 4546 }, { "epoch": 1.79, "learning_rate": 3.7710124440940837e-07, "loss": 0.0292, "step": 4547 }, { "epoch": 1.79, "learning_rate": 3.7708937958520206e-07, "loss": 0.03, "step": 4548 }, { "epoch": 1.79, "learning_rate": 3.77077511874708e-07, "loss": 0.0348, "step": 4549 }, { "epoch": 1.79, "learning_rate": 3.770656412781196e-07, "loss": 0.0323, "step": 4550 }, { "epoch": 1.79, "learning_rate": 3.7705376779563034e-07, "loss": 0.0277, "step": 4551 }, { "epoch": 1.79, "learning_rate": 3.770418914274338e-07, "loss": 0.0278, "step": 4552 }, { "epoch": 1.79, "learning_rate": 3.770300121737235e-07, "loss": 0.0322, "step": 4553 }, { "epoch": 1.79, "learning_rate": 3.77018130034693e-07, "loss": 0.0317, "step": 4554 }, { "epoch": 1.8, "learning_rate": 3.770062450105361e-07, "loss": 0.0303, "step": 4555 }, { "epoch": 1.8, "learning_rate": 3.769943571014464e-07, "loss": 0.0286, "step": 4556 }, { "epoch": 1.8, "learning_rate": 3.7698246630761767e-07, "loss": 0.0311, "step": 4557 }, { "epoch": 1.8, "learning_rate": 3.769705726292437e-07, "loss": 0.0307, "step": 4558 }, { "epoch": 1.8, "learning_rate": 3.7695867606651837e-07, "loss": 0.0279, "step": 4559 }, { "epoch": 1.8, "learning_rate": 3.7694677661963557e-07, "loss": 0.0263, "step": 4560 }, { "epoch": 1.8, "learning_rate": 3.769348742887893e-07, "loss": 0.0251, "step": 4561 }, { "epoch": 1.8, "learning_rate": 3.769229690741734e-07, "loss": 0.0285, "step": 4562 }, { "epoch": 1.8, "learning_rate": 3.7691106097598205e-07, "loss": 0.0265, "step": 4563 }, { "epoch": 1.8, "learning_rate": 3.768991499944093e-07, "loss": 0.0336, "step": 4564 }, { "epoch": 1.8, "learning_rate": 3.7688723612964915e-07, "loss": 0.0273, "step": 4565 }, { "epoch": 1.8, "learning_rate": 3.76875319381896e-07, "loss": 0.0281, "step": 4566 }, { "epoch": 1.8, "learning_rate": 3.768633997513439e-07, "loss": 0.0282, "step": 4567 }, { "epoch": 1.8, "learning_rate": 3.768514772381872e-07, "loss": 0.0326, "step": 4568 }, { "epoch": 1.8, "learning_rate": 3.768395518426202e-07, "loss": 0.0289, "step": 4569 }, { "epoch": 1.8, "learning_rate": 3.768276235648373e-07, "loss": 0.0324, "step": 4570 }, { "epoch": 1.8, "learning_rate": 3.768156924050329e-07, "loss": 0.0299, "step": 4571 }, { "epoch": 1.8, "learning_rate": 3.768037583634014e-07, "loss": 0.0278, "step": 4572 }, { "epoch": 1.8, "learning_rate": 3.767918214401374e-07, "loss": 0.0266, "step": 4573 }, { "epoch": 1.8, "learning_rate": 3.7677988163543534e-07, "loss": 0.0315, "step": 4574 }, { "epoch": 1.8, "learning_rate": 3.767679389494899e-07, "loss": 0.0322, "step": 4575 }, { "epoch": 1.8, "learning_rate": 3.767559933824957e-07, "loss": 0.0307, "step": 4576 }, { "epoch": 1.8, "learning_rate": 3.7674404493464747e-07, "loss": 0.0275, "step": 4577 }, { "epoch": 1.8, "learning_rate": 3.7673209360613993e-07, "loss": 0.026, "step": 4578 }, { "epoch": 1.8, "learning_rate": 3.7672013939716785e-07, "loss": 0.0286, "step": 4579 }, { "epoch": 1.81, "learning_rate": 3.767081823079261e-07, "loss": 0.0323, "step": 4580 }, { "epoch": 1.81, "learning_rate": 3.7669622233860945e-07, "loss": 0.0278, "step": 4581 }, { "epoch": 1.81, "learning_rate": 3.766842594894131e-07, "loss": 0.0321, "step": 4582 }, { "epoch": 1.81, "learning_rate": 3.766722937605317e-07, "loss": 0.028, "step": 4583 }, { "epoch": 1.81, "learning_rate": 3.766603251521605e-07, "loss": 0.0274, "step": 4584 }, { "epoch": 1.81, "learning_rate": 3.7664835366449447e-07, "loss": 0.0303, "step": 4585 }, { "epoch": 1.81, "learning_rate": 3.7663637929772876e-07, "loss": 0.0272, "step": 4586 }, { "epoch": 1.81, "learning_rate": 3.766244020520585e-07, "loss": 0.0313, "step": 4587 }, { "epoch": 1.81, "learning_rate": 3.7661242192767893e-07, "loss": 0.0263, "step": 4588 }, { "epoch": 1.81, "learning_rate": 3.766004389247853e-07, "loss": 0.0245, "step": 4589 }, { "epoch": 1.81, "learning_rate": 3.76588453043573e-07, "loss": 0.0295, "step": 4590 }, { "epoch": 1.81, "learning_rate": 3.765764642842372e-07, "loss": 0.0258, "step": 4591 }, { "epoch": 1.81, "learning_rate": 3.765644726469734e-07, "loss": 0.0292, "step": 4592 }, { "epoch": 1.81, "learning_rate": 3.7655247813197706e-07, "loss": 0.0349, "step": 4593 }, { "epoch": 1.81, "learning_rate": 3.7654048073944367e-07, "loss": 0.0256, "step": 4594 }, { "epoch": 1.81, "learning_rate": 3.7652848046956874e-07, "loss": 0.0285, "step": 4595 }, { "epoch": 1.81, "learning_rate": 3.765164773225479e-07, "loss": 0.0302, "step": 4596 }, { "epoch": 1.81, "learning_rate": 3.765044712985767e-07, "loss": 0.0292, "step": 4597 }, { "epoch": 1.81, "learning_rate": 3.7649246239785094e-07, "loss": 0.0275, "step": 4598 }, { "epoch": 1.81, "learning_rate": 3.7648045062056623e-07, "loss": 0.0286, "step": 4599 }, { "epoch": 1.81, "learning_rate": 3.764684359669184e-07, "loss": 0.0241, "step": 4600 }, { "epoch": 1.81, "learning_rate": 3.764564184371033e-07, "loss": 0.0294, "step": 4601 }, { "epoch": 1.81, "learning_rate": 3.7644439803131673e-07, "loss": 0.0294, "step": 4602 }, { "epoch": 1.81, "learning_rate": 3.7643237474975463e-07, "loss": 0.029, "step": 4603 }, { "epoch": 1.81, "learning_rate": 3.76420348592613e-07, "loss": 0.0264, "step": 4604 }, { "epoch": 1.81, "learning_rate": 3.764083195600878e-07, "loss": 0.0275, "step": 4605 }, { "epoch": 1.82, "learning_rate": 3.763962876523751e-07, "loss": 0.026, "step": 4606 }, { "epoch": 1.82, "learning_rate": 3.7638425286967097e-07, "loss": 0.0266, "step": 4607 }, { "epoch": 1.82, "learning_rate": 3.7637221521217166e-07, "loss": 0.0278, "step": 4608 }, { "epoch": 1.82, "learning_rate": 3.7636017468007324e-07, "loss": 0.023, "step": 4609 }, { "epoch": 1.82, "learning_rate": 3.7634813127357206e-07, "loss": 0.0289, "step": 4610 }, { "epoch": 1.82, "learning_rate": 3.763360849928643e-07, "loss": 0.0315, "step": 4611 }, { "epoch": 1.82, "learning_rate": 3.7632403583814635e-07, "loss": 0.0312, "step": 4612 }, { "epoch": 1.82, "learning_rate": 3.7631198380961464e-07, "loss": 0.0297, "step": 4613 }, { "epoch": 1.82, "learning_rate": 3.762999289074656e-07, "loss": 0.0268, "step": 4614 }, { "epoch": 1.82, "learning_rate": 3.7628787113189557e-07, "loss": 0.0311, "step": 4615 }, { "epoch": 1.82, "learning_rate": 3.762758104831012e-07, "loss": 0.0284, "step": 4616 }, { "epoch": 1.82, "learning_rate": 3.7626374696127905e-07, "loss": 0.0291, "step": 4617 }, { "epoch": 1.82, "learning_rate": 3.7625168056662566e-07, "loss": 0.0279, "step": 4618 }, { "epoch": 1.82, "learning_rate": 3.7623961129933785e-07, "loss": 0.0313, "step": 4619 }, { "epoch": 1.82, "learning_rate": 3.7622753915961213e-07, "loss": 0.0286, "step": 4620 }, { "epoch": 1.82, "learning_rate": 3.7621546414764544e-07, "loss": 0.0321, "step": 4621 }, { "epoch": 1.82, "learning_rate": 3.762033862636344e-07, "loss": 0.0277, "step": 4622 }, { "epoch": 1.82, "learning_rate": 3.761913055077761e-07, "loss": 0.0273, "step": 4623 }, { "epoch": 1.82, "learning_rate": 3.7617922188026714e-07, "loss": 0.0296, "step": 4624 }, { "epoch": 1.82, "learning_rate": 3.7616713538130474e-07, "loss": 0.0303, "step": 4625 }, { "epoch": 1.82, "learning_rate": 3.7615504601108573e-07, "loss": 0.0292, "step": 4626 }, { "epoch": 1.82, "learning_rate": 3.761429537698072e-07, "loss": 0.028, "step": 4627 }, { "epoch": 1.82, "learning_rate": 3.7613085865766625e-07, "loss": 0.0282, "step": 4628 }, { "epoch": 1.82, "learning_rate": 3.7611876067485996e-07, "loss": 0.0292, "step": 4629 }, { "epoch": 1.82, "learning_rate": 3.761066598215856e-07, "loss": 0.0309, "step": 4630 }, { "epoch": 1.83, "learning_rate": 3.760945560980403e-07, "loss": 0.0276, "step": 4631 }, { "epoch": 1.83, "learning_rate": 3.760824495044213e-07, "loss": 0.0259, "step": 4632 }, { "epoch": 1.83, "learning_rate": 3.7607034004092606e-07, "loss": 0.0345, "step": 4633 }, { "epoch": 1.83, "learning_rate": 3.760582277077519e-07, "loss": 0.0275, "step": 4634 }, { "epoch": 1.83, "learning_rate": 3.760461125050962e-07, "loss": 0.0232, "step": 4635 }, { "epoch": 1.83, "learning_rate": 3.760339944331563e-07, "loss": 0.0304, "step": 4636 }, { "epoch": 1.83, "learning_rate": 3.7602187349212997e-07, "loss": 0.0267, "step": 4637 }, { "epoch": 1.83, "learning_rate": 3.760097496822146e-07, "loss": 0.0239, "step": 4638 }, { "epoch": 1.83, "learning_rate": 3.7599762300360776e-07, "loss": 0.0292, "step": 4639 }, { "epoch": 1.83, "learning_rate": 3.759854934565072e-07, "loss": 0.0266, "step": 4640 }, { "epoch": 1.83, "learning_rate": 3.7597336104111055e-07, "loss": 0.0298, "step": 4641 }, { "epoch": 1.83, "learning_rate": 3.7596122575761553e-07, "loss": 0.0254, "step": 4642 }, { "epoch": 1.83, "learning_rate": 3.7594908760622e-07, "loss": 0.0286, "step": 4643 }, { "epoch": 1.83, "learning_rate": 3.7593694658712174e-07, "loss": 0.0307, "step": 4644 }, { "epoch": 1.83, "learning_rate": 3.759248027005187e-07, "loss": 0.0272, "step": 4645 }, { "epoch": 1.83, "learning_rate": 3.759126559466087e-07, "loss": 0.0231, "step": 4646 }, { "epoch": 1.83, "learning_rate": 3.7590050632558973e-07, "loss": 0.0242, "step": 4647 }, { "epoch": 1.83, "learning_rate": 3.758883538376599e-07, "loss": 0.029, "step": 4648 }, { "epoch": 1.83, "learning_rate": 3.758761984830172e-07, "loss": 0.0302, "step": 4649 }, { "epoch": 1.83, "learning_rate": 3.758640402618598e-07, "loss": 0.0252, "step": 4650 }, { "epoch": 1.83, "learning_rate": 3.758518791743857e-07, "loss": 0.0329, "step": 4651 }, { "epoch": 1.83, "learning_rate": 3.7583971522079336e-07, "loss": 0.0262, "step": 4652 }, { "epoch": 1.83, "learning_rate": 3.758275484012809e-07, "loss": 0.0302, "step": 4653 }, { "epoch": 1.83, "learning_rate": 3.758153787160466e-07, "loss": 0.0244, "step": 4654 }, { "epoch": 1.83, "learning_rate": 3.7580320616528884e-07, "loss": 0.0279, "step": 4655 }, { "epoch": 1.83, "learning_rate": 3.7579103074920597e-07, "loss": 0.0292, "step": 4656 }, { "epoch": 1.84, "learning_rate": 3.7577885246799657e-07, "loss": 0.0269, "step": 4657 }, { "epoch": 1.84, "learning_rate": 3.7576667132185894e-07, "loss": 0.0311, "step": 4658 }, { "epoch": 1.84, "learning_rate": 3.7575448731099173e-07, "loss": 0.0261, "step": 4659 }, { "epoch": 1.84, "learning_rate": 3.757423004355935e-07, "loss": 0.0262, "step": 4660 }, { "epoch": 1.84, "learning_rate": 3.7573011069586284e-07, "loss": 0.0289, "step": 4661 }, { "epoch": 1.84, "learning_rate": 3.757179180919985e-07, "loss": 0.0309, "step": 4662 }, { "epoch": 1.84, "learning_rate": 3.7570572262419913e-07, "loss": 0.0265, "step": 4663 }, { "epoch": 1.84, "learning_rate": 3.7569352429266356e-07, "loss": 0.0276, "step": 4664 }, { "epoch": 1.84, "learning_rate": 3.756813230975905e-07, "loss": 0.0314, "step": 4665 }, { "epoch": 1.84, "learning_rate": 3.7566911903917894e-07, "loss": 0.0276, "step": 4666 }, { "epoch": 1.84, "learning_rate": 3.7565691211762766e-07, "loss": 0.0231, "step": 4667 }, { "epoch": 1.84, "learning_rate": 3.7564470233313573e-07, "loss": 0.0275, "step": 4668 }, { "epoch": 1.84, "learning_rate": 3.756324896859021e-07, "loss": 0.0255, "step": 4669 }, { "epoch": 1.84, "learning_rate": 3.756202741761258e-07, "loss": 0.0266, "step": 4670 }, { "epoch": 1.84, "learning_rate": 3.756080558040059e-07, "loss": 0.0292, "step": 4671 }, { "epoch": 1.84, "learning_rate": 3.755958345697417e-07, "loss": 0.026, "step": 4672 }, { "epoch": 1.84, "learning_rate": 3.755836104735322e-07, "loss": 0.0291, "step": 4673 }, { "epoch": 1.84, "learning_rate": 3.7557138351557673e-07, "loss": 0.0288, "step": 4674 }, { "epoch": 1.84, "learning_rate": 3.755591536960745e-07, "loss": 0.0291, "step": 4675 }, { "epoch": 1.84, "learning_rate": 3.755469210152249e-07, "loss": 0.0279, "step": 4676 }, { "epoch": 1.84, "learning_rate": 3.7553468547322733e-07, "loss": 0.026, "step": 4677 }, { "epoch": 1.84, "learning_rate": 3.755224470702811e-07, "loss": 0.0264, "step": 4678 }, { "epoch": 1.84, "learning_rate": 3.755102058065858e-07, "loss": 0.0262, "step": 4679 }, { "epoch": 1.84, "learning_rate": 3.7549796168234085e-07, "loss": 0.0289, "step": 4680 }, { "epoch": 1.84, "learning_rate": 3.754857146977459e-07, "loss": 0.0294, "step": 4681 }, { "epoch": 1.85, "learning_rate": 3.754734648530005e-07, "loss": 0.0285, "step": 4682 }, { "epoch": 1.85, "learning_rate": 3.754612121483043e-07, "loss": 0.0258, "step": 4683 }, { "epoch": 1.85, "learning_rate": 3.75448956583857e-07, "loss": 0.0279, "step": 4684 }, { "epoch": 1.85, "learning_rate": 3.754366981598584e-07, "loss": 0.0281, "step": 4685 }, { "epoch": 1.85, "learning_rate": 3.7542443687650817e-07, "loss": 0.0313, "step": 4686 }, { "epoch": 1.85, "learning_rate": 3.7541217273400634e-07, "loss": 0.0281, "step": 4687 }, { "epoch": 1.85, "learning_rate": 3.753999057325526e-07, "loss": 0.0326, "step": 4688 }, { "epoch": 1.85, "learning_rate": 3.7538763587234703e-07, "loss": 0.0266, "step": 4689 }, { "epoch": 1.85, "learning_rate": 3.753753631535895e-07, "loss": 0.0261, "step": 4690 }, { "epoch": 1.85, "learning_rate": 3.753630875764801e-07, "loss": 0.0303, "step": 4691 }, { "epoch": 1.85, "learning_rate": 3.75350809141219e-07, "loss": 0.0242, "step": 4692 }, { "epoch": 1.85, "learning_rate": 3.7533852784800606e-07, "loss": 0.0267, "step": 4693 }, { "epoch": 1.85, "learning_rate": 3.7532624369704167e-07, "loss": 0.028, "step": 4694 }, { "epoch": 1.85, "learning_rate": 3.7531395668852597e-07, "loss": 0.0295, "step": 4695 }, { "epoch": 1.85, "learning_rate": 3.7530166682265924e-07, "loss": 0.033, "step": 4696 }, { "epoch": 1.85, "learning_rate": 3.752893740996417e-07, "loss": 0.0271, "step": 4697 }, { "epoch": 1.85, "learning_rate": 3.752770785196739e-07, "loss": 0.0251, "step": 4698 }, { "epoch": 1.85, "learning_rate": 3.7526478008295604e-07, "loss": 0.0293, "step": 4699 }, { "epoch": 1.85, "learning_rate": 3.752524787896886e-07, "loss": 0.0309, "step": 4700 }, { "epoch": 1.85, "learning_rate": 3.7524017464007216e-07, "loss": 0.0243, "step": 4701 }, { "epoch": 1.85, "learning_rate": 3.752278676343072e-07, "loss": 0.029, "step": 4702 }, { "epoch": 1.85, "learning_rate": 3.752155577725943e-07, "loss": 0.0291, "step": 4703 }, { "epoch": 1.85, "learning_rate": 3.7520324505513405e-07, "loss": 0.024, "step": 4704 }, { "epoch": 1.85, "learning_rate": 3.7519092948212723e-07, "loss": 0.0285, "step": 4705 }, { "epoch": 1.85, "learning_rate": 3.7517861105377457e-07, "loss": 0.0292, "step": 4706 }, { "epoch": 1.86, "learning_rate": 3.7516628977027676e-07, "loss": 0.0339, "step": 4707 }, { "epoch": 1.86, "learning_rate": 3.7515396563183465e-07, "loss": 0.0316, "step": 4708 }, { "epoch": 1.86, "learning_rate": 3.7514163863864905e-07, "loss": 0.029, "step": 4709 }, { "epoch": 1.86, "learning_rate": 3.75129308790921e-07, "loss": 0.0243, "step": 4710 }, { "epoch": 1.86, "learning_rate": 3.7511697608885135e-07, "loss": 0.0294, "step": 4711 }, { "epoch": 1.86, "learning_rate": 3.751046405326411e-07, "loss": 0.0308, "step": 4712 }, { "epoch": 1.86, "learning_rate": 3.750923021224914e-07, "loss": 0.0284, "step": 4713 }, { "epoch": 1.86, "learning_rate": 3.7507996085860327e-07, "loss": 0.0313, "step": 4714 }, { "epoch": 1.86, "learning_rate": 3.7506761674117783e-07, "loss": 0.0278, "step": 4715 }, { "epoch": 1.86, "learning_rate": 3.7505526977041634e-07, "loss": 0.0257, "step": 4716 }, { "epoch": 1.86, "learning_rate": 3.7504291994652004e-07, "loss": 0.0275, "step": 4717 }, { "epoch": 1.86, "learning_rate": 3.7503056726969014e-07, "loss": 0.0272, "step": 4718 }, { "epoch": 1.86, "learning_rate": 3.7501821174012797e-07, "loss": 0.0293, "step": 4719 }, { "epoch": 1.86, "learning_rate": 3.75005853358035e-07, "loss": 0.0275, "step": 4720 }, { "epoch": 1.86, "learning_rate": 3.749934921236126e-07, "loss": 0.0281, "step": 4721 }, { "epoch": 1.86, "learning_rate": 3.749811280370622e-07, "loss": 0.0245, "step": 4722 }, { "epoch": 1.86, "learning_rate": 3.749687610985854e-07, "loss": 0.0305, "step": 4723 }, { "epoch": 1.86, "learning_rate": 3.749563913083837e-07, "loss": 0.027, "step": 4724 }, { "epoch": 1.86, "learning_rate": 3.7494401866665867e-07, "loss": 0.0312, "step": 4725 }, { "epoch": 1.86, "learning_rate": 3.7493164317361207e-07, "loss": 0.0256, "step": 4726 }, { "epoch": 1.86, "learning_rate": 3.749192648294456e-07, "loss": 0.0252, "step": 4727 }, { "epoch": 1.86, "learning_rate": 3.7490688363436085e-07, "loss": 0.0257, "step": 4728 }, { "epoch": 1.86, "learning_rate": 3.7489449958855984e-07, "loss": 0.0278, "step": 4729 }, { "epoch": 1.86, "learning_rate": 3.748821126922442e-07, "loss": 0.0286, "step": 4730 }, { "epoch": 1.86, "learning_rate": 3.7486972294561595e-07, "loss": 0.0292, "step": 4731 }, { "epoch": 1.86, "learning_rate": 3.7485733034887705e-07, "loss": 0.0266, "step": 4732 }, { "epoch": 1.87, "learning_rate": 3.7484493490222936e-07, "loss": 0.0289, "step": 4733 }, { "epoch": 1.87, "learning_rate": 3.7483253660587495e-07, "loss": 0.027, "step": 4734 }, { "epoch": 1.87, "learning_rate": 3.748201354600159e-07, "loss": 0.0274, "step": 4735 }, { "epoch": 1.87, "learning_rate": 3.7480773146485436e-07, "loss": 0.0242, "step": 4736 }, { "epoch": 1.87, "learning_rate": 3.747953246205925e-07, "loss": 0.027, "step": 4737 }, { "epoch": 1.87, "learning_rate": 3.747829149274325e-07, "loss": 0.0326, "step": 4738 }, { "epoch": 1.87, "learning_rate": 3.747705023855766e-07, "loss": 0.0285, "step": 4739 }, { "epoch": 1.87, "learning_rate": 3.747580869952271e-07, "loss": 0.0283, "step": 4740 }, { "epoch": 1.87, "learning_rate": 3.747456687565864e-07, "loss": 0.0218, "step": 4741 }, { "epoch": 1.87, "learning_rate": 3.7473324766985694e-07, "loss": 0.028, "step": 4742 }, { "epoch": 1.87, "learning_rate": 3.747208237352411e-07, "loss": 0.0322, "step": 4743 }, { "epoch": 1.87, "learning_rate": 3.747083969529413e-07, "loss": 0.0272, "step": 4744 }, { "epoch": 1.87, "learning_rate": 3.7469596732316024e-07, "loss": 0.0259, "step": 4745 }, { "epoch": 1.87, "learning_rate": 3.746835348461004e-07, "loss": 0.0325, "step": 4746 }, { "epoch": 1.87, "learning_rate": 3.746710995219644e-07, "loss": 0.033, "step": 4747 }, { "epoch": 1.87, "learning_rate": 3.7465866135095493e-07, "loss": 0.0289, "step": 4748 }, { "epoch": 1.87, "learning_rate": 3.746462203332748e-07, "loss": 0.026, "step": 4749 }, { "epoch": 1.87, "learning_rate": 3.7463377646912666e-07, "loss": 0.027, "step": 4750 }, { "epoch": 1.87, "learning_rate": 3.746213297587134e-07, "loss": 0.0276, "step": 4751 }, { "epoch": 1.87, "learning_rate": 3.7460888020223786e-07, "loss": 0.0307, "step": 4752 }, { "epoch": 1.87, "learning_rate": 3.7459642779990286e-07, "loss": 0.0308, "step": 4753 }, { "epoch": 1.87, "learning_rate": 3.745839725519116e-07, "loss": 0.0271, "step": 4754 }, { "epoch": 1.87, "learning_rate": 3.7457151445846684e-07, "loss": 0.0237, "step": 4755 }, { "epoch": 1.87, "learning_rate": 3.745590535197717e-07, "loss": 0.0262, "step": 4756 }, { "epoch": 1.87, "learning_rate": 3.7454658973602927e-07, "loss": 0.028, "step": 4757 }, { "epoch": 1.88, "learning_rate": 3.745341231074428e-07, "loss": 0.0277, "step": 4758 }, { "epoch": 1.88, "learning_rate": 3.745216536342153e-07, "loss": 0.0218, "step": 4759 }, { "epoch": 1.88, "learning_rate": 3.745091813165501e-07, "loss": 0.0241, "step": 4760 }, { "epoch": 1.88, "learning_rate": 3.7449670615465047e-07, "loss": 0.0302, "step": 4761 }, { "epoch": 1.88, "learning_rate": 3.7448422814871976e-07, "loss": 0.0259, "step": 4762 }, { "epoch": 1.88, "learning_rate": 3.744717472989613e-07, "loss": 0.0295, "step": 4763 }, { "epoch": 1.88, "learning_rate": 3.7445926360557854e-07, "loss": 0.0306, "step": 4764 }, { "epoch": 1.88, "learning_rate": 3.744467770687749e-07, "loss": 0.0284, "step": 4765 }, { "epoch": 1.88, "learning_rate": 3.74434287688754e-07, "loss": 0.0318, "step": 4766 }, { "epoch": 1.88, "learning_rate": 3.744217954657192e-07, "loss": 0.0262, "step": 4767 }, { "epoch": 1.88, "learning_rate": 3.744093003998743e-07, "loss": 0.0304, "step": 4768 }, { "epoch": 1.88, "learning_rate": 3.7439680249142285e-07, "loss": 0.029, "step": 4769 }, { "epoch": 1.88, "learning_rate": 3.7438430174056856e-07, "loss": 0.0286, "step": 4770 }, { "epoch": 1.88, "learning_rate": 3.7437179814751523e-07, "loss": 0.0308, "step": 4771 }, { "epoch": 1.88, "learning_rate": 3.743592917124666e-07, "loss": 0.0237, "step": 4772 }, { "epoch": 1.88, "learning_rate": 3.743467824356265e-07, "loss": 0.0285, "step": 4773 }, { "epoch": 1.88, "learning_rate": 3.743342703171988e-07, "loss": 0.0227, "step": 4774 }, { "epoch": 1.88, "learning_rate": 3.7432175535738744e-07, "loss": 0.0254, "step": 4775 }, { "epoch": 1.88, "learning_rate": 3.7430923755639647e-07, "loss": 0.0306, "step": 4776 }, { "epoch": 1.88, "learning_rate": 3.7429671691442983e-07, "loss": 0.0321, "step": 4777 }, { "epoch": 1.88, "learning_rate": 3.7428419343169156e-07, "loss": 0.0273, "step": 4778 }, { "epoch": 1.88, "learning_rate": 3.7427166710838585e-07, "loss": 0.0257, "step": 4779 }, { "epoch": 1.88, "learning_rate": 3.742591379447168e-07, "loss": 0.0303, "step": 4780 }, { "epoch": 1.88, "learning_rate": 3.7424660594088867e-07, "loss": 0.0305, "step": 4781 }, { "epoch": 1.88, "learning_rate": 3.7423407109710574e-07, "loss": 0.0276, "step": 4782 }, { "epoch": 1.89, "learning_rate": 3.742215334135721e-07, "loss": 0.0308, "step": 4783 }, { "epoch": 1.89, "learning_rate": 3.742089928904924e-07, "loss": 0.0315, "step": 4784 }, { "epoch": 1.89, "learning_rate": 3.7419644952807087e-07, "loss": 0.0264, "step": 4785 }, { "epoch": 1.89, "learning_rate": 3.741839033265119e-07, "loss": 0.0289, "step": 4786 }, { "epoch": 1.89, "learning_rate": 3.741713542860201e-07, "loss": 0.0277, "step": 4787 }, { "epoch": 1.89, "learning_rate": 3.7415880240679986e-07, "loss": 0.0267, "step": 4788 }, { "epoch": 1.89, "learning_rate": 3.741462476890559e-07, "loss": 0.0273, "step": 4789 }, { "epoch": 1.89, "learning_rate": 3.741336901329927e-07, "loss": 0.0253, "step": 4790 }, { "epoch": 1.89, "learning_rate": 3.7412112973881505e-07, "loss": 0.026, "step": 4791 }, { "epoch": 1.89, "learning_rate": 3.7410856650672766e-07, "loss": 0.0271, "step": 4792 }, { "epoch": 1.89, "learning_rate": 3.740960004369352e-07, "loss": 0.0278, "step": 4793 }, { "epoch": 1.89, "learning_rate": 3.7408343152964253e-07, "loss": 0.0276, "step": 4794 }, { "epoch": 1.89, "learning_rate": 3.7407085978505444e-07, "loss": 0.0331, "step": 4795 }, { "epoch": 1.89, "learning_rate": 3.7405828520337603e-07, "loss": 0.0259, "step": 4796 }, { "epoch": 1.89, "learning_rate": 3.7404570778481206e-07, "loss": 0.0362, "step": 4797 }, { "epoch": 1.89, "learning_rate": 3.740331275295675e-07, "loss": 0.0266, "step": 4798 }, { "epoch": 1.89, "learning_rate": 3.740205444378475e-07, "loss": 0.0265, "step": 4799 }, { "epoch": 1.89, "learning_rate": 3.7400795850985714e-07, "loss": 0.0261, "step": 4800 }, { "epoch": 1.89, "learning_rate": 3.739953697458015e-07, "loss": 0.0282, "step": 4801 }, { "epoch": 1.89, "learning_rate": 3.7398277814588574e-07, "loss": 0.0287, "step": 4802 }, { "epoch": 1.89, "learning_rate": 3.739701837103152e-07, "loss": 0.0269, "step": 4803 }, { "epoch": 1.89, "learning_rate": 3.7395758643929503e-07, "loss": 0.0284, "step": 4804 }, { "epoch": 1.89, "learning_rate": 3.7394498633303054e-07, "loss": 0.0267, "step": 4805 }, { "epoch": 1.89, "learning_rate": 3.7393238339172714e-07, "loss": 0.0217, "step": 4806 }, { "epoch": 1.89, "learning_rate": 3.7391977761559034e-07, "loss": 0.027, "step": 4807 }, { "epoch": 1.89, "learning_rate": 3.739071690048254e-07, "loss": 0.0283, "step": 4808 }, { "epoch": 1.9, "learning_rate": 3.7389455755963793e-07, "loss": 0.0282, "step": 4809 }, { "epoch": 1.9, "learning_rate": 3.7388194328023346e-07, "loss": 0.0245, "step": 4810 }, { "epoch": 1.9, "learning_rate": 3.738693261668176e-07, "loss": 0.0296, "step": 4811 }, { "epoch": 1.9, "learning_rate": 3.73856706219596e-07, "loss": 0.0284, "step": 4812 }, { "epoch": 1.9, "learning_rate": 3.738440834387743e-07, "loss": 0.0253, "step": 4813 }, { "epoch": 1.9, "learning_rate": 3.7383145782455825e-07, "loss": 0.0303, "step": 4814 }, { "epoch": 1.9, "learning_rate": 3.738188293771536e-07, "loss": 0.0239, "step": 4815 }, { "epoch": 1.9, "learning_rate": 3.7380619809676627e-07, "loss": 0.0224, "step": 4816 }, { "epoch": 1.9, "learning_rate": 3.73793563983602e-07, "loss": 0.0234, "step": 4817 }, { "epoch": 1.9, "learning_rate": 3.737809270378668e-07, "loss": 0.0244, "step": 4818 }, { "epoch": 1.9, "learning_rate": 3.737682872597666e-07, "loss": 0.0286, "step": 4819 }, { "epoch": 1.9, "learning_rate": 3.737556446495074e-07, "loss": 0.0268, "step": 4820 }, { "epoch": 1.9, "learning_rate": 3.737429992072953e-07, "loss": 0.0249, "step": 4821 }, { "epoch": 1.9, "learning_rate": 3.7373035093333634e-07, "loss": 0.0288, "step": 4822 }, { "epoch": 1.9, "learning_rate": 3.737176998278367e-07, "loss": 0.0298, "step": 4823 }, { "epoch": 1.9, "learning_rate": 3.7370504589100264e-07, "loss": 0.0283, "step": 4824 }, { "epoch": 1.9, "learning_rate": 3.736923891230403e-07, "loss": 0.0264, "step": 4825 }, { "epoch": 1.9, "learning_rate": 3.7367972952415597e-07, "loss": 0.0255, "step": 4826 }, { "epoch": 1.9, "learning_rate": 3.73667067094556e-07, "loss": 0.0274, "step": 4827 }, { "epoch": 1.9, "learning_rate": 3.736544018344468e-07, "loss": 0.0332, "step": 4828 }, { "epoch": 1.9, "learning_rate": 3.736417337440348e-07, "loss": 0.0299, "step": 4829 }, { "epoch": 1.9, "learning_rate": 3.7362906282352644e-07, "loss": 0.0272, "step": 4830 }, { "epoch": 1.9, "learning_rate": 3.7361638907312823e-07, "loss": 0.0313, "step": 4831 }, { "epoch": 1.9, "learning_rate": 3.7360371249304674e-07, "loss": 0.0267, "step": 4832 }, { "epoch": 1.9, "learning_rate": 3.735910330834886e-07, "loss": 0.0278, "step": 4833 }, { "epoch": 1.91, "learning_rate": 3.7357835084466045e-07, "loss": 0.0261, "step": 4834 }, { "epoch": 1.91, "learning_rate": 3.7356566577676894e-07, "loss": 0.0314, "step": 4835 }, { "epoch": 1.91, "learning_rate": 3.735529778800209e-07, "loss": 0.0307, "step": 4836 }, { "epoch": 1.91, "learning_rate": 3.735402871546231e-07, "loss": 0.0283, "step": 4837 }, { "epoch": 1.91, "learning_rate": 3.7352759360078235e-07, "loss": 0.0295, "step": 4838 }, { "epoch": 1.91, "learning_rate": 3.735148972187056e-07, "loss": 0.0265, "step": 4839 }, { "epoch": 1.91, "learning_rate": 3.735021980085997e-07, "loss": 0.0262, "step": 4840 }, { "epoch": 1.91, "learning_rate": 3.7348949597067166e-07, "loss": 0.0321, "step": 4841 }, { "epoch": 1.91, "learning_rate": 3.7347679110512853e-07, "loss": 0.0268, "step": 4842 }, { "epoch": 1.91, "learning_rate": 3.734640834121774e-07, "loss": 0.0244, "step": 4843 }, { "epoch": 1.91, "learning_rate": 3.7345137289202527e-07, "loss": 0.0273, "step": 4844 }, { "epoch": 1.91, "learning_rate": 3.734386595448794e-07, "loss": 0.0227, "step": 4845 }, { "epoch": 1.91, "learning_rate": 3.73425943370947e-07, "loss": 0.027, "step": 4846 }, { "epoch": 1.91, "learning_rate": 3.7341322437043525e-07, "loss": 0.0245, "step": 4847 }, { "epoch": 1.91, "learning_rate": 3.734005025435515e-07, "loss": 0.0264, "step": 4848 }, { "epoch": 1.91, "learning_rate": 3.733877778905032e-07, "loss": 0.0271, "step": 4849 }, { "epoch": 1.91, "learning_rate": 3.733750504114975e-07, "loss": 0.0286, "step": 4850 }, { "epoch": 1.91, "learning_rate": 3.733623201067421e-07, "loss": 0.0323, "step": 4851 }, { "epoch": 1.91, "learning_rate": 3.733495869764443e-07, "loss": 0.0286, "step": 4852 }, { "epoch": 1.91, "learning_rate": 3.733368510208117e-07, "loss": 0.0269, "step": 4853 }, { "epoch": 1.91, "learning_rate": 3.7332411224005183e-07, "loss": 0.0282, "step": 4854 }, { "epoch": 1.91, "learning_rate": 3.733113706343724e-07, "loss": 0.0288, "step": 4855 }, { "epoch": 1.91, "learning_rate": 3.73298626203981e-07, "loss": 0.0267, "step": 4856 }, { "epoch": 1.91, "learning_rate": 3.732858789490854e-07, "loss": 0.0262, "step": 4857 }, { "epoch": 1.91, "learning_rate": 3.732731288698934e-07, "loss": 0.0286, "step": 4858 }, { "epoch": 1.91, "learning_rate": 3.7326037596661264e-07, "loss": 0.0293, "step": 4859 }, { "epoch": 1.92, "learning_rate": 3.732476202394511e-07, "loss": 0.024, "step": 4860 }, { "epoch": 1.92, "learning_rate": 3.732348616886167e-07, "loss": 0.0265, "step": 4861 }, { "epoch": 1.92, "learning_rate": 3.7322210031431735e-07, "loss": 0.0306, "step": 4862 }, { "epoch": 1.92, "learning_rate": 3.73209336116761e-07, "loss": 0.0275, "step": 4863 }, { "epoch": 1.92, "learning_rate": 3.731965690961557e-07, "loss": 0.0279, "step": 4864 }, { "epoch": 1.92, "learning_rate": 3.7318379925270955e-07, "loss": 0.0298, "step": 4865 }, { "epoch": 1.92, "learning_rate": 3.7317102658663074e-07, "loss": 0.026, "step": 4866 }, { "epoch": 1.92, "learning_rate": 3.7315825109812733e-07, "loss": 0.0286, "step": 4867 }, { "epoch": 1.92, "learning_rate": 3.731454727874076e-07, "loss": 0.0265, "step": 4868 }, { "epoch": 1.92, "learning_rate": 3.7313269165467984e-07, "loss": 0.0281, "step": 4869 }, { "epoch": 1.92, "learning_rate": 3.7311990770015233e-07, "loss": 0.0257, "step": 4870 }, { "epoch": 1.92, "learning_rate": 3.7310712092403345e-07, "loss": 0.0267, "step": 4871 }, { "epoch": 1.92, "learning_rate": 3.730943313265316e-07, "loss": 0.0301, "step": 4872 }, { "epoch": 1.92, "learning_rate": 3.730815389078552e-07, "loss": 0.0296, "step": 4873 }, { "epoch": 1.92, "learning_rate": 3.7306874366821274e-07, "loss": 0.0254, "step": 4874 }, { "epoch": 1.92, "learning_rate": 3.7305594560781283e-07, "loss": 0.0282, "step": 4875 }, { "epoch": 1.92, "learning_rate": 3.73043144726864e-07, "loss": 0.03, "step": 4876 }, { "epoch": 1.92, "learning_rate": 3.7303034102557496e-07, "loss": 0.0262, "step": 4877 }, { "epoch": 1.92, "learning_rate": 3.7301753450415426e-07, "loss": 0.027, "step": 4878 }, { "epoch": 1.92, "learning_rate": 3.7300472516281075e-07, "loss": 0.0246, "step": 4879 }, { "epoch": 1.92, "learning_rate": 3.7299191300175316e-07, "loss": 0.0274, "step": 4880 }, { "epoch": 1.92, "learning_rate": 3.7297909802119023e-07, "loss": 0.0266, "step": 4881 }, { "epoch": 1.92, "learning_rate": 3.72966280221331e-07, "loss": 0.0326, "step": 4882 }, { "epoch": 1.92, "learning_rate": 3.729534596023842e-07, "loss": 0.0253, "step": 4883 }, { "epoch": 1.92, "learning_rate": 3.7294063616455895e-07, "loss": 0.0303, "step": 4884 }, { "epoch": 1.93, "learning_rate": 3.729278099080641e-07, "loss": 0.0277, "step": 4885 }, { "epoch": 1.93, "learning_rate": 3.729149808331088e-07, "loss": 0.0336, "step": 4886 }, { "epoch": 1.93, "learning_rate": 3.729021489399021e-07, "loss": 0.0289, "step": 4887 }, { "epoch": 1.93, "learning_rate": 3.728893142286532e-07, "loss": 0.0291, "step": 4888 }, { "epoch": 1.93, "learning_rate": 3.728764766995712e-07, "loss": 0.025, "step": 4889 }, { "epoch": 1.93, "learning_rate": 3.728636363528654e-07, "loss": 0.0249, "step": 4890 }, { "epoch": 1.93, "learning_rate": 3.7285079318874496e-07, "loss": 0.0252, "step": 4891 }, { "epoch": 1.93, "learning_rate": 3.7283794720741943e-07, "loss": 0.0279, "step": 4892 }, { "epoch": 1.93, "learning_rate": 3.7282509840909795e-07, "loss": 0.0306, "step": 4893 }, { "epoch": 1.93, "learning_rate": 3.7281224679399006e-07, "loss": 0.0271, "step": 4894 }, { "epoch": 1.93, "learning_rate": 3.7279939236230513e-07, "loss": 0.0279, "step": 4895 }, { "epoch": 1.93, "learning_rate": 3.727865351142528e-07, "loss": 0.0269, "step": 4896 }, { "epoch": 1.93, "learning_rate": 3.7277367505004256e-07, "loss": 0.028, "step": 4897 }, { "epoch": 1.93, "learning_rate": 3.7276081216988394e-07, "loss": 0.0243, "step": 4898 }, { "epoch": 1.93, "learning_rate": 3.7274794647398673e-07, "loss": 0.0285, "step": 4899 }, { "epoch": 1.93, "learning_rate": 3.727350779625605e-07, "loss": 0.0313, "step": 4900 }, { "epoch": 1.93, "learning_rate": 3.7272220663581504e-07, "loss": 0.0234, "step": 4901 }, { "epoch": 1.93, "learning_rate": 3.7270933249396004e-07, "loss": 0.0234, "step": 4902 }, { "epoch": 1.93, "learning_rate": 3.726964555372055e-07, "loss": 0.0289, "step": 4903 }, { "epoch": 1.93, "learning_rate": 3.7268357576576117e-07, "loss": 0.0326, "step": 4904 }, { "epoch": 1.93, "learning_rate": 3.72670693179837e-07, "loss": 0.0241, "step": 4905 }, { "epoch": 1.93, "learning_rate": 3.7265780777964294e-07, "loss": 0.0247, "step": 4906 }, { "epoch": 1.93, "learning_rate": 3.726449195653891e-07, "loss": 0.0267, "step": 4907 }, { "epoch": 1.93, "learning_rate": 3.726320285372854e-07, "loss": 0.0329, "step": 4908 }, { "epoch": 1.93, "learning_rate": 3.72619134695542e-07, "loss": 0.0264, "step": 4909 }, { "epoch": 1.94, "learning_rate": 3.7260623804036907e-07, "loss": 0.0289, "step": 4910 }, { "epoch": 1.94, "learning_rate": 3.725933385719768e-07, "loss": 0.0309, "step": 4911 }, { "epoch": 1.94, "learning_rate": 3.725804362905754e-07, "loss": 0.0259, "step": 4912 }, { "epoch": 1.94, "learning_rate": 3.725675311963752e-07, "loss": 0.0338, "step": 4913 }, { "epoch": 1.94, "learning_rate": 3.725546232895865e-07, "loss": 0.0279, "step": 4914 }, { "epoch": 1.94, "learning_rate": 3.725417125704197e-07, "loss": 0.0278, "step": 4915 }, { "epoch": 1.94, "learning_rate": 3.725287990390853e-07, "loss": 0.029, "step": 4916 }, { "epoch": 1.94, "learning_rate": 3.7251588269579357e-07, "loss": 0.0238, "step": 4917 }, { "epoch": 1.94, "learning_rate": 3.7250296354075513e-07, "loss": 0.0271, "step": 4918 }, { "epoch": 1.94, "learning_rate": 3.7249004157418067e-07, "loss": 0.0266, "step": 4919 }, { "epoch": 1.94, "learning_rate": 3.724771167962806e-07, "loss": 0.0293, "step": 4920 }, { "epoch": 1.94, "learning_rate": 3.724641892072657e-07, "loss": 0.0298, "step": 4921 }, { "epoch": 1.94, "learning_rate": 3.7245125880734664e-07, "loss": 0.0284, "step": 4922 }, { "epoch": 1.94, "learning_rate": 3.7243832559673416e-07, "loss": 0.0278, "step": 4923 }, { "epoch": 1.94, "learning_rate": 3.72425389575639e-07, "loss": 0.0238, "step": 4924 }, { "epoch": 1.94, "learning_rate": 3.7241245074427205e-07, "loss": 0.0325, "step": 4925 }, { "epoch": 1.94, "learning_rate": 3.723995091028443e-07, "loss": 0.0272, "step": 4926 }, { "epoch": 1.94, "learning_rate": 3.7238656465156643e-07, "loss": 0.0277, "step": 4927 }, { "epoch": 1.94, "learning_rate": 3.723736173906496e-07, "loss": 0.0316, "step": 4928 }, { "epoch": 1.94, "learning_rate": 3.723606673203048e-07, "loss": 0.0248, "step": 4929 }, { "epoch": 1.94, "learning_rate": 3.7234771444074305e-07, "loss": 0.0269, "step": 4930 }, { "epoch": 1.94, "learning_rate": 3.7233475875217553e-07, "loss": 0.0321, "step": 4931 }, { "epoch": 1.94, "learning_rate": 3.723218002548134e-07, "loss": 0.0235, "step": 4932 }, { "epoch": 1.94, "learning_rate": 3.7230883894886774e-07, "loss": 0.0299, "step": 4933 }, { "epoch": 1.94, "learning_rate": 3.7229587483454994e-07, "loss": 0.0268, "step": 4934 }, { "epoch": 1.94, "learning_rate": 3.722829079120712e-07, "loss": 0.0299, "step": 4935 }, { "epoch": 1.95, "learning_rate": 3.7226993818164296e-07, "loss": 0.0292, "step": 4936 }, { "epoch": 1.95, "learning_rate": 3.7225696564347645e-07, "loss": 0.0278, "step": 4937 }, { "epoch": 1.95, "learning_rate": 3.722439902977833e-07, "loss": 0.0253, "step": 4938 }, { "epoch": 1.95, "learning_rate": 3.722310121447748e-07, "loss": 0.0231, "step": 4939 }, { "epoch": 1.95, "learning_rate": 3.7221803118466264e-07, "loss": 0.0316, "step": 4940 }, { "epoch": 1.95, "learning_rate": 3.7220504741765824e-07, "loss": 0.0296, "step": 4941 }, { "epoch": 1.95, "learning_rate": 3.7219206084397337e-07, "loss": 0.0263, "step": 4942 }, { "epoch": 1.95, "learning_rate": 3.721790714638195e-07, "loss": 0.0288, "step": 4943 }, { "epoch": 1.95, "learning_rate": 3.721660792774086e-07, "loss": 0.0285, "step": 4944 }, { "epoch": 1.95, "learning_rate": 3.721530842849522e-07, "loss": 0.0233, "step": 4945 }, { "epoch": 1.95, "learning_rate": 3.721400864866621e-07, "loss": 0.0232, "step": 4946 }, { "epoch": 1.95, "learning_rate": 3.7212708588275026e-07, "loss": 0.0271, "step": 4947 }, { "epoch": 1.95, "learning_rate": 3.721140824734285e-07, "loss": 0.0292, "step": 4948 }, { "epoch": 1.95, "learning_rate": 3.721010762589089e-07, "loss": 0.0269, "step": 4949 }, { "epoch": 1.95, "learning_rate": 3.720880672394032e-07, "loss": 0.0257, "step": 4950 }, { "epoch": 1.95, "learning_rate": 3.720750554151236e-07, "loss": 0.0283, "step": 4951 }, { "epoch": 1.95, "learning_rate": 3.720620407862821e-07, "loss": 0.0266, "step": 4952 }, { "epoch": 1.95, "learning_rate": 3.720490233530908e-07, "loss": 0.0264, "step": 4953 }, { "epoch": 1.95, "learning_rate": 3.72036003115762e-07, "loss": 0.0294, "step": 4954 }, { "epoch": 1.95, "learning_rate": 3.720229800745077e-07, "loss": 0.0276, "step": 4955 }, { "epoch": 1.95, "learning_rate": 3.720099542295403e-07, "loss": 0.0267, "step": 4956 }, { "epoch": 1.95, "learning_rate": 3.7199692558107213e-07, "loss": 0.0252, "step": 4957 }, { "epoch": 1.95, "learning_rate": 3.719838941293155e-07, "loss": 0.0273, "step": 4958 }, { "epoch": 1.95, "learning_rate": 3.7197085987448265e-07, "loss": 0.0265, "step": 4959 }, { "epoch": 1.95, "learning_rate": 3.7195782281678624e-07, "loss": 0.0239, "step": 4960 }, { "epoch": 1.96, "learning_rate": 3.7194478295643865e-07, "loss": 0.0299, "step": 4961 }, { "epoch": 1.96, "learning_rate": 3.7193174029365243e-07, "loss": 0.0279, "step": 4962 }, { "epoch": 1.96, "learning_rate": 3.719186948286402e-07, "loss": 0.0261, "step": 4963 }, { "epoch": 1.96, "learning_rate": 3.7190564656161443e-07, "loss": 0.0262, "step": 4964 }, { "epoch": 1.96, "learning_rate": 3.718925954927879e-07, "loss": 0.0272, "step": 4965 }, { "epoch": 1.96, "learning_rate": 3.7187954162237333e-07, "loss": 0.0306, "step": 4966 }, { "epoch": 1.96, "learning_rate": 3.718664849505835e-07, "loss": 0.0274, "step": 4967 }, { "epoch": 1.96, "learning_rate": 3.7185342547763115e-07, "loss": 0.0227, "step": 4968 }, { "epoch": 1.96, "learning_rate": 3.718403632037291e-07, "loss": 0.0253, "step": 4969 }, { "epoch": 1.96, "learning_rate": 3.7182729812909037e-07, "loss": 0.0302, "step": 4970 }, { "epoch": 1.96, "learning_rate": 3.718142302539278e-07, "loss": 0.0256, "step": 4971 }, { "epoch": 1.96, "learning_rate": 3.7180115957845437e-07, "loss": 0.025, "step": 4972 }, { "epoch": 1.96, "learning_rate": 3.7178808610288323e-07, "loss": 0.0248, "step": 4973 }, { "epoch": 1.96, "learning_rate": 3.717750098274273e-07, "loss": 0.0275, "step": 4974 }, { "epoch": 1.96, "learning_rate": 3.717619307522998e-07, "loss": 0.0267, "step": 4975 }, { "epoch": 1.96, "learning_rate": 3.717488488777139e-07, "loss": 0.0265, "step": 4976 }, { "epoch": 1.96, "learning_rate": 3.7173576420388274e-07, "loss": 0.0267, "step": 4977 }, { "epoch": 1.96, "learning_rate": 3.717226767310197e-07, "loss": 0.0223, "step": 4978 }, { "epoch": 1.96, "learning_rate": 3.7170958645933794e-07, "loss": 0.0303, "step": 4979 }, { "epoch": 1.96, "learning_rate": 3.71696493389051e-07, "loss": 0.0277, "step": 4980 }, { "epoch": 1.96, "learning_rate": 3.7168339752037204e-07, "loss": 0.0241, "step": 4981 }, { "epoch": 1.96, "learning_rate": 3.716702988535147e-07, "loss": 0.0276, "step": 4982 }, { "epoch": 1.96, "learning_rate": 3.716571973886924e-07, "loss": 0.0228, "step": 4983 }, { "epoch": 1.96, "learning_rate": 3.7164409312611864e-07, "loss": 0.0233, "step": 4984 }, { "epoch": 1.96, "learning_rate": 3.71630986066007e-07, "loss": 0.0283, "step": 4985 }, { "epoch": 1.97, "learning_rate": 3.716178762085712e-07, "loss": 0.0323, "step": 4986 }, { "epoch": 1.97, "learning_rate": 3.7160476355402486e-07, "loss": 0.0328, "step": 4987 }, { "epoch": 1.97, "learning_rate": 3.715916481025816e-07, "loss": 0.0232, "step": 4988 }, { "epoch": 1.97, "learning_rate": 3.7157852985445535e-07, "loss": 0.026, "step": 4989 }, { "epoch": 1.97, "learning_rate": 3.715654088098598e-07, "loss": 0.0271, "step": 4990 }, { "epoch": 1.97, "learning_rate": 3.715522849690088e-07, "loss": 0.0302, "step": 4991 }, { "epoch": 1.97, "learning_rate": 3.715391583321163e-07, "loss": 0.0298, "step": 4992 }, { "epoch": 1.97, "learning_rate": 3.715260288993963e-07, "loss": 0.0283, "step": 4993 }, { "epoch": 1.97, "learning_rate": 3.7151289667106264e-07, "loss": 0.0281, "step": 4994 }, { "epoch": 1.97, "learning_rate": 3.7149976164732946e-07, "loss": 0.0271, "step": 4995 }, { "epoch": 1.97, "learning_rate": 3.7148662382841085e-07, "loss": 0.0273, "step": 4996 }, { "epoch": 1.97, "learning_rate": 3.7147348321452085e-07, "loss": 0.0259, "step": 4997 }, { "epoch": 1.97, "learning_rate": 3.714603398058737e-07, "loss": 0.0252, "step": 4998 }, { "epoch": 1.97, "learning_rate": 3.7144719360268357e-07, "loss": 0.0265, "step": 4999 }, { "epoch": 1.97, "learning_rate": 3.714340446051648e-07, "loss": 0.0288, "step": 5000 }, { "epoch": 1.97, "learning_rate": 3.714208928135316e-07, "loss": 0.0269, "step": 5001 }, { "epoch": 1.97, "learning_rate": 3.7140773822799847e-07, "loss": 0.0286, "step": 5002 }, { "epoch": 1.97, "learning_rate": 3.713945808487796e-07, "loss": 0.0278, "step": 5003 }, { "epoch": 1.97, "learning_rate": 3.713814206760896e-07, "loss": 0.0301, "step": 5004 }, { "epoch": 1.97, "learning_rate": 3.71368257710143e-07, "loss": 0.0229, "step": 5005 }, { "epoch": 1.97, "learning_rate": 3.7135509195115415e-07, "loss": 0.0298, "step": 5006 }, { "epoch": 1.97, "learning_rate": 3.7134192339933774e-07, "loss": 0.0253, "step": 5007 }, { "epoch": 1.97, "learning_rate": 3.713287520549084e-07, "loss": 0.0226, "step": 5008 }, { "epoch": 1.97, "learning_rate": 3.713155779180808e-07, "loss": 0.0301, "step": 5009 }, { "epoch": 1.97, "learning_rate": 3.7130240098906965e-07, "loss": 0.03, "step": 5010 }, { "epoch": 1.97, "learning_rate": 3.712892212680897e-07, "loss": 0.0336, "step": 5011 }, { "epoch": 1.98, "learning_rate": 3.7127603875535577e-07, "loss": 0.0302, "step": 5012 }, { "epoch": 1.98, "learning_rate": 3.7126285345108277e-07, "loss": 0.0304, "step": 5013 }, { "epoch": 1.98, "learning_rate": 3.7124966535548554e-07, "loss": 0.0275, "step": 5014 }, { "epoch": 1.98, "learning_rate": 3.7123647446877895e-07, "loss": 0.0214, "step": 5015 }, { "epoch": 1.98, "learning_rate": 3.712232807911781e-07, "loss": 0.0238, "step": 5016 }, { "epoch": 1.98, "learning_rate": 3.71210084322898e-07, "loss": 0.0278, "step": 5017 }, { "epoch": 1.98, "learning_rate": 3.711968850641538e-07, "loss": 0.0278, "step": 5018 }, { "epoch": 1.98, "learning_rate": 3.711836830151605e-07, "loss": 0.0279, "step": 5019 }, { "epoch": 1.98, "learning_rate": 3.711704781761334e-07, "loss": 0.0297, "step": 5020 }, { "epoch": 1.98, "learning_rate": 3.7115727054728757e-07, "loss": 0.0248, "step": 5021 }, { "epoch": 1.98, "learning_rate": 3.7114406012883846e-07, "loss": 0.0248, "step": 5022 }, { "epoch": 1.98, "learning_rate": 3.7113084692100117e-07, "loss": 0.0328, "step": 5023 }, { "epoch": 1.98, "learning_rate": 3.7111763092399127e-07, "loss": 0.0254, "step": 5024 }, { "epoch": 1.98, "learning_rate": 3.71104412138024e-07, "loss": 0.0241, "step": 5025 }, { "epoch": 1.98, "learning_rate": 3.7109119056331487e-07, "loss": 0.0258, "step": 5026 }, { "epoch": 1.98, "learning_rate": 3.7107796620007934e-07, "loss": 0.0248, "step": 5027 }, { "epoch": 1.98, "learning_rate": 3.71064739048533e-07, "loss": 0.0276, "step": 5028 }, { "epoch": 1.98, "learning_rate": 3.710515091088914e-07, "loss": 0.0299, "step": 5029 }, { "epoch": 1.98, "learning_rate": 3.710382763813701e-07, "loss": 0.0259, "step": 5030 }, { "epoch": 1.98, "learning_rate": 3.7102504086618496e-07, "loss": 0.0287, "step": 5031 }, { "epoch": 1.98, "learning_rate": 3.710118025635515e-07, "loss": 0.0272, "step": 5032 }, { "epoch": 1.98, "learning_rate": 3.709985614736856e-07, "loss": 0.0268, "step": 5033 }, { "epoch": 1.98, "learning_rate": 3.70985317596803e-07, "loss": 0.0271, "step": 5034 }, { "epoch": 1.98, "learning_rate": 3.7097207093311964e-07, "loss": 0.0251, "step": 5035 }, { "epoch": 1.98, "learning_rate": 3.7095882148285134e-07, "loss": 0.026, "step": 5036 }, { "epoch": 1.99, "learning_rate": 3.709455692462141e-07, "loss": 0.0254, "step": 5037 }, { "epoch": 1.99, "learning_rate": 3.7093231422342387e-07, "loss": 0.0321, "step": 5038 }, { "epoch": 1.99, "learning_rate": 3.709190564146967e-07, "loss": 0.0266, "step": 5039 }, { "epoch": 1.99, "learning_rate": 3.709057958202487e-07, "loss": 0.0293, "step": 5040 }, { "epoch": 1.99, "learning_rate": 3.708925324402959e-07, "loss": 0.0271, "step": 5041 }, { "epoch": 1.99, "learning_rate": 3.7087926627505466e-07, "loss": 0.0259, "step": 5042 }, { "epoch": 1.99, "learning_rate": 3.70865997324741e-07, "loss": 0.0299, "step": 5043 }, { "epoch": 1.99, "learning_rate": 3.7085272558957135e-07, "loss": 0.0293, "step": 5044 }, { "epoch": 1.99, "learning_rate": 3.7083945106976187e-07, "loss": 0.0258, "step": 5045 }, { "epoch": 1.99, "learning_rate": 3.7082617376552904e-07, "loss": 0.026, "step": 5046 }, { "epoch": 1.99, "learning_rate": 3.7081289367708916e-07, "loss": 0.0273, "step": 5047 }, { "epoch": 1.99, "learning_rate": 3.707996108046587e-07, "loss": 0.0249, "step": 5048 }, { "epoch": 1.99, "learning_rate": 3.707863251484542e-07, "loss": 0.0234, "step": 5049 }, { "epoch": 1.99, "learning_rate": 3.707730367086922e-07, "loss": 0.029, "step": 5050 }, { "epoch": 1.99, "learning_rate": 3.707597454855892e-07, "loss": 0.0292, "step": 5051 }, { "epoch": 1.99, "learning_rate": 3.707464514793619e-07, "loss": 0.0277, "step": 5052 }, { "epoch": 1.99, "learning_rate": 3.707331546902269e-07, "loss": 0.0277, "step": 5053 }, { "epoch": 1.99, "learning_rate": 3.7071985511840105e-07, "loss": 0.0297, "step": 5054 }, { "epoch": 1.99, "learning_rate": 3.7070655276410096e-07, "loss": 0.0238, "step": 5055 }, { "epoch": 1.99, "learning_rate": 3.706932476275435e-07, "loss": 0.0296, "step": 5056 }, { "epoch": 1.99, "learning_rate": 3.7067993970894553e-07, "loss": 0.0265, "step": 5057 }, { "epoch": 1.99, "learning_rate": 3.7066662900852393e-07, "loss": 0.0256, "step": 5058 }, { "epoch": 1.99, "learning_rate": 3.7065331552649573e-07, "loss": 0.0262, "step": 5059 }, { "epoch": 1.99, "learning_rate": 3.706399992630778e-07, "loss": 0.0253, "step": 5060 }, { "epoch": 1.99, "learning_rate": 3.7062668021848723e-07, "loss": 0.0296, "step": 5061 }, { "epoch": 1.99, "learning_rate": 3.706133583929411e-07, "loss": 0.0287, "step": 5062 }, { "epoch": 2.0, "learning_rate": 3.706000337866565e-07, "loss": 0.0257, "step": 5063 }, { "epoch": 2.0, "learning_rate": 3.7058670639985074e-07, "loss": 0.0296, "step": 5064 }, { "epoch": 2.0, "learning_rate": 3.7057337623274076e-07, "loss": 0.0224, "step": 5065 }, { "epoch": 2.0, "learning_rate": 3.705600432855441e-07, "loss": 0.0289, "step": 5066 }, { "epoch": 2.0, "learning_rate": 3.7054670755847795e-07, "loss": 0.0223, "step": 5067 }, { "epoch": 2.0, "learning_rate": 3.705333690517596e-07, "loss": 0.0274, "step": 5068 }, { "epoch": 2.0, "learning_rate": 3.7052002776560667e-07, "loss": 0.0266, "step": 5069 }, { "epoch": 2.0, "learning_rate": 3.705066837002363e-07, "loss": 0.0304, "step": 5070 }, { "epoch": 2.0, "learning_rate": 3.704933368558662e-07, "loss": 0.0273, "step": 5071 }, { "epoch": 2.0, "learning_rate": 3.704799872327138e-07, "loss": 0.0256, "step": 5072 }, { "epoch": 2.0, "learning_rate": 3.704666348309967e-07, "loss": 0.0251, "step": 5073 }, { "epoch": 2.0, "learning_rate": 3.7045327965093256e-07, "loss": 0.0278, "step": 5074 }, { "epoch": 2.0, "learning_rate": 3.704399216927389e-07, "loss": 0.0283, "step": 5075 }, { "epoch": 2.0, "learning_rate": 3.704265609566337e-07, "loss": 0.0265, "step": 5076 }, { "epoch": 2.0, "learning_rate": 3.7041319744283454e-07, "loss": 0.0278, "step": 5077 }, { "epoch": 2.0, "learning_rate": 3.7039983115155924e-07, "loss": 0.0214, "step": 5078 }, { "epoch": 2.0, "learning_rate": 3.7038646208302564e-07, "loss": 0.0231, "step": 5079 }, { "epoch": 2.0, "learning_rate": 3.7037309023745175e-07, "loss": 0.0243, "step": 5080 }, { "epoch": 2.0, "learning_rate": 3.703597156150554e-07, "loss": 0.0272, "step": 5081 }, { "epoch": 2.0, "learning_rate": 3.703463382160546e-07, "loss": 0.0287, "step": 5082 }, { "epoch": 2.0, "learning_rate": 3.703329580406673e-07, "loss": 0.0205, "step": 5083 }, { "epoch": 2.0, "learning_rate": 3.703195750891118e-07, "loss": 0.0264, "step": 5084 }, { "epoch": 2.0, "learning_rate": 3.70306189361606e-07, "loss": 0.0222, "step": 5085 }, { "epoch": 2.0, "learning_rate": 3.702928008583681e-07, "loss": 0.0242, "step": 5086 }, { "epoch": 2.0, "learning_rate": 3.7027940957961647e-07, "loss": 0.0253, "step": 5087 }, { "epoch": 2.01, "learning_rate": 3.702660155255692e-07, "loss": 0.0207, "step": 5088 }, { "epoch": 2.01, "learning_rate": 3.7025261869644467e-07, "loss": 0.0256, "step": 5089 }, { "epoch": 2.01, "learning_rate": 3.7023921909246115e-07, "loss": 0.0243, "step": 5090 }, { "epoch": 2.01, "learning_rate": 3.702258167138372e-07, "loss": 0.0265, "step": 5091 }, { "epoch": 2.01, "learning_rate": 3.702124115607911e-07, "loss": 0.0246, "step": 5092 }, { "epoch": 2.01, "learning_rate": 3.7019900363354137e-07, "loss": 0.0293, "step": 5093 }, { "epoch": 2.01, "learning_rate": 3.7018559293230657e-07, "loss": 0.0261, "step": 5094 }, { "epoch": 2.01, "learning_rate": 3.7017217945730525e-07, "loss": 0.0235, "step": 5095 }, { "epoch": 2.01, "learning_rate": 3.7015876320875597e-07, "loss": 0.0285, "step": 5096 }, { "epoch": 2.01, "learning_rate": 3.701453441868776e-07, "loss": 0.0197, "step": 5097 }, { "epoch": 2.01, "learning_rate": 3.7013192239188856e-07, "loss": 0.0264, "step": 5098 }, { "epoch": 2.01, "learning_rate": 3.701184978240079e-07, "loss": 0.0243, "step": 5099 }, { "epoch": 2.01, "learning_rate": 3.701050704834542e-07, "loss": 0.0289, "step": 5100 }, { "epoch": 2.01, "learning_rate": 3.700916403704464e-07, "loss": 0.0358, "step": 5101 }, { "epoch": 2.01, "learning_rate": 3.7007820748520334e-07, "loss": 0.0235, "step": 5102 }, { "epoch": 2.01, "learning_rate": 3.70064771827944e-07, "loss": 0.0244, "step": 5103 }, { "epoch": 2.01, "learning_rate": 3.700513333988874e-07, "loss": 0.0262, "step": 5104 }, { "epoch": 2.01, "learning_rate": 3.7003789219825243e-07, "loss": 0.029, "step": 5105 }, { "epoch": 2.01, "learning_rate": 3.700244482262583e-07, "loss": 0.0242, "step": 5106 }, { "epoch": 2.01, "learning_rate": 3.7001100148312403e-07, "loss": 0.0226, "step": 5107 }, { "epoch": 2.01, "learning_rate": 3.6999755196906886e-07, "loss": 0.0235, "step": 5108 }, { "epoch": 2.01, "learning_rate": 3.6998409968431196e-07, "loss": 0.0246, "step": 5109 }, { "epoch": 2.01, "learning_rate": 3.699706446290726e-07, "loss": 0.0253, "step": 5110 }, { "epoch": 2.01, "learning_rate": 3.6995718680356996e-07, "loss": 0.0266, "step": 5111 }, { "epoch": 2.01, "learning_rate": 3.699437262080236e-07, "loss": 0.0234, "step": 5112 }, { "epoch": 2.02, "learning_rate": 3.6993026284265276e-07, "loss": 0.024, "step": 5113 }, { "epoch": 2.02, "learning_rate": 3.699167967076769e-07, "loss": 0.0231, "step": 5114 }, { "epoch": 2.02, "learning_rate": 3.6990332780331545e-07, "loss": 0.0272, "step": 5115 }, { "epoch": 2.02, "learning_rate": 3.69889856129788e-07, "loss": 0.0272, "step": 5116 }, { "epoch": 2.02, "learning_rate": 3.698763816873141e-07, "loss": 0.0249, "step": 5117 }, { "epoch": 2.02, "learning_rate": 3.6986290447611336e-07, "loss": 0.0278, "step": 5118 }, { "epoch": 2.02, "learning_rate": 3.6984942449640543e-07, "loss": 0.0288, "step": 5119 }, { "epoch": 2.02, "learning_rate": 3.6983594174841e-07, "loss": 0.0269, "step": 5120 }, { "epoch": 2.02, "learning_rate": 3.6982245623234694e-07, "loss": 0.0263, "step": 5121 }, { "epoch": 2.02, "learning_rate": 3.698089679484358e-07, "loss": 0.0291, "step": 5122 }, { "epoch": 2.02, "learning_rate": 3.6979547689689666e-07, "loss": 0.0271, "step": 5123 }, { "epoch": 2.02, "learning_rate": 3.697819830779493e-07, "loss": 0.0226, "step": 5124 }, { "epoch": 2.02, "learning_rate": 3.6976848649181366e-07, "loss": 0.0238, "step": 5125 }, { "epoch": 2.02, "learning_rate": 3.697549871387097e-07, "loss": 0.0255, "step": 5126 }, { "epoch": 2.02, "learning_rate": 3.697414850188574e-07, "loss": 0.029, "step": 5127 }, { "epoch": 2.02, "learning_rate": 3.697279801324769e-07, "loss": 0.024, "step": 5128 }, { "epoch": 2.02, "learning_rate": 3.6971447247978833e-07, "loss": 0.0274, "step": 5129 }, { "epoch": 2.02, "learning_rate": 3.697009620610117e-07, "loss": 0.0237, "step": 5130 }, { "epoch": 2.02, "learning_rate": 3.696874488763674e-07, "loss": 0.0265, "step": 5131 }, { "epoch": 2.02, "learning_rate": 3.6967393292607554e-07, "loss": 0.0256, "step": 5132 }, { "epoch": 2.02, "learning_rate": 3.696604142103564e-07, "loss": 0.0271, "step": 5133 }, { "epoch": 2.02, "learning_rate": 3.6964689272943045e-07, "loss": 0.0261, "step": 5134 }, { "epoch": 2.02, "learning_rate": 3.6963336848351796e-07, "loss": 0.0273, "step": 5135 }, { "epoch": 2.02, "learning_rate": 3.6961984147283936e-07, "loss": 0.0244, "step": 5136 }, { "epoch": 2.02, "learning_rate": 3.696063116976151e-07, "loss": 0.0201, "step": 5137 }, { "epoch": 2.02, "learning_rate": 3.695927791580658e-07, "loss": 0.0228, "step": 5138 }, { "epoch": 2.03, "learning_rate": 3.695792438544119e-07, "loss": 0.0276, "step": 5139 }, { "epoch": 2.03, "learning_rate": 3.6956570578687405e-07, "loss": 0.0231, "step": 5140 }, { "epoch": 2.03, "learning_rate": 3.6955216495567296e-07, "loss": 0.0252, "step": 5141 }, { "epoch": 2.03, "learning_rate": 3.695386213610292e-07, "loss": 0.0264, "step": 5142 }, { "epoch": 2.03, "learning_rate": 3.6952507500316364e-07, "loss": 0.0251, "step": 5143 }, { "epoch": 2.03, "learning_rate": 3.69511525882297e-07, "loss": 0.0218, "step": 5144 }, { "epoch": 2.03, "learning_rate": 3.6949797399865007e-07, "loss": 0.0244, "step": 5145 }, { "epoch": 2.03, "learning_rate": 3.694844193524438e-07, "loss": 0.0254, "step": 5146 }, { "epoch": 2.03, "learning_rate": 3.694708619438991e-07, "loss": 0.0263, "step": 5147 }, { "epoch": 2.03, "learning_rate": 3.6945730177323687e-07, "loss": 0.0252, "step": 5148 }, { "epoch": 2.03, "learning_rate": 3.694437388406782e-07, "loss": 0.0194, "step": 5149 }, { "epoch": 2.03, "learning_rate": 3.694301731464441e-07, "loss": 0.0233, "step": 5150 }, { "epoch": 2.03, "learning_rate": 3.694166046907557e-07, "loss": 0.0258, "step": 5151 }, { "epoch": 2.03, "learning_rate": 3.6940303347383415e-07, "loss": 0.0243, "step": 5152 }, { "epoch": 2.03, "learning_rate": 3.6938945949590057e-07, "loss": 0.026, "step": 5153 }, { "epoch": 2.03, "learning_rate": 3.6937588275717623e-07, "loss": 0.0245, "step": 5154 }, { "epoch": 2.03, "learning_rate": 3.693623032578824e-07, "loss": 0.0251, "step": 5155 }, { "epoch": 2.03, "learning_rate": 3.693487209982406e-07, "loss": 0.0204, "step": 5156 }, { "epoch": 2.03, "learning_rate": 3.6933513597847185e-07, "loss": 0.0268, "step": 5157 }, { "epoch": 2.03, "learning_rate": 3.6932154819879776e-07, "loss": 0.0246, "step": 5158 }, { "epoch": 2.03, "learning_rate": 3.693079576594399e-07, "loss": 0.0261, "step": 5159 }, { "epoch": 2.03, "learning_rate": 3.692943643606195e-07, "loss": 0.0277, "step": 5160 }, { "epoch": 2.03, "learning_rate": 3.6928076830255833e-07, "loss": 0.0252, "step": 5161 }, { "epoch": 2.03, "learning_rate": 3.6926716948547786e-07, "loss": 0.0237, "step": 5162 }, { "epoch": 2.03, "learning_rate": 3.6925356790959985e-07, "loss": 0.0274, "step": 5163 }, { "epoch": 2.04, "learning_rate": 3.692399635751459e-07, "loss": 0.0248, "step": 5164 }, { "epoch": 2.04, "learning_rate": 3.6922635648233777e-07, "loss": 0.0253, "step": 5165 }, { "epoch": 2.04, "learning_rate": 3.692127466313972e-07, "loss": 0.024, "step": 5166 }, { "epoch": 2.04, "learning_rate": 3.6919913402254604e-07, "loss": 0.0301, "step": 5167 }, { "epoch": 2.04, "learning_rate": 3.691855186560062e-07, "loss": 0.0241, "step": 5168 }, { "epoch": 2.04, "learning_rate": 3.6917190053199945e-07, "loss": 0.0215, "step": 5169 }, { "epoch": 2.04, "learning_rate": 3.691582796507479e-07, "loss": 0.0252, "step": 5170 }, { "epoch": 2.04, "learning_rate": 3.691446560124734e-07, "loss": 0.0224, "step": 5171 }, { "epoch": 2.04, "learning_rate": 3.691310296173982e-07, "loss": 0.0231, "step": 5172 }, { "epoch": 2.04, "learning_rate": 3.6911740046574414e-07, "loss": 0.0229, "step": 5173 }, { "epoch": 2.04, "learning_rate": 3.6910376855773356e-07, "loss": 0.0258, "step": 5174 }, { "epoch": 2.04, "learning_rate": 3.690901338935885e-07, "loss": 0.023, "step": 5175 }, { "epoch": 2.04, "learning_rate": 3.6907649647353135e-07, "loss": 0.0247, "step": 5176 }, { "epoch": 2.04, "learning_rate": 3.6906285629778415e-07, "loss": 0.027, "step": 5177 }, { "epoch": 2.04, "learning_rate": 3.6904921336656943e-07, "loss": 0.0255, "step": 5178 }, { "epoch": 2.04, "learning_rate": 3.6903556768010937e-07, "loss": 0.0212, "step": 5179 }, { "epoch": 2.04, "learning_rate": 3.690219192386266e-07, "loss": 0.0237, "step": 5180 }, { "epoch": 2.04, "learning_rate": 3.690082680423433e-07, "loss": 0.0248, "step": 5181 }, { "epoch": 2.04, "learning_rate": 3.6899461409148213e-07, "loss": 0.0233, "step": 5182 }, { "epoch": 2.04, "learning_rate": 3.6898095738626565e-07, "loss": 0.0259, "step": 5183 }, { "epoch": 2.04, "learning_rate": 3.689672979269163e-07, "loss": 0.0241, "step": 5184 }, { "epoch": 2.04, "learning_rate": 3.6895363571365685e-07, "loss": 0.0277, "step": 5185 }, { "epoch": 2.04, "learning_rate": 3.689399707467099e-07, "loss": 0.0289, "step": 5186 }, { "epoch": 2.04, "learning_rate": 3.689263030262982e-07, "loss": 0.0239, "step": 5187 }, { "epoch": 2.04, "learning_rate": 3.689126325526445e-07, "loss": 0.0232, "step": 5188 }, { "epoch": 2.05, "learning_rate": 3.6889895932597153e-07, "loss": 0.0278, "step": 5189 }, { "epoch": 2.05, "learning_rate": 3.688852833465023e-07, "loss": 0.0233, "step": 5190 }, { "epoch": 2.05, "learning_rate": 3.688716046144596e-07, "loss": 0.03, "step": 5191 }, { "epoch": 2.05, "learning_rate": 3.6885792313006643e-07, "loss": 0.0293, "step": 5192 }, { "epoch": 2.05, "learning_rate": 3.6884423889354573e-07, "loss": 0.021, "step": 5193 }, { "epoch": 2.05, "learning_rate": 3.688305519051205e-07, "loss": 0.0254, "step": 5194 }, { "epoch": 2.05, "learning_rate": 3.688168621650139e-07, "loss": 0.0276, "step": 5195 }, { "epoch": 2.05, "learning_rate": 3.688031696734491e-07, "loss": 0.0253, "step": 5196 }, { "epoch": 2.05, "learning_rate": 3.687894744306491e-07, "loss": 0.0211, "step": 5197 }, { "epoch": 2.05, "learning_rate": 3.6877577643683726e-07, "loss": 0.023, "step": 5198 }, { "epoch": 2.05, "learning_rate": 3.6876207569223665e-07, "loss": 0.0276, "step": 5199 }, { "epoch": 2.05, "learning_rate": 3.6874837219707083e-07, "loss": 0.024, "step": 5200 }, { "epoch": 2.05, "learning_rate": 3.6873466595156293e-07, "loss": 0.0247, "step": 5201 }, { "epoch": 2.05, "learning_rate": 3.6872095695593646e-07, "loss": 0.0253, "step": 5202 }, { "epoch": 2.05, "learning_rate": 3.687072452104148e-07, "loss": 0.0245, "step": 5203 }, { "epoch": 2.05, "learning_rate": 3.686935307152215e-07, "loss": 0.0211, "step": 5204 }, { "epoch": 2.05, "learning_rate": 3.6867981347057995e-07, "loss": 0.0246, "step": 5205 }, { "epoch": 2.05, "learning_rate": 3.686660934767138e-07, "loss": 0.0302, "step": 5206 }, { "epoch": 2.05, "learning_rate": 3.686523707338467e-07, "loss": 0.0215, "step": 5207 }, { "epoch": 2.05, "learning_rate": 3.6863864524220235e-07, "loss": 0.0228, "step": 5208 }, { "epoch": 2.05, "learning_rate": 3.686249170020043e-07, "loss": 0.0269, "step": 5209 }, { "epoch": 2.05, "learning_rate": 3.6861118601347643e-07, "loss": 0.024, "step": 5210 }, { "epoch": 2.05, "learning_rate": 3.685974522768425e-07, "loss": 0.0252, "step": 5211 }, { "epoch": 2.05, "learning_rate": 3.685837157923263e-07, "loss": 0.0202, "step": 5212 }, { "epoch": 2.05, "learning_rate": 3.685699765601517e-07, "loss": 0.0245, "step": 5213 }, { "epoch": 2.05, "learning_rate": 3.685562345805428e-07, "loss": 0.0283, "step": 5214 }, { "epoch": 2.06, "learning_rate": 3.685424898537233e-07, "loss": 0.0212, "step": 5215 }, { "epoch": 2.06, "learning_rate": 3.685287423799175e-07, "loss": 0.0243, "step": 5216 }, { "epoch": 2.06, "learning_rate": 3.6851499215934925e-07, "loss": 0.0286, "step": 5217 }, { "epoch": 2.06, "learning_rate": 3.685012391922428e-07, "loss": 0.0285, "step": 5218 }, { "epoch": 2.06, "learning_rate": 3.6848748347882223e-07, "loss": 0.0252, "step": 5219 }, { "epoch": 2.06, "learning_rate": 3.684737250193117e-07, "loss": 0.0247, "step": 5220 }, { "epoch": 2.06, "learning_rate": 3.684599638139356e-07, "loss": 0.0279, "step": 5221 }, { "epoch": 2.06, "learning_rate": 3.68446199862918e-07, "loss": 0.0265, "step": 5222 }, { "epoch": 2.06, "learning_rate": 3.6843243316648343e-07, "loss": 0.0236, "step": 5223 }, { "epoch": 2.06, "learning_rate": 3.6841866372485615e-07, "loss": 0.0234, "step": 5224 }, { "epoch": 2.06, "learning_rate": 3.684048915382607e-07, "loss": 0.0281, "step": 5225 }, { "epoch": 2.06, "learning_rate": 3.6839111660692134e-07, "loss": 0.0235, "step": 5226 }, { "epoch": 2.06, "learning_rate": 3.683773389310628e-07, "loss": 0.0266, "step": 5227 }, { "epoch": 2.06, "learning_rate": 3.683635585109095e-07, "loss": 0.0199, "step": 5228 }, { "epoch": 2.06, "learning_rate": 3.683497753466861e-07, "loss": 0.0191, "step": 5229 }, { "epoch": 2.06, "learning_rate": 3.6833598943861725e-07, "loss": 0.023, "step": 5230 }, { "epoch": 2.06, "learning_rate": 3.6832220078692753e-07, "loss": 0.0247, "step": 5231 }, { "epoch": 2.06, "learning_rate": 3.6830840939184183e-07, "loss": 0.0222, "step": 5232 }, { "epoch": 2.06, "learning_rate": 3.682946152535848e-07, "loss": 0.022, "step": 5233 }, { "epoch": 2.06, "learning_rate": 3.682808183723813e-07, "loss": 0.022, "step": 5234 }, { "epoch": 2.06, "learning_rate": 3.6826701874845635e-07, "loss": 0.0247, "step": 5235 }, { "epoch": 2.06, "learning_rate": 3.682532163820347e-07, "loss": 0.0241, "step": 5236 }, { "epoch": 2.06, "learning_rate": 3.6823941127334124e-07, "loss": 0.0219, "step": 5237 }, { "epoch": 2.06, "learning_rate": 3.682256034226011e-07, "loss": 0.0215, "step": 5238 }, { "epoch": 2.06, "learning_rate": 3.6821179283003927e-07, "loss": 0.0245, "step": 5239 }, { "epoch": 2.07, "learning_rate": 3.6819797949588094e-07, "loss": 0.0232, "step": 5240 }, { "epoch": 2.07, "learning_rate": 3.6818416342035113e-07, "loss": 0.0241, "step": 5241 }, { "epoch": 2.07, "learning_rate": 3.681703446036751e-07, "loss": 0.0218, "step": 5242 }, { "epoch": 2.07, "learning_rate": 3.6815652304607803e-07, "loss": 0.0213, "step": 5243 }, { "epoch": 2.07, "learning_rate": 3.6814269874778526e-07, "loss": 0.0292, "step": 5244 }, { "epoch": 2.07, "learning_rate": 3.6812887170902193e-07, "loss": 0.0296, "step": 5245 }, { "epoch": 2.07, "learning_rate": 3.681150419300136e-07, "loss": 0.0264, "step": 5246 }, { "epoch": 2.07, "learning_rate": 3.681012094109856e-07, "loss": 0.0256, "step": 5247 }, { "epoch": 2.07, "learning_rate": 3.680873741521633e-07, "loss": 0.0241, "step": 5248 }, { "epoch": 2.07, "learning_rate": 3.680735361537723e-07, "loss": 0.0233, "step": 5249 }, { "epoch": 2.07, "learning_rate": 3.680596954160381e-07, "loss": 0.0255, "step": 5250 }, { "epoch": 2.07, "learning_rate": 3.6804585193918627e-07, "loss": 0.0231, "step": 5251 }, { "epoch": 2.07, "learning_rate": 3.680320057234425e-07, "loss": 0.0256, "step": 5252 }, { "epoch": 2.07, "learning_rate": 3.680181567690324e-07, "loss": 0.0271, "step": 5253 }, { "epoch": 2.07, "learning_rate": 3.6800430507618173e-07, "loss": 0.0238, "step": 5254 }, { "epoch": 2.07, "learning_rate": 3.679904506451161e-07, "loss": 0.0244, "step": 5255 }, { "epoch": 2.07, "learning_rate": 3.679765934760616e-07, "loss": 0.0222, "step": 5256 }, { "epoch": 2.07, "learning_rate": 3.679627335692438e-07, "loss": 0.0246, "step": 5257 }, { "epoch": 2.07, "learning_rate": 3.679488709248888e-07, "loss": 0.0221, "step": 5258 }, { "epoch": 2.07, "learning_rate": 3.6793500554322244e-07, "loss": 0.0197, "step": 5259 }, { "epoch": 2.07, "learning_rate": 3.679211374244707e-07, "loss": 0.0266, "step": 5260 }, { "epoch": 2.07, "learning_rate": 3.679072665688596e-07, "loss": 0.0227, "step": 5261 }, { "epoch": 2.07, "learning_rate": 3.6789339297661535e-07, "loss": 0.0215, "step": 5262 }, { "epoch": 2.07, "learning_rate": 3.6787951664796387e-07, "loss": 0.0232, "step": 5263 }, { "epoch": 2.07, "learning_rate": 3.6786563758313144e-07, "loss": 0.0236, "step": 5264 }, { "epoch": 2.07, "learning_rate": 3.6785175578234426e-07, "loss": 0.027, "step": 5265 }, { "epoch": 2.08, "learning_rate": 3.6783787124582855e-07, "loss": 0.0242, "step": 5266 }, { "epoch": 2.08, "learning_rate": 3.678239839738106e-07, "loss": 0.0245, "step": 5267 }, { "epoch": 2.08, "learning_rate": 3.6781009396651677e-07, "loss": 0.0232, "step": 5268 }, { "epoch": 2.08, "learning_rate": 3.677962012241735e-07, "loss": 0.0259, "step": 5269 }, { "epoch": 2.08, "learning_rate": 3.677823057470071e-07, "loss": 0.0239, "step": 5270 }, { "epoch": 2.08, "learning_rate": 3.677684075352442e-07, "loss": 0.0261, "step": 5271 }, { "epoch": 2.08, "learning_rate": 3.6775450658911124e-07, "loss": 0.0295, "step": 5272 }, { "epoch": 2.08, "learning_rate": 3.677406029088347e-07, "loss": 0.0229, "step": 5273 }, { "epoch": 2.08, "learning_rate": 3.6772669649464136e-07, "loss": 0.0251, "step": 5274 }, { "epoch": 2.08, "learning_rate": 3.677127873467577e-07, "loss": 0.0239, "step": 5275 }, { "epoch": 2.08, "learning_rate": 3.6769887546541055e-07, "loss": 0.0218, "step": 5276 }, { "epoch": 2.08, "learning_rate": 3.6768496085082656e-07, "loss": 0.0232, "step": 5277 }, { "epoch": 2.08, "learning_rate": 3.676710435032326e-07, "loss": 0.0237, "step": 5278 }, { "epoch": 2.08, "learning_rate": 3.6765712342285546e-07, "loss": 0.0241, "step": 5279 }, { "epoch": 2.08, "learning_rate": 3.67643200609922e-07, "loss": 0.026, "step": 5280 }, { "epoch": 2.08, "learning_rate": 3.6762927506465914e-07, "loss": 0.0229, "step": 5281 }, { "epoch": 2.08, "learning_rate": 3.676153467872939e-07, "loss": 0.0265, "step": 5282 }, { "epoch": 2.08, "learning_rate": 3.676014157780532e-07, "loss": 0.0226, "step": 5283 }, { "epoch": 2.08, "learning_rate": 3.6758748203716425e-07, "loss": 0.0256, "step": 5284 }, { "epoch": 2.08, "learning_rate": 3.67573545564854e-07, "loss": 0.0277, "step": 5285 }, { "epoch": 2.08, "learning_rate": 3.675596063613496e-07, "loss": 0.0228, "step": 5286 }, { "epoch": 2.08, "learning_rate": 3.6754566442687827e-07, "loss": 0.0233, "step": 5287 }, { "epoch": 2.08, "learning_rate": 3.6753171976166727e-07, "loss": 0.0191, "step": 5288 }, { "epoch": 2.08, "learning_rate": 3.675177723659439e-07, "loss": 0.0282, "step": 5289 }, { "epoch": 2.08, "learning_rate": 3.675038222399354e-07, "loss": 0.0286, "step": 5290 }, { "epoch": 2.09, "learning_rate": 3.674898693838692e-07, "loss": 0.0235, "step": 5291 }, { "epoch": 2.09, "learning_rate": 3.6747591379797264e-07, "loss": 0.0263, "step": 5292 }, { "epoch": 2.09, "learning_rate": 3.6746195548247323e-07, "loss": 0.0269, "step": 5293 }, { "epoch": 2.09, "learning_rate": 3.674479944375985e-07, "loss": 0.0249, "step": 5294 }, { "epoch": 2.09, "learning_rate": 3.6743403066357594e-07, "loss": 0.0221, "step": 5295 }, { "epoch": 2.09, "learning_rate": 3.6742006416063314e-07, "loss": 0.0272, "step": 5296 }, { "epoch": 2.09, "learning_rate": 3.674060949289977e-07, "loss": 0.0253, "step": 5297 }, { "epoch": 2.09, "learning_rate": 3.673921229688974e-07, "loss": 0.0231, "step": 5298 }, { "epoch": 2.09, "learning_rate": 3.6737814828055987e-07, "loss": 0.0283, "step": 5299 }, { "epoch": 2.09, "learning_rate": 3.673641708642129e-07, "loss": 0.0273, "step": 5300 }, { "epoch": 2.09, "learning_rate": 3.673501907200843e-07, "loss": 0.0248, "step": 5301 }, { "epoch": 2.09, "learning_rate": 3.6733620784840203e-07, "loss": 0.0212, "step": 5302 }, { "epoch": 2.09, "learning_rate": 3.673222222493938e-07, "loss": 0.0245, "step": 5303 }, { "epoch": 2.09, "learning_rate": 3.673082339232877e-07, "loss": 0.0243, "step": 5304 }, { "epoch": 2.09, "learning_rate": 3.672942428703116e-07, "loss": 0.0219, "step": 5305 }, { "epoch": 2.09, "learning_rate": 3.672802490906937e-07, "loss": 0.0286, "step": 5306 }, { "epoch": 2.09, "learning_rate": 3.672662525846619e-07, "loss": 0.0223, "step": 5307 }, { "epoch": 2.09, "learning_rate": 3.672522533524444e-07, "loss": 0.0237, "step": 5308 }, { "epoch": 2.09, "learning_rate": 3.672382513942694e-07, "loss": 0.0227, "step": 5309 }, { "epoch": 2.09, "learning_rate": 3.6722424671036505e-07, "loss": 0.0259, "step": 5310 }, { "epoch": 2.09, "learning_rate": 3.672102393009596e-07, "loss": 0.0234, "step": 5311 }, { "epoch": 2.09, "learning_rate": 3.6719622916628145e-07, "loss": 0.0273, "step": 5312 }, { "epoch": 2.09, "learning_rate": 3.6718221630655883e-07, "loss": 0.0299, "step": 5313 }, { "epoch": 2.09, "learning_rate": 3.6716820072202016e-07, "loss": 0.0254, "step": 5314 }, { "epoch": 2.09, "learning_rate": 3.671541824128939e-07, "loss": 0.0235, "step": 5315 }, { "epoch": 2.1, "learning_rate": 3.671401613794085e-07, "loss": 0.0234, "step": 5316 }, { "epoch": 2.1, "learning_rate": 3.6712613762179254e-07, "loss": 0.0241, "step": 5317 }, { "epoch": 2.1, "learning_rate": 3.6711211114027447e-07, "loss": 0.0214, "step": 5318 }, { "epoch": 2.1, "learning_rate": 3.6709808193508305e-07, "loss": 0.0239, "step": 5319 }, { "epoch": 2.1, "learning_rate": 3.670840500064468e-07, "loss": 0.0233, "step": 5320 }, { "epoch": 2.1, "learning_rate": 3.670700153545945e-07, "loss": 0.0194, "step": 5321 }, { "epoch": 2.1, "learning_rate": 3.6705597797975483e-07, "loss": 0.0279, "step": 5322 }, { "epoch": 2.1, "learning_rate": 3.6704193788215667e-07, "loss": 0.0239, "step": 5323 }, { "epoch": 2.1, "learning_rate": 3.6702789506202876e-07, "loss": 0.0207, "step": 5324 }, { "epoch": 2.1, "learning_rate": 3.670138495196e-07, "loss": 0.0205, "step": 5325 }, { "epoch": 2.1, "learning_rate": 3.6699980125509933e-07, "loss": 0.0217, "step": 5326 }, { "epoch": 2.1, "learning_rate": 3.6698575026875574e-07, "loss": 0.0225, "step": 5327 }, { "epoch": 2.1, "learning_rate": 3.669716965607982e-07, "loss": 0.0249, "step": 5328 }, { "epoch": 2.1, "learning_rate": 3.6695764013145577e-07, "loss": 0.0247, "step": 5329 }, { "epoch": 2.1, "learning_rate": 3.669435809809576e-07, "loss": 0.0247, "step": 5330 }, { "epoch": 2.1, "learning_rate": 3.669295191095327e-07, "loss": 0.0243, "step": 5331 }, { "epoch": 2.1, "learning_rate": 3.6691545451741037e-07, "loss": 0.0259, "step": 5332 }, { "epoch": 2.1, "learning_rate": 3.669013872048198e-07, "loss": 0.0243, "step": 5333 }, { "epoch": 2.1, "learning_rate": 3.668873171719903e-07, "loss": 0.0259, "step": 5334 }, { "epoch": 2.1, "learning_rate": 3.6687324441915115e-07, "loss": 0.0249, "step": 5335 }, { "epoch": 2.1, "learning_rate": 3.668591689465318e-07, "loss": 0.0226, "step": 5336 }, { "epoch": 2.1, "learning_rate": 3.668450907543615e-07, "loss": 0.0278, "step": 5337 }, { "epoch": 2.1, "learning_rate": 3.6683100984286985e-07, "loss": 0.0277, "step": 5338 }, { "epoch": 2.1, "learning_rate": 3.6681692621228633e-07, "loss": 0.0289, "step": 5339 }, { "epoch": 2.1, "learning_rate": 3.6680283986284033e-07, "loss": 0.0226, "step": 5340 }, { "epoch": 2.1, "learning_rate": 3.667887507947617e-07, "loss": 0.024, "step": 5341 }, { "epoch": 2.11, "learning_rate": 3.667746590082798e-07, "loss": 0.0237, "step": 5342 }, { "epoch": 2.11, "learning_rate": 3.6676056450362447e-07, "loss": 0.0234, "step": 5343 }, { "epoch": 2.11, "learning_rate": 3.667464672810254e-07, "loss": 0.0233, "step": 5344 }, { "epoch": 2.11, "learning_rate": 3.6673236734071235e-07, "loss": 0.0231, "step": 5345 }, { "epoch": 2.11, "learning_rate": 3.6671826468291507e-07, "loss": 0.0238, "step": 5346 }, { "epoch": 2.11, "learning_rate": 3.667041593078634e-07, "loss": 0.0254, "step": 5347 }, { "epoch": 2.11, "learning_rate": 3.666900512157874e-07, "loss": 0.0208, "step": 5348 }, { "epoch": 2.11, "learning_rate": 3.666759404069169e-07, "loss": 0.0218, "step": 5349 }, { "epoch": 2.11, "learning_rate": 3.6666182688148187e-07, "loss": 0.0254, "step": 5350 }, { "epoch": 2.11, "learning_rate": 3.6664771063971234e-07, "loss": 0.0233, "step": 5351 }, { "epoch": 2.11, "learning_rate": 3.666335916818385e-07, "loss": 0.0229, "step": 5352 }, { "epoch": 2.11, "learning_rate": 3.6661947000809024e-07, "loss": 0.0266, "step": 5353 }, { "epoch": 2.11, "learning_rate": 3.666053456186979e-07, "loss": 0.0239, "step": 5354 }, { "epoch": 2.11, "learning_rate": 3.6659121851389163e-07, "loss": 0.0245, "step": 5355 }, { "epoch": 2.11, "learning_rate": 3.665770886939017e-07, "loss": 0.0263, "step": 5356 }, { "epoch": 2.11, "learning_rate": 3.665629561589584e-07, "loss": 0.0282, "step": 5357 }, { "epoch": 2.11, "learning_rate": 3.6654882090929206e-07, "loss": 0.0208, "step": 5358 }, { "epoch": 2.11, "learning_rate": 3.66534682945133e-07, "loss": 0.0276, "step": 5359 }, { "epoch": 2.11, "learning_rate": 3.6652054226671176e-07, "loss": 0.0218, "step": 5360 }, { "epoch": 2.11, "learning_rate": 3.665063988742588e-07, "loss": 0.025, "step": 5361 }, { "epoch": 2.11, "learning_rate": 3.6649225276800457e-07, "loss": 0.0225, "step": 5362 }, { "epoch": 2.11, "learning_rate": 3.664781039481796e-07, "loss": 0.0244, "step": 5363 }, { "epoch": 2.11, "learning_rate": 3.6646395241501465e-07, "loss": 0.0231, "step": 5364 }, { "epoch": 2.11, "learning_rate": 3.664497981687402e-07, "loss": 0.0263, "step": 5365 }, { "epoch": 2.11, "learning_rate": 3.664356412095871e-07, "loss": 0.0261, "step": 5366 }, { "epoch": 2.12, "learning_rate": 3.6642148153778587e-07, "loss": 0.0235, "step": 5367 }, { "epoch": 2.12, "learning_rate": 3.664073191535675e-07, "loss": 0.0237, "step": 5368 }, { "epoch": 2.12, "learning_rate": 3.6639315405716275e-07, "loss": 0.0237, "step": 5369 }, { "epoch": 2.12, "learning_rate": 3.663789862488025e-07, "loss": 0.0251, "step": 5370 }, { "epoch": 2.12, "learning_rate": 3.663648157287176e-07, "loss": 0.0277, "step": 5371 }, { "epoch": 2.12, "learning_rate": 3.6635064249713904e-07, "loss": 0.0242, "step": 5372 }, { "epoch": 2.12, "learning_rate": 3.6633646655429786e-07, "loss": 0.0233, "step": 5373 }, { "epoch": 2.12, "learning_rate": 3.6632228790042503e-07, "loss": 0.0242, "step": 5374 }, { "epoch": 2.12, "learning_rate": 3.6630810653575174e-07, "loss": 0.0283, "step": 5375 }, { "epoch": 2.12, "learning_rate": 3.6629392246050907e-07, "loss": 0.0253, "step": 5376 }, { "epoch": 2.12, "learning_rate": 3.6627973567492815e-07, "loss": 0.0245, "step": 5377 }, { "epoch": 2.12, "learning_rate": 3.6626554617924027e-07, "loss": 0.0259, "step": 5378 }, { "epoch": 2.12, "learning_rate": 3.662513539736767e-07, "loss": 0.0237, "step": 5379 }, { "epoch": 2.12, "learning_rate": 3.662371590584687e-07, "loss": 0.0251, "step": 5380 }, { "epoch": 2.12, "learning_rate": 3.662229614338478e-07, "loss": 0.0263, "step": 5381 }, { "epoch": 2.12, "learning_rate": 3.6620876110004507e-07, "loss": 0.0269, "step": 5382 }, { "epoch": 2.12, "learning_rate": 3.6619455805729224e-07, "loss": 0.0238, "step": 5383 }, { "epoch": 2.12, "learning_rate": 3.6618035230582074e-07, "loss": 0.0282, "step": 5384 }, { "epoch": 2.12, "learning_rate": 3.6616614384586196e-07, "loss": 0.0261, "step": 5385 }, { "epoch": 2.12, "learning_rate": 3.661519326776476e-07, "loss": 0.0223, "step": 5386 }, { "epoch": 2.12, "learning_rate": 3.6613771880140927e-07, "loss": 0.0206, "step": 5387 }, { "epoch": 2.12, "learning_rate": 3.6612350221737867e-07, "loss": 0.0228, "step": 5388 }, { "epoch": 2.12, "learning_rate": 3.661092829257874e-07, "loss": 0.0254, "step": 5389 }, { "epoch": 2.12, "learning_rate": 3.6609506092686734e-07, "loss": 0.0253, "step": 5390 }, { "epoch": 2.12, "learning_rate": 3.6608083622085015e-07, "loss": 0.0266, "step": 5391 }, { "epoch": 2.13, "learning_rate": 3.660666088079678e-07, "loss": 0.0288, "step": 5392 }, { "epoch": 2.13, "learning_rate": 3.6605237868845215e-07, "loss": 0.0234, "step": 5393 }, { "epoch": 2.13, "learning_rate": 3.66038145862535e-07, "loss": 0.0207, "step": 5394 }, { "epoch": 2.13, "learning_rate": 3.6602391033044845e-07, "loss": 0.0236, "step": 5395 }, { "epoch": 2.13, "learning_rate": 3.660096720924245e-07, "loss": 0.0244, "step": 5396 }, { "epoch": 2.13, "learning_rate": 3.659954311486952e-07, "loss": 0.0214, "step": 5397 }, { "epoch": 2.13, "learning_rate": 3.6598118749949265e-07, "loss": 0.0241, "step": 5398 }, { "epoch": 2.13, "learning_rate": 3.6596694114504905e-07, "loss": 0.0273, "step": 5399 }, { "epoch": 2.13, "learning_rate": 3.6595269208559653e-07, "loss": 0.0206, "step": 5400 }, { "epoch": 2.13, "learning_rate": 3.6593844032136733e-07, "loss": 0.0245, "step": 5401 }, { "epoch": 2.13, "learning_rate": 3.6592418585259374e-07, "loss": 0.0225, "step": 5402 }, { "epoch": 2.13, "learning_rate": 3.659099286795081e-07, "loss": 0.0231, "step": 5403 }, { "epoch": 2.13, "learning_rate": 3.658956688023428e-07, "loss": 0.0279, "step": 5404 }, { "epoch": 2.13, "learning_rate": 3.6588140622133025e-07, "loss": 0.0194, "step": 5405 }, { "epoch": 2.13, "learning_rate": 3.6586714093670285e-07, "loss": 0.0247, "step": 5406 }, { "epoch": 2.13, "learning_rate": 3.6585287294869315e-07, "loss": 0.0196, "step": 5407 }, { "epoch": 2.13, "learning_rate": 3.658386022575337e-07, "loss": 0.0264, "step": 5408 }, { "epoch": 2.13, "learning_rate": 3.658243288634571e-07, "loss": 0.0248, "step": 5409 }, { "epoch": 2.13, "learning_rate": 3.658100527666959e-07, "loss": 0.0304, "step": 5410 }, { "epoch": 2.13, "learning_rate": 3.6579577396748286e-07, "loss": 0.0249, "step": 5411 }, { "epoch": 2.13, "learning_rate": 3.657814924660507e-07, "loss": 0.0234, "step": 5412 }, { "epoch": 2.13, "learning_rate": 3.657672082626322e-07, "loss": 0.0257, "step": 5413 }, { "epoch": 2.13, "learning_rate": 3.657529213574601e-07, "loss": 0.0218, "step": 5414 }, { "epoch": 2.13, "learning_rate": 3.6573863175076734e-07, "loss": 0.0257, "step": 5415 }, { "epoch": 2.13, "learning_rate": 3.657243394427867e-07, "loss": 0.03, "step": 5416 }, { "epoch": 2.13, "learning_rate": 3.657100444337513e-07, "loss": 0.0244, "step": 5417 }, { "epoch": 2.14, "learning_rate": 3.656957467238939e-07, "loss": 0.0243, "step": 5418 }, { "epoch": 2.14, "learning_rate": 3.6568144631344775e-07, "loss": 0.0269, "step": 5419 }, { "epoch": 2.14, "learning_rate": 3.6566714320264583e-07, "loss": 0.0219, "step": 5420 }, { "epoch": 2.14, "learning_rate": 3.6565283739172126e-07, "loss": 0.0284, "step": 5421 }, { "epoch": 2.14, "learning_rate": 3.6563852888090716e-07, "loss": 0.0222, "step": 5422 }, { "epoch": 2.14, "learning_rate": 3.656242176704368e-07, "loss": 0.0285, "step": 5423 }, { "epoch": 2.14, "learning_rate": 3.6560990376054344e-07, "loss": 0.0217, "step": 5424 }, { "epoch": 2.14, "learning_rate": 3.655955871514603e-07, "loss": 0.0194, "step": 5425 }, { "epoch": 2.14, "learning_rate": 3.6558126784342084e-07, "loss": 0.0228, "step": 5426 }, { "epoch": 2.14, "learning_rate": 3.6556694583665827e-07, "loss": 0.0247, "step": 5427 }, { "epoch": 2.14, "learning_rate": 3.6555262113140617e-07, "loss": 0.0223, "step": 5428 }, { "epoch": 2.14, "learning_rate": 3.6553829372789795e-07, "loss": 0.0215, "step": 5429 }, { "epoch": 2.14, "learning_rate": 3.655239636263672e-07, "loss": 0.0274, "step": 5430 }, { "epoch": 2.14, "learning_rate": 3.655096308270473e-07, "loss": 0.0272, "step": 5431 }, { "epoch": 2.14, "learning_rate": 3.65495295330172e-07, "loss": 0.025, "step": 5432 }, { "epoch": 2.14, "learning_rate": 3.654809571359749e-07, "loss": 0.0222, "step": 5433 }, { "epoch": 2.14, "learning_rate": 3.654666162446897e-07, "loss": 0.0205, "step": 5434 }, { "epoch": 2.14, "learning_rate": 3.654522726565502e-07, "loss": 0.0234, "step": 5435 }, { "epoch": 2.14, "learning_rate": 3.6543792637179004e-07, "loss": 0.0268, "step": 5436 }, { "epoch": 2.14, "learning_rate": 3.654235773906432e-07, "loss": 0.028, "step": 5437 }, { "epoch": 2.14, "learning_rate": 3.654092257133434e-07, "loss": 0.0254, "step": 5438 }, { "epoch": 2.14, "learning_rate": 3.653948713401246e-07, "loss": 0.0252, "step": 5439 }, { "epoch": 2.14, "learning_rate": 3.6538051427122076e-07, "loss": 0.0262, "step": 5440 }, { "epoch": 2.14, "learning_rate": 3.6536615450686597e-07, "loss": 0.027, "step": 5441 }, { "epoch": 2.14, "learning_rate": 3.653517920472941e-07, "loss": 0.0241, "step": 5442 }, { "epoch": 2.15, "learning_rate": 3.653374268927393e-07, "loss": 0.0243, "step": 5443 }, { "epoch": 2.15, "learning_rate": 3.653230590434358e-07, "loss": 0.0218, "step": 5444 }, { "epoch": 2.15, "learning_rate": 3.653086884996177e-07, "loss": 0.0224, "step": 5445 }, { "epoch": 2.15, "learning_rate": 3.652943152615192e-07, "loss": 0.0215, "step": 5446 }, { "epoch": 2.15, "learning_rate": 3.6527993932937463e-07, "loss": 0.0233, "step": 5447 }, { "epoch": 2.15, "learning_rate": 3.652655607034182e-07, "loss": 0.023, "step": 5448 }, { "epoch": 2.15, "learning_rate": 3.652511793838843e-07, "loss": 0.0237, "step": 5449 }, { "epoch": 2.15, "learning_rate": 3.6523679537100734e-07, "loss": 0.0304, "step": 5450 }, { "epoch": 2.15, "learning_rate": 3.652224086650217e-07, "loss": 0.0277, "step": 5451 }, { "epoch": 2.15, "learning_rate": 3.65208019266162e-07, "loss": 0.0204, "step": 5452 }, { "epoch": 2.15, "learning_rate": 3.651936271746626e-07, "loss": 0.0245, "step": 5453 }, { "epoch": 2.15, "learning_rate": 3.651792323907581e-07, "loss": 0.025, "step": 5454 }, { "epoch": 2.15, "learning_rate": 3.651648349146833e-07, "loss": 0.0239, "step": 5455 }, { "epoch": 2.15, "learning_rate": 3.651504347466726e-07, "loss": 0.0267, "step": 5456 }, { "epoch": 2.15, "learning_rate": 3.651360318869608e-07, "loss": 0.0215, "step": 5457 }, { "epoch": 2.15, "learning_rate": 3.6512162633578276e-07, "loss": 0.0245, "step": 5458 }, { "epoch": 2.15, "learning_rate": 3.651072180933731e-07, "loss": 0.0239, "step": 5459 }, { "epoch": 2.15, "learning_rate": 3.6509280715996674e-07, "loss": 0.0252, "step": 5460 }, { "epoch": 2.15, "learning_rate": 3.650783935357985e-07, "loss": 0.0243, "step": 5461 }, { "epoch": 2.15, "learning_rate": 3.650639772211034e-07, "loss": 0.0248, "step": 5462 }, { "epoch": 2.15, "learning_rate": 3.650495582161163e-07, "loss": 0.0195, "step": 5463 }, { "epoch": 2.15, "learning_rate": 3.650351365210722e-07, "loss": 0.0234, "step": 5464 }, { "epoch": 2.15, "learning_rate": 3.6502071213620624e-07, "loss": 0.0213, "step": 5465 }, { "epoch": 2.15, "learning_rate": 3.650062850617535e-07, "loss": 0.0212, "step": 5466 }, { "epoch": 2.15, "learning_rate": 3.649918552979491e-07, "loss": 0.0245, "step": 5467 }, { "epoch": 2.15, "learning_rate": 3.649774228450282e-07, "loss": 0.0259, "step": 5468 }, { "epoch": 2.16, "learning_rate": 3.64962987703226e-07, "loss": 0.0293, "step": 5469 }, { "epoch": 2.16, "learning_rate": 3.6494854987277785e-07, "loss": 0.0227, "step": 5470 }, { "epoch": 2.16, "learning_rate": 3.64934109353919e-07, "loss": 0.0242, "step": 5471 }, { "epoch": 2.16, "learning_rate": 3.649196661468849e-07, "loss": 0.0275, "step": 5472 }, { "epoch": 2.16, "learning_rate": 3.649052202519109e-07, "loss": 0.0235, "step": 5473 }, { "epoch": 2.16, "learning_rate": 3.6489077166923234e-07, "loss": 0.0225, "step": 5474 }, { "epoch": 2.16, "learning_rate": 3.648763203990849e-07, "loss": 0.0206, "step": 5475 }, { "epoch": 2.16, "learning_rate": 3.6486186644170396e-07, "loss": 0.0213, "step": 5476 }, { "epoch": 2.16, "learning_rate": 3.648474097973252e-07, "loss": 0.0244, "step": 5477 }, { "epoch": 2.16, "learning_rate": 3.648329504661842e-07, "loss": 0.0266, "step": 5478 }, { "epoch": 2.16, "learning_rate": 3.6481848844851657e-07, "loss": 0.0262, "step": 5479 }, { "epoch": 2.16, "learning_rate": 3.6480402374455813e-07, "loss": 0.026, "step": 5480 }, { "epoch": 2.16, "learning_rate": 3.647895563545446e-07, "loss": 0.0269, "step": 5481 }, { "epoch": 2.16, "learning_rate": 3.6477508627871167e-07, "loss": 0.0228, "step": 5482 }, { "epoch": 2.16, "learning_rate": 3.6476061351729535e-07, "loss": 0.0234, "step": 5483 }, { "epoch": 2.16, "learning_rate": 3.647461380705313e-07, "loss": 0.0229, "step": 5484 }, { "epoch": 2.16, "learning_rate": 3.647316599386557e-07, "loss": 0.0237, "step": 5485 }, { "epoch": 2.16, "learning_rate": 3.647171791219044e-07, "loss": 0.0292, "step": 5486 }, { "epoch": 2.16, "learning_rate": 3.6470269562051337e-07, "loss": 0.0232, "step": 5487 }, { "epoch": 2.16, "learning_rate": 3.646882094347188e-07, "loss": 0.0238, "step": 5488 }, { "epoch": 2.16, "learning_rate": 3.646737205647566e-07, "loss": 0.0257, "step": 5489 }, { "epoch": 2.16, "learning_rate": 3.6465922901086314e-07, "loss": 0.0285, "step": 5490 }, { "epoch": 2.16, "learning_rate": 3.6464473477327446e-07, "loss": 0.0242, "step": 5491 }, { "epoch": 2.16, "learning_rate": 3.6463023785222686e-07, "loss": 0.0217, "step": 5492 }, { "epoch": 2.16, "learning_rate": 3.6461573824795655e-07, "loss": 0.0232, "step": 5493 }, { "epoch": 2.17, "learning_rate": 3.646012359606999e-07, "loss": 0.0244, "step": 5494 }, { "epoch": 2.17, "learning_rate": 3.645867309906933e-07, "loss": 0.0217, "step": 5495 }, { "epoch": 2.17, "learning_rate": 3.645722233381731e-07, "loss": 0.0263, "step": 5496 }, { "epoch": 2.17, "learning_rate": 3.645577130033758e-07, "loss": 0.0241, "step": 5497 }, { "epoch": 2.17, "learning_rate": 3.645431999865378e-07, "loss": 0.0246, "step": 5498 }, { "epoch": 2.17, "learning_rate": 3.645286842878958e-07, "loss": 0.0222, "step": 5499 }, { "epoch": 2.17, "learning_rate": 3.6451416590768627e-07, "loss": 0.0234, "step": 5500 }, { "epoch": 2.17, "learning_rate": 3.644996448461459e-07, "loss": 0.0243, "step": 5501 }, { "epoch": 2.17, "learning_rate": 3.644851211035113e-07, "loss": 0.0237, "step": 5502 }, { "epoch": 2.17, "learning_rate": 3.6447059468001924e-07, "loss": 0.0232, "step": 5503 }, { "epoch": 2.17, "learning_rate": 3.644560655759064e-07, "loss": 0.0227, "step": 5504 }, { "epoch": 2.17, "learning_rate": 3.6444153379140967e-07, "loss": 0.0229, "step": 5505 }, { "epoch": 2.17, "learning_rate": 3.6442699932676584e-07, "loss": 0.0246, "step": 5506 }, { "epoch": 2.17, "learning_rate": 3.644124621822119e-07, "loss": 0.027, "step": 5507 }, { "epoch": 2.17, "learning_rate": 3.6439792235798464e-07, "loss": 0.0239, "step": 5508 }, { "epoch": 2.17, "learning_rate": 3.643833798543211e-07, "loss": 0.0237, "step": 5509 }, { "epoch": 2.17, "learning_rate": 3.643688346714583e-07, "loss": 0.0301, "step": 5510 }, { "epoch": 2.17, "learning_rate": 3.6435428680963336e-07, "loss": 0.0225, "step": 5511 }, { "epoch": 2.17, "learning_rate": 3.643397362690833e-07, "loss": 0.0253, "step": 5512 }, { "epoch": 2.17, "learning_rate": 3.643251830500452e-07, "loss": 0.0264, "step": 5513 }, { "epoch": 2.17, "learning_rate": 3.643106271527565e-07, "loss": 0.0212, "step": 5514 }, { "epoch": 2.17, "learning_rate": 3.642960685774543e-07, "loss": 0.0262, "step": 5515 }, { "epoch": 2.17, "learning_rate": 3.642815073243758e-07, "loss": 0.0251, "step": 5516 }, { "epoch": 2.17, "learning_rate": 3.642669433937585e-07, "loss": 0.0284, "step": 5517 }, { "epoch": 2.17, "learning_rate": 3.642523767858396e-07, "loss": 0.0239, "step": 5518 }, { "epoch": 2.18, "learning_rate": 3.6423780750085664e-07, "loss": 0.0216, "step": 5519 }, { "epoch": 2.18, "learning_rate": 3.6422323553904705e-07, "loss": 0.0248, "step": 5520 }, { "epoch": 2.18, "learning_rate": 3.642086609006483e-07, "loss": 0.025, "step": 5521 }, { "epoch": 2.18, "learning_rate": 3.641940835858979e-07, "loss": 0.0275, "step": 5522 }, { "epoch": 2.18, "learning_rate": 3.641795035950335e-07, "loss": 0.0228, "step": 5523 }, { "epoch": 2.18, "learning_rate": 3.6416492092829274e-07, "loss": 0.0236, "step": 5524 }, { "epoch": 2.18, "learning_rate": 3.641503355859132e-07, "loss": 0.0251, "step": 5525 }, { "epoch": 2.18, "learning_rate": 3.6413574756813275e-07, "loss": 0.0194, "step": 5526 }, { "epoch": 2.18, "learning_rate": 3.641211568751891e-07, "loss": 0.0207, "step": 5527 }, { "epoch": 2.18, "learning_rate": 3.641065635073199e-07, "loss": 0.0224, "step": 5528 }, { "epoch": 2.18, "learning_rate": 3.6409196746476323e-07, "loss": 0.0245, "step": 5529 }, { "epoch": 2.18, "learning_rate": 3.6407736874775687e-07, "loss": 0.026, "step": 5530 }, { "epoch": 2.18, "learning_rate": 3.6406276735653874e-07, "loss": 0.0214, "step": 5531 }, { "epoch": 2.18, "learning_rate": 3.6404816329134687e-07, "loss": 0.024, "step": 5532 }, { "epoch": 2.18, "learning_rate": 3.640335565524193e-07, "loss": 0.022, "step": 5533 }, { "epoch": 2.18, "learning_rate": 3.6401894713999394e-07, "loss": 0.0263, "step": 5534 }, { "epoch": 2.18, "learning_rate": 3.6400433505430914e-07, "loss": 0.0215, "step": 5535 }, { "epoch": 2.18, "learning_rate": 3.639897202956029e-07, "loss": 0.0225, "step": 5536 }, { "epoch": 2.18, "learning_rate": 3.6397510286411345e-07, "loss": 0.025, "step": 5537 }, { "epoch": 2.18, "learning_rate": 3.63960482760079e-07, "loss": 0.025, "step": 5538 }, { "epoch": 2.18, "learning_rate": 3.6394585998373796e-07, "loss": 0.0245, "step": 5539 }, { "epoch": 2.18, "learning_rate": 3.6393123453532846e-07, "loss": 0.023, "step": 5540 }, { "epoch": 2.18, "learning_rate": 3.639166064150891e-07, "loss": 0.0284, "step": 5541 }, { "epoch": 2.18, "learning_rate": 3.639019756232581e-07, "loss": 0.0212, "step": 5542 }, { "epoch": 2.18, "learning_rate": 3.6388734216007404e-07, "loss": 0.0215, "step": 5543 }, { "epoch": 2.18, "learning_rate": 3.6387270602577533e-07, "loss": 0.0204, "step": 5544 }, { "epoch": 2.19, "learning_rate": 3.6385806722060063e-07, "loss": 0.0216, "step": 5545 }, { "epoch": 2.19, "learning_rate": 3.6384342574478844e-07, "loss": 0.0205, "step": 5546 }, { "epoch": 2.19, "learning_rate": 3.6382878159857745e-07, "loss": 0.023, "step": 5547 }, { "epoch": 2.19, "learning_rate": 3.638141347822063e-07, "loss": 0.0215, "step": 5548 }, { "epoch": 2.19, "learning_rate": 3.637994852959137e-07, "loss": 0.0281, "step": 5549 }, { "epoch": 2.19, "learning_rate": 3.637848331399384e-07, "loss": 0.0234, "step": 5550 }, { "epoch": 2.19, "learning_rate": 3.637701783145193e-07, "loss": 0.0225, "step": 5551 }, { "epoch": 2.19, "learning_rate": 3.637555208198952e-07, "loss": 0.0234, "step": 5552 }, { "epoch": 2.19, "learning_rate": 3.63740860656305e-07, "loss": 0.0271, "step": 5553 }, { "epoch": 2.19, "learning_rate": 3.6372619782398763e-07, "loss": 0.0204, "step": 5554 }, { "epoch": 2.19, "learning_rate": 3.6371153232318205e-07, "loss": 0.0267, "step": 5555 }, { "epoch": 2.19, "learning_rate": 3.636968641541273e-07, "loss": 0.0235, "step": 5556 }, { "epoch": 2.19, "learning_rate": 3.6368219331706243e-07, "loss": 0.024, "step": 5557 }, { "epoch": 2.19, "learning_rate": 3.6366751981222664e-07, "loss": 0.0255, "step": 5558 }, { "epoch": 2.19, "learning_rate": 3.6365284363985895e-07, "loss": 0.0269, "step": 5559 }, { "epoch": 2.19, "learning_rate": 3.6363816480019865e-07, "loss": 0.0245, "step": 5560 }, { "epoch": 2.19, "learning_rate": 3.63623483293485e-07, "loss": 0.0287, "step": 5561 }, { "epoch": 2.19, "learning_rate": 3.636087991199573e-07, "loss": 0.0266, "step": 5562 }, { "epoch": 2.19, "learning_rate": 3.635941122798547e-07, "loss": 0.025, "step": 5563 }, { "epoch": 2.19, "learning_rate": 3.635794227734168e-07, "loss": 0.0225, "step": 5564 }, { "epoch": 2.19, "learning_rate": 3.635647306008829e-07, "loss": 0.0277, "step": 5565 }, { "epoch": 2.19, "learning_rate": 3.6355003576249246e-07, "loss": 0.0218, "step": 5566 }, { "epoch": 2.19, "learning_rate": 3.6353533825848504e-07, "loss": 0.0249, "step": 5567 }, { "epoch": 2.19, "learning_rate": 3.6352063808910014e-07, "loss": 0.0209, "step": 5568 }, { "epoch": 2.19, "learning_rate": 3.635059352545773e-07, "loss": 0.0208, "step": 5569 }, { "epoch": 2.2, "learning_rate": 3.634912297551563e-07, "loss": 0.0246, "step": 5570 }, { "epoch": 2.2, "learning_rate": 3.634765215910767e-07, "loss": 0.0268, "step": 5571 }, { "epoch": 2.2, "learning_rate": 3.6346181076257826e-07, "loss": 0.0192, "step": 5572 }, { "epoch": 2.2, "learning_rate": 3.6344709726990074e-07, "loss": 0.0241, "step": 5573 }, { "epoch": 2.2, "learning_rate": 3.634323811132839e-07, "loss": 0.0231, "step": 5574 }, { "epoch": 2.2, "learning_rate": 3.634176622929677e-07, "loss": 0.0219, "step": 5575 }, { "epoch": 2.2, "learning_rate": 3.634029408091919e-07, "loss": 0.024, "step": 5576 }, { "epoch": 2.2, "learning_rate": 3.6338821666219655e-07, "loss": 0.0227, "step": 5577 }, { "epoch": 2.2, "learning_rate": 3.633734898522216e-07, "loss": 0.018, "step": 5578 }, { "epoch": 2.2, "learning_rate": 3.633587603795071e-07, "loss": 0.0248, "step": 5579 }, { "epoch": 2.2, "learning_rate": 3.6334402824429306e-07, "loss": 0.0247, "step": 5580 }, { "epoch": 2.2, "learning_rate": 3.633292934468196e-07, "loss": 0.0255, "step": 5581 }, { "epoch": 2.2, "learning_rate": 3.6331455598732685e-07, "loss": 0.0228, "step": 5582 }, { "epoch": 2.2, "learning_rate": 3.632998158660551e-07, "loss": 0.0261, "step": 5583 }, { "epoch": 2.2, "learning_rate": 3.632850730832445e-07, "loss": 0.0244, "step": 5584 }, { "epoch": 2.2, "learning_rate": 3.632703276391354e-07, "loss": 0.0275, "step": 5585 }, { "epoch": 2.2, "learning_rate": 3.6325557953396813e-07, "loss": 0.0237, "step": 5586 }, { "epoch": 2.2, "learning_rate": 3.6324082876798296e-07, "loss": 0.0226, "step": 5587 }, { "epoch": 2.2, "learning_rate": 3.632260753414204e-07, "loss": 0.0249, "step": 5588 }, { "epoch": 2.2, "learning_rate": 3.632113192545209e-07, "loss": 0.0233, "step": 5589 }, { "epoch": 2.2, "learning_rate": 3.631965605075249e-07, "loss": 0.0238, "step": 5590 }, { "epoch": 2.2, "learning_rate": 3.6318179910067316e-07, "loss": 0.0247, "step": 5591 }, { "epoch": 2.2, "learning_rate": 3.631670350342059e-07, "loss": 0.026, "step": 5592 }, { "epoch": 2.2, "learning_rate": 3.63152268308364e-07, "loss": 0.0201, "step": 5593 }, { "epoch": 2.2, "learning_rate": 3.6313749892338824e-07, "loss": 0.0216, "step": 5594 }, { "epoch": 2.21, "learning_rate": 3.63122726879519e-07, "loss": 0.0243, "step": 5595 }, { "epoch": 2.21, "learning_rate": 3.6310795217699734e-07, "loss": 0.0237, "step": 5596 }, { "epoch": 2.21, "learning_rate": 3.630931748160639e-07, "loss": 0.0221, "step": 5597 }, { "epoch": 2.21, "learning_rate": 3.6307839479695963e-07, "loss": 0.0296, "step": 5598 }, { "epoch": 2.21, "learning_rate": 3.630636121199254e-07, "loss": 0.0278, "step": 5599 }, { "epoch": 2.21, "learning_rate": 3.6304882678520207e-07, "loss": 0.0235, "step": 5600 }, { "epoch": 2.21, "learning_rate": 3.6303403879303065e-07, "loss": 0.0203, "step": 5601 }, { "epoch": 2.21, "learning_rate": 3.6301924814365224e-07, "loss": 0.0245, "step": 5602 }, { "epoch": 2.21, "learning_rate": 3.630044548373078e-07, "loss": 0.0248, "step": 5603 }, { "epoch": 2.21, "learning_rate": 3.629896588742385e-07, "loss": 0.0263, "step": 5604 }, { "epoch": 2.21, "learning_rate": 3.629748602546854e-07, "loss": 0.0235, "step": 5605 }, { "epoch": 2.21, "learning_rate": 3.629600589788899e-07, "loss": 0.027, "step": 5606 }, { "epoch": 2.21, "learning_rate": 3.6294525504709307e-07, "loss": 0.025, "step": 5607 }, { "epoch": 2.21, "learning_rate": 3.629304484595362e-07, "loss": 0.0249, "step": 5608 }, { "epoch": 2.21, "learning_rate": 3.629156392164606e-07, "loss": 0.022, "step": 5609 }, { "epoch": 2.21, "learning_rate": 3.6290082731810776e-07, "loss": 0.024, "step": 5610 }, { "epoch": 2.21, "learning_rate": 3.62886012764719e-07, "loss": 0.02, "step": 5611 }, { "epoch": 2.21, "learning_rate": 3.628711955565358e-07, "loss": 0.0235, "step": 5612 }, { "epoch": 2.21, "learning_rate": 3.6285637569379965e-07, "loss": 0.0241, "step": 5613 }, { "epoch": 2.21, "learning_rate": 3.6284155317675207e-07, "loss": 0.0271, "step": 5614 }, { "epoch": 2.21, "learning_rate": 3.628267280056347e-07, "loss": 0.0232, "step": 5615 }, { "epoch": 2.21, "learning_rate": 3.6281190018068903e-07, "loss": 0.0292, "step": 5616 }, { "epoch": 2.21, "learning_rate": 3.6279706970215696e-07, "loss": 0.0201, "step": 5617 }, { "epoch": 2.21, "learning_rate": 3.6278223657028e-07, "loss": 0.0269, "step": 5618 }, { "epoch": 2.21, "learning_rate": 3.627674007853e-07, "loss": 0.0202, "step": 5619 }, { "epoch": 2.21, "learning_rate": 3.6275256234745876e-07, "loss": 0.0235, "step": 5620 }, { "epoch": 2.22, "learning_rate": 3.6273772125699813e-07, "loss": 0.0216, "step": 5621 }, { "epoch": 2.22, "learning_rate": 3.627228775141599e-07, "loss": 0.0226, "step": 5622 }, { "epoch": 2.22, "learning_rate": 3.627080311191861e-07, "loss": 0.0258, "step": 5623 }, { "epoch": 2.22, "learning_rate": 3.6269318207231875e-07, "loss": 0.0223, "step": 5624 }, { "epoch": 2.22, "learning_rate": 3.626783303737997e-07, "loss": 0.0217, "step": 5625 }, { "epoch": 2.22, "learning_rate": 3.626634760238712e-07, "loss": 0.0233, "step": 5626 }, { "epoch": 2.22, "learning_rate": 3.6264861902277523e-07, "loss": 0.0247, "step": 5627 }, { "epoch": 2.22, "learning_rate": 3.6263375937075393e-07, "loss": 0.0256, "step": 5628 }, { "epoch": 2.22, "learning_rate": 3.626188970680496e-07, "loss": 0.0206, "step": 5629 }, { "epoch": 2.22, "learning_rate": 3.626040321149043e-07, "loss": 0.0258, "step": 5630 }, { "epoch": 2.22, "learning_rate": 3.6258916451156047e-07, "loss": 0.0285, "step": 5631 }, { "epoch": 2.22, "learning_rate": 3.6257429425826035e-07, "loss": 0.0204, "step": 5632 }, { "epoch": 2.22, "learning_rate": 3.625594213552463e-07, "loss": 0.0239, "step": 5633 }, { "epoch": 2.22, "learning_rate": 3.625445458027608e-07, "loss": 0.0251, "step": 5634 }, { "epoch": 2.22, "learning_rate": 3.6252966760104623e-07, "loss": 0.0244, "step": 5635 }, { "epoch": 2.22, "learning_rate": 3.62514786750345e-07, "loss": 0.0209, "step": 5636 }, { "epoch": 2.22, "learning_rate": 3.624999032508999e-07, "loss": 0.0202, "step": 5637 }, { "epoch": 2.22, "learning_rate": 3.624850171029532e-07, "loss": 0.0259, "step": 5638 }, { "epoch": 2.22, "learning_rate": 3.6247012830674777e-07, "loss": 0.0248, "step": 5639 }, { "epoch": 2.22, "learning_rate": 3.6245523686252614e-07, "loss": 0.0281, "step": 5640 }, { "epoch": 2.22, "learning_rate": 3.6244034277053103e-07, "loss": 0.0235, "step": 5641 }, { "epoch": 2.22, "learning_rate": 3.624254460310052e-07, "loss": 0.0323, "step": 5642 }, { "epoch": 2.22, "learning_rate": 3.6241054664419153e-07, "loss": 0.0237, "step": 5643 }, { "epoch": 2.22, "learning_rate": 3.623956446103327e-07, "loss": 0.021, "step": 5644 }, { "epoch": 2.22, "learning_rate": 3.6238073992967173e-07, "loss": 0.0248, "step": 5645 }, { "epoch": 2.23, "learning_rate": 3.623658326024515e-07, "loss": 0.0217, "step": 5646 }, { "epoch": 2.23, "learning_rate": 3.6235092262891494e-07, "loss": 0.0226, "step": 5647 }, { "epoch": 2.23, "learning_rate": 3.623360100093051e-07, "loss": 0.0231, "step": 5648 }, { "epoch": 2.23, "learning_rate": 3.6232109474386505e-07, "loss": 0.0224, "step": 5649 }, { "epoch": 2.23, "learning_rate": 3.6230617683283775e-07, "loss": 0.0232, "step": 5650 }, { "epoch": 2.23, "learning_rate": 3.622912562764665e-07, "loss": 0.0226, "step": 5651 }, { "epoch": 2.23, "learning_rate": 3.622763330749945e-07, "loss": 0.0223, "step": 5652 }, { "epoch": 2.23, "learning_rate": 3.6226140722866484e-07, "loss": 0.0271, "step": 5653 }, { "epoch": 2.23, "learning_rate": 3.6224647873772084e-07, "loss": 0.027, "step": 5654 }, { "epoch": 2.23, "learning_rate": 3.6223154760240584e-07, "loss": 0.0252, "step": 5655 }, { "epoch": 2.23, "learning_rate": 3.622166138229632e-07, "loss": 0.0258, "step": 5656 }, { "epoch": 2.23, "learning_rate": 3.6220167739963623e-07, "loss": 0.0246, "step": 5657 }, { "epoch": 2.23, "learning_rate": 3.621867383326685e-07, "loss": 0.0302, "step": 5658 }, { "epoch": 2.23, "learning_rate": 3.6217179662230343e-07, "loss": 0.0222, "step": 5659 }, { "epoch": 2.23, "learning_rate": 3.621568522687845e-07, "loss": 0.0257, "step": 5660 }, { "epoch": 2.23, "learning_rate": 3.6214190527235536e-07, "loss": 0.0232, "step": 5661 }, { "epoch": 2.23, "learning_rate": 3.6212695563325956e-07, "loss": 0.0227, "step": 5662 }, { "epoch": 2.23, "learning_rate": 3.621120033517408e-07, "loss": 0.0218, "step": 5663 }, { "epoch": 2.23, "learning_rate": 3.620970484280428e-07, "loss": 0.0272, "step": 5664 }, { "epoch": 2.23, "learning_rate": 3.620820908624092e-07, "loss": 0.0207, "step": 5665 }, { "epoch": 2.23, "learning_rate": 3.6206713065508393e-07, "loss": 0.027, "step": 5666 }, { "epoch": 2.23, "learning_rate": 3.6205216780631074e-07, "loss": 0.0202, "step": 5667 }, { "epoch": 2.23, "learning_rate": 3.620372023163335e-07, "loss": 0.0272, "step": 5668 }, { "epoch": 2.23, "learning_rate": 3.620222341853961e-07, "loss": 0.0244, "step": 5669 }, { "epoch": 2.23, "learning_rate": 3.6200726341374253e-07, "loss": 0.0269, "step": 5670 }, { "epoch": 2.23, "learning_rate": 3.619922900016168e-07, "loss": 0.0221, "step": 5671 }, { "epoch": 2.24, "learning_rate": 3.6197731394926295e-07, "loss": 0.0222, "step": 5672 }, { "epoch": 2.24, "learning_rate": 3.619623352569251e-07, "loss": 0.0221, "step": 5673 }, { "epoch": 2.24, "learning_rate": 3.6194735392484723e-07, "loss": 0.0214, "step": 5674 }, { "epoch": 2.24, "learning_rate": 3.6193236995327367e-07, "loss": 0.0232, "step": 5675 }, { "epoch": 2.24, "learning_rate": 3.619173833424487e-07, "loss": 0.0235, "step": 5676 }, { "epoch": 2.24, "learning_rate": 3.619023940926163e-07, "loss": 0.0234, "step": 5677 }, { "epoch": 2.24, "learning_rate": 3.6188740220402103e-07, "loss": 0.0256, "step": 5678 }, { "epoch": 2.24, "learning_rate": 3.618724076769071e-07, "loss": 0.0257, "step": 5679 }, { "epoch": 2.24, "learning_rate": 3.61857410511519e-07, "loss": 0.0217, "step": 5680 }, { "epoch": 2.24, "learning_rate": 3.6184241070810104e-07, "loss": 0.0252, "step": 5681 }, { "epoch": 2.24, "learning_rate": 3.6182740826689783e-07, "loss": 0.0218, "step": 5682 }, { "epoch": 2.24, "learning_rate": 3.618124031881538e-07, "loss": 0.0245, "step": 5683 }, { "epoch": 2.24, "learning_rate": 3.6179739547211345e-07, "loss": 0.0272, "step": 5684 }, { "epoch": 2.24, "learning_rate": 3.6178238511902155e-07, "loss": 0.0231, "step": 5685 }, { "epoch": 2.24, "learning_rate": 3.6176737212912254e-07, "loss": 0.0247, "step": 5686 }, { "epoch": 2.24, "learning_rate": 3.617523565026613e-07, "loss": 0.0253, "step": 5687 }, { "epoch": 2.24, "learning_rate": 3.617373382398825e-07, "loss": 0.0248, "step": 5688 }, { "epoch": 2.24, "learning_rate": 3.6172231734103084e-07, "loss": 0.0262, "step": 5689 }, { "epoch": 2.24, "learning_rate": 3.6170729380635123e-07, "loss": 0.0251, "step": 5690 }, { "epoch": 2.24, "learning_rate": 3.6169226763608857e-07, "loss": 0.0228, "step": 5691 }, { "epoch": 2.24, "learning_rate": 3.616772388304876e-07, "loss": 0.0254, "step": 5692 }, { "epoch": 2.24, "learning_rate": 3.616622073897933e-07, "loss": 0.0246, "step": 5693 }, { "epoch": 2.24, "learning_rate": 3.616471733142509e-07, "loss": 0.0261, "step": 5694 }, { "epoch": 2.24, "learning_rate": 3.616321366041051e-07, "loss": 0.0263, "step": 5695 }, { "epoch": 2.24, "learning_rate": 3.616170972596012e-07, "loss": 0.0235, "step": 5696 }, { "epoch": 2.25, "learning_rate": 3.616020552809842e-07, "loss": 0.0227, "step": 5697 }, { "epoch": 2.25, "learning_rate": 3.615870106684993e-07, "loss": 0.0202, "step": 5698 }, { "epoch": 2.25, "learning_rate": 3.615719634223918e-07, "loss": 0.0268, "step": 5699 }, { "epoch": 2.25, "learning_rate": 3.6155691354290676e-07, "loss": 0.0225, "step": 5700 }, { "epoch": 2.25, "learning_rate": 3.615418610302896e-07, "loss": 0.0282, "step": 5701 }, { "epoch": 2.25, "learning_rate": 3.615268058847856e-07, "loss": 0.0243, "step": 5702 }, { "epoch": 2.25, "learning_rate": 3.6151174810664023e-07, "loss": 0.022, "step": 5703 }, { "epoch": 2.25, "learning_rate": 3.6149668769609877e-07, "loss": 0.0221, "step": 5704 }, { "epoch": 2.25, "learning_rate": 3.614816246534067e-07, "loss": 0.0258, "step": 5705 }, { "epoch": 2.25, "learning_rate": 3.614665589788097e-07, "loss": 0.0193, "step": 5706 }, { "epoch": 2.25, "learning_rate": 3.6145149067255306e-07, "loss": 0.0265, "step": 5707 }, { "epoch": 2.25, "learning_rate": 3.6143641973488263e-07, "loss": 0.0228, "step": 5708 }, { "epoch": 2.25, "learning_rate": 3.614213461660438e-07, "loss": 0.0222, "step": 5709 }, { "epoch": 2.25, "learning_rate": 3.614062699662824e-07, "loss": 0.0258, "step": 5710 }, { "epoch": 2.25, "learning_rate": 3.6139119113584414e-07, "loss": 0.0262, "step": 5711 }, { "epoch": 2.25, "learning_rate": 3.613761096749747e-07, "loss": 0.0285, "step": 5712 }, { "epoch": 2.25, "learning_rate": 3.6136102558391996e-07, "loss": 0.0233, "step": 5713 }, { "epoch": 2.25, "learning_rate": 3.6134593886292575e-07, "loss": 0.0226, "step": 5714 }, { "epoch": 2.25, "learning_rate": 3.613308495122379e-07, "loss": 0.0233, "step": 5715 }, { "epoch": 2.25, "learning_rate": 3.6131575753210247e-07, "loss": 0.0239, "step": 5716 }, { "epoch": 2.25, "learning_rate": 3.613006629227653e-07, "loss": 0.0197, "step": 5717 }, { "epoch": 2.25, "learning_rate": 3.612855656844726e-07, "loss": 0.0231, "step": 5718 }, { "epoch": 2.25, "learning_rate": 3.612704658174702e-07, "loss": 0.0228, "step": 5719 }, { "epoch": 2.25, "learning_rate": 3.612553633220043e-07, "loss": 0.0278, "step": 5720 }, { "epoch": 2.25, "learning_rate": 3.612402581983211e-07, "loss": 0.025, "step": 5721 }, { "epoch": 2.26, "learning_rate": 3.612251504466667e-07, "loss": 0.0235, "step": 5722 }, { "epoch": 2.26, "learning_rate": 3.6121004006728743e-07, "loss": 0.0238, "step": 5723 }, { "epoch": 2.26, "learning_rate": 3.6119492706042955e-07, "loss": 0.0209, "step": 5724 }, { "epoch": 2.26, "learning_rate": 3.611798114263393e-07, "loss": 0.0277, "step": 5725 }, { "epoch": 2.26, "learning_rate": 3.6116469316526303e-07, "loss": 0.0255, "step": 5726 }, { "epoch": 2.26, "learning_rate": 3.611495722774473e-07, "loss": 0.018, "step": 5727 }, { "epoch": 2.26, "learning_rate": 3.611344487631384e-07, "loss": 0.0225, "step": 5728 }, { "epoch": 2.26, "learning_rate": 3.611193226225829e-07, "loss": 0.0301, "step": 5729 }, { "epoch": 2.26, "learning_rate": 3.611041938560273e-07, "loss": 0.0246, "step": 5730 }, { "epoch": 2.26, "learning_rate": 3.6108906246371824e-07, "loss": 0.027, "step": 5731 }, { "epoch": 2.26, "learning_rate": 3.6107392844590223e-07, "loss": 0.0261, "step": 5732 }, { "epoch": 2.26, "learning_rate": 3.61058791802826e-07, "loss": 0.0284, "step": 5733 }, { "epoch": 2.26, "learning_rate": 3.6104365253473626e-07, "loss": 0.0292, "step": 5734 }, { "epoch": 2.26, "learning_rate": 3.6102851064187976e-07, "loss": 0.0233, "step": 5735 }, { "epoch": 2.26, "learning_rate": 3.6101336612450326e-07, "loss": 0.0212, "step": 5736 }, { "epoch": 2.26, "learning_rate": 3.6099821898285363e-07, "loss": 0.0252, "step": 5737 }, { "epoch": 2.26, "learning_rate": 3.6098306921717767e-07, "loss": 0.0219, "step": 5738 }, { "epoch": 2.26, "learning_rate": 3.609679168277224e-07, "loss": 0.0207, "step": 5739 }, { "epoch": 2.26, "learning_rate": 3.6095276181473465e-07, "loss": 0.0306, "step": 5740 }, { "epoch": 2.26, "learning_rate": 3.609376041784616e-07, "loss": 0.0216, "step": 5741 }, { "epoch": 2.26, "learning_rate": 3.6092244391915007e-07, "loss": 0.0221, "step": 5742 }, { "epoch": 2.26, "learning_rate": 3.6090728103704737e-07, "loss": 0.027, "step": 5743 }, { "epoch": 2.26, "learning_rate": 3.6089211553240054e-07, "loss": 0.0258, "step": 5744 }, { "epoch": 2.26, "learning_rate": 3.6087694740545675e-07, "loss": 0.0199, "step": 5745 }, { "epoch": 2.26, "learning_rate": 3.6086177665646324e-07, "loss": 0.0212, "step": 5746 }, { "epoch": 2.26, "learning_rate": 3.6084660328566717e-07, "loss": 0.0193, "step": 5747 }, { "epoch": 2.27, "learning_rate": 3.6083142729331605e-07, "loss": 0.0257, "step": 5748 }, { "epoch": 2.27, "learning_rate": 3.6081624867965703e-07, "loss": 0.0223, "step": 5749 }, { "epoch": 2.27, "learning_rate": 3.608010674449375e-07, "loss": 0.0243, "step": 5750 }, { "epoch": 2.27, "learning_rate": 3.607858835894051e-07, "loss": 0.0232, "step": 5751 }, { "epoch": 2.27, "learning_rate": 3.607706971133071e-07, "loss": 0.023, "step": 5752 }, { "epoch": 2.27, "learning_rate": 3.60755508016891e-07, "loss": 0.0276, "step": 5753 }, { "epoch": 2.27, "learning_rate": 3.6074031630040454e-07, "loss": 0.0267, "step": 5754 }, { "epoch": 2.27, "learning_rate": 3.607251219640952e-07, "loss": 0.0224, "step": 5755 }, { "epoch": 2.27, "learning_rate": 3.6070992500821066e-07, "loss": 0.0235, "step": 5756 }, { "epoch": 2.27, "learning_rate": 3.606947254329985e-07, "loss": 0.0218, "step": 5757 }, { "epoch": 2.27, "learning_rate": 3.606795232387067e-07, "loss": 0.0296, "step": 5758 }, { "epoch": 2.27, "learning_rate": 3.6066431842558275e-07, "loss": 0.0206, "step": 5759 }, { "epoch": 2.27, "learning_rate": 3.606491109938747e-07, "loss": 0.0224, "step": 5760 }, { "epoch": 2.27, "learning_rate": 3.6063390094383016e-07, "loss": 0.0234, "step": 5761 }, { "epoch": 2.27, "learning_rate": 3.606186882756973e-07, "loss": 0.0219, "step": 5762 }, { "epoch": 2.27, "learning_rate": 3.606034729897239e-07, "loss": 0.0258, "step": 5763 }, { "epoch": 2.27, "learning_rate": 3.605882550861579e-07, "loss": 0.0229, "step": 5764 }, { "epoch": 2.27, "learning_rate": 3.605730345652475e-07, "loss": 0.0242, "step": 5765 }, { "epoch": 2.27, "learning_rate": 3.605578114272406e-07, "loss": 0.0233, "step": 5766 }, { "epoch": 2.27, "learning_rate": 3.605425856723855e-07, "loss": 0.0219, "step": 5767 }, { "epoch": 2.27, "learning_rate": 3.6052735730093017e-07, "loss": 0.0258, "step": 5768 }, { "epoch": 2.27, "learning_rate": 3.6051212631312285e-07, "loss": 0.0241, "step": 5769 }, { "epoch": 2.27, "learning_rate": 3.604968927092119e-07, "loss": 0.0247, "step": 5770 }, { "epoch": 2.27, "learning_rate": 3.6048165648944553e-07, "loss": 0.0197, "step": 5771 }, { "epoch": 2.27, "learning_rate": 3.6046641765407204e-07, "loss": 0.0241, "step": 5772 }, { "epoch": 2.28, "learning_rate": 3.604511762033398e-07, "loss": 0.024, "step": 5773 }, { "epoch": 2.28, "learning_rate": 3.604359321374973e-07, "loss": 0.0259, "step": 5774 }, { "epoch": 2.28, "learning_rate": 3.604206854567929e-07, "loss": 0.022, "step": 5775 }, { "epoch": 2.28, "learning_rate": 3.604054361614751e-07, "loss": 0.0226, "step": 5776 }, { "epoch": 2.28, "learning_rate": 3.6039018425179255e-07, "loss": 0.0246, "step": 5777 }, { "epoch": 2.28, "learning_rate": 3.603749297279937e-07, "loss": 0.0205, "step": 5778 }, { "epoch": 2.28, "learning_rate": 3.603596725903273e-07, "loss": 0.0248, "step": 5779 }, { "epoch": 2.28, "learning_rate": 3.6034441283904196e-07, "loss": 0.0221, "step": 5780 }, { "epoch": 2.28, "learning_rate": 3.603291504743863e-07, "loss": 0.0209, "step": 5781 }, { "epoch": 2.28, "learning_rate": 3.6031388549660926e-07, "loss": 0.0266, "step": 5782 }, { "epoch": 2.28, "learning_rate": 3.602986179059595e-07, "loss": 0.0239, "step": 5783 }, { "epoch": 2.28, "learning_rate": 3.602833477026859e-07, "loss": 0.0256, "step": 5784 }, { "epoch": 2.28, "learning_rate": 3.602680748870373e-07, "loss": 0.0228, "step": 5785 }, { "epoch": 2.28, "learning_rate": 3.6025279945926264e-07, "loss": 0.0265, "step": 5786 }, { "epoch": 2.28, "learning_rate": 3.6023752141961097e-07, "loss": 0.0238, "step": 5787 }, { "epoch": 2.28, "learning_rate": 3.602222407683312e-07, "loss": 0.0237, "step": 5788 }, { "epoch": 2.28, "learning_rate": 3.602069575056724e-07, "loss": 0.0202, "step": 5789 }, { "epoch": 2.28, "learning_rate": 3.601916716318837e-07, "loss": 0.0201, "step": 5790 }, { "epoch": 2.28, "learning_rate": 3.6017638314721427e-07, "loss": 0.026, "step": 5791 }, { "epoch": 2.28, "learning_rate": 3.601610920519131e-07, "loss": 0.0235, "step": 5792 }, { "epoch": 2.28, "learning_rate": 3.6014579834622965e-07, "loss": 0.0265, "step": 5793 }, { "epoch": 2.28, "learning_rate": 3.6013050203041304e-07, "loss": 0.0272, "step": 5794 }, { "epoch": 2.28, "learning_rate": 3.601152031047127e-07, "loss": 0.0249, "step": 5795 }, { "epoch": 2.28, "learning_rate": 3.6009990156937784e-07, "loss": 0.0238, "step": 5796 }, { "epoch": 2.28, "learning_rate": 3.600845974246579e-07, "loss": 0.0285, "step": 5797 }, { "epoch": 2.29, "learning_rate": 3.600692906708023e-07, "loss": 0.027, "step": 5798 }, { "epoch": 2.29, "learning_rate": 3.600539813080606e-07, "loss": 0.023, "step": 5799 }, { "epoch": 2.29, "learning_rate": 3.600386693366822e-07, "loss": 0.0263, "step": 5800 }, { "epoch": 2.29, "learning_rate": 3.600233547569167e-07, "loss": 0.0242, "step": 5801 }, { "epoch": 2.29, "learning_rate": 3.600080375690138e-07, "loss": 0.0204, "step": 5802 }, { "epoch": 2.29, "learning_rate": 3.59992717773223e-07, "loss": 0.0295, "step": 5803 }, { "epoch": 2.29, "learning_rate": 3.5997739536979415e-07, "loss": 0.0236, "step": 5804 }, { "epoch": 2.29, "learning_rate": 3.5996207035897684e-07, "loss": 0.0267, "step": 5805 }, { "epoch": 2.29, "learning_rate": 3.5994674274102093e-07, "loss": 0.0227, "step": 5806 }, { "epoch": 2.29, "learning_rate": 3.599314125161761e-07, "loss": 0.0217, "step": 5807 }, { "epoch": 2.29, "learning_rate": 3.5991607968469244e-07, "loss": 0.0224, "step": 5808 }, { "epoch": 2.29, "learning_rate": 3.5990074424681967e-07, "loss": 0.025, "step": 5809 }, { "epoch": 2.29, "learning_rate": 3.598854062028078e-07, "loss": 0.0226, "step": 5810 }, { "epoch": 2.29, "learning_rate": 3.5987006555290677e-07, "loss": 0.0263, "step": 5811 }, { "epoch": 2.29, "learning_rate": 3.598547222973667e-07, "loss": 0.0217, "step": 5812 }, { "epoch": 2.29, "learning_rate": 3.598393764364376e-07, "loss": 0.0258, "step": 5813 }, { "epoch": 2.29, "learning_rate": 3.598240279703695e-07, "loss": 0.0272, "step": 5814 }, { "epoch": 2.29, "learning_rate": 3.598086768994128e-07, "loss": 0.0242, "step": 5815 }, { "epoch": 2.29, "learning_rate": 3.5979332322381745e-07, "loss": 0.0241, "step": 5816 }, { "epoch": 2.29, "learning_rate": 3.597779669438338e-07, "loss": 0.024, "step": 5817 }, { "epoch": 2.29, "learning_rate": 3.5976260805971214e-07, "loss": 0.0259, "step": 5818 }, { "epoch": 2.29, "learning_rate": 3.5974724657170276e-07, "loss": 0.0249, "step": 5819 }, { "epoch": 2.29, "learning_rate": 3.597318824800561e-07, "loss": 0.0223, "step": 5820 }, { "epoch": 2.29, "learning_rate": 3.5971651578502246e-07, "loss": 0.0228, "step": 5821 }, { "epoch": 2.29, "learning_rate": 3.5970114648685236e-07, "loss": 0.0288, "step": 5822 }, { "epoch": 2.29, "learning_rate": 3.596857745857963e-07, "loss": 0.0232, "step": 5823 }, { "epoch": 2.3, "learning_rate": 3.596704000821048e-07, "loss": 0.025, "step": 5824 }, { "epoch": 2.3, "learning_rate": 3.596550229760285e-07, "loss": 0.0264, "step": 5825 }, { "epoch": 2.3, "learning_rate": 3.59639643267818e-07, "loss": 0.0254, "step": 5826 }, { "epoch": 2.3, "learning_rate": 3.5962426095772383e-07, "loss": 0.025, "step": 5827 }, { "epoch": 2.3, "learning_rate": 3.596088760459969e-07, "loss": 0.023, "step": 5828 }, { "epoch": 2.3, "learning_rate": 3.5959348853288776e-07, "loss": 0.023, "step": 5829 }, { "epoch": 2.3, "learning_rate": 3.595780984186474e-07, "loss": 0.0297, "step": 5830 }, { "epoch": 2.3, "learning_rate": 3.5956270570352655e-07, "loss": 0.0219, "step": 5831 }, { "epoch": 2.3, "learning_rate": 3.595473103877761e-07, "loss": 0.0253, "step": 5832 }, { "epoch": 2.3, "learning_rate": 3.5953191247164694e-07, "loss": 0.0244, "step": 5833 }, { "epoch": 2.3, "learning_rate": 3.595165119553901e-07, "loss": 0.0245, "step": 5834 }, { "epoch": 2.3, "learning_rate": 3.5950110883925656e-07, "loss": 0.0245, "step": 5835 }, { "epoch": 2.3, "learning_rate": 3.594857031234973e-07, "loss": 0.0216, "step": 5836 }, { "epoch": 2.3, "learning_rate": 3.5947029480836355e-07, "loss": 0.0255, "step": 5837 }, { "epoch": 2.3, "learning_rate": 3.5945488389410636e-07, "loss": 0.0256, "step": 5838 }, { "epoch": 2.3, "learning_rate": 3.594394703809768e-07, "loss": 0.0248, "step": 5839 }, { "epoch": 2.3, "learning_rate": 3.594240542692263e-07, "loss": 0.0235, "step": 5840 }, { "epoch": 2.3, "learning_rate": 3.5940863555910605e-07, "loss": 0.0265, "step": 5841 }, { "epoch": 2.3, "learning_rate": 3.593932142508672e-07, "loss": 0.0231, "step": 5842 }, { "epoch": 2.3, "learning_rate": 3.5937779034476127e-07, "loss": 0.0251, "step": 5843 }, { "epoch": 2.3, "learning_rate": 3.5936236384103956e-07, "loss": 0.0255, "step": 5844 }, { "epoch": 2.3, "learning_rate": 3.593469347399535e-07, "loss": 0.0189, "step": 5845 }, { "epoch": 2.3, "learning_rate": 3.593315030417546e-07, "loss": 0.0215, "step": 5846 }, { "epoch": 2.3, "learning_rate": 3.5931606874669437e-07, "loss": 0.0246, "step": 5847 }, { "epoch": 2.3, "learning_rate": 3.593006318550244e-07, "loss": 0.0253, "step": 5848 }, { "epoch": 2.31, "learning_rate": 3.5928519236699615e-07, "loss": 0.0221, "step": 5849 }, { "epoch": 2.31, "learning_rate": 3.592697502828614e-07, "loss": 0.0219, "step": 5850 }, { "epoch": 2.31, "learning_rate": 3.5925430560287177e-07, "loss": 0.0253, "step": 5851 }, { "epoch": 2.31, "learning_rate": 3.59238858327279e-07, "loss": 0.0283, "step": 5852 }, { "epoch": 2.31, "learning_rate": 3.5922340845633486e-07, "loss": 0.0242, "step": 5853 }, { "epoch": 2.31, "learning_rate": 3.5920795599029114e-07, "loss": 0.0218, "step": 5854 }, { "epoch": 2.31, "learning_rate": 3.5919250092939975e-07, "loss": 0.0237, "step": 5855 }, { "epoch": 2.31, "learning_rate": 3.591770432739125e-07, "loss": 0.0259, "step": 5856 }, { "epoch": 2.31, "learning_rate": 3.591615830240813e-07, "loss": 0.0197, "step": 5857 }, { "epoch": 2.31, "learning_rate": 3.591461201801583e-07, "loss": 0.0226, "step": 5858 }, { "epoch": 2.31, "learning_rate": 3.5913065474239537e-07, "loss": 0.0208, "step": 5859 }, { "epoch": 2.31, "learning_rate": 3.591151867110446e-07, "loss": 0.0214, "step": 5860 }, { "epoch": 2.31, "learning_rate": 3.5909971608635814e-07, "loss": 0.0227, "step": 5861 }, { "epoch": 2.31, "learning_rate": 3.590842428685882e-07, "loss": 0.021, "step": 5862 }, { "epoch": 2.31, "learning_rate": 3.5906876705798674e-07, "loss": 0.0199, "step": 5863 }, { "epoch": 2.31, "learning_rate": 3.590532886548062e-07, "loss": 0.0257, "step": 5864 }, { "epoch": 2.31, "learning_rate": 3.5903780765929874e-07, "loss": 0.0271, "step": 5865 }, { "epoch": 2.31, "learning_rate": 3.5902232407171675e-07, "loss": 0.0235, "step": 5866 }, { "epoch": 2.31, "learning_rate": 3.590068378923126e-07, "loss": 0.0232, "step": 5867 }, { "epoch": 2.31, "learning_rate": 3.589913491213386e-07, "loss": 0.0229, "step": 5868 }, { "epoch": 2.31, "learning_rate": 3.5897585775904736e-07, "loss": 0.024, "step": 5869 }, { "epoch": 2.31, "learning_rate": 3.5896036380569115e-07, "loss": 0.0251, "step": 5870 }, { "epoch": 2.31, "learning_rate": 3.589448672615226e-07, "loss": 0.0264, "step": 5871 }, { "epoch": 2.31, "learning_rate": 3.589293681267943e-07, "loss": 0.0224, "step": 5872 }, { "epoch": 2.31, "learning_rate": 3.5891386640175887e-07, "loss": 0.0238, "step": 5873 }, { "epoch": 2.31, "learning_rate": 3.5889836208666895e-07, "loss": 0.0253, "step": 5874 }, { "epoch": 2.32, "learning_rate": 3.588828551817771e-07, "loss": 0.0242, "step": 5875 }, { "epoch": 2.32, "learning_rate": 3.588673456873363e-07, "loss": 0.0254, "step": 5876 }, { "epoch": 2.32, "learning_rate": 3.588518336035992e-07, "loss": 0.0217, "step": 5877 }, { "epoch": 2.32, "learning_rate": 3.5883631893081863e-07, "loss": 0.0274, "step": 5878 }, { "epoch": 2.32, "learning_rate": 3.588208016692475e-07, "loss": 0.023, "step": 5879 }, { "epoch": 2.32, "learning_rate": 3.5880528181913867e-07, "loss": 0.0217, "step": 5880 }, { "epoch": 2.32, "learning_rate": 3.587897593807451e-07, "loss": 0.026, "step": 5881 }, { "epoch": 2.32, "learning_rate": 3.5877423435431974e-07, "loss": 0.0235, "step": 5882 }, { "epoch": 2.32, "learning_rate": 3.5875870674011567e-07, "loss": 0.0233, "step": 5883 }, { "epoch": 2.32, "learning_rate": 3.5874317653838603e-07, "loss": 0.0229, "step": 5884 }, { "epoch": 2.32, "learning_rate": 3.587276437493838e-07, "loss": 0.0236, "step": 5885 }, { "epoch": 2.32, "learning_rate": 3.587121083733622e-07, "loss": 0.0248, "step": 5886 }, { "epoch": 2.32, "learning_rate": 3.5869657041057455e-07, "loss": 0.0207, "step": 5887 }, { "epoch": 2.32, "learning_rate": 3.5868102986127396e-07, "loss": 0.025, "step": 5888 }, { "epoch": 2.32, "learning_rate": 3.5866548672571375e-07, "loss": 0.0213, "step": 5889 }, { "epoch": 2.32, "learning_rate": 3.5864994100414723e-07, "loss": 0.0199, "step": 5890 }, { "epoch": 2.32, "learning_rate": 3.586343926968278e-07, "loss": 0.0277, "step": 5891 }, { "epoch": 2.32, "learning_rate": 3.5861884180400886e-07, "loss": 0.0255, "step": 5892 }, { "epoch": 2.32, "learning_rate": 3.5860328832594385e-07, "loss": 0.0268, "step": 5893 }, { "epoch": 2.32, "learning_rate": 3.585877322628863e-07, "loss": 0.0225, "step": 5894 }, { "epoch": 2.32, "learning_rate": 3.5857217361508976e-07, "loss": 0.0247, "step": 5895 }, { "epoch": 2.32, "learning_rate": 3.585566123828078e-07, "loss": 0.0211, "step": 5896 }, { "epoch": 2.32, "learning_rate": 3.58541048566294e-07, "loss": 0.026, "step": 5897 }, { "epoch": 2.32, "learning_rate": 3.585254821658021e-07, "loss": 0.0262, "step": 5898 }, { "epoch": 2.32, "learning_rate": 3.585099131815857e-07, "loss": 0.0256, "step": 5899 }, { "epoch": 2.33, "learning_rate": 3.584943416138987e-07, "loss": 0.0239, "step": 5900 }, { "epoch": 2.33, "learning_rate": 3.5847876746299487e-07, "loss": 0.0252, "step": 5901 }, { "epoch": 2.33, "learning_rate": 3.584631907291279e-07, "loss": 0.0212, "step": 5902 }, { "epoch": 2.33, "learning_rate": 3.5844761141255174e-07, "loss": 0.0291, "step": 5903 }, { "epoch": 2.33, "learning_rate": 3.5843202951352036e-07, "loss": 0.0212, "step": 5904 }, { "epoch": 2.33, "learning_rate": 3.584164450322877e-07, "loss": 0.0177, "step": 5905 }, { "epoch": 2.33, "learning_rate": 3.584008579691077e-07, "loss": 0.0222, "step": 5906 }, { "epoch": 2.33, "learning_rate": 3.5838526832423453e-07, "loss": 0.0198, "step": 5907 }, { "epoch": 2.33, "learning_rate": 3.5836967609792213e-07, "loss": 0.0215, "step": 5908 }, { "epoch": 2.33, "learning_rate": 3.583540812904248e-07, "loss": 0.0239, "step": 5909 }, { "epoch": 2.33, "learning_rate": 3.583384839019965e-07, "loss": 0.0193, "step": 5910 }, { "epoch": 2.33, "learning_rate": 3.5832288393289167e-07, "loss": 0.0258, "step": 5911 }, { "epoch": 2.33, "learning_rate": 3.583072813833643e-07, "loss": 0.023, "step": 5912 }, { "epoch": 2.33, "learning_rate": 3.5829167625366906e-07, "loss": 0.022, "step": 5913 }, { "epoch": 2.33, "learning_rate": 3.5827606854405994e-07, "loss": 0.0262, "step": 5914 }, { "epoch": 2.33, "learning_rate": 3.5826045825479143e-07, "loss": 0.0217, "step": 5915 }, { "epoch": 2.33, "learning_rate": 3.58244845386118e-07, "loss": 0.0248, "step": 5916 }, { "epoch": 2.33, "learning_rate": 3.5822922993829414e-07, "loss": 0.0238, "step": 5917 }, { "epoch": 2.33, "learning_rate": 3.582136119115743e-07, "loss": 0.0191, "step": 5918 }, { "epoch": 2.33, "learning_rate": 3.5819799130621303e-07, "loss": 0.0235, "step": 5919 }, { "epoch": 2.33, "learning_rate": 3.581823681224649e-07, "loss": 0.0202, "step": 5920 }, { "epoch": 2.33, "learning_rate": 3.5816674236058465e-07, "loss": 0.0234, "step": 5921 }, { "epoch": 2.33, "learning_rate": 3.581511140208269e-07, "loss": 0.023, "step": 5922 }, { "epoch": 2.33, "learning_rate": 3.5813548310344626e-07, "loss": 0.0226, "step": 5923 }, { "epoch": 2.33, "learning_rate": 3.5811984960869764e-07, "loss": 0.0227, "step": 5924 }, { "epoch": 2.34, "learning_rate": 3.581042135368357e-07, "loss": 0.0254, "step": 5925 }, { "epoch": 2.34, "learning_rate": 3.5808857488811544e-07, "loss": 0.0267, "step": 5926 }, { "epoch": 2.34, "learning_rate": 3.580729336627917e-07, "loss": 0.0245, "step": 5927 }, { "epoch": 2.34, "learning_rate": 3.580572898611194e-07, "loss": 0.0267, "step": 5928 }, { "epoch": 2.34, "learning_rate": 3.5804164348335345e-07, "loss": 0.0241, "step": 5929 }, { "epoch": 2.34, "learning_rate": 3.5802599452974887e-07, "loss": 0.0217, "step": 5930 }, { "epoch": 2.34, "learning_rate": 3.5801034300056083e-07, "loss": 0.0224, "step": 5931 }, { "epoch": 2.34, "learning_rate": 3.5799468889604434e-07, "loss": 0.0226, "step": 5932 }, { "epoch": 2.34, "learning_rate": 3.5797903221645447e-07, "loss": 0.0198, "step": 5933 }, { "epoch": 2.34, "learning_rate": 3.5796337296204654e-07, "loss": 0.0249, "step": 5934 }, { "epoch": 2.34, "learning_rate": 3.579477111330757e-07, "loss": 0.0251, "step": 5935 }, { "epoch": 2.34, "learning_rate": 3.579320467297972e-07, "loss": 0.0238, "step": 5936 }, { "epoch": 2.34, "learning_rate": 3.579163797524664e-07, "loss": 0.0213, "step": 5937 }, { "epoch": 2.34, "learning_rate": 3.579007102013386e-07, "loss": 0.0217, "step": 5938 }, { "epoch": 2.34, "learning_rate": 3.5788503807666917e-07, "loss": 0.0242, "step": 5939 }, { "epoch": 2.34, "learning_rate": 3.578693633787136e-07, "loss": 0.0238, "step": 5940 }, { "epoch": 2.34, "learning_rate": 3.578536861077274e-07, "loss": 0.023, "step": 5941 }, { "epoch": 2.34, "learning_rate": 3.5783800626396596e-07, "loss": 0.0224, "step": 5942 }, { "epoch": 2.34, "learning_rate": 3.578223238476849e-07, "loss": 0.0259, "step": 5943 }, { "epoch": 2.34, "learning_rate": 3.578066388591399e-07, "loss": 0.0228, "step": 5944 }, { "epoch": 2.34, "learning_rate": 3.577909512985864e-07, "loss": 0.022, "step": 5945 }, { "epoch": 2.34, "learning_rate": 3.5777526116628033e-07, "loss": 0.0235, "step": 5946 }, { "epoch": 2.34, "learning_rate": 3.577595684624772e-07, "loss": 0.0242, "step": 5947 }, { "epoch": 2.34, "learning_rate": 3.5774387318743287e-07, "loss": 0.0241, "step": 5948 }, { "epoch": 2.34, "learning_rate": 3.5772817534140326e-07, "loss": 0.0233, "step": 5949 }, { "epoch": 2.34, "learning_rate": 3.57712474924644e-07, "loss": 0.024, "step": 5950 }, { "epoch": 2.35, "learning_rate": 3.5769677193741117e-07, "loss": 0.024, "step": 5951 }, { "epoch": 2.35, "learning_rate": 3.576810663799605e-07, "loss": 0.0233, "step": 5952 }, { "epoch": 2.35, "learning_rate": 3.576653582525482e-07, "loss": 0.0246, "step": 5953 }, { "epoch": 2.35, "learning_rate": 3.5764964755543023e-07, "loss": 0.02, "step": 5954 }, { "epoch": 2.35, "learning_rate": 3.576339342888625e-07, "loss": 0.0194, "step": 5955 }, { "epoch": 2.35, "learning_rate": 3.576182184531013e-07, "loss": 0.0253, "step": 5956 }, { "epoch": 2.35, "learning_rate": 3.5760250004840263e-07, "loss": 0.0197, "step": 5957 }, { "epoch": 2.35, "learning_rate": 3.575867790750228e-07, "loss": 0.0195, "step": 5958 }, { "epoch": 2.35, "learning_rate": 3.575710555332179e-07, "loss": 0.0214, "step": 5959 }, { "epoch": 2.35, "learning_rate": 3.575553294232444e-07, "loss": 0.0213, "step": 5960 }, { "epoch": 2.35, "learning_rate": 3.575396007453584e-07, "loss": 0.0235, "step": 5961 }, { "epoch": 2.35, "learning_rate": 3.575238694998163e-07, "loss": 0.0264, "step": 5962 }, { "epoch": 2.35, "learning_rate": 3.5750813568687457e-07, "loss": 0.0218, "step": 5963 }, { "epoch": 2.35, "learning_rate": 3.5749239930678964e-07, "loss": 0.0252, "step": 5964 }, { "epoch": 2.35, "learning_rate": 3.574766603598179e-07, "loss": 0.0235, "step": 5965 }, { "epoch": 2.35, "learning_rate": 3.57460918846216e-07, "loss": 0.0206, "step": 5966 }, { "epoch": 2.35, "learning_rate": 3.5744517476624043e-07, "loss": 0.0224, "step": 5967 }, { "epoch": 2.35, "learning_rate": 3.574294281201478e-07, "loss": 0.0263, "step": 5968 }, { "epoch": 2.35, "learning_rate": 3.5741367890819473e-07, "loss": 0.0228, "step": 5969 }, { "epoch": 2.35, "learning_rate": 3.5739792713063795e-07, "loss": 0.0328, "step": 5970 }, { "epoch": 2.35, "learning_rate": 3.5738217278773416e-07, "loss": 0.0217, "step": 5971 }, { "epoch": 2.35, "learning_rate": 3.5736641587974017e-07, "loss": 0.0255, "step": 5972 }, { "epoch": 2.35, "learning_rate": 3.573506564069127e-07, "loss": 0.0302, "step": 5973 }, { "epoch": 2.35, "learning_rate": 3.5733489436950876e-07, "loss": 0.0262, "step": 5974 }, { "epoch": 2.35, "learning_rate": 3.5731912976778517e-07, "loss": 0.0211, "step": 5975 }, { "epoch": 2.36, "learning_rate": 3.573033626019988e-07, "loss": 0.0202, "step": 5976 }, { "epoch": 2.36, "learning_rate": 3.572875928724067e-07, "loss": 0.0201, "step": 5977 }, { "epoch": 2.36, "learning_rate": 3.5727182057926593e-07, "loss": 0.0193, "step": 5978 }, { "epoch": 2.36, "learning_rate": 3.5725604572283347e-07, "loss": 0.0217, "step": 5979 }, { "epoch": 2.36, "learning_rate": 3.572402683033665e-07, "loss": 0.0216, "step": 5980 }, { "epoch": 2.36, "learning_rate": 3.572244883211221e-07, "loss": 0.0209, "step": 5981 }, { "epoch": 2.36, "learning_rate": 3.5720870577635755e-07, "loss": 0.0245, "step": 5982 }, { "epoch": 2.36, "learning_rate": 3.5719292066933e-07, "loss": 0.0254, "step": 5983 }, { "epoch": 2.36, "learning_rate": 3.5717713300029673e-07, "loss": 0.0224, "step": 5984 }, { "epoch": 2.36, "learning_rate": 3.5716134276951503e-07, "loss": 0.0228, "step": 5985 }, { "epoch": 2.36, "learning_rate": 3.5714554997724234e-07, "loss": 0.0259, "step": 5986 }, { "epoch": 2.36, "learning_rate": 3.5712975462373603e-07, "loss": 0.02, "step": 5987 }, { "epoch": 2.36, "learning_rate": 3.571139567092536e-07, "loss": 0.0236, "step": 5988 }, { "epoch": 2.36, "learning_rate": 3.570981562340523e-07, "loss": 0.0201, "step": 5989 }, { "epoch": 2.36, "learning_rate": 3.570823531983899e-07, "loss": 0.0202, "step": 5990 }, { "epoch": 2.36, "learning_rate": 3.5706654760252394e-07, "loss": 0.0196, "step": 5991 }, { "epoch": 2.36, "learning_rate": 3.5705073944671196e-07, "loss": 0.0227, "step": 5992 }, { "epoch": 2.36, "learning_rate": 3.5703492873121157e-07, "loss": 0.0244, "step": 5993 }, { "epoch": 2.36, "learning_rate": 3.5701911545628047e-07, "loss": 0.025, "step": 5994 }, { "epoch": 2.36, "learning_rate": 3.570032996221765e-07, "loss": 0.022, "step": 5995 }, { "epoch": 2.36, "learning_rate": 3.569874812291573e-07, "loss": 0.0189, "step": 5996 }, { "epoch": 2.36, "learning_rate": 3.5697166027748073e-07, "loss": 0.0233, "step": 5997 }, { "epoch": 2.36, "learning_rate": 3.5695583676740473e-07, "loss": 0.0251, "step": 5998 }, { "epoch": 2.36, "learning_rate": 3.5694001069918716e-07, "loss": 0.0213, "step": 5999 }, { "epoch": 2.36, "learning_rate": 3.569241820730859e-07, "loss": 0.0249, "step": 6000 }, { "epoch": 2.37, "learning_rate": 3.5690835088935895e-07, "loss": 0.0234, "step": 6001 }, { "epoch": 2.37, "learning_rate": 3.568925171482644e-07, "loss": 0.0224, "step": 6002 }, { "epoch": 2.37, "learning_rate": 3.568766808500602e-07, "loss": 0.0259, "step": 6003 }, { "epoch": 2.37, "learning_rate": 3.5686084199500464e-07, "loss": 0.0191, "step": 6004 }, { "epoch": 2.37, "learning_rate": 3.5684500058335566e-07, "loss": 0.0176, "step": 6005 }, { "epoch": 2.37, "learning_rate": 3.5682915661537156e-07, "loss": 0.0233, "step": 6006 }, { "epoch": 2.37, "learning_rate": 3.568133100913106e-07, "loss": 0.0233, "step": 6007 }, { "epoch": 2.37, "learning_rate": 3.5679746101143094e-07, "loss": 0.0261, "step": 6008 }, { "epoch": 2.37, "learning_rate": 3.5678160937599105e-07, "loss": 0.0209, "step": 6009 }, { "epoch": 2.37, "learning_rate": 3.567657551852492e-07, "loss": 0.0215, "step": 6010 }, { "epoch": 2.37, "learning_rate": 3.567498984394638e-07, "loss": 0.0197, "step": 6011 }, { "epoch": 2.37, "learning_rate": 3.567340391388932e-07, "loss": 0.0214, "step": 6012 }, { "epoch": 2.37, "learning_rate": 3.5671817728379604e-07, "loss": 0.0252, "step": 6013 }, { "epoch": 2.37, "learning_rate": 3.567023128744308e-07, "loss": 0.0272, "step": 6014 }, { "epoch": 2.37, "learning_rate": 3.5668644591105595e-07, "loss": 0.0253, "step": 6015 }, { "epoch": 2.37, "learning_rate": 3.566705763939302e-07, "loss": 0.0216, "step": 6016 }, { "epoch": 2.37, "learning_rate": 3.566547043233122e-07, "loss": 0.0254, "step": 6017 }, { "epoch": 2.37, "learning_rate": 3.566388296994606e-07, "loss": 0.0258, "step": 6018 }, { "epoch": 2.37, "learning_rate": 3.5662295252263403e-07, "loss": 0.027, "step": 6019 }, { "epoch": 2.37, "learning_rate": 3.5660707279309154e-07, "loss": 0.0211, "step": 6020 }, { "epoch": 2.37, "learning_rate": 3.5659119051109165e-07, "loss": 0.0218, "step": 6021 }, { "epoch": 2.37, "learning_rate": 3.565753056768934e-07, "loss": 0.024, "step": 6022 }, { "epoch": 2.37, "learning_rate": 3.565594182907556e-07, "loss": 0.0206, "step": 6023 }, { "epoch": 2.37, "learning_rate": 3.565435283529373e-07, "loss": 0.0236, "step": 6024 }, { "epoch": 2.37, "learning_rate": 3.5652763586369734e-07, "loss": 0.0252, "step": 6025 }, { "epoch": 2.37, "learning_rate": 3.5651174082329487e-07, "loss": 0.0261, "step": 6026 }, { "epoch": 2.38, "learning_rate": 3.564958432319888e-07, "loss": 0.0223, "step": 6027 }, { "epoch": 2.38, "learning_rate": 3.564799430900384e-07, "loss": 0.0204, "step": 6028 }, { "epoch": 2.38, "learning_rate": 3.5646404039770277e-07, "loss": 0.0213, "step": 6029 }, { "epoch": 2.38, "learning_rate": 3.56448135155241e-07, "loss": 0.0243, "step": 6030 }, { "epoch": 2.38, "learning_rate": 3.5643222736291254e-07, "loss": 0.0197, "step": 6031 }, { "epoch": 2.38, "learning_rate": 3.564163170209764e-07, "loss": 0.0194, "step": 6032 }, { "epoch": 2.38, "learning_rate": 3.56400404129692e-07, "loss": 0.0221, "step": 6033 }, { "epoch": 2.38, "learning_rate": 3.563844886893188e-07, "loss": 0.0237, "step": 6034 }, { "epoch": 2.38, "learning_rate": 3.563685707001161e-07, "loss": 0.0242, "step": 6035 }, { "epoch": 2.38, "learning_rate": 3.563526501623433e-07, "loss": 0.0234, "step": 6036 }, { "epoch": 2.38, "learning_rate": 3.563367270762599e-07, "loss": 0.025, "step": 6037 }, { "epoch": 2.38, "learning_rate": 3.5632080144212557e-07, "loss": 0.023, "step": 6038 }, { "epoch": 2.38, "learning_rate": 3.563048732601997e-07, "loss": 0.0251, "step": 6039 }, { "epoch": 2.38, "learning_rate": 3.5628894253074195e-07, "loss": 0.0252, "step": 6040 }, { "epoch": 2.38, "learning_rate": 3.5627300925401196e-07, "loss": 0.0256, "step": 6041 }, { "epoch": 2.38, "learning_rate": 3.562570734302694e-07, "loss": 0.0242, "step": 6042 }, { "epoch": 2.38, "learning_rate": 3.5624113505977406e-07, "loss": 0.0213, "step": 6043 }, { "epoch": 2.38, "learning_rate": 3.562251941427857e-07, "loss": 0.0278, "step": 6044 }, { "epoch": 2.38, "learning_rate": 3.5620925067956407e-07, "loss": 0.0212, "step": 6045 }, { "epoch": 2.38, "learning_rate": 3.561933046703691e-07, "loss": 0.025, "step": 6046 }, { "epoch": 2.38, "learning_rate": 3.5617735611546064e-07, "loss": 0.0217, "step": 6047 }, { "epoch": 2.38, "learning_rate": 3.5616140501509857e-07, "loss": 0.0212, "step": 6048 }, { "epoch": 2.38, "learning_rate": 3.5614545136954304e-07, "loss": 0.0228, "step": 6049 }, { "epoch": 2.38, "learning_rate": 3.5612949517905395e-07, "loss": 0.0225, "step": 6050 }, { "epoch": 2.38, "learning_rate": 3.561135364438913e-07, "loss": 0.0177, "step": 6051 }, { "epoch": 2.39, "learning_rate": 3.560975751643154e-07, "loss": 0.0229, "step": 6052 }, { "epoch": 2.39, "learning_rate": 3.5608161134058614e-07, "loss": 0.0215, "step": 6053 }, { "epoch": 2.39, "learning_rate": 3.560656449729639e-07, "loss": 0.0242, "step": 6054 }, { "epoch": 2.39, "learning_rate": 3.560496760617088e-07, "loss": 0.0246, "step": 6055 }, { "epoch": 2.39, "learning_rate": 3.5603370460708115e-07, "loss": 0.0213, "step": 6056 }, { "epoch": 2.39, "learning_rate": 3.560177306093413e-07, "loss": 0.0222, "step": 6057 }, { "epoch": 2.39, "learning_rate": 3.560017540687495e-07, "loss": 0.0214, "step": 6058 }, { "epoch": 2.39, "learning_rate": 3.5598577498556627e-07, "loss": 0.0239, "step": 6059 }, { "epoch": 2.39, "learning_rate": 3.559697933600519e-07, "loss": 0.0226, "step": 6060 }, { "epoch": 2.39, "learning_rate": 3.55953809192467e-07, "loss": 0.0248, "step": 6061 }, { "epoch": 2.39, "learning_rate": 3.559378224830721e-07, "loss": 0.0278, "step": 6062 }, { "epoch": 2.39, "learning_rate": 3.559218332321276e-07, "loss": 0.0243, "step": 6063 }, { "epoch": 2.39, "learning_rate": 3.559058414398942e-07, "loss": 0.022, "step": 6064 }, { "epoch": 2.39, "learning_rate": 3.558898471066325e-07, "loss": 0.0227, "step": 6065 }, { "epoch": 2.39, "learning_rate": 3.558738502326033e-07, "loss": 0.0222, "step": 6066 }, { "epoch": 2.39, "learning_rate": 3.558578508180672e-07, "loss": 0.0205, "step": 6067 }, { "epoch": 2.39, "learning_rate": 3.55841848863285e-07, "loss": 0.0227, "step": 6068 }, { "epoch": 2.39, "learning_rate": 3.558258443685176e-07, "loss": 0.0217, "step": 6069 }, { "epoch": 2.39, "learning_rate": 3.5580983733402564e-07, "loss": 0.0275, "step": 6070 }, { "epoch": 2.39, "learning_rate": 3.5579382776007024e-07, "loss": 0.0248, "step": 6071 }, { "epoch": 2.39, "learning_rate": 3.5577781564691214e-07, "loss": 0.0181, "step": 6072 }, { "epoch": 2.39, "learning_rate": 3.557618009948125e-07, "loss": 0.0227, "step": 6073 }, { "epoch": 2.39, "learning_rate": 3.557457838040321e-07, "loss": 0.023, "step": 6074 }, { "epoch": 2.39, "learning_rate": 3.557297640748322e-07, "loss": 0.0191, "step": 6075 }, { "epoch": 2.39, "learning_rate": 3.5571374180747383e-07, "loss": 0.0211, "step": 6076 }, { "epoch": 2.39, "learning_rate": 3.5569771700221813e-07, "loss": 0.0237, "step": 6077 }, { "epoch": 2.4, "learning_rate": 3.556816896593263e-07, "loss": 0.0213, "step": 6078 }, { "epoch": 2.4, "learning_rate": 3.556656597790595e-07, "loss": 0.0232, "step": 6079 }, { "epoch": 2.4, "learning_rate": 3.5564962736167905e-07, "loss": 0.0223, "step": 6080 }, { "epoch": 2.4, "learning_rate": 3.556335924074462e-07, "loss": 0.022, "step": 6081 }, { "epoch": 2.4, "learning_rate": 3.5561755491662234e-07, "loss": 0.0253, "step": 6082 }, { "epoch": 2.4, "learning_rate": 3.556015148894689e-07, "loss": 0.0231, "step": 6083 }, { "epoch": 2.4, "learning_rate": 3.555854723262472e-07, "loss": 0.0178, "step": 6084 }, { "epoch": 2.4, "learning_rate": 3.5556942722721875e-07, "loss": 0.0242, "step": 6085 }, { "epoch": 2.4, "learning_rate": 3.5555337959264514e-07, "loss": 0.0189, "step": 6086 }, { "epoch": 2.4, "learning_rate": 3.555373294227878e-07, "loss": 0.023, "step": 6087 }, { "epoch": 2.4, "learning_rate": 3.5552127671790835e-07, "loss": 0.0288, "step": 6088 }, { "epoch": 2.4, "learning_rate": 3.5550522147826854e-07, "loss": 0.0243, "step": 6089 }, { "epoch": 2.4, "learning_rate": 3.554891637041299e-07, "loss": 0.0193, "step": 6090 }, { "epoch": 2.4, "learning_rate": 3.5547310339575425e-07, "loss": 0.0246, "step": 6091 }, { "epoch": 2.4, "learning_rate": 3.554570405534033e-07, "loss": 0.0277, "step": 6092 }, { "epoch": 2.4, "learning_rate": 3.5544097517733884e-07, "loss": 0.0203, "step": 6093 }, { "epoch": 2.4, "learning_rate": 3.554249072678227e-07, "loss": 0.0254, "step": 6094 }, { "epoch": 2.4, "learning_rate": 3.554088368251168e-07, "loss": 0.0217, "step": 6095 }, { "epoch": 2.4, "learning_rate": 3.5539276384948307e-07, "loss": 0.0266, "step": 6096 }, { "epoch": 2.4, "learning_rate": 3.553766883411835e-07, "loss": 0.0251, "step": 6097 }, { "epoch": 2.4, "learning_rate": 3.5536061030048003e-07, "loss": 0.0238, "step": 6098 }, { "epoch": 2.4, "learning_rate": 3.5534452972763473e-07, "loss": 0.0196, "step": 6099 }, { "epoch": 2.4, "learning_rate": 3.5532844662290964e-07, "loss": 0.0219, "step": 6100 }, { "epoch": 2.4, "learning_rate": 3.55312360986567e-07, "loss": 0.0226, "step": 6101 }, { "epoch": 2.4, "learning_rate": 3.55296272818869e-07, "loss": 0.0221, "step": 6102 }, { "epoch": 2.41, "learning_rate": 3.5528018212007763e-07, "loss": 0.0256, "step": 6103 }, { "epoch": 2.41, "learning_rate": 3.552640888904554e-07, "loss": 0.0193, "step": 6104 }, { "epoch": 2.41, "learning_rate": 3.552479931302644e-07, "loss": 0.0225, "step": 6105 }, { "epoch": 2.41, "learning_rate": 3.552318948397671e-07, "loss": 0.0232, "step": 6106 }, { "epoch": 2.41, "learning_rate": 3.552157940192258e-07, "loss": 0.024, "step": 6107 }, { "epoch": 2.41, "learning_rate": 3.55199690668903e-07, "loss": 0.0257, "step": 6108 }, { "epoch": 2.41, "learning_rate": 3.551835847890611e-07, "loss": 0.0272, "step": 6109 }, { "epoch": 2.41, "learning_rate": 3.551674763799627e-07, "loss": 0.0185, "step": 6110 }, { "epoch": 2.41, "learning_rate": 3.551513654418702e-07, "loss": 0.0246, "step": 6111 }, { "epoch": 2.41, "learning_rate": 3.551352519750463e-07, "loss": 0.0291, "step": 6112 }, { "epoch": 2.41, "learning_rate": 3.551191359797535e-07, "loss": 0.0245, "step": 6113 }, { "epoch": 2.41, "learning_rate": 3.551030174562545e-07, "loss": 0.0257, "step": 6114 }, { "epoch": 2.41, "learning_rate": 3.550868964048121e-07, "loss": 0.0254, "step": 6115 }, { "epoch": 2.41, "learning_rate": 3.5507077282568905e-07, "loss": 0.0217, "step": 6116 }, { "epoch": 2.41, "learning_rate": 3.55054646719148e-07, "loss": 0.0193, "step": 6117 }, { "epoch": 2.41, "learning_rate": 3.550385180854519e-07, "loss": 0.0257, "step": 6118 }, { "epoch": 2.41, "learning_rate": 3.5502238692486356e-07, "loss": 0.02, "step": 6119 }, { "epoch": 2.41, "learning_rate": 3.5500625323764596e-07, "loss": 0.0232, "step": 6120 }, { "epoch": 2.41, "learning_rate": 3.5499011702406196e-07, "loss": 0.0245, "step": 6121 }, { "epoch": 2.41, "learning_rate": 3.549739782843747e-07, "loss": 0.0249, "step": 6122 }, { "epoch": 2.41, "learning_rate": 3.5495783701884705e-07, "loss": 0.0268, "step": 6123 }, { "epoch": 2.41, "learning_rate": 3.5494169322774215e-07, "loss": 0.0237, "step": 6124 }, { "epoch": 2.41, "learning_rate": 3.5492554691132323e-07, "loss": 0.0242, "step": 6125 }, { "epoch": 2.41, "learning_rate": 3.5490939806985325e-07, "loss": 0.0225, "step": 6126 }, { "epoch": 2.41, "learning_rate": 3.548932467035956e-07, "loss": 0.0251, "step": 6127 }, { "epoch": 2.42, "learning_rate": 3.5487709281281334e-07, "loss": 0.0196, "step": 6128 }, { "epoch": 2.42, "learning_rate": 3.5486093639776995e-07, "loss": 0.0217, "step": 6129 }, { "epoch": 2.42, "learning_rate": 3.548447774587286e-07, "loss": 0.0233, "step": 6130 }, { "epoch": 2.42, "learning_rate": 3.548286159959528e-07, "loss": 0.0233, "step": 6131 }, { "epoch": 2.42, "learning_rate": 3.5481245200970577e-07, "loss": 0.0233, "step": 6132 }, { "epoch": 2.42, "learning_rate": 3.5479628550025114e-07, "loss": 0.019, "step": 6133 }, { "epoch": 2.42, "learning_rate": 3.5478011646785224e-07, "loss": 0.0191, "step": 6134 }, { "epoch": 2.42, "learning_rate": 3.547639449127727e-07, "loss": 0.0175, "step": 6135 }, { "epoch": 2.42, "learning_rate": 3.547477708352761e-07, "loss": 0.0251, "step": 6136 }, { "epoch": 2.42, "learning_rate": 3.5473159423562605e-07, "loss": 0.0236, "step": 6137 }, { "epoch": 2.42, "learning_rate": 3.5471541511408616e-07, "loss": 0.0217, "step": 6138 }, { "epoch": 2.42, "learning_rate": 3.546992334709201e-07, "loss": 0.025, "step": 6139 }, { "epoch": 2.42, "learning_rate": 3.5468304930639166e-07, "loss": 0.0257, "step": 6140 }, { "epoch": 2.42, "learning_rate": 3.546668626207647e-07, "loss": 0.0215, "step": 6141 }, { "epoch": 2.42, "learning_rate": 3.5465067341430284e-07, "loss": 0.0267, "step": 6142 }, { "epoch": 2.42, "learning_rate": 3.5463448168727004e-07, "loss": 0.0216, "step": 6143 }, { "epoch": 2.42, "learning_rate": 3.546182874399303e-07, "loss": 0.0207, "step": 6144 }, { "epoch": 2.42, "learning_rate": 3.5460209067254736e-07, "loss": 0.0211, "step": 6145 }, { "epoch": 2.42, "learning_rate": 3.5458589138538535e-07, "loss": 0.0251, "step": 6146 }, { "epoch": 2.42, "learning_rate": 3.545696895787083e-07, "loss": 0.0257, "step": 6147 }, { "epoch": 2.42, "learning_rate": 3.545534852527802e-07, "loss": 0.021, "step": 6148 }, { "epoch": 2.42, "learning_rate": 3.545372784078652e-07, "loss": 0.0224, "step": 6149 }, { "epoch": 2.42, "learning_rate": 3.545210690442274e-07, "loss": 0.0249, "step": 6150 }, { "epoch": 2.42, "learning_rate": 3.54504857162131e-07, "loss": 0.0191, "step": 6151 }, { "epoch": 2.42, "learning_rate": 3.5448864276184027e-07, "loss": 0.0213, "step": 6152 }, { "epoch": 2.42, "learning_rate": 3.544724258436195e-07, "loss": 0.027, "step": 6153 }, { "epoch": 2.43, "learning_rate": 3.5445620640773293e-07, "loss": 0.0206, "step": 6154 }, { "epoch": 2.43, "learning_rate": 3.544399844544449e-07, "loss": 0.0213, "step": 6155 }, { "epoch": 2.43, "learning_rate": 3.544237599840199e-07, "loss": 0.0205, "step": 6156 }, { "epoch": 2.43, "learning_rate": 3.544075329967223e-07, "loss": 0.0271, "step": 6157 }, { "epoch": 2.43, "learning_rate": 3.543913034928166e-07, "loss": 0.0218, "step": 6158 }, { "epoch": 2.43, "learning_rate": 3.5437507147256724e-07, "loss": 0.02, "step": 6159 }, { "epoch": 2.43, "learning_rate": 3.5435883693623893e-07, "loss": 0.0239, "step": 6160 }, { "epoch": 2.43, "learning_rate": 3.543425998840961e-07, "loss": 0.0256, "step": 6161 }, { "epoch": 2.43, "learning_rate": 3.5432636031640344e-07, "loss": 0.0223, "step": 6162 }, { "epoch": 2.43, "learning_rate": 3.5431011823342577e-07, "loss": 0.0227, "step": 6163 }, { "epoch": 2.43, "learning_rate": 3.5429387363542765e-07, "loss": 0.0216, "step": 6164 }, { "epoch": 2.43, "learning_rate": 3.542776265226738e-07, "loss": 0.0244, "step": 6165 }, { "epoch": 2.43, "learning_rate": 3.5426137689542925e-07, "loss": 0.0233, "step": 6166 }, { "epoch": 2.43, "learning_rate": 3.5424512475395856e-07, "loss": 0.0254, "step": 6167 }, { "epoch": 2.43, "learning_rate": 3.5422887009852687e-07, "loss": 0.0257, "step": 6168 }, { "epoch": 2.43, "learning_rate": 3.54212612929399e-07, "loss": 0.0215, "step": 6169 }, { "epoch": 2.43, "learning_rate": 3.541963532468399e-07, "loss": 0.0284, "step": 6170 }, { "epoch": 2.43, "learning_rate": 3.5418009105111455e-07, "loss": 0.0226, "step": 6171 }, { "epoch": 2.43, "learning_rate": 3.541638263424881e-07, "loss": 0.0188, "step": 6172 }, { "epoch": 2.43, "learning_rate": 3.541475591212255e-07, "loss": 0.0259, "step": 6173 }, { "epoch": 2.43, "learning_rate": 3.541312893875921e-07, "loss": 0.0295, "step": 6174 }, { "epoch": 2.43, "learning_rate": 3.541150171418528e-07, "loss": 0.0214, "step": 6175 }, { "epoch": 2.43, "learning_rate": 3.54098742384273e-07, "loss": 0.0228, "step": 6176 }, { "epoch": 2.43, "learning_rate": 3.540824651151179e-07, "loss": 0.0244, "step": 6177 }, { "epoch": 2.43, "learning_rate": 3.5406618533465284e-07, "loss": 0.026, "step": 6178 }, { "epoch": 2.44, "learning_rate": 3.540499030431431e-07, "loss": 0.0242, "step": 6179 }, { "epoch": 2.44, "learning_rate": 3.5403361824085403e-07, "loss": 0.025, "step": 6180 }, { "epoch": 2.44, "learning_rate": 3.5401733092805106e-07, "loss": 0.0236, "step": 6181 }, { "epoch": 2.44, "learning_rate": 3.540010411049997e-07, "loss": 0.0246, "step": 6182 }, { "epoch": 2.44, "learning_rate": 3.5398474877196546e-07, "loss": 0.0288, "step": 6183 }, { "epoch": 2.44, "learning_rate": 3.539684539292138e-07, "loss": 0.023, "step": 6184 }, { "epoch": 2.44, "learning_rate": 3.539521565770104e-07, "loss": 0.0249, "step": 6185 }, { "epoch": 2.44, "learning_rate": 3.539358567156208e-07, "loss": 0.0241, "step": 6186 }, { "epoch": 2.44, "learning_rate": 3.539195543453107e-07, "loss": 0.0256, "step": 6187 }, { "epoch": 2.44, "learning_rate": 3.539032494663458e-07, "loss": 0.0175, "step": 6188 }, { "epoch": 2.44, "learning_rate": 3.5388694207899183e-07, "loss": 0.0241, "step": 6189 }, { "epoch": 2.44, "learning_rate": 3.5387063218351455e-07, "loss": 0.0212, "step": 6190 }, { "epoch": 2.44, "learning_rate": 3.5385431978017987e-07, "loss": 0.0202, "step": 6191 }, { "epoch": 2.44, "learning_rate": 3.538380048692536e-07, "loss": 0.0176, "step": 6192 }, { "epoch": 2.44, "learning_rate": 3.5382168745100164e-07, "loss": 0.0242, "step": 6193 }, { "epoch": 2.44, "learning_rate": 3.5380536752569e-07, "loss": 0.0216, "step": 6194 }, { "epoch": 2.44, "learning_rate": 3.5378904509358456e-07, "loss": 0.0235, "step": 6195 }, { "epoch": 2.44, "learning_rate": 3.5377272015495143e-07, "loss": 0.0235, "step": 6196 }, { "epoch": 2.44, "learning_rate": 3.537563927100567e-07, "loss": 0.0243, "step": 6197 }, { "epoch": 2.44, "learning_rate": 3.5374006275916646e-07, "loss": 0.0182, "step": 6198 }, { "epoch": 2.44, "learning_rate": 3.5372373030254677e-07, "loss": 0.0219, "step": 6199 }, { "epoch": 2.44, "learning_rate": 3.5370739534046403e-07, "loss": 0.0219, "step": 6200 }, { "epoch": 2.44, "learning_rate": 3.536910578731843e-07, "loss": 0.0242, "step": 6201 }, { "epoch": 2.44, "learning_rate": 3.5367471790097394e-07, "loss": 0.0197, "step": 6202 }, { "epoch": 2.44, "learning_rate": 3.5365837542409917e-07, "loss": 0.0248, "step": 6203 }, { "epoch": 2.45, "learning_rate": 3.5364203044282643e-07, "loss": 0.021, "step": 6204 }, { "epoch": 2.45, "learning_rate": 3.536256829574221e-07, "loss": 0.024, "step": 6205 }, { "epoch": 2.45, "learning_rate": 3.536093329681527e-07, "loss": 0.0217, "step": 6206 }, { "epoch": 2.45, "learning_rate": 3.535929804752846e-07, "loss": 0.0205, "step": 6207 }, { "epoch": 2.45, "learning_rate": 3.5357662547908433e-07, "loss": 0.0264, "step": 6208 }, { "epoch": 2.45, "learning_rate": 3.5356026797981845e-07, "loss": 0.0231, "step": 6209 }, { "epoch": 2.45, "learning_rate": 3.535439079777536e-07, "loss": 0.0266, "step": 6210 }, { "epoch": 2.45, "learning_rate": 3.535275454731565e-07, "loss": 0.0272, "step": 6211 }, { "epoch": 2.45, "learning_rate": 3.535111804662936e-07, "loss": 0.0228, "step": 6212 }, { "epoch": 2.45, "learning_rate": 3.5349481295743184e-07, "loss": 0.0227, "step": 6213 }, { "epoch": 2.45, "learning_rate": 3.534784429468379e-07, "loss": 0.0208, "step": 6214 }, { "epoch": 2.45, "learning_rate": 3.534620704347786e-07, "loss": 0.0202, "step": 6215 }, { "epoch": 2.45, "learning_rate": 3.534456954215208e-07, "loss": 0.0202, "step": 6216 }, { "epoch": 2.45, "learning_rate": 3.534293179073314e-07, "loss": 0.0221, "step": 6217 }, { "epoch": 2.45, "learning_rate": 3.534129378924772e-07, "loss": 0.0222, "step": 6218 }, { "epoch": 2.45, "learning_rate": 3.533965553772254e-07, "loss": 0.0252, "step": 6219 }, { "epoch": 2.45, "learning_rate": 3.5338017036184283e-07, "loss": 0.0257, "step": 6220 }, { "epoch": 2.45, "learning_rate": 3.5336378284659656e-07, "loss": 0.02, "step": 6221 }, { "epoch": 2.45, "learning_rate": 3.533473928317538e-07, "loss": 0.0215, "step": 6222 }, { "epoch": 2.45, "learning_rate": 3.5333100031758153e-07, "loss": 0.0262, "step": 6223 }, { "epoch": 2.45, "learning_rate": 3.5331460530434705e-07, "loss": 0.0223, "step": 6224 }, { "epoch": 2.45, "learning_rate": 3.532982077923174e-07, "loss": 0.0222, "step": 6225 }, { "epoch": 2.45, "learning_rate": 3.5328180778176005e-07, "loss": 0.0192, "step": 6226 }, { "epoch": 2.45, "learning_rate": 3.5326540527294217e-07, "loss": 0.0215, "step": 6227 }, { "epoch": 2.45, "learning_rate": 3.532490002661311e-07, "loss": 0.0249, "step": 6228 }, { "epoch": 2.45, "learning_rate": 3.532325927615943e-07, "loss": 0.0237, "step": 6229 }, { "epoch": 2.46, "learning_rate": 3.5321618275959903e-07, "loss": 0.0212, "step": 6230 }, { "epoch": 2.46, "learning_rate": 3.531997702604129e-07, "loss": 0.0221, "step": 6231 }, { "epoch": 2.46, "learning_rate": 3.531833552643034e-07, "loss": 0.0238, "step": 6232 }, { "epoch": 2.46, "learning_rate": 3.5316693777153794e-07, "loss": 0.024, "step": 6233 }, { "epoch": 2.46, "learning_rate": 3.531505177823842e-07, "loss": 0.02, "step": 6234 }, { "epoch": 2.46, "learning_rate": 3.5313409529710974e-07, "loss": 0.0216, "step": 6235 }, { "epoch": 2.46, "learning_rate": 3.5311767031598235e-07, "loss": 0.024, "step": 6236 }, { "epoch": 2.46, "learning_rate": 3.531012428392696e-07, "loss": 0.0201, "step": 6237 }, { "epoch": 2.46, "learning_rate": 3.530848128672393e-07, "loss": 0.025, "step": 6238 }, { "epoch": 2.46, "learning_rate": 3.530683804001592e-07, "loss": 0.0252, "step": 6239 }, { "epoch": 2.46, "learning_rate": 3.5305194543829716e-07, "loss": 0.0254, "step": 6240 }, { "epoch": 2.46, "learning_rate": 3.53035507981921e-07, "loss": 0.0246, "step": 6241 }, { "epoch": 2.46, "learning_rate": 3.5301906803129864e-07, "loss": 0.0242, "step": 6242 }, { "epoch": 2.46, "learning_rate": 3.53002625586698e-07, "loss": 0.0251, "step": 6243 }, { "epoch": 2.46, "learning_rate": 3.5298618064838715e-07, "loss": 0.0269, "step": 6244 }, { "epoch": 2.46, "learning_rate": 3.529697332166341e-07, "loss": 0.0226, "step": 6245 }, { "epoch": 2.46, "learning_rate": 3.529532832917068e-07, "loss": 0.022, "step": 6246 }, { "epoch": 2.46, "learning_rate": 3.529368308738735e-07, "loss": 0.0244, "step": 6247 }, { "epoch": 2.46, "learning_rate": 3.529203759634023e-07, "loss": 0.0203, "step": 6248 }, { "epoch": 2.46, "learning_rate": 3.5290391856056134e-07, "loss": 0.0211, "step": 6249 }, { "epoch": 2.46, "learning_rate": 3.5288745866561895e-07, "loss": 0.0233, "step": 6250 }, { "epoch": 2.46, "learning_rate": 3.5287099627884327e-07, "loss": 0.019, "step": 6251 }, { "epoch": 2.46, "learning_rate": 3.5285453140050276e-07, "loss": 0.0208, "step": 6252 }, { "epoch": 2.46, "learning_rate": 3.528380640308657e-07, "loss": 0.0235, "step": 6253 }, { "epoch": 2.46, "learning_rate": 3.5282159417020044e-07, "loss": 0.0254, "step": 6254 }, { "epoch": 2.47, "learning_rate": 3.5280512181877543e-07, "loss": 0.0236, "step": 6255 }, { "epoch": 2.47, "learning_rate": 3.5278864697685926e-07, "loss": 0.0212, "step": 6256 }, { "epoch": 2.47, "learning_rate": 3.527721696447203e-07, "loss": 0.0242, "step": 6257 }, { "epoch": 2.47, "learning_rate": 3.5275568982262714e-07, "loss": 0.0217, "step": 6258 }, { "epoch": 2.47, "learning_rate": 3.5273920751084843e-07, "loss": 0.0252, "step": 6259 }, { "epoch": 2.47, "learning_rate": 3.527227227096528e-07, "loss": 0.0244, "step": 6260 }, { "epoch": 2.47, "learning_rate": 3.527062354193089e-07, "loss": 0.0285, "step": 6261 }, { "epoch": 2.47, "learning_rate": 3.5268974564008537e-07, "loss": 0.0224, "step": 6262 }, { "epoch": 2.47, "learning_rate": 3.526732533722511e-07, "loss": 0.0238, "step": 6263 }, { "epoch": 2.47, "learning_rate": 3.526567586160749e-07, "loss": 0.0214, "step": 6264 }, { "epoch": 2.47, "learning_rate": 3.526402613718255e-07, "loss": 0.0229, "step": 6265 }, { "epoch": 2.47, "learning_rate": 3.5262376163977174e-07, "loss": 0.0256, "step": 6266 }, { "epoch": 2.47, "learning_rate": 3.526072594201827e-07, "loss": 0.0235, "step": 6267 }, { "epoch": 2.47, "learning_rate": 3.525907547133273e-07, "loss": 0.0182, "step": 6268 }, { "epoch": 2.47, "learning_rate": 3.525742475194745e-07, "loss": 0.0211, "step": 6269 }, { "epoch": 2.47, "learning_rate": 3.525577378388933e-07, "loss": 0.0241, "step": 6270 }, { "epoch": 2.47, "learning_rate": 3.525412256718529e-07, "loss": 0.0222, "step": 6271 }, { "epoch": 2.47, "learning_rate": 3.525247110186223e-07, "loss": 0.0191, "step": 6272 }, { "epoch": 2.47, "learning_rate": 3.5250819387947074e-07, "loss": 0.0221, "step": 6273 }, { "epoch": 2.47, "learning_rate": 3.524916742546674e-07, "loss": 0.0236, "step": 6274 }, { "epoch": 2.47, "learning_rate": 3.524751521444815e-07, "loss": 0.0225, "step": 6275 }, { "epoch": 2.47, "learning_rate": 3.524586275491824e-07, "loss": 0.0225, "step": 6276 }, { "epoch": 2.47, "learning_rate": 3.524421004690394e-07, "loss": 0.0192, "step": 6277 }, { "epoch": 2.47, "learning_rate": 3.5242557090432177e-07, "loss": 0.0234, "step": 6278 }, { "epoch": 2.47, "learning_rate": 3.5240903885529906e-07, "loss": 0.0233, "step": 6279 }, { "epoch": 2.47, "learning_rate": 3.523925043222407e-07, "loss": 0.0234, "step": 6280 }, { "epoch": 2.48, "learning_rate": 3.52375967305416e-07, "loss": 0.0161, "step": 6281 }, { "epoch": 2.48, "learning_rate": 3.5235942780509476e-07, "loss": 0.0237, "step": 6282 }, { "epoch": 2.48, "learning_rate": 3.523428858215463e-07, "loss": 0.0204, "step": 6283 }, { "epoch": 2.48, "learning_rate": 3.5232634135504037e-07, "loss": 0.0256, "step": 6284 }, { "epoch": 2.48, "learning_rate": 3.523097944058466e-07, "loss": 0.0222, "step": 6285 }, { "epoch": 2.48, "learning_rate": 3.5229324497423464e-07, "loss": 0.0225, "step": 6286 }, { "epoch": 2.48, "learning_rate": 3.5227669306047423e-07, "loss": 0.0214, "step": 6287 }, { "epoch": 2.48, "learning_rate": 3.522601386648352e-07, "loss": 0.0229, "step": 6288 }, { "epoch": 2.48, "learning_rate": 3.522435817875873e-07, "loss": 0.0234, "step": 6289 }, { "epoch": 2.48, "learning_rate": 3.522270224290004e-07, "loss": 0.0251, "step": 6290 }, { "epoch": 2.48, "learning_rate": 3.522104605893444e-07, "loss": 0.0242, "step": 6291 }, { "epoch": 2.48, "learning_rate": 3.521938962688892e-07, "loss": 0.0245, "step": 6292 }, { "epoch": 2.48, "learning_rate": 3.5217732946790483e-07, "loss": 0.0234, "step": 6293 }, { "epoch": 2.48, "learning_rate": 3.521607601866613e-07, "loss": 0.0264, "step": 6294 }, { "epoch": 2.48, "learning_rate": 3.5214418842542857e-07, "loss": 0.0253, "step": 6295 }, { "epoch": 2.48, "learning_rate": 3.5212761418447683e-07, "loss": 0.0234, "step": 6296 }, { "epoch": 2.48, "learning_rate": 3.5211103746407617e-07, "loss": 0.0231, "step": 6297 }, { "epoch": 2.48, "learning_rate": 3.520944582644968e-07, "loss": 0.0251, "step": 6298 }, { "epoch": 2.48, "learning_rate": 3.520778765860089e-07, "loss": 0.0229, "step": 6299 }, { "epoch": 2.48, "learning_rate": 3.5206129242888276e-07, "loss": 0.0189, "step": 6300 }, { "epoch": 2.48, "learning_rate": 3.5204470579338865e-07, "loss": 0.022, "step": 6301 }, { "epoch": 2.48, "learning_rate": 3.520281166797969e-07, "loss": 0.0237, "step": 6302 }, { "epoch": 2.48, "learning_rate": 3.5201152508837795e-07, "loss": 0.0226, "step": 6303 }, { "epoch": 2.48, "learning_rate": 3.5199493101940215e-07, "loss": 0.0232, "step": 6304 }, { "epoch": 2.48, "learning_rate": 3.5197833447313995e-07, "loss": 0.0194, "step": 6305 }, { "epoch": 2.49, "learning_rate": 3.519617354498619e-07, "loss": 0.0242, "step": 6306 }, { "epoch": 2.49, "learning_rate": 3.5194513394983855e-07, "loss": 0.0227, "step": 6307 }, { "epoch": 2.49, "learning_rate": 3.519285299733404e-07, "loss": 0.0195, "step": 6308 }, { "epoch": 2.49, "learning_rate": 3.5191192352063814e-07, "loss": 0.023, "step": 6309 }, { "epoch": 2.49, "learning_rate": 3.518953145920024e-07, "loss": 0.0216, "step": 6310 }, { "epoch": 2.49, "learning_rate": 3.5187870318770394e-07, "loss": 0.0242, "step": 6311 }, { "epoch": 2.49, "learning_rate": 3.5186208930801335e-07, "loss": 0.0208, "step": 6312 }, { "epoch": 2.49, "learning_rate": 3.518454729532016e-07, "loss": 0.022, "step": 6313 }, { "epoch": 2.49, "learning_rate": 3.5182885412353935e-07, "loss": 0.0245, "step": 6314 }, { "epoch": 2.49, "learning_rate": 3.5181223281929756e-07, "loss": 0.0193, "step": 6315 }, { "epoch": 2.49, "learning_rate": 3.517956090407471e-07, "loss": 0.0231, "step": 6316 }, { "epoch": 2.49, "learning_rate": 3.517789827881589e-07, "loss": 0.025, "step": 6317 }, { "epoch": 2.49, "learning_rate": 3.5176235406180395e-07, "loss": 0.0205, "step": 6318 }, { "epoch": 2.49, "learning_rate": 3.5174572286195333e-07, "loss": 0.0223, "step": 6319 }, { "epoch": 2.49, "learning_rate": 3.5172908918887804e-07, "loss": 0.0236, "step": 6320 }, { "epoch": 2.49, "learning_rate": 3.517124530428492e-07, "loss": 0.0201, "step": 6321 }, { "epoch": 2.49, "learning_rate": 3.5169581442413795e-07, "loss": 0.0206, "step": 6322 }, { "epoch": 2.49, "learning_rate": 3.516791733330154e-07, "loss": 0.0203, "step": 6323 }, { "epoch": 2.49, "learning_rate": 3.5166252976975297e-07, "loss": 0.0181, "step": 6324 }, { "epoch": 2.49, "learning_rate": 3.516458837346217e-07, "loss": 0.0219, "step": 6325 }, { "epoch": 2.49, "learning_rate": 3.5162923522789306e-07, "loss": 0.023, "step": 6326 }, { "epoch": 2.49, "learning_rate": 3.5161258424983836e-07, "loss": 0.0229, "step": 6327 }, { "epoch": 2.49, "learning_rate": 3.515959308007289e-07, "loss": 0.0207, "step": 6328 }, { "epoch": 2.49, "learning_rate": 3.515792748808362e-07, "loss": 0.0224, "step": 6329 }, { "epoch": 2.49, "learning_rate": 3.5156261649043167e-07, "loss": 0.0224, "step": 6330 }, { "epoch": 2.5, "learning_rate": 3.515459556297869e-07, "loss": 0.0203, "step": 6331 }, { "epoch": 2.5, "learning_rate": 3.515292922991733e-07, "loss": 0.024, "step": 6332 }, { "epoch": 2.5, "learning_rate": 3.5151262649886254e-07, "loss": 0.02, "step": 6333 }, { "epoch": 2.5, "learning_rate": 3.5149595822912627e-07, "loss": 0.0222, "step": 6334 }, { "epoch": 2.5, "learning_rate": 3.514792874902361e-07, "loss": 0.0214, "step": 6335 }, { "epoch": 2.5, "learning_rate": 3.5146261428246384e-07, "loss": 0.0219, "step": 6336 }, { "epoch": 2.5, "learning_rate": 3.514459386060811e-07, "loss": 0.0254, "step": 6337 }, { "epoch": 2.5, "learning_rate": 3.5142926046135977e-07, "loss": 0.0214, "step": 6338 }, { "epoch": 2.5, "learning_rate": 3.514125798485716e-07, "loss": 0.0241, "step": 6339 }, { "epoch": 2.5, "learning_rate": 3.513958967679885e-07, "loss": 0.023, "step": 6340 }, { "epoch": 2.5, "learning_rate": 3.513792112198824e-07, "loss": 0.0226, "step": 6341 }, { "epoch": 2.5, "learning_rate": 3.5136252320452525e-07, "loss": 0.0226, "step": 6342 }, { "epoch": 2.5, "learning_rate": 3.513458327221889e-07, "loss": 0.0195, "step": 6343 }, { "epoch": 2.5, "learning_rate": 3.5132913977314564e-07, "loss": 0.0172, "step": 6344 }, { "epoch": 2.5, "learning_rate": 3.5131244435766735e-07, "loss": 0.0264, "step": 6345 }, { "epoch": 2.5, "learning_rate": 3.5129574647602617e-07, "loss": 0.0217, "step": 6346 }, { "epoch": 2.5, "learning_rate": 3.512790461284943e-07, "loss": 0.0247, "step": 6347 }, { "epoch": 2.5, "learning_rate": 3.512623433153438e-07, "loss": 0.0263, "step": 6348 }, { "epoch": 2.5, "learning_rate": 3.5124563803684713e-07, "loss": 0.0205, "step": 6349 }, { "epoch": 2.5, "learning_rate": 3.512289302932763e-07, "loss": 0.0235, "step": 6350 }, { "epoch": 2.5, "learning_rate": 3.512122200849038e-07, "loss": 0.0221, "step": 6351 }, { "epoch": 2.5, "learning_rate": 3.51195507412002e-07, "loss": 0.0252, "step": 6352 }, { "epoch": 2.5, "learning_rate": 3.5117879227484314e-07, "loss": 0.0221, "step": 6353 }, { "epoch": 2.5, "learning_rate": 3.511620746736997e-07, "loss": 0.0235, "step": 6354 }, { "epoch": 2.5, "learning_rate": 3.5114535460884426e-07, "loss": 0.0179, "step": 6355 }, { "epoch": 2.5, "learning_rate": 3.5112863208054924e-07, "loss": 0.0238, "step": 6356 }, { "epoch": 2.51, "learning_rate": 3.511119070890872e-07, "loss": 0.0199, "step": 6357 }, { "epoch": 2.51, "learning_rate": 3.510951796347307e-07, "loss": 0.0233, "step": 6358 }, { "epoch": 2.51, "learning_rate": 3.5107844971775246e-07, "loss": 0.0246, "step": 6359 }, { "epoch": 2.51, "learning_rate": 3.510617173384251e-07, "loss": 0.0231, "step": 6360 }, { "epoch": 2.51, "learning_rate": 3.5104498249702127e-07, "loss": 0.0258, "step": 6361 }, { "epoch": 2.51, "learning_rate": 3.510282451938138e-07, "loss": 0.0257, "step": 6362 }, { "epoch": 2.51, "learning_rate": 3.510115054290756e-07, "loss": 0.0273, "step": 6363 }, { "epoch": 2.51, "learning_rate": 3.5099476320307923e-07, "loss": 0.0238, "step": 6364 }, { "epoch": 2.51, "learning_rate": 3.509780185160977e-07, "loss": 0.0201, "step": 6365 }, { "epoch": 2.51, "learning_rate": 3.5096127136840403e-07, "loss": 0.0259, "step": 6366 }, { "epoch": 2.51, "learning_rate": 3.50944521760271e-07, "loss": 0.0208, "step": 6367 }, { "epoch": 2.51, "learning_rate": 3.5092776969197175e-07, "loss": 0.0251, "step": 6368 }, { "epoch": 2.51, "learning_rate": 3.509110151637791e-07, "loss": 0.0226, "step": 6369 }, { "epoch": 2.51, "learning_rate": 3.5089425817596636e-07, "loss": 0.0209, "step": 6370 }, { "epoch": 2.51, "learning_rate": 3.5087749872880653e-07, "loss": 0.0202, "step": 6371 }, { "epoch": 2.51, "learning_rate": 3.508607368225728e-07, "loss": 0.0214, "step": 6372 }, { "epoch": 2.51, "learning_rate": 3.508439724575383e-07, "loss": 0.0218, "step": 6373 }, { "epoch": 2.51, "learning_rate": 3.508272056339763e-07, "loss": 0.0241, "step": 6374 }, { "epoch": 2.51, "learning_rate": 3.5081043635216014e-07, "loss": 0.0259, "step": 6375 }, { "epoch": 2.51, "learning_rate": 3.50793664612363e-07, "loss": 0.0243, "step": 6376 }, { "epoch": 2.51, "learning_rate": 3.507768904148584e-07, "loss": 0.0248, "step": 6377 }, { "epoch": 2.51, "learning_rate": 3.507601137599195e-07, "loss": 0.0225, "step": 6378 }, { "epoch": 2.51, "learning_rate": 3.5074333464782e-07, "loss": 0.0228, "step": 6379 }, { "epoch": 2.51, "learning_rate": 3.5072655307883326e-07, "loss": 0.0206, "step": 6380 }, { "epoch": 2.51, "learning_rate": 3.507097690532327e-07, "loss": 0.0229, "step": 6381 }, { "epoch": 2.52, "learning_rate": 3.5069298257129195e-07, "loss": 0.0235, "step": 6382 }, { "epoch": 2.52, "learning_rate": 3.506761936332847e-07, "loss": 0.0204, "step": 6383 }, { "epoch": 2.52, "learning_rate": 3.5065940223948437e-07, "loss": 0.021, "step": 6384 }, { "epoch": 2.52, "learning_rate": 3.506426083901648e-07, "loss": 0.021, "step": 6385 }, { "epoch": 2.52, "learning_rate": 3.506258120855998e-07, "loss": 0.0214, "step": 6386 }, { "epoch": 2.52, "learning_rate": 3.5060901332606285e-07, "loss": 0.0254, "step": 6387 }, { "epoch": 2.52, "learning_rate": 3.505922121118279e-07, "loss": 0.023, "step": 6388 }, { "epoch": 2.52, "learning_rate": 3.505754084431688e-07, "loss": 0.0221, "step": 6389 }, { "epoch": 2.52, "learning_rate": 3.5055860232035934e-07, "loss": 0.0233, "step": 6390 }, { "epoch": 2.52, "learning_rate": 3.5054179374367355e-07, "loss": 0.0218, "step": 6391 }, { "epoch": 2.52, "learning_rate": 3.505249827133853e-07, "loss": 0.0227, "step": 6392 }, { "epoch": 2.52, "learning_rate": 3.5050816922976855e-07, "loss": 0.022, "step": 6393 }, { "epoch": 2.52, "learning_rate": 3.504913532930974e-07, "loss": 0.0239, "step": 6394 }, { "epoch": 2.52, "learning_rate": 3.50474534903646e-07, "loss": 0.0236, "step": 6395 }, { "epoch": 2.52, "learning_rate": 3.504577140616883e-07, "loss": 0.0204, "step": 6396 }, { "epoch": 2.52, "learning_rate": 3.504408907674985e-07, "loss": 0.0219, "step": 6397 }, { "epoch": 2.52, "learning_rate": 3.504240650213509e-07, "loss": 0.0211, "step": 6398 }, { "epoch": 2.52, "learning_rate": 3.5040723682351964e-07, "loss": 0.0211, "step": 6399 }, { "epoch": 2.52, "learning_rate": 3.50390406174279e-07, "loss": 0.0253, "step": 6400 }, { "epoch": 2.52, "learning_rate": 3.503735730739033e-07, "loss": 0.0277, "step": 6401 }, { "epoch": 2.52, "learning_rate": 3.503567375226669e-07, "loss": 0.0209, "step": 6402 }, { "epoch": 2.52, "learning_rate": 3.503398995208442e-07, "loss": 0.0198, "step": 6403 }, { "epoch": 2.52, "learning_rate": 3.5032305906870963e-07, "loss": 0.0223, "step": 6404 }, { "epoch": 2.52, "learning_rate": 3.503062161665377e-07, "loss": 0.0247, "step": 6405 }, { "epoch": 2.52, "learning_rate": 3.502893708146028e-07, "loss": 0.0196, "step": 6406 }, { "epoch": 2.53, "learning_rate": 3.502725230131796e-07, "loss": 0.0193, "step": 6407 }, { "epoch": 2.53, "learning_rate": 3.5025567276254267e-07, "loss": 0.0188, "step": 6408 }, { "epoch": 2.53, "learning_rate": 3.5023882006296663e-07, "loss": 0.0229, "step": 6409 }, { "epoch": 2.53, "learning_rate": 3.5022196491472623e-07, "loss": 0.0206, "step": 6410 }, { "epoch": 2.53, "learning_rate": 3.5020510731809607e-07, "loss": 0.0204, "step": 6411 }, { "epoch": 2.53, "learning_rate": 3.501882472733509e-07, "loss": 0.024, "step": 6412 }, { "epoch": 2.53, "learning_rate": 3.501713847807656e-07, "loss": 0.0213, "step": 6413 }, { "epoch": 2.53, "learning_rate": 3.5015451984061496e-07, "loss": 0.026, "step": 6414 }, { "epoch": 2.53, "learning_rate": 3.5013765245317384e-07, "loss": 0.0226, "step": 6415 }, { "epoch": 2.53, "learning_rate": 3.5012078261871717e-07, "loss": 0.021, "step": 6416 }, { "epoch": 2.53, "learning_rate": 3.5010391033751996e-07, "loss": 0.0222, "step": 6417 }, { "epoch": 2.53, "learning_rate": 3.50087035609857e-07, "loss": 0.0206, "step": 6418 }, { "epoch": 2.53, "learning_rate": 3.500701584360036e-07, "loss": 0.0205, "step": 6419 }, { "epoch": 2.53, "learning_rate": 3.5005327881623464e-07, "loss": 0.0217, "step": 6420 }, { "epoch": 2.53, "learning_rate": 3.500363967508253e-07, "loss": 0.0238, "step": 6421 }, { "epoch": 2.53, "learning_rate": 3.500195122400508e-07, "loss": 0.0202, "step": 6422 }, { "epoch": 2.53, "learning_rate": 3.5000262528418615e-07, "loss": 0.0246, "step": 6423 }, { "epoch": 2.53, "learning_rate": 3.499857358835067e-07, "loss": 0.0223, "step": 6424 }, { "epoch": 2.53, "learning_rate": 3.4996884403828773e-07, "loss": 0.0204, "step": 6425 }, { "epoch": 2.53, "learning_rate": 3.4995194974880456e-07, "loss": 0.0216, "step": 6426 }, { "epoch": 2.53, "learning_rate": 3.4993505301533243e-07, "loss": 0.0228, "step": 6427 }, { "epoch": 2.53, "learning_rate": 3.499181538381468e-07, "loss": 0.0226, "step": 6428 }, { "epoch": 2.53, "learning_rate": 3.499012522175232e-07, "loss": 0.0224, "step": 6429 }, { "epoch": 2.53, "learning_rate": 3.4988434815373696e-07, "loss": 0.0239, "step": 6430 }, { "epoch": 2.53, "learning_rate": 3.498674416470637e-07, "loss": 0.0229, "step": 6431 }, { "epoch": 2.53, "learning_rate": 3.4985053269777886e-07, "loss": 0.0194, "step": 6432 }, { "epoch": 2.54, "learning_rate": 3.498336213061581e-07, "loss": 0.0196, "step": 6433 }, { "epoch": 2.54, "learning_rate": 3.4981670747247705e-07, "loss": 0.0255, "step": 6434 }, { "epoch": 2.54, "learning_rate": 3.4979979119701135e-07, "loss": 0.0229, "step": 6435 }, { "epoch": 2.54, "learning_rate": 3.4978287248003675e-07, "loss": 0.0248, "step": 6436 }, { "epoch": 2.54, "learning_rate": 3.4976595132182886e-07, "loss": 0.0254, "step": 6437 }, { "epoch": 2.54, "learning_rate": 3.497490277226637e-07, "loss": 0.0222, "step": 6438 }, { "epoch": 2.54, "learning_rate": 3.497321016828169e-07, "loss": 0.0248, "step": 6439 }, { "epoch": 2.54, "learning_rate": 3.497151732025645e-07, "loss": 0.0216, "step": 6440 }, { "epoch": 2.54, "learning_rate": 3.496982422821823e-07, "loss": 0.0198, "step": 6441 }, { "epoch": 2.54, "learning_rate": 3.496813089219462e-07, "loss": 0.0235, "step": 6442 }, { "epoch": 2.54, "learning_rate": 3.496643731221323e-07, "loss": 0.0237, "step": 6443 }, { "epoch": 2.54, "learning_rate": 3.4964743488301657e-07, "loss": 0.0269, "step": 6444 }, { "epoch": 2.54, "learning_rate": 3.4963049420487507e-07, "loss": 0.0216, "step": 6445 }, { "epoch": 2.54, "learning_rate": 3.4961355108798394e-07, "loss": 0.021, "step": 6446 }, { "epoch": 2.54, "learning_rate": 3.495966055326193e-07, "loss": 0.0273, "step": 6447 }, { "epoch": 2.54, "learning_rate": 3.495796575390574e-07, "loss": 0.0249, "step": 6448 }, { "epoch": 2.54, "learning_rate": 3.495627071075744e-07, "loss": 0.021, "step": 6449 }, { "epoch": 2.54, "learning_rate": 3.4954575423844656e-07, "loss": 0.0278, "step": 6450 }, { "epoch": 2.54, "learning_rate": 3.495287989319502e-07, "loss": 0.0237, "step": 6451 }, { "epoch": 2.54, "learning_rate": 3.495118411883617e-07, "loss": 0.0208, "step": 6452 }, { "epoch": 2.54, "learning_rate": 3.494948810079574e-07, "loss": 0.0212, "step": 6453 }, { "epoch": 2.54, "learning_rate": 3.494779183910138e-07, "loss": 0.0182, "step": 6454 }, { "epoch": 2.54, "learning_rate": 3.494609533378073e-07, "loss": 0.0217, "step": 6455 }, { "epoch": 2.54, "learning_rate": 3.4944398584861435e-07, "loss": 0.0221, "step": 6456 }, { "epoch": 2.54, "learning_rate": 3.494270159237116e-07, "loss": 0.0194, "step": 6457 }, { "epoch": 2.55, "learning_rate": 3.494100435633756e-07, "loss": 0.0247, "step": 6458 }, { "epoch": 2.55, "learning_rate": 3.493930687678829e-07, "loss": 0.0227, "step": 6459 }, { "epoch": 2.55, "learning_rate": 3.4937609153751036e-07, "loss": 0.0176, "step": 6460 }, { "epoch": 2.55, "learning_rate": 3.4935911187253447e-07, "loss": 0.0209, "step": 6461 }, { "epoch": 2.55, "learning_rate": 3.49342129773232e-07, "loss": 0.026, "step": 6462 }, { "epoch": 2.55, "learning_rate": 3.493251452398799e-07, "loss": 0.0204, "step": 6463 }, { "epoch": 2.55, "learning_rate": 3.493081582727548e-07, "loss": 0.0188, "step": 6464 }, { "epoch": 2.55, "learning_rate": 3.492911688721337e-07, "loss": 0.0214, "step": 6465 }, { "epoch": 2.55, "learning_rate": 3.492741770382934e-07, "loss": 0.0222, "step": 6466 }, { "epoch": 2.55, "learning_rate": 3.4925718277151095e-07, "loss": 0.0246, "step": 6467 }, { "epoch": 2.55, "learning_rate": 3.492401860720632e-07, "loss": 0.0207, "step": 6468 }, { "epoch": 2.55, "learning_rate": 3.4922318694022723e-07, "loss": 0.0216, "step": 6469 }, { "epoch": 2.55, "learning_rate": 3.4920618537628017e-07, "loss": 0.02, "step": 6470 }, { "epoch": 2.55, "learning_rate": 3.49189181380499e-07, "loss": 0.0209, "step": 6471 }, { "epoch": 2.55, "learning_rate": 3.49172174953161e-07, "loss": 0.0249, "step": 6472 }, { "epoch": 2.55, "learning_rate": 3.4915516609454314e-07, "loss": 0.0218, "step": 6473 }, { "epoch": 2.55, "learning_rate": 3.491381548049228e-07, "loss": 0.0184, "step": 6474 }, { "epoch": 2.55, "learning_rate": 3.491211410845772e-07, "loss": 0.0185, "step": 6475 }, { "epoch": 2.55, "learning_rate": 3.4910412493378373e-07, "loss": 0.019, "step": 6476 }, { "epoch": 2.55, "learning_rate": 3.4908710635281955e-07, "loss": 0.0204, "step": 6477 }, { "epoch": 2.55, "learning_rate": 3.4907008534196216e-07, "loss": 0.0195, "step": 6478 }, { "epoch": 2.55, "learning_rate": 3.4905306190148886e-07, "loss": 0.0217, "step": 6479 }, { "epoch": 2.55, "learning_rate": 3.4903603603167726e-07, "loss": 0.0184, "step": 6480 }, { "epoch": 2.55, "learning_rate": 3.490190077328048e-07, "loss": 0.023, "step": 6481 }, { "epoch": 2.55, "learning_rate": 3.4900197700514894e-07, "loss": 0.0234, "step": 6482 }, { "epoch": 2.56, "learning_rate": 3.489849438489873e-07, "loss": 0.0233, "step": 6483 }, { "epoch": 2.56, "learning_rate": 3.489679082645976e-07, "loss": 0.0238, "step": 6484 }, { "epoch": 2.56, "learning_rate": 3.489508702522573e-07, "loss": 0.0225, "step": 6485 }, { "epoch": 2.56, "learning_rate": 3.489338298122442e-07, "loss": 0.0199, "step": 6486 }, { "epoch": 2.56, "learning_rate": 3.4891678694483615e-07, "loss": 0.0229, "step": 6487 }, { "epoch": 2.56, "learning_rate": 3.488997416503107e-07, "loss": 0.0204, "step": 6488 }, { "epoch": 2.56, "learning_rate": 3.488826939289457e-07, "loss": 0.0234, "step": 6489 }, { "epoch": 2.56, "learning_rate": 3.488656437810192e-07, "loss": 0.0244, "step": 6490 }, { "epoch": 2.56, "learning_rate": 3.488485912068089e-07, "loss": 0.0225, "step": 6491 }, { "epoch": 2.56, "learning_rate": 3.4883153620659276e-07, "loss": 0.023, "step": 6492 }, { "epoch": 2.56, "learning_rate": 3.4881447878064875e-07, "loss": 0.0259, "step": 6493 }, { "epoch": 2.56, "learning_rate": 3.4879741892925497e-07, "loss": 0.0213, "step": 6494 }, { "epoch": 2.56, "learning_rate": 3.487803566526894e-07, "loss": 0.0188, "step": 6495 }, { "epoch": 2.56, "learning_rate": 3.4876329195123016e-07, "loss": 0.0259, "step": 6496 }, { "epoch": 2.56, "learning_rate": 3.487462248251553e-07, "loss": 0.0237, "step": 6497 }, { "epoch": 2.56, "learning_rate": 3.4872915527474306e-07, "loss": 0.0231, "step": 6498 }, { "epoch": 2.56, "learning_rate": 3.4871208330027163e-07, "loss": 0.0252, "step": 6499 }, { "epoch": 2.56, "learning_rate": 3.486950089020193e-07, "loss": 0.021, "step": 6500 }, { "epoch": 2.56, "learning_rate": 3.4867793208026435e-07, "loss": 0.0193, "step": 6501 }, { "epoch": 2.56, "learning_rate": 3.48660852835285e-07, "loss": 0.024, "step": 6502 }, { "epoch": 2.56, "learning_rate": 3.486437711673597e-07, "loss": 0.0281, "step": 6503 }, { "epoch": 2.56, "learning_rate": 3.486266870767668e-07, "loss": 0.024, "step": 6504 }, { "epoch": 2.56, "learning_rate": 3.486096005637848e-07, "loss": 0.0259, "step": 6505 }, { "epoch": 2.56, "learning_rate": 3.4859251162869224e-07, "loss": 0.0232, "step": 6506 }, { "epoch": 2.56, "learning_rate": 3.485754202717675e-07, "loss": 0.0215, "step": 6507 }, { "epoch": 2.56, "learning_rate": 3.4855832649328923e-07, "loss": 0.0202, "step": 6508 }, { "epoch": 2.57, "learning_rate": 3.48541230293536e-07, "loss": 0.0212, "step": 6509 }, { "epoch": 2.57, "learning_rate": 3.485241316727866e-07, "loss": 0.0175, "step": 6510 }, { "epoch": 2.57, "learning_rate": 3.4850703063131945e-07, "loss": 0.0235, "step": 6511 }, { "epoch": 2.57, "learning_rate": 3.4848992716941344e-07, "loss": 0.0224, "step": 6512 }, { "epoch": 2.57, "learning_rate": 3.484728212873473e-07, "loss": 0.0216, "step": 6513 }, { "epoch": 2.57, "learning_rate": 3.4845571298539985e-07, "loss": 0.0218, "step": 6514 }, { "epoch": 2.57, "learning_rate": 3.484386022638499e-07, "loss": 0.0287, "step": 6515 }, { "epoch": 2.57, "learning_rate": 3.484214891229763e-07, "loss": 0.0212, "step": 6516 }, { "epoch": 2.57, "learning_rate": 3.48404373563058e-07, "loss": 0.0237, "step": 6517 }, { "epoch": 2.57, "learning_rate": 3.4838725558437407e-07, "loss": 0.0248, "step": 6518 }, { "epoch": 2.57, "learning_rate": 3.483701351872033e-07, "loss": 0.0227, "step": 6519 }, { "epoch": 2.57, "learning_rate": 3.483530123718248e-07, "loss": 0.022, "step": 6520 }, { "epoch": 2.57, "learning_rate": 3.483358871385177e-07, "loss": 0.0212, "step": 6521 }, { "epoch": 2.57, "learning_rate": 3.483187594875611e-07, "loss": 0.0251, "step": 6522 }, { "epoch": 2.57, "learning_rate": 3.4830162941923413e-07, "loss": 0.017, "step": 6523 }, { "epoch": 2.57, "learning_rate": 3.4828449693381597e-07, "loss": 0.0192, "step": 6524 }, { "epoch": 2.57, "learning_rate": 3.482673620315859e-07, "loss": 0.0227, "step": 6525 }, { "epoch": 2.57, "learning_rate": 3.4825022471282315e-07, "loss": 0.0189, "step": 6526 }, { "epoch": 2.57, "learning_rate": 3.4823308497780706e-07, "loss": 0.0241, "step": 6527 }, { "epoch": 2.57, "learning_rate": 3.48215942826817e-07, "loss": 0.0183, "step": 6528 }, { "epoch": 2.57, "learning_rate": 3.481987982601323e-07, "loss": 0.0195, "step": 6529 }, { "epoch": 2.57, "learning_rate": 3.4818165127803247e-07, "loss": 0.0233, "step": 6530 }, { "epoch": 2.57, "learning_rate": 3.481645018807969e-07, "loss": 0.0228, "step": 6531 }, { "epoch": 2.57, "learning_rate": 3.481473500687051e-07, "loss": 0.024, "step": 6532 }, { "epoch": 2.57, "learning_rate": 3.481301958420367e-07, "loss": 0.0199, "step": 6533 }, { "epoch": 2.58, "learning_rate": 3.481130392010712e-07, "loss": 0.0216, "step": 6534 }, { "epoch": 2.58, "learning_rate": 3.4809588014608834e-07, "loss": 0.0177, "step": 6535 }, { "epoch": 2.58, "learning_rate": 3.480787186773677e-07, "loss": 0.0208, "step": 6536 }, { "epoch": 2.58, "learning_rate": 3.4806155479518895e-07, "loss": 0.02, "step": 6537 }, { "epoch": 2.58, "learning_rate": 3.480443884998319e-07, "loss": 0.0225, "step": 6538 }, { "epoch": 2.58, "learning_rate": 3.480272197915763e-07, "loss": 0.0227, "step": 6539 }, { "epoch": 2.58, "learning_rate": 3.48010048670702e-07, "loss": 0.0226, "step": 6540 }, { "epoch": 2.58, "learning_rate": 3.4799287513748883e-07, "loss": 0.0219, "step": 6541 }, { "epoch": 2.58, "learning_rate": 3.4797569919221685e-07, "loss": 0.0213, "step": 6542 }, { "epoch": 2.58, "learning_rate": 3.4795852083516573e-07, "loss": 0.0248, "step": 6543 }, { "epoch": 2.58, "learning_rate": 3.479413400666156e-07, "loss": 0.0232, "step": 6544 }, { "epoch": 2.58, "learning_rate": 3.479241568868465e-07, "loss": 0.023, "step": 6545 }, { "epoch": 2.58, "learning_rate": 3.4790697129613846e-07, "loss": 0.0232, "step": 6546 }, { "epoch": 2.58, "learning_rate": 3.4788978329477163e-07, "loss": 0.0265, "step": 6547 }, { "epoch": 2.58, "learning_rate": 3.4787259288302606e-07, "loss": 0.0251, "step": 6548 }, { "epoch": 2.58, "learning_rate": 3.4785540006118197e-07, "loss": 0.0201, "step": 6549 }, { "epoch": 2.58, "learning_rate": 3.478382048295196e-07, "loss": 0.0213, "step": 6550 }, { "epoch": 2.58, "learning_rate": 3.4782100718831914e-07, "loss": 0.0238, "step": 6551 }, { "epoch": 2.58, "learning_rate": 3.4780380713786094e-07, "loss": 0.0226, "step": 6552 }, { "epoch": 2.58, "learning_rate": 3.4778660467842536e-07, "loss": 0.0216, "step": 6553 }, { "epoch": 2.58, "learning_rate": 3.4776939981029267e-07, "loss": 0.0228, "step": 6554 }, { "epoch": 2.58, "learning_rate": 3.4775219253374334e-07, "loss": 0.0209, "step": 6555 }, { "epoch": 2.58, "learning_rate": 3.4773498284905786e-07, "loss": 0.0217, "step": 6556 }, { "epoch": 2.58, "learning_rate": 3.4771777075651663e-07, "loss": 0.0239, "step": 6557 }, { "epoch": 2.58, "learning_rate": 3.4770055625640034e-07, "loss": 0.0194, "step": 6558 }, { "epoch": 2.58, "learning_rate": 3.4768333934898937e-07, "loss": 0.0209, "step": 6559 }, { "epoch": 2.59, "learning_rate": 3.476661200345645e-07, "loss": 0.0169, "step": 6560 }, { "epoch": 2.59, "learning_rate": 3.4764889831340626e-07, "loss": 0.0244, "step": 6561 }, { "epoch": 2.59, "learning_rate": 3.4763167418579544e-07, "loss": 0.0229, "step": 6562 }, { "epoch": 2.59, "learning_rate": 3.476144476520126e-07, "loss": 0.0221, "step": 6563 }, { "epoch": 2.59, "learning_rate": 3.4759721871233876e-07, "loss": 0.0239, "step": 6564 }, { "epoch": 2.59, "learning_rate": 3.475799873670544e-07, "loss": 0.0248, "step": 6565 }, { "epoch": 2.59, "learning_rate": 3.475627536164406e-07, "loss": 0.0238, "step": 6566 }, { "epoch": 2.59, "learning_rate": 3.4754551746077825e-07, "loss": 0.0239, "step": 6567 }, { "epoch": 2.59, "learning_rate": 3.475282789003482e-07, "loss": 0.023, "step": 6568 }, { "epoch": 2.59, "learning_rate": 3.4751103793543134e-07, "loss": 0.0225, "step": 6569 }, { "epoch": 2.59, "learning_rate": 3.474937945663088e-07, "loss": 0.0204, "step": 6570 }, { "epoch": 2.59, "learning_rate": 3.474765487932616e-07, "loss": 0.0211, "step": 6571 }, { "epoch": 2.59, "learning_rate": 3.4745930061657076e-07, "loss": 0.023, "step": 6572 }, { "epoch": 2.59, "learning_rate": 3.474420500365174e-07, "loss": 0.0246, "step": 6573 }, { "epoch": 2.59, "learning_rate": 3.4742479705338283e-07, "loss": 0.0252, "step": 6574 }, { "epoch": 2.59, "learning_rate": 3.474075416674481e-07, "loss": 0.0246, "step": 6575 }, { "epoch": 2.59, "learning_rate": 3.473902838789943e-07, "loss": 0.0243, "step": 6576 }, { "epoch": 2.59, "learning_rate": 3.473730236883031e-07, "loss": 0.022, "step": 6577 }, { "epoch": 2.59, "learning_rate": 3.4735576109565547e-07, "loss": 0.0256, "step": 6578 }, { "epoch": 2.59, "learning_rate": 3.473384961013329e-07, "loss": 0.0203, "step": 6579 }, { "epoch": 2.59, "learning_rate": 3.473212287056168e-07, "loss": 0.0251, "step": 6580 }, { "epoch": 2.59, "learning_rate": 3.473039589087886e-07, "loss": 0.0223, "step": 6581 }, { "epoch": 2.59, "learning_rate": 3.4728668671112966e-07, "loss": 0.0238, "step": 6582 }, { "epoch": 2.59, "learning_rate": 3.4726941211292166e-07, "loss": 0.02, "step": 6583 }, { "epoch": 2.59, "learning_rate": 3.4725213511444607e-07, "loss": 0.0221, "step": 6584 }, { "epoch": 2.6, "learning_rate": 3.4723485571598435e-07, "loss": 0.0218, "step": 6585 }, { "epoch": 2.6, "learning_rate": 3.472175739178183e-07, "loss": 0.0236, "step": 6586 }, { "epoch": 2.6, "learning_rate": 3.4720028972022965e-07, "loss": 0.0233, "step": 6587 }, { "epoch": 2.6, "learning_rate": 3.471830031234999e-07, "loss": 0.0221, "step": 6588 }, { "epoch": 2.6, "learning_rate": 3.4716571412791094e-07, "loss": 0.0253, "step": 6589 }, { "epoch": 2.6, "learning_rate": 3.471484227337445e-07, "loss": 0.0201, "step": 6590 }, { "epoch": 2.6, "learning_rate": 3.471311289412824e-07, "loss": 0.0263, "step": 6591 }, { "epoch": 2.6, "learning_rate": 3.471138327508065e-07, "loss": 0.0246, "step": 6592 }, { "epoch": 2.6, "learning_rate": 3.470965341625987e-07, "loss": 0.0211, "step": 6593 }, { "epoch": 2.6, "learning_rate": 3.47079233176941e-07, "loss": 0.0257, "step": 6594 }, { "epoch": 2.6, "learning_rate": 3.470619297941153e-07, "loss": 0.0239, "step": 6595 }, { "epoch": 2.6, "learning_rate": 3.470446240144037e-07, "loss": 0.0178, "step": 6596 }, { "epoch": 2.6, "learning_rate": 3.470273158380881e-07, "loss": 0.0219, "step": 6597 }, { "epoch": 2.6, "learning_rate": 3.470100052654508e-07, "loss": 0.0196, "step": 6598 }, { "epoch": 2.6, "learning_rate": 3.469926922967738e-07, "loss": 0.0191, "step": 6599 }, { "epoch": 2.6, "learning_rate": 3.469753769323394e-07, "loss": 0.0265, "step": 6600 }, { "epoch": 2.6, "learning_rate": 3.469580591724297e-07, "loss": 0.0187, "step": 6601 }, { "epoch": 2.6, "learning_rate": 3.469407390173269e-07, "loss": 0.0218, "step": 6602 }, { "epoch": 2.6, "learning_rate": 3.4692341646731347e-07, "loss": 0.0186, "step": 6603 }, { "epoch": 2.6, "learning_rate": 3.4690609152267156e-07, "loss": 0.025, "step": 6604 }, { "epoch": 2.6, "learning_rate": 3.468887641836837e-07, "loss": 0.0214, "step": 6605 }, { "epoch": 2.6, "learning_rate": 3.468714344506322e-07, "loss": 0.0244, "step": 6606 }, { "epoch": 2.6, "learning_rate": 3.468541023237996e-07, "loss": 0.0229, "step": 6607 }, { "epoch": 2.6, "learning_rate": 3.4683676780346827e-07, "loss": 0.018, "step": 6608 }, { "epoch": 2.6, "learning_rate": 3.468194308899208e-07, "loss": 0.0245, "step": 6609 }, { "epoch": 2.61, "learning_rate": 3.4680209158343976e-07, "loss": 0.0219, "step": 6610 }, { "epoch": 2.61, "learning_rate": 3.467847498843077e-07, "loss": 0.0218, "step": 6611 }, { "epoch": 2.61, "learning_rate": 3.4676740579280736e-07, "loss": 0.0253, "step": 6612 }, { "epoch": 2.61, "learning_rate": 3.467500593092213e-07, "loss": 0.0241, "step": 6613 }, { "epoch": 2.61, "learning_rate": 3.4673271043383235e-07, "loss": 0.0214, "step": 6614 }, { "epoch": 2.61, "learning_rate": 3.467153591669232e-07, "loss": 0.0264, "step": 6615 }, { "epoch": 2.61, "learning_rate": 3.4669800550877676e-07, "loss": 0.0209, "step": 6616 }, { "epoch": 2.61, "learning_rate": 3.4668064945967573e-07, "loss": 0.0196, "step": 6617 }, { "epoch": 2.61, "learning_rate": 3.46663291019903e-07, "loss": 0.0167, "step": 6618 }, { "epoch": 2.61, "learning_rate": 3.466459301897416e-07, "loss": 0.0232, "step": 6619 }, { "epoch": 2.61, "learning_rate": 3.466285669694744e-07, "loss": 0.0169, "step": 6620 }, { "epoch": 2.61, "learning_rate": 3.4661120135938435e-07, "loss": 0.0208, "step": 6621 }, { "epoch": 2.61, "learning_rate": 3.4659383335975455e-07, "loss": 0.02, "step": 6622 }, { "epoch": 2.61, "learning_rate": 3.465764629708681e-07, "loss": 0.0239, "step": 6623 }, { "epoch": 2.61, "learning_rate": 3.465590901930081e-07, "loss": 0.0215, "step": 6624 }, { "epoch": 2.61, "learning_rate": 3.465417150264576e-07, "loss": 0.0233, "step": 6625 }, { "epoch": 2.61, "learning_rate": 3.4652433747149996e-07, "loss": 0.0182, "step": 6626 }, { "epoch": 2.61, "learning_rate": 3.4650695752841825e-07, "loss": 0.02, "step": 6627 }, { "epoch": 2.61, "learning_rate": 3.464895751974958e-07, "loss": 0.0194, "step": 6628 }, { "epoch": 2.61, "learning_rate": 3.464721904790159e-07, "loss": 0.0203, "step": 6629 }, { "epoch": 2.61, "learning_rate": 3.464548033732619e-07, "loss": 0.0239, "step": 6630 }, { "epoch": 2.61, "learning_rate": 3.464374138805172e-07, "loss": 0.0222, "step": 6631 }, { "epoch": 2.61, "learning_rate": 3.464200220010653e-07, "loss": 0.0252, "step": 6632 }, { "epoch": 2.61, "learning_rate": 3.464026277351895e-07, "loss": 0.0214, "step": 6633 }, { "epoch": 2.61, "learning_rate": 3.463852310831733e-07, "loss": 0.0202, "step": 6634 }, { "epoch": 2.61, "learning_rate": 3.463678320453004e-07, "loss": 0.023, "step": 6635 }, { "epoch": 2.62, "learning_rate": 3.4635043062185427e-07, "loss": 0.0207, "step": 6636 }, { "epoch": 2.62, "learning_rate": 3.4633302681311857e-07, "loss": 0.0199, "step": 6637 }, { "epoch": 2.62, "learning_rate": 3.463156206193769e-07, "loss": 0.024, "step": 6638 }, { "epoch": 2.62, "learning_rate": 3.46298212040913e-07, "loss": 0.0212, "step": 6639 }, { "epoch": 2.62, "learning_rate": 3.462808010780106e-07, "loss": 0.026, "step": 6640 }, { "epoch": 2.62, "learning_rate": 3.4626338773095347e-07, "loss": 0.0207, "step": 6641 }, { "epoch": 2.62, "learning_rate": 3.4624597200002543e-07, "loss": 0.02, "step": 6642 }, { "epoch": 2.62, "learning_rate": 3.462285538855103e-07, "loss": 0.022, "step": 6643 }, { "epoch": 2.62, "learning_rate": 3.4621113338769196e-07, "loss": 0.0234, "step": 6644 }, { "epoch": 2.62, "learning_rate": 3.4619371050685437e-07, "loss": 0.0252, "step": 6645 }, { "epoch": 2.62, "learning_rate": 3.461762852432815e-07, "loss": 0.023, "step": 6646 }, { "epoch": 2.62, "learning_rate": 3.4615885759725734e-07, "loss": 0.0237, "step": 6647 }, { "epoch": 2.62, "learning_rate": 3.4614142756906595e-07, "loss": 0.0216, "step": 6648 }, { "epoch": 2.62, "learning_rate": 3.461239951589914e-07, "loss": 0.0222, "step": 6649 }, { "epoch": 2.62, "learning_rate": 3.461065603673178e-07, "loss": 0.0209, "step": 6650 }, { "epoch": 2.62, "learning_rate": 3.460891231943293e-07, "loss": 0.0204, "step": 6651 }, { "epoch": 2.62, "learning_rate": 3.4607168364031015e-07, "loss": 0.0247, "step": 6652 }, { "epoch": 2.62, "learning_rate": 3.4605424170554455e-07, "loss": 0.022, "step": 6653 }, { "epoch": 2.62, "learning_rate": 3.460367973903168e-07, "loss": 0.0254, "step": 6654 }, { "epoch": 2.62, "learning_rate": 3.460193506949111e-07, "loss": 0.0224, "step": 6655 }, { "epoch": 2.62, "learning_rate": 3.460019016196121e-07, "loss": 0.0198, "step": 6656 }, { "epoch": 2.62, "learning_rate": 3.4598445016470384e-07, "loss": 0.0203, "step": 6657 }, { "epoch": 2.62, "learning_rate": 3.4596699633047094e-07, "loss": 0.0193, "step": 6658 }, { "epoch": 2.62, "learning_rate": 3.4594954011719794e-07, "loss": 0.0195, "step": 6659 }, { "epoch": 2.62, "learning_rate": 3.4593208152516917e-07, "loss": 0.0185, "step": 6660 }, { "epoch": 2.63, "learning_rate": 3.459146205546693e-07, "loss": 0.0192, "step": 6661 }, { "epoch": 2.63, "learning_rate": 3.4589715720598285e-07, "loss": 0.024, "step": 6662 }, { "epoch": 2.63, "learning_rate": 3.458796914793945e-07, "loss": 0.021, "step": 6663 }, { "epoch": 2.63, "learning_rate": 3.4586222337518887e-07, "loss": 0.0217, "step": 6664 }, { "epoch": 2.63, "learning_rate": 3.4584475289365063e-07, "loss": 0.0212, "step": 6665 }, { "epoch": 2.63, "learning_rate": 3.458272800350647e-07, "loss": 0.0218, "step": 6666 }, { "epoch": 2.63, "learning_rate": 3.458098047997157e-07, "loss": 0.0189, "step": 6667 }, { "epoch": 2.63, "learning_rate": 3.457923271878884e-07, "loss": 0.0217, "step": 6668 }, { "epoch": 2.63, "learning_rate": 3.457748471998678e-07, "loss": 0.0218, "step": 6669 }, { "epoch": 2.63, "learning_rate": 3.457573648359388e-07, "loss": 0.0236, "step": 6670 }, { "epoch": 2.63, "learning_rate": 3.457398800963862e-07, "loss": 0.0243, "step": 6671 }, { "epoch": 2.63, "learning_rate": 3.457223929814951e-07, "loss": 0.0223, "step": 6672 }, { "epoch": 2.63, "learning_rate": 3.4570490349155045e-07, "loss": 0.0235, "step": 6673 }, { "epoch": 2.63, "learning_rate": 3.4568741162683736e-07, "loss": 0.0236, "step": 6674 }, { "epoch": 2.63, "learning_rate": 3.456699173876408e-07, "loss": 0.0217, "step": 6675 }, { "epoch": 2.63, "learning_rate": 3.4565242077424605e-07, "loss": 0.0254, "step": 6676 }, { "epoch": 2.63, "learning_rate": 3.4563492178693815e-07, "loss": 0.0253, "step": 6677 }, { "epoch": 2.63, "learning_rate": 3.4561742042600236e-07, "loss": 0.0208, "step": 6678 }, { "epoch": 2.63, "learning_rate": 3.45599916691724e-07, "loss": 0.0211, "step": 6679 }, { "epoch": 2.63, "learning_rate": 3.455824105843882e-07, "loss": 0.0247, "step": 6680 }, { "epoch": 2.63, "learning_rate": 3.455649021042804e-07, "loss": 0.0242, "step": 6681 }, { "epoch": 2.63, "learning_rate": 3.4554739125168597e-07, "loss": 0.0214, "step": 6682 }, { "epoch": 2.63, "learning_rate": 3.4552987802689026e-07, "loss": 0.0231, "step": 6683 }, { "epoch": 2.63, "learning_rate": 3.4551236243017867e-07, "loss": 0.0221, "step": 6684 }, { "epoch": 2.63, "learning_rate": 3.454948444618368e-07, "loss": 0.0176, "step": 6685 }, { "epoch": 2.64, "learning_rate": 3.4547732412215e-07, "loss": 0.0187, "step": 6686 }, { "epoch": 2.64, "learning_rate": 3.45459801411404e-07, "loss": 0.0252, "step": 6687 }, { "epoch": 2.64, "learning_rate": 3.454422763298843e-07, "loss": 0.0212, "step": 6688 }, { "epoch": 2.64, "learning_rate": 3.4542474887787643e-07, "loss": 0.0193, "step": 6689 }, { "epoch": 2.64, "learning_rate": 3.4540721905566626e-07, "loss": 0.0254, "step": 6690 }, { "epoch": 2.64, "learning_rate": 3.453896868635394e-07, "loss": 0.022, "step": 6691 }, { "epoch": 2.64, "learning_rate": 3.453721523017817e-07, "loss": 0.0216, "step": 6692 }, { "epoch": 2.64, "learning_rate": 3.453546153706788e-07, "loss": 0.0188, "step": 6693 }, { "epoch": 2.64, "learning_rate": 3.453370760705165e-07, "loss": 0.0203, "step": 6694 }, { "epoch": 2.64, "learning_rate": 3.4531953440158084e-07, "loss": 0.0261, "step": 6695 }, { "epoch": 2.64, "learning_rate": 3.453019903641576e-07, "loss": 0.0242, "step": 6696 }, { "epoch": 2.64, "learning_rate": 3.452844439585327e-07, "loss": 0.0238, "step": 6697 }, { "epoch": 2.64, "learning_rate": 3.4526689518499224e-07, "loss": 0.0214, "step": 6698 }, { "epoch": 2.64, "learning_rate": 3.4524934404382215e-07, "loss": 0.0275, "step": 6699 }, { "epoch": 2.64, "learning_rate": 3.452317905353086e-07, "loss": 0.0205, "step": 6700 }, { "epoch": 2.64, "learning_rate": 3.452142346597374e-07, "loss": 0.02, "step": 6701 }, { "epoch": 2.64, "learning_rate": 3.45196676417395e-07, "loss": 0.0198, "step": 6702 }, { "epoch": 2.64, "learning_rate": 3.451791158085674e-07, "loss": 0.0212, "step": 6703 }, { "epoch": 2.64, "learning_rate": 3.4516155283354095e-07, "loss": 0.0256, "step": 6704 }, { "epoch": 2.64, "learning_rate": 3.451439874926017e-07, "loss": 0.0243, "step": 6705 }, { "epoch": 2.64, "learning_rate": 3.451264197860361e-07, "loss": 0.024, "step": 6706 }, { "epoch": 2.64, "learning_rate": 3.4510884971413036e-07, "loss": 0.0182, "step": 6707 }, { "epoch": 2.64, "learning_rate": 3.45091277277171e-07, "loss": 0.0207, "step": 6708 }, { "epoch": 2.64, "learning_rate": 3.450737024754443e-07, "loss": 0.0207, "step": 6709 }, { "epoch": 2.64, "learning_rate": 3.450561253092367e-07, "loss": 0.021, "step": 6710 }, { "epoch": 2.64, "learning_rate": 3.450385457788347e-07, "loss": 0.023, "step": 6711 }, { "epoch": 2.65, "learning_rate": 3.4502096388452496e-07, "loss": 0.0181, "step": 6712 }, { "epoch": 2.65, "learning_rate": 3.450033796265938e-07, "loss": 0.0238, "step": 6713 }, { "epoch": 2.65, "learning_rate": 3.44985793005328e-07, "loss": 0.0184, "step": 6714 }, { "epoch": 2.65, "learning_rate": 3.449682040210141e-07, "loss": 0.0226, "step": 6715 }, { "epoch": 2.65, "learning_rate": 3.449506126739388e-07, "loss": 0.0268, "step": 6716 }, { "epoch": 2.65, "learning_rate": 3.449330189643888e-07, "loss": 0.0199, "step": 6717 }, { "epoch": 2.65, "learning_rate": 3.4491542289265087e-07, "loss": 0.0203, "step": 6718 }, { "epoch": 2.65, "learning_rate": 3.4489782445901183e-07, "loss": 0.0235, "step": 6719 }, { "epoch": 2.65, "learning_rate": 3.448802236637584e-07, "loss": 0.0205, "step": 6720 }, { "epoch": 2.65, "learning_rate": 3.4486262050717755e-07, "loss": 0.0238, "step": 6721 }, { "epoch": 2.65, "learning_rate": 3.448450149895562e-07, "loss": 0.0233, "step": 6722 }, { "epoch": 2.65, "learning_rate": 3.448274071111812e-07, "loss": 0.0194, "step": 6723 }, { "epoch": 2.65, "learning_rate": 3.448097968723395e-07, "loss": 0.0229, "step": 6724 }, { "epoch": 2.65, "learning_rate": 3.4479218427331836e-07, "loss": 0.0233, "step": 6725 }, { "epoch": 2.65, "learning_rate": 3.4477456931440454e-07, "loss": 0.0237, "step": 6726 }, { "epoch": 2.65, "learning_rate": 3.447569519958853e-07, "loss": 0.0217, "step": 6727 }, { "epoch": 2.65, "learning_rate": 3.4473933231804776e-07, "loss": 0.0191, "step": 6728 }, { "epoch": 2.65, "learning_rate": 3.447217102811791e-07, "loss": 0.0214, "step": 6729 }, { "epoch": 2.65, "learning_rate": 3.4470408588556655e-07, "loss": 0.0208, "step": 6730 }, { "epoch": 2.65, "learning_rate": 3.446864591314973e-07, "loss": 0.0208, "step": 6731 }, { "epoch": 2.65, "learning_rate": 3.4466883001925864e-07, "loss": 0.0209, "step": 6732 }, { "epoch": 2.65, "learning_rate": 3.4465119854913793e-07, "loss": 0.0228, "step": 6733 }, { "epoch": 2.65, "learning_rate": 3.4463356472142246e-07, "loss": 0.0245, "step": 6734 }, { "epoch": 2.65, "learning_rate": 3.446159285363998e-07, "loss": 0.0216, "step": 6735 }, { "epoch": 2.65, "learning_rate": 3.4459828999435727e-07, "loss": 0.019, "step": 6736 }, { "epoch": 2.66, "learning_rate": 3.445806490955824e-07, "loss": 0.0241, "step": 6737 }, { "epoch": 2.66, "learning_rate": 3.445630058403626e-07, "loss": 0.0239, "step": 6738 }, { "epoch": 2.66, "learning_rate": 3.4454536022898556e-07, "loss": 0.0211, "step": 6739 }, { "epoch": 2.66, "learning_rate": 3.445277122617389e-07, "loss": 0.0226, "step": 6740 }, { "epoch": 2.66, "learning_rate": 3.4451006193891015e-07, "loss": 0.0214, "step": 6741 }, { "epoch": 2.66, "learning_rate": 3.44492409260787e-07, "loss": 0.0201, "step": 6742 }, { "epoch": 2.66, "learning_rate": 3.4447475422765715e-07, "loss": 0.0221, "step": 6743 }, { "epoch": 2.66, "learning_rate": 3.444570968398084e-07, "loss": 0.0229, "step": 6744 }, { "epoch": 2.66, "learning_rate": 3.4443943709752854e-07, "loss": 0.0206, "step": 6745 }, { "epoch": 2.66, "learning_rate": 3.444217750011054e-07, "loss": 0.0246, "step": 6746 }, { "epoch": 2.66, "learning_rate": 3.444041105508268e-07, "loss": 0.0251, "step": 6747 }, { "epoch": 2.66, "learning_rate": 3.4438644374698066e-07, "loss": 0.0193, "step": 6748 }, { "epoch": 2.66, "learning_rate": 3.443687745898549e-07, "loss": 0.0213, "step": 6749 }, { "epoch": 2.66, "learning_rate": 3.443511030797376e-07, "loss": 0.0184, "step": 6750 }, { "epoch": 2.66, "learning_rate": 3.443334292169167e-07, "loss": 0.021, "step": 6751 }, { "epoch": 2.66, "learning_rate": 3.443157530016802e-07, "loss": 0.0195, "step": 6752 }, { "epoch": 2.66, "learning_rate": 3.442980744343163e-07, "loss": 0.0177, "step": 6753 }, { "epoch": 2.66, "learning_rate": 3.442803935151131e-07, "loss": 0.0209, "step": 6754 }, { "epoch": 2.66, "learning_rate": 3.4426271024435876e-07, "loss": 0.0311, "step": 6755 }, { "epoch": 2.66, "learning_rate": 3.442450246223415e-07, "loss": 0.0199, "step": 6756 }, { "epoch": 2.66, "learning_rate": 3.4422733664934957e-07, "loss": 0.0264, "step": 6757 }, { "epoch": 2.66, "learning_rate": 3.442096463256712e-07, "loss": 0.0232, "step": 6758 }, { "epoch": 2.66, "learning_rate": 3.441919536515948e-07, "loss": 0.0191, "step": 6759 }, { "epoch": 2.66, "learning_rate": 3.441742586274087e-07, "loss": 0.0181, "step": 6760 }, { "epoch": 2.66, "learning_rate": 3.441565612534014e-07, "loss": 0.0216, "step": 6761 }, { "epoch": 2.66, "learning_rate": 3.4413886152986115e-07, "loss": 0.0226, "step": 6762 }, { "epoch": 2.67, "learning_rate": 3.441211594570765e-07, "loss": 0.0223, "step": 6763 }, { "epoch": 2.67, "learning_rate": 3.44103455035336e-07, "loss": 0.0259, "step": 6764 }, { "epoch": 2.67, "learning_rate": 3.4408574826492823e-07, "loss": 0.0226, "step": 6765 }, { "epoch": 2.67, "learning_rate": 3.440680391461418e-07, "loss": 0.022, "step": 6766 }, { "epoch": 2.67, "learning_rate": 3.4405032767926524e-07, "loss": 0.0199, "step": 6767 }, { "epoch": 2.67, "learning_rate": 3.440326138645872e-07, "loss": 0.0197, "step": 6768 }, { "epoch": 2.67, "learning_rate": 3.440148977023965e-07, "loss": 0.0181, "step": 6769 }, { "epoch": 2.67, "learning_rate": 3.4399717919298195e-07, "loss": 0.0234, "step": 6770 }, { "epoch": 2.67, "learning_rate": 3.4397945833663215e-07, "loss": 0.0201, "step": 6771 }, { "epoch": 2.67, "learning_rate": 3.43961735133636e-07, "loss": 0.0218, "step": 6772 }, { "epoch": 2.67, "learning_rate": 3.439440095842823e-07, "loss": 0.023, "step": 6773 }, { "epoch": 2.67, "learning_rate": 3.4392628168886007e-07, "loss": 0.02, "step": 6774 }, { "epoch": 2.67, "learning_rate": 3.439085514476582e-07, "loss": 0.0187, "step": 6775 }, { "epoch": 2.67, "learning_rate": 3.4389081886096563e-07, "loss": 0.0216, "step": 6776 }, { "epoch": 2.67, "learning_rate": 3.4387308392907146e-07, "loss": 0.021, "step": 6777 }, { "epoch": 2.67, "learning_rate": 3.438553466522646e-07, "loss": 0.0221, "step": 6778 }, { "epoch": 2.67, "learning_rate": 3.4383760703083426e-07, "loss": 0.0226, "step": 6779 }, { "epoch": 2.67, "learning_rate": 3.438198650650696e-07, "loss": 0.0235, "step": 6780 }, { "epoch": 2.67, "learning_rate": 3.438021207552597e-07, "loss": 0.022, "step": 6781 }, { "epoch": 2.67, "learning_rate": 3.4378437410169373e-07, "loss": 0.0216, "step": 6782 }, { "epoch": 2.67, "learning_rate": 3.4376662510466103e-07, "loss": 0.0198, "step": 6783 }, { "epoch": 2.67, "learning_rate": 3.437488737644508e-07, "loss": 0.0193, "step": 6784 }, { "epoch": 2.67, "learning_rate": 3.437311200813524e-07, "loss": 0.0271, "step": 6785 }, { "epoch": 2.67, "learning_rate": 3.4371336405565517e-07, "loss": 0.0214, "step": 6786 }, { "epoch": 2.67, "learning_rate": 3.436956056876486e-07, "loss": 0.021, "step": 6787 }, { "epoch": 2.68, "learning_rate": 3.43677844977622e-07, "loss": 0.0202, "step": 6788 }, { "epoch": 2.68, "learning_rate": 3.4366008192586495e-07, "loss": 0.0252, "step": 6789 }, { "epoch": 2.68, "learning_rate": 3.4364231653266684e-07, "loss": 0.0241, "step": 6790 }, { "epoch": 2.68, "learning_rate": 3.436245487983173e-07, "loss": 0.0217, "step": 6791 }, { "epoch": 2.68, "learning_rate": 3.4360677872310594e-07, "loss": 0.0238, "step": 6792 }, { "epoch": 2.68, "learning_rate": 3.4358900630732226e-07, "loss": 0.0231, "step": 6793 }, { "epoch": 2.68, "learning_rate": 3.4357123155125605e-07, "loss": 0.0225, "step": 6794 }, { "epoch": 2.68, "learning_rate": 3.4355345445519695e-07, "loss": 0.0177, "step": 6795 }, { "epoch": 2.68, "learning_rate": 3.435356750194348e-07, "loss": 0.0185, "step": 6796 }, { "epoch": 2.68, "learning_rate": 3.435178932442592e-07, "loss": 0.0246, "step": 6797 }, { "epoch": 2.68, "learning_rate": 3.435001091299601e-07, "loss": 0.0211, "step": 6798 }, { "epoch": 2.68, "learning_rate": 3.434823226768273e-07, "loss": 0.0206, "step": 6799 }, { "epoch": 2.68, "learning_rate": 3.434645338851507e-07, "loss": 0.0212, "step": 6800 }, { "epoch": 2.68, "learning_rate": 3.434467427552203e-07, "loss": 0.0215, "step": 6801 }, { "epoch": 2.68, "learning_rate": 3.43428949287326e-07, "loss": 0.0208, "step": 6802 }, { "epoch": 2.68, "learning_rate": 3.4341115348175775e-07, "loss": 0.0254, "step": 6803 }, { "epoch": 2.68, "learning_rate": 3.4339335533880566e-07, "loss": 0.0208, "step": 6804 }, { "epoch": 2.68, "learning_rate": 3.4337555485875987e-07, "loss": 0.02, "step": 6805 }, { "epoch": 2.68, "learning_rate": 3.4335775204191037e-07, "loss": 0.0255, "step": 6806 }, { "epoch": 2.68, "learning_rate": 3.433399468885475e-07, "loss": 0.0203, "step": 6807 }, { "epoch": 2.68, "learning_rate": 3.433221393989612e-07, "loss": 0.0203, "step": 6808 }, { "epoch": 2.68, "learning_rate": 3.4330432957344193e-07, "loss": 0.0178, "step": 6809 }, { "epoch": 2.68, "learning_rate": 3.432865174122799e-07, "loss": 0.022, "step": 6810 }, { "epoch": 2.68, "learning_rate": 3.432687029157654e-07, "loss": 0.0217, "step": 6811 }, { "epoch": 2.68, "learning_rate": 3.432508860841888e-07, "loss": 0.0205, "step": 6812 }, { "epoch": 2.69, "learning_rate": 3.4323306691784044e-07, "loss": 0.0201, "step": 6813 }, { "epoch": 2.69, "learning_rate": 3.4321524541701077e-07, "loss": 0.0202, "step": 6814 }, { "epoch": 2.69, "learning_rate": 3.431974215819903e-07, "loss": 0.0224, "step": 6815 }, { "epoch": 2.69, "learning_rate": 3.431795954130695e-07, "loss": 0.0239, "step": 6816 }, { "epoch": 2.69, "learning_rate": 3.431617669105389e-07, "loss": 0.0243, "step": 6817 }, { "epoch": 2.69, "learning_rate": 3.43143936074689e-07, "loss": 0.0193, "step": 6818 }, { "epoch": 2.69, "learning_rate": 3.4312610290581055e-07, "loss": 0.0202, "step": 6819 }, { "epoch": 2.69, "learning_rate": 3.431082674041942e-07, "loss": 0.0193, "step": 6820 }, { "epoch": 2.69, "learning_rate": 3.430904295701305e-07, "loss": 0.0203, "step": 6821 }, { "epoch": 2.69, "learning_rate": 3.4307258940391034e-07, "loss": 0.0221, "step": 6822 }, { "epoch": 2.69, "learning_rate": 3.4305474690582445e-07, "loss": 0.0211, "step": 6823 }, { "epoch": 2.69, "learning_rate": 3.430369020761635e-07, "loss": 0.0247, "step": 6824 }, { "epoch": 2.69, "learning_rate": 3.4301905491521843e-07, "loss": 0.0252, "step": 6825 }, { "epoch": 2.69, "learning_rate": 3.430012054232803e-07, "loss": 0.0237, "step": 6826 }, { "epoch": 2.69, "learning_rate": 3.429833536006397e-07, "loss": 0.0211, "step": 6827 }, { "epoch": 2.69, "learning_rate": 3.4296549944758776e-07, "loss": 0.018, "step": 6828 }, { "epoch": 2.69, "learning_rate": 3.429476429644155e-07, "loss": 0.025, "step": 6829 }, { "epoch": 2.69, "learning_rate": 3.4292978415141394e-07, "loss": 0.0231, "step": 6830 }, { "epoch": 2.69, "learning_rate": 3.42911923008874e-07, "loss": 0.0232, "step": 6831 }, { "epoch": 2.69, "learning_rate": 3.4289405953708705e-07, "loss": 0.0209, "step": 6832 }, { "epoch": 2.69, "learning_rate": 3.428761937363441e-07, "loss": 0.018, "step": 6833 }, { "epoch": 2.69, "learning_rate": 3.4285832560693627e-07, "loss": 0.0191, "step": 6834 }, { "epoch": 2.69, "learning_rate": 3.4284045514915485e-07, "loss": 0.0232, "step": 6835 }, { "epoch": 2.69, "learning_rate": 3.4282258236329115e-07, "loss": 0.0211, "step": 6836 }, { "epoch": 2.69, "learning_rate": 3.428047072496364e-07, "loss": 0.0281, "step": 6837 }, { "epoch": 2.69, "learning_rate": 3.42786829808482e-07, "loss": 0.0236, "step": 6838 }, { "epoch": 2.7, "learning_rate": 3.4276895004011923e-07, "loss": 0.0178, "step": 6839 }, { "epoch": 2.7, "learning_rate": 3.427510679448396e-07, "loss": 0.0237, "step": 6840 }, { "epoch": 2.7, "learning_rate": 3.427331835229345e-07, "loss": 0.0225, "step": 6841 }, { "epoch": 2.7, "learning_rate": 3.427152967746954e-07, "loss": 0.0209, "step": 6842 }, { "epoch": 2.7, "learning_rate": 3.426974077004139e-07, "loss": 0.0203, "step": 6843 }, { "epoch": 2.7, "learning_rate": 3.4267951630038153e-07, "loss": 0.0177, "step": 6844 }, { "epoch": 2.7, "learning_rate": 3.426616225748899e-07, "loss": 0.0217, "step": 6845 }, { "epoch": 2.7, "learning_rate": 3.426437265242306e-07, "loss": 0.0238, "step": 6846 }, { "epoch": 2.7, "learning_rate": 3.426258281486954e-07, "loss": 0.0267, "step": 6847 }, { "epoch": 2.7, "learning_rate": 3.4260792744857596e-07, "loss": 0.0225, "step": 6848 }, { "epoch": 2.7, "learning_rate": 3.4259002442416406e-07, "loss": 0.0256, "step": 6849 }, { "epoch": 2.7, "learning_rate": 3.4257211907575147e-07, "loss": 0.021, "step": 6850 }, { "epoch": 2.7, "learning_rate": 3.4255421140363e-07, "loss": 0.0183, "step": 6851 }, { "epoch": 2.7, "learning_rate": 3.4253630140809155e-07, "loss": 0.0214, "step": 6852 }, { "epoch": 2.7, "learning_rate": 3.4251838908942794e-07, "loss": 0.0212, "step": 6853 }, { "epoch": 2.7, "learning_rate": 3.4250047444793133e-07, "loss": 0.0225, "step": 6854 }, { "epoch": 2.7, "learning_rate": 3.4248255748389345e-07, "loss": 0.023, "step": 6855 }, { "epoch": 2.7, "learning_rate": 3.424646381976065e-07, "loss": 0.0194, "step": 6856 }, { "epoch": 2.7, "learning_rate": 3.424467165893625e-07, "loss": 0.0224, "step": 6857 }, { "epoch": 2.7, "learning_rate": 3.424287926594534e-07, "loss": 0.0244, "step": 6858 }, { "epoch": 2.7, "learning_rate": 3.424108664081715e-07, "loss": 0.0208, "step": 6859 }, { "epoch": 2.7, "learning_rate": 3.4239293783580896e-07, "loss": 0.0265, "step": 6860 }, { "epoch": 2.7, "learning_rate": 3.423750069426579e-07, "loss": 0.022, "step": 6861 }, { "epoch": 2.7, "learning_rate": 3.423570737290106e-07, "loss": 0.0202, "step": 6862 }, { "epoch": 2.7, "learning_rate": 3.423391381951594e-07, "loss": 0.0246, "step": 6863 }, { "epoch": 2.71, "learning_rate": 3.4232120034139655e-07, "loss": 0.0215, "step": 6864 }, { "epoch": 2.71, "learning_rate": 3.4230326016801445e-07, "loss": 0.0216, "step": 6865 }, { "epoch": 2.71, "learning_rate": 3.4228531767530547e-07, "loss": 0.0189, "step": 6866 }, { "epoch": 2.71, "learning_rate": 3.4226737286356206e-07, "loss": 0.0227, "step": 6867 }, { "epoch": 2.71, "learning_rate": 3.422494257330767e-07, "loss": 0.0211, "step": 6868 }, { "epoch": 2.71, "learning_rate": 3.4223147628414194e-07, "loss": 0.0219, "step": 6869 }, { "epoch": 2.71, "learning_rate": 3.422135245170502e-07, "loss": 0.0206, "step": 6870 }, { "epoch": 2.71, "learning_rate": 3.4219557043209417e-07, "loss": 0.024, "step": 6871 }, { "epoch": 2.71, "learning_rate": 3.421776140295665e-07, "loss": 0.0216, "step": 6872 }, { "epoch": 2.71, "learning_rate": 3.421596553097598e-07, "loss": 0.0198, "step": 6873 }, { "epoch": 2.71, "learning_rate": 3.421416942729667e-07, "loss": 0.0209, "step": 6874 }, { "epoch": 2.71, "learning_rate": 3.4212373091948007e-07, "loss": 0.0211, "step": 6875 }, { "epoch": 2.71, "learning_rate": 3.421057652495927e-07, "loss": 0.0199, "step": 6876 }, { "epoch": 2.71, "learning_rate": 3.4208779726359725e-07, "loss": 0.0239, "step": 6877 }, { "epoch": 2.71, "learning_rate": 3.4206982696178664e-07, "loss": 0.0219, "step": 6878 }, { "epoch": 2.71, "learning_rate": 3.420518543444538e-07, "loss": 0.0227, "step": 6879 }, { "epoch": 2.71, "learning_rate": 3.420338794118916e-07, "loss": 0.0193, "step": 6880 }, { "epoch": 2.71, "learning_rate": 3.420159021643931e-07, "loss": 0.0221, "step": 6881 }, { "epoch": 2.71, "learning_rate": 3.419979226022512e-07, "loss": 0.0231, "step": 6882 }, { "epoch": 2.71, "learning_rate": 3.4197994072575894e-07, "loss": 0.0222, "step": 6883 }, { "epoch": 2.71, "learning_rate": 3.419619565352095e-07, "loss": 0.0176, "step": 6884 }, { "epoch": 2.71, "learning_rate": 3.419439700308959e-07, "loss": 0.0202, "step": 6885 }, { "epoch": 2.71, "learning_rate": 3.419259812131113e-07, "loss": 0.0263, "step": 6886 }, { "epoch": 2.71, "learning_rate": 3.419079900821489e-07, "loss": 0.0208, "step": 6887 }, { "epoch": 2.71, "learning_rate": 3.4188999663830194e-07, "loss": 0.0214, "step": 6888 }, { "epoch": 2.72, "learning_rate": 3.4187200088186366e-07, "loss": 0.0236, "step": 6889 }, { "epoch": 2.72, "learning_rate": 3.418540028131274e-07, "loss": 0.0227, "step": 6890 }, { "epoch": 2.72, "learning_rate": 3.4183600243238655e-07, "loss": 0.0223, "step": 6891 }, { "epoch": 2.72, "learning_rate": 3.418179997399344e-07, "loss": 0.023, "step": 6892 }, { "epoch": 2.72, "learning_rate": 3.417999947360643e-07, "loss": 0.0215, "step": 6893 }, { "epoch": 2.72, "learning_rate": 3.417819874210699e-07, "loss": 0.0243, "step": 6894 }, { "epoch": 2.72, "learning_rate": 3.417639777952446e-07, "loss": 0.0187, "step": 6895 }, { "epoch": 2.72, "learning_rate": 3.417459658588818e-07, "loss": 0.022, "step": 6896 }, { "epoch": 2.72, "learning_rate": 3.417279516122753e-07, "loss": 0.0198, "step": 6897 }, { "epoch": 2.72, "learning_rate": 3.417099350557185e-07, "loss": 0.0226, "step": 6898 }, { "epoch": 2.72, "learning_rate": 3.4169191618950525e-07, "loss": 0.0205, "step": 6899 }, { "epoch": 2.72, "learning_rate": 3.4167389501392904e-07, "loss": 0.0239, "step": 6900 }, { "epoch": 2.72, "learning_rate": 3.4165587152928363e-07, "loss": 0.0222, "step": 6901 }, { "epoch": 2.72, "learning_rate": 3.416378457358629e-07, "loss": 0.0216, "step": 6902 }, { "epoch": 2.72, "learning_rate": 3.4161981763396047e-07, "loss": 0.0164, "step": 6903 }, { "epoch": 2.72, "learning_rate": 3.4160178722387024e-07, "loss": 0.0224, "step": 6904 }, { "epoch": 2.72, "learning_rate": 3.4158375450588617e-07, "loss": 0.0255, "step": 6905 }, { "epoch": 2.72, "learning_rate": 3.4156571948030203e-07, "loss": 0.0233, "step": 6906 }, { "epoch": 2.72, "learning_rate": 3.415476821474118e-07, "loss": 0.0239, "step": 6907 }, { "epoch": 2.72, "learning_rate": 3.415296425075095e-07, "loss": 0.0234, "step": 6908 }, { "epoch": 2.72, "learning_rate": 3.4151160056088915e-07, "loss": 0.0186, "step": 6909 }, { "epoch": 2.72, "learning_rate": 3.4149355630784476e-07, "loss": 0.0224, "step": 6910 }, { "epoch": 2.72, "learning_rate": 3.414755097486704e-07, "loss": 0.0194, "step": 6911 }, { "epoch": 2.72, "learning_rate": 3.414574608836604e-07, "loss": 0.0221, "step": 6912 }, { "epoch": 2.72, "learning_rate": 3.414394097131086e-07, "loss": 0.0229, "step": 6913 }, { "epoch": 2.72, "learning_rate": 3.4142135623730945e-07, "loss": 0.0222, "step": 6914 }, { "epoch": 2.73, "learning_rate": 3.414033004565572e-07, "loss": 0.0234, "step": 6915 }, { "epoch": 2.73, "learning_rate": 3.41385242371146e-07, "loss": 0.0206, "step": 6916 }, { "epoch": 2.73, "learning_rate": 3.413671819813702e-07, "loss": 0.0206, "step": 6917 }, { "epoch": 2.73, "learning_rate": 3.413491192875242e-07, "loss": 0.0206, "step": 6918 }, { "epoch": 2.73, "learning_rate": 3.4133105428990244e-07, "loss": 0.0229, "step": 6919 }, { "epoch": 2.73, "learning_rate": 3.4131298698879927e-07, "loss": 0.0206, "step": 6920 }, { "epoch": 2.73, "learning_rate": 3.412949173845092e-07, "loss": 0.0228, "step": 6921 }, { "epoch": 2.73, "learning_rate": 3.4127684547732666e-07, "loss": 0.0217, "step": 6922 }, { "epoch": 2.73, "learning_rate": 3.412587712675463e-07, "loss": 0.0214, "step": 6923 }, { "epoch": 2.73, "learning_rate": 3.4124069475546263e-07, "loss": 0.0196, "step": 6924 }, { "epoch": 2.73, "learning_rate": 3.412226159413703e-07, "loss": 0.0249, "step": 6925 }, { "epoch": 2.73, "learning_rate": 3.41204534825564e-07, "loss": 0.0266, "step": 6926 }, { "epoch": 2.73, "learning_rate": 3.411864514083383e-07, "loss": 0.0229, "step": 6927 }, { "epoch": 2.73, "learning_rate": 3.411683656899881e-07, "loss": 0.0232, "step": 6928 }, { "epoch": 2.73, "learning_rate": 3.41150277670808e-07, "loss": 0.0223, "step": 6929 }, { "epoch": 2.73, "learning_rate": 3.41132187351093e-07, "loss": 0.0221, "step": 6930 }, { "epoch": 2.73, "learning_rate": 3.411140947311378e-07, "loss": 0.0207, "step": 6931 }, { "epoch": 2.73, "learning_rate": 3.410959998112374e-07, "loss": 0.0219, "step": 6932 }, { "epoch": 2.73, "learning_rate": 3.4107790259168655e-07, "loss": 0.02, "step": 6933 }, { "epoch": 2.73, "learning_rate": 3.4105980307278034e-07, "loss": 0.025, "step": 6934 }, { "epoch": 2.73, "learning_rate": 3.4104170125481373e-07, "loss": 0.0211, "step": 6935 }, { "epoch": 2.73, "learning_rate": 3.4102359713808177e-07, "loss": 0.0228, "step": 6936 }, { "epoch": 2.73, "learning_rate": 3.4100549072287947e-07, "loss": 0.0185, "step": 6937 }, { "epoch": 2.73, "learning_rate": 3.4098738200950193e-07, "loss": 0.0258, "step": 6938 }, { "epoch": 2.73, "learning_rate": 3.409692709982444e-07, "loss": 0.0248, "step": 6939 }, { "epoch": 2.74, "learning_rate": 3.40951157689402e-07, "loss": 0.0243, "step": 6940 }, { "epoch": 2.74, "learning_rate": 3.409330420832699e-07, "loss": 0.0181, "step": 6941 }, { "epoch": 2.74, "learning_rate": 3.409149241801435e-07, "loss": 0.0231, "step": 6942 }, { "epoch": 2.74, "learning_rate": 3.408968039803179e-07, "loss": 0.0225, "step": 6943 }, { "epoch": 2.74, "learning_rate": 3.4087868148408865e-07, "loss": 0.0254, "step": 6944 }, { "epoch": 2.74, "learning_rate": 3.408605566917509e-07, "loss": 0.0184, "step": 6945 }, { "epoch": 2.74, "learning_rate": 3.408424296036002e-07, "loss": 0.0183, "step": 6946 }, { "epoch": 2.74, "learning_rate": 3.4082430021993193e-07, "loss": 0.0197, "step": 6947 }, { "epoch": 2.74, "learning_rate": 3.408061685410416e-07, "loss": 0.0232, "step": 6948 }, { "epoch": 2.74, "learning_rate": 3.4078803456722476e-07, "loss": 0.0199, "step": 6949 }, { "epoch": 2.74, "learning_rate": 3.407698982987769e-07, "loss": 0.0241, "step": 6950 }, { "epoch": 2.74, "learning_rate": 3.4075175973599366e-07, "loss": 0.0228, "step": 6951 }, { "epoch": 2.74, "learning_rate": 3.4073361887917065e-07, "loss": 0.0219, "step": 6952 }, { "epoch": 2.74, "learning_rate": 3.407154757286035e-07, "loss": 0.0187, "step": 6953 }, { "epoch": 2.74, "learning_rate": 3.4069733028458806e-07, "loss": 0.0227, "step": 6954 }, { "epoch": 2.74, "learning_rate": 3.4067918254741987e-07, "loss": 0.024, "step": 6955 }, { "epoch": 2.74, "learning_rate": 3.4066103251739485e-07, "loss": 0.0214, "step": 6956 }, { "epoch": 2.74, "learning_rate": 3.4064288019480877e-07, "loss": 0.0212, "step": 6957 }, { "epoch": 2.74, "learning_rate": 3.4062472557995745e-07, "loss": 0.021, "step": 6958 }, { "epoch": 2.74, "learning_rate": 3.4060656867313695e-07, "loss": 0.026, "step": 6959 }, { "epoch": 2.74, "learning_rate": 3.4058840947464297e-07, "loss": 0.0195, "step": 6960 }, { "epoch": 2.74, "learning_rate": 3.4057024798477166e-07, "loss": 0.0224, "step": 6961 }, { "epoch": 2.74, "learning_rate": 3.405520842038189e-07, "loss": 0.022, "step": 6962 }, { "epoch": 2.74, "learning_rate": 3.4053391813208083e-07, "loss": 0.022, "step": 6963 }, { "epoch": 2.74, "learning_rate": 3.4051574976985344e-07, "loss": 0.0231, "step": 6964 }, { "epoch": 2.74, "learning_rate": 3.404975791174329e-07, "loss": 0.0201, "step": 6965 }, { "epoch": 2.75, "learning_rate": 3.4047940617511533e-07, "loss": 0.0198, "step": 6966 }, { "epoch": 2.75, "learning_rate": 3.4046123094319695e-07, "loss": 0.0209, "step": 6967 }, { "epoch": 2.75, "learning_rate": 3.40443053421974e-07, "loss": 0.0195, "step": 6968 }, { "epoch": 2.75, "learning_rate": 3.4042487361174273e-07, "loss": 0.0198, "step": 6969 }, { "epoch": 2.75, "learning_rate": 3.4040669151279944e-07, "loss": 0.0213, "step": 6970 }, { "epoch": 2.75, "learning_rate": 3.403885071254404e-07, "loss": 0.0271, "step": 6971 }, { "epoch": 2.75, "learning_rate": 3.403703204499622e-07, "loss": 0.0249, "step": 6972 }, { "epoch": 2.75, "learning_rate": 3.4035213148666103e-07, "loss": 0.0236, "step": 6973 }, { "epoch": 2.75, "learning_rate": 3.4033394023583343e-07, "loss": 0.0176, "step": 6974 }, { "epoch": 2.75, "learning_rate": 3.403157466977759e-07, "loss": 0.0267, "step": 6975 }, { "epoch": 2.75, "learning_rate": 3.40297550872785e-07, "loss": 0.0219, "step": 6976 }, { "epoch": 2.75, "learning_rate": 3.4027935276115715e-07, "loss": 0.0226, "step": 6977 }, { "epoch": 2.75, "learning_rate": 3.402611523631891e-07, "loss": 0.024, "step": 6978 }, { "epoch": 2.75, "learning_rate": 3.402429496791774e-07, "loss": 0.0197, "step": 6979 }, { "epoch": 2.75, "learning_rate": 3.402247447094188e-07, "loss": 0.0225, "step": 6980 }, { "epoch": 2.75, "learning_rate": 3.4020653745420997e-07, "loss": 0.0243, "step": 6981 }, { "epoch": 2.75, "learning_rate": 3.4018832791384775e-07, "loss": 0.0222, "step": 6982 }, { "epoch": 2.75, "learning_rate": 3.401701160886287e-07, "loss": 0.0216, "step": 6983 }, { "epoch": 2.75, "learning_rate": 3.401519019788499e-07, "loss": 0.0209, "step": 6984 }, { "epoch": 2.75, "learning_rate": 3.401336855848081e-07, "loss": 0.0229, "step": 6985 }, { "epoch": 2.75, "learning_rate": 3.401154669068002e-07, "loss": 0.0222, "step": 6986 }, { "epoch": 2.75, "learning_rate": 3.4009724594512306e-07, "loss": 0.0251, "step": 6987 }, { "epoch": 2.75, "learning_rate": 3.400790227000738e-07, "loss": 0.021, "step": 6988 }, { "epoch": 2.75, "learning_rate": 3.4006079717194935e-07, "loss": 0.0199, "step": 6989 }, { "epoch": 2.75, "learning_rate": 3.400425693610468e-07, "loss": 0.0263, "step": 6990 }, { "epoch": 2.76, "learning_rate": 3.4002433926766317e-07, "loss": 0.0186, "step": 6991 }, { "epoch": 2.76, "learning_rate": 3.400061068920956e-07, "loss": 0.0233, "step": 6992 }, { "epoch": 2.76, "learning_rate": 3.3998787223464133e-07, "loss": 0.0248, "step": 6993 }, { "epoch": 2.76, "learning_rate": 3.399696352955975e-07, "loss": 0.0221, "step": 6994 }, { "epoch": 2.76, "learning_rate": 3.3995139607526137e-07, "loss": 0.0215, "step": 6995 }, { "epoch": 2.76, "learning_rate": 3.399331545739301e-07, "loss": 0.0208, "step": 6996 }, { "epoch": 2.76, "learning_rate": 3.3991491079190116e-07, "loss": 0.0275, "step": 6997 }, { "epoch": 2.76, "learning_rate": 3.398966647294718e-07, "loss": 0.019, "step": 6998 }, { "epoch": 2.76, "learning_rate": 3.398784163869394e-07, "loss": 0.0196, "step": 6999 }, { "epoch": 2.76, "learning_rate": 3.3986016576460144e-07, "loss": 0.0224, "step": 7000 }, { "epoch": 2.76, "learning_rate": 3.3984191286275526e-07, "loss": 0.0208, "step": 7001 }, { "epoch": 2.76, "learning_rate": 3.398236576816985e-07, "loss": 0.0232, "step": 7002 }, { "epoch": 2.76, "learning_rate": 3.3980540022172857e-07, "loss": 0.0232, "step": 7003 }, { "epoch": 2.76, "learning_rate": 3.397871404831432e-07, "loss": 0.0188, "step": 7004 }, { "epoch": 2.76, "learning_rate": 3.3976887846623985e-07, "loss": 0.0191, "step": 7005 }, { "epoch": 2.76, "learning_rate": 3.397506141713162e-07, "loss": 0.0177, "step": 7006 }, { "epoch": 2.76, "learning_rate": 3.397323475986699e-07, "loss": 0.0177, "step": 7007 }, { "epoch": 2.76, "learning_rate": 3.3971407874859876e-07, "loss": 0.021, "step": 7008 }, { "epoch": 2.76, "learning_rate": 3.396958076214004e-07, "loss": 0.0236, "step": 7009 }, { "epoch": 2.76, "learning_rate": 3.396775342173727e-07, "loss": 0.0253, "step": 7010 }, { "epoch": 2.76, "learning_rate": 3.396592585368135e-07, "loss": 0.0213, "step": 7011 }, { "epoch": 2.76, "learning_rate": 3.396409805800207e-07, "loss": 0.0234, "step": 7012 }, { "epoch": 2.76, "learning_rate": 3.396227003472921e-07, "loss": 0.0183, "step": 7013 }, { "epoch": 2.76, "learning_rate": 3.396044178389257e-07, "loss": 0.0205, "step": 7014 }, { "epoch": 2.76, "learning_rate": 3.395861330552195e-07, "loss": 0.0192, "step": 7015 }, { "epoch": 2.77, "learning_rate": 3.395678459964714e-07, "loss": 0.0205, "step": 7016 }, { "epoch": 2.77, "learning_rate": 3.395495566629796e-07, "loss": 0.0247, "step": 7017 }, { "epoch": 2.77, "learning_rate": 3.39531265055042e-07, "loss": 0.0206, "step": 7018 }, { "epoch": 2.77, "learning_rate": 3.3951297117295696e-07, "loss": 0.0224, "step": 7019 }, { "epoch": 2.77, "learning_rate": 3.3949467501702245e-07, "loss": 0.0171, "step": 7020 }, { "epoch": 2.77, "learning_rate": 3.3947637658753684e-07, "loss": 0.0186, "step": 7021 }, { "epoch": 2.77, "learning_rate": 3.394580758847982e-07, "loss": 0.0239, "step": 7022 }, { "epoch": 2.77, "learning_rate": 3.3943977290910484e-07, "loss": 0.0219, "step": 7023 }, { "epoch": 2.77, "learning_rate": 3.394214676607552e-07, "loss": 0.0239, "step": 7024 }, { "epoch": 2.77, "learning_rate": 3.394031601400475e-07, "loss": 0.023, "step": 7025 }, { "epoch": 2.77, "learning_rate": 3.3938485034728015e-07, "loss": 0.0194, "step": 7026 }, { "epoch": 2.77, "learning_rate": 3.3936653828275155e-07, "loss": 0.0201, "step": 7027 }, { "epoch": 2.77, "learning_rate": 3.3934822394676016e-07, "loss": 0.0202, "step": 7028 }, { "epoch": 2.77, "learning_rate": 3.3932990733960456e-07, "loss": 0.0208, "step": 7029 }, { "epoch": 2.77, "learning_rate": 3.393115884615833e-07, "loss": 0.0224, "step": 7030 }, { "epoch": 2.77, "learning_rate": 3.392932673129947e-07, "loss": 0.0222, "step": 7031 }, { "epoch": 2.77, "learning_rate": 3.392749438941377e-07, "loss": 0.023, "step": 7032 }, { "epoch": 2.77, "learning_rate": 3.392566182053107e-07, "loss": 0.0204, "step": 7033 }, { "epoch": 2.77, "learning_rate": 3.3923829024681254e-07, "loss": 0.0232, "step": 7034 }, { "epoch": 2.77, "learning_rate": 3.392199600189419e-07, "loss": 0.0264, "step": 7035 }, { "epoch": 2.77, "learning_rate": 3.392016275219974e-07, "loss": 0.019, "step": 7036 }, { "epoch": 2.77, "learning_rate": 3.3918329275627795e-07, "loss": 0.023, "step": 7037 }, { "epoch": 2.77, "learning_rate": 3.391649557220823e-07, "loss": 0.0214, "step": 7038 }, { "epoch": 2.77, "learning_rate": 3.391466164197095e-07, "loss": 0.0289, "step": 7039 }, { "epoch": 2.77, "learning_rate": 3.391282748494583e-07, "loss": 0.0217, "step": 7040 }, { "epoch": 2.77, "learning_rate": 3.391099310116277e-07, "loss": 0.0255, "step": 7041 }, { "epoch": 2.78, "learning_rate": 3.390915849065165e-07, "loss": 0.0269, "step": 7042 }, { "epoch": 2.78, "learning_rate": 3.3907323653442397e-07, "loss": 0.0234, "step": 7043 }, { "epoch": 2.78, "learning_rate": 3.390548858956491e-07, "loss": 0.0197, "step": 7044 }, { "epoch": 2.78, "learning_rate": 3.3903653299049086e-07, "loss": 0.0215, "step": 7045 }, { "epoch": 2.78, "learning_rate": 3.3901817781924845e-07, "loss": 0.0212, "step": 7046 }, { "epoch": 2.78, "learning_rate": 3.38999820382221e-07, "loss": 0.0223, "step": 7047 }, { "epoch": 2.78, "learning_rate": 3.389814606797078e-07, "loss": 0.0206, "step": 7048 }, { "epoch": 2.78, "learning_rate": 3.3896309871200803e-07, "loss": 0.0224, "step": 7049 }, { "epoch": 2.78, "learning_rate": 3.3894473447942087e-07, "loss": 0.0269, "step": 7050 }, { "epoch": 2.78, "learning_rate": 3.3892636798224575e-07, "loss": 0.0206, "step": 7051 }, { "epoch": 2.78, "learning_rate": 3.3890799922078195e-07, "loss": 0.0234, "step": 7052 }, { "epoch": 2.78, "learning_rate": 3.388896281953289e-07, "loss": 0.0187, "step": 7053 }, { "epoch": 2.78, "learning_rate": 3.3887125490618603e-07, "loss": 0.0229, "step": 7054 }, { "epoch": 2.78, "learning_rate": 3.388528793536527e-07, "loss": 0.0237, "step": 7055 }, { "epoch": 2.78, "learning_rate": 3.388345015380285e-07, "loss": 0.0216, "step": 7056 }, { "epoch": 2.78, "learning_rate": 3.388161214596129e-07, "loss": 0.0217, "step": 7057 }, { "epoch": 2.78, "learning_rate": 3.3879773911870557e-07, "loss": 0.0224, "step": 7058 }, { "epoch": 2.78, "learning_rate": 3.38779354515606e-07, "loss": 0.0217, "step": 7059 }, { "epoch": 2.78, "learning_rate": 3.3876096765061385e-07, "loss": 0.0244, "step": 7060 }, { "epoch": 2.78, "learning_rate": 3.387425785240289e-07, "loss": 0.0225, "step": 7061 }, { "epoch": 2.78, "learning_rate": 3.387241871361507e-07, "loss": 0.0238, "step": 7062 }, { "epoch": 2.78, "learning_rate": 3.387057934872791e-07, "loss": 0.0176, "step": 7063 }, { "epoch": 2.78, "learning_rate": 3.386873975777139e-07, "loss": 0.0229, "step": 7064 }, { "epoch": 2.78, "learning_rate": 3.3866899940775485e-07, "loss": 0.0253, "step": 7065 }, { "epoch": 2.78, "learning_rate": 3.386505989777019e-07, "loss": 0.0195, "step": 7066 }, { "epoch": 2.79, "learning_rate": 3.3863219628785486e-07, "loss": 0.0198, "step": 7067 }, { "epoch": 2.79, "learning_rate": 3.3861379133851383e-07, "loss": 0.0238, "step": 7068 }, { "epoch": 2.79, "learning_rate": 3.385953841299786e-07, "loss": 0.0208, "step": 7069 }, { "epoch": 2.79, "learning_rate": 3.3857697466254926e-07, "loss": 0.0224, "step": 7070 }, { "epoch": 2.79, "learning_rate": 3.3855856293652585e-07, "loss": 0.0194, "step": 7071 }, { "epoch": 2.79, "learning_rate": 3.385401489522084e-07, "loss": 0.0205, "step": 7072 }, { "epoch": 2.79, "learning_rate": 3.385217327098971e-07, "loss": 0.0189, "step": 7073 }, { "epoch": 2.79, "learning_rate": 3.3850331420989214e-07, "loss": 0.0214, "step": 7074 }, { "epoch": 2.79, "learning_rate": 3.3848489345249364e-07, "loss": 0.0184, "step": 7075 }, { "epoch": 2.79, "learning_rate": 3.384664704380018e-07, "loss": 0.0221, "step": 7076 }, { "epoch": 2.79, "learning_rate": 3.38448045166717e-07, "loss": 0.02, "step": 7077 }, { "epoch": 2.79, "learning_rate": 3.384296176389394e-07, "loss": 0.0209, "step": 7078 }, { "epoch": 2.79, "learning_rate": 3.384111878549695e-07, "loss": 0.0255, "step": 7079 }, { "epoch": 2.79, "learning_rate": 3.383927558151075e-07, "loss": 0.018, "step": 7080 }, { "epoch": 2.79, "learning_rate": 3.3837432151965406e-07, "loss": 0.0199, "step": 7081 }, { "epoch": 2.79, "learning_rate": 3.383558849689094e-07, "loss": 0.0247, "step": 7082 }, { "epoch": 2.79, "learning_rate": 3.3833744616317407e-07, "loss": 0.0226, "step": 7083 }, { "epoch": 2.79, "learning_rate": 3.383190051027487e-07, "loss": 0.0211, "step": 7084 }, { "epoch": 2.79, "learning_rate": 3.383005617879337e-07, "loss": 0.0221, "step": 7085 }, { "epoch": 2.79, "learning_rate": 3.3828211621902973e-07, "loss": 0.0212, "step": 7086 }, { "epoch": 2.79, "learning_rate": 3.3826366839633744e-07, "loss": 0.0204, "step": 7087 }, { "epoch": 2.79, "learning_rate": 3.382452183201575e-07, "loss": 0.0203, "step": 7088 }, { "epoch": 2.79, "learning_rate": 3.382267659907906e-07, "loss": 0.0221, "step": 7089 }, { "epoch": 2.79, "learning_rate": 3.382083114085375e-07, "loss": 0.0249, "step": 7090 }, { "epoch": 2.79, "learning_rate": 3.3818985457369895e-07, "loss": 0.0188, "step": 7091 }, { "epoch": 2.8, "learning_rate": 3.3817139548657585e-07, "loss": 0.0195, "step": 7092 }, { "epoch": 2.8, "learning_rate": 3.3815293414746896e-07, "loss": 0.0199, "step": 7093 }, { "epoch": 2.8, "learning_rate": 3.3813447055667925e-07, "loss": 0.0229, "step": 7094 }, { "epoch": 2.8, "learning_rate": 3.381160047145076e-07, "loss": 0.0196, "step": 7095 }, { "epoch": 2.8, "learning_rate": 3.3809753662125494e-07, "loss": 0.0208, "step": 7096 }, { "epoch": 2.8, "learning_rate": 3.3807906627722236e-07, "loss": 0.0203, "step": 7097 }, { "epoch": 2.8, "learning_rate": 3.380605936827108e-07, "loss": 0.0208, "step": 7098 }, { "epoch": 2.8, "learning_rate": 3.380421188380215e-07, "loss": 0.0235, "step": 7099 }, { "epoch": 2.8, "learning_rate": 3.3802364174345535e-07, "loss": 0.0189, "step": 7100 }, { "epoch": 2.8, "learning_rate": 3.380051623993137e-07, "loss": 0.0248, "step": 7101 }, { "epoch": 2.8, "learning_rate": 3.379866808058976e-07, "loss": 0.022, "step": 7102 }, { "epoch": 2.8, "learning_rate": 3.379681969635083e-07, "loss": 0.0223, "step": 7103 }, { "epoch": 2.8, "learning_rate": 3.3794971087244715e-07, "loss": 0.0182, "step": 7104 }, { "epoch": 2.8, "learning_rate": 3.3793122253301533e-07, "loss": 0.0193, "step": 7105 }, { "epoch": 2.8, "learning_rate": 3.379127319455142e-07, "loss": 0.0246, "step": 7106 }, { "epoch": 2.8, "learning_rate": 3.378942391102451e-07, "loss": 0.0185, "step": 7107 }, { "epoch": 2.8, "learning_rate": 3.3787574402750954e-07, "loss": 0.0212, "step": 7108 }, { "epoch": 2.8, "learning_rate": 3.3785724669760887e-07, "loss": 0.0181, "step": 7109 }, { "epoch": 2.8, "learning_rate": 3.378387471208446e-07, "loss": 0.0209, "step": 7110 }, { "epoch": 2.8, "learning_rate": 3.3782024529751824e-07, "loss": 0.0234, "step": 7111 }, { "epoch": 2.8, "learning_rate": 3.378017412279314e-07, "loss": 0.0214, "step": 7112 }, { "epoch": 2.8, "learning_rate": 3.3778323491238554e-07, "loss": 0.0214, "step": 7113 }, { "epoch": 2.8, "learning_rate": 3.377647263511824e-07, "loss": 0.0242, "step": 7114 }, { "epoch": 2.8, "learning_rate": 3.3774621554462365e-07, "loss": 0.0232, "step": 7115 }, { "epoch": 2.8, "learning_rate": 3.377277024930109e-07, "loss": 0.0207, "step": 7116 }, { "epoch": 2.8, "learning_rate": 3.377091871966459e-07, "loss": 0.0205, "step": 7117 }, { "epoch": 2.81, "learning_rate": 3.376906696558304e-07, "loss": 0.0258, "step": 7118 }, { "epoch": 2.81, "learning_rate": 3.376721498708663e-07, "loss": 0.0257, "step": 7119 }, { "epoch": 2.81, "learning_rate": 3.376536278420554e-07, "loss": 0.0165, "step": 7120 }, { "epoch": 2.81, "learning_rate": 3.376351035696995e-07, "loss": 0.0215, "step": 7121 }, { "epoch": 2.81, "learning_rate": 3.376165770541007e-07, "loss": 0.0223, "step": 7122 }, { "epoch": 2.81, "learning_rate": 3.375980482955608e-07, "loss": 0.0221, "step": 7123 }, { "epoch": 2.81, "learning_rate": 3.3757951729438183e-07, "loss": 0.0174, "step": 7124 }, { "epoch": 2.81, "learning_rate": 3.3756098405086585e-07, "loss": 0.0222, "step": 7125 }, { "epoch": 2.81, "learning_rate": 3.3754244856531485e-07, "loss": 0.0228, "step": 7126 }, { "epoch": 2.81, "learning_rate": 3.3752391083803107e-07, "loss": 0.017, "step": 7127 }, { "epoch": 2.81, "learning_rate": 3.375053708693165e-07, "loss": 0.0203, "step": 7128 }, { "epoch": 2.81, "learning_rate": 3.374868286594733e-07, "loss": 0.023, "step": 7129 }, { "epoch": 2.81, "learning_rate": 3.374682842088038e-07, "loss": 0.027, "step": 7130 }, { "epoch": 2.81, "learning_rate": 3.3744973751761023e-07, "loss": 0.0204, "step": 7131 }, { "epoch": 2.81, "learning_rate": 3.3743118858619486e-07, "loss": 0.0181, "step": 7132 }, { "epoch": 2.81, "learning_rate": 3.374126374148599e-07, "loss": 0.0208, "step": 7133 }, { "epoch": 2.81, "learning_rate": 3.373940840039078e-07, "loss": 0.0209, "step": 7134 }, { "epoch": 2.81, "learning_rate": 3.37375528353641e-07, "loss": 0.0196, "step": 7135 }, { "epoch": 2.81, "learning_rate": 3.373569704643619e-07, "loss": 0.0208, "step": 7136 }, { "epoch": 2.81, "learning_rate": 3.3733841033637286e-07, "loss": 0.0296, "step": 7137 }, { "epoch": 2.81, "learning_rate": 3.3731984796997653e-07, "loss": 0.0226, "step": 7138 }, { "epoch": 2.81, "learning_rate": 3.3730128336547537e-07, "loss": 0.02, "step": 7139 }, { "epoch": 2.81, "learning_rate": 3.3728271652317193e-07, "loss": 0.0234, "step": 7140 }, { "epoch": 2.81, "learning_rate": 3.3726414744336886e-07, "loss": 0.0234, "step": 7141 }, { "epoch": 2.81, "learning_rate": 3.372455761263688e-07, "loss": 0.0189, "step": 7142 }, { "epoch": 2.82, "learning_rate": 3.372270025724745e-07, "loss": 0.0255, "step": 7143 }, { "epoch": 2.82, "learning_rate": 3.372084267819886e-07, "loss": 0.0251, "step": 7144 }, { "epoch": 2.82, "learning_rate": 3.371898487552138e-07, "loss": 0.0193, "step": 7145 }, { "epoch": 2.82, "learning_rate": 3.3717126849245303e-07, "loss": 0.0236, "step": 7146 }, { "epoch": 2.82, "learning_rate": 3.3715268599400907e-07, "loss": 0.0208, "step": 7147 }, { "epoch": 2.82, "learning_rate": 3.3713410126018475e-07, "loss": 0.0242, "step": 7148 }, { "epoch": 2.82, "learning_rate": 3.3711551429128295e-07, "loss": 0.0184, "step": 7149 }, { "epoch": 2.82, "learning_rate": 3.370969250876068e-07, "loss": 0.0196, "step": 7150 }, { "epoch": 2.82, "learning_rate": 3.3707833364945895e-07, "loss": 0.0225, "step": 7151 }, { "epoch": 2.82, "learning_rate": 3.3705973997714275e-07, "loss": 0.0207, "step": 7152 }, { "epoch": 2.82, "learning_rate": 3.37041144070961e-07, "loss": 0.0195, "step": 7153 }, { "epoch": 2.82, "learning_rate": 3.3702254593121696e-07, "loss": 0.0217, "step": 7154 }, { "epoch": 2.82, "learning_rate": 3.3700394555821364e-07, "loss": 0.02, "step": 7155 }, { "epoch": 2.82, "learning_rate": 3.369853429522542e-07, "loss": 0.0205, "step": 7156 }, { "epoch": 2.82, "learning_rate": 3.369667381136419e-07, "loss": 0.0217, "step": 7157 }, { "epoch": 2.82, "learning_rate": 3.3694813104267993e-07, "loss": 0.0224, "step": 7158 }, { "epoch": 2.82, "learning_rate": 3.3692952173967157e-07, "loss": 0.0215, "step": 7159 }, { "epoch": 2.82, "learning_rate": 3.3691091020492007e-07, "loss": 0.0181, "step": 7160 }, { "epoch": 2.82, "learning_rate": 3.3689229643872883e-07, "loss": 0.0191, "step": 7161 }, { "epoch": 2.82, "learning_rate": 3.368736804414012e-07, "loss": 0.0219, "step": 7162 }, { "epoch": 2.82, "learning_rate": 3.3685506221324064e-07, "loss": 0.0178, "step": 7163 }, { "epoch": 2.82, "learning_rate": 3.3683644175455055e-07, "loss": 0.0216, "step": 7164 }, { "epoch": 2.82, "learning_rate": 3.368178190656344e-07, "loss": 0.0244, "step": 7165 }, { "epoch": 2.82, "learning_rate": 3.3679919414679577e-07, "loss": 0.0236, "step": 7166 }, { "epoch": 2.82, "learning_rate": 3.3678056699833814e-07, "loss": 0.0213, "step": 7167 }, { "epoch": 2.82, "learning_rate": 3.3676193762056523e-07, "loss": 0.0212, "step": 7168 }, { "epoch": 2.83, "learning_rate": 3.3674330601378055e-07, "loss": 0.0237, "step": 7169 }, { "epoch": 2.83, "learning_rate": 3.3672467217828786e-07, "loss": 0.0227, "step": 7170 }, { "epoch": 2.83, "learning_rate": 3.367060361143907e-07, "loss": 0.0211, "step": 7171 }, { "epoch": 2.83, "learning_rate": 3.36687397822393e-07, "loss": 0.0225, "step": 7172 }, { "epoch": 2.83, "learning_rate": 3.3666875730259834e-07, "loss": 0.023, "step": 7173 }, { "epoch": 2.83, "learning_rate": 3.366501145553108e-07, "loss": 0.0237, "step": 7174 }, { "epoch": 2.83, "learning_rate": 3.3663146958083405e-07, "loss": 0.0201, "step": 7175 }, { "epoch": 2.83, "learning_rate": 3.3661282237947194e-07, "loss": 0.0269, "step": 7176 }, { "epoch": 2.83, "learning_rate": 3.3659417295152846e-07, "loss": 0.0219, "step": 7177 }, { "epoch": 2.83, "learning_rate": 3.365755212973076e-07, "loss": 0.0223, "step": 7178 }, { "epoch": 2.83, "learning_rate": 3.3655686741711334e-07, "loss": 0.0238, "step": 7179 }, { "epoch": 2.83, "learning_rate": 3.365382113112496e-07, "loss": 0.022, "step": 7180 }, { "epoch": 2.83, "learning_rate": 3.365195529800206e-07, "loss": 0.0254, "step": 7181 }, { "epoch": 2.83, "learning_rate": 3.365008924237303e-07, "loss": 0.0208, "step": 7182 }, { "epoch": 2.83, "learning_rate": 3.3648222964268296e-07, "loss": 0.0205, "step": 7183 }, { "epoch": 2.83, "learning_rate": 3.3646356463718275e-07, "loss": 0.0253, "step": 7184 }, { "epoch": 2.83, "learning_rate": 3.3644489740753375e-07, "loss": 0.025, "step": 7185 }, { "epoch": 2.83, "learning_rate": 3.3642622795404034e-07, "loss": 0.0233, "step": 7186 }, { "epoch": 2.83, "learning_rate": 3.3640755627700673e-07, "loss": 0.0239, "step": 7187 }, { "epoch": 2.83, "learning_rate": 3.3638888237673726e-07, "loss": 0.0244, "step": 7188 }, { "epoch": 2.83, "learning_rate": 3.3637020625353633e-07, "loss": 0.0196, "step": 7189 }, { "epoch": 2.83, "learning_rate": 3.363515279077083e-07, "loss": 0.0192, "step": 7190 }, { "epoch": 2.83, "learning_rate": 3.363328473395576e-07, "loss": 0.0235, "step": 7191 }, { "epoch": 2.83, "learning_rate": 3.363141645493887e-07, "loss": 0.0237, "step": 7192 }, { "epoch": 2.83, "learning_rate": 3.36295479537506e-07, "loss": 0.0211, "step": 7193 }, { "epoch": 2.84, "learning_rate": 3.362767923042142e-07, "loss": 0.0256, "step": 7194 }, { "epoch": 2.84, "learning_rate": 3.3625810284981785e-07, "loss": 0.0203, "step": 7195 }, { "epoch": 2.84, "learning_rate": 3.362394111746214e-07, "loss": 0.0231, "step": 7196 }, { "epoch": 2.84, "learning_rate": 3.3622071727892973e-07, "loss": 0.0203, "step": 7197 }, { "epoch": 2.84, "learning_rate": 3.362020211630473e-07, "loss": 0.0227, "step": 7198 }, { "epoch": 2.84, "learning_rate": 3.36183322827279e-07, "loss": 0.0215, "step": 7199 }, { "epoch": 2.84, "learning_rate": 3.361646222719295e-07, "loss": 0.0258, "step": 7200 }, { "epoch": 2.84, "learning_rate": 3.3614591949730355e-07, "loss": 0.0193, "step": 7201 }, { "epoch": 2.84, "learning_rate": 3.36127214503706e-07, "loss": 0.0191, "step": 7202 }, { "epoch": 2.84, "learning_rate": 3.3610850729144185e-07, "loss": 0.0202, "step": 7203 }, { "epoch": 2.84, "learning_rate": 3.3608979786081586e-07, "loss": 0.0194, "step": 7204 }, { "epoch": 2.84, "learning_rate": 3.36071086212133e-07, "loss": 0.0183, "step": 7205 }, { "epoch": 2.84, "learning_rate": 3.360523723456982e-07, "loss": 0.0207, "step": 7206 }, { "epoch": 2.84, "learning_rate": 3.3603365626181656e-07, "loss": 0.022, "step": 7207 }, { "epoch": 2.84, "learning_rate": 3.36014937960793e-07, "loss": 0.0203, "step": 7208 }, { "epoch": 2.84, "learning_rate": 3.359962174429327e-07, "loss": 0.0203, "step": 7209 }, { "epoch": 2.84, "learning_rate": 3.359774947085408e-07, "loss": 0.0228, "step": 7210 }, { "epoch": 2.84, "learning_rate": 3.359587697579223e-07, "loss": 0.0215, "step": 7211 }, { "epoch": 2.84, "learning_rate": 3.359400425913826e-07, "loss": 0.019, "step": 7212 }, { "epoch": 2.84, "learning_rate": 3.359213132092267e-07, "loss": 0.0188, "step": 7213 }, { "epoch": 2.84, "learning_rate": 3.3590258161176005e-07, "loss": 0.0254, "step": 7214 }, { "epoch": 2.84, "learning_rate": 3.3588384779928786e-07, "loss": 0.02, "step": 7215 }, { "epoch": 2.84, "learning_rate": 3.3586511177211543e-07, "loss": 0.0212, "step": 7216 }, { "epoch": 2.84, "learning_rate": 3.3584637353054814e-07, "loss": 0.0173, "step": 7217 }, { "epoch": 2.84, "learning_rate": 3.3582763307489153e-07, "loss": 0.0232, "step": 7218 }, { "epoch": 2.85, "learning_rate": 3.358088904054509e-07, "loss": 0.0172, "step": 7219 }, { "epoch": 2.85, "learning_rate": 3.357901455225317e-07, "loss": 0.0223, "step": 7220 }, { "epoch": 2.85, "learning_rate": 3.3577139842643957e-07, "loss": 0.0228, "step": 7221 }, { "epoch": 2.85, "learning_rate": 3.357526491174799e-07, "loss": 0.0206, "step": 7222 }, { "epoch": 2.85, "learning_rate": 3.357338975959585e-07, "loss": 0.0203, "step": 7223 }, { "epoch": 2.85, "learning_rate": 3.357151438621808e-07, "loss": 0.0193, "step": 7224 }, { "epoch": 2.85, "learning_rate": 3.3569638791645253e-07, "loss": 0.0234, "step": 7225 }, { "epoch": 2.85, "learning_rate": 3.3567762975907935e-07, "loss": 0.0245, "step": 7226 }, { "epoch": 2.85, "learning_rate": 3.3565886939036705e-07, "loss": 0.0199, "step": 7227 }, { "epoch": 2.85, "learning_rate": 3.356401068106213e-07, "loss": 0.0174, "step": 7228 }, { "epoch": 2.85, "learning_rate": 3.35621342020148e-07, "loss": 0.021, "step": 7229 }, { "epoch": 2.85, "learning_rate": 3.356025750192529e-07, "loss": 0.0154, "step": 7230 }, { "epoch": 2.85, "learning_rate": 3.35583805808242e-07, "loss": 0.0185, "step": 7231 }, { "epoch": 2.85, "learning_rate": 3.3556503438742103e-07, "loss": 0.022, "step": 7232 }, { "epoch": 2.85, "learning_rate": 3.3554626075709615e-07, "loss": 0.0195, "step": 7233 }, { "epoch": 2.85, "learning_rate": 3.3552748491757313e-07, "loss": 0.0225, "step": 7234 }, { "epoch": 2.85, "learning_rate": 3.355087068691581e-07, "loss": 0.0213, "step": 7235 }, { "epoch": 2.85, "learning_rate": 3.3548992661215716e-07, "loss": 0.0202, "step": 7236 }, { "epoch": 2.85, "learning_rate": 3.3547114414687625e-07, "loss": 0.0244, "step": 7237 }, { "epoch": 2.85, "learning_rate": 3.3545235947362163e-07, "loss": 0.0224, "step": 7238 }, { "epoch": 2.85, "learning_rate": 3.3543357259269947e-07, "loss": 0.0221, "step": 7239 }, { "epoch": 2.85, "learning_rate": 3.3541478350441585e-07, "loss": 0.018, "step": 7240 }, { "epoch": 2.85, "learning_rate": 3.35395992209077e-07, "loss": 0.0195, "step": 7241 }, { "epoch": 2.85, "learning_rate": 3.353771987069893e-07, "loss": 0.0272, "step": 7242 }, { "epoch": 2.85, "learning_rate": 3.3535840299845907e-07, "loss": 0.0236, "step": 7243 }, { "epoch": 2.85, "learning_rate": 3.3533960508379256e-07, "loss": 0.0193, "step": 7244 }, { "epoch": 2.86, "learning_rate": 3.353208049632962e-07, "loss": 0.0225, "step": 7245 }, { "epoch": 2.86, "learning_rate": 3.353020026372763e-07, "loss": 0.0213, "step": 7246 }, { "epoch": 2.86, "learning_rate": 3.3528319810603945e-07, "loss": 0.0236, "step": 7247 }, { "epoch": 2.86, "learning_rate": 3.3526439136989213e-07, "loss": 0.0237, "step": 7248 }, { "epoch": 2.86, "learning_rate": 3.352455824291407e-07, "loss": 0.0265, "step": 7249 }, { "epoch": 2.86, "learning_rate": 3.3522677128409194e-07, "loss": 0.0177, "step": 7250 }, { "epoch": 2.86, "learning_rate": 3.3520795793505227e-07, "loss": 0.0247, "step": 7251 }, { "epoch": 2.86, "learning_rate": 3.3518914238232837e-07, "loss": 0.0248, "step": 7252 }, { "epoch": 2.86, "learning_rate": 3.3517032462622693e-07, "loss": 0.0249, "step": 7253 }, { "epoch": 2.86, "learning_rate": 3.3515150466705463e-07, "loss": 0.0198, "step": 7254 }, { "epoch": 2.86, "learning_rate": 3.351326825051182e-07, "loss": 0.0228, "step": 7255 }, { "epoch": 2.86, "learning_rate": 3.3511385814072447e-07, "loss": 0.0222, "step": 7256 }, { "epoch": 2.86, "learning_rate": 3.3509503157418013e-07, "loss": 0.023, "step": 7257 }, { "epoch": 2.86, "learning_rate": 3.350762028057921e-07, "loss": 0.0187, "step": 7258 }, { "epoch": 2.86, "learning_rate": 3.350573718358673e-07, "loss": 0.0212, "step": 7259 }, { "epoch": 2.86, "learning_rate": 3.350385386647126e-07, "loss": 0.0244, "step": 7260 }, { "epoch": 2.86, "learning_rate": 3.3501970329263493e-07, "loss": 0.0216, "step": 7261 }, { "epoch": 2.86, "learning_rate": 3.350008657199413e-07, "loss": 0.0188, "step": 7262 }, { "epoch": 2.86, "learning_rate": 3.349820259469387e-07, "loss": 0.0194, "step": 7263 }, { "epoch": 2.86, "learning_rate": 3.3496318397393435e-07, "loss": 0.0195, "step": 7264 }, { "epoch": 2.86, "learning_rate": 3.349443398012351e-07, "loss": 0.0225, "step": 7265 }, { "epoch": 2.86, "learning_rate": 3.349254934291483e-07, "loss": 0.0206, "step": 7266 }, { "epoch": 2.86, "learning_rate": 3.34906644857981e-07, "loss": 0.0182, "step": 7267 }, { "epoch": 2.86, "learning_rate": 3.3488779408804033e-07, "loss": 0.0206, "step": 7268 }, { "epoch": 2.86, "learning_rate": 3.348689411196337e-07, "loss": 0.0208, "step": 7269 }, { "epoch": 2.87, "learning_rate": 3.3485008595306834e-07, "loss": 0.0204, "step": 7270 }, { "epoch": 2.87, "learning_rate": 3.348312285886514e-07, "loss": 0.0235, "step": 7271 }, { "epoch": 2.87, "learning_rate": 3.348123690266905e-07, "loss": 0.0234, "step": 7272 }, { "epoch": 2.87, "learning_rate": 3.347935072674928e-07, "loss": 0.0222, "step": 7273 }, { "epoch": 2.87, "learning_rate": 3.347746433113658e-07, "loss": 0.0252, "step": 7274 }, { "epoch": 2.87, "learning_rate": 3.3475577715861697e-07, "loss": 0.0234, "step": 7275 }, { "epoch": 2.87, "learning_rate": 3.3473690880955375e-07, "loss": 0.025, "step": 7276 }, { "epoch": 2.87, "learning_rate": 3.347180382644837e-07, "loss": 0.0216, "step": 7277 }, { "epoch": 2.87, "learning_rate": 3.3469916552371437e-07, "loss": 0.0232, "step": 7278 }, { "epoch": 2.87, "learning_rate": 3.3468029058755337e-07, "loss": 0.0246, "step": 7279 }, { "epoch": 2.87, "learning_rate": 3.346614134563083e-07, "loss": 0.0203, "step": 7280 }, { "epoch": 2.87, "learning_rate": 3.346425341302869e-07, "loss": 0.0206, "step": 7281 }, { "epoch": 2.87, "learning_rate": 3.346236526097968e-07, "loss": 0.0204, "step": 7282 }, { "epoch": 2.87, "learning_rate": 3.3460476889514577e-07, "loss": 0.0243, "step": 7283 }, { "epoch": 2.87, "learning_rate": 3.3458588298664163e-07, "loss": 0.0196, "step": 7284 }, { "epoch": 2.87, "learning_rate": 3.345669948845921e-07, "loss": 0.0191, "step": 7285 }, { "epoch": 2.87, "learning_rate": 3.3454810458930506e-07, "loss": 0.0221, "step": 7286 }, { "epoch": 2.87, "learning_rate": 3.3452921210108837e-07, "loss": 0.0177, "step": 7287 }, { "epoch": 2.87, "learning_rate": 3.3451031742025003e-07, "loss": 0.0211, "step": 7288 }, { "epoch": 2.87, "learning_rate": 3.3449142054709793e-07, "loss": 0.0194, "step": 7289 }, { "epoch": 2.87, "learning_rate": 3.3447252148194014e-07, "loss": 0.0223, "step": 7290 }, { "epoch": 2.87, "learning_rate": 3.3445362022508454e-07, "loss": 0.0226, "step": 7291 }, { "epoch": 2.87, "learning_rate": 3.3443471677683935e-07, "loss": 0.0233, "step": 7292 }, { "epoch": 2.87, "learning_rate": 3.344158111375125e-07, "loss": 0.0197, "step": 7293 }, { "epoch": 2.87, "learning_rate": 3.343969033074123e-07, "loss": 0.0229, "step": 7294 }, { "epoch": 2.88, "learning_rate": 3.3437799328684684e-07, "loss": 0.0246, "step": 7295 }, { "epoch": 2.88, "learning_rate": 3.3435908107612427e-07, "loss": 0.0219, "step": 7296 }, { "epoch": 2.88, "learning_rate": 3.343401666755529e-07, "loss": 0.0198, "step": 7297 }, { "epoch": 2.88, "learning_rate": 3.34321250085441e-07, "loss": 0.0216, "step": 7298 }, { "epoch": 2.88, "learning_rate": 3.3430233130609686e-07, "loss": 0.024, "step": 7299 }, { "epoch": 2.88, "learning_rate": 3.342834103378288e-07, "loss": 0.018, "step": 7300 }, { "epoch": 2.88, "learning_rate": 3.342644871809453e-07, "loss": 0.0225, "step": 7301 }, { "epoch": 2.88, "learning_rate": 3.3424556183575466e-07, "loss": 0.0197, "step": 7302 }, { "epoch": 2.88, "learning_rate": 3.342266343025654e-07, "loss": 0.0212, "step": 7303 }, { "epoch": 2.88, "learning_rate": 3.342077045816861e-07, "loss": 0.0221, "step": 7304 }, { "epoch": 2.88, "learning_rate": 3.3418877267342507e-07, "loss": 0.0225, "step": 7305 }, { "epoch": 2.88, "learning_rate": 3.34169838578091e-07, "loss": 0.02, "step": 7306 }, { "epoch": 2.88, "learning_rate": 3.3415090229599254e-07, "loss": 0.0201, "step": 7307 }, { "epoch": 2.88, "learning_rate": 3.341319638274382e-07, "loss": 0.0215, "step": 7308 }, { "epoch": 2.88, "learning_rate": 3.341130231727368e-07, "loss": 0.0189, "step": 7309 }, { "epoch": 2.88, "learning_rate": 3.340940803321968e-07, "loss": 0.0208, "step": 7310 }, { "epoch": 2.88, "learning_rate": 3.3407513530612715e-07, "loss": 0.0178, "step": 7311 }, { "epoch": 2.88, "learning_rate": 3.340561880948366e-07, "loss": 0.0185, "step": 7312 }, { "epoch": 2.88, "learning_rate": 3.340372386986339e-07, "loss": 0.0152, "step": 7313 }, { "epoch": 2.88, "learning_rate": 3.3401828711782795e-07, "loss": 0.0232, "step": 7314 }, { "epoch": 2.88, "learning_rate": 3.339993333527276e-07, "loss": 0.0194, "step": 7315 }, { "epoch": 2.88, "learning_rate": 3.3398037740364177e-07, "loss": 0.0196, "step": 7316 }, { "epoch": 2.88, "learning_rate": 3.3396141927087936e-07, "loss": 0.0207, "step": 7317 }, { "epoch": 2.88, "learning_rate": 3.339424589547495e-07, "loss": 0.0209, "step": 7318 }, { "epoch": 2.88, "learning_rate": 3.3392349645556104e-07, "loss": 0.023, "step": 7319 }, { "epoch": 2.88, "learning_rate": 3.339045317736232e-07, "loss": 0.0241, "step": 7320 }, { "epoch": 2.89, "learning_rate": 3.33885564909245e-07, "loss": 0.0216, "step": 7321 }, { "epoch": 2.89, "learning_rate": 3.3386659586273556e-07, "loss": 0.0207, "step": 7322 }, { "epoch": 2.89, "learning_rate": 3.3384762463440407e-07, "loss": 0.024, "step": 7323 }, { "epoch": 2.89, "learning_rate": 3.338286512245597e-07, "loss": 0.0216, "step": 7324 }, { "epoch": 2.89, "learning_rate": 3.338096756335117e-07, "loss": 0.0193, "step": 7325 }, { "epoch": 2.89, "learning_rate": 3.3379069786156937e-07, "loss": 0.0193, "step": 7326 }, { "epoch": 2.89, "learning_rate": 3.33771717909042e-07, "loss": 0.0229, "step": 7327 }, { "epoch": 2.89, "learning_rate": 3.33752735776239e-07, "loss": 0.0214, "step": 7328 }, { "epoch": 2.89, "learning_rate": 3.337337514634696e-07, "loss": 0.0179, "step": 7329 }, { "epoch": 2.89, "learning_rate": 3.337147649710433e-07, "loss": 0.023, "step": 7330 }, { "epoch": 2.89, "learning_rate": 3.336957762992696e-07, "loss": 0.0232, "step": 7331 }, { "epoch": 2.89, "learning_rate": 3.3367678544845793e-07, "loss": 0.0182, "step": 7332 }, { "epoch": 2.89, "learning_rate": 3.3365779241891785e-07, "loss": 0.0251, "step": 7333 }, { "epoch": 2.89, "learning_rate": 3.3363879721095884e-07, "loss": 0.0261, "step": 7334 }, { "epoch": 2.89, "learning_rate": 3.336197998248906e-07, "loss": 0.0238, "step": 7335 }, { "epoch": 2.89, "learning_rate": 3.3360080026102264e-07, "loss": 0.0208, "step": 7336 }, { "epoch": 2.89, "learning_rate": 3.3358179851966473e-07, "loss": 0.0208, "step": 7337 }, { "epoch": 2.89, "learning_rate": 3.335627946011264e-07, "loss": 0.0221, "step": 7338 }, { "epoch": 2.89, "learning_rate": 3.3354378850571767e-07, "loss": 0.0224, "step": 7339 }, { "epoch": 2.89, "learning_rate": 3.335247802337481e-07, "loss": 0.0207, "step": 7340 }, { "epoch": 2.89, "learning_rate": 3.3350576978552753e-07, "loss": 0.0242, "step": 7341 }, { "epoch": 2.89, "learning_rate": 3.3348675716136586e-07, "loss": 0.0206, "step": 7342 }, { "epoch": 2.89, "learning_rate": 3.334677423615729e-07, "loss": 0.0218, "step": 7343 }, { "epoch": 2.89, "learning_rate": 3.3344872538645856e-07, "loss": 0.0204, "step": 7344 }, { "epoch": 2.89, "learning_rate": 3.3342970623633287e-07, "loss": 0.0206, "step": 7345 }, { "epoch": 2.9, "learning_rate": 3.3341068491150577e-07, "loss": 0.0221, "step": 7346 }, { "epoch": 2.9, "learning_rate": 3.333916614122872e-07, "loss": 0.0231, "step": 7347 }, { "epoch": 2.9, "learning_rate": 3.333726357389873e-07, "loss": 0.0239, "step": 7348 }, { "epoch": 2.9, "learning_rate": 3.333536078919161e-07, "loss": 0.022, "step": 7349 }, { "epoch": 2.9, "learning_rate": 3.3333457787138384e-07, "loss": 0.0215, "step": 7350 }, { "epoch": 2.9, "learning_rate": 3.333155456777006e-07, "loss": 0.0212, "step": 7351 }, { "epoch": 2.9, "learning_rate": 3.3329651131117655e-07, "loss": 0.0199, "step": 7352 }, { "epoch": 2.9, "learning_rate": 3.3327747477212196e-07, "loss": 0.0243, "step": 7353 }, { "epoch": 2.9, "learning_rate": 3.332584360608471e-07, "loss": 0.0225, "step": 7354 }, { "epoch": 2.9, "learning_rate": 3.3323939517766226e-07, "loss": 0.0195, "step": 7355 }, { "epoch": 2.9, "learning_rate": 3.332203521228777e-07, "loss": 0.0223, "step": 7356 }, { "epoch": 2.9, "learning_rate": 3.33201306896804e-07, "loss": 0.0242, "step": 7357 }, { "epoch": 2.9, "learning_rate": 3.3318225949975136e-07, "loss": 0.0225, "step": 7358 }, { "epoch": 2.9, "learning_rate": 3.331632099320303e-07, "loss": 0.0193, "step": 7359 }, { "epoch": 2.9, "learning_rate": 3.331441581939513e-07, "loss": 0.0184, "step": 7360 }, { "epoch": 2.9, "learning_rate": 3.3312510428582496e-07, "loss": 0.0166, "step": 7361 }, { "epoch": 2.9, "learning_rate": 3.331060482079617e-07, "loss": 0.024, "step": 7362 }, { "epoch": 2.9, "learning_rate": 3.330869899606721e-07, "loss": 0.0227, "step": 7363 }, { "epoch": 2.9, "learning_rate": 3.330679295442669e-07, "loss": 0.0206, "step": 7364 }, { "epoch": 2.9, "learning_rate": 3.3304886695905665e-07, "loss": 0.0189, "step": 7365 }, { "epoch": 2.9, "learning_rate": 3.3302980220535216e-07, "loss": 0.019, "step": 7366 }, { "epoch": 2.9, "learning_rate": 3.33010735283464e-07, "loss": 0.0219, "step": 7367 }, { "epoch": 2.9, "learning_rate": 3.32991666193703e-07, "loss": 0.0205, "step": 7368 }, { "epoch": 2.9, "learning_rate": 3.3297259493638e-07, "loss": 0.0153, "step": 7369 }, { "epoch": 2.9, "learning_rate": 3.3295352151180583e-07, "loss": 0.0212, "step": 7370 }, { "epoch": 2.9, "learning_rate": 3.329344459202912e-07, "loss": 0.0165, "step": 7371 }, { "epoch": 2.91, "learning_rate": 3.329153681621473e-07, "loss": 0.02, "step": 7372 }, { "epoch": 2.91, "learning_rate": 3.3289628823768475e-07, "loss": 0.0216, "step": 7373 }, { "epoch": 2.91, "learning_rate": 3.328772061472148e-07, "loss": 0.0182, "step": 7374 }, { "epoch": 2.91, "learning_rate": 3.328581218910483e-07, "loss": 0.0227, "step": 7375 }, { "epoch": 2.91, "learning_rate": 3.3283903546949634e-07, "loss": 0.0196, "step": 7376 }, { "epoch": 2.91, "learning_rate": 3.3281994688286996e-07, "loss": 0.019, "step": 7377 }, { "epoch": 2.91, "learning_rate": 3.328008561314803e-07, "loss": 0.0174, "step": 7378 }, { "epoch": 2.91, "learning_rate": 3.3278176321563864e-07, "loss": 0.0181, "step": 7379 }, { "epoch": 2.91, "learning_rate": 3.327626681356559e-07, "loss": 0.0223, "step": 7380 }, { "epoch": 2.91, "learning_rate": 3.327435708918435e-07, "loss": 0.0206, "step": 7381 }, { "epoch": 2.91, "learning_rate": 3.3272447148451264e-07, "loss": 0.0223, "step": 7382 }, { "epoch": 2.91, "learning_rate": 3.327053699139746e-07, "loss": 0.0189, "step": 7383 }, { "epoch": 2.91, "learning_rate": 3.326862661805408e-07, "loss": 0.0246, "step": 7384 }, { "epoch": 2.91, "learning_rate": 3.3266716028452244e-07, "loss": 0.0211, "step": 7385 }, { "epoch": 2.91, "learning_rate": 3.32648052226231e-07, "loss": 0.0222, "step": 7386 }, { "epoch": 2.91, "learning_rate": 3.326289420059779e-07, "loss": 0.0214, "step": 7387 }, { "epoch": 2.91, "learning_rate": 3.3260982962407463e-07, "loss": 0.0186, "step": 7388 }, { "epoch": 2.91, "learning_rate": 3.3259071508083277e-07, "loss": 0.0218, "step": 7389 }, { "epoch": 2.91, "learning_rate": 3.3257159837656363e-07, "loss": 0.0193, "step": 7390 }, { "epoch": 2.91, "learning_rate": 3.32552479511579e-07, "loss": 0.023, "step": 7391 }, { "epoch": 2.91, "learning_rate": 3.3253335848619043e-07, "loss": 0.0217, "step": 7392 }, { "epoch": 2.91, "learning_rate": 3.3251423530070947e-07, "loss": 0.0197, "step": 7393 }, { "epoch": 2.91, "learning_rate": 3.324951099554479e-07, "loss": 0.0239, "step": 7394 }, { "epoch": 2.91, "learning_rate": 3.3247598245071746e-07, "loss": 0.0201, "step": 7395 }, { "epoch": 2.91, "learning_rate": 3.324568527868298e-07, "loss": 0.021, "step": 7396 }, { "epoch": 2.92, "learning_rate": 3.324377209640968e-07, "loss": 0.0199, "step": 7397 }, { "epoch": 2.92, "learning_rate": 3.3241858698283013e-07, "loss": 0.0226, "step": 7398 }, { "epoch": 2.92, "learning_rate": 3.323994508433418e-07, "loss": 0.0204, "step": 7399 }, { "epoch": 2.92, "learning_rate": 3.3238031254594363e-07, "loss": 0.0196, "step": 7400 }, { "epoch": 2.92, "learning_rate": 3.3236117209094763e-07, "loss": 0.0184, "step": 7401 }, { "epoch": 2.92, "learning_rate": 3.323420294786656e-07, "loss": 0.018, "step": 7402 }, { "epoch": 2.92, "learning_rate": 3.323228847094096e-07, "loss": 0.0205, "step": 7403 }, { "epoch": 2.92, "learning_rate": 3.323037377834918e-07, "loss": 0.0232, "step": 7404 }, { "epoch": 2.92, "learning_rate": 3.3228458870122414e-07, "loss": 0.018, "step": 7405 }, { "epoch": 2.92, "learning_rate": 3.322654374629187e-07, "loss": 0.02, "step": 7406 }, { "epoch": 2.92, "learning_rate": 3.3224628406888765e-07, "loss": 0.0213, "step": 7407 }, { "epoch": 2.92, "learning_rate": 3.3222712851944313e-07, "loss": 0.024, "step": 7408 }, { "epoch": 2.92, "learning_rate": 3.3220797081489745e-07, "loss": 0.0232, "step": 7409 }, { "epoch": 2.92, "learning_rate": 3.321888109555627e-07, "loss": 0.0233, "step": 7410 }, { "epoch": 2.92, "learning_rate": 3.321696489417513e-07, "loss": 0.0226, "step": 7411 }, { "epoch": 2.92, "learning_rate": 3.321504847737755e-07, "loss": 0.0199, "step": 7412 }, { "epoch": 2.92, "learning_rate": 3.321313184519475e-07, "loss": 0.0201, "step": 7413 }, { "epoch": 2.92, "learning_rate": 3.3211214997658e-07, "loss": 0.0203, "step": 7414 }, { "epoch": 2.92, "learning_rate": 3.320929793479852e-07, "loss": 0.0219, "step": 7415 }, { "epoch": 2.92, "learning_rate": 3.3207380656647554e-07, "loss": 0.02, "step": 7416 }, { "epoch": 2.92, "learning_rate": 3.320546316323636e-07, "loss": 0.0208, "step": 7417 }, { "epoch": 2.92, "learning_rate": 3.320354545459619e-07, "loss": 0.019, "step": 7418 }, { "epoch": 2.92, "learning_rate": 3.320162753075829e-07, "loss": 0.0184, "step": 7419 }, { "epoch": 2.92, "learning_rate": 3.3199709391753923e-07, "loss": 0.0182, "step": 7420 }, { "epoch": 2.92, "learning_rate": 3.3197791037614364e-07, "loss": 0.0199, "step": 7421 }, { "epoch": 2.93, "learning_rate": 3.319587246837086e-07, "loss": 0.017, "step": 7422 }, { "epoch": 2.93, "learning_rate": 3.3193953684054696e-07, "loss": 0.0204, "step": 7423 }, { "epoch": 2.93, "learning_rate": 3.319203468469714e-07, "loss": 0.0218, "step": 7424 }, { "epoch": 2.93, "learning_rate": 3.3190115470329464e-07, "loss": 0.0183, "step": 7425 }, { "epoch": 2.93, "learning_rate": 3.3188196040982953e-07, "loss": 0.021, "step": 7426 }, { "epoch": 2.93, "learning_rate": 3.3186276396688893e-07, "loss": 0.0199, "step": 7427 }, { "epoch": 2.93, "learning_rate": 3.3184356537478566e-07, "loss": 0.0216, "step": 7428 }, { "epoch": 2.93, "learning_rate": 3.3182436463383265e-07, "loss": 0.0204, "step": 7429 }, { "epoch": 2.93, "learning_rate": 3.3180516174434293e-07, "loss": 0.0224, "step": 7430 }, { "epoch": 2.93, "learning_rate": 3.3178595670662927e-07, "loss": 0.02, "step": 7431 }, { "epoch": 2.93, "learning_rate": 3.317667495210049e-07, "loss": 0.023, "step": 7432 }, { "epoch": 2.93, "learning_rate": 3.3174754018778266e-07, "loss": 0.0231, "step": 7433 }, { "epoch": 2.93, "learning_rate": 3.317283287072759e-07, "loss": 0.0229, "step": 7434 }, { "epoch": 2.93, "learning_rate": 3.317091150797975e-07, "loss": 0.0203, "step": 7435 }, { "epoch": 2.93, "learning_rate": 3.316898993056607e-07, "loss": 0.0205, "step": 7436 }, { "epoch": 2.93, "learning_rate": 3.3167068138517874e-07, "loss": 0.0231, "step": 7437 }, { "epoch": 2.93, "learning_rate": 3.3165146131866474e-07, "loss": 0.0195, "step": 7438 }, { "epoch": 2.93, "learning_rate": 3.31632239106432e-07, "loss": 0.0197, "step": 7439 }, { "epoch": 2.93, "learning_rate": 3.316130147487938e-07, "loss": 0.0216, "step": 7440 }, { "epoch": 2.93, "learning_rate": 3.315937882460636e-07, "loss": 0.0203, "step": 7441 }, { "epoch": 2.93, "learning_rate": 3.315745595985545e-07, "loss": 0.0229, "step": 7442 }, { "epoch": 2.93, "learning_rate": 3.3155532880658017e-07, "loss": 0.0245, "step": 7443 }, { "epoch": 2.93, "learning_rate": 3.3153609587045387e-07, "loss": 0.0206, "step": 7444 }, { "epoch": 2.93, "learning_rate": 3.3151686079048914e-07, "loss": 0.0191, "step": 7445 }, { "epoch": 2.93, "learning_rate": 3.3149762356699944e-07, "loss": 0.0198, "step": 7446 }, { "epoch": 2.93, "learning_rate": 3.314783842002983e-07, "loss": 0.0183, "step": 7447 }, { "epoch": 2.94, "learning_rate": 3.314591426906994e-07, "loss": 0.0211, "step": 7448 }, { "epoch": 2.94, "learning_rate": 3.3143989903851624e-07, "loss": 0.0183, "step": 7449 }, { "epoch": 2.94, "learning_rate": 3.3142065324406246e-07, "loss": 0.0258, "step": 7450 }, { "epoch": 2.94, "learning_rate": 3.3140140530765177e-07, "loss": 0.0177, "step": 7451 }, { "epoch": 2.94, "learning_rate": 3.3138215522959794e-07, "loss": 0.016, "step": 7452 }, { "epoch": 2.94, "learning_rate": 3.313629030102146e-07, "loss": 0.0193, "step": 7453 }, { "epoch": 2.94, "learning_rate": 3.3134364864981567e-07, "loss": 0.0175, "step": 7454 }, { "epoch": 2.94, "learning_rate": 3.3132439214871477e-07, "loss": 0.0263, "step": 7455 }, { "epoch": 2.94, "learning_rate": 3.3130513350722593e-07, "loss": 0.0199, "step": 7456 }, { "epoch": 2.94, "learning_rate": 3.31285872725663e-07, "loss": 0.0235, "step": 7457 }, { "epoch": 2.94, "learning_rate": 3.3126660980433985e-07, "loss": 0.0237, "step": 7458 }, { "epoch": 2.94, "learning_rate": 3.312473447435704e-07, "loss": 0.0194, "step": 7459 }, { "epoch": 2.94, "learning_rate": 3.3122807754366873e-07, "loss": 0.0229, "step": 7460 }, { "epoch": 2.94, "learning_rate": 3.312088082049489e-07, "loss": 0.0231, "step": 7461 }, { "epoch": 2.94, "learning_rate": 3.3118953672772486e-07, "loss": 0.022, "step": 7462 }, { "epoch": 2.94, "learning_rate": 3.3117026311231077e-07, "loss": 0.0198, "step": 7463 }, { "epoch": 2.94, "learning_rate": 3.311509873590208e-07, "loss": 0.0241, "step": 7464 }, { "epoch": 2.94, "learning_rate": 3.3113170946816896e-07, "loss": 0.0204, "step": 7465 }, { "epoch": 2.94, "learning_rate": 3.3111242944006964e-07, "loss": 0.0172, "step": 7466 }, { "epoch": 2.94, "learning_rate": 3.3109314727503696e-07, "loss": 0.0179, "step": 7467 }, { "epoch": 2.94, "learning_rate": 3.310738629733852e-07, "loss": 0.0179, "step": 7468 }, { "epoch": 2.94, "learning_rate": 3.3105457653542866e-07, "loss": 0.0208, "step": 7469 }, { "epoch": 2.94, "learning_rate": 3.310352879614818e-07, "loss": 0.0218, "step": 7470 }, { "epoch": 2.94, "learning_rate": 3.3101599725185883e-07, "loss": 0.0198, "step": 7471 }, { "epoch": 2.94, "learning_rate": 3.3099670440687423e-07, "loss": 0.0212, "step": 7472 }, { "epoch": 2.95, "learning_rate": 3.309774094268424e-07, "loss": 0.0195, "step": 7473 }, { "epoch": 2.95, "learning_rate": 3.3095811231207793e-07, "loss": 0.0223, "step": 7474 }, { "epoch": 2.95, "learning_rate": 3.3093881306289527e-07, "loss": 0.0158, "step": 7475 }, { "epoch": 2.95, "learning_rate": 3.309195116796089e-07, "loss": 0.027, "step": 7476 }, { "epoch": 2.95, "learning_rate": 3.3090020816253354e-07, "loss": 0.0218, "step": 7477 }, { "epoch": 2.95, "learning_rate": 3.308809025119836e-07, "loss": 0.0195, "step": 7478 }, { "epoch": 2.95, "learning_rate": 3.3086159472827395e-07, "loss": 0.0238, "step": 7479 }, { "epoch": 2.95, "learning_rate": 3.308422848117192e-07, "loss": 0.0207, "step": 7480 }, { "epoch": 2.95, "learning_rate": 3.308229727626341e-07, "loss": 0.0201, "step": 7481 }, { "epoch": 2.95, "learning_rate": 3.308036585813333e-07, "loss": 0.0199, "step": 7482 }, { "epoch": 2.95, "learning_rate": 3.3078434226813164e-07, "loss": 0.019, "step": 7483 }, { "epoch": 2.95, "learning_rate": 3.3076502382334407e-07, "loss": 0.0167, "step": 7484 }, { "epoch": 2.95, "learning_rate": 3.307457032472853e-07, "loss": 0.022, "step": 7485 }, { "epoch": 2.95, "learning_rate": 3.3072638054027027e-07, "loss": 0.0234, "step": 7486 }, { "epoch": 2.95, "learning_rate": 3.307070557026139e-07, "loss": 0.0233, "step": 7487 }, { "epoch": 2.95, "learning_rate": 3.306877287346312e-07, "loss": 0.0209, "step": 7488 }, { "epoch": 2.95, "learning_rate": 3.306683996366372e-07, "loss": 0.0181, "step": 7489 }, { "epoch": 2.95, "learning_rate": 3.306490684089468e-07, "loss": 0.0219, "step": 7490 }, { "epoch": 2.95, "learning_rate": 3.3062973505187514e-07, "loss": 0.0233, "step": 7491 }, { "epoch": 2.95, "learning_rate": 3.306103995657374e-07, "loss": 0.0248, "step": 7492 }, { "epoch": 2.95, "learning_rate": 3.305910619508486e-07, "loss": 0.0201, "step": 7493 }, { "epoch": 2.95, "learning_rate": 3.3057172220752405e-07, "loss": 0.0234, "step": 7494 }, { "epoch": 2.95, "learning_rate": 3.3055238033607875e-07, "loss": 0.0191, "step": 7495 }, { "epoch": 2.95, "learning_rate": 3.3053303633682814e-07, "loss": 0.0218, "step": 7496 }, { "epoch": 2.95, "learning_rate": 3.305136902100874e-07, "loss": 0.0249, "step": 7497 }, { "epoch": 2.96, "learning_rate": 3.304943419561719e-07, "loss": 0.0212, "step": 7498 }, { "epoch": 2.96, "learning_rate": 3.3047499157539697e-07, "loss": 0.0198, "step": 7499 }, { "epoch": 2.96, "learning_rate": 3.304556390680779e-07, "loss": 0.0206, "step": 7500 }, { "epoch": 2.96, "learning_rate": 3.3043628443453024e-07, "loss": 0.0193, "step": 7501 }, { "epoch": 2.96, "learning_rate": 3.304169276750693e-07, "loss": 0.0257, "step": 7502 }, { "epoch": 2.96, "learning_rate": 3.3039756879001075e-07, "loss": 0.0236, "step": 7503 }, { "epoch": 2.96, "learning_rate": 3.3037820777967e-07, "loss": 0.0181, "step": 7504 }, { "epoch": 2.96, "learning_rate": 3.3035884464436257e-07, "loss": 0.0283, "step": 7505 }, { "epoch": 2.96, "learning_rate": 3.3033947938440414e-07, "loss": 0.0207, "step": 7506 }, { "epoch": 2.96, "learning_rate": 3.3032011200011026e-07, "loss": 0.0209, "step": 7507 }, { "epoch": 2.96, "learning_rate": 3.303007424917966e-07, "loss": 0.0218, "step": 7508 }, { "epoch": 2.96, "learning_rate": 3.302813708597789e-07, "loss": 0.021, "step": 7509 }, { "epoch": 2.96, "learning_rate": 3.3026199710437287e-07, "loss": 0.0212, "step": 7510 }, { "epoch": 2.96, "learning_rate": 3.302426212258942e-07, "loss": 0.0228, "step": 7511 }, { "epoch": 2.96, "learning_rate": 3.302232432246589e-07, "loss": 0.0191, "step": 7512 }, { "epoch": 2.96, "learning_rate": 3.302038631009825e-07, "loss": 0.0192, "step": 7513 }, { "epoch": 2.96, "learning_rate": 3.3018448085518104e-07, "loss": 0.0242, "step": 7514 }, { "epoch": 2.96, "learning_rate": 3.301650964875704e-07, "loss": 0.0194, "step": 7515 }, { "epoch": 2.96, "learning_rate": 3.301457099984666e-07, "loss": 0.0238, "step": 7516 }, { "epoch": 2.96, "learning_rate": 3.3012632138818546e-07, "loss": 0.0196, "step": 7517 }, { "epoch": 2.96, "learning_rate": 3.3010693065704305e-07, "loss": 0.0177, "step": 7518 }, { "epoch": 2.96, "learning_rate": 3.300875378053554e-07, "loss": 0.0226, "step": 7519 }, { "epoch": 2.96, "learning_rate": 3.300681428334386e-07, "loss": 0.0232, "step": 7520 }, { "epoch": 2.96, "learning_rate": 3.300487457416088e-07, "loss": 0.0194, "step": 7521 }, { "epoch": 2.96, "learning_rate": 3.3002934653018207e-07, "loss": 0.0186, "step": 7522 }, { "epoch": 2.96, "learning_rate": 3.300099451994746e-07, "loss": 0.0211, "step": 7523 }, { "epoch": 2.97, "learning_rate": 3.2999054174980263e-07, "loss": 0.0209, "step": 7524 }, { "epoch": 2.97, "learning_rate": 3.2997113618148234e-07, "loss": 0.0221, "step": 7525 }, { "epoch": 2.97, "learning_rate": 3.2995172849483017e-07, "loss": 0.0205, "step": 7526 }, { "epoch": 2.97, "learning_rate": 3.299323186901622e-07, "loss": 0.0173, "step": 7527 }, { "epoch": 2.97, "learning_rate": 3.2991290676779505e-07, "loss": 0.0227, "step": 7528 }, { "epoch": 2.97, "learning_rate": 3.298934927280449e-07, "loss": 0.02, "step": 7529 }, { "epoch": 2.97, "learning_rate": 3.298740765712282e-07, "loss": 0.0263, "step": 7530 }, { "epoch": 2.97, "learning_rate": 3.298546582976614e-07, "loss": 0.0216, "step": 7531 }, { "epoch": 2.97, "learning_rate": 3.298352379076611e-07, "loss": 0.0206, "step": 7532 }, { "epoch": 2.97, "learning_rate": 3.2981581540154373e-07, "loss": 0.0212, "step": 7533 }, { "epoch": 2.97, "learning_rate": 3.297963907796259e-07, "loss": 0.0194, "step": 7534 }, { "epoch": 2.97, "learning_rate": 3.297769640422242e-07, "loss": 0.0246, "step": 7535 }, { "epoch": 2.97, "learning_rate": 3.297575351896551e-07, "loss": 0.0211, "step": 7536 }, { "epoch": 2.97, "learning_rate": 3.297381042222355e-07, "loss": 0.02, "step": 7537 }, { "epoch": 2.97, "learning_rate": 3.297186711402819e-07, "loss": 0.0196, "step": 7538 }, { "epoch": 2.97, "learning_rate": 3.2969923594411115e-07, "loss": 0.019, "step": 7539 }, { "epoch": 2.97, "learning_rate": 3.2967979863404e-07, "loss": 0.0202, "step": 7540 }, { "epoch": 2.97, "learning_rate": 3.296603592103851e-07, "loss": 0.0234, "step": 7541 }, { "epoch": 2.97, "learning_rate": 3.2964091767346353e-07, "loss": 0.0195, "step": 7542 }, { "epoch": 2.97, "learning_rate": 3.2962147402359205e-07, "loss": 0.021, "step": 7543 }, { "epoch": 2.97, "learning_rate": 3.2960202826108746e-07, "loss": 0.0192, "step": 7544 }, { "epoch": 2.97, "learning_rate": 3.2958258038626683e-07, "loss": 0.0236, "step": 7545 }, { "epoch": 2.97, "learning_rate": 3.2956313039944705e-07, "loss": 0.0181, "step": 7546 }, { "epoch": 2.97, "learning_rate": 3.2954367830094517e-07, "loss": 0.0174, "step": 7547 }, { "epoch": 2.97, "learning_rate": 3.2952422409107825e-07, "loss": 0.0164, "step": 7548 }, { "epoch": 2.98, "learning_rate": 3.295047677701633e-07, "loss": 0.0223, "step": 7549 }, { "epoch": 2.98, "learning_rate": 3.2948530933851746e-07, "loss": 0.0199, "step": 7550 }, { "epoch": 2.98, "learning_rate": 3.294658487964579e-07, "loss": 0.019, "step": 7551 }, { "epoch": 2.98, "learning_rate": 3.294463861443017e-07, "loss": 0.0202, "step": 7552 }, { "epoch": 2.98, "learning_rate": 3.2942692138236614e-07, "loss": 0.0251, "step": 7553 }, { "epoch": 2.98, "learning_rate": 3.2940745451096856e-07, "loss": 0.0199, "step": 7554 }, { "epoch": 2.98, "learning_rate": 3.293879855304261e-07, "loss": 0.0194, "step": 7555 }, { "epoch": 2.98, "learning_rate": 3.2936851444105614e-07, "loss": 0.0219, "step": 7556 }, { "epoch": 2.98, "learning_rate": 3.29349041243176e-07, "loss": 0.0183, "step": 7557 }, { "epoch": 2.98, "learning_rate": 3.293295659371031e-07, "loss": 0.0205, "step": 7558 }, { "epoch": 2.98, "learning_rate": 3.2931008852315484e-07, "loss": 0.0169, "step": 7559 }, { "epoch": 2.98, "learning_rate": 3.292906090016486e-07, "loss": 0.0164, "step": 7560 }, { "epoch": 2.98, "learning_rate": 3.2927112737290196e-07, "loss": 0.024, "step": 7561 }, { "epoch": 2.98, "learning_rate": 3.292516436372324e-07, "loss": 0.0188, "step": 7562 }, { "epoch": 2.98, "learning_rate": 3.2923215779495755e-07, "loss": 0.0219, "step": 7563 }, { "epoch": 2.98, "learning_rate": 3.292126698463949e-07, "loss": 0.0212, "step": 7564 }, { "epoch": 2.98, "learning_rate": 3.291931797918621e-07, "loss": 0.022, "step": 7565 }, { "epoch": 2.98, "learning_rate": 3.2917368763167684e-07, "loss": 0.0224, "step": 7566 }, { "epoch": 2.98, "learning_rate": 3.2915419336615675e-07, "loss": 0.0159, "step": 7567 }, { "epoch": 2.98, "learning_rate": 3.2913469699561965e-07, "loss": 0.0213, "step": 7568 }, { "epoch": 2.98, "learning_rate": 3.2911519852038324e-07, "loss": 0.0222, "step": 7569 }, { "epoch": 2.98, "learning_rate": 3.290956979407653e-07, "loss": 0.021, "step": 7570 }, { "epoch": 2.98, "learning_rate": 3.290761952570837e-07, "loss": 0.0253, "step": 7571 }, { "epoch": 2.98, "learning_rate": 3.290566904696563e-07, "loss": 0.0174, "step": 7572 }, { "epoch": 2.98, "learning_rate": 3.2903718357880094e-07, "loss": 0.0207, "step": 7573 }, { "epoch": 2.98, "learning_rate": 3.290176745848357e-07, "loss": 0.0205, "step": 7574 }, { "epoch": 2.99, "learning_rate": 3.2899816348807844e-07, "loss": 0.0188, "step": 7575 }, { "epoch": 2.99, "learning_rate": 3.289786502888471e-07, "loss": 0.0221, "step": 7576 }, { "epoch": 2.99, "learning_rate": 3.289591349874598e-07, "loss": 0.0182, "step": 7577 }, { "epoch": 2.99, "learning_rate": 3.289396175842346e-07, "loss": 0.0219, "step": 7578 }, { "epoch": 2.99, "learning_rate": 3.289200980794896e-07, "loss": 0.0212, "step": 7579 }, { "epoch": 2.99, "learning_rate": 3.2890057647354294e-07, "loss": 0.02, "step": 7580 }, { "epoch": 2.99, "learning_rate": 3.288810527667128e-07, "loss": 0.0214, "step": 7581 }, { "epoch": 2.99, "learning_rate": 3.2886152695931735e-07, "loss": 0.0238, "step": 7582 }, { "epoch": 2.99, "learning_rate": 3.288419990516749e-07, "loss": 0.022, "step": 7583 }, { "epoch": 2.99, "learning_rate": 3.288224690441037e-07, "loss": 0.0217, "step": 7584 }, { "epoch": 2.99, "learning_rate": 3.288029369369219e-07, "loss": 0.02, "step": 7585 }, { "epoch": 2.99, "learning_rate": 3.2878340273044814e-07, "loss": 0.0214, "step": 7586 }, { "epoch": 2.99, "learning_rate": 3.2876386642500056e-07, "loss": 0.0234, "step": 7587 }, { "epoch": 2.99, "learning_rate": 3.287443280208977e-07, "loss": 0.0194, "step": 7588 }, { "epoch": 2.99, "learning_rate": 3.2872478751845796e-07, "loss": 0.0199, "step": 7589 }, { "epoch": 2.99, "learning_rate": 3.287052449179998e-07, "loss": 0.0191, "step": 7590 }, { "epoch": 2.99, "learning_rate": 3.2868570021984177e-07, "loss": 0.0219, "step": 7591 }, { "epoch": 2.99, "learning_rate": 3.286661534243024e-07, "loss": 0.0193, "step": 7592 }, { "epoch": 2.99, "learning_rate": 3.286466045317003e-07, "loss": 0.0194, "step": 7593 }, { "epoch": 2.99, "learning_rate": 3.286270535423541e-07, "loss": 0.0207, "step": 7594 }, { "epoch": 2.99, "learning_rate": 3.286075004565823e-07, "loss": 0.0207, "step": 7595 }, { "epoch": 2.99, "learning_rate": 3.285879452747038e-07, "loss": 0.02, "step": 7596 }, { "epoch": 2.99, "learning_rate": 3.285683879970372e-07, "loss": 0.02, "step": 7597 }, { "epoch": 2.99, "learning_rate": 3.285488286239013e-07, "loss": 0.0209, "step": 7598 }, { "epoch": 2.99, "learning_rate": 3.285292671556148e-07, "loss": 0.0231, "step": 7599 }, { "epoch": 3.0, "learning_rate": 3.2850970359249664e-07, "loss": 0.026, "step": 7600 }, { "epoch": 3.0, "learning_rate": 3.2849013793486557e-07, "loss": 0.0163, "step": 7601 }, { "epoch": 3.0, "learning_rate": 3.2847057018304057e-07, "loss": 0.0239, "step": 7602 }, { "epoch": 3.0, "learning_rate": 3.2845100033734054e-07, "loss": 0.018, "step": 7603 }, { "epoch": 3.0, "learning_rate": 3.2843142839808444e-07, "loss": 0.0231, "step": 7604 }, { "epoch": 3.0, "learning_rate": 3.2841185436559124e-07, "loss": 0.0193, "step": 7605 }, { "epoch": 3.0, "learning_rate": 3.2839227824017995e-07, "loss": 0.0218, "step": 7606 }, { "epoch": 3.0, "learning_rate": 3.283727000221697e-07, "loss": 0.0232, "step": 7607 }, { "epoch": 3.0, "learning_rate": 3.283531197118795e-07, "loss": 0.0204, "step": 7608 }, { "epoch": 3.0, "learning_rate": 3.2833353730962855e-07, "loss": 0.0155, "step": 7609 }, { "epoch": 3.0, "learning_rate": 3.28313952815736e-07, "loss": 0.0207, "step": 7610 }, { "epoch": 3.0, "learning_rate": 3.28294366230521e-07, "loss": 0.0206, "step": 7611 }, { "epoch": 3.0, "learning_rate": 3.282747775543028e-07, "loss": 0.0197, "step": 7612 }, { "epoch": 3.0, "learning_rate": 3.2825518678740067e-07, "loss": 0.023, "step": 7613 }, { "epoch": 3.0, "learning_rate": 3.28235593930134e-07, "loss": 0.0179, "step": 7614 }, { "epoch": 3.0, "learning_rate": 3.28215998982822e-07, "loss": 0.0191, "step": 7615 }, { "epoch": 3.0, "learning_rate": 3.281964019457841e-07, "loss": 0.0216, "step": 7616 }, { "epoch": 3.0, "learning_rate": 3.2817680281933964e-07, "loss": 0.0195, "step": 7617 }, { "epoch": 3.0, "learning_rate": 3.2815720160380813e-07, "loss": 0.017, "step": 7618 }, { "epoch": 3.0, "learning_rate": 3.2813759829950904e-07, "loss": 0.0208, "step": 7619 }, { "epoch": 3.0, "learning_rate": 3.281179929067618e-07, "loss": 0.0157, "step": 7620 }, { "epoch": 3.0, "learning_rate": 3.28098385425886e-07, "loss": 0.0199, "step": 7621 }, { "epoch": 3.0, "learning_rate": 3.280787758572012e-07, "loss": 0.0205, "step": 7622 }, { "epoch": 3.0, "learning_rate": 3.2805916420102704e-07, "loss": 0.0203, "step": 7623 }, { "epoch": 3.0, "learning_rate": 3.280395504576831e-07, "loss": 0.0206, "step": 7624 }, { "epoch": 3.01, "learning_rate": 3.280199346274891e-07, "loss": 0.0214, "step": 7625 }, { "epoch": 3.01, "learning_rate": 3.280003167107648e-07, "loss": 0.0188, "step": 7626 }, { "epoch": 3.01, "learning_rate": 3.2798069670782974e-07, "loss": 0.0182, "step": 7627 }, { "epoch": 3.01, "learning_rate": 3.279610746190039e-07, "loss": 0.0199, "step": 7628 }, { "epoch": 3.01, "learning_rate": 3.2794145044460704e-07, "loss": 0.0167, "step": 7629 }, { "epoch": 3.01, "learning_rate": 3.2792182418495895e-07, "loss": 0.0154, "step": 7630 }, { "epoch": 3.01, "learning_rate": 3.2790219584037954e-07, "loss": 0.0195, "step": 7631 }, { "epoch": 3.01, "learning_rate": 3.278825654111888e-07, "loss": 0.0189, "step": 7632 }, { "epoch": 3.01, "learning_rate": 3.2786293289770655e-07, "loss": 0.0199, "step": 7633 }, { "epoch": 3.01, "learning_rate": 3.278432983002528e-07, "loss": 0.0171, "step": 7634 }, { "epoch": 3.01, "learning_rate": 3.2782366161914765e-07, "loss": 0.0189, "step": 7635 }, { "epoch": 3.01, "learning_rate": 3.278040228547111e-07, "loss": 0.0182, "step": 7636 }, { "epoch": 3.01, "learning_rate": 3.2778438200726315e-07, "loss": 0.0202, "step": 7637 }, { "epoch": 3.01, "learning_rate": 3.2776473907712394e-07, "loss": 0.0232, "step": 7638 }, { "epoch": 3.01, "learning_rate": 3.277450940646138e-07, "loss": 0.0173, "step": 7639 }, { "epoch": 3.01, "learning_rate": 3.2772544697005266e-07, "loss": 0.0182, "step": 7640 }, { "epoch": 3.01, "learning_rate": 3.277057977937609e-07, "loss": 0.0162, "step": 7641 }, { "epoch": 3.01, "learning_rate": 3.2768614653605866e-07, "loss": 0.0172, "step": 7642 }, { "epoch": 3.01, "learning_rate": 3.2766649319726636e-07, "loss": 0.0181, "step": 7643 }, { "epoch": 3.01, "learning_rate": 3.2764683777770427e-07, "loss": 0.02, "step": 7644 }, { "epoch": 3.01, "learning_rate": 3.276271802776927e-07, "loss": 0.0197, "step": 7645 }, { "epoch": 3.01, "learning_rate": 3.276075206975521e-07, "loss": 0.0212, "step": 7646 }, { "epoch": 3.01, "learning_rate": 3.2758785903760276e-07, "loss": 0.0153, "step": 7647 }, { "epoch": 3.01, "learning_rate": 3.275681952981653e-07, "loss": 0.0171, "step": 7648 }, { "epoch": 3.01, "learning_rate": 3.275485294795601e-07, "loss": 0.0215, "step": 7649 }, { "epoch": 3.01, "learning_rate": 3.275288615821078e-07, "loss": 0.017, "step": 7650 }, { "epoch": 3.02, "learning_rate": 3.2750919160612876e-07, "loss": 0.0196, "step": 7651 }, { "epoch": 3.02, "learning_rate": 3.2748951955194375e-07, "loss": 0.0202, "step": 7652 }, { "epoch": 3.02, "learning_rate": 3.2746984541987336e-07, "loss": 0.0189, "step": 7653 }, { "epoch": 3.02, "learning_rate": 3.274501692102382e-07, "loss": 0.019, "step": 7654 }, { "epoch": 3.02, "learning_rate": 3.2743049092335897e-07, "loss": 0.0196, "step": 7655 }, { "epoch": 3.02, "learning_rate": 3.2741081055955644e-07, "loss": 0.0214, "step": 7656 }, { "epoch": 3.02, "learning_rate": 3.2739112811915126e-07, "loss": 0.0196, "step": 7657 }, { "epoch": 3.02, "learning_rate": 3.273714436024644e-07, "loss": 0.0203, "step": 7658 }, { "epoch": 3.02, "learning_rate": 3.2735175700981653e-07, "loss": 0.0196, "step": 7659 }, { "epoch": 3.02, "learning_rate": 3.273320683415286e-07, "loss": 0.0179, "step": 7660 }, { "epoch": 3.02, "learning_rate": 3.273123775979215e-07, "loss": 0.017, "step": 7661 }, { "epoch": 3.02, "learning_rate": 3.2729268477931606e-07, "loss": 0.0183, "step": 7662 }, { "epoch": 3.02, "learning_rate": 3.272729898860334e-07, "loss": 0.0168, "step": 7663 }, { "epoch": 3.02, "learning_rate": 3.272532929183944e-07, "loss": 0.0199, "step": 7664 }, { "epoch": 3.02, "learning_rate": 3.2723359387672014e-07, "loss": 0.0193, "step": 7665 }, { "epoch": 3.02, "learning_rate": 3.2721389276133165e-07, "loss": 0.0213, "step": 7666 }, { "epoch": 3.02, "learning_rate": 3.2719418957255013e-07, "loss": 0.02, "step": 7667 }, { "epoch": 3.02, "learning_rate": 3.2717448431069655e-07, "loss": 0.0201, "step": 7668 }, { "epoch": 3.02, "learning_rate": 3.2715477697609216e-07, "loss": 0.019, "step": 7669 }, { "epoch": 3.02, "learning_rate": 3.2713506756905823e-07, "loss": 0.0209, "step": 7670 }, { "epoch": 3.02, "learning_rate": 3.271153560899158e-07, "loss": 0.0224, "step": 7671 }, { "epoch": 3.02, "learning_rate": 3.2709564253898637e-07, "loss": 0.0211, "step": 7672 }, { "epoch": 3.02, "learning_rate": 3.2707592691659106e-07, "loss": 0.0194, "step": 7673 }, { "epoch": 3.02, "learning_rate": 3.270562092230514e-07, "loss": 0.015, "step": 7674 }, { "epoch": 3.02, "learning_rate": 3.2703648945868853e-07, "loss": 0.0151, "step": 7675 }, { "epoch": 3.03, "learning_rate": 3.2701676762382393e-07, "loss": 0.0203, "step": 7676 }, { "epoch": 3.03, "learning_rate": 3.269970437187791e-07, "loss": 0.0182, "step": 7677 }, { "epoch": 3.03, "learning_rate": 3.2697731774387544e-07, "loss": 0.023, "step": 7678 }, { "epoch": 3.03, "learning_rate": 3.2695758969943454e-07, "loss": 0.0203, "step": 7679 }, { "epoch": 3.03, "learning_rate": 3.269378595857778e-07, "loss": 0.0148, "step": 7680 }, { "epoch": 3.03, "learning_rate": 3.2691812740322696e-07, "loss": 0.0218, "step": 7681 }, { "epoch": 3.03, "learning_rate": 3.268983931521035e-07, "loss": 0.0215, "step": 7682 }, { "epoch": 3.03, "learning_rate": 3.2687865683272907e-07, "loss": 0.0223, "step": 7683 }, { "epoch": 3.03, "learning_rate": 3.2685891844542545e-07, "loss": 0.0218, "step": 7684 }, { "epoch": 3.03, "learning_rate": 3.2683917799051415e-07, "loss": 0.0191, "step": 7685 }, { "epoch": 3.03, "learning_rate": 3.268194354683171e-07, "loss": 0.0205, "step": 7686 }, { "epoch": 3.03, "learning_rate": 3.267996908791559e-07, "loss": 0.0189, "step": 7687 }, { "epoch": 3.03, "learning_rate": 3.2677994422335256e-07, "loss": 0.0162, "step": 7688 }, { "epoch": 3.03, "learning_rate": 3.267601955012288e-07, "loss": 0.018, "step": 7689 }, { "epoch": 3.03, "learning_rate": 3.267404447131065e-07, "loss": 0.0214, "step": 7690 }, { "epoch": 3.03, "learning_rate": 3.267206918593075e-07, "loss": 0.0178, "step": 7691 }, { "epoch": 3.03, "learning_rate": 3.2670093694015394e-07, "loss": 0.0158, "step": 7692 }, { "epoch": 3.03, "learning_rate": 3.266811799559676e-07, "loss": 0.0194, "step": 7693 }, { "epoch": 3.03, "learning_rate": 3.2666142090707054e-07, "loss": 0.0195, "step": 7694 }, { "epoch": 3.03, "learning_rate": 3.266416597937849e-07, "loss": 0.0201, "step": 7695 }, { "epoch": 3.03, "learning_rate": 3.2662189661643256e-07, "loss": 0.0168, "step": 7696 }, { "epoch": 3.03, "learning_rate": 3.2660213137533584e-07, "loss": 0.0162, "step": 7697 }, { "epoch": 3.03, "learning_rate": 3.265823640708167e-07, "loss": 0.0169, "step": 7698 }, { "epoch": 3.03, "learning_rate": 3.2656259470319756e-07, "loss": 0.0183, "step": 7699 }, { "epoch": 3.03, "learning_rate": 3.265428232728004e-07, "loss": 0.0236, "step": 7700 }, { "epoch": 3.04, "learning_rate": 3.265230497799475e-07, "loss": 0.0216, "step": 7701 }, { "epoch": 3.04, "learning_rate": 3.265032742249612e-07, "loss": 0.0178, "step": 7702 }, { "epoch": 3.04, "learning_rate": 3.264834966081638e-07, "loss": 0.0205, "step": 7703 }, { "epoch": 3.04, "learning_rate": 3.264637169298777e-07, "loss": 0.0203, "step": 7704 }, { "epoch": 3.04, "learning_rate": 3.264439351904251e-07, "loss": 0.0189, "step": 7705 }, { "epoch": 3.04, "learning_rate": 3.2642415139012866e-07, "loss": 0.0235, "step": 7706 }, { "epoch": 3.04, "learning_rate": 3.264043655293106e-07, "loss": 0.0191, "step": 7707 }, { "epoch": 3.04, "learning_rate": 3.2638457760829356e-07, "loss": 0.017, "step": 7708 }, { "epoch": 3.04, "learning_rate": 3.263647876273999e-07, "loss": 0.0194, "step": 7709 }, { "epoch": 3.04, "learning_rate": 3.2634499558695234e-07, "loss": 0.0165, "step": 7710 }, { "epoch": 3.04, "learning_rate": 3.263252014872733e-07, "loss": 0.0176, "step": 7711 }, { "epoch": 3.04, "learning_rate": 3.263054053286855e-07, "loss": 0.0179, "step": 7712 }, { "epoch": 3.04, "learning_rate": 3.262856071115116e-07, "loss": 0.0215, "step": 7713 }, { "epoch": 3.04, "learning_rate": 3.262658068360742e-07, "loss": 0.0184, "step": 7714 }, { "epoch": 3.04, "learning_rate": 3.2624600450269606e-07, "loss": 0.017, "step": 7715 }, { "epoch": 3.04, "learning_rate": 3.2622620011169983e-07, "loss": 0.0188, "step": 7716 }, { "epoch": 3.04, "learning_rate": 3.2620639366340854e-07, "loss": 0.0189, "step": 7717 }, { "epoch": 3.04, "learning_rate": 3.2618658515814474e-07, "loss": 0.0176, "step": 7718 }, { "epoch": 3.04, "learning_rate": 3.2616677459623143e-07, "loss": 0.0184, "step": 7719 }, { "epoch": 3.04, "learning_rate": 3.2614696197799137e-07, "loss": 0.018, "step": 7720 }, { "epoch": 3.04, "learning_rate": 3.2612714730374765e-07, "loss": 0.0184, "step": 7721 }, { "epoch": 3.04, "learning_rate": 3.2610733057382305e-07, "loss": 0.0207, "step": 7722 }, { "epoch": 3.04, "learning_rate": 3.2608751178854066e-07, "loss": 0.0173, "step": 7723 }, { "epoch": 3.04, "learning_rate": 3.260676909482234e-07, "loss": 0.023, "step": 7724 }, { "epoch": 3.04, "learning_rate": 3.2604786805319444e-07, "loss": 0.0239, "step": 7725 }, { "epoch": 3.04, "learning_rate": 3.260280431037768e-07, "loss": 0.02, "step": 7726 }, { "epoch": 3.05, "learning_rate": 3.260082161002936e-07, "loss": 0.0204, "step": 7727 }, { "epoch": 3.05, "learning_rate": 3.2598838704306797e-07, "loss": 0.0199, "step": 7728 }, { "epoch": 3.05, "learning_rate": 3.25968555932423e-07, "loss": 0.0208, "step": 7729 }, { "epoch": 3.05, "learning_rate": 3.2594872276868224e-07, "loss": 0.0218, "step": 7730 }, { "epoch": 3.05, "learning_rate": 3.259288875521686e-07, "loss": 0.0234, "step": 7731 }, { "epoch": 3.05, "learning_rate": 3.259090502832055e-07, "loss": 0.0167, "step": 7732 }, { "epoch": 3.05, "learning_rate": 3.258892109621162e-07, "loss": 0.0189, "step": 7733 }, { "epoch": 3.05, "learning_rate": 3.2586936958922416e-07, "loss": 0.0209, "step": 7734 }, { "epoch": 3.05, "learning_rate": 3.2584952616485267e-07, "loss": 0.0176, "step": 7735 }, { "epoch": 3.05, "learning_rate": 3.258296806893251e-07, "loss": 0.0185, "step": 7736 }, { "epoch": 3.05, "learning_rate": 3.2580983316296507e-07, "loss": 0.0199, "step": 7737 }, { "epoch": 3.05, "learning_rate": 3.257899835860959e-07, "loss": 0.0176, "step": 7738 }, { "epoch": 3.05, "learning_rate": 3.257701319590413e-07, "loss": 0.0176, "step": 7739 }, { "epoch": 3.05, "learning_rate": 3.257502782821245e-07, "loss": 0.0163, "step": 7740 }, { "epoch": 3.05, "learning_rate": 3.2573042255566943e-07, "loss": 0.0191, "step": 7741 }, { "epoch": 3.05, "learning_rate": 3.257105647799995e-07, "loss": 0.0166, "step": 7742 }, { "epoch": 3.05, "learning_rate": 3.2569070495543846e-07, "loss": 0.0168, "step": 7743 }, { "epoch": 3.05, "learning_rate": 3.2567084308230997e-07, "loss": 0.0205, "step": 7744 }, { "epoch": 3.05, "learning_rate": 3.2565097916093765e-07, "loss": 0.0164, "step": 7745 }, { "epoch": 3.05, "learning_rate": 3.2563111319164544e-07, "loss": 0.0188, "step": 7746 }, { "epoch": 3.05, "learning_rate": 3.2561124517475696e-07, "loss": 0.0192, "step": 7747 }, { "epoch": 3.05, "learning_rate": 3.255913751105961e-07, "loss": 0.0192, "step": 7748 }, { "epoch": 3.05, "learning_rate": 3.2557150299948667e-07, "loss": 0.0222, "step": 7749 }, { "epoch": 3.05, "learning_rate": 3.255516288417526e-07, "loss": 0.0204, "step": 7750 }, { "epoch": 3.05, "learning_rate": 3.2553175263771785e-07, "loss": 0.0142, "step": 7751 }, { "epoch": 3.06, "learning_rate": 3.2551187438770633e-07, "loss": 0.0201, "step": 7752 }, { "epoch": 3.06, "learning_rate": 3.2549199409204195e-07, "loss": 0.018, "step": 7753 }, { "epoch": 3.06, "learning_rate": 3.254721117510488e-07, "loss": 0.0218, "step": 7754 }, { "epoch": 3.06, "learning_rate": 3.25452227365051e-07, "loss": 0.0176, "step": 7755 }, { "epoch": 3.06, "learning_rate": 3.2543234093437246e-07, "loss": 0.0153, "step": 7756 }, { "epoch": 3.06, "learning_rate": 3.2541245245933745e-07, "loss": 0.0162, "step": 7757 }, { "epoch": 3.06, "learning_rate": 3.2539256194027005e-07, "loss": 0.0177, "step": 7758 }, { "epoch": 3.06, "learning_rate": 3.253726693774944e-07, "loss": 0.0193, "step": 7759 }, { "epoch": 3.06, "learning_rate": 3.2535277477133487e-07, "loss": 0.0174, "step": 7760 }, { "epoch": 3.06, "learning_rate": 3.2533287812211565e-07, "loss": 0.0209, "step": 7761 }, { "epoch": 3.06, "learning_rate": 3.2531297943016097e-07, "loss": 0.0175, "step": 7762 }, { "epoch": 3.06, "learning_rate": 3.2529307869579513e-07, "loss": 0.0159, "step": 7763 }, { "epoch": 3.06, "learning_rate": 3.252731759193426e-07, "loss": 0.0239, "step": 7764 }, { "epoch": 3.06, "learning_rate": 3.2525327110112767e-07, "loss": 0.0188, "step": 7765 }, { "epoch": 3.06, "learning_rate": 3.252333642414748e-07, "loss": 0.0149, "step": 7766 }, { "epoch": 3.06, "learning_rate": 3.252134553407084e-07, "loss": 0.0166, "step": 7767 }, { "epoch": 3.06, "learning_rate": 3.25193544399153e-07, "loss": 0.0164, "step": 7768 }, { "epoch": 3.06, "learning_rate": 3.2517363141713306e-07, "loss": 0.0188, "step": 7769 }, { "epoch": 3.06, "learning_rate": 3.2515371639497316e-07, "loss": 0.0186, "step": 7770 }, { "epoch": 3.06, "learning_rate": 3.2513379933299794e-07, "loss": 0.02, "step": 7771 }, { "epoch": 3.06, "learning_rate": 3.25113880231532e-07, "loss": 0.0196, "step": 7772 }, { "epoch": 3.06, "learning_rate": 3.250939590908999e-07, "loss": 0.0185, "step": 7773 }, { "epoch": 3.06, "learning_rate": 3.250740359114264e-07, "loss": 0.0199, "step": 7774 }, { "epoch": 3.06, "learning_rate": 3.250541106934363e-07, "loss": 0.018, "step": 7775 }, { "epoch": 3.06, "learning_rate": 3.250341834372541e-07, "loss": 0.0171, "step": 7776 }, { "epoch": 3.06, "learning_rate": 3.250142541432049e-07, "loss": 0.019, "step": 7777 }, { "epoch": 3.07, "learning_rate": 3.249943228116132e-07, "loss": 0.0175, "step": 7778 }, { "epoch": 3.07, "learning_rate": 3.2497438944280407e-07, "loss": 0.0193, "step": 7779 }, { "epoch": 3.07, "learning_rate": 3.2495445403710234e-07, "loss": 0.019, "step": 7780 }, { "epoch": 3.07, "learning_rate": 3.2493451659483297e-07, "loss": 0.0211, "step": 7781 }, { "epoch": 3.07, "learning_rate": 3.2491457711632075e-07, "loss": 0.0159, "step": 7782 }, { "epoch": 3.07, "learning_rate": 3.2489463560189084e-07, "loss": 0.0185, "step": 7783 }, { "epoch": 3.07, "learning_rate": 3.248746920518682e-07, "loss": 0.0184, "step": 7784 }, { "epoch": 3.07, "learning_rate": 3.248547464665778e-07, "loss": 0.0189, "step": 7785 }, { "epoch": 3.07, "learning_rate": 3.2483479884634486e-07, "loss": 0.0185, "step": 7786 }, { "epoch": 3.07, "learning_rate": 3.2481484919149437e-07, "loss": 0.02, "step": 7787 }, { "epoch": 3.07, "learning_rate": 3.247948975023516e-07, "loss": 0.0174, "step": 7788 }, { "epoch": 3.07, "learning_rate": 3.247749437792416e-07, "loss": 0.0197, "step": 7789 }, { "epoch": 3.07, "learning_rate": 3.247549880224897e-07, "loss": 0.0203, "step": 7790 }, { "epoch": 3.07, "learning_rate": 3.247350302324211e-07, "loss": 0.0185, "step": 7791 }, { "epoch": 3.07, "learning_rate": 3.2471507040936104e-07, "loss": 0.0192, "step": 7792 }, { "epoch": 3.07, "learning_rate": 3.2469510855363495e-07, "loss": 0.0173, "step": 7793 }, { "epoch": 3.07, "learning_rate": 3.24675144665568e-07, "loss": 0.018, "step": 7794 }, { "epoch": 3.07, "learning_rate": 3.246551787454858e-07, "loss": 0.0169, "step": 7795 }, { "epoch": 3.07, "learning_rate": 3.2463521079371356e-07, "loss": 0.0213, "step": 7796 }, { "epoch": 3.07, "learning_rate": 3.246152408105768e-07, "loss": 0.0172, "step": 7797 }, { "epoch": 3.07, "learning_rate": 3.2459526879640105e-07, "loss": 0.016, "step": 7798 }, { "epoch": 3.07, "learning_rate": 3.2457529475151176e-07, "loss": 0.0214, "step": 7799 }, { "epoch": 3.07, "learning_rate": 3.245553186762345e-07, "loss": 0.019, "step": 7800 }, { "epoch": 3.07, "learning_rate": 3.2453534057089486e-07, "loss": 0.0212, "step": 7801 }, { "epoch": 3.07, "learning_rate": 3.245153604358184e-07, "loss": 0.02, "step": 7802 }, { "epoch": 3.08, "learning_rate": 3.2449537827133085e-07, "loss": 0.017, "step": 7803 }, { "epoch": 3.08, "learning_rate": 3.2447539407775786e-07, "loss": 0.0194, "step": 7804 }, { "epoch": 3.08, "learning_rate": 3.244554078554251e-07, "loss": 0.018, "step": 7805 }, { "epoch": 3.08, "learning_rate": 3.244354196046583e-07, "loss": 0.0159, "step": 7806 }, { "epoch": 3.08, "learning_rate": 3.2441542932578336e-07, "loss": 0.0154, "step": 7807 }, { "epoch": 3.08, "learning_rate": 3.2439543701912594e-07, "loss": 0.0218, "step": 7808 }, { "epoch": 3.08, "learning_rate": 3.24375442685012e-07, "loss": 0.0193, "step": 7809 }, { "epoch": 3.08, "learning_rate": 3.2435544632376736e-07, "loss": 0.0212, "step": 7810 }, { "epoch": 3.08, "learning_rate": 3.24335447935718e-07, "loss": 0.0185, "step": 7811 }, { "epoch": 3.08, "learning_rate": 3.243154475211897e-07, "loss": 0.0178, "step": 7812 }, { "epoch": 3.08, "learning_rate": 3.2429544508050863e-07, "loss": 0.0178, "step": 7813 }, { "epoch": 3.08, "learning_rate": 3.2427544061400065e-07, "loss": 0.0211, "step": 7814 }, { "epoch": 3.08, "learning_rate": 3.242554341219919e-07, "loss": 0.0217, "step": 7815 }, { "epoch": 3.08, "learning_rate": 3.242354256048084e-07, "loss": 0.0208, "step": 7816 }, { "epoch": 3.08, "learning_rate": 3.242154150627763e-07, "loss": 0.0198, "step": 7817 }, { "epoch": 3.08, "learning_rate": 3.241954024962217e-07, "loss": 0.0214, "step": 7818 }, { "epoch": 3.08, "learning_rate": 3.241753879054708e-07, "loss": 0.0169, "step": 7819 }, { "epoch": 3.08, "learning_rate": 3.241553712908498e-07, "loss": 0.0213, "step": 7820 }, { "epoch": 3.08, "learning_rate": 3.2413535265268496e-07, "loss": 0.0229, "step": 7821 }, { "epoch": 3.08, "learning_rate": 3.2411533199130253e-07, "loss": 0.0219, "step": 7822 }, { "epoch": 3.08, "learning_rate": 3.240953093070288e-07, "loss": 0.0209, "step": 7823 }, { "epoch": 3.08, "learning_rate": 3.240752846001902e-07, "loss": 0.0206, "step": 7824 }, { "epoch": 3.08, "learning_rate": 3.24055257871113e-07, "loss": 0.0192, "step": 7825 }, { "epoch": 3.08, "learning_rate": 3.240352291201236e-07, "loss": 0.0157, "step": 7826 }, { "epoch": 3.08, "learning_rate": 3.240151983475485e-07, "loss": 0.0229, "step": 7827 }, { "epoch": 3.09, "learning_rate": 3.2399516555371415e-07, "loss": 0.0204, "step": 7828 }, { "epoch": 3.09, "learning_rate": 3.23975130738947e-07, "loss": 0.0167, "step": 7829 }, { "epoch": 3.09, "learning_rate": 3.2395509390357376e-07, "loss": 0.0168, "step": 7830 }, { "epoch": 3.09, "learning_rate": 3.2393505504792077e-07, "loss": 0.019, "step": 7831 }, { "epoch": 3.09, "learning_rate": 3.239150141723148e-07, "loss": 0.0207, "step": 7832 }, { "epoch": 3.09, "learning_rate": 3.2389497127708243e-07, "loss": 0.0197, "step": 7833 }, { "epoch": 3.09, "learning_rate": 3.238749263625503e-07, "loss": 0.0182, "step": 7834 }, { "epoch": 3.09, "learning_rate": 3.2385487942904516e-07, "loss": 0.0197, "step": 7835 }, { "epoch": 3.09, "learning_rate": 3.238348304768937e-07, "loss": 0.0177, "step": 7836 }, { "epoch": 3.09, "learning_rate": 3.2381477950642275e-07, "loss": 0.0204, "step": 7837 }, { "epoch": 3.09, "learning_rate": 3.2379472651795896e-07, "loss": 0.0178, "step": 7838 }, { "epoch": 3.09, "learning_rate": 3.237746715118294e-07, "loss": 0.018, "step": 7839 }, { "epoch": 3.09, "learning_rate": 3.2375461448836074e-07, "loss": 0.0166, "step": 7840 }, { "epoch": 3.09, "learning_rate": 3.2373455544788e-07, "loss": 0.0202, "step": 7841 }, { "epoch": 3.09, "learning_rate": 3.2371449439071404e-07, "loss": 0.0211, "step": 7842 }, { "epoch": 3.09, "learning_rate": 3.2369443131718985e-07, "loss": 0.0198, "step": 7843 }, { "epoch": 3.09, "learning_rate": 3.2367436622763443e-07, "loss": 0.0183, "step": 7844 }, { "epoch": 3.09, "learning_rate": 3.2365429912237476e-07, "loss": 0.0166, "step": 7845 }, { "epoch": 3.09, "learning_rate": 3.2363423000173804e-07, "loss": 0.0171, "step": 7846 }, { "epoch": 3.09, "learning_rate": 3.2361415886605113e-07, "loss": 0.02, "step": 7847 }, { "epoch": 3.09, "learning_rate": 3.235940857156414e-07, "loss": 0.0166, "step": 7848 }, { "epoch": 3.09, "learning_rate": 3.235740105508359e-07, "loss": 0.0186, "step": 7849 }, { "epoch": 3.09, "learning_rate": 3.235539333719618e-07, "loss": 0.0159, "step": 7850 }, { "epoch": 3.09, "learning_rate": 3.235338541793464e-07, "loss": 0.0169, "step": 7851 }, { "epoch": 3.09, "learning_rate": 3.2351377297331697e-07, "loss": 0.0174, "step": 7852 }, { "epoch": 3.09, "learning_rate": 3.234936897542007e-07, "loss": 0.0183, "step": 7853 }, { "epoch": 3.1, "learning_rate": 3.2347360452232495e-07, "loss": 0.0178, "step": 7854 }, { "epoch": 3.1, "learning_rate": 3.234535172780172e-07, "loss": 0.0198, "step": 7855 }, { "epoch": 3.1, "learning_rate": 3.2343342802160466e-07, "loss": 0.0198, "step": 7856 }, { "epoch": 3.1, "learning_rate": 3.2341333675341494e-07, "loss": 0.0174, "step": 7857 }, { "epoch": 3.1, "learning_rate": 3.233932434737753e-07, "loss": 0.02, "step": 7858 }, { "epoch": 3.1, "learning_rate": 3.233731481830134e-07, "loss": 0.021, "step": 7859 }, { "epoch": 3.1, "learning_rate": 3.2335305088145657e-07, "loss": 0.0184, "step": 7860 }, { "epoch": 3.1, "learning_rate": 3.2333295156943254e-07, "loss": 0.018, "step": 7861 }, { "epoch": 3.1, "learning_rate": 3.233128502472689e-07, "loss": 0.0215, "step": 7862 }, { "epoch": 3.1, "learning_rate": 3.2329274691529325e-07, "loss": 0.0162, "step": 7863 }, { "epoch": 3.1, "learning_rate": 3.232726415738331e-07, "loss": 0.0204, "step": 7864 }, { "epoch": 3.1, "learning_rate": 3.232525342232163e-07, "loss": 0.0193, "step": 7865 }, { "epoch": 3.1, "learning_rate": 3.232324248637705e-07, "loss": 0.0161, "step": 7866 }, { "epoch": 3.1, "learning_rate": 3.232123134958234e-07, "loss": 0.0215, "step": 7867 }, { "epoch": 3.1, "learning_rate": 3.2319220011970297e-07, "loss": 0.0229, "step": 7868 }, { "epoch": 3.1, "learning_rate": 3.2317208473573683e-07, "loss": 0.0161, "step": 7869 }, { "epoch": 3.1, "learning_rate": 3.2315196734425295e-07, "loss": 0.0188, "step": 7870 }, { "epoch": 3.1, "learning_rate": 3.231318479455792e-07, "loss": 0.0201, "step": 7871 }, { "epoch": 3.1, "learning_rate": 3.2311172654004343e-07, "loss": 0.0201, "step": 7872 }, { "epoch": 3.1, "learning_rate": 3.230916031279736e-07, "loss": 0.0201, "step": 7873 }, { "epoch": 3.1, "learning_rate": 3.230714777096978e-07, "loss": 0.0193, "step": 7874 }, { "epoch": 3.1, "learning_rate": 3.230513502855438e-07, "loss": 0.0163, "step": 7875 }, { "epoch": 3.1, "learning_rate": 3.2303122085583995e-07, "loss": 0.0163, "step": 7876 }, { "epoch": 3.1, "learning_rate": 3.2301108942091414e-07, "loss": 0.0201, "step": 7877 }, { "epoch": 3.1, "learning_rate": 3.229909559810945e-07, "loss": 0.017, "step": 7878 }, { "epoch": 3.11, "learning_rate": 3.2297082053670924e-07, "loss": 0.0232, "step": 7879 }, { "epoch": 3.11, "learning_rate": 3.229506830880865e-07, "loss": 0.0203, "step": 7880 }, { "epoch": 3.11, "learning_rate": 3.2293054363555445e-07, "loss": 0.0157, "step": 7881 }, { "epoch": 3.11, "learning_rate": 3.2291040217944144e-07, "loss": 0.0172, "step": 7882 }, { "epoch": 3.11, "learning_rate": 3.228902587200756e-07, "loss": 0.0229, "step": 7883 }, { "epoch": 3.11, "learning_rate": 3.228701132577854e-07, "loss": 0.0171, "step": 7884 }, { "epoch": 3.11, "learning_rate": 3.2284996579289906e-07, "loss": 0.021, "step": 7885 }, { "epoch": 3.11, "learning_rate": 3.2282981632574495e-07, "loss": 0.0207, "step": 7886 }, { "epoch": 3.11, "learning_rate": 3.2280966485665155e-07, "loss": 0.0197, "step": 7887 }, { "epoch": 3.11, "learning_rate": 3.2278951138594725e-07, "loss": 0.0174, "step": 7888 }, { "epoch": 3.11, "learning_rate": 3.2276935591396053e-07, "loss": 0.0168, "step": 7889 }, { "epoch": 3.11, "learning_rate": 3.227491984410199e-07, "loss": 0.0192, "step": 7890 }, { "epoch": 3.11, "learning_rate": 3.2272903896745386e-07, "loss": 0.0212, "step": 7891 }, { "epoch": 3.11, "learning_rate": 3.227088774935911e-07, "loss": 0.0177, "step": 7892 }, { "epoch": 3.11, "learning_rate": 3.2268871401976004e-07, "loss": 0.0199, "step": 7893 }, { "epoch": 3.11, "learning_rate": 3.2266854854628944e-07, "loss": 0.0185, "step": 7894 }, { "epoch": 3.11, "learning_rate": 3.22648381073508e-07, "loss": 0.0239, "step": 7895 }, { "epoch": 3.11, "learning_rate": 3.2262821160174427e-07, "loss": 0.0166, "step": 7896 }, { "epoch": 3.11, "learning_rate": 3.226080401313271e-07, "loss": 0.0142, "step": 7897 }, { "epoch": 3.11, "learning_rate": 3.225878666625852e-07, "loss": 0.0173, "step": 7898 }, { "epoch": 3.11, "learning_rate": 3.225676911958474e-07, "loss": 0.0193, "step": 7899 }, { "epoch": 3.11, "learning_rate": 3.2254751373144245e-07, "loss": 0.0209, "step": 7900 }, { "epoch": 3.11, "learning_rate": 3.225273342696994e-07, "loss": 0.018, "step": 7901 }, { "epoch": 3.11, "learning_rate": 3.225071528109469e-07, "loss": 0.0178, "step": 7902 }, { "epoch": 3.11, "learning_rate": 3.22486969355514e-07, "loss": 0.0172, "step": 7903 }, { "epoch": 3.12, "learning_rate": 3.2246678390372976e-07, "loss": 0.0191, "step": 7904 }, { "epoch": 3.12, "learning_rate": 3.22446596455923e-07, "loss": 0.0184, "step": 7905 }, { "epoch": 3.12, "learning_rate": 3.2242640701242285e-07, "loss": 0.0201, "step": 7906 }, { "epoch": 3.12, "learning_rate": 3.224062155735583e-07, "loss": 0.013, "step": 7907 }, { "epoch": 3.12, "learning_rate": 3.2238602213965857e-07, "loss": 0.0198, "step": 7908 }, { "epoch": 3.12, "learning_rate": 3.2236582671105255e-07, "loss": 0.0238, "step": 7909 }, { "epoch": 3.12, "learning_rate": 3.223456292880696e-07, "loss": 0.0145, "step": 7910 }, { "epoch": 3.12, "learning_rate": 3.2232542987103875e-07, "loss": 0.0204, "step": 7911 }, { "epoch": 3.12, "learning_rate": 3.223052284602894e-07, "loss": 0.019, "step": 7912 }, { "epoch": 3.12, "learning_rate": 3.222850250561508e-07, "loss": 0.0183, "step": 7913 }, { "epoch": 3.12, "learning_rate": 3.222648196589519e-07, "loss": 0.0171, "step": 7914 }, { "epoch": 3.12, "learning_rate": 3.222446122690225e-07, "loss": 0.0204, "step": 7915 }, { "epoch": 3.12, "learning_rate": 3.222244028866915e-07, "loss": 0.0184, "step": 7916 }, { "epoch": 3.12, "learning_rate": 3.222041915122886e-07, "loss": 0.0202, "step": 7917 }, { "epoch": 3.12, "learning_rate": 3.2218397814614317e-07, "loss": 0.0214, "step": 7918 }, { "epoch": 3.12, "learning_rate": 3.221637627885845e-07, "loss": 0.018, "step": 7919 }, { "epoch": 3.12, "learning_rate": 3.2214354543994213e-07, "loss": 0.0153, "step": 7920 }, { "epoch": 3.12, "learning_rate": 3.221233261005457e-07, "loss": 0.0184, "step": 7921 }, { "epoch": 3.12, "learning_rate": 3.221031047707246e-07, "loss": 0.0204, "step": 7922 }, { "epoch": 3.12, "learning_rate": 3.2208288145080846e-07, "loss": 0.0158, "step": 7923 }, { "epoch": 3.12, "learning_rate": 3.22062656141127e-07, "loss": 0.0198, "step": 7924 }, { "epoch": 3.12, "learning_rate": 3.220424288420097e-07, "loss": 0.017, "step": 7925 }, { "epoch": 3.12, "learning_rate": 3.220221995537863e-07, "loss": 0.0156, "step": 7926 }, { "epoch": 3.12, "learning_rate": 3.2200196827678643e-07, "loss": 0.0174, "step": 7927 }, { "epoch": 3.12, "learning_rate": 3.2198173501134e-07, "loss": 0.0212, "step": 7928 }, { "epoch": 3.12, "learning_rate": 3.219614997577766e-07, "loss": 0.0191, "step": 7929 }, { "epoch": 3.13, "learning_rate": 3.2194126251642615e-07, "loss": 0.0175, "step": 7930 }, { "epoch": 3.13, "learning_rate": 3.219210232876184e-07, "loss": 0.0193, "step": 7931 }, { "epoch": 3.13, "learning_rate": 3.2190078207168336e-07, "loss": 0.0148, "step": 7932 }, { "epoch": 3.13, "learning_rate": 3.218805388689508e-07, "loss": 0.0224, "step": 7933 }, { "epoch": 3.13, "learning_rate": 3.2186029367975067e-07, "loss": 0.0196, "step": 7934 }, { "epoch": 3.13, "learning_rate": 3.21840046504413e-07, "loss": 0.0235, "step": 7935 }, { "epoch": 3.13, "learning_rate": 3.2181979734326763e-07, "loss": 0.02, "step": 7936 }, { "epoch": 3.13, "learning_rate": 3.2179954619664487e-07, "loss": 0.0203, "step": 7937 }, { "epoch": 3.13, "learning_rate": 3.2177929306487456e-07, "loss": 0.0185, "step": 7938 }, { "epoch": 3.13, "learning_rate": 3.2175903794828683e-07, "loss": 0.0154, "step": 7939 }, { "epoch": 3.13, "learning_rate": 3.217387808472118e-07, "loss": 0.0181, "step": 7940 }, { "epoch": 3.13, "learning_rate": 3.2171852176197976e-07, "loss": 0.0168, "step": 7941 }, { "epoch": 3.13, "learning_rate": 3.2169826069292077e-07, "loss": 0.0193, "step": 7942 }, { "epoch": 3.13, "learning_rate": 3.216779976403651e-07, "loss": 0.0213, "step": 7943 }, { "epoch": 3.13, "learning_rate": 3.21657732604643e-07, "loss": 0.0191, "step": 7944 }, { "epoch": 3.13, "learning_rate": 3.216374655860847e-07, "loss": 0.0214, "step": 7945 }, { "epoch": 3.13, "learning_rate": 3.216171965850206e-07, "loss": 0.0211, "step": 7946 }, { "epoch": 3.13, "learning_rate": 3.2159692560178105e-07, "loss": 0.0182, "step": 7947 }, { "epoch": 3.13, "learning_rate": 3.2157665263669643e-07, "loss": 0.0186, "step": 7948 }, { "epoch": 3.13, "learning_rate": 3.2155637769009713e-07, "loss": 0.0224, "step": 7949 }, { "epoch": 3.13, "learning_rate": 3.2153610076231366e-07, "loss": 0.0176, "step": 7950 }, { "epoch": 3.13, "learning_rate": 3.215158218536764e-07, "loss": 0.0193, "step": 7951 }, { "epoch": 3.13, "learning_rate": 3.214955409645159e-07, "loss": 0.024, "step": 7952 }, { "epoch": 3.13, "learning_rate": 3.214752580951629e-07, "loss": 0.0227, "step": 7953 }, { "epoch": 3.13, "learning_rate": 3.2145497324594766e-07, "loss": 0.0174, "step": 7954 }, { "epoch": 3.14, "learning_rate": 3.21434686417201e-07, "loss": 0.0187, "step": 7955 }, { "epoch": 3.14, "learning_rate": 3.2141439760925347e-07, "loss": 0.0174, "step": 7956 }, { "epoch": 3.14, "learning_rate": 3.2139410682243576e-07, "loss": 0.0194, "step": 7957 }, { "epoch": 3.14, "learning_rate": 3.213738140570787e-07, "loss": 0.0202, "step": 7958 }, { "epoch": 3.14, "learning_rate": 3.213535193135129e-07, "loss": 0.0183, "step": 7959 }, { "epoch": 3.14, "learning_rate": 3.2133322259206917e-07, "loss": 0.0168, "step": 7960 }, { "epoch": 3.14, "learning_rate": 3.2131292389307827e-07, "loss": 0.0196, "step": 7961 }, { "epoch": 3.14, "learning_rate": 3.2129262321687117e-07, "loss": 0.0185, "step": 7962 }, { "epoch": 3.14, "learning_rate": 3.2127232056377863e-07, "loss": 0.0199, "step": 7963 }, { "epoch": 3.14, "learning_rate": 3.2125201593413154e-07, "loss": 0.019, "step": 7964 }, { "epoch": 3.14, "learning_rate": 3.212317093282609e-07, "loss": 0.0189, "step": 7965 }, { "epoch": 3.14, "learning_rate": 3.2121140074649765e-07, "loss": 0.0158, "step": 7966 }, { "epoch": 3.14, "learning_rate": 3.211910901891728e-07, "loss": 0.0161, "step": 7967 }, { "epoch": 3.14, "learning_rate": 3.2117077765661737e-07, "loss": 0.0188, "step": 7968 }, { "epoch": 3.14, "learning_rate": 3.2115046314916237e-07, "loss": 0.0158, "step": 7969 }, { "epoch": 3.14, "learning_rate": 3.21130146667139e-07, "loss": 0.0185, "step": 7970 }, { "epoch": 3.14, "learning_rate": 3.2110982821087833e-07, "loss": 0.0185, "step": 7971 }, { "epoch": 3.14, "learning_rate": 3.210895077807115e-07, "loss": 0.0171, "step": 7972 }, { "epoch": 3.14, "learning_rate": 3.2106918537696975e-07, "loss": 0.0204, "step": 7973 }, { "epoch": 3.14, "learning_rate": 3.2104886099998425e-07, "loss": 0.0179, "step": 7974 }, { "epoch": 3.14, "learning_rate": 3.2102853465008634e-07, "loss": 0.0151, "step": 7975 }, { "epoch": 3.14, "learning_rate": 3.210082063276072e-07, "loss": 0.0185, "step": 7976 }, { "epoch": 3.14, "learning_rate": 3.2098787603287826e-07, "loss": 0.0204, "step": 7977 }, { "epoch": 3.14, "learning_rate": 3.209675437662308e-07, "loss": 0.018, "step": 7978 }, { "epoch": 3.14, "learning_rate": 3.2094720952799625e-07, "loss": 0.0196, "step": 7979 }, { "epoch": 3.14, "learning_rate": 3.20926873318506e-07, "loss": 0.0178, "step": 7980 }, { "epoch": 3.15, "learning_rate": 3.209065351380914e-07, "loss": 0.018, "step": 7981 }, { "epoch": 3.15, "learning_rate": 3.2088619498708415e-07, "loss": 0.021, "step": 7982 }, { "epoch": 3.15, "learning_rate": 3.208658528658156e-07, "loss": 0.0204, "step": 7983 }, { "epoch": 3.15, "learning_rate": 3.208455087746174e-07, "loss": 0.0187, "step": 7984 }, { "epoch": 3.15, "learning_rate": 3.2082516271382105e-07, "loss": 0.0207, "step": 7985 }, { "epoch": 3.15, "learning_rate": 3.2080481468375814e-07, "loss": 0.0196, "step": 7986 }, { "epoch": 3.15, "learning_rate": 3.2078446468476037e-07, "loss": 0.0178, "step": 7987 }, { "epoch": 3.15, "learning_rate": 3.2076411271715935e-07, "loss": 0.0256, "step": 7988 }, { "epoch": 3.15, "learning_rate": 3.207437587812869e-07, "loss": 0.0175, "step": 7989 }, { "epoch": 3.15, "learning_rate": 3.2072340287747466e-07, "loss": 0.0198, "step": 7990 }, { "epoch": 3.15, "learning_rate": 3.2070304500605447e-07, "loss": 0.0164, "step": 7991 }, { "epoch": 3.15, "learning_rate": 3.206826851673581e-07, "loss": 0.0198, "step": 7992 }, { "epoch": 3.15, "learning_rate": 3.206623233617173e-07, "loss": 0.0206, "step": 7993 }, { "epoch": 3.15, "learning_rate": 3.2064195958946405e-07, "loss": 0.0166, "step": 7994 }, { "epoch": 3.15, "learning_rate": 3.2062159385093023e-07, "loss": 0.0226, "step": 7995 }, { "epoch": 3.15, "learning_rate": 3.2060122614644776e-07, "loss": 0.0166, "step": 7996 }, { "epoch": 3.15, "learning_rate": 3.205808564763486e-07, "loss": 0.0185, "step": 7997 }, { "epoch": 3.15, "learning_rate": 3.205604848409647e-07, "loss": 0.0245, "step": 7998 }, { "epoch": 3.15, "learning_rate": 3.2054011124062815e-07, "loss": 0.0162, "step": 7999 }, { "epoch": 3.15, "learning_rate": 3.20519735675671e-07, "loss": 0.0176, "step": 8000 }, { "epoch": 3.15, "learning_rate": 3.2049935814642536e-07, "loss": 0.0194, "step": 8001 }, { "epoch": 3.15, "learning_rate": 3.2047897865322325e-07, "loss": 0.0174, "step": 8002 }, { "epoch": 3.15, "learning_rate": 3.204585971963969e-07, "loss": 0.0149, "step": 8003 }, { "epoch": 3.15, "learning_rate": 3.204382137762786e-07, "loss": 0.0213, "step": 8004 }, { "epoch": 3.15, "learning_rate": 3.204178283932003e-07, "loss": 0.0168, "step": 8005 }, { "epoch": 3.16, "learning_rate": 3.2039744104749453e-07, "loss": 0.0173, "step": 8006 }, { "epoch": 3.16, "learning_rate": 3.203770517394934e-07, "loss": 0.0153, "step": 8007 }, { "epoch": 3.16, "learning_rate": 3.2035666046952926e-07, "loss": 0.0168, "step": 8008 }, { "epoch": 3.16, "learning_rate": 3.203362672379345e-07, "loss": 0.0217, "step": 8009 }, { "epoch": 3.16, "learning_rate": 3.203158720450415e-07, "loss": 0.0176, "step": 8010 }, { "epoch": 3.16, "learning_rate": 3.202954748911826e-07, "loss": 0.0194, "step": 8011 }, { "epoch": 3.16, "learning_rate": 3.2027507577669034e-07, "loss": 0.0174, "step": 8012 }, { "epoch": 3.16, "learning_rate": 3.2025467470189707e-07, "loss": 0.0185, "step": 8013 }, { "epoch": 3.16, "learning_rate": 3.2023427166713547e-07, "loss": 0.0183, "step": 8014 }, { "epoch": 3.16, "learning_rate": 3.2021386667273784e-07, "loss": 0.0196, "step": 8015 }, { "epoch": 3.16, "learning_rate": 3.20193459719037e-07, "loss": 0.0225, "step": 8016 }, { "epoch": 3.16, "learning_rate": 3.2017305080636545e-07, "loss": 0.0175, "step": 8017 }, { "epoch": 3.16, "learning_rate": 3.201526399350557e-07, "loss": 0.0171, "step": 8018 }, { "epoch": 3.16, "learning_rate": 3.2013222710544065e-07, "loss": 0.0205, "step": 8019 }, { "epoch": 3.16, "learning_rate": 3.201118123178528e-07, "loss": 0.0186, "step": 8020 }, { "epoch": 3.16, "learning_rate": 3.20091395572625e-07, "loss": 0.0209, "step": 8021 }, { "epoch": 3.16, "learning_rate": 3.2007097687008995e-07, "loss": 0.0147, "step": 8022 }, { "epoch": 3.16, "learning_rate": 3.200505562105805e-07, "loss": 0.0188, "step": 8023 }, { "epoch": 3.16, "learning_rate": 3.200301335944294e-07, "loss": 0.0181, "step": 8024 }, { "epoch": 3.16, "learning_rate": 3.2000970902196957e-07, "loss": 0.0192, "step": 8025 }, { "epoch": 3.16, "learning_rate": 3.199892824935339e-07, "loss": 0.0153, "step": 8026 }, { "epoch": 3.16, "learning_rate": 3.199688540094552e-07, "loss": 0.0159, "step": 8027 }, { "epoch": 3.16, "learning_rate": 3.1994842357006657e-07, "loss": 0.016, "step": 8028 }, { "epoch": 3.16, "learning_rate": 3.199279911757009e-07, "loss": 0.0166, "step": 8029 }, { "epoch": 3.16, "learning_rate": 3.1990755682669133e-07, "loss": 0.0191, "step": 8030 }, { "epoch": 3.17, "learning_rate": 3.1988712052337067e-07, "loss": 0.0197, "step": 8031 }, { "epoch": 3.17, "learning_rate": 3.1986668226607227e-07, "loss": 0.0219, "step": 8032 }, { "epoch": 3.17, "learning_rate": 3.1984624205512916e-07, "loss": 0.0169, "step": 8033 }, { "epoch": 3.17, "learning_rate": 3.1982579989087435e-07, "loss": 0.019, "step": 8034 }, { "epoch": 3.17, "learning_rate": 3.1980535577364113e-07, "loss": 0.0195, "step": 8035 }, { "epoch": 3.17, "learning_rate": 3.197849097037627e-07, "loss": 0.0181, "step": 8036 }, { "epoch": 3.17, "learning_rate": 3.197644616815724e-07, "loss": 0.0195, "step": 8037 }, { "epoch": 3.17, "learning_rate": 3.197440117074033e-07, "loss": 0.0203, "step": 8038 }, { "epoch": 3.17, "learning_rate": 3.197235597815888e-07, "loss": 0.0194, "step": 8039 }, { "epoch": 3.17, "learning_rate": 3.1970310590446223e-07, "loss": 0.0188, "step": 8040 }, { "epoch": 3.17, "learning_rate": 3.19682650076357e-07, "loss": 0.0241, "step": 8041 }, { "epoch": 3.17, "learning_rate": 3.196621922976064e-07, "loss": 0.0163, "step": 8042 }, { "epoch": 3.17, "learning_rate": 3.1964173256854403e-07, "loss": 0.0161, "step": 8043 }, { "epoch": 3.17, "learning_rate": 3.196212708895032e-07, "loss": 0.0164, "step": 8044 }, { "epoch": 3.17, "learning_rate": 3.196008072608174e-07, "loss": 0.0202, "step": 8045 }, { "epoch": 3.17, "learning_rate": 3.195803416828203e-07, "loss": 0.0188, "step": 8046 }, { "epoch": 3.17, "learning_rate": 3.1955987415584533e-07, "loss": 0.0156, "step": 8047 }, { "epoch": 3.17, "learning_rate": 3.195394046802262e-07, "loss": 0.0194, "step": 8048 }, { "epoch": 3.17, "learning_rate": 3.195189332562964e-07, "loss": 0.0171, "step": 8049 }, { "epoch": 3.17, "learning_rate": 3.194984598843896e-07, "loss": 0.0221, "step": 8050 }, { "epoch": 3.17, "learning_rate": 3.194779845648395e-07, "loss": 0.0184, "step": 8051 }, { "epoch": 3.17, "learning_rate": 3.1945750729797993e-07, "loss": 0.0216, "step": 8052 }, { "epoch": 3.17, "learning_rate": 3.1943702808414454e-07, "loss": 0.0167, "step": 8053 }, { "epoch": 3.17, "learning_rate": 3.19416546923667e-07, "loss": 0.0173, "step": 8054 }, { "epoch": 3.17, "learning_rate": 3.1939606381688135e-07, "loss": 0.0179, "step": 8055 }, { "epoch": 3.17, "learning_rate": 3.193755787641213e-07, "loss": 0.0198, "step": 8056 }, { "epoch": 3.18, "learning_rate": 3.193550917657207e-07, "loss": 0.0195, "step": 8057 }, { "epoch": 3.18, "learning_rate": 3.1933460282201356e-07, "loss": 0.0192, "step": 8058 }, { "epoch": 3.18, "learning_rate": 3.193141119333337e-07, "loss": 0.0159, "step": 8059 }, { "epoch": 3.18, "learning_rate": 3.1929361910001516e-07, "loss": 0.0198, "step": 8060 }, { "epoch": 3.18, "learning_rate": 3.19273124322392e-07, "loss": 0.0171, "step": 8061 }, { "epoch": 3.18, "learning_rate": 3.192526276007981e-07, "loss": 0.0203, "step": 8062 }, { "epoch": 3.18, "learning_rate": 3.1923212893556766e-07, "loss": 0.0172, "step": 8063 }, { "epoch": 3.18, "learning_rate": 3.192116283270348e-07, "loss": 0.0176, "step": 8064 }, { "epoch": 3.18, "learning_rate": 3.1919112577553346e-07, "loss": 0.0225, "step": 8065 }, { "epoch": 3.18, "learning_rate": 3.1917062128139795e-07, "loss": 0.0184, "step": 8066 }, { "epoch": 3.18, "learning_rate": 3.1915011484496247e-07, "loss": 0.0214, "step": 8067 }, { "epoch": 3.18, "learning_rate": 3.191296064665612e-07, "loss": 0.0179, "step": 8068 }, { "epoch": 3.18, "learning_rate": 3.1910909614652834e-07, "loss": 0.0195, "step": 8069 }, { "epoch": 3.18, "learning_rate": 3.190885838851983e-07, "loss": 0.0154, "step": 8070 }, { "epoch": 3.18, "learning_rate": 3.190680696829053e-07, "loss": 0.0188, "step": 8071 }, { "epoch": 3.18, "learning_rate": 3.190475535399838e-07, "loss": 0.0186, "step": 8072 }, { "epoch": 3.18, "learning_rate": 3.190270354567681e-07, "loss": 0.0177, "step": 8073 }, { "epoch": 3.18, "learning_rate": 3.190065154335925e-07, "loss": 0.0172, "step": 8074 }, { "epoch": 3.18, "learning_rate": 3.1898599347079165e-07, "loss": 0.0167, "step": 8075 }, { "epoch": 3.18, "learning_rate": 3.1896546956869996e-07, "loss": 0.0194, "step": 8076 }, { "epoch": 3.18, "learning_rate": 3.189449437276519e-07, "loss": 0.0227, "step": 8077 }, { "epoch": 3.18, "learning_rate": 3.1892441594798204e-07, "loss": 0.0189, "step": 8078 }, { "epoch": 3.18, "learning_rate": 3.189038862300249e-07, "loss": 0.0211, "step": 8079 }, { "epoch": 3.18, "learning_rate": 3.1888335457411517e-07, "loss": 0.0241, "step": 8080 }, { "epoch": 3.18, "learning_rate": 3.188628209805875e-07, "loss": 0.0171, "step": 8081 }, { "epoch": 3.19, "learning_rate": 3.188422854497764e-07, "loss": 0.0187, "step": 8082 }, { "epoch": 3.19, "learning_rate": 3.1882174798201673e-07, "loss": 0.0242, "step": 8083 }, { "epoch": 3.19, "learning_rate": 3.188012085776432e-07, "loss": 0.0185, "step": 8084 }, { "epoch": 3.19, "learning_rate": 3.187806672369904e-07, "loss": 0.0172, "step": 8085 }, { "epoch": 3.19, "learning_rate": 3.1876012396039327e-07, "loss": 0.0176, "step": 8086 }, { "epoch": 3.19, "learning_rate": 3.1873957874818665e-07, "loss": 0.0204, "step": 8087 }, { "epoch": 3.19, "learning_rate": 3.187190316007054e-07, "loss": 0.0174, "step": 8088 }, { "epoch": 3.19, "learning_rate": 3.186984825182843e-07, "loss": 0.0169, "step": 8089 }, { "epoch": 3.19, "learning_rate": 3.186779315012584e-07, "loss": 0.024, "step": 8090 }, { "epoch": 3.19, "learning_rate": 3.1865737854996255e-07, "loss": 0.0212, "step": 8091 }, { "epoch": 3.19, "learning_rate": 3.1863682366473173e-07, "loss": 0.0194, "step": 8092 }, { "epoch": 3.19, "learning_rate": 3.186162668459011e-07, "loss": 0.0184, "step": 8093 }, { "epoch": 3.19, "learning_rate": 3.1859570809380546e-07, "loss": 0.0187, "step": 8094 }, { "epoch": 3.19, "learning_rate": 3.185751474087802e-07, "loss": 0.0176, "step": 8095 }, { "epoch": 3.19, "learning_rate": 3.1855458479116007e-07, "loss": 0.0199, "step": 8096 }, { "epoch": 3.19, "learning_rate": 3.1853402024128046e-07, "loss": 0.0189, "step": 8097 }, { "epoch": 3.19, "learning_rate": 3.1851345375947647e-07, "loss": 0.0171, "step": 8098 }, { "epoch": 3.19, "learning_rate": 3.184928853460833e-07, "loss": 0.019, "step": 8099 }, { "epoch": 3.19, "learning_rate": 3.1847231500143614e-07, "loss": 0.014, "step": 8100 }, { "epoch": 3.19, "learning_rate": 3.184517427258704e-07, "loss": 0.0192, "step": 8101 }, { "epoch": 3.19, "learning_rate": 3.1843116851972123e-07, "loss": 0.0221, "step": 8102 }, { "epoch": 3.19, "learning_rate": 3.18410592383324e-07, "loss": 0.0198, "step": 8103 }, { "epoch": 3.19, "learning_rate": 3.1839001431701413e-07, "loss": 0.0205, "step": 8104 }, { "epoch": 3.19, "learning_rate": 3.183694343211269e-07, "loss": 0.0195, "step": 8105 }, { "epoch": 3.19, "learning_rate": 3.183488523959978e-07, "loss": 0.0193, "step": 8106 }, { "epoch": 3.2, "learning_rate": 3.1832826854196226e-07, "loss": 0.0164, "step": 8107 }, { "epoch": 3.2, "learning_rate": 3.183076827593559e-07, "loss": 0.0225, "step": 8108 }, { "epoch": 3.2, "learning_rate": 3.1828709504851394e-07, "loss": 0.0177, "step": 8109 }, { "epoch": 3.2, "learning_rate": 3.1826650540977223e-07, "loss": 0.0148, "step": 8110 }, { "epoch": 3.2, "learning_rate": 3.182459138434662e-07, "loss": 0.0176, "step": 8111 }, { "epoch": 3.2, "learning_rate": 3.182253203499314e-07, "loss": 0.0196, "step": 8112 }, { "epoch": 3.2, "learning_rate": 3.182047249295037e-07, "loss": 0.0182, "step": 8113 }, { "epoch": 3.2, "learning_rate": 3.181841275825186e-07, "loss": 0.0195, "step": 8114 }, { "epoch": 3.2, "learning_rate": 3.181635283093118e-07, "loss": 0.0177, "step": 8115 }, { "epoch": 3.2, "learning_rate": 3.1814292711021904e-07, "loss": 0.0213, "step": 8116 }, { "epoch": 3.2, "learning_rate": 3.181223239855762e-07, "loss": 0.0247, "step": 8117 }, { "epoch": 3.2, "learning_rate": 3.181017189357189e-07, "loss": 0.0179, "step": 8118 }, { "epoch": 3.2, "learning_rate": 3.180811119609832e-07, "loss": 0.0223, "step": 8119 }, { "epoch": 3.2, "learning_rate": 3.1806050306170476e-07, "loss": 0.0194, "step": 8120 }, { "epoch": 3.2, "learning_rate": 3.180398922382196e-07, "loss": 0.0215, "step": 8121 }, { "epoch": 3.2, "learning_rate": 3.1801927949086355e-07, "loss": 0.0159, "step": 8122 }, { "epoch": 3.2, "learning_rate": 3.179986648199727e-07, "loss": 0.0146, "step": 8123 }, { "epoch": 3.2, "learning_rate": 3.179780482258829e-07, "loss": 0.0189, "step": 8124 }, { "epoch": 3.2, "learning_rate": 3.1795742970893017e-07, "loss": 0.0181, "step": 8125 }, { "epoch": 3.2, "learning_rate": 3.179368092694506e-07, "loss": 0.0203, "step": 8126 }, { "epoch": 3.2, "learning_rate": 3.179161869077804e-07, "loss": 0.0191, "step": 8127 }, { "epoch": 3.2, "learning_rate": 3.1789556262425545e-07, "loss": 0.0194, "step": 8128 }, { "epoch": 3.2, "learning_rate": 3.1787493641921205e-07, "loss": 0.0165, "step": 8129 }, { "epoch": 3.2, "learning_rate": 3.178543082929864e-07, "loss": 0.0178, "step": 8130 }, { "epoch": 3.2, "learning_rate": 3.1783367824591456e-07, "loss": 0.02, "step": 8131 }, { "epoch": 3.2, "learning_rate": 3.1781304627833285e-07, "loss": 0.0219, "step": 8132 }, { "epoch": 3.21, "learning_rate": 3.177924123905776e-07, "loss": 0.0172, "step": 8133 }, { "epoch": 3.21, "learning_rate": 3.1777177658298504e-07, "loss": 0.0159, "step": 8134 }, { "epoch": 3.21, "learning_rate": 3.177511388558915e-07, "loss": 0.0197, "step": 8135 }, { "epoch": 3.21, "learning_rate": 3.1773049920963333e-07, "loss": 0.0163, "step": 8136 }, { "epoch": 3.21, "learning_rate": 3.17709857644547e-07, "loss": 0.0203, "step": 8137 }, { "epoch": 3.21, "learning_rate": 3.176892141609688e-07, "loss": 0.0171, "step": 8138 }, { "epoch": 3.21, "learning_rate": 3.1766856875923534e-07, "loss": 0.0177, "step": 8139 }, { "epoch": 3.21, "learning_rate": 3.1764792143968307e-07, "loss": 0.0217, "step": 8140 }, { "epoch": 3.21, "learning_rate": 3.1762727220264847e-07, "loss": 0.0169, "step": 8141 }, { "epoch": 3.21, "learning_rate": 3.176066210484681e-07, "loss": 0.0203, "step": 8142 }, { "epoch": 3.21, "learning_rate": 3.1758596797747846e-07, "loss": 0.0214, "step": 8143 }, { "epoch": 3.21, "learning_rate": 3.175653129900164e-07, "loss": 0.0219, "step": 8144 }, { "epoch": 3.21, "learning_rate": 3.1754465608641835e-07, "loss": 0.0207, "step": 8145 }, { "epoch": 3.21, "learning_rate": 3.1752399726702104e-07, "loss": 0.0183, "step": 8146 }, { "epoch": 3.21, "learning_rate": 3.175033365321612e-07, "loss": 0.0201, "step": 8147 }, { "epoch": 3.21, "learning_rate": 3.174826738821756e-07, "loss": 0.0173, "step": 8148 }, { "epoch": 3.21, "learning_rate": 3.174620093174009e-07, "loss": 0.0168, "step": 8149 }, { "epoch": 3.21, "learning_rate": 3.17441342838174e-07, "loss": 0.0152, "step": 8150 }, { "epoch": 3.21, "learning_rate": 3.174206744448317e-07, "loss": 0.0182, "step": 8151 }, { "epoch": 3.21, "learning_rate": 3.1740000413771087e-07, "loss": 0.0178, "step": 8152 }, { "epoch": 3.21, "learning_rate": 3.173793319171484e-07, "loss": 0.0175, "step": 8153 }, { "epoch": 3.21, "learning_rate": 3.173586577834812e-07, "loss": 0.0194, "step": 8154 }, { "epoch": 3.21, "learning_rate": 3.1733798173704624e-07, "loss": 0.0188, "step": 8155 }, { "epoch": 3.21, "learning_rate": 3.1731730377818055e-07, "loss": 0.0164, "step": 8156 }, { "epoch": 3.21, "learning_rate": 3.1729662390722114e-07, "loss": 0.0209, "step": 8157 }, { "epoch": 3.22, "learning_rate": 3.1727594212450494e-07, "loss": 0.0178, "step": 8158 }, { "epoch": 3.22, "learning_rate": 3.172552584303692e-07, "loss": 0.0169, "step": 8159 }, { "epoch": 3.22, "learning_rate": 3.1723457282515086e-07, "loss": 0.0184, "step": 8160 }, { "epoch": 3.22, "learning_rate": 3.172138853091872e-07, "loss": 0.0244, "step": 8161 }, { "epoch": 3.22, "learning_rate": 3.1719319588281537e-07, "loss": 0.0176, "step": 8162 }, { "epoch": 3.22, "learning_rate": 3.1717250454637257e-07, "loss": 0.0156, "step": 8163 }, { "epoch": 3.22, "learning_rate": 3.1715181130019606e-07, "loss": 0.0182, "step": 8164 }, { "epoch": 3.22, "learning_rate": 3.1713111614462305e-07, "loss": 0.0205, "step": 8165 }, { "epoch": 3.22, "learning_rate": 3.1711041907999077e-07, "loss": 0.019, "step": 8166 }, { "epoch": 3.22, "learning_rate": 3.1708972010663677e-07, "loss": 0.0156, "step": 8167 }, { "epoch": 3.22, "learning_rate": 3.170690192248982e-07, "loss": 0.0212, "step": 8168 }, { "epoch": 3.22, "learning_rate": 3.1704831643511265e-07, "loss": 0.0194, "step": 8169 }, { "epoch": 3.22, "learning_rate": 3.170276117376174e-07, "loss": 0.0194, "step": 8170 }, { "epoch": 3.22, "learning_rate": 3.170069051327498e-07, "loss": 0.0201, "step": 8171 }, { "epoch": 3.22, "learning_rate": 3.169861966208477e-07, "loss": 0.0161, "step": 8172 }, { "epoch": 3.22, "learning_rate": 3.169654862022482e-07, "loss": 0.019, "step": 8173 }, { "epoch": 3.22, "learning_rate": 3.1694477387728916e-07, "loss": 0.0201, "step": 8174 }, { "epoch": 3.22, "learning_rate": 3.16924059646308e-07, "loss": 0.0197, "step": 8175 }, { "epoch": 3.22, "learning_rate": 3.1690334350964235e-07, "loss": 0.0181, "step": 8176 }, { "epoch": 3.22, "learning_rate": 3.168826254676299e-07, "loss": 0.0197, "step": 8177 }, { "epoch": 3.22, "learning_rate": 3.168619055206083e-07, "loss": 0.0201, "step": 8178 }, { "epoch": 3.22, "learning_rate": 3.1684118366891523e-07, "loss": 0.0173, "step": 8179 }, { "epoch": 3.22, "learning_rate": 3.1682045991288846e-07, "loss": 0.0185, "step": 8180 }, { "epoch": 3.22, "learning_rate": 3.1679973425286577e-07, "loss": 0.018, "step": 8181 }, { "epoch": 3.22, "learning_rate": 3.1677900668918487e-07, "loss": 0.023, "step": 8182 }, { "epoch": 3.22, "learning_rate": 3.167582772221837e-07, "loss": 0.0195, "step": 8183 }, { "epoch": 3.23, "learning_rate": 3.1673754585220006e-07, "loss": 0.0187, "step": 8184 }, { "epoch": 3.23, "learning_rate": 3.167168125795718e-07, "loss": 0.0175, "step": 8185 }, { "epoch": 3.23, "learning_rate": 3.166960774046369e-07, "loss": 0.0174, "step": 8186 }, { "epoch": 3.23, "learning_rate": 3.166753403277333e-07, "loss": 0.0214, "step": 8187 }, { "epoch": 3.23, "learning_rate": 3.16654601349199e-07, "loss": 0.0217, "step": 8188 }, { "epoch": 3.23, "learning_rate": 3.166338604693719e-07, "loss": 0.0136, "step": 8189 }, { "epoch": 3.23, "learning_rate": 3.1661311768859025e-07, "loss": 0.0189, "step": 8190 }, { "epoch": 3.23, "learning_rate": 3.1659237300719193e-07, "loss": 0.0187, "step": 8191 }, { "epoch": 3.23, "learning_rate": 3.1657162642551516e-07, "loss": 0.0183, "step": 8192 }, { "epoch": 3.23, "learning_rate": 3.1655087794389806e-07, "loss": 0.0192, "step": 8193 }, { "epoch": 3.23, "learning_rate": 3.1653012756267877e-07, "loss": 0.0167, "step": 8194 }, { "epoch": 3.23, "learning_rate": 3.1650937528219557e-07, "loss": 0.0182, "step": 8195 }, { "epoch": 3.23, "learning_rate": 3.164886211027865e-07, "loss": 0.0216, "step": 8196 }, { "epoch": 3.23, "learning_rate": 3.164678650247901e-07, "loss": 0.021, "step": 8197 }, { "epoch": 3.23, "learning_rate": 3.164471070485444e-07, "loss": 0.0187, "step": 8198 }, { "epoch": 3.23, "learning_rate": 3.164263471743879e-07, "loss": 0.0166, "step": 8199 }, { "epoch": 3.23, "learning_rate": 3.164055854026588e-07, "loss": 0.0209, "step": 8200 }, { "epoch": 3.23, "learning_rate": 3.163848217336957e-07, "loss": 0.0203, "step": 8201 }, { "epoch": 3.23, "learning_rate": 3.163640561678368e-07, "loss": 0.0171, "step": 8202 }, { "epoch": 3.23, "learning_rate": 3.1634328870542063e-07, "loss": 0.0184, "step": 8203 }, { "epoch": 3.23, "learning_rate": 3.163225193467857e-07, "loss": 0.0163, "step": 8204 }, { "epoch": 3.23, "learning_rate": 3.163017480922705e-07, "loss": 0.0196, "step": 8205 }, { "epoch": 3.23, "learning_rate": 3.1628097494221357e-07, "loss": 0.019, "step": 8206 }, { "epoch": 3.23, "learning_rate": 3.162601998969535e-07, "loss": 0.0146, "step": 8207 }, { "epoch": 3.23, "learning_rate": 3.162394229568288e-07, "loss": 0.0209, "step": 8208 }, { "epoch": 3.24, "learning_rate": 3.162186441221782e-07, "loss": 0.0188, "step": 8209 }, { "epoch": 3.24, "learning_rate": 3.1619786339334036e-07, "loss": 0.0176, "step": 8210 }, { "epoch": 3.24, "learning_rate": 3.1617708077065393e-07, "loss": 0.0184, "step": 8211 }, { "epoch": 3.24, "learning_rate": 3.1615629625445765e-07, "loss": 0.0173, "step": 8212 }, { "epoch": 3.24, "learning_rate": 3.1613550984509026e-07, "loss": 0.0202, "step": 8213 }, { "epoch": 3.24, "learning_rate": 3.1611472154289064e-07, "loss": 0.0177, "step": 8214 }, { "epoch": 3.24, "learning_rate": 3.1609393134819744e-07, "loss": 0.0191, "step": 8215 }, { "epoch": 3.24, "learning_rate": 3.1607313926134964e-07, "loss": 0.0143, "step": 8216 }, { "epoch": 3.24, "learning_rate": 3.160523452826861e-07, "loss": 0.0158, "step": 8217 }, { "epoch": 3.24, "learning_rate": 3.1603154941254566e-07, "loss": 0.0191, "step": 8218 }, { "epoch": 3.24, "learning_rate": 3.1601075165126733e-07, "loss": 0.0184, "step": 8219 }, { "epoch": 3.24, "learning_rate": 3.159899519991901e-07, "loss": 0.021, "step": 8220 }, { "epoch": 3.24, "learning_rate": 3.159691504566529e-07, "loss": 0.0194, "step": 8221 }, { "epoch": 3.24, "learning_rate": 3.159483470239948e-07, "loss": 0.0197, "step": 8222 }, { "epoch": 3.24, "learning_rate": 3.1592754170155485e-07, "loss": 0.0222, "step": 8223 }, { "epoch": 3.24, "learning_rate": 3.159067344896722e-07, "loss": 0.016, "step": 8224 }, { "epoch": 3.24, "learning_rate": 3.1588592538868594e-07, "loss": 0.0195, "step": 8225 }, { "epoch": 3.24, "learning_rate": 3.1586511439893515e-07, "loss": 0.0198, "step": 8226 }, { "epoch": 3.24, "learning_rate": 3.158443015207592e-07, "loss": 0.0187, "step": 8227 }, { "epoch": 3.24, "learning_rate": 3.158234867544971e-07, "loss": 0.0189, "step": 8228 }, { "epoch": 3.24, "learning_rate": 3.158026701004883e-07, "loss": 0.0159, "step": 8229 }, { "epoch": 3.24, "learning_rate": 3.157818515590719e-07, "loss": 0.019, "step": 8230 }, { "epoch": 3.24, "learning_rate": 3.157610311305872e-07, "loss": 0.0185, "step": 8231 }, { "epoch": 3.24, "learning_rate": 3.1574020881537383e-07, "loss": 0.0175, "step": 8232 }, { "epoch": 3.24, "learning_rate": 3.1571938461377083e-07, "loss": 0.0166, "step": 8233 }, { "epoch": 3.25, "learning_rate": 3.156985585261178e-07, "loss": 0.0147, "step": 8234 }, { "epoch": 3.25, "learning_rate": 3.1567773055275406e-07, "loss": 0.0205, "step": 8235 }, { "epoch": 3.25, "learning_rate": 3.1565690069401913e-07, "loss": 0.0179, "step": 8236 }, { "epoch": 3.25, "learning_rate": 3.1563606895025256e-07, "loss": 0.0209, "step": 8237 }, { "epoch": 3.25, "learning_rate": 3.156152353217937e-07, "loss": 0.0202, "step": 8238 }, { "epoch": 3.25, "learning_rate": 3.1559439980898227e-07, "loss": 0.0161, "step": 8239 }, { "epoch": 3.25, "learning_rate": 3.1557356241215783e-07, "loss": 0.0155, "step": 8240 }, { "epoch": 3.25, "learning_rate": 3.1555272313165996e-07, "loss": 0.0168, "step": 8241 }, { "epoch": 3.25, "learning_rate": 3.1553188196782833e-07, "loss": 0.0154, "step": 8242 }, { "epoch": 3.25, "learning_rate": 3.155110389210026e-07, "loss": 0.0148, "step": 8243 }, { "epoch": 3.25, "learning_rate": 3.154901939915225e-07, "loss": 0.0149, "step": 8244 }, { "epoch": 3.25, "learning_rate": 3.1546934717972776e-07, "loss": 0.0158, "step": 8245 }, { "epoch": 3.25, "learning_rate": 3.1544849848595814e-07, "loss": 0.0201, "step": 8246 }, { "epoch": 3.25, "learning_rate": 3.1542764791055343e-07, "loss": 0.0197, "step": 8247 }, { "epoch": 3.25, "learning_rate": 3.1540679545385355e-07, "loss": 0.0175, "step": 8248 }, { "epoch": 3.25, "learning_rate": 3.1538594111619827e-07, "loss": 0.0167, "step": 8249 }, { "epoch": 3.25, "learning_rate": 3.1536508489792756e-07, "loss": 0.0209, "step": 8250 }, { "epoch": 3.25, "learning_rate": 3.1534422679938125e-07, "loss": 0.0179, "step": 8251 }, { "epoch": 3.25, "learning_rate": 3.153233668208994e-07, "loss": 0.0186, "step": 8252 }, { "epoch": 3.25, "learning_rate": 3.153025049628219e-07, "loss": 0.0172, "step": 8253 }, { "epoch": 3.25, "learning_rate": 3.1528164122548877e-07, "loss": 0.0198, "step": 8254 }, { "epoch": 3.25, "learning_rate": 3.152607756092402e-07, "loss": 0.0177, "step": 8255 }, { "epoch": 3.25, "learning_rate": 3.1523990811441604e-07, "loss": 0.0164, "step": 8256 }, { "epoch": 3.25, "learning_rate": 3.152190387413566e-07, "loss": 0.0119, "step": 8257 }, { "epoch": 3.25, "learning_rate": 3.151981674904019e-07, "loss": 0.0201, "step": 8258 }, { "epoch": 3.25, "learning_rate": 3.1517729436189215e-07, "loss": 0.014, "step": 8259 }, { "epoch": 3.26, "learning_rate": 3.151564193561676e-07, "loss": 0.0176, "step": 8260 }, { "epoch": 3.26, "learning_rate": 3.151355424735684e-07, "loss": 0.0216, "step": 8261 }, { "epoch": 3.26, "learning_rate": 3.1511466371443486e-07, "loss": 0.0136, "step": 8262 }, { "epoch": 3.26, "learning_rate": 3.150937830791073e-07, "loss": 0.0216, "step": 8263 }, { "epoch": 3.26, "learning_rate": 3.150729005679259e-07, "loss": 0.0157, "step": 8264 }, { "epoch": 3.26, "learning_rate": 3.150520161812312e-07, "loss": 0.02, "step": 8265 }, { "epoch": 3.26, "learning_rate": 3.150311299193634e-07, "loss": 0.0202, "step": 8266 }, { "epoch": 3.26, "learning_rate": 3.150102417826631e-07, "loss": 0.0186, "step": 8267 }, { "epoch": 3.26, "learning_rate": 3.1498935177147064e-07, "loss": 0.0168, "step": 8268 }, { "epoch": 3.26, "learning_rate": 3.1496845988612647e-07, "loss": 0.0181, "step": 8269 }, { "epoch": 3.26, "learning_rate": 3.1494756612697115e-07, "loss": 0.017, "step": 8270 }, { "epoch": 3.26, "learning_rate": 3.149266704943452e-07, "loss": 0.0174, "step": 8271 }, { "epoch": 3.26, "learning_rate": 3.149057729885892e-07, "loss": 0.0181, "step": 8272 }, { "epoch": 3.26, "learning_rate": 3.1488487361004374e-07, "loss": 0.0192, "step": 8273 }, { "epoch": 3.26, "learning_rate": 3.1486397235904944e-07, "loss": 0.0173, "step": 8274 }, { "epoch": 3.26, "learning_rate": 3.1484306923594695e-07, "loss": 0.0171, "step": 8275 }, { "epoch": 3.26, "learning_rate": 3.14822164241077e-07, "loss": 0.0205, "step": 8276 }, { "epoch": 3.26, "learning_rate": 3.1480125737478026e-07, "loss": 0.0184, "step": 8277 }, { "epoch": 3.26, "learning_rate": 3.147803486373975e-07, "loss": 0.017, "step": 8278 }, { "epoch": 3.26, "learning_rate": 3.147594380292695e-07, "loss": 0.0204, "step": 8279 }, { "epoch": 3.26, "learning_rate": 3.1473852555073703e-07, "loss": 0.0232, "step": 8280 }, { "epoch": 3.26, "learning_rate": 3.1471761120214104e-07, "loss": 0.019, "step": 8281 }, { "epoch": 3.26, "learning_rate": 3.1469669498382237e-07, "loss": 0.0164, "step": 8282 }, { "epoch": 3.26, "learning_rate": 3.146757768961219e-07, "loss": 0.0195, "step": 8283 }, { "epoch": 3.26, "learning_rate": 3.146548569393805e-07, "loss": 0.019, "step": 8284 }, { "epoch": 3.27, "learning_rate": 3.1463393511393915e-07, "loss": 0.0214, "step": 8285 }, { "epoch": 3.27, "learning_rate": 3.1461301142013893e-07, "loss": 0.0192, "step": 8286 }, { "epoch": 3.27, "learning_rate": 3.1459208585832083e-07, "loss": 0.0181, "step": 8287 }, { "epoch": 3.27, "learning_rate": 3.145711584288259e-07, "loss": 0.0186, "step": 8288 }, { "epoch": 3.27, "learning_rate": 3.145502291319951e-07, "loss": 0.0185, "step": 8289 }, { "epoch": 3.27, "learning_rate": 3.1452929796816977e-07, "loss": 0.0273, "step": 8290 }, { "epoch": 3.27, "learning_rate": 3.145083649376909e-07, "loss": 0.0185, "step": 8291 }, { "epoch": 3.27, "learning_rate": 3.144874300408998e-07, "loss": 0.0195, "step": 8292 }, { "epoch": 3.27, "learning_rate": 3.1446649327813753e-07, "loss": 0.019, "step": 8293 }, { "epoch": 3.27, "learning_rate": 3.1444555464974535e-07, "loss": 0.0206, "step": 8294 }, { "epoch": 3.27, "learning_rate": 3.1442461415606464e-07, "loss": 0.0169, "step": 8295 }, { "epoch": 3.27, "learning_rate": 3.144036717974366e-07, "loss": 0.0198, "step": 8296 }, { "epoch": 3.27, "learning_rate": 3.1438272757420256e-07, "loss": 0.0152, "step": 8297 }, { "epoch": 3.27, "learning_rate": 3.143617814867039e-07, "loss": 0.0152, "step": 8298 }, { "epoch": 3.27, "learning_rate": 3.14340833535282e-07, "loss": 0.016, "step": 8299 }, { "epoch": 3.27, "learning_rate": 3.1431988372027836e-07, "loss": 0.0167, "step": 8300 }, { "epoch": 3.27, "learning_rate": 3.1429893204203427e-07, "loss": 0.0185, "step": 8301 }, { "epoch": 3.27, "learning_rate": 3.1427797850089135e-07, "loss": 0.0203, "step": 8302 }, { "epoch": 3.27, "learning_rate": 3.142570230971911e-07, "loss": 0.0181, "step": 8303 }, { "epoch": 3.27, "learning_rate": 3.14236065831275e-07, "loss": 0.017, "step": 8304 }, { "epoch": 3.27, "learning_rate": 3.1421510670348463e-07, "loss": 0.0165, "step": 8305 }, { "epoch": 3.27, "learning_rate": 3.141941457141616e-07, "loss": 0.0162, "step": 8306 }, { "epoch": 3.27, "learning_rate": 3.141731828636475e-07, "loss": 0.0185, "step": 8307 }, { "epoch": 3.27, "learning_rate": 3.141522181522842e-07, "loss": 0.0173, "step": 8308 }, { "epoch": 3.27, "learning_rate": 3.141312515804131e-07, "loss": 0.0148, "step": 8309 }, { "epoch": 3.28, "learning_rate": 3.141102831483761e-07, "loss": 0.0211, "step": 8310 }, { "epoch": 3.28, "learning_rate": 3.1408931285651487e-07, "loss": 0.0171, "step": 8311 }, { "epoch": 3.28, "learning_rate": 3.140683407051713e-07, "loss": 0.0183, "step": 8312 }, { "epoch": 3.28, "learning_rate": 3.1404736669468715e-07, "loss": 0.0236, "step": 8313 }, { "epoch": 3.28, "learning_rate": 3.1402639082540417e-07, "loss": 0.0211, "step": 8314 }, { "epoch": 3.28, "learning_rate": 3.1400541309766436e-07, "loss": 0.0177, "step": 8315 }, { "epoch": 3.28, "learning_rate": 3.1398443351180965e-07, "loss": 0.0193, "step": 8316 }, { "epoch": 3.28, "learning_rate": 3.1396345206818183e-07, "loss": 0.021, "step": 8317 }, { "epoch": 3.28, "learning_rate": 3.13942468767123e-07, "loss": 0.0173, "step": 8318 }, { "epoch": 3.28, "learning_rate": 3.1392148360897505e-07, "loss": 0.0155, "step": 8319 }, { "epoch": 3.28, "learning_rate": 3.139004965940801e-07, "loss": 0.0168, "step": 8320 }, { "epoch": 3.28, "learning_rate": 3.138795077227801e-07, "loss": 0.0156, "step": 8321 }, { "epoch": 3.28, "learning_rate": 3.1385851699541723e-07, "loss": 0.0218, "step": 8322 }, { "epoch": 3.28, "learning_rate": 3.1383752441233365e-07, "loss": 0.015, "step": 8323 }, { "epoch": 3.28, "learning_rate": 3.1381652997387136e-07, "loss": 0.0185, "step": 8324 }, { "epoch": 3.28, "learning_rate": 3.1379553368037266e-07, "loss": 0.0178, "step": 8325 }, { "epoch": 3.28, "learning_rate": 3.1377453553217965e-07, "loss": 0.0209, "step": 8326 }, { "epoch": 3.28, "learning_rate": 3.1375353552963464e-07, "loss": 0.0241, "step": 8327 }, { "epoch": 3.28, "learning_rate": 3.1373253367307993e-07, "loss": 0.0199, "step": 8328 }, { "epoch": 3.28, "learning_rate": 3.137115299628577e-07, "loss": 0.0164, "step": 8329 }, { "epoch": 3.28, "learning_rate": 3.1369052439931054e-07, "loss": 0.0184, "step": 8330 }, { "epoch": 3.28, "learning_rate": 3.136695169827804e-07, "loss": 0.0182, "step": 8331 }, { "epoch": 3.28, "learning_rate": 3.1364850771361e-07, "loss": 0.019, "step": 8332 }, { "epoch": 3.28, "learning_rate": 3.136274965921417e-07, "loss": 0.0199, "step": 8333 }, { "epoch": 3.28, "learning_rate": 3.136064836187177e-07, "loss": 0.0198, "step": 8334 }, { "epoch": 3.28, "learning_rate": 3.135854687936809e-07, "loss": 0.0242, "step": 8335 }, { "epoch": 3.29, "learning_rate": 3.135644521173734e-07, "loss": 0.0235, "step": 8336 }, { "epoch": 3.29, "learning_rate": 3.1354343359013807e-07, "loss": 0.0192, "step": 8337 }, { "epoch": 3.29, "learning_rate": 3.135224132123173e-07, "loss": 0.0213, "step": 8338 }, { "epoch": 3.29, "learning_rate": 3.135013909842537e-07, "loss": 0.0179, "step": 8339 }, { "epoch": 3.29, "learning_rate": 3.1348036690628997e-07, "loss": 0.0171, "step": 8340 }, { "epoch": 3.29, "learning_rate": 3.1345934097876874e-07, "loss": 0.0188, "step": 8341 }, { "epoch": 3.29, "learning_rate": 3.134383132020326e-07, "loss": 0.0183, "step": 8342 }, { "epoch": 3.29, "learning_rate": 3.1341728357642445e-07, "loss": 0.0181, "step": 8343 }, { "epoch": 3.29, "learning_rate": 3.1339625210228695e-07, "loss": 0.0185, "step": 8344 }, { "epoch": 3.29, "learning_rate": 3.1337521877996285e-07, "loss": 0.0196, "step": 8345 }, { "epoch": 3.29, "learning_rate": 3.13354183609795e-07, "loss": 0.0154, "step": 8346 }, { "epoch": 3.29, "learning_rate": 3.1333314659212626e-07, "loss": 0.0171, "step": 8347 }, { "epoch": 3.29, "learning_rate": 3.133121077272994e-07, "loss": 0.0174, "step": 8348 }, { "epoch": 3.29, "learning_rate": 3.132910670156575e-07, "loss": 0.0185, "step": 8349 }, { "epoch": 3.29, "learning_rate": 3.1327002445754336e-07, "loss": 0.0173, "step": 8350 }, { "epoch": 3.29, "learning_rate": 3.1324898005329995e-07, "loss": 0.0186, "step": 8351 }, { "epoch": 3.29, "learning_rate": 3.132279338032703e-07, "loss": 0.0182, "step": 8352 }, { "epoch": 3.29, "learning_rate": 3.132068857077974e-07, "loss": 0.0188, "step": 8353 }, { "epoch": 3.29, "learning_rate": 3.131858357672243e-07, "loss": 0.0164, "step": 8354 }, { "epoch": 3.29, "learning_rate": 3.1316478398189416e-07, "loss": 0.0168, "step": 8355 }, { "epoch": 3.29, "learning_rate": 3.131437303521499e-07, "loss": 0.0188, "step": 8356 }, { "epoch": 3.29, "learning_rate": 3.1312267487833494e-07, "loss": 0.0192, "step": 8357 }, { "epoch": 3.29, "learning_rate": 3.1310161756079225e-07, "loss": 0.021, "step": 8358 }, { "epoch": 3.29, "learning_rate": 3.1308055839986504e-07, "loss": 0.0209, "step": 8359 }, { "epoch": 3.29, "learning_rate": 3.130594973958967e-07, "loss": 0.0202, "step": 8360 }, { "epoch": 3.3, "learning_rate": 3.1303843454923034e-07, "loss": 0.0195, "step": 8361 }, { "epoch": 3.3, "learning_rate": 3.1301736986020925e-07, "loss": 0.0201, "step": 8362 }, { "epoch": 3.3, "learning_rate": 3.1299630332917685e-07, "loss": 0.0221, "step": 8363 }, { "epoch": 3.3, "learning_rate": 3.1297523495647644e-07, "loss": 0.0183, "step": 8364 }, { "epoch": 3.3, "learning_rate": 3.1295416474245135e-07, "loss": 0.018, "step": 8365 }, { "epoch": 3.3, "learning_rate": 3.1293309268744514e-07, "loss": 0.0185, "step": 8366 }, { "epoch": 3.3, "learning_rate": 3.1291201879180103e-07, "loss": 0.021, "step": 8367 }, { "epoch": 3.3, "learning_rate": 3.128909430558627e-07, "loss": 0.0206, "step": 8368 }, { "epoch": 3.3, "learning_rate": 3.128698654799735e-07, "loss": 0.0201, "step": 8369 }, { "epoch": 3.3, "learning_rate": 3.128487860644771e-07, "loss": 0.0212, "step": 8370 }, { "epoch": 3.3, "learning_rate": 3.1282770480971703e-07, "loss": 0.0185, "step": 8371 }, { "epoch": 3.3, "learning_rate": 3.1280662171603684e-07, "loss": 0.0201, "step": 8372 }, { "epoch": 3.3, "learning_rate": 3.127855367837801e-07, "loss": 0.0235, "step": 8373 }, { "epoch": 3.3, "learning_rate": 3.127644500132906e-07, "loss": 0.0175, "step": 8374 }, { "epoch": 3.3, "learning_rate": 3.1274336140491187e-07, "loss": 0.0213, "step": 8375 }, { "epoch": 3.3, "learning_rate": 3.127222709589877e-07, "loss": 0.0167, "step": 8376 }, { "epoch": 3.3, "learning_rate": 3.1270117867586196e-07, "loss": 0.0175, "step": 8377 }, { "epoch": 3.3, "learning_rate": 3.1268008455587814e-07, "loss": 0.0179, "step": 8378 }, { "epoch": 3.3, "learning_rate": 3.126589885993803e-07, "loss": 0.0173, "step": 8379 }, { "epoch": 3.3, "learning_rate": 3.1263789080671206e-07, "loss": 0.017, "step": 8380 }, { "epoch": 3.3, "learning_rate": 3.126167911782175e-07, "loss": 0.0209, "step": 8381 }, { "epoch": 3.3, "learning_rate": 3.125956897142403e-07, "loss": 0.0195, "step": 8382 }, { "epoch": 3.3, "learning_rate": 3.1257458641512456e-07, "loss": 0.0231, "step": 8383 }, { "epoch": 3.3, "learning_rate": 3.1255348128121415e-07, "loss": 0.0208, "step": 8384 }, { "epoch": 3.3, "learning_rate": 3.12532374312853e-07, "loss": 0.0213, "step": 8385 }, { "epoch": 3.3, "learning_rate": 3.1251126551038523e-07, "loss": 0.0165, "step": 8386 }, { "epoch": 3.31, "learning_rate": 3.1249015487415483e-07, "loss": 0.023, "step": 8387 }, { "epoch": 3.31, "learning_rate": 3.124690424045059e-07, "loss": 0.0178, "step": 8388 }, { "epoch": 3.31, "learning_rate": 3.124479281017824e-07, "loss": 0.0172, "step": 8389 }, { "epoch": 3.31, "learning_rate": 3.124268119663287e-07, "loss": 0.0181, "step": 8390 }, { "epoch": 3.31, "learning_rate": 3.1240569399848876e-07, "loss": 0.02, "step": 8391 }, { "epoch": 3.31, "learning_rate": 3.123845741986068e-07, "loss": 0.0158, "step": 8392 }, { "epoch": 3.31, "learning_rate": 3.123634525670272e-07, "loss": 0.017, "step": 8393 }, { "epoch": 3.31, "learning_rate": 3.1234232910409403e-07, "loss": 0.018, "step": 8394 }, { "epoch": 3.31, "learning_rate": 3.123212038101517e-07, "loss": 0.0198, "step": 8395 }, { "epoch": 3.31, "learning_rate": 3.123000766855444e-07, "loss": 0.0193, "step": 8396 }, { "epoch": 3.31, "learning_rate": 3.122789477306165e-07, "loss": 0.0174, "step": 8397 }, { "epoch": 3.31, "learning_rate": 3.1225781694571243e-07, "loss": 0.0164, "step": 8398 }, { "epoch": 3.31, "learning_rate": 3.122366843311766e-07, "loss": 0.0203, "step": 8399 }, { "epoch": 3.31, "learning_rate": 3.122155498873533e-07, "loss": 0.0185, "step": 8400 }, { "epoch": 3.31, "learning_rate": 3.1219441361458716e-07, "loss": 0.0204, "step": 8401 }, { "epoch": 3.31, "learning_rate": 3.1217327551322255e-07, "loss": 0.0181, "step": 8402 }, { "epoch": 3.31, "learning_rate": 3.121521355836041e-07, "loss": 0.0192, "step": 8403 }, { "epoch": 3.31, "learning_rate": 3.121309938260762e-07, "loss": 0.0223, "step": 8404 }, { "epoch": 3.31, "learning_rate": 3.121098502409835e-07, "loss": 0.0208, "step": 8405 }, { "epoch": 3.31, "learning_rate": 3.1208870482867075e-07, "loss": 0.0216, "step": 8406 }, { "epoch": 3.31, "learning_rate": 3.120675575894824e-07, "loss": 0.0161, "step": 8407 }, { "epoch": 3.31, "learning_rate": 3.120464085237632e-07, "loss": 0.0186, "step": 8408 }, { "epoch": 3.31, "learning_rate": 3.1202525763185785e-07, "loss": 0.0246, "step": 8409 }, { "epoch": 3.31, "learning_rate": 3.12004104914111e-07, "loss": 0.0159, "step": 8410 }, { "epoch": 3.31, "learning_rate": 3.1198295037086747e-07, "loss": 0.0162, "step": 8411 }, { "epoch": 3.32, "learning_rate": 3.119617940024721e-07, "loss": 0.0184, "step": 8412 }, { "epoch": 3.32, "learning_rate": 3.119406358092696e-07, "loss": 0.0173, "step": 8413 }, { "epoch": 3.32, "learning_rate": 3.119194757916049e-07, "loss": 0.022, "step": 8414 }, { "epoch": 3.32, "learning_rate": 3.118983139498229e-07, "loss": 0.0197, "step": 8415 }, { "epoch": 3.32, "learning_rate": 3.1187715028426833e-07, "loss": 0.0189, "step": 8416 }, { "epoch": 3.32, "learning_rate": 3.118559847952863e-07, "loss": 0.0174, "step": 8417 }, { "epoch": 3.32, "learning_rate": 3.118348174832217e-07, "loss": 0.0164, "step": 8418 }, { "epoch": 3.32, "learning_rate": 3.118136483484196e-07, "loss": 0.0214, "step": 8419 }, { "epoch": 3.32, "learning_rate": 3.1179247739122493e-07, "loss": 0.0154, "step": 8420 }, { "epoch": 3.32, "learning_rate": 3.1177130461198274e-07, "loss": 0.0171, "step": 8421 }, { "epoch": 3.32, "learning_rate": 3.117501300110382e-07, "loss": 0.02, "step": 8422 }, { "epoch": 3.32, "learning_rate": 3.117289535887364e-07, "loss": 0.0182, "step": 8423 }, { "epoch": 3.32, "learning_rate": 3.117077753454224e-07, "loss": 0.0171, "step": 8424 }, { "epoch": 3.32, "learning_rate": 3.1168659528144156e-07, "loss": 0.0224, "step": 8425 }, { "epoch": 3.32, "learning_rate": 3.1166541339713883e-07, "loss": 0.0205, "step": 8426 }, { "epoch": 3.32, "learning_rate": 3.1164422969285965e-07, "loss": 0.0186, "step": 8427 }, { "epoch": 3.32, "learning_rate": 3.116230441689492e-07, "loss": 0.0206, "step": 8428 }, { "epoch": 3.32, "learning_rate": 3.1160185682575275e-07, "loss": 0.0185, "step": 8429 }, { "epoch": 3.32, "learning_rate": 3.1158066766361566e-07, "loss": 0.0196, "step": 8430 }, { "epoch": 3.32, "learning_rate": 3.115594766828833e-07, "loss": 0.0174, "step": 8431 }, { "epoch": 3.32, "learning_rate": 3.11538283883901e-07, "loss": 0.019, "step": 8432 }, { "epoch": 3.32, "learning_rate": 3.1151708926701417e-07, "loss": 0.0182, "step": 8433 }, { "epoch": 3.32, "learning_rate": 3.1149589283256825e-07, "loss": 0.0193, "step": 8434 }, { "epoch": 3.32, "learning_rate": 3.114746945809088e-07, "loss": 0.0176, "step": 8435 }, { "epoch": 3.32, "learning_rate": 3.114534945123812e-07, "loss": 0.0192, "step": 8436 }, { "epoch": 3.33, "learning_rate": 3.114322926273311e-07, "loss": 0.0167, "step": 8437 }, { "epoch": 3.33, "learning_rate": 3.114110889261039e-07, "loss": 0.0164, "step": 8438 }, { "epoch": 3.33, "learning_rate": 3.113898834090453e-07, "loss": 0.0208, "step": 8439 }, { "epoch": 3.33, "learning_rate": 3.1136867607650087e-07, "loss": 0.0164, "step": 8440 }, { "epoch": 3.33, "learning_rate": 3.1134746692881637e-07, "loss": 0.0192, "step": 8441 }, { "epoch": 3.33, "learning_rate": 3.113262559663373e-07, "loss": 0.0226, "step": 8442 }, { "epoch": 3.33, "learning_rate": 3.113050431894095e-07, "loss": 0.02, "step": 8443 }, { "epoch": 3.33, "learning_rate": 3.112838285983786e-07, "loss": 0.0184, "step": 8444 }, { "epoch": 3.33, "learning_rate": 3.112626121935905e-07, "loss": 0.0167, "step": 8445 }, { "epoch": 3.33, "learning_rate": 3.112413939753909e-07, "loss": 0.0157, "step": 8446 }, { "epoch": 3.33, "learning_rate": 3.1122017394412563e-07, "loss": 0.0192, "step": 8447 }, { "epoch": 3.33, "learning_rate": 3.1119895210014057e-07, "loss": 0.018, "step": 8448 }, { "epoch": 3.33, "learning_rate": 3.111777284437816e-07, "loss": 0.0171, "step": 8449 }, { "epoch": 3.33, "learning_rate": 3.1115650297539465e-07, "loss": 0.0182, "step": 8450 }, { "epoch": 3.33, "learning_rate": 3.1113527569532557e-07, "loss": 0.0163, "step": 8451 }, { "epoch": 3.33, "learning_rate": 3.111140466039204e-07, "loss": 0.0182, "step": 8452 }, { "epoch": 3.33, "learning_rate": 3.1109281570152524e-07, "loss": 0.0167, "step": 8453 }, { "epoch": 3.33, "learning_rate": 3.1107158298848593e-07, "loss": 0.018, "step": 8454 }, { "epoch": 3.33, "learning_rate": 3.1105034846514874e-07, "loss": 0.0199, "step": 8455 }, { "epoch": 3.33, "learning_rate": 3.1102911213185953e-07, "loss": 0.0179, "step": 8456 }, { "epoch": 3.33, "learning_rate": 3.110078739889646e-07, "loss": 0.0151, "step": 8457 }, { "epoch": 3.33, "learning_rate": 3.1098663403681e-07, "loss": 0.0193, "step": 8458 }, { "epoch": 3.33, "learning_rate": 3.1096539227574194e-07, "loss": 0.021, "step": 8459 }, { "epoch": 3.33, "learning_rate": 3.1094414870610664e-07, "loss": 0.0163, "step": 8460 }, { "epoch": 3.33, "learning_rate": 3.109229033282504e-07, "loss": 0.0172, "step": 8461 }, { "epoch": 3.33, "learning_rate": 3.1090165614251936e-07, "loss": 0.0218, "step": 8462 }, { "epoch": 3.34, "learning_rate": 3.1088040714925984e-07, "loss": 0.0204, "step": 8463 }, { "epoch": 3.34, "learning_rate": 3.1085915634881825e-07, "loss": 0.0199, "step": 8464 }, { "epoch": 3.34, "learning_rate": 3.108379037415409e-07, "loss": 0.0171, "step": 8465 }, { "epoch": 3.34, "learning_rate": 3.1081664932777416e-07, "loss": 0.0215, "step": 8466 }, { "epoch": 3.34, "learning_rate": 3.107953931078645e-07, "loss": 0.0151, "step": 8467 }, { "epoch": 3.34, "learning_rate": 3.107741350821583e-07, "loss": 0.0195, "step": 8468 }, { "epoch": 3.34, "learning_rate": 3.10752875251002e-07, "loss": 0.0146, "step": 8469 }, { "epoch": 3.34, "learning_rate": 3.1073161361474227e-07, "loss": 0.0181, "step": 8470 }, { "epoch": 3.34, "learning_rate": 3.107103501737254e-07, "loss": 0.0176, "step": 8471 }, { "epoch": 3.34, "learning_rate": 3.1068908492829813e-07, "loss": 0.0155, "step": 8472 }, { "epoch": 3.34, "learning_rate": 3.106678178788071e-07, "loss": 0.0188, "step": 8473 }, { "epoch": 3.34, "learning_rate": 3.106465490255987e-07, "loss": 0.0183, "step": 8474 }, { "epoch": 3.34, "learning_rate": 3.106252783690198e-07, "loss": 0.018, "step": 8475 }, { "epoch": 3.34, "learning_rate": 3.1060400590941694e-07, "loss": 0.0183, "step": 8476 }, { "epoch": 3.34, "learning_rate": 3.105827316471369e-07, "loss": 0.0189, "step": 8477 }, { "epoch": 3.34, "learning_rate": 3.105614555825264e-07, "loss": 0.0203, "step": 8478 }, { "epoch": 3.34, "learning_rate": 3.1054017771593224e-07, "loss": 0.0156, "step": 8479 }, { "epoch": 3.34, "learning_rate": 3.1051889804770117e-07, "loss": 0.0158, "step": 8480 }, { "epoch": 3.34, "learning_rate": 3.1049761657818e-07, "loss": 0.0212, "step": 8481 }, { "epoch": 3.34, "learning_rate": 3.1047633330771563e-07, "loss": 0.0144, "step": 8482 }, { "epoch": 3.34, "learning_rate": 3.104550482366549e-07, "loss": 0.0178, "step": 8483 }, { "epoch": 3.34, "learning_rate": 3.104337613653448e-07, "loss": 0.018, "step": 8484 }, { "epoch": 3.34, "learning_rate": 3.1041247269413223e-07, "loss": 0.0212, "step": 8485 }, { "epoch": 3.34, "learning_rate": 3.1039118222336414e-07, "loss": 0.0164, "step": 8486 }, { "epoch": 3.34, "learning_rate": 3.103698899533876e-07, "loss": 0.0171, "step": 8487 }, { "epoch": 3.35, "learning_rate": 3.1034859588454953e-07, "loss": 0.0181, "step": 8488 }, { "epoch": 3.35, "learning_rate": 3.1032730001719715e-07, "loss": 0.0185, "step": 8489 }, { "epoch": 3.35, "learning_rate": 3.1030600235167737e-07, "loss": 0.0227, "step": 8490 }, { "epoch": 3.35, "learning_rate": 3.1028470288833746e-07, "loss": 0.016, "step": 8491 }, { "epoch": 3.35, "learning_rate": 3.1026340162752444e-07, "loss": 0.0205, "step": 8492 }, { "epoch": 3.35, "learning_rate": 3.102420985695856e-07, "loss": 0.0161, "step": 8493 }, { "epoch": 3.35, "learning_rate": 3.102207937148681e-07, "loss": 0.0182, "step": 8494 }, { "epoch": 3.35, "learning_rate": 3.1019948706371917e-07, "loss": 0.0195, "step": 8495 }, { "epoch": 3.35, "learning_rate": 3.1017817861648603e-07, "loss": 0.0178, "step": 8496 }, { "epoch": 3.35, "learning_rate": 3.101568683735161e-07, "loss": 0.0181, "step": 8497 }, { "epoch": 3.35, "learning_rate": 3.101355563351566e-07, "loss": 0.0175, "step": 8498 }, { "epoch": 3.35, "learning_rate": 3.10114242501755e-07, "loss": 0.0187, "step": 8499 }, { "epoch": 3.35, "learning_rate": 3.1009292687365855e-07, "loss": 0.0176, "step": 8500 }, { "epoch": 3.35, "learning_rate": 3.1007160945121465e-07, "loss": 0.0172, "step": 8501 }, { "epoch": 3.35, "learning_rate": 3.100502902347709e-07, "loss": 0.0218, "step": 8502 }, { "epoch": 3.35, "learning_rate": 3.100289692246746e-07, "loss": 0.0183, "step": 8503 }, { "epoch": 3.35, "learning_rate": 3.100076464212734e-07, "loss": 0.0164, "step": 8504 }, { "epoch": 3.35, "learning_rate": 3.0998632182491473e-07, "loss": 0.017, "step": 8505 }, { "epoch": 3.35, "learning_rate": 3.099649954359461e-07, "loss": 0.0208, "step": 8506 }, { "epoch": 3.35, "learning_rate": 3.099436672547153e-07, "loss": 0.0197, "step": 8507 }, { "epoch": 3.35, "learning_rate": 3.0992233728156976e-07, "loss": 0.0222, "step": 8508 }, { "epoch": 3.35, "learning_rate": 3.0990100551685716e-07, "loss": 0.0194, "step": 8509 }, { "epoch": 3.35, "learning_rate": 3.098796719609252e-07, "loss": 0.0176, "step": 8510 }, { "epoch": 3.35, "learning_rate": 3.098583366141216e-07, "loss": 0.0192, "step": 8511 }, { "epoch": 3.35, "learning_rate": 3.0983699947679405e-07, "loss": 0.0198, "step": 8512 }, { "epoch": 3.36, "learning_rate": 3.098156605492904e-07, "loss": 0.0203, "step": 8513 }, { "epoch": 3.36, "learning_rate": 3.097943198319584e-07, "loss": 0.0175, "step": 8514 }, { "epoch": 3.36, "learning_rate": 3.097729773251458e-07, "loss": 0.0191, "step": 8515 }, { "epoch": 3.36, "learning_rate": 3.097516330292005e-07, "loss": 0.0168, "step": 8516 }, { "epoch": 3.36, "learning_rate": 3.097302869444704e-07, "loss": 0.0188, "step": 8517 }, { "epoch": 3.36, "learning_rate": 3.097089390713034e-07, "loss": 0.0206, "step": 8518 }, { "epoch": 3.36, "learning_rate": 3.0968758941004744e-07, "loss": 0.0202, "step": 8519 }, { "epoch": 3.36, "learning_rate": 3.096662379610504e-07, "loss": 0.0217, "step": 8520 }, { "epoch": 3.36, "learning_rate": 3.096448847246605e-07, "loss": 0.0157, "step": 8521 }, { "epoch": 3.36, "learning_rate": 3.096235297012255e-07, "loss": 0.0212, "step": 8522 }, { "epoch": 3.36, "learning_rate": 3.0960217289109365e-07, "loss": 0.0171, "step": 8523 }, { "epoch": 3.36, "learning_rate": 3.095808142946129e-07, "loss": 0.0184, "step": 8524 }, { "epoch": 3.36, "learning_rate": 3.0955945391213145e-07, "loss": 0.0182, "step": 8525 }, { "epoch": 3.36, "learning_rate": 3.095380917439974e-07, "loss": 0.0181, "step": 8526 }, { "epoch": 3.36, "learning_rate": 3.095167277905589e-07, "loss": 0.0157, "step": 8527 }, { "epoch": 3.36, "learning_rate": 3.094953620521643e-07, "loss": 0.0197, "step": 8528 }, { "epoch": 3.36, "learning_rate": 3.094739945291616e-07, "loss": 0.0188, "step": 8529 }, { "epoch": 3.36, "learning_rate": 3.0945262522189923e-07, "loss": 0.0151, "step": 8530 }, { "epoch": 3.36, "learning_rate": 3.0943125413072533e-07, "loss": 0.021, "step": 8531 }, { "epoch": 3.36, "learning_rate": 3.094098812559884e-07, "loss": 0.018, "step": 8532 }, { "epoch": 3.36, "learning_rate": 3.093885065980367e-07, "loss": 0.0171, "step": 8533 }, { "epoch": 3.36, "learning_rate": 3.093671301572185e-07, "loss": 0.0178, "step": 8534 }, { "epoch": 3.36, "learning_rate": 3.093457519338824e-07, "loss": 0.0189, "step": 8535 }, { "epoch": 3.36, "learning_rate": 3.0932437192837664e-07, "loss": 0.018, "step": 8536 }, { "epoch": 3.36, "learning_rate": 3.093029901410498e-07, "loss": 0.0218, "step": 8537 }, { "epoch": 3.36, "learning_rate": 3.0928160657225035e-07, "loss": 0.0143, "step": 8538 }, { "epoch": 3.37, "learning_rate": 3.0926022122232684e-07, "loss": 0.0196, "step": 8539 }, { "epoch": 3.37, "learning_rate": 3.092388340916277e-07, "loss": 0.0177, "step": 8540 }, { "epoch": 3.37, "learning_rate": 3.092174451805016e-07, "loss": 0.0177, "step": 8541 }, { "epoch": 3.37, "learning_rate": 3.091960544892973e-07, "loss": 0.0167, "step": 8542 }, { "epoch": 3.37, "learning_rate": 3.09174662018363e-07, "loss": 0.0188, "step": 8543 }, { "epoch": 3.37, "learning_rate": 3.0915326776804784e-07, "loss": 0.0182, "step": 8544 }, { "epoch": 3.37, "learning_rate": 3.091318717387002e-07, "loss": 0.0195, "step": 8545 }, { "epoch": 3.37, "learning_rate": 3.09110473930669e-07, "loss": 0.0179, "step": 8546 }, { "epoch": 3.37, "learning_rate": 3.0908907434430285e-07, "loss": 0.0181, "step": 8547 }, { "epoch": 3.37, "learning_rate": 3.090676729799506e-07, "loss": 0.0176, "step": 8548 }, { "epoch": 3.37, "learning_rate": 3.09046269837961e-07, "loss": 0.0203, "step": 8549 }, { "epoch": 3.37, "learning_rate": 3.09024864918683e-07, "loss": 0.0194, "step": 8550 }, { "epoch": 3.37, "learning_rate": 3.0900345822246534e-07, "loss": 0.0168, "step": 8551 }, { "epoch": 3.37, "learning_rate": 3.0898204974965706e-07, "loss": 0.0194, "step": 8552 }, { "epoch": 3.37, "learning_rate": 3.089606395006069e-07, "loss": 0.0206, "step": 8553 }, { "epoch": 3.37, "learning_rate": 3.0893922747566397e-07, "loss": 0.0129, "step": 8554 }, { "epoch": 3.37, "learning_rate": 3.089178136751772e-07, "loss": 0.0207, "step": 8555 }, { "epoch": 3.37, "learning_rate": 3.088963980994956e-07, "loss": 0.0209, "step": 8556 }, { "epoch": 3.37, "learning_rate": 3.0887498074896824e-07, "loss": 0.016, "step": 8557 }, { "epoch": 3.37, "learning_rate": 3.0885356162394407e-07, "loss": 0.0213, "step": 8558 }, { "epoch": 3.37, "learning_rate": 3.088321407247724e-07, "loss": 0.0173, "step": 8559 }, { "epoch": 3.37, "learning_rate": 3.088107180518022e-07, "loss": 0.0183, "step": 8560 }, { "epoch": 3.37, "learning_rate": 3.087892936053827e-07, "loss": 0.0159, "step": 8561 }, { "epoch": 3.37, "learning_rate": 3.08767867385863e-07, "loss": 0.0179, "step": 8562 }, { "epoch": 3.37, "learning_rate": 3.0874643939359234e-07, "loss": 0.0198, "step": 8563 }, { "epoch": 3.38, "learning_rate": 3.087250096289201e-07, "loss": 0.0176, "step": 8564 }, { "epoch": 3.38, "learning_rate": 3.0870357809219537e-07, "loss": 0.0183, "step": 8565 }, { "epoch": 3.38, "learning_rate": 3.0868214478376756e-07, "loss": 0.0175, "step": 8566 }, { "epoch": 3.38, "learning_rate": 3.0866070970398596e-07, "loss": 0.0195, "step": 8567 }, { "epoch": 3.38, "learning_rate": 3.086392728531999e-07, "loss": 0.0183, "step": 8568 }, { "epoch": 3.38, "learning_rate": 3.0861783423175885e-07, "loss": 0.0162, "step": 8569 }, { "epoch": 3.38, "learning_rate": 3.0859639384001214e-07, "loss": 0.0186, "step": 8570 }, { "epoch": 3.38, "learning_rate": 3.085749516783093e-07, "loss": 0.0137, "step": 8571 }, { "epoch": 3.38, "learning_rate": 3.0855350774699974e-07, "loss": 0.0184, "step": 8572 }, { "epoch": 3.38, "learning_rate": 3.08532062046433e-07, "loss": 0.0166, "step": 8573 }, { "epoch": 3.38, "learning_rate": 3.085106145769586e-07, "loss": 0.0209, "step": 8574 }, { "epoch": 3.38, "learning_rate": 3.084891653389261e-07, "loss": 0.0161, "step": 8575 }, { "epoch": 3.38, "learning_rate": 3.084677143326851e-07, "loss": 0.0165, "step": 8576 }, { "epoch": 3.38, "learning_rate": 3.0844626155858517e-07, "loss": 0.0143, "step": 8577 }, { "epoch": 3.38, "learning_rate": 3.08424807016976e-07, "loss": 0.0164, "step": 8578 }, { "epoch": 3.38, "learning_rate": 3.084033507082073e-07, "loss": 0.0188, "step": 8579 }, { "epoch": 3.38, "learning_rate": 3.083818926326287e-07, "loss": 0.0149, "step": 8580 }, { "epoch": 3.38, "learning_rate": 3.0836043279059e-07, "loss": 0.0166, "step": 8581 }, { "epoch": 3.38, "learning_rate": 3.0833897118244094e-07, "loss": 0.0171, "step": 8582 }, { "epoch": 3.38, "learning_rate": 3.083175078085313e-07, "loss": 0.0192, "step": 8583 }, { "epoch": 3.38, "learning_rate": 3.082960426692109e-07, "loss": 0.0151, "step": 8584 }, { "epoch": 3.38, "learning_rate": 3.082745757648296e-07, "loss": 0.02, "step": 8585 }, { "epoch": 3.38, "learning_rate": 3.0825310709573724e-07, "loss": 0.0168, "step": 8586 }, { "epoch": 3.38, "learning_rate": 3.082316366622838e-07, "loss": 0.0187, "step": 8587 }, { "epoch": 3.38, "learning_rate": 3.082101644648191e-07, "loss": 0.0223, "step": 8588 }, { "epoch": 3.38, "learning_rate": 3.0818869050369325e-07, "loss": 0.0159, "step": 8589 }, { "epoch": 3.39, "learning_rate": 3.081672147792562e-07, "loss": 0.0201, "step": 8590 }, { "epoch": 3.39, "learning_rate": 3.0814573729185787e-07, "loss": 0.0165, "step": 8591 }, { "epoch": 3.39, "learning_rate": 3.0812425804184843e-07, "loss": 0.0182, "step": 8592 }, { "epoch": 3.39, "learning_rate": 3.0810277702957794e-07, "loss": 0.0171, "step": 8593 }, { "epoch": 3.39, "learning_rate": 3.0808129425539645e-07, "loss": 0.0154, "step": 8594 }, { "epoch": 3.39, "learning_rate": 3.0805980971965414e-07, "loss": 0.0206, "step": 8595 }, { "epoch": 3.39, "learning_rate": 3.0803832342270115e-07, "loss": 0.0165, "step": 8596 }, { "epoch": 3.39, "learning_rate": 3.080168353648877e-07, "loss": 0.0169, "step": 8597 }, { "epoch": 3.39, "learning_rate": 3.079953455465639e-07, "loss": 0.0192, "step": 8598 }, { "epoch": 3.39, "learning_rate": 3.079738539680802e-07, "loss": 0.016, "step": 8599 }, { "epoch": 3.39, "learning_rate": 3.0795236062978677e-07, "loss": 0.0169, "step": 8600 }, { "epoch": 3.39, "learning_rate": 3.0793086553203393e-07, "loss": 0.019, "step": 8601 }, { "epoch": 3.39, "learning_rate": 3.0790936867517207e-07, "loss": 0.0188, "step": 8602 }, { "epoch": 3.39, "learning_rate": 3.0788787005955143e-07, "loss": 0.0189, "step": 8603 }, { "epoch": 3.39, "learning_rate": 3.0786636968552253e-07, "loss": 0.018, "step": 8604 }, { "epoch": 3.39, "learning_rate": 3.0784486755343567e-07, "loss": 0.019, "step": 8605 }, { "epoch": 3.39, "learning_rate": 3.0782336366364146e-07, "loss": 0.0153, "step": 8606 }, { "epoch": 3.39, "learning_rate": 3.078018580164902e-07, "loss": 0.0134, "step": 8607 }, { "epoch": 3.39, "learning_rate": 3.077803506123326e-07, "loss": 0.0185, "step": 8608 }, { "epoch": 3.39, "learning_rate": 3.0775884145151893e-07, "loss": 0.0192, "step": 8609 }, { "epoch": 3.39, "learning_rate": 3.0773733053440005e-07, "loss": 0.0197, "step": 8610 }, { "epoch": 3.39, "learning_rate": 3.0771581786132644e-07, "loss": 0.0198, "step": 8611 }, { "epoch": 3.39, "learning_rate": 3.076943034326486e-07, "loss": 0.0193, "step": 8612 }, { "epoch": 3.39, "learning_rate": 3.0767278724871746e-07, "loss": 0.0164, "step": 8613 }, { "epoch": 3.39, "learning_rate": 3.0765126930988343e-07, "loss": 0.0192, "step": 8614 }, { "epoch": 3.4, "learning_rate": 3.0762974961649726e-07, "loss": 0.0157, "step": 8615 }, { "epoch": 3.4, "learning_rate": 3.076082281689099e-07, "loss": 0.0179, "step": 8616 }, { "epoch": 3.4, "learning_rate": 3.0758670496747185e-07, "loss": 0.0187, "step": 8617 }, { "epoch": 3.4, "learning_rate": 3.07565180012534e-07, "loss": 0.0195, "step": 8618 }, { "epoch": 3.4, "learning_rate": 3.0754365330444733e-07, "loss": 0.0182, "step": 8619 }, { "epoch": 3.4, "learning_rate": 3.075221248435624e-07, "loss": 0.0215, "step": 8620 }, { "epoch": 3.4, "learning_rate": 3.0750059463023037e-07, "loss": 0.0182, "step": 8621 }, { "epoch": 3.4, "learning_rate": 3.0747906266480194e-07, "loss": 0.0182, "step": 8622 }, { "epoch": 3.4, "learning_rate": 3.0745752894762823e-07, "loss": 0.0157, "step": 8623 }, { "epoch": 3.4, "learning_rate": 3.0743599347906005e-07, "loss": 0.0166, "step": 8624 }, { "epoch": 3.4, "learning_rate": 3.0741445625944846e-07, "loss": 0.0197, "step": 8625 }, { "epoch": 3.4, "learning_rate": 3.0739291728914454e-07, "loss": 0.0163, "step": 8626 }, { "epoch": 3.4, "learning_rate": 3.0737137656849923e-07, "loss": 0.0183, "step": 8627 }, { "epoch": 3.4, "learning_rate": 3.073498340978637e-07, "loss": 0.0202, "step": 8628 }, { "epoch": 3.4, "learning_rate": 3.07328289877589e-07, "loss": 0.0181, "step": 8629 }, { "epoch": 3.4, "learning_rate": 3.0730674390802626e-07, "loss": 0.0183, "step": 8630 }, { "epoch": 3.4, "learning_rate": 3.0728519618952673e-07, "loss": 0.0174, "step": 8631 }, { "epoch": 3.4, "learning_rate": 3.0726364672244155e-07, "loss": 0.0167, "step": 8632 }, { "epoch": 3.4, "learning_rate": 3.07242095507122e-07, "loss": 0.0162, "step": 8633 }, { "epoch": 3.4, "learning_rate": 3.0722054254391927e-07, "loss": 0.0165, "step": 8634 }, { "epoch": 3.4, "learning_rate": 3.0719898783318457e-07, "loss": 0.0171, "step": 8635 }, { "epoch": 3.4, "learning_rate": 3.0717743137526934e-07, "loss": 0.0206, "step": 8636 }, { "epoch": 3.4, "learning_rate": 3.0715587317052485e-07, "loss": 0.0183, "step": 8637 }, { "epoch": 3.4, "learning_rate": 3.071343132193026e-07, "loss": 0.0247, "step": 8638 }, { "epoch": 3.4, "learning_rate": 3.0711275152195375e-07, "loss": 0.0193, "step": 8639 }, { "epoch": 3.41, "learning_rate": 3.0709118807882986e-07, "loss": 0.0202, "step": 8640 }, { "epoch": 3.41, "learning_rate": 3.070696228902824e-07, "loss": 0.0176, "step": 8641 }, { "epoch": 3.41, "learning_rate": 3.070480559566628e-07, "loss": 0.0164, "step": 8642 }, { "epoch": 3.41, "learning_rate": 3.0702648727832256e-07, "loss": 0.0175, "step": 8643 }, { "epoch": 3.41, "learning_rate": 3.0700491685561326e-07, "loss": 0.0211, "step": 8644 }, { "epoch": 3.41, "learning_rate": 3.069833446888864e-07, "loss": 0.0172, "step": 8645 }, { "epoch": 3.41, "learning_rate": 3.069617707784937e-07, "loss": 0.0197, "step": 8646 }, { "epoch": 3.41, "learning_rate": 3.0694019512478663e-07, "loss": 0.0188, "step": 8647 }, { "epoch": 3.41, "learning_rate": 3.0691861772811694e-07, "loss": 0.018, "step": 8648 }, { "epoch": 3.41, "learning_rate": 3.068970385888362e-07, "loss": 0.0194, "step": 8649 }, { "epoch": 3.41, "learning_rate": 3.0687545770729627e-07, "loss": 0.0164, "step": 8650 }, { "epoch": 3.41, "learning_rate": 3.0685387508384885e-07, "loss": 0.016, "step": 8651 }, { "epoch": 3.41, "learning_rate": 3.068322907188456e-07, "loss": 0.017, "step": 8652 }, { "epoch": 3.41, "learning_rate": 3.0681070461263844e-07, "loss": 0.021, "step": 8653 }, { "epoch": 3.41, "learning_rate": 3.0678911676557906e-07, "loss": 0.0163, "step": 8654 }, { "epoch": 3.41, "learning_rate": 3.067675271780194e-07, "loss": 0.0176, "step": 8655 }, { "epoch": 3.41, "learning_rate": 3.067459358503113e-07, "loss": 0.0164, "step": 8656 }, { "epoch": 3.41, "learning_rate": 3.0672434278280674e-07, "loss": 0.015, "step": 8657 }, { "epoch": 3.41, "learning_rate": 3.067027479758576e-07, "loss": 0.0229, "step": 8658 }, { "epoch": 3.41, "learning_rate": 3.066811514298158e-07, "loss": 0.0193, "step": 8659 }, { "epoch": 3.41, "learning_rate": 3.066595531450333e-07, "loss": 0.0153, "step": 8660 }, { "epoch": 3.41, "learning_rate": 3.066379531218623e-07, "loss": 0.0162, "step": 8661 }, { "epoch": 3.41, "learning_rate": 3.066163513606547e-07, "loss": 0.0201, "step": 8662 }, { "epoch": 3.41, "learning_rate": 3.065947478617626e-07, "loss": 0.0191, "step": 8663 }, { "epoch": 3.41, "learning_rate": 3.065731426255381e-07, "loss": 0.0168, "step": 8664 }, { "epoch": 3.41, "learning_rate": 3.0655153565233333e-07, "loss": 0.0193, "step": 8665 }, { "epoch": 3.42, "learning_rate": 3.065299269425005e-07, "loss": 0.017, "step": 8666 }, { "epoch": 3.42, "learning_rate": 3.0650831649639174e-07, "loss": 0.0186, "step": 8667 }, { "epoch": 3.42, "learning_rate": 3.064867043143594e-07, "loss": 0.0193, "step": 8668 }, { "epoch": 3.42, "learning_rate": 3.0646509039675547e-07, "loss": 0.0159, "step": 8669 }, { "epoch": 3.42, "learning_rate": 3.064434747439325e-07, "loss": 0.0172, "step": 8670 }, { "epoch": 3.42, "learning_rate": 3.0642185735624255e-07, "loss": 0.02, "step": 8671 }, { "epoch": 3.42, "learning_rate": 3.064002382340382e-07, "loss": 0.0211, "step": 8672 }, { "epoch": 3.42, "learning_rate": 3.063786173776716e-07, "loss": 0.0212, "step": 8673 }, { "epoch": 3.42, "learning_rate": 3.063569947874952e-07, "loss": 0.0181, "step": 8674 }, { "epoch": 3.42, "learning_rate": 3.063353704638614e-07, "loss": 0.0179, "step": 8675 }, { "epoch": 3.42, "learning_rate": 3.0631374440712276e-07, "loss": 0.0176, "step": 8676 }, { "epoch": 3.42, "learning_rate": 3.062921166176316e-07, "loss": 0.0196, "step": 8677 }, { "epoch": 3.42, "learning_rate": 3.062704870957406e-07, "loss": 0.0189, "step": 8678 }, { "epoch": 3.42, "learning_rate": 3.06248855841802e-07, "loss": 0.0203, "step": 8679 }, { "epoch": 3.42, "learning_rate": 3.0622722285616867e-07, "loss": 0.0164, "step": 8680 }, { "epoch": 3.42, "learning_rate": 3.06205588139193e-07, "loss": 0.0156, "step": 8681 }, { "epoch": 3.42, "learning_rate": 3.061839516912277e-07, "loss": 0.0183, "step": 8682 }, { "epoch": 3.42, "learning_rate": 3.0616231351262525e-07, "loss": 0.0164, "step": 8683 }, { "epoch": 3.42, "learning_rate": 3.0614067360373853e-07, "loss": 0.0189, "step": 8684 }, { "epoch": 3.42, "learning_rate": 3.061190319649201e-07, "loss": 0.0162, "step": 8685 }, { "epoch": 3.42, "learning_rate": 3.0609738859652276e-07, "loss": 0.0135, "step": 8686 }, { "epoch": 3.42, "learning_rate": 3.060757434988993e-07, "loss": 0.0153, "step": 8687 }, { "epoch": 3.42, "learning_rate": 3.0605409667240233e-07, "loss": 0.0178, "step": 8688 }, { "epoch": 3.42, "learning_rate": 3.0603244811738477e-07, "loss": 0.0168, "step": 8689 }, { "epoch": 3.42, "learning_rate": 3.060107978341995e-07, "loss": 0.0193, "step": 8690 }, { "epoch": 3.43, "learning_rate": 3.0598914582319937e-07, "loss": 0.0161, "step": 8691 }, { "epoch": 3.43, "learning_rate": 3.059674920847372e-07, "loss": 0.016, "step": 8692 }, { "epoch": 3.43, "learning_rate": 3.0594583661916604e-07, "loss": 0.02, "step": 8693 }, { "epoch": 3.43, "learning_rate": 3.0592417942683864e-07, "loss": 0.0151, "step": 8694 }, { "epoch": 3.43, "learning_rate": 3.0590252050810816e-07, "loss": 0.0178, "step": 8695 }, { "epoch": 3.43, "learning_rate": 3.0588085986332756e-07, "loss": 0.0191, "step": 8696 }, { "epoch": 3.43, "learning_rate": 3.0585919749284984e-07, "loss": 0.0185, "step": 8697 }, { "epoch": 3.43, "learning_rate": 3.0583753339702815e-07, "loss": 0.0162, "step": 8698 }, { "epoch": 3.43, "learning_rate": 3.058158675762154e-07, "loss": 0.0172, "step": 8699 }, { "epoch": 3.43, "learning_rate": 3.0579420003076495e-07, "loss": 0.0176, "step": 8700 }, { "epoch": 3.43, "learning_rate": 3.0577253076102976e-07, "loss": 0.0175, "step": 8701 }, { "epoch": 3.43, "learning_rate": 3.057508597673632e-07, "loss": 0.0175, "step": 8702 }, { "epoch": 3.43, "learning_rate": 3.0572918705011824e-07, "loss": 0.0194, "step": 8703 }, { "epoch": 3.43, "learning_rate": 3.057075126096482e-07, "loss": 0.0207, "step": 8704 }, { "epoch": 3.43, "learning_rate": 3.056858364463064e-07, "loss": 0.0177, "step": 8705 }, { "epoch": 3.43, "learning_rate": 3.0566415856044607e-07, "loss": 0.0166, "step": 8706 }, { "epoch": 3.43, "learning_rate": 3.0564247895242054e-07, "loss": 0.019, "step": 8707 }, { "epoch": 3.43, "learning_rate": 3.056207976225832e-07, "loss": 0.016, "step": 8708 }, { "epoch": 3.43, "learning_rate": 3.0559911457128735e-07, "loss": 0.0231, "step": 8709 }, { "epoch": 3.43, "learning_rate": 3.0557742979888643e-07, "loss": 0.0182, "step": 8710 }, { "epoch": 3.43, "learning_rate": 3.055557433057338e-07, "loss": 0.0183, "step": 8711 }, { "epoch": 3.43, "learning_rate": 3.055340550921831e-07, "loss": 0.0173, "step": 8712 }, { "epoch": 3.43, "learning_rate": 3.055123651585877e-07, "loss": 0.0254, "step": 8713 }, { "epoch": 3.43, "learning_rate": 3.0549067350530097e-07, "loss": 0.0176, "step": 8714 }, { "epoch": 3.43, "learning_rate": 3.054689801326767e-07, "loss": 0.0148, "step": 8715 }, { "epoch": 3.44, "learning_rate": 3.0544728504106826e-07, "loss": 0.0158, "step": 8716 }, { "epoch": 3.44, "learning_rate": 3.0542558823082943e-07, "loss": 0.0169, "step": 8717 }, { "epoch": 3.44, "learning_rate": 3.0540388970231366e-07, "loss": 0.022, "step": 8718 }, { "epoch": 3.44, "learning_rate": 3.0538218945587477e-07, "loss": 0.0171, "step": 8719 }, { "epoch": 3.44, "learning_rate": 3.0536048749186626e-07, "loss": 0.0156, "step": 8720 }, { "epoch": 3.44, "learning_rate": 3.0533878381064207e-07, "loss": 0.0188, "step": 8721 }, { "epoch": 3.44, "learning_rate": 3.053170784125557e-07, "loss": 0.0199, "step": 8722 }, { "epoch": 3.44, "learning_rate": 3.0529537129796105e-07, "loss": 0.0181, "step": 8723 }, { "epoch": 3.44, "learning_rate": 3.052736624672118e-07, "loss": 0.0196, "step": 8724 }, { "epoch": 3.44, "learning_rate": 3.0525195192066197e-07, "loss": 0.0176, "step": 8725 }, { "epoch": 3.44, "learning_rate": 3.052302396586652e-07, "loss": 0.0198, "step": 8726 }, { "epoch": 3.44, "learning_rate": 3.052085256815755e-07, "loss": 0.0223, "step": 8727 }, { "epoch": 3.44, "learning_rate": 3.051868099897467e-07, "loss": 0.0139, "step": 8728 }, { "epoch": 3.44, "learning_rate": 3.0516509258353276e-07, "loss": 0.0206, "step": 8729 }, { "epoch": 3.44, "learning_rate": 3.0514337346328763e-07, "loss": 0.0154, "step": 8730 }, { "epoch": 3.44, "learning_rate": 3.0512165262936536e-07, "loss": 0.0195, "step": 8731 }, { "epoch": 3.44, "learning_rate": 3.050999300821199e-07, "loss": 0.019, "step": 8732 }, { "epoch": 3.44, "learning_rate": 3.050782058219053e-07, "loss": 0.0213, "step": 8733 }, { "epoch": 3.44, "learning_rate": 3.0505647984907567e-07, "loss": 0.0203, "step": 8734 }, { "epoch": 3.44, "learning_rate": 3.050347521639851e-07, "loss": 0.0163, "step": 8735 }, { "epoch": 3.44, "learning_rate": 3.050130227669877e-07, "loss": 0.018, "step": 8736 }, { "epoch": 3.44, "learning_rate": 3.0499129165843757e-07, "loss": 0.0196, "step": 8737 }, { "epoch": 3.44, "learning_rate": 3.04969558838689e-07, "loss": 0.0173, "step": 8738 }, { "epoch": 3.44, "learning_rate": 3.049478243080961e-07, "loss": 0.0164, "step": 8739 }, { "epoch": 3.44, "learning_rate": 3.0492608806701324e-07, "loss": 0.0154, "step": 8740 }, { "epoch": 3.44, "learning_rate": 3.0490435011579463e-07, "loss": 0.0178, "step": 8741 }, { "epoch": 3.45, "learning_rate": 3.048826104547945e-07, "loss": 0.02, "step": 8742 }, { "epoch": 3.45, "learning_rate": 3.048608690843672e-07, "loss": 0.0198, "step": 8743 }, { "epoch": 3.45, "learning_rate": 3.0483912600486713e-07, "loss": 0.0198, "step": 8744 }, { "epoch": 3.45, "learning_rate": 3.048173812166487e-07, "loss": 0.0167, "step": 8745 }, { "epoch": 3.45, "learning_rate": 3.047956347200662e-07, "loss": 0.017, "step": 8746 }, { "epoch": 3.45, "learning_rate": 3.0477388651547414e-07, "loss": 0.0201, "step": 8747 }, { "epoch": 3.45, "learning_rate": 3.04752136603227e-07, "loss": 0.0184, "step": 8748 }, { "epoch": 3.45, "learning_rate": 3.0473038498367913e-07, "loss": 0.0175, "step": 8749 }, { "epoch": 3.45, "learning_rate": 3.047086316571852e-07, "loss": 0.0199, "step": 8750 }, { "epoch": 3.45, "learning_rate": 3.0468687662409976e-07, "loss": 0.0178, "step": 8751 }, { "epoch": 3.45, "learning_rate": 3.0466511988477733e-07, "loss": 0.0204, "step": 8752 }, { "epoch": 3.45, "learning_rate": 3.0464336143957254e-07, "loss": 0.0158, "step": 8753 }, { "epoch": 3.45, "learning_rate": 3.046216012888399e-07, "loss": 0.0148, "step": 8754 }, { "epoch": 3.45, "learning_rate": 3.045998394329342e-07, "loss": 0.0178, "step": 8755 }, { "epoch": 3.45, "learning_rate": 3.0457807587221015e-07, "loss": 0.0233, "step": 8756 }, { "epoch": 3.45, "learning_rate": 3.045563106070224e-07, "loss": 0.0169, "step": 8757 }, { "epoch": 3.45, "learning_rate": 3.045345436377256e-07, "loss": 0.0187, "step": 8758 }, { "epoch": 3.45, "learning_rate": 3.0451277496467473e-07, "loss": 0.0218, "step": 8759 }, { "epoch": 3.45, "learning_rate": 3.0449100458822445e-07, "loss": 0.0159, "step": 8760 }, { "epoch": 3.45, "learning_rate": 3.044692325087295e-07, "loss": 0.0258, "step": 8761 }, { "epoch": 3.45, "learning_rate": 3.0444745872654494e-07, "loss": 0.0196, "step": 8762 }, { "epoch": 3.45, "learning_rate": 3.044256832420255e-07, "loss": 0.0222, "step": 8763 }, { "epoch": 3.45, "learning_rate": 3.0440390605552617e-07, "loss": 0.0216, "step": 8764 }, { "epoch": 3.45, "learning_rate": 3.043821271674018e-07, "loss": 0.0163, "step": 8765 }, { "epoch": 3.45, "learning_rate": 3.043603465780074e-07, "loss": 0.0179, "step": 8766 }, { "epoch": 3.46, "learning_rate": 3.0433856428769804e-07, "loss": 0.0166, "step": 8767 }, { "epoch": 3.46, "learning_rate": 3.043167802968286e-07, "loss": 0.023, "step": 8768 }, { "epoch": 3.46, "learning_rate": 3.042949946057542e-07, "loss": 0.014, "step": 8769 }, { "epoch": 3.46, "learning_rate": 3.042732072148299e-07, "loss": 0.0181, "step": 8770 }, { "epoch": 3.46, "learning_rate": 3.042514181244107e-07, "loss": 0.0167, "step": 8771 }, { "epoch": 3.46, "learning_rate": 3.042296273348519e-07, "loss": 0.0177, "step": 8772 }, { "epoch": 3.46, "learning_rate": 3.0420783484650866e-07, "loss": 0.0165, "step": 8773 }, { "epoch": 3.46, "learning_rate": 3.0418604065973597e-07, "loss": 0.0183, "step": 8774 }, { "epoch": 3.46, "learning_rate": 3.0416424477488923e-07, "loss": 0.0171, "step": 8775 }, { "epoch": 3.46, "learning_rate": 3.0414244719232355e-07, "loss": 0.0174, "step": 8776 }, { "epoch": 3.46, "learning_rate": 3.0412064791239437e-07, "loss": 0.0159, "step": 8777 }, { "epoch": 3.46, "learning_rate": 3.040988469354568e-07, "loss": 0.0168, "step": 8778 }, { "epoch": 3.46, "learning_rate": 3.0407704426186625e-07, "loss": 0.0167, "step": 8779 }, { "epoch": 3.46, "learning_rate": 3.040552398919781e-07, "loss": 0.0178, "step": 8780 }, { "epoch": 3.46, "learning_rate": 3.040334338261476e-07, "loss": 0.0198, "step": 8781 }, { "epoch": 3.46, "learning_rate": 3.040116260647303e-07, "loss": 0.0182, "step": 8782 }, { "epoch": 3.46, "learning_rate": 3.0398981660808154e-07, "loss": 0.0217, "step": 8783 }, { "epoch": 3.46, "learning_rate": 3.039680054565568e-07, "loss": 0.0189, "step": 8784 }, { "epoch": 3.46, "learning_rate": 3.039461926105116e-07, "loss": 0.0156, "step": 8785 }, { "epoch": 3.46, "learning_rate": 3.039243780703015e-07, "loss": 0.0152, "step": 8786 }, { "epoch": 3.46, "learning_rate": 3.039025618362819e-07, "loss": 0.0229, "step": 8787 }, { "epoch": 3.46, "learning_rate": 3.0388074390880853e-07, "loss": 0.0158, "step": 8788 }, { "epoch": 3.46, "learning_rate": 3.038589242882369e-07, "loss": 0.0172, "step": 8789 }, { "epoch": 3.46, "learning_rate": 3.038371029749226e-07, "loss": 0.0174, "step": 8790 }, { "epoch": 3.46, "learning_rate": 3.0381527996922137e-07, "loss": 0.0174, "step": 8791 }, { "epoch": 3.46, "learning_rate": 3.0379345527148886e-07, "loss": 0.0184, "step": 8792 }, { "epoch": 3.47, "learning_rate": 3.037716288820808e-07, "loss": 0.0168, "step": 8793 }, { "epoch": 3.47, "learning_rate": 3.0374980080135285e-07, "loss": 0.0146, "step": 8794 }, { "epoch": 3.47, "learning_rate": 3.037279710296609e-07, "loss": 0.0163, "step": 8795 }, { "epoch": 3.47, "learning_rate": 3.037061395673606e-07, "loss": 0.0199, "step": 8796 }, { "epoch": 3.47, "learning_rate": 3.036843064148079e-07, "loss": 0.0208, "step": 8797 }, { "epoch": 3.47, "learning_rate": 3.036624715723585e-07, "loss": 0.0179, "step": 8798 }, { "epoch": 3.47, "learning_rate": 3.036406350403684e-07, "loss": 0.0166, "step": 8799 }, { "epoch": 3.47, "learning_rate": 3.036187968191935e-07, "loss": 0.0207, "step": 8800 }, { "epoch": 3.47, "learning_rate": 3.0359695690918967e-07, "loss": 0.0135, "step": 8801 }, { "epoch": 3.47, "learning_rate": 3.035751153107129e-07, "loss": 0.0182, "step": 8802 }, { "epoch": 3.47, "learning_rate": 3.0355327202411916e-07, "loss": 0.0177, "step": 8803 }, { "epoch": 3.47, "learning_rate": 3.035314270497645e-07, "loss": 0.019, "step": 8804 }, { "epoch": 3.47, "learning_rate": 3.035095803880049e-07, "loss": 0.0161, "step": 8805 }, { "epoch": 3.47, "learning_rate": 3.034877320391964e-07, "loss": 0.0189, "step": 8806 }, { "epoch": 3.47, "learning_rate": 3.0346588200369523e-07, "loss": 0.0172, "step": 8807 }, { "epoch": 3.47, "learning_rate": 3.034440302818574e-07, "loss": 0.0198, "step": 8808 }, { "epoch": 3.47, "learning_rate": 3.034221768740391e-07, "loss": 0.0213, "step": 8809 }, { "epoch": 3.47, "learning_rate": 3.0340032178059653e-07, "loss": 0.0173, "step": 8810 }, { "epoch": 3.47, "learning_rate": 3.0337846500188586e-07, "loss": 0.0211, "step": 8811 }, { "epoch": 3.47, "learning_rate": 3.033566065382633e-07, "loss": 0.0203, "step": 8812 }, { "epoch": 3.47, "learning_rate": 3.033347463900852e-07, "loss": 0.016, "step": 8813 }, { "epoch": 3.47, "learning_rate": 3.033128845577077e-07, "loss": 0.0146, "step": 8814 }, { "epoch": 3.47, "learning_rate": 3.032910210414872e-07, "loss": 0.0182, "step": 8815 }, { "epoch": 3.47, "learning_rate": 3.0326915584178007e-07, "loss": 0.0189, "step": 8816 }, { "epoch": 3.47, "learning_rate": 3.0324728895894265e-07, "loss": 0.0171, "step": 8817 }, { "epoch": 3.48, "learning_rate": 3.032254203933313e-07, "loss": 0.0184, "step": 8818 }, { "epoch": 3.48, "learning_rate": 3.0320355014530255e-07, "loss": 0.0196, "step": 8819 }, { "epoch": 3.48, "learning_rate": 3.0318167821521275e-07, "loss": 0.019, "step": 8820 }, { "epoch": 3.48, "learning_rate": 3.0315980460341844e-07, "loss": 0.0194, "step": 8821 }, { "epoch": 3.48, "learning_rate": 3.0313792931027606e-07, "loss": 0.0146, "step": 8822 }, { "epoch": 3.48, "learning_rate": 3.0311605233614217e-07, "loss": 0.0178, "step": 8823 }, { "epoch": 3.48, "learning_rate": 3.030941736813734e-07, "loss": 0.0182, "step": 8824 }, { "epoch": 3.48, "learning_rate": 3.0307229334632624e-07, "loss": 0.0161, "step": 8825 }, { "epoch": 3.48, "learning_rate": 3.030504113313574e-07, "loss": 0.0216, "step": 8826 }, { "epoch": 3.48, "learning_rate": 3.030285276368234e-07, "loss": 0.0175, "step": 8827 }, { "epoch": 3.48, "learning_rate": 3.030066422630811e-07, "loss": 0.0176, "step": 8828 }, { "epoch": 3.48, "learning_rate": 3.0298475521048697e-07, "loss": 0.0178, "step": 8829 }, { "epoch": 3.48, "learning_rate": 3.0296286647939786e-07, "loss": 0.0192, "step": 8830 }, { "epoch": 3.48, "learning_rate": 3.029409760701706e-07, "loss": 0.0169, "step": 8831 }, { "epoch": 3.48, "learning_rate": 3.029190839831618e-07, "loss": 0.0175, "step": 8832 }, { "epoch": 3.48, "learning_rate": 3.028971902187284e-07, "loss": 0.0164, "step": 8833 }, { "epoch": 3.48, "learning_rate": 3.0287529477722716e-07, "loss": 0.0209, "step": 8834 }, { "epoch": 3.48, "learning_rate": 3.02853397659015e-07, "loss": 0.0178, "step": 8835 }, { "epoch": 3.48, "learning_rate": 3.028314988644487e-07, "loss": 0.0174, "step": 8836 }, { "epoch": 3.48, "learning_rate": 3.028095983938853e-07, "loss": 0.0155, "step": 8837 }, { "epoch": 3.48, "learning_rate": 3.0278769624768173e-07, "loss": 0.0157, "step": 8838 }, { "epoch": 3.48, "learning_rate": 3.0276579242619494e-07, "loss": 0.0189, "step": 8839 }, { "epoch": 3.48, "learning_rate": 3.0274388692978175e-07, "loss": 0.0181, "step": 8840 }, { "epoch": 3.48, "learning_rate": 3.027219797587995e-07, "loss": 0.0177, "step": 8841 }, { "epoch": 3.48, "learning_rate": 3.027000709136051e-07, "loss": 0.0142, "step": 8842 }, { "epoch": 3.49, "learning_rate": 3.0267816039455563e-07, "loss": 0.019, "step": 8843 }, { "epoch": 3.49, "learning_rate": 3.0265624820200823e-07, "loss": 0.019, "step": 8844 }, { "epoch": 3.49, "learning_rate": 3.026343343363199e-07, "loss": 0.0212, "step": 8845 }, { "epoch": 3.49, "learning_rate": 3.02612418797848e-07, "loss": 0.0192, "step": 8846 }, { "epoch": 3.49, "learning_rate": 3.0259050158694955e-07, "loss": 0.0213, "step": 8847 }, { "epoch": 3.49, "learning_rate": 3.025685827039819e-07, "loss": 0.0177, "step": 8848 }, { "epoch": 3.49, "learning_rate": 3.025466621493022e-07, "loss": 0.0162, "step": 8849 }, { "epoch": 3.49, "learning_rate": 3.025247399232678e-07, "loss": 0.0191, "step": 8850 }, { "epoch": 3.49, "learning_rate": 3.0250281602623596e-07, "loss": 0.0201, "step": 8851 }, { "epoch": 3.49, "learning_rate": 3.0248089045856394e-07, "loss": 0.0182, "step": 8852 }, { "epoch": 3.49, "learning_rate": 3.024589632206092e-07, "loss": 0.0197, "step": 8853 }, { "epoch": 3.49, "learning_rate": 3.024370343127291e-07, "loss": 0.0224, "step": 8854 }, { "epoch": 3.49, "learning_rate": 3.0241510373528103e-07, "loss": 0.0234, "step": 8855 }, { "epoch": 3.49, "learning_rate": 3.023931714886224e-07, "loss": 0.018, "step": 8856 }, { "epoch": 3.49, "learning_rate": 3.023712375731107e-07, "loss": 0.0146, "step": 8857 }, { "epoch": 3.49, "learning_rate": 3.023493019891034e-07, "loss": 0.0176, "step": 8858 }, { "epoch": 3.49, "learning_rate": 3.0232736473695805e-07, "loss": 0.0179, "step": 8859 }, { "epoch": 3.49, "learning_rate": 3.0230542581703214e-07, "loss": 0.0211, "step": 8860 }, { "epoch": 3.49, "learning_rate": 3.0228348522968333e-07, "loss": 0.0182, "step": 8861 }, { "epoch": 3.49, "learning_rate": 3.0226154297526917e-07, "loss": 0.0157, "step": 8862 }, { "epoch": 3.49, "learning_rate": 3.0223959905414723e-07, "loss": 0.0158, "step": 8863 }, { "epoch": 3.49, "learning_rate": 3.0221765346667526e-07, "loss": 0.0181, "step": 8864 }, { "epoch": 3.49, "learning_rate": 3.0219570621321077e-07, "loss": 0.0211, "step": 8865 }, { "epoch": 3.49, "learning_rate": 3.0217375729411174e-07, "loss": 0.0206, "step": 8866 }, { "epoch": 3.49, "learning_rate": 3.0215180670973566e-07, "loss": 0.0189, "step": 8867 }, { "epoch": 3.49, "learning_rate": 3.021298544604404e-07, "loss": 0.0163, "step": 8868 }, { "epoch": 3.5, "learning_rate": 3.0210790054658375e-07, "loss": 0.0183, "step": 8869 }, { "epoch": 3.5, "learning_rate": 3.020859449685235e-07, "loss": 0.0161, "step": 8870 }, { "epoch": 3.5, "learning_rate": 3.020639877266175e-07, "loss": 0.0173, "step": 8871 }, { "epoch": 3.5, "learning_rate": 3.0204202882122356e-07, "loss": 0.0143, "step": 8872 }, { "epoch": 3.5, "learning_rate": 3.0202006825269977e-07, "loss": 0.0157, "step": 8873 }, { "epoch": 3.5, "learning_rate": 3.019981060214039e-07, "loss": 0.0168, "step": 8874 }, { "epoch": 3.5, "learning_rate": 3.019761421276938e-07, "loss": 0.0141, "step": 8875 }, { "epoch": 3.5, "learning_rate": 3.0195417657192764e-07, "loss": 0.0149, "step": 8876 }, { "epoch": 3.5, "learning_rate": 3.019322093544634e-07, "loss": 0.0178, "step": 8877 }, { "epoch": 3.5, "learning_rate": 3.01910240475659e-07, "loss": 0.0172, "step": 8878 }, { "epoch": 3.5, "learning_rate": 3.018882699358726e-07, "loss": 0.0176, "step": 8879 }, { "epoch": 3.5, "learning_rate": 3.018662977354622e-07, "loss": 0.0172, "step": 8880 }, { "epoch": 3.5, "learning_rate": 3.01844323874786e-07, "loss": 0.0169, "step": 8881 }, { "epoch": 3.5, "learning_rate": 3.018223483542021e-07, "loss": 0.0136, "step": 8882 }, { "epoch": 3.5, "learning_rate": 3.018003711740687e-07, "loss": 0.02, "step": 8883 }, { "epoch": 3.5, "learning_rate": 3.017783923347439e-07, "loss": 0.0246, "step": 8884 }, { "epoch": 3.5, "learning_rate": 3.0175641183658604e-07, "loss": 0.0218, "step": 8885 }, { "epoch": 3.5, "learning_rate": 3.017344296799534e-07, "loss": 0.0174, "step": 8886 }, { "epoch": 3.5, "learning_rate": 3.017124458652041e-07, "loss": 0.0231, "step": 8887 }, { "epoch": 3.5, "learning_rate": 3.0169046039269646e-07, "loss": 0.0196, "step": 8888 }, { "epoch": 3.5, "learning_rate": 3.0166847326278884e-07, "loss": 0.0182, "step": 8889 }, { "epoch": 3.5, "learning_rate": 3.016464844758397e-07, "loss": 0.0184, "step": 8890 }, { "epoch": 3.5, "learning_rate": 3.016244940322074e-07, "loss": 0.0164, "step": 8891 }, { "epoch": 3.5, "learning_rate": 3.0160250193225025e-07, "loss": 0.0209, "step": 8892 }, { "epoch": 3.5, "learning_rate": 3.0158050817632676e-07, "loss": 0.0212, "step": 8893 }, { "epoch": 3.51, "learning_rate": 3.015585127647953e-07, "loss": 0.0171, "step": 8894 }, { "epoch": 3.51, "learning_rate": 3.0153651569801454e-07, "loss": 0.0162, "step": 8895 }, { "epoch": 3.51, "learning_rate": 3.015145169763428e-07, "loss": 0.0181, "step": 8896 }, { "epoch": 3.51, "learning_rate": 3.0149251660013873e-07, "loss": 0.0156, "step": 8897 }, { "epoch": 3.51, "learning_rate": 3.0147051456976097e-07, "loss": 0.0174, "step": 8898 }, { "epoch": 3.51, "learning_rate": 3.0144851088556795e-07, "loss": 0.0143, "step": 8899 }, { "epoch": 3.51, "learning_rate": 3.014265055479184e-07, "loss": 0.0201, "step": 8900 }, { "epoch": 3.51, "learning_rate": 3.0140449855717107e-07, "loss": 0.0148, "step": 8901 }, { "epoch": 3.51, "learning_rate": 3.0138248991368443e-07, "loss": 0.0158, "step": 8902 }, { "epoch": 3.51, "learning_rate": 3.0136047961781724e-07, "loss": 0.0142, "step": 8903 }, { "epoch": 3.51, "learning_rate": 3.0133846766992833e-07, "loss": 0.0184, "step": 8904 }, { "epoch": 3.51, "learning_rate": 3.013164540703764e-07, "loss": 0.0147, "step": 8905 }, { "epoch": 3.51, "learning_rate": 3.012944388195203e-07, "loss": 0.0181, "step": 8906 }, { "epoch": 3.51, "learning_rate": 3.012724219177188e-07, "loss": 0.0167, "step": 8907 }, { "epoch": 3.51, "learning_rate": 3.012504033653307e-07, "loss": 0.0165, "step": 8908 }, { "epoch": 3.51, "learning_rate": 3.012283831627149e-07, "loss": 0.0189, "step": 8909 }, { "epoch": 3.51, "learning_rate": 3.0120636131023033e-07, "loss": 0.0179, "step": 8910 }, { "epoch": 3.51, "learning_rate": 3.0118433780823583e-07, "loss": 0.0153, "step": 8911 }, { "epoch": 3.51, "learning_rate": 3.0116231265709047e-07, "loss": 0.0181, "step": 8912 }, { "epoch": 3.51, "learning_rate": 3.011402858571532e-07, "loss": 0.0164, "step": 8913 }, { "epoch": 3.51, "learning_rate": 3.011182574087829e-07, "loss": 0.0156, "step": 8914 }, { "epoch": 3.51, "learning_rate": 3.010962273123387e-07, "loss": 0.0178, "step": 8915 }, { "epoch": 3.51, "learning_rate": 3.0107419556817965e-07, "loss": 0.0224, "step": 8916 }, { "epoch": 3.51, "learning_rate": 3.0105216217666483e-07, "loss": 0.0177, "step": 8917 }, { "epoch": 3.51, "learning_rate": 3.0103012713815343e-07, "loss": 0.0195, "step": 8918 }, { "epoch": 3.52, "learning_rate": 3.010080904530044e-07, "loss": 0.0196, "step": 8919 }, { "epoch": 3.52, "learning_rate": 3.0098605212157704e-07, "loss": 0.0165, "step": 8920 }, { "epoch": 3.52, "learning_rate": 3.0096401214423056e-07, "loss": 0.0195, "step": 8921 }, { "epoch": 3.52, "learning_rate": 3.009419705213241e-07, "loss": 0.016, "step": 8922 }, { "epoch": 3.52, "learning_rate": 3.0091992725321687e-07, "loss": 0.0184, "step": 8923 }, { "epoch": 3.52, "learning_rate": 3.0089788234026827e-07, "loss": 0.0183, "step": 8924 }, { "epoch": 3.52, "learning_rate": 3.008758357828375e-07, "loss": 0.0175, "step": 8925 }, { "epoch": 3.52, "learning_rate": 3.0085378758128395e-07, "loss": 0.0225, "step": 8926 }, { "epoch": 3.52, "learning_rate": 3.0083173773596695e-07, "loss": 0.0196, "step": 8927 }, { "epoch": 3.52, "learning_rate": 3.008096862472459e-07, "loss": 0.0184, "step": 8928 }, { "epoch": 3.52, "learning_rate": 3.007876331154801e-07, "loss": 0.019, "step": 8929 }, { "epoch": 3.52, "learning_rate": 3.007655783410291e-07, "loss": 0.0154, "step": 8930 }, { "epoch": 3.52, "learning_rate": 3.007435219242523e-07, "loss": 0.0219, "step": 8931 }, { "epoch": 3.52, "learning_rate": 3.0072146386550914e-07, "loss": 0.0207, "step": 8932 }, { "epoch": 3.52, "learning_rate": 3.0069940416515935e-07, "loss": 0.0168, "step": 8933 }, { "epoch": 3.52, "learning_rate": 3.006773428235622e-07, "loss": 0.0147, "step": 8934 }, { "epoch": 3.52, "learning_rate": 3.006552798410774e-07, "loss": 0.0152, "step": 8935 }, { "epoch": 3.52, "learning_rate": 3.0063321521806453e-07, "loss": 0.0156, "step": 8936 }, { "epoch": 3.52, "learning_rate": 3.0061114895488324e-07, "loss": 0.019, "step": 8937 }, { "epoch": 3.52, "learning_rate": 3.0058908105189307e-07, "loss": 0.0132, "step": 8938 }, { "epoch": 3.52, "learning_rate": 3.005670115094538e-07, "loss": 0.0178, "step": 8939 }, { "epoch": 3.52, "learning_rate": 3.005449403279251e-07, "loss": 0.0154, "step": 8940 }, { "epoch": 3.52, "learning_rate": 3.005228675076666e-07, "loss": 0.0186, "step": 8941 }, { "epoch": 3.52, "learning_rate": 3.005007930490382e-07, "loss": 0.0134, "step": 8942 }, { "epoch": 3.52, "learning_rate": 3.004787169523996e-07, "loss": 0.021, "step": 8943 }, { "epoch": 3.52, "learning_rate": 3.004566392181107e-07, "loss": 0.0189, "step": 8944 }, { "epoch": 3.53, "learning_rate": 3.004345598465311e-07, "loss": 0.0207, "step": 8945 }, { "epoch": 3.53, "learning_rate": 3.0041247883802096e-07, "loss": 0.0201, "step": 8946 }, { "epoch": 3.53, "learning_rate": 3.0039039619293993e-07, "loss": 0.0221, "step": 8947 }, { "epoch": 3.53, "learning_rate": 3.0036831191164806e-07, "loss": 0.0207, "step": 8948 }, { "epoch": 3.53, "learning_rate": 3.003462259945052e-07, "loss": 0.0184, "step": 8949 }, { "epoch": 3.53, "learning_rate": 3.0032413844187147e-07, "loss": 0.0218, "step": 8950 }, { "epoch": 3.53, "learning_rate": 3.003020492541067e-07, "loss": 0.0187, "step": 8951 }, { "epoch": 3.53, "learning_rate": 3.0027995843157096e-07, "loss": 0.0197, "step": 8952 }, { "epoch": 3.53, "learning_rate": 3.0025786597462433e-07, "loss": 0.0161, "step": 8953 }, { "epoch": 3.53, "learning_rate": 3.002357718836268e-07, "loss": 0.0147, "step": 8954 }, { "epoch": 3.53, "learning_rate": 3.002136761589386e-07, "loss": 0.0155, "step": 8955 }, { "epoch": 3.53, "learning_rate": 3.001915788009197e-07, "loss": 0.0208, "step": 8956 }, { "epoch": 3.53, "learning_rate": 3.0016947980993045e-07, "loss": 0.0192, "step": 8957 }, { "epoch": 3.53, "learning_rate": 3.001473791863308e-07, "loss": 0.017, "step": 8958 }, { "epoch": 3.53, "learning_rate": 3.001252769304812e-07, "loss": 0.0167, "step": 8959 }, { "epoch": 3.53, "learning_rate": 3.001031730427416e-07, "loss": 0.017, "step": 8960 }, { "epoch": 3.53, "learning_rate": 3.000810675234726e-07, "loss": 0.0183, "step": 8961 }, { "epoch": 3.53, "learning_rate": 3.000589603730342e-07, "loss": 0.0152, "step": 8962 }, { "epoch": 3.53, "learning_rate": 3.0003685159178687e-07, "loss": 0.0206, "step": 8963 }, { "epoch": 3.53, "learning_rate": 3.000147411800909e-07, "loss": 0.0196, "step": 8964 }, { "epoch": 3.53, "learning_rate": 2.9999262913830654e-07, "loss": 0.0179, "step": 8965 }, { "epoch": 3.53, "learning_rate": 2.999705154667944e-07, "loss": 0.0183, "step": 8966 }, { "epoch": 3.53, "learning_rate": 2.9994840016591475e-07, "loss": 0.0177, "step": 8967 }, { "epoch": 3.53, "learning_rate": 2.9992628323602814e-07, "loss": 0.0165, "step": 8968 }, { "epoch": 3.53, "learning_rate": 2.999041646774949e-07, "loss": 0.0207, "step": 8969 }, { "epoch": 3.54, "learning_rate": 2.9988204449067566e-07, "loss": 0.0154, "step": 8970 }, { "epoch": 3.54, "learning_rate": 2.998599226759309e-07, "loss": 0.0166, "step": 8971 }, { "epoch": 3.54, "learning_rate": 2.998377992336212e-07, "loss": 0.016, "step": 8972 }, { "epoch": 3.54, "learning_rate": 2.998156741641071e-07, "loss": 0.0192, "step": 8973 }, { "epoch": 3.54, "learning_rate": 2.9979354746774913e-07, "loss": 0.0163, "step": 8974 }, { "epoch": 3.54, "learning_rate": 2.997714191449081e-07, "loss": 0.0146, "step": 8975 }, { "epoch": 3.54, "learning_rate": 2.997492891959445e-07, "loss": 0.0174, "step": 8976 }, { "epoch": 3.54, "learning_rate": 2.9972715762121916e-07, "loss": 0.0187, "step": 8977 }, { "epoch": 3.54, "learning_rate": 2.9970502442109267e-07, "loss": 0.0149, "step": 8978 }, { "epoch": 3.54, "learning_rate": 2.9968288959592584e-07, "loss": 0.0198, "step": 8979 }, { "epoch": 3.54, "learning_rate": 2.9966075314607936e-07, "loss": 0.0158, "step": 8980 }, { "epoch": 3.54, "learning_rate": 2.996386150719141e-07, "loss": 0.0202, "step": 8981 }, { "epoch": 3.54, "learning_rate": 2.9961647537379093e-07, "loss": 0.0182, "step": 8982 }, { "epoch": 3.54, "learning_rate": 2.9959433405207055e-07, "loss": 0.017, "step": 8983 }, { "epoch": 3.54, "learning_rate": 2.9957219110711386e-07, "loss": 0.0171, "step": 8984 }, { "epoch": 3.54, "learning_rate": 2.995500465392818e-07, "loss": 0.0149, "step": 8985 }, { "epoch": 3.54, "learning_rate": 2.995279003489353e-07, "loss": 0.0187, "step": 8986 }, { "epoch": 3.54, "learning_rate": 2.995057525364353e-07, "loss": 0.0156, "step": 8987 }, { "epoch": 3.54, "learning_rate": 2.994836031021427e-07, "loss": 0.0162, "step": 8988 }, { "epoch": 3.54, "learning_rate": 2.994614520464186e-07, "loss": 0.0165, "step": 8989 }, { "epoch": 3.54, "learning_rate": 2.9943929936962404e-07, "loss": 0.0159, "step": 8990 }, { "epoch": 3.54, "learning_rate": 2.9941714507211997e-07, "loss": 0.0145, "step": 8991 }, { "epoch": 3.54, "learning_rate": 2.993949891542676e-07, "loss": 0.02, "step": 8992 }, { "epoch": 3.54, "learning_rate": 2.993728316164279e-07, "loss": 0.0212, "step": 8993 }, { "epoch": 3.54, "learning_rate": 2.993506724589621e-07, "loss": 0.0203, "step": 8994 }, { "epoch": 3.55, "learning_rate": 2.993285116822313e-07, "loss": 0.0167, "step": 8995 }, { "epoch": 3.55, "learning_rate": 2.9930634928659676e-07, "loss": 0.0137, "step": 8996 }, { "epoch": 3.55, "learning_rate": 2.9928418527241967e-07, "loss": 0.0203, "step": 8997 }, { "epoch": 3.55, "learning_rate": 2.992620196400612e-07, "loss": 0.0239, "step": 8998 }, { "epoch": 3.55, "learning_rate": 2.992398523898827e-07, "loss": 0.02, "step": 8999 }, { "epoch": 3.55, "learning_rate": 2.9921768352224543e-07, "loss": 0.0211, "step": 9000 }, { "epoch": 3.55, "learning_rate": 2.991955130375108e-07, "loss": 0.0187, "step": 9001 }, { "epoch": 3.55, "learning_rate": 2.9917334093603995e-07, "loss": 0.0176, "step": 9002 }, { "epoch": 3.55, "learning_rate": 2.991511672181944e-07, "loss": 0.0163, "step": 9003 }, { "epoch": 3.55, "learning_rate": 2.9912899188433554e-07, "loss": 0.0175, "step": 9004 }, { "epoch": 3.55, "learning_rate": 2.9910681493482476e-07, "loss": 0.0178, "step": 9005 }, { "epoch": 3.55, "learning_rate": 2.990846363700235e-07, "loss": 0.0153, "step": 9006 }, { "epoch": 3.55, "learning_rate": 2.9906245619029336e-07, "loss": 0.0183, "step": 9007 }, { "epoch": 3.55, "learning_rate": 2.9904027439599563e-07, "loss": 0.0226, "step": 9008 }, { "epoch": 3.55, "learning_rate": 2.990180909874919e-07, "loss": 0.0195, "step": 9009 }, { "epoch": 3.55, "learning_rate": 2.989959059651439e-07, "loss": 0.0166, "step": 9010 }, { "epoch": 3.55, "learning_rate": 2.9897371932931304e-07, "loss": 0.0185, "step": 9011 }, { "epoch": 3.55, "learning_rate": 2.9895153108036104e-07, "loss": 0.015, "step": 9012 }, { "epoch": 3.55, "learning_rate": 2.9892934121864936e-07, "loss": 0.0185, "step": 9013 }, { "epoch": 3.55, "learning_rate": 2.9890714974453985e-07, "loss": 0.0187, "step": 9014 }, { "epoch": 3.55, "learning_rate": 2.988849566583941e-07, "loss": 0.0161, "step": 9015 }, { "epoch": 3.55, "learning_rate": 2.988627619605738e-07, "loss": 0.0211, "step": 9016 }, { "epoch": 3.55, "learning_rate": 2.9884056565144084e-07, "loss": 0.0184, "step": 9017 }, { "epoch": 3.55, "learning_rate": 2.9881836773135676e-07, "loss": 0.0153, "step": 9018 }, { "epoch": 3.55, "learning_rate": 2.9879616820068357e-07, "loss": 0.019, "step": 9019 }, { "epoch": 3.55, "learning_rate": 2.987739670597829e-07, "loss": 0.0131, "step": 9020 }, { "epoch": 3.56, "learning_rate": 2.987517643090168e-07, "loss": 0.0163, "step": 9021 }, { "epoch": 3.56, "learning_rate": 2.9872955994874696e-07, "loss": 0.0191, "step": 9022 }, { "epoch": 3.56, "learning_rate": 2.987073539793353e-07, "loss": 0.015, "step": 9023 }, { "epoch": 3.56, "learning_rate": 2.9868514640114385e-07, "loss": 0.0167, "step": 9024 }, { "epoch": 3.56, "learning_rate": 2.986629372145345e-07, "loss": 0.0172, "step": 9025 }, { "epoch": 3.56, "learning_rate": 2.9864072641986917e-07, "loss": 0.0184, "step": 9026 }, { "epoch": 3.56, "learning_rate": 2.9861851401751e-07, "loss": 0.0204, "step": 9027 }, { "epoch": 3.56, "learning_rate": 2.985963000078189e-07, "loss": 0.0181, "step": 9028 }, { "epoch": 3.56, "learning_rate": 2.9857408439115796e-07, "loss": 0.0174, "step": 9029 }, { "epoch": 3.56, "learning_rate": 2.9855186716788927e-07, "loss": 0.0157, "step": 9030 }, { "epoch": 3.56, "learning_rate": 2.985296483383749e-07, "loss": 0.0218, "step": 9031 }, { "epoch": 3.56, "learning_rate": 2.9850742790297696e-07, "loss": 0.0177, "step": 9032 }, { "epoch": 3.56, "learning_rate": 2.9848520586205777e-07, "loss": 0.019, "step": 9033 }, { "epoch": 3.56, "learning_rate": 2.9846298221597927e-07, "loss": 0.0214, "step": 9034 }, { "epoch": 3.56, "learning_rate": 2.9844075696510395e-07, "loss": 0.0178, "step": 9035 }, { "epoch": 3.56, "learning_rate": 2.984185301097938e-07, "loss": 0.0166, "step": 9036 }, { "epoch": 3.56, "learning_rate": 2.983963016504113e-07, "loss": 0.0179, "step": 9037 }, { "epoch": 3.56, "learning_rate": 2.9837407158731845e-07, "loss": 0.0184, "step": 9038 }, { "epoch": 3.56, "learning_rate": 2.983518399208779e-07, "loss": 0.0181, "step": 9039 }, { "epoch": 3.56, "learning_rate": 2.9832960665145177e-07, "loss": 0.018, "step": 9040 }, { "epoch": 3.56, "learning_rate": 2.983073717794025e-07, "loss": 0.0173, "step": 9041 }, { "epoch": 3.56, "learning_rate": 2.982851353050925e-07, "loss": 0.0179, "step": 9042 }, { "epoch": 3.56, "learning_rate": 2.982628972288842e-07, "loss": 0.0181, "step": 9043 }, { "epoch": 3.56, "learning_rate": 2.9824065755114e-07, "loss": 0.0186, "step": 9044 }, { "epoch": 3.56, "learning_rate": 2.982184162722223e-07, "loss": 0.0195, "step": 9045 }, { "epoch": 3.57, "learning_rate": 2.9819617339249384e-07, "loss": 0.0176, "step": 9046 }, { "epoch": 3.57, "learning_rate": 2.981739289123169e-07, "loss": 0.0182, "step": 9047 }, { "epoch": 3.57, "learning_rate": 2.981516828320541e-07, "loss": 0.0167, "step": 9048 }, { "epoch": 3.57, "learning_rate": 2.981294351520681e-07, "loss": 0.0187, "step": 9049 }, { "epoch": 3.57, "learning_rate": 2.9810718587272146e-07, "loss": 0.0173, "step": 9050 }, { "epoch": 3.57, "learning_rate": 2.9808493499437675e-07, "loss": 0.0205, "step": 9051 }, { "epoch": 3.57, "learning_rate": 2.980626825173967e-07, "loss": 0.0169, "step": 9052 }, { "epoch": 3.57, "learning_rate": 2.9804042844214395e-07, "loss": 0.015, "step": 9053 }, { "epoch": 3.57, "learning_rate": 2.9801817276898125e-07, "loss": 0.0178, "step": 9054 }, { "epoch": 3.57, "learning_rate": 2.9799591549827126e-07, "loss": 0.0165, "step": 9055 }, { "epoch": 3.57, "learning_rate": 2.979736566303768e-07, "loss": 0.0196, "step": 9056 }, { "epoch": 3.57, "learning_rate": 2.9795139616566067e-07, "loss": 0.0169, "step": 9057 }, { "epoch": 3.57, "learning_rate": 2.979291341044856e-07, "loss": 0.0185, "step": 9058 }, { "epoch": 3.57, "learning_rate": 2.9790687044721447e-07, "loss": 0.0217, "step": 9059 }, { "epoch": 3.57, "learning_rate": 2.978846051942102e-07, "loss": 0.0187, "step": 9060 }, { "epoch": 3.57, "learning_rate": 2.9786233834583556e-07, "loss": 0.0171, "step": 9061 }, { "epoch": 3.57, "learning_rate": 2.9784006990245364e-07, "loss": 0.014, "step": 9062 }, { "epoch": 3.57, "learning_rate": 2.978177998644272e-07, "loss": 0.0172, "step": 9063 }, { "epoch": 3.57, "learning_rate": 2.9779552823211933e-07, "loss": 0.0149, "step": 9064 }, { "epoch": 3.57, "learning_rate": 2.9777325500589297e-07, "loss": 0.0187, "step": 9065 }, { "epoch": 3.57, "learning_rate": 2.977509801861112e-07, "loss": 0.019, "step": 9066 }, { "epoch": 3.57, "learning_rate": 2.9772870377313696e-07, "loss": 0.0176, "step": 9067 }, { "epoch": 3.57, "learning_rate": 2.977064257673334e-07, "loss": 0.0182, "step": 9068 }, { "epoch": 3.57, "learning_rate": 2.976841461690636e-07, "loss": 0.0142, "step": 9069 }, { "epoch": 3.57, "learning_rate": 2.976618649786907e-07, "loss": 0.0166, "step": 9070 }, { "epoch": 3.57, "learning_rate": 2.9763958219657774e-07, "loss": 0.0137, "step": 9071 }, { "epoch": 3.58, "learning_rate": 2.976172978230881e-07, "loss": 0.0138, "step": 9072 }, { "epoch": 3.58, "learning_rate": 2.975950118585848e-07, "loss": 0.0155, "step": 9073 }, { "epoch": 3.58, "learning_rate": 2.975727243034311e-07, "loss": 0.018, "step": 9074 }, { "epoch": 3.58, "learning_rate": 2.975504351579904e-07, "loss": 0.0172, "step": 9075 }, { "epoch": 3.58, "learning_rate": 2.9752814442262575e-07, "loss": 0.0212, "step": 9076 }, { "epoch": 3.58, "learning_rate": 2.9750585209770065e-07, "loss": 0.0162, "step": 9077 }, { "epoch": 3.58, "learning_rate": 2.9748355818357833e-07, "loss": 0.0166, "step": 9078 }, { "epoch": 3.58, "learning_rate": 2.9746126268062214e-07, "loss": 0.0185, "step": 9079 }, { "epoch": 3.58, "learning_rate": 2.9743896558919554e-07, "loss": 0.0185, "step": 9080 }, { "epoch": 3.58, "learning_rate": 2.974166669096618e-07, "loss": 0.0183, "step": 9081 }, { "epoch": 3.58, "learning_rate": 2.973943666423846e-07, "loss": 0.0191, "step": 9082 }, { "epoch": 3.58, "learning_rate": 2.973720647877271e-07, "loss": 0.0167, "step": 9083 }, { "epoch": 3.58, "learning_rate": 2.97349761346053e-07, "loss": 0.0173, "step": 9084 }, { "epoch": 3.58, "learning_rate": 2.973274563177258e-07, "loss": 0.0189, "step": 9085 }, { "epoch": 3.58, "learning_rate": 2.973051497031089e-07, "loss": 0.0177, "step": 9086 }, { "epoch": 3.58, "learning_rate": 2.9728284150256594e-07, "loss": 0.018, "step": 9087 }, { "epoch": 3.58, "learning_rate": 2.972605317164605e-07, "loss": 0.0154, "step": 9088 }, { "epoch": 3.58, "learning_rate": 2.9723822034515627e-07, "loss": 0.0197, "step": 9089 }, { "epoch": 3.58, "learning_rate": 2.972159073890168e-07, "loss": 0.017, "step": 9090 }, { "epoch": 3.58, "learning_rate": 2.9719359284840587e-07, "loss": 0.019, "step": 9091 }, { "epoch": 3.58, "learning_rate": 2.9717127672368703e-07, "loss": 0.0199, "step": 9092 }, { "epoch": 3.58, "learning_rate": 2.9714895901522406e-07, "loss": 0.0177, "step": 9093 }, { "epoch": 3.58, "learning_rate": 2.971266397233807e-07, "loss": 0.0162, "step": 9094 }, { "epoch": 3.58, "learning_rate": 2.971043188485208e-07, "loss": 0.0149, "step": 9095 }, { "epoch": 3.58, "learning_rate": 2.97081996391008e-07, "loss": 0.019, "step": 9096 }, { "epoch": 3.59, "learning_rate": 2.9705967235120624e-07, "loss": 0.0191, "step": 9097 }, { "epoch": 3.59, "learning_rate": 2.970373467294793e-07, "loss": 0.0221, "step": 9098 }, { "epoch": 3.59, "learning_rate": 2.9701501952619114e-07, "loss": 0.0176, "step": 9099 }, { "epoch": 3.59, "learning_rate": 2.9699269074170554e-07, "loss": 0.0153, "step": 9100 }, { "epoch": 3.59, "learning_rate": 2.9697036037638653e-07, "loss": 0.0215, "step": 9101 }, { "epoch": 3.59, "learning_rate": 2.9694802843059806e-07, "loss": 0.0197, "step": 9102 }, { "epoch": 3.59, "learning_rate": 2.9692569490470403e-07, "loss": 0.0203, "step": 9103 }, { "epoch": 3.59, "learning_rate": 2.9690335979906846e-07, "loss": 0.0165, "step": 9104 }, { "epoch": 3.59, "learning_rate": 2.968810231140554e-07, "loss": 0.0186, "step": 9105 }, { "epoch": 3.59, "learning_rate": 2.9685868485002897e-07, "loss": 0.0183, "step": 9106 }, { "epoch": 3.59, "learning_rate": 2.968363450073531e-07, "loss": 0.0155, "step": 9107 }, { "epoch": 3.59, "learning_rate": 2.968140035863921e-07, "loss": 0.0172, "step": 9108 }, { "epoch": 3.59, "learning_rate": 2.967916605875099e-07, "loss": 0.0181, "step": 9109 }, { "epoch": 3.59, "learning_rate": 2.9676931601107076e-07, "loss": 0.0188, "step": 9110 }, { "epoch": 3.59, "learning_rate": 2.967469698574389e-07, "loss": 0.018, "step": 9111 }, { "epoch": 3.59, "learning_rate": 2.9672462212697835e-07, "loss": 0.0207, "step": 9112 }, { "epoch": 3.59, "learning_rate": 2.9670227282005356e-07, "loss": 0.0196, "step": 9113 }, { "epoch": 3.59, "learning_rate": 2.966799219370286e-07, "loss": 0.0168, "step": 9114 }, { "epoch": 3.59, "learning_rate": 2.966575694782679e-07, "loss": 0.0172, "step": 9115 }, { "epoch": 3.59, "learning_rate": 2.9663521544413574e-07, "loss": 0.0142, "step": 9116 }, { "epoch": 3.59, "learning_rate": 2.9661285983499645e-07, "loss": 0.0187, "step": 9117 }, { "epoch": 3.59, "learning_rate": 2.9659050265121434e-07, "loss": 0.0186, "step": 9118 }, { "epoch": 3.59, "learning_rate": 2.9656814389315384e-07, "loss": 0.0158, "step": 9119 }, { "epoch": 3.59, "learning_rate": 2.965457835611794e-07, "loss": 0.0198, "step": 9120 }, { "epoch": 3.59, "learning_rate": 2.9652342165565536e-07, "loss": 0.0178, "step": 9121 }, { "epoch": 3.6, "learning_rate": 2.965010581769463e-07, "loss": 0.0168, "step": 9122 }, { "epoch": 3.6, "learning_rate": 2.9647869312541656e-07, "loss": 0.0168, "step": 9123 }, { "epoch": 3.6, "learning_rate": 2.9645632650143086e-07, "loss": 0.0169, "step": 9124 }, { "epoch": 3.6, "learning_rate": 2.964339583053536e-07, "loss": 0.0186, "step": 9125 }, { "epoch": 3.6, "learning_rate": 2.964115885375494e-07, "loss": 0.0196, "step": 9126 }, { "epoch": 3.6, "learning_rate": 2.9638921719838273e-07, "loss": 0.0244, "step": 9127 }, { "epoch": 3.6, "learning_rate": 2.9636684428821846e-07, "loss": 0.0202, "step": 9128 }, { "epoch": 3.6, "learning_rate": 2.96344469807421e-07, "loss": 0.0182, "step": 9129 }, { "epoch": 3.6, "learning_rate": 2.963220937563552e-07, "loss": 0.0173, "step": 9130 }, { "epoch": 3.6, "learning_rate": 2.962997161353855e-07, "loss": 0.0161, "step": 9131 }, { "epoch": 3.6, "learning_rate": 2.96277336944877e-07, "loss": 0.0192, "step": 9132 }, { "epoch": 3.6, "learning_rate": 2.962549561851941e-07, "loss": 0.0162, "step": 9133 }, { "epoch": 3.6, "learning_rate": 2.962325738567017e-07, "loss": 0.0169, "step": 9134 }, { "epoch": 3.6, "learning_rate": 2.962101899597646e-07, "loss": 0.0193, "step": 9135 }, { "epoch": 3.6, "learning_rate": 2.961878044947477e-07, "loss": 0.0165, "step": 9136 }, { "epoch": 3.6, "learning_rate": 2.9616541746201574e-07, "loss": 0.0175, "step": 9137 }, { "epoch": 3.6, "learning_rate": 2.961430288619336e-07, "loss": 0.0178, "step": 9138 }, { "epoch": 3.6, "learning_rate": 2.9612063869486625e-07, "loss": 0.0172, "step": 9139 }, { "epoch": 3.6, "learning_rate": 2.960982469611785e-07, "loss": 0.0176, "step": 9140 }, { "epoch": 3.6, "learning_rate": 2.9607585366123546e-07, "loss": 0.0168, "step": 9141 }, { "epoch": 3.6, "learning_rate": 2.9605345879540206e-07, "loss": 0.0174, "step": 9142 }, { "epoch": 3.6, "learning_rate": 2.9603106236404314e-07, "loss": 0.0174, "step": 9143 }, { "epoch": 3.6, "learning_rate": 2.9600866436752393e-07, "loss": 0.0157, "step": 9144 }, { "epoch": 3.6, "learning_rate": 2.959862648062094e-07, "loss": 0.0189, "step": 9145 }, { "epoch": 3.6, "learning_rate": 2.959638636804646e-07, "loss": 0.0172, "step": 9146 }, { "epoch": 3.6, "learning_rate": 2.9594146099065473e-07, "loss": 0.0182, "step": 9147 }, { "epoch": 3.61, "learning_rate": 2.959190567371448e-07, "loss": 0.0181, "step": 9148 }, { "epoch": 3.61, "learning_rate": 2.9589665092030005e-07, "loss": 0.0192, "step": 9149 }, { "epoch": 3.61, "learning_rate": 2.9587424354048564e-07, "loss": 0.0156, "step": 9150 }, { "epoch": 3.61, "learning_rate": 2.9585183459806675e-07, "loss": 0.0198, "step": 9151 }, { "epoch": 3.61, "learning_rate": 2.958294240934087e-07, "loss": 0.0182, "step": 9152 }, { "epoch": 3.61, "learning_rate": 2.9580701202687663e-07, "loss": 0.0167, "step": 9153 }, { "epoch": 3.61, "learning_rate": 2.957845983988359e-07, "loss": 0.016, "step": 9154 }, { "epoch": 3.61, "learning_rate": 2.9576218320965175e-07, "loss": 0.0213, "step": 9155 }, { "epoch": 3.61, "learning_rate": 2.9573976645968957e-07, "loss": 0.0165, "step": 9156 }, { "epoch": 3.61, "learning_rate": 2.957173481493148e-07, "loss": 0.0146, "step": 9157 }, { "epoch": 3.61, "learning_rate": 2.956949282788926e-07, "loss": 0.0166, "step": 9158 }, { "epoch": 3.61, "learning_rate": 2.9567250684878857e-07, "loss": 0.0142, "step": 9159 }, { "epoch": 3.61, "learning_rate": 2.9565008385936804e-07, "loss": 0.0192, "step": 9160 }, { "epoch": 3.61, "learning_rate": 2.9562765931099655e-07, "loss": 0.0176, "step": 9161 }, { "epoch": 3.61, "learning_rate": 2.9560523320403955e-07, "loss": 0.0168, "step": 9162 }, { "epoch": 3.61, "learning_rate": 2.955828055388625e-07, "loss": 0.0209, "step": 9163 }, { "epoch": 3.61, "learning_rate": 2.955603763158311e-07, "loss": 0.0179, "step": 9164 }, { "epoch": 3.61, "learning_rate": 2.955379455353107e-07, "loss": 0.0176, "step": 9165 }, { "epoch": 3.61, "learning_rate": 2.95515513197667e-07, "loss": 0.0156, "step": 9166 }, { "epoch": 3.61, "learning_rate": 2.9549307930326565e-07, "loss": 0.0163, "step": 9167 }, { "epoch": 3.61, "learning_rate": 2.954706438524722e-07, "loss": 0.0206, "step": 9168 }, { "epoch": 3.61, "learning_rate": 2.9544820684565236e-07, "loss": 0.0206, "step": 9169 }, { "epoch": 3.61, "learning_rate": 2.9542576828317184e-07, "loss": 0.0208, "step": 9170 }, { "epoch": 3.61, "learning_rate": 2.9540332816539627e-07, "loss": 0.0159, "step": 9171 }, { "epoch": 3.61, "learning_rate": 2.9538088649269146e-07, "loss": 0.0204, "step": 9172 }, { "epoch": 3.62, "learning_rate": 2.9535844326542326e-07, "loss": 0.0189, "step": 9173 }, { "epoch": 3.62, "learning_rate": 2.9533599848395724e-07, "loss": 0.0203, "step": 9174 }, { "epoch": 3.62, "learning_rate": 2.953135521486594e-07, "loss": 0.0214, "step": 9175 }, { "epoch": 3.62, "learning_rate": 2.9529110425989546e-07, "loss": 0.0191, "step": 9176 }, { "epoch": 3.62, "learning_rate": 2.9526865481803145e-07, "loss": 0.0186, "step": 9177 }, { "epoch": 3.62, "learning_rate": 2.9524620382343306e-07, "loss": 0.0141, "step": 9178 }, { "epoch": 3.62, "learning_rate": 2.9522375127646644e-07, "loss": 0.0129, "step": 9179 }, { "epoch": 3.62, "learning_rate": 2.952012971774973e-07, "loss": 0.0161, "step": 9180 }, { "epoch": 3.62, "learning_rate": 2.9517884152689166e-07, "loss": 0.0199, "step": 9181 }, { "epoch": 3.62, "learning_rate": 2.9515638432501566e-07, "loss": 0.0197, "step": 9182 }, { "epoch": 3.62, "learning_rate": 2.951339255722351e-07, "loss": 0.015, "step": 9183 }, { "epoch": 3.62, "learning_rate": 2.9511146526891625e-07, "loss": 0.0153, "step": 9184 }, { "epoch": 3.62, "learning_rate": 2.95089003415425e-07, "loss": 0.0142, "step": 9185 }, { "epoch": 3.62, "learning_rate": 2.9506654001212756e-07, "loss": 0.0161, "step": 9186 }, { "epoch": 3.62, "learning_rate": 2.950440750593899e-07, "loss": 0.0151, "step": 9187 }, { "epoch": 3.62, "learning_rate": 2.950216085575784e-07, "loss": 0.0198, "step": 9188 }, { "epoch": 3.62, "learning_rate": 2.94999140507059e-07, "loss": 0.0183, "step": 9189 }, { "epoch": 3.62, "learning_rate": 2.949766709081981e-07, "loss": 0.0159, "step": 9190 }, { "epoch": 3.62, "learning_rate": 2.949541997613617e-07, "loss": 0.0186, "step": 9191 }, { "epoch": 3.62, "learning_rate": 2.949317270669162e-07, "loss": 0.0248, "step": 9192 }, { "epoch": 3.62, "learning_rate": 2.949092528252279e-07, "loss": 0.0216, "step": 9193 }, { "epoch": 3.62, "learning_rate": 2.948867770366629e-07, "loss": 0.0157, "step": 9194 }, { "epoch": 3.62, "learning_rate": 2.948642997015877e-07, "loss": 0.0172, "step": 9195 }, { "epoch": 3.62, "learning_rate": 2.948418208203686e-07, "loss": 0.018, "step": 9196 }, { "epoch": 3.62, "learning_rate": 2.94819340393372e-07, "loss": 0.0199, "step": 9197 }, { "epoch": 3.63, "learning_rate": 2.947968584209642e-07, "loss": 0.0179, "step": 9198 }, { "epoch": 3.63, "learning_rate": 2.947743749035117e-07, "loss": 0.0178, "step": 9199 }, { "epoch": 3.63, "learning_rate": 2.9475188984138095e-07, "loss": 0.0188, "step": 9200 }, { "epoch": 3.63, "learning_rate": 2.947294032349384e-07, "loss": 0.0174, "step": 9201 }, { "epoch": 3.63, "learning_rate": 2.9470691508455057e-07, "loss": 0.0172, "step": 9202 }, { "epoch": 3.63, "learning_rate": 2.946844253905839e-07, "loss": 0.0181, "step": 9203 }, { "epoch": 3.63, "learning_rate": 2.9466193415340497e-07, "loss": 0.0173, "step": 9204 }, { "epoch": 3.63, "learning_rate": 2.9463944137338047e-07, "loss": 0.019, "step": 9205 }, { "epoch": 3.63, "learning_rate": 2.946169470508769e-07, "loss": 0.0163, "step": 9206 }, { "epoch": 3.63, "learning_rate": 2.9459445118626084e-07, "loss": 0.0201, "step": 9207 }, { "epoch": 3.63, "learning_rate": 2.9457195377989904e-07, "loss": 0.0141, "step": 9208 }, { "epoch": 3.63, "learning_rate": 2.945494548321581e-07, "loss": 0.0188, "step": 9209 }, { "epoch": 3.63, "learning_rate": 2.945269543434048e-07, "loss": 0.0174, "step": 9210 }, { "epoch": 3.63, "learning_rate": 2.945044523140057e-07, "loss": 0.0149, "step": 9211 }, { "epoch": 3.63, "learning_rate": 2.9448194874432777e-07, "loss": 0.0163, "step": 9212 }, { "epoch": 3.63, "learning_rate": 2.9445944363473763e-07, "loss": 0.016, "step": 9213 }, { "epoch": 3.63, "learning_rate": 2.9443693698560207e-07, "loss": 0.0169, "step": 9214 }, { "epoch": 3.63, "learning_rate": 2.94414428797288e-07, "loss": 0.0163, "step": 9215 }, { "epoch": 3.63, "learning_rate": 2.9439191907016223e-07, "loss": 0.017, "step": 9216 }, { "epoch": 3.63, "learning_rate": 2.943694078045917e-07, "loss": 0.0172, "step": 9217 }, { "epoch": 3.63, "learning_rate": 2.9434689500094317e-07, "loss": 0.0155, "step": 9218 }, { "epoch": 3.63, "learning_rate": 2.9432438065958367e-07, "loss": 0.0191, "step": 9219 }, { "epoch": 3.63, "learning_rate": 2.9430186478088017e-07, "loss": 0.0186, "step": 9220 }, { "epoch": 3.63, "learning_rate": 2.9427934736519955e-07, "loss": 0.0191, "step": 9221 }, { "epoch": 3.63, "learning_rate": 2.9425682841290884e-07, "loss": 0.0186, "step": 9222 }, { "epoch": 3.63, "learning_rate": 2.942343079243751e-07, "loss": 0.0178, "step": 9223 }, { "epoch": 3.64, "learning_rate": 2.9421178589996536e-07, "loss": 0.0153, "step": 9224 }, { "epoch": 3.64, "learning_rate": 2.941892623400467e-07, "loss": 0.0173, "step": 9225 }, { "epoch": 3.64, "learning_rate": 2.9416673724498624e-07, "loss": 0.0175, "step": 9226 }, { "epoch": 3.64, "learning_rate": 2.94144210615151e-07, "loss": 0.0194, "step": 9227 }, { "epoch": 3.64, "learning_rate": 2.9412168245090825e-07, "loss": 0.0164, "step": 9228 }, { "epoch": 3.64, "learning_rate": 2.9409915275262516e-07, "loss": 0.0217, "step": 9229 }, { "epoch": 3.64, "learning_rate": 2.9407662152066884e-07, "loss": 0.0153, "step": 9230 }, { "epoch": 3.64, "learning_rate": 2.940540887554066e-07, "loss": 0.0189, "step": 9231 }, { "epoch": 3.64, "learning_rate": 2.9403155445720564e-07, "loss": 0.0153, "step": 9232 }, { "epoch": 3.64, "learning_rate": 2.940090186264333e-07, "loss": 0.0185, "step": 9233 }, { "epoch": 3.64, "learning_rate": 2.9398648126345673e-07, "loss": 0.0177, "step": 9234 }, { "epoch": 3.64, "learning_rate": 2.939639423686435e-07, "loss": 0.0189, "step": 9235 }, { "epoch": 3.64, "learning_rate": 2.9394140194236073e-07, "loss": 0.0174, "step": 9236 }, { "epoch": 3.64, "learning_rate": 2.939188599849759e-07, "loss": 0.0157, "step": 9237 }, { "epoch": 3.64, "learning_rate": 2.938963164968564e-07, "loss": 0.0145, "step": 9238 }, { "epoch": 3.64, "learning_rate": 2.9387377147836963e-07, "loss": 0.021, "step": 9239 }, { "epoch": 3.64, "learning_rate": 2.9385122492988303e-07, "loss": 0.0239, "step": 9240 }, { "epoch": 3.64, "learning_rate": 2.9382867685176415e-07, "loss": 0.0168, "step": 9241 }, { "epoch": 3.64, "learning_rate": 2.9380612724438043e-07, "loss": 0.0165, "step": 9242 }, { "epoch": 3.64, "learning_rate": 2.9378357610809943e-07, "loss": 0.0165, "step": 9243 }, { "epoch": 3.64, "learning_rate": 2.9376102344328867e-07, "loss": 0.0158, "step": 9244 }, { "epoch": 3.64, "learning_rate": 2.937384692503157e-07, "loss": 0.0156, "step": 9245 }, { "epoch": 3.64, "learning_rate": 2.937159135295482e-07, "loss": 0.0175, "step": 9246 }, { "epoch": 3.64, "learning_rate": 2.936933562813536e-07, "loss": 0.0168, "step": 9247 }, { "epoch": 3.64, "learning_rate": 2.9367079750609984e-07, "loss": 0.0162, "step": 9248 }, { "epoch": 3.65, "learning_rate": 2.9364823720415437e-07, "loss": 0.0186, "step": 9249 }, { "epoch": 3.65, "learning_rate": 2.93625675375885e-07, "loss": 0.0171, "step": 9250 }, { "epoch": 3.65, "learning_rate": 2.9360311202165937e-07, "loss": 0.0161, "step": 9251 }, { "epoch": 3.65, "learning_rate": 2.935805471418453e-07, "loss": 0.0162, "step": 9252 }, { "epoch": 3.65, "learning_rate": 2.935579807368106e-07, "loss": 0.0181, "step": 9253 }, { "epoch": 3.65, "learning_rate": 2.9353541280692287e-07, "loss": 0.0194, "step": 9254 }, { "epoch": 3.65, "learning_rate": 2.935128433525502e-07, "loss": 0.0182, "step": 9255 }, { "epoch": 3.65, "learning_rate": 2.9349027237406023e-07, "loss": 0.017, "step": 9256 }, { "epoch": 3.65, "learning_rate": 2.934676998718209e-07, "loss": 0.0159, "step": 9257 }, { "epoch": 3.65, "learning_rate": 2.934451258462001e-07, "loss": 0.0167, "step": 9258 }, { "epoch": 3.65, "learning_rate": 2.934225502975659e-07, "loss": 0.0203, "step": 9259 }, { "epoch": 3.65, "learning_rate": 2.9339997322628594e-07, "loss": 0.0208, "step": 9260 }, { "epoch": 3.65, "learning_rate": 2.9337739463272855e-07, "loss": 0.0154, "step": 9261 }, { "epoch": 3.65, "learning_rate": 2.9335481451726143e-07, "loss": 0.0169, "step": 9262 }, { "epoch": 3.65, "learning_rate": 2.9333223288025267e-07, "loss": 0.0194, "step": 9263 }, { "epoch": 3.65, "learning_rate": 2.9330964972207046e-07, "loss": 0.0141, "step": 9264 }, { "epoch": 3.65, "learning_rate": 2.932870650430828e-07, "loss": 0.0188, "step": 9265 }, { "epoch": 3.65, "learning_rate": 2.932644788436577e-07, "loss": 0.0122, "step": 9266 }, { "epoch": 3.65, "learning_rate": 2.9324189112416336e-07, "loss": 0.0154, "step": 9267 }, { "epoch": 3.65, "learning_rate": 2.932193018849679e-07, "loss": 0.0156, "step": 9268 }, { "epoch": 3.65, "learning_rate": 2.931967111264394e-07, "loss": 0.0164, "step": 9269 }, { "epoch": 3.65, "learning_rate": 2.931741188489463e-07, "loss": 0.0193, "step": 9270 }, { "epoch": 3.65, "learning_rate": 2.9315152505285663e-07, "loss": 0.0175, "step": 9271 }, { "epoch": 3.65, "learning_rate": 2.931289297385386e-07, "loss": 0.0197, "step": 9272 }, { "epoch": 3.65, "learning_rate": 2.9310633290636066e-07, "loss": 0.0188, "step": 9273 }, { "epoch": 3.65, "learning_rate": 2.9308373455669093e-07, "loss": 0.016, "step": 9274 }, { "epoch": 3.66, "learning_rate": 2.9306113468989784e-07, "loss": 0.0198, "step": 9275 }, { "epoch": 3.66, "learning_rate": 2.9303853330634964e-07, "loss": 0.0164, "step": 9276 }, { "epoch": 3.66, "learning_rate": 2.930159304064148e-07, "loss": 0.0178, "step": 9277 }, { "epoch": 3.66, "learning_rate": 2.929933259904616e-07, "loss": 0.0155, "step": 9278 }, { "epoch": 3.66, "learning_rate": 2.929707200588586e-07, "loss": 0.0226, "step": 9279 }, { "epoch": 3.66, "learning_rate": 2.9294811261197407e-07, "loss": 0.0222, "step": 9280 }, { "epoch": 3.66, "learning_rate": 2.929255036501767e-07, "loss": 0.0169, "step": 9281 }, { "epoch": 3.66, "learning_rate": 2.9290289317383477e-07, "loss": 0.0171, "step": 9282 }, { "epoch": 3.66, "learning_rate": 2.928802811833168e-07, "loss": 0.0171, "step": 9283 }, { "epoch": 3.66, "learning_rate": 2.928576676789915e-07, "loss": 0.0179, "step": 9284 }, { "epoch": 3.66, "learning_rate": 2.9283505266122734e-07, "loss": 0.0173, "step": 9285 }, { "epoch": 3.66, "learning_rate": 2.928124361303929e-07, "loss": 0.0174, "step": 9286 }, { "epoch": 3.66, "learning_rate": 2.9278981808685687e-07, "loss": 0.0173, "step": 9287 }, { "epoch": 3.66, "learning_rate": 2.927671985309878e-07, "loss": 0.019, "step": 9288 }, { "epoch": 3.66, "learning_rate": 2.927445774631543e-07, "loss": 0.0218, "step": 9289 }, { "epoch": 3.66, "learning_rate": 2.9272195488372523e-07, "loss": 0.0171, "step": 9290 }, { "epoch": 3.66, "learning_rate": 2.9269933079306924e-07, "loss": 0.019, "step": 9291 }, { "epoch": 3.66, "learning_rate": 2.92676705191555e-07, "loss": 0.0176, "step": 9292 }, { "epoch": 3.66, "learning_rate": 2.9265407807955126e-07, "loss": 0.0181, "step": 9293 }, { "epoch": 3.66, "learning_rate": 2.926314494574269e-07, "loss": 0.0176, "step": 9294 }, { "epoch": 3.66, "learning_rate": 2.9260881932555074e-07, "loss": 0.0147, "step": 9295 }, { "epoch": 3.66, "learning_rate": 2.9258618768429154e-07, "loss": 0.0166, "step": 9296 }, { "epoch": 3.66, "learning_rate": 2.925635545340182e-07, "loss": 0.0177, "step": 9297 }, { "epoch": 3.66, "learning_rate": 2.925409198750996e-07, "loss": 0.0176, "step": 9298 }, { "epoch": 3.66, "learning_rate": 2.9251828370790466e-07, "loss": 0.0202, "step": 9299 }, { "epoch": 3.67, "learning_rate": 2.924956460328023e-07, "loss": 0.015, "step": 9300 }, { "epoch": 3.67, "learning_rate": 2.9247300685016154e-07, "loss": 0.0151, "step": 9301 }, { "epoch": 3.67, "learning_rate": 2.924503661603513e-07, "loss": 0.0169, "step": 9302 }, { "epoch": 3.67, "learning_rate": 2.924277239637405e-07, "loss": 0.017, "step": 9303 }, { "epoch": 3.67, "learning_rate": 2.924050802606984e-07, "loss": 0.0148, "step": 9304 }, { "epoch": 3.67, "learning_rate": 2.923824350515938e-07, "loss": 0.0187, "step": 9305 }, { "epoch": 3.67, "learning_rate": 2.92359788336796e-07, "loss": 0.0178, "step": 9306 }, { "epoch": 3.67, "learning_rate": 2.92337140116674e-07, "loss": 0.0216, "step": 9307 }, { "epoch": 3.67, "learning_rate": 2.92314490391597e-07, "loss": 0.0146, "step": 9308 }, { "epoch": 3.67, "learning_rate": 2.9229183916193404e-07, "loss": 0.0162, "step": 9309 }, { "epoch": 3.67, "learning_rate": 2.922691864280544e-07, "loss": 0.0179, "step": 9310 }, { "epoch": 3.67, "learning_rate": 2.9224653219032725e-07, "loss": 0.0184, "step": 9311 }, { "epoch": 3.67, "learning_rate": 2.922238764491218e-07, "loss": 0.0216, "step": 9312 }, { "epoch": 3.67, "learning_rate": 2.922012192048074e-07, "loss": 0.0212, "step": 9313 }, { "epoch": 3.67, "learning_rate": 2.9217856045775315e-07, "loss": 0.0179, "step": 9314 }, { "epoch": 3.67, "learning_rate": 2.9215590020832854e-07, "loss": 0.0183, "step": 9315 }, { "epoch": 3.67, "learning_rate": 2.9213323845690276e-07, "loss": 0.0188, "step": 9316 }, { "epoch": 3.67, "learning_rate": 2.921105752038453e-07, "loss": 0.0158, "step": 9317 }, { "epoch": 3.67, "learning_rate": 2.9208791044952537e-07, "loss": 0.0184, "step": 9318 }, { "epoch": 3.67, "learning_rate": 2.9206524419431253e-07, "loss": 0.0181, "step": 9319 }, { "epoch": 3.67, "learning_rate": 2.9204257643857607e-07, "loss": 0.0171, "step": 9320 }, { "epoch": 3.67, "learning_rate": 2.920199071826855e-07, "loss": 0.0187, "step": 9321 }, { "epoch": 3.67, "learning_rate": 2.919972364270104e-07, "loss": 0.0166, "step": 9322 }, { "epoch": 3.67, "learning_rate": 2.919745641719201e-07, "loss": 0.0175, "step": 9323 }, { "epoch": 3.67, "learning_rate": 2.919518904177841e-07, "loss": 0.0197, "step": 9324 }, { "epoch": 3.68, "learning_rate": 2.9192921516497217e-07, "loss": 0.0147, "step": 9325 }, { "epoch": 3.68, "learning_rate": 2.919065384138537e-07, "loss": 0.0176, "step": 9326 }, { "epoch": 3.68, "learning_rate": 2.9188386016479827e-07, "loss": 0.0175, "step": 9327 }, { "epoch": 3.68, "learning_rate": 2.918611804181757e-07, "loss": 0.017, "step": 9328 }, { "epoch": 3.68, "learning_rate": 2.9183849917435544e-07, "loss": 0.016, "step": 9329 }, { "epoch": 3.68, "learning_rate": 2.918158164337072e-07, "loss": 0.0157, "step": 9330 }, { "epoch": 3.68, "learning_rate": 2.917931321966007e-07, "loss": 0.0201, "step": 9331 }, { "epoch": 3.68, "learning_rate": 2.917704464634057e-07, "loss": 0.0192, "step": 9332 }, { "epoch": 3.68, "learning_rate": 2.917477592344919e-07, "loss": 0.0183, "step": 9333 }, { "epoch": 3.68, "learning_rate": 2.9172507051022897e-07, "loss": 0.0181, "step": 9334 }, { "epoch": 3.68, "learning_rate": 2.917023802909869e-07, "loss": 0.0217, "step": 9335 }, { "epoch": 3.68, "learning_rate": 2.9167968857713533e-07, "loss": 0.0153, "step": 9336 }, { "epoch": 3.68, "learning_rate": 2.916569953690442e-07, "loss": 0.0155, "step": 9337 }, { "epoch": 3.68, "learning_rate": 2.9163430066708334e-07, "loss": 0.0192, "step": 9338 }, { "epoch": 3.68, "learning_rate": 2.9161160447162265e-07, "loss": 0.0174, "step": 9339 }, { "epoch": 3.68, "learning_rate": 2.9158890678303207e-07, "loss": 0.0186, "step": 9340 }, { "epoch": 3.68, "learning_rate": 2.915662076016815e-07, "loss": 0.0187, "step": 9341 }, { "epoch": 3.68, "learning_rate": 2.9154350692794094e-07, "loss": 0.0169, "step": 9342 }, { "epoch": 3.68, "learning_rate": 2.915208047621802e-07, "loss": 0.018, "step": 9343 }, { "epoch": 3.68, "learning_rate": 2.914981011047696e-07, "loss": 0.0151, "step": 9344 }, { "epoch": 3.68, "learning_rate": 2.914753959560789e-07, "loss": 0.0199, "step": 9345 }, { "epoch": 3.68, "learning_rate": 2.914526893164784e-07, "loss": 0.0181, "step": 9346 }, { "epoch": 3.68, "learning_rate": 2.914299811863379e-07, "loss": 0.0162, "step": 9347 }, { "epoch": 3.68, "learning_rate": 2.914072715660278e-07, "loss": 0.0181, "step": 9348 }, { "epoch": 3.68, "learning_rate": 2.91384560455918e-07, "loss": 0.0162, "step": 9349 }, { "epoch": 3.68, "learning_rate": 2.9136184785637886e-07, "loss": 0.0155, "step": 9350 }, { "epoch": 3.69, "learning_rate": 2.9133913376778035e-07, "loss": 0.0156, "step": 9351 }, { "epoch": 3.69, "learning_rate": 2.9131641819049286e-07, "loss": 0.0188, "step": 9352 }, { "epoch": 3.69, "learning_rate": 2.912937011248865e-07, "loss": 0.0144, "step": 9353 }, { "epoch": 3.69, "learning_rate": 2.9127098257133156e-07, "loss": 0.0183, "step": 9354 }, { "epoch": 3.69, "learning_rate": 2.912482625301984e-07, "loss": 0.015, "step": 9355 }, { "epoch": 3.69, "learning_rate": 2.912255410018571e-07, "loss": 0.0203, "step": 9356 }, { "epoch": 3.69, "learning_rate": 2.912028179866782e-07, "loss": 0.0194, "step": 9357 }, { "epoch": 3.69, "learning_rate": 2.91180093485032e-07, "loss": 0.018, "step": 9358 }, { "epoch": 3.69, "learning_rate": 2.911573674972888e-07, "loss": 0.0151, "step": 9359 }, { "epoch": 3.69, "learning_rate": 2.9113464002381903e-07, "loss": 0.0192, "step": 9360 }, { "epoch": 3.69, "learning_rate": 2.911119110649932e-07, "loss": 0.0147, "step": 9361 }, { "epoch": 3.69, "learning_rate": 2.910891806211817e-07, "loss": 0.0177, "step": 9362 }, { "epoch": 3.69, "learning_rate": 2.9106644869275495e-07, "loss": 0.0164, "step": 9363 }, { "epoch": 3.69, "learning_rate": 2.9104371528008353e-07, "loss": 0.0165, "step": 9364 }, { "epoch": 3.69, "learning_rate": 2.910209803835379e-07, "loss": 0.0201, "step": 9365 }, { "epoch": 3.69, "learning_rate": 2.9099824400348865e-07, "loss": 0.017, "step": 9366 }, { "epoch": 3.69, "learning_rate": 2.909755061403063e-07, "loss": 0.0142, "step": 9367 }, { "epoch": 3.69, "learning_rate": 2.909527667943615e-07, "loss": 0.0217, "step": 9368 }, { "epoch": 3.69, "learning_rate": 2.909300259660248e-07, "loss": 0.0169, "step": 9369 }, { "epoch": 3.69, "learning_rate": 2.909072836556668e-07, "loss": 0.0187, "step": 9370 }, { "epoch": 3.69, "learning_rate": 2.9088453986365837e-07, "loss": 0.018, "step": 9371 }, { "epoch": 3.69, "learning_rate": 2.9086179459037e-07, "loss": 0.0197, "step": 9372 }, { "epoch": 3.69, "learning_rate": 2.908390478361725e-07, "loss": 0.0198, "step": 9373 }, { "epoch": 3.69, "learning_rate": 2.908162996014366e-07, "loss": 0.0153, "step": 9374 }, { "epoch": 3.69, "learning_rate": 2.9079354988653304e-07, "loss": 0.0158, "step": 9375 }, { "epoch": 3.7, "learning_rate": 2.9077079869183257e-07, "loss": 0.0168, "step": 9376 }, { "epoch": 3.7, "learning_rate": 2.9074804601770604e-07, "loss": 0.0165, "step": 9377 }, { "epoch": 3.7, "learning_rate": 2.907252918645243e-07, "loss": 0.0184, "step": 9378 }, { "epoch": 3.7, "learning_rate": 2.907025362326582e-07, "loss": 0.0204, "step": 9379 }, { "epoch": 3.7, "learning_rate": 2.9067977912247853e-07, "loss": 0.0137, "step": 9380 }, { "epoch": 3.7, "learning_rate": 2.9065702053435637e-07, "loss": 0.0199, "step": 9381 }, { "epoch": 3.7, "learning_rate": 2.9063426046866253e-07, "loss": 0.0191, "step": 9382 }, { "epoch": 3.7, "learning_rate": 2.9061149892576796e-07, "loss": 0.0169, "step": 9383 }, { "epoch": 3.7, "learning_rate": 2.905887359060437e-07, "loss": 0.0186, "step": 9384 }, { "epoch": 3.7, "learning_rate": 2.905659714098607e-07, "loss": 0.0171, "step": 9385 }, { "epoch": 3.7, "learning_rate": 2.9054320543759007e-07, "loss": 0.0155, "step": 9386 }, { "epoch": 3.7, "learning_rate": 2.9052043798960276e-07, "loss": 0.0174, "step": 9387 }, { "epoch": 3.7, "learning_rate": 2.904976690662699e-07, "loss": 0.0172, "step": 9388 }, { "epoch": 3.7, "learning_rate": 2.9047489866796257e-07, "loss": 0.0173, "step": 9389 }, { "epoch": 3.7, "learning_rate": 2.904521267950519e-07, "loss": 0.0165, "step": 9390 }, { "epoch": 3.7, "learning_rate": 2.9042935344790903e-07, "loss": 0.0187, "step": 9391 }, { "epoch": 3.7, "learning_rate": 2.9040657862690506e-07, "loss": 0.015, "step": 9392 }, { "epoch": 3.7, "learning_rate": 2.903838023324114e-07, "loss": 0.0157, "step": 9393 }, { "epoch": 3.7, "learning_rate": 2.9036102456479905e-07, "loss": 0.016, "step": 9394 }, { "epoch": 3.7, "learning_rate": 2.903382453244394e-07, "loss": 0.0173, "step": 9395 }, { "epoch": 3.7, "learning_rate": 2.9031546461170356e-07, "loss": 0.0169, "step": 9396 }, { "epoch": 3.7, "learning_rate": 2.9029268242696297e-07, "loss": 0.0162, "step": 9397 }, { "epoch": 3.7, "learning_rate": 2.902698987705888e-07, "loss": 0.0184, "step": 9398 }, { "epoch": 3.7, "learning_rate": 2.9024711364295254e-07, "loss": 0.0149, "step": 9399 }, { "epoch": 3.7, "learning_rate": 2.902243270444255e-07, "loss": 0.0162, "step": 9400 }, { "epoch": 3.71, "learning_rate": 2.9020153897537904e-07, "loss": 0.0166, "step": 9401 }, { "epoch": 3.71, "learning_rate": 2.9017874943618463e-07, "loss": 0.0211, "step": 9402 }, { "epoch": 3.71, "learning_rate": 2.9015595842721356e-07, "loss": 0.0163, "step": 9403 }, { "epoch": 3.71, "learning_rate": 2.901331659488375e-07, "loss": 0.0163, "step": 9404 }, { "epoch": 3.71, "learning_rate": 2.901103720014277e-07, "loss": 0.0219, "step": 9405 }, { "epoch": 3.71, "learning_rate": 2.9008757658535585e-07, "loss": 0.0192, "step": 9406 }, { "epoch": 3.71, "learning_rate": 2.900647797009934e-07, "loss": 0.0167, "step": 9407 }, { "epoch": 3.71, "learning_rate": 2.90041981348712e-07, "loss": 0.0165, "step": 9408 }, { "epoch": 3.71, "learning_rate": 2.90019181528883e-07, "loss": 0.0176, "step": 9409 }, { "epoch": 3.71, "learning_rate": 2.8999638024187833e-07, "loss": 0.0183, "step": 9410 }, { "epoch": 3.71, "learning_rate": 2.8997357748806933e-07, "loss": 0.0203, "step": 9411 }, { "epoch": 3.71, "learning_rate": 2.8995077326782777e-07, "loss": 0.0193, "step": 9412 }, { "epoch": 3.71, "learning_rate": 2.899279675815254e-07, "loss": 0.0188, "step": 9413 }, { "epoch": 3.71, "learning_rate": 2.8990516042953374e-07, "loss": 0.0154, "step": 9414 }, { "epoch": 3.71, "learning_rate": 2.8988235181222466e-07, "loss": 0.0177, "step": 9415 }, { "epoch": 3.71, "learning_rate": 2.898595417299698e-07, "loss": 0.0179, "step": 9416 }, { "epoch": 3.71, "learning_rate": 2.89836730183141e-07, "loss": 0.0172, "step": 9417 }, { "epoch": 3.71, "learning_rate": 2.898139171721101e-07, "loss": 0.0199, "step": 9418 }, { "epoch": 3.71, "learning_rate": 2.8979110269724875e-07, "loss": 0.0214, "step": 9419 }, { "epoch": 3.71, "learning_rate": 2.8976828675892893e-07, "loss": 0.0153, "step": 9420 }, { "epoch": 3.71, "learning_rate": 2.897454693575225e-07, "loss": 0.0182, "step": 9421 }, { "epoch": 3.71, "learning_rate": 2.897226504934013e-07, "loss": 0.0155, "step": 9422 }, { "epoch": 3.71, "learning_rate": 2.8969983016693725e-07, "loss": 0.0173, "step": 9423 }, { "epoch": 3.71, "learning_rate": 2.8967700837850227e-07, "loss": 0.018, "step": 9424 }, { "epoch": 3.71, "learning_rate": 2.8965418512846837e-07, "loss": 0.0197, "step": 9425 }, { "epoch": 3.71, "learning_rate": 2.896313604172075e-07, "loss": 0.0185, "step": 9426 }, { "epoch": 3.72, "learning_rate": 2.896085342450917e-07, "loss": 0.0177, "step": 9427 }, { "epoch": 3.72, "learning_rate": 2.89585706612493e-07, "loss": 0.0215, "step": 9428 }, { "epoch": 3.72, "learning_rate": 2.8956287751978337e-07, "loss": 0.0195, "step": 9429 }, { "epoch": 3.72, "learning_rate": 2.89540046967335e-07, "loss": 0.0139, "step": 9430 }, { "epoch": 3.72, "learning_rate": 2.8951721495552e-07, "loss": 0.0186, "step": 9431 }, { "epoch": 3.72, "learning_rate": 2.8949438148471036e-07, "loss": 0.0178, "step": 9432 }, { "epoch": 3.72, "learning_rate": 2.8947154655527833e-07, "loss": 0.0137, "step": 9433 }, { "epoch": 3.72, "learning_rate": 2.8944871016759613e-07, "loss": 0.0146, "step": 9434 }, { "epoch": 3.72, "learning_rate": 2.8942587232203586e-07, "loss": 0.0209, "step": 9435 }, { "epoch": 3.72, "learning_rate": 2.8940303301896975e-07, "loss": 0.0174, "step": 9436 }, { "epoch": 3.72, "learning_rate": 2.893801922587702e-07, "loss": 0.0183, "step": 9437 }, { "epoch": 3.72, "learning_rate": 2.8935735004180924e-07, "loss": 0.0166, "step": 9438 }, { "epoch": 3.72, "learning_rate": 2.8933450636845933e-07, "loss": 0.0186, "step": 9439 }, { "epoch": 3.72, "learning_rate": 2.893116612390927e-07, "loss": 0.0181, "step": 9440 }, { "epoch": 3.72, "learning_rate": 2.8928881465408176e-07, "loss": 0.0174, "step": 9441 }, { "epoch": 3.72, "learning_rate": 2.8926596661379885e-07, "loss": 0.0178, "step": 9442 }, { "epoch": 3.72, "learning_rate": 2.892431171186163e-07, "loss": 0.0199, "step": 9443 }, { "epoch": 3.72, "learning_rate": 2.892202661689067e-07, "loss": 0.0179, "step": 9444 }, { "epoch": 3.72, "learning_rate": 2.8919741376504223e-07, "loss": 0.0151, "step": 9445 }, { "epoch": 3.72, "learning_rate": 2.8917455990739554e-07, "loss": 0.017, "step": 9446 }, { "epoch": 3.72, "learning_rate": 2.89151704596339e-07, "loss": 0.0201, "step": 9447 }, { "epoch": 3.72, "learning_rate": 2.8912884783224523e-07, "loss": 0.0187, "step": 9448 }, { "epoch": 3.72, "learning_rate": 2.891059896154867e-07, "loss": 0.0186, "step": 9449 }, { "epoch": 3.72, "learning_rate": 2.8908312994643593e-07, "loss": 0.0204, "step": 9450 }, { "epoch": 3.72, "learning_rate": 2.8906026882546557e-07, "loss": 0.0204, "step": 9451 }, { "epoch": 3.73, "learning_rate": 2.8903740625294814e-07, "loss": 0.0183, "step": 9452 }, { "epoch": 3.73, "learning_rate": 2.8901454222925633e-07, "loss": 0.0173, "step": 9453 }, { "epoch": 3.73, "learning_rate": 2.889916767547627e-07, "loss": 0.0171, "step": 9454 }, { "epoch": 3.73, "learning_rate": 2.8896880982984013e-07, "loss": 0.0186, "step": 9455 }, { "epoch": 3.73, "learning_rate": 2.889459414548611e-07, "loss": 0.017, "step": 9456 }, { "epoch": 3.73, "learning_rate": 2.8892307163019843e-07, "loss": 0.021, "step": 9457 }, { "epoch": 3.73, "learning_rate": 2.889002003562249e-07, "loss": 0.014, "step": 9458 }, { "epoch": 3.73, "learning_rate": 2.8887732763331306e-07, "loss": 0.0194, "step": 9459 }, { "epoch": 3.73, "learning_rate": 2.88854453461836e-07, "loss": 0.0179, "step": 9460 }, { "epoch": 3.73, "learning_rate": 2.888315778421663e-07, "loss": 0.0171, "step": 9461 }, { "epoch": 3.73, "learning_rate": 2.8880870077467696e-07, "loss": 0.0165, "step": 9462 }, { "epoch": 3.73, "learning_rate": 2.8878582225974066e-07, "loss": 0.0181, "step": 9463 }, { "epoch": 3.73, "learning_rate": 2.887629422977305e-07, "loss": 0.0225, "step": 9464 }, { "epoch": 3.73, "learning_rate": 2.8874006088901924e-07, "loss": 0.0149, "step": 9465 }, { "epoch": 3.73, "learning_rate": 2.887171780339799e-07, "loss": 0.0156, "step": 9466 }, { "epoch": 3.73, "learning_rate": 2.886942937329853e-07, "loss": 0.0141, "step": 9467 }, { "epoch": 3.73, "learning_rate": 2.8867140798640864e-07, "loss": 0.0202, "step": 9468 }, { "epoch": 3.73, "learning_rate": 2.886485207946227e-07, "loss": 0.0174, "step": 9469 }, { "epoch": 3.73, "learning_rate": 2.8862563215800065e-07, "loss": 0.0204, "step": 9470 }, { "epoch": 3.73, "learning_rate": 2.886027420769155e-07, "loss": 0.0165, "step": 9471 }, { "epoch": 3.73, "learning_rate": 2.885798505517402e-07, "loss": 0.0115, "step": 9472 }, { "epoch": 3.73, "learning_rate": 2.885569575828481e-07, "loss": 0.0195, "step": 9473 }, { "epoch": 3.73, "learning_rate": 2.885340631706121e-07, "loss": 0.0182, "step": 9474 }, { "epoch": 3.73, "learning_rate": 2.8851116731540546e-07, "loss": 0.0206, "step": 9475 }, { "epoch": 3.73, "learning_rate": 2.8848827001760133e-07, "loss": 0.018, "step": 9476 }, { "epoch": 3.73, "learning_rate": 2.884653712775729e-07, "loss": 0.0171, "step": 9477 }, { "epoch": 3.74, "learning_rate": 2.8844247109569325e-07, "loss": 0.018, "step": 9478 }, { "epoch": 3.74, "learning_rate": 2.8841956947233586e-07, "loss": 0.0152, "step": 9479 }, { "epoch": 3.74, "learning_rate": 2.8839666640787383e-07, "loss": 0.0148, "step": 9480 }, { "epoch": 3.74, "learning_rate": 2.8837376190268054e-07, "loss": 0.0202, "step": 9481 }, { "epoch": 3.74, "learning_rate": 2.883508559571291e-07, "loss": 0.0144, "step": 9482 }, { "epoch": 3.74, "learning_rate": 2.8832794857159314e-07, "loss": 0.0164, "step": 9483 }, { "epoch": 3.74, "learning_rate": 2.8830503974644585e-07, "loss": 0.0179, "step": 9484 }, { "epoch": 3.74, "learning_rate": 2.882821294820605e-07, "loss": 0.0201, "step": 9485 }, { "epoch": 3.74, "learning_rate": 2.882592177788107e-07, "loss": 0.0181, "step": 9486 }, { "epoch": 3.74, "learning_rate": 2.8823630463706976e-07, "loss": 0.0153, "step": 9487 }, { "epoch": 3.74, "learning_rate": 2.8821339005721117e-07, "loss": 0.0206, "step": 9488 }, { "epoch": 3.74, "learning_rate": 2.881904740396084e-07, "loss": 0.0153, "step": 9489 }, { "epoch": 3.74, "learning_rate": 2.8816755658463495e-07, "loss": 0.017, "step": 9490 }, { "epoch": 3.74, "learning_rate": 2.881446376926643e-07, "loss": 0.0214, "step": 9491 }, { "epoch": 3.74, "learning_rate": 2.8812171736407e-07, "loss": 0.022, "step": 9492 }, { "epoch": 3.74, "learning_rate": 2.880987955992257e-07, "loss": 0.018, "step": 9493 }, { "epoch": 3.74, "learning_rate": 2.880758723985049e-07, "loss": 0.0153, "step": 9494 }, { "epoch": 3.74, "learning_rate": 2.880529477622812e-07, "loss": 0.0171, "step": 9495 }, { "epoch": 3.74, "learning_rate": 2.880300216909283e-07, "loss": 0.0211, "step": 9496 }, { "epoch": 3.74, "learning_rate": 2.880070941848198e-07, "loss": 0.0159, "step": 9497 }, { "epoch": 3.74, "learning_rate": 2.879841652443295e-07, "loss": 0.0181, "step": 9498 }, { "epoch": 3.74, "learning_rate": 2.87961234869831e-07, "loss": 0.0192, "step": 9499 }, { "epoch": 3.74, "learning_rate": 2.87938303061698e-07, "loss": 0.0163, "step": 9500 }, { "epoch": 3.74, "learning_rate": 2.879153698203044e-07, "loss": 0.0187, "step": 9501 }, { "epoch": 3.74, "learning_rate": 2.878924351460239e-07, "loss": 0.0161, "step": 9502 }, { "epoch": 3.75, "learning_rate": 2.8786949903923015e-07, "loss": 0.0173, "step": 9503 }, { "epoch": 3.75, "learning_rate": 2.8784656150029725e-07, "loss": 0.0155, "step": 9504 }, { "epoch": 3.75, "learning_rate": 2.8782362252959883e-07, "loss": 0.0221, "step": 9505 }, { "epoch": 3.75, "learning_rate": 2.8780068212750885e-07, "loss": 0.0159, "step": 9506 }, { "epoch": 3.75, "learning_rate": 2.8777774029440124e-07, "loss": 0.0171, "step": 9507 }, { "epoch": 3.75, "learning_rate": 2.8775479703064984e-07, "loss": 0.0181, "step": 9508 }, { "epoch": 3.75, "learning_rate": 2.877318523366286e-07, "loss": 0.0166, "step": 9509 }, { "epoch": 3.75, "learning_rate": 2.877089062127115e-07, "loss": 0.0176, "step": 9510 }, { "epoch": 3.75, "learning_rate": 2.876859586592726e-07, "loss": 0.0177, "step": 9511 }, { "epoch": 3.75, "learning_rate": 2.8766300967668576e-07, "loss": 0.0193, "step": 9512 }, { "epoch": 3.75, "learning_rate": 2.8764005926532513e-07, "loss": 0.0172, "step": 9513 }, { "epoch": 3.75, "learning_rate": 2.876171074255647e-07, "loss": 0.0136, "step": 9514 }, { "epoch": 3.75, "learning_rate": 2.8759415415777866e-07, "loss": 0.0128, "step": 9515 }, { "epoch": 3.75, "learning_rate": 2.875711994623409e-07, "loss": 0.0189, "step": 9516 }, { "epoch": 3.75, "learning_rate": 2.8754824333962584e-07, "loss": 0.0192, "step": 9517 }, { "epoch": 3.75, "learning_rate": 2.875252857900074e-07, "loss": 0.0172, "step": 9518 }, { "epoch": 3.75, "learning_rate": 2.875023268138598e-07, "loss": 0.0155, "step": 9519 }, { "epoch": 3.75, "learning_rate": 2.874793664115573e-07, "loss": 0.0198, "step": 9520 }, { "epoch": 3.75, "learning_rate": 2.874564045834741e-07, "loss": 0.0166, "step": 9521 }, { "epoch": 3.75, "learning_rate": 2.874334413299843e-07, "loss": 0.0173, "step": 9522 }, { "epoch": 3.75, "learning_rate": 2.874104766514624e-07, "loss": 0.0198, "step": 9523 }, { "epoch": 3.75, "learning_rate": 2.873875105482826e-07, "loss": 0.0224, "step": 9524 }, { "epoch": 3.75, "learning_rate": 2.8736454302081913e-07, "loss": 0.0182, "step": 9525 }, { "epoch": 3.75, "learning_rate": 2.8734157406944645e-07, "loss": 0.0162, "step": 9526 }, { "epoch": 3.75, "learning_rate": 2.873186036945388e-07, "loss": 0.017, "step": 9527 }, { "epoch": 3.76, "learning_rate": 2.872956318964706e-07, "loss": 0.014, "step": 9528 }, { "epoch": 3.76, "learning_rate": 2.872726586756163e-07, "loss": 0.0209, "step": 9529 }, { "epoch": 3.76, "learning_rate": 2.8724968403235036e-07, "loss": 0.019, "step": 9530 }, { "epoch": 3.76, "learning_rate": 2.872267079670471e-07, "loss": 0.0195, "step": 9531 }, { "epoch": 3.76, "learning_rate": 2.872037304800811e-07, "loss": 0.017, "step": 9532 }, { "epoch": 3.76, "learning_rate": 2.8718075157182684e-07, "loss": 0.016, "step": 9533 }, { "epoch": 3.76, "learning_rate": 2.8715777124265874e-07, "loss": 0.0185, "step": 9534 }, { "epoch": 3.76, "learning_rate": 2.8713478949295154e-07, "loss": 0.0168, "step": 9535 }, { "epoch": 3.76, "learning_rate": 2.8711180632307967e-07, "loss": 0.0162, "step": 9536 }, { "epoch": 3.76, "learning_rate": 2.8708882173341774e-07, "loss": 0.0171, "step": 9537 }, { "epoch": 3.76, "learning_rate": 2.8706583572434035e-07, "loss": 0.0175, "step": 9538 }, { "epoch": 3.76, "learning_rate": 2.8704284829622226e-07, "loss": 0.019, "step": 9539 }, { "epoch": 3.76, "learning_rate": 2.870198594494379e-07, "loss": 0.0176, "step": 9540 }, { "epoch": 3.76, "learning_rate": 2.869968691843622e-07, "loss": 0.018, "step": 9541 }, { "epoch": 3.76, "learning_rate": 2.869738775013697e-07, "loss": 0.0184, "step": 9542 }, { "epoch": 3.76, "learning_rate": 2.8695088440083514e-07, "loss": 0.0168, "step": 9543 }, { "epoch": 3.76, "learning_rate": 2.869278898831334e-07, "loss": 0.0156, "step": 9544 }, { "epoch": 3.76, "learning_rate": 2.869048939486391e-07, "loss": 0.0175, "step": 9545 }, { "epoch": 3.76, "learning_rate": 2.8688189659772715e-07, "loss": 0.0186, "step": 9546 }, { "epoch": 3.76, "learning_rate": 2.8685889783077233e-07, "loss": 0.0193, "step": 9547 }, { "epoch": 3.76, "learning_rate": 2.8683589764814954e-07, "loss": 0.0163, "step": 9548 }, { "epoch": 3.76, "learning_rate": 2.868128960502335e-07, "loss": 0.0205, "step": 9549 }, { "epoch": 3.76, "learning_rate": 2.8678989303739923e-07, "loss": 0.0171, "step": 9550 }, { "epoch": 3.76, "learning_rate": 2.8676688861002165e-07, "loss": 0.0184, "step": 9551 }, { "epoch": 3.76, "learning_rate": 2.867438827684755e-07, "loss": 0.016, "step": 9552 }, { "epoch": 3.76, "learning_rate": 2.8672087551313607e-07, "loss": 0.0204, "step": 9553 }, { "epoch": 3.77, "learning_rate": 2.8669786684437806e-07, "loss": 0.0177, "step": 9554 }, { "epoch": 3.77, "learning_rate": 2.8667485676257667e-07, "loss": 0.0223, "step": 9555 }, { "epoch": 3.77, "learning_rate": 2.866518452681068e-07, "loss": 0.0144, "step": 9556 }, { "epoch": 3.77, "learning_rate": 2.8662883236134353e-07, "loss": 0.0179, "step": 9557 }, { "epoch": 3.77, "learning_rate": 2.86605818042662e-07, "loss": 0.0162, "step": 9558 }, { "epoch": 3.77, "learning_rate": 2.865828023124372e-07, "loss": 0.0179, "step": 9559 }, { "epoch": 3.77, "learning_rate": 2.8655978517104433e-07, "loss": 0.016, "step": 9560 }, { "epoch": 3.77, "learning_rate": 2.8653676661885857e-07, "loss": 0.0138, "step": 9561 }, { "epoch": 3.77, "learning_rate": 2.8651374665625505e-07, "loss": 0.0154, "step": 9562 }, { "epoch": 3.77, "learning_rate": 2.8649072528360887e-07, "loss": 0.0194, "step": 9563 }, { "epoch": 3.77, "learning_rate": 2.8646770250129535e-07, "loss": 0.0162, "step": 9564 }, { "epoch": 3.77, "learning_rate": 2.8644467830968964e-07, "loss": 0.0191, "step": 9565 }, { "epoch": 3.77, "learning_rate": 2.8642165270916717e-07, "loss": 0.0226, "step": 9566 }, { "epoch": 3.77, "learning_rate": 2.863986257001031e-07, "loss": 0.017, "step": 9567 }, { "epoch": 3.77, "learning_rate": 2.863755972828726e-07, "loss": 0.0161, "step": 9568 }, { "epoch": 3.77, "learning_rate": 2.8635256745785123e-07, "loss": 0.018, "step": 9569 }, { "epoch": 3.77, "learning_rate": 2.8632953622541427e-07, "loss": 0.02, "step": 9570 }, { "epoch": 3.77, "learning_rate": 2.863065035859371e-07, "loss": 0.0173, "step": 9571 }, { "epoch": 3.77, "learning_rate": 2.86283469539795e-07, "loss": 0.0173, "step": 9572 }, { "epoch": 3.77, "learning_rate": 2.8626043408736357e-07, "loss": 0.0178, "step": 9573 }, { "epoch": 3.77, "learning_rate": 2.862373972290181e-07, "loss": 0.0157, "step": 9574 }, { "epoch": 3.77, "learning_rate": 2.8621435896513417e-07, "loss": 0.0174, "step": 9575 }, { "epoch": 3.77, "learning_rate": 2.8619131929608716e-07, "loss": 0.0142, "step": 9576 }, { "epoch": 3.77, "learning_rate": 2.861682782222527e-07, "loss": 0.017, "step": 9577 }, { "epoch": 3.77, "learning_rate": 2.8614523574400624e-07, "loss": 0.0136, "step": 9578 }, { "epoch": 3.78, "learning_rate": 2.8612219186172333e-07, "loss": 0.0168, "step": 9579 }, { "epoch": 3.78, "learning_rate": 2.860991465757796e-07, "loss": 0.0179, "step": 9580 }, { "epoch": 3.78, "learning_rate": 2.8607609988655065e-07, "loss": 0.0178, "step": 9581 }, { "epoch": 3.78, "learning_rate": 2.8605305179441214e-07, "loss": 0.0205, "step": 9582 }, { "epoch": 3.78, "learning_rate": 2.860300022997396e-07, "loss": 0.0217, "step": 9583 }, { "epoch": 3.78, "learning_rate": 2.860069514029088e-07, "loss": 0.0172, "step": 9584 }, { "epoch": 3.78, "learning_rate": 2.859838991042954e-07, "loss": 0.015, "step": 9585 }, { "epoch": 3.78, "learning_rate": 2.8596084540427516e-07, "loss": 0.0159, "step": 9586 }, { "epoch": 3.78, "learning_rate": 2.859377903032237e-07, "loss": 0.0185, "step": 9587 }, { "epoch": 3.78, "learning_rate": 2.8591473380151696e-07, "loss": 0.0177, "step": 9588 }, { "epoch": 3.78, "learning_rate": 2.8589167589953064e-07, "loss": 0.0176, "step": 9589 }, { "epoch": 3.78, "learning_rate": 2.858686165976405e-07, "loss": 0.0178, "step": 9590 }, { "epoch": 3.78, "learning_rate": 2.8584555589622243e-07, "loss": 0.0185, "step": 9591 }, { "epoch": 3.78, "learning_rate": 2.8582249379565224e-07, "loss": 0.0182, "step": 9592 }, { "epoch": 3.78, "learning_rate": 2.8579943029630586e-07, "loss": 0.0195, "step": 9593 }, { "epoch": 3.78, "learning_rate": 2.857763653985592e-07, "loss": 0.0135, "step": 9594 }, { "epoch": 3.78, "learning_rate": 2.857532991027881e-07, "loss": 0.0155, "step": 9595 }, { "epoch": 3.78, "learning_rate": 2.857302314093685e-07, "loss": 0.0175, "step": 9596 }, { "epoch": 3.78, "learning_rate": 2.857071623186765e-07, "loss": 0.0151, "step": 9597 }, { "epoch": 3.78, "learning_rate": 2.85684091831088e-07, "loss": 0.0199, "step": 9598 }, { "epoch": 3.78, "learning_rate": 2.8566101994697905e-07, "loss": 0.0177, "step": 9599 }, { "epoch": 3.78, "learning_rate": 2.8563794666672565e-07, "loss": 0.0205, "step": 9600 }, { "epoch": 3.78, "learning_rate": 2.8561487199070383e-07, "loss": 0.0162, "step": 9601 }, { "epoch": 3.78, "learning_rate": 2.8559179591928984e-07, "loss": 0.0171, "step": 9602 }, { "epoch": 3.78, "learning_rate": 2.855687184528595e-07, "loss": 0.0133, "step": 9603 }, { "epoch": 3.79, "learning_rate": 2.8554563959178917e-07, "loss": 0.0195, "step": 9604 }, { "epoch": 3.79, "learning_rate": 2.8552255933645493e-07, "loss": 0.0175, "step": 9605 }, { "epoch": 3.79, "learning_rate": 2.8549947768723295e-07, "loss": 0.0175, "step": 9606 }, { "epoch": 3.79, "learning_rate": 2.8547639464449946e-07, "loss": 0.0139, "step": 9607 }, { "epoch": 3.79, "learning_rate": 2.854533102086306e-07, "loss": 0.0183, "step": 9608 }, { "epoch": 3.79, "learning_rate": 2.854302243800027e-07, "loss": 0.0166, "step": 9609 }, { "epoch": 3.79, "learning_rate": 2.8540713715899195e-07, "loss": 0.0164, "step": 9610 }, { "epoch": 3.79, "learning_rate": 2.8538404854597467e-07, "loss": 0.0202, "step": 9611 }, { "epoch": 3.79, "learning_rate": 2.853609585413272e-07, "loss": 0.0196, "step": 9612 }, { "epoch": 3.79, "learning_rate": 2.8533786714542585e-07, "loss": 0.0135, "step": 9613 }, { "epoch": 3.79, "learning_rate": 2.8531477435864684e-07, "loss": 0.0178, "step": 9614 }, { "epoch": 3.79, "learning_rate": 2.8529168018136684e-07, "loss": 0.0184, "step": 9615 }, { "epoch": 3.79, "learning_rate": 2.85268584613962e-07, "loss": 0.0177, "step": 9616 }, { "epoch": 3.79, "learning_rate": 2.852454876568088e-07, "loss": 0.0153, "step": 9617 }, { "epoch": 3.79, "learning_rate": 2.852223893102837e-07, "loss": 0.0169, "step": 9618 }, { "epoch": 3.79, "learning_rate": 2.8519928957476326e-07, "loss": 0.0163, "step": 9619 }, { "epoch": 3.79, "learning_rate": 2.8517618845062376e-07, "loss": 0.0152, "step": 9620 }, { "epoch": 3.79, "learning_rate": 2.85153085938242e-07, "loss": 0.0156, "step": 9621 }, { "epoch": 3.79, "learning_rate": 2.8512998203799425e-07, "loss": 0.0175, "step": 9622 }, { "epoch": 3.79, "learning_rate": 2.851068767502572e-07, "loss": 0.0182, "step": 9623 }, { "epoch": 3.79, "learning_rate": 2.850837700754074e-07, "loss": 0.016, "step": 9624 }, { "epoch": 3.79, "learning_rate": 2.8506066201382147e-07, "loss": 0.0171, "step": 9625 }, { "epoch": 3.79, "learning_rate": 2.8503755256587606e-07, "loss": 0.0198, "step": 9626 }, { "epoch": 3.79, "learning_rate": 2.850144417319477e-07, "loss": 0.013, "step": 9627 }, { "epoch": 3.79, "learning_rate": 2.849913295124132e-07, "loss": 0.0205, "step": 9628 }, { "epoch": 3.79, "learning_rate": 2.849682159076492e-07, "loss": 0.0196, "step": 9629 }, { "epoch": 3.8, "learning_rate": 2.849451009180324e-07, "loss": 0.0181, "step": 9630 }, { "epoch": 3.8, "learning_rate": 2.849219845439396e-07, "loss": 0.0226, "step": 9631 }, { "epoch": 3.8, "learning_rate": 2.848988667857475e-07, "loss": 0.0157, "step": 9632 }, { "epoch": 3.8, "learning_rate": 2.8487574764383284e-07, "loss": 0.0212, "step": 9633 }, { "epoch": 3.8, "learning_rate": 2.848526271185726e-07, "loss": 0.0188, "step": 9634 }, { "epoch": 3.8, "learning_rate": 2.848295052103435e-07, "loss": 0.0182, "step": 9635 }, { "epoch": 3.8, "learning_rate": 2.848063819195223e-07, "loss": 0.0145, "step": 9636 }, { "epoch": 3.8, "learning_rate": 2.84783257246486e-07, "loss": 0.013, "step": 9637 }, { "epoch": 3.8, "learning_rate": 2.847601311916114e-07, "loss": 0.0155, "step": 9638 }, { "epoch": 3.8, "learning_rate": 2.847370037552756e-07, "loss": 0.0147, "step": 9639 }, { "epoch": 3.8, "learning_rate": 2.8471387493785535e-07, "loss": 0.0167, "step": 9640 }, { "epoch": 3.8, "learning_rate": 2.8469074473972773e-07, "loss": 0.0177, "step": 9641 }, { "epoch": 3.8, "learning_rate": 2.8466761316126966e-07, "loss": 0.0207, "step": 9642 }, { "epoch": 3.8, "learning_rate": 2.846444802028582e-07, "loss": 0.0149, "step": 9643 }, { "epoch": 3.8, "learning_rate": 2.846213458648703e-07, "loss": 0.0158, "step": 9644 }, { "epoch": 3.8, "learning_rate": 2.845982101476831e-07, "loss": 0.0192, "step": 9645 }, { "epoch": 3.8, "learning_rate": 2.8457507305167366e-07, "loss": 0.0185, "step": 9646 }, { "epoch": 3.8, "learning_rate": 2.845519345772191e-07, "loss": 0.0162, "step": 9647 }, { "epoch": 3.8, "learning_rate": 2.8452879472469644e-07, "loss": 0.0156, "step": 9648 }, { "epoch": 3.8, "learning_rate": 2.845056534944829e-07, "loss": 0.0189, "step": 9649 }, { "epoch": 3.8, "learning_rate": 2.844825108869557e-07, "loss": 0.0164, "step": 9650 }, { "epoch": 3.8, "learning_rate": 2.844593669024919e-07, "loss": 0.0149, "step": 9651 }, { "epoch": 3.8, "learning_rate": 2.844362215414689e-07, "loss": 0.0153, "step": 9652 }, { "epoch": 3.8, "learning_rate": 2.8441307480426374e-07, "loss": 0.0168, "step": 9653 }, { "epoch": 3.8, "learning_rate": 2.8438992669125376e-07, "loss": 0.0151, "step": 9654 }, { "epoch": 3.81, "learning_rate": 2.843667772028162e-07, "loss": 0.0189, "step": 9655 }, { "epoch": 3.81, "learning_rate": 2.8434362633932843e-07, "loss": 0.015, "step": 9656 }, { "epoch": 3.81, "learning_rate": 2.843204741011678e-07, "loss": 0.0174, "step": 9657 }, { "epoch": 3.81, "learning_rate": 2.842973204887115e-07, "loss": 0.0157, "step": 9658 }, { "epoch": 3.81, "learning_rate": 2.8427416550233703e-07, "loss": 0.0193, "step": 9659 }, { "epoch": 3.81, "learning_rate": 2.8425100914242174e-07, "loss": 0.017, "step": 9660 }, { "epoch": 3.81, "learning_rate": 2.84227851409343e-07, "loss": 0.0211, "step": 9661 }, { "epoch": 3.81, "learning_rate": 2.842046923034784e-07, "loss": 0.0179, "step": 9662 }, { "epoch": 3.81, "learning_rate": 2.8418153182520516e-07, "loss": 0.0178, "step": 9663 }, { "epoch": 3.81, "learning_rate": 2.84158369974901e-07, "loss": 0.0163, "step": 9664 }, { "epoch": 3.81, "learning_rate": 2.841352067529433e-07, "loss": 0.0174, "step": 9665 }, { "epoch": 3.81, "learning_rate": 2.8411204215970956e-07, "loss": 0.02, "step": 9666 }, { "epoch": 3.81, "learning_rate": 2.8408887619557736e-07, "loss": 0.0188, "step": 9667 }, { "epoch": 3.81, "learning_rate": 2.840657088609243e-07, "loss": 0.017, "step": 9668 }, { "epoch": 3.81, "learning_rate": 2.840425401561279e-07, "loss": 0.0177, "step": 9669 }, { "epoch": 3.81, "learning_rate": 2.8401937008156595e-07, "loss": 0.0171, "step": 9670 }, { "epoch": 3.81, "learning_rate": 2.839961986376159e-07, "loss": 0.0194, "step": 9671 }, { "epoch": 3.81, "learning_rate": 2.8397302582465543e-07, "loss": 0.0165, "step": 9672 }, { "epoch": 3.81, "learning_rate": 2.839498516430623e-07, "loss": 0.0178, "step": 9673 }, { "epoch": 3.81, "learning_rate": 2.8392667609321416e-07, "loss": 0.0152, "step": 9674 }, { "epoch": 3.81, "learning_rate": 2.839034991754888e-07, "loss": 0.0165, "step": 9675 }, { "epoch": 3.81, "learning_rate": 2.838803208902639e-07, "loss": 0.0185, "step": 9676 }, { "epoch": 3.81, "learning_rate": 2.8385714123791727e-07, "loss": 0.018, "step": 9677 }, { "epoch": 3.81, "learning_rate": 2.8383396021882666e-07, "loss": 0.0197, "step": 9678 }, { "epoch": 3.81, "learning_rate": 2.838107778333699e-07, "loss": 0.0185, "step": 9679 }, { "epoch": 3.81, "learning_rate": 2.8378759408192485e-07, "loss": 0.0197, "step": 9680 }, { "epoch": 3.82, "learning_rate": 2.8376440896486936e-07, "loss": 0.0223, "step": 9681 }, { "epoch": 3.82, "learning_rate": 2.8374122248258136e-07, "loss": 0.0174, "step": 9682 }, { "epoch": 3.82, "learning_rate": 2.837180346354386e-07, "loss": 0.0172, "step": 9683 }, { "epoch": 3.82, "learning_rate": 2.836948454238192e-07, "loss": 0.0167, "step": 9684 }, { "epoch": 3.82, "learning_rate": 2.836716548481009e-07, "loss": 0.0188, "step": 9685 }, { "epoch": 3.82, "learning_rate": 2.8364846290866196e-07, "loss": 0.0159, "step": 9686 }, { "epoch": 3.82, "learning_rate": 2.8362526960588007e-07, "loss": 0.0199, "step": 9687 }, { "epoch": 3.82, "learning_rate": 2.8360207494013344e-07, "loss": 0.0134, "step": 9688 }, { "epoch": 3.82, "learning_rate": 2.8357887891180006e-07, "loss": 0.0197, "step": 9689 }, { "epoch": 3.82, "learning_rate": 2.8355568152125797e-07, "loss": 0.0195, "step": 9690 }, { "epoch": 3.82, "learning_rate": 2.8353248276888527e-07, "loss": 0.0187, "step": 9691 }, { "epoch": 3.82, "learning_rate": 2.8350928265506004e-07, "loss": 0.0194, "step": 9692 }, { "epoch": 3.82, "learning_rate": 2.8348608118016044e-07, "loss": 0.0163, "step": 9693 }, { "epoch": 3.82, "learning_rate": 2.8346287834456457e-07, "loss": 0.0173, "step": 9694 }, { "epoch": 3.82, "learning_rate": 2.8343967414865067e-07, "loss": 0.0167, "step": 9695 }, { "epoch": 3.82, "learning_rate": 2.8341646859279683e-07, "loss": 0.016, "step": 9696 }, { "epoch": 3.82, "learning_rate": 2.833932616773814e-07, "loss": 0.0168, "step": 9697 }, { "epoch": 3.82, "learning_rate": 2.8337005340278255e-07, "loss": 0.0157, "step": 9698 }, { "epoch": 3.82, "learning_rate": 2.833468437693785e-07, "loss": 0.0166, "step": 9699 }, { "epoch": 3.82, "learning_rate": 2.833236327775475e-07, "loss": 0.0196, "step": 9700 }, { "epoch": 3.82, "learning_rate": 2.8330042042766804e-07, "loss": 0.0149, "step": 9701 }, { "epoch": 3.82, "learning_rate": 2.8327720672011836e-07, "loss": 0.0208, "step": 9702 }, { "epoch": 3.82, "learning_rate": 2.832539916552767e-07, "loss": 0.0183, "step": 9703 }, { "epoch": 3.82, "learning_rate": 2.832307752335215e-07, "loss": 0.0171, "step": 9704 }, { "epoch": 3.82, "learning_rate": 2.8320755745523117e-07, "loss": 0.019, "step": 9705 }, { "epoch": 3.83, "learning_rate": 2.831843383207841e-07, "loss": 0.018, "step": 9706 }, { "epoch": 3.83, "learning_rate": 2.8316111783055877e-07, "loss": 0.0165, "step": 9707 }, { "epoch": 3.83, "learning_rate": 2.831378959849336e-07, "loss": 0.0134, "step": 9708 }, { "epoch": 3.83, "learning_rate": 2.831146727842871e-07, "loss": 0.0199, "step": 9709 }, { "epoch": 3.83, "learning_rate": 2.8309144822899773e-07, "loss": 0.0146, "step": 9710 }, { "epoch": 3.83, "learning_rate": 2.83068222319444e-07, "loss": 0.0194, "step": 9711 }, { "epoch": 3.83, "learning_rate": 2.8304499505600453e-07, "loss": 0.017, "step": 9712 }, { "epoch": 3.83, "learning_rate": 2.830217664390579e-07, "loss": 0.0172, "step": 9713 }, { "epoch": 3.83, "learning_rate": 2.8299853646898257e-07, "loss": 0.0174, "step": 9714 }, { "epoch": 3.83, "learning_rate": 2.829753051461573e-07, "loss": 0.0171, "step": 9715 }, { "epoch": 3.83, "learning_rate": 2.8295207247096067e-07, "loss": 0.0207, "step": 9716 }, { "epoch": 3.83, "learning_rate": 2.8292883844377133e-07, "loss": 0.0146, "step": 9717 }, { "epoch": 3.83, "learning_rate": 2.829056030649679e-07, "loss": 0.0204, "step": 9718 }, { "epoch": 3.83, "learning_rate": 2.8288236633492917e-07, "loss": 0.0154, "step": 9719 }, { "epoch": 3.83, "learning_rate": 2.828591282540339e-07, "loss": 0.0179, "step": 9720 }, { "epoch": 3.83, "learning_rate": 2.8283588882266076e-07, "loss": 0.0167, "step": 9721 }, { "epoch": 3.83, "learning_rate": 2.8281264804118846e-07, "loss": 0.0163, "step": 9722 }, { "epoch": 3.83, "learning_rate": 2.827894059099959e-07, "loss": 0.0209, "step": 9723 }, { "epoch": 3.83, "learning_rate": 2.827661624294618e-07, "loss": 0.0191, "step": 9724 }, { "epoch": 3.83, "learning_rate": 2.827429175999651e-07, "loss": 0.0178, "step": 9725 }, { "epoch": 3.83, "learning_rate": 2.8271967142188456e-07, "loss": 0.0156, "step": 9726 }, { "epoch": 3.83, "learning_rate": 2.826964238955991e-07, "loss": 0.0189, "step": 9727 }, { "epoch": 3.83, "learning_rate": 2.8267317502148764e-07, "loss": 0.0193, "step": 9728 }, { "epoch": 3.83, "learning_rate": 2.82649924799929e-07, "loss": 0.016, "step": 9729 }, { "epoch": 3.83, "learning_rate": 2.826266732313023e-07, "loss": 0.0133, "step": 9730 }, { "epoch": 3.84, "learning_rate": 2.826034203159864e-07, "loss": 0.0194, "step": 9731 }, { "epoch": 3.84, "learning_rate": 2.825801660543602e-07, "loss": 0.0147, "step": 9732 }, { "epoch": 3.84, "learning_rate": 2.825569104468029e-07, "loss": 0.0176, "step": 9733 }, { "epoch": 3.84, "learning_rate": 2.8253365349369336e-07, "loss": 0.016, "step": 9734 }, { "epoch": 3.84, "learning_rate": 2.825103951954107e-07, "loss": 0.0162, "step": 9735 }, { "epoch": 3.84, "learning_rate": 2.8248713555233406e-07, "loss": 0.0168, "step": 9736 }, { "epoch": 3.84, "learning_rate": 2.8246387456484247e-07, "loss": 0.0191, "step": 9737 }, { "epoch": 3.84, "learning_rate": 2.82440612233315e-07, "loss": 0.0168, "step": 9738 }, { "epoch": 3.84, "learning_rate": 2.8241734855813095e-07, "loss": 0.0228, "step": 9739 }, { "epoch": 3.84, "learning_rate": 2.8239408353966927e-07, "loss": 0.0198, "step": 9740 }, { "epoch": 3.84, "learning_rate": 2.8237081717830933e-07, "loss": 0.016, "step": 9741 }, { "epoch": 3.84, "learning_rate": 2.823475494744302e-07, "loss": 0.0185, "step": 9742 }, { "epoch": 3.84, "learning_rate": 2.823242804284112e-07, "loss": 0.0157, "step": 9743 }, { "epoch": 3.84, "learning_rate": 2.823010100406316e-07, "loss": 0.0183, "step": 9744 }, { "epoch": 3.84, "learning_rate": 2.8227773831147044e-07, "loss": 0.0176, "step": 9745 }, { "epoch": 3.84, "learning_rate": 2.822544652413073e-07, "loss": 0.0185, "step": 9746 }, { "epoch": 3.84, "learning_rate": 2.8223119083052145e-07, "loss": 0.0221, "step": 9747 }, { "epoch": 3.84, "learning_rate": 2.8220791507949213e-07, "loss": 0.0189, "step": 9748 }, { "epoch": 3.84, "learning_rate": 2.8218463798859866e-07, "loss": 0.0174, "step": 9749 }, { "epoch": 3.84, "learning_rate": 2.821613595582206e-07, "loss": 0.0167, "step": 9750 }, { "epoch": 3.84, "learning_rate": 2.8213807978873726e-07, "loss": 0.0169, "step": 9751 }, { "epoch": 3.84, "learning_rate": 2.8211479868052794e-07, "loss": 0.0198, "step": 9752 }, { "epoch": 3.84, "learning_rate": 2.8209151623397226e-07, "loss": 0.0192, "step": 9753 }, { "epoch": 3.84, "learning_rate": 2.8206823244944963e-07, "loss": 0.0161, "step": 9754 }, { "epoch": 3.84, "learning_rate": 2.8204494732733954e-07, "loss": 0.0173, "step": 9755 }, { "epoch": 3.84, "learning_rate": 2.820216608680215e-07, "loss": 0.019, "step": 9756 }, { "epoch": 3.85, "learning_rate": 2.8199837307187507e-07, "loss": 0.0152, "step": 9757 }, { "epoch": 3.85, "learning_rate": 2.8197508393927975e-07, "loss": 0.016, "step": 9758 }, { "epoch": 3.85, "learning_rate": 2.8195179347061516e-07, "loss": 0.0141, "step": 9759 }, { "epoch": 3.85, "learning_rate": 2.8192850166626087e-07, "loss": 0.0164, "step": 9760 }, { "epoch": 3.85, "learning_rate": 2.8190520852659657e-07, "loss": 0.0182, "step": 9761 }, { "epoch": 3.85, "learning_rate": 2.818819140520019e-07, "loss": 0.0193, "step": 9762 }, { "epoch": 3.85, "learning_rate": 2.8185861824285634e-07, "loss": 0.0191, "step": 9763 }, { "epoch": 3.85, "learning_rate": 2.818353210995398e-07, "loss": 0.0151, "step": 9764 }, { "epoch": 3.85, "learning_rate": 2.818120226224319e-07, "loss": 0.0134, "step": 9765 }, { "epoch": 3.85, "learning_rate": 2.817887228119123e-07, "loss": 0.0182, "step": 9766 }, { "epoch": 3.85, "learning_rate": 2.817654216683609e-07, "loss": 0.0173, "step": 9767 }, { "epoch": 3.85, "learning_rate": 2.817421191921574e-07, "loss": 0.0175, "step": 9768 }, { "epoch": 3.85, "learning_rate": 2.817188153836815e-07, "loss": 0.0191, "step": 9769 }, { "epoch": 3.85, "learning_rate": 2.8169551024331324e-07, "loss": 0.0132, "step": 9770 }, { "epoch": 3.85, "learning_rate": 2.816722037714322e-07, "loss": 0.0134, "step": 9771 }, { "epoch": 3.85, "learning_rate": 2.8164889596841843e-07, "loss": 0.0172, "step": 9772 }, { "epoch": 3.85, "learning_rate": 2.816255868346517e-07, "loss": 0.0178, "step": 9773 }, { "epoch": 3.85, "learning_rate": 2.8160227637051204e-07, "loss": 0.016, "step": 9774 }, { "epoch": 3.85, "learning_rate": 2.8157896457637925e-07, "loss": 0.0167, "step": 9775 }, { "epoch": 3.85, "learning_rate": 2.8155565145263327e-07, "loss": 0.0159, "step": 9776 }, { "epoch": 3.85, "learning_rate": 2.8153233699965427e-07, "loss": 0.0194, "step": 9777 }, { "epoch": 3.85, "learning_rate": 2.8150902121782197e-07, "loss": 0.0202, "step": 9778 }, { "epoch": 3.85, "learning_rate": 2.8148570410751653e-07, "loss": 0.0153, "step": 9779 }, { "epoch": 3.85, "learning_rate": 2.814623856691179e-07, "loss": 0.0142, "step": 9780 }, { "epoch": 3.85, "learning_rate": 2.8143906590300627e-07, "loss": 0.0174, "step": 9781 }, { "epoch": 3.86, "learning_rate": 2.8141574480956165e-07, "loss": 0.0181, "step": 9782 }, { "epoch": 3.86, "learning_rate": 2.8139242238916405e-07, "loss": 0.0168, "step": 9783 }, { "epoch": 3.86, "learning_rate": 2.8136909864219375e-07, "loss": 0.017, "step": 9784 }, { "epoch": 3.86, "learning_rate": 2.8134577356903084e-07, "loss": 0.0223, "step": 9785 }, { "epoch": 3.86, "learning_rate": 2.813224471700554e-07, "loss": 0.0145, "step": 9786 }, { "epoch": 3.86, "learning_rate": 2.812991194456476e-07, "loss": 0.0212, "step": 9787 }, { "epoch": 3.86, "learning_rate": 2.812757903961878e-07, "loss": 0.0145, "step": 9788 }, { "epoch": 3.86, "learning_rate": 2.812524600220561e-07, "loss": 0.0172, "step": 9789 }, { "epoch": 3.86, "learning_rate": 2.8122912832363276e-07, "loss": 0.0179, "step": 9790 }, { "epoch": 3.86, "learning_rate": 2.8120579530129815e-07, "loss": 0.0181, "step": 9791 }, { "epoch": 3.86, "learning_rate": 2.811824609554325e-07, "loss": 0.0188, "step": 9792 }, { "epoch": 3.86, "learning_rate": 2.8115912528641605e-07, "loss": 0.0194, "step": 9793 }, { "epoch": 3.86, "learning_rate": 2.8113578829462925e-07, "loss": 0.0149, "step": 9794 }, { "epoch": 3.86, "learning_rate": 2.8111244998045244e-07, "loss": 0.0195, "step": 9795 }, { "epoch": 3.86, "learning_rate": 2.810891103442659e-07, "loss": 0.0186, "step": 9796 }, { "epoch": 3.86, "learning_rate": 2.8106576938645017e-07, "loss": 0.0176, "step": 9797 }, { "epoch": 3.86, "learning_rate": 2.810424271073855e-07, "loss": 0.0166, "step": 9798 }, { "epoch": 3.86, "learning_rate": 2.8101908350745254e-07, "loss": 0.0206, "step": 9799 }, { "epoch": 3.86, "learning_rate": 2.809957385870316e-07, "loss": 0.0179, "step": 9800 }, { "epoch": 3.86, "learning_rate": 2.809723923465033e-07, "loss": 0.0153, "step": 9801 }, { "epoch": 3.86, "learning_rate": 2.80949044786248e-07, "loss": 0.0175, "step": 9802 }, { "epoch": 3.86, "learning_rate": 2.809256959066463e-07, "loss": 0.0151, "step": 9803 }, { "epoch": 3.86, "learning_rate": 2.809023457080788e-07, "loss": 0.0167, "step": 9804 }, { "epoch": 3.86, "learning_rate": 2.8087899419092595e-07, "loss": 0.0165, "step": 9805 }, { "epoch": 3.86, "learning_rate": 2.808556413555685e-07, "loss": 0.0164, "step": 9806 }, { "epoch": 3.87, "learning_rate": 2.808322872023869e-07, "loss": 0.0177, "step": 9807 }, { "epoch": 3.87, "learning_rate": 2.8080893173176195e-07, "loss": 0.0185, "step": 9808 }, { "epoch": 3.87, "learning_rate": 2.807855749440741e-07, "loss": 0.0193, "step": 9809 }, { "epoch": 3.87, "learning_rate": 2.8076221683970434e-07, "loss": 0.0221, "step": 9810 }, { "epoch": 3.87, "learning_rate": 2.807388574190331e-07, "loss": 0.0157, "step": 9811 }, { "epoch": 3.87, "learning_rate": 2.807154966824411e-07, "loss": 0.0167, "step": 9812 }, { "epoch": 3.87, "learning_rate": 2.806921346303093e-07, "loss": 0.0158, "step": 9813 }, { "epoch": 3.87, "learning_rate": 2.806687712630183e-07, "loss": 0.0152, "step": 9814 }, { "epoch": 3.87, "learning_rate": 2.8064540658094896e-07, "loss": 0.0172, "step": 9815 }, { "epoch": 3.87, "learning_rate": 2.80622040584482e-07, "loss": 0.0171, "step": 9816 }, { "epoch": 3.87, "learning_rate": 2.805986732739984e-07, "loss": 0.0176, "step": 9817 }, { "epoch": 3.87, "learning_rate": 2.805753046498788e-07, "loss": 0.0188, "step": 9818 }, { "epoch": 3.87, "learning_rate": 2.805519347125043e-07, "loss": 0.0156, "step": 9819 }, { "epoch": 3.87, "learning_rate": 2.8052856346225564e-07, "loss": 0.0177, "step": 9820 }, { "epoch": 3.87, "learning_rate": 2.8050519089951377e-07, "loss": 0.0198, "step": 9821 }, { "epoch": 3.87, "learning_rate": 2.8048181702465973e-07, "loss": 0.0146, "step": 9822 }, { "epoch": 3.87, "learning_rate": 2.804584418380743e-07, "loss": 0.0198, "step": 9823 }, { "epoch": 3.87, "learning_rate": 2.8043506534013855e-07, "loss": 0.0165, "step": 9824 }, { "epoch": 3.87, "learning_rate": 2.804116875312335e-07, "loss": 0.0213, "step": 9825 }, { "epoch": 3.87, "learning_rate": 2.803883084117402e-07, "loss": 0.0157, "step": 9826 }, { "epoch": 3.87, "learning_rate": 2.8036492798203966e-07, "loss": 0.0211, "step": 9827 }, { "epoch": 3.87, "learning_rate": 2.8034154624251287e-07, "loss": 0.0191, "step": 9828 }, { "epoch": 3.87, "learning_rate": 2.80318163193541e-07, "loss": 0.0186, "step": 9829 }, { "epoch": 3.87, "learning_rate": 2.8029477883550514e-07, "loss": 0.0171, "step": 9830 }, { "epoch": 3.87, "learning_rate": 2.8027139316878643e-07, "loss": 0.0202, "step": 9831 }, { "epoch": 3.87, "learning_rate": 2.80248006193766e-07, "loss": 0.0162, "step": 9832 }, { "epoch": 3.88, "learning_rate": 2.8022461791082503e-07, "loss": 0.0159, "step": 9833 }, { "epoch": 3.88, "learning_rate": 2.802012283203447e-07, "loss": 0.015, "step": 9834 }, { "epoch": 3.88, "learning_rate": 2.801778374227063e-07, "loss": 0.0179, "step": 9835 }, { "epoch": 3.88, "learning_rate": 2.8015444521829093e-07, "loss": 0.0197, "step": 9836 }, { "epoch": 3.88, "learning_rate": 2.8013105170748e-07, "loss": 0.0193, "step": 9837 }, { "epoch": 3.88, "learning_rate": 2.801076568906546e-07, "loss": 0.0134, "step": 9838 }, { "epoch": 3.88, "learning_rate": 2.8008426076819625e-07, "loss": 0.0166, "step": 9839 }, { "epoch": 3.88, "learning_rate": 2.800608633404861e-07, "loss": 0.018, "step": 9840 }, { "epoch": 3.88, "learning_rate": 2.800374646079056e-07, "loss": 0.0172, "step": 9841 }, { "epoch": 3.88, "learning_rate": 2.800140645708361e-07, "loss": 0.0166, "step": 9842 }, { "epoch": 3.88, "learning_rate": 2.799906632296588e-07, "loss": 0.0175, "step": 9843 }, { "epoch": 3.88, "learning_rate": 2.7996726058475544e-07, "loss": 0.0145, "step": 9844 }, { "epoch": 3.88, "learning_rate": 2.7994385663650716e-07, "loss": 0.0172, "step": 9845 }, { "epoch": 3.88, "learning_rate": 2.7992045138529557e-07, "loss": 0.0176, "step": 9846 }, { "epoch": 3.88, "learning_rate": 2.7989704483150205e-07, "loss": 0.0201, "step": 9847 }, { "epoch": 3.88, "learning_rate": 2.7987363697550814e-07, "loss": 0.0191, "step": 9848 }, { "epoch": 3.88, "learning_rate": 2.7985022781769536e-07, "loss": 0.0152, "step": 9849 }, { "epoch": 3.88, "learning_rate": 2.7982681735844526e-07, "loss": 0.0206, "step": 9850 }, { "epoch": 3.88, "learning_rate": 2.798034055981393e-07, "loss": 0.0175, "step": 9851 }, { "epoch": 3.88, "learning_rate": 2.797799925371592e-07, "loss": 0.019, "step": 9852 }, { "epoch": 3.88, "learning_rate": 2.797565781758864e-07, "loss": 0.017, "step": 9853 }, { "epoch": 3.88, "learning_rate": 2.7973316251470255e-07, "loss": 0.0157, "step": 9854 }, { "epoch": 3.88, "learning_rate": 2.797097455539894e-07, "loss": 0.0199, "step": 9855 }, { "epoch": 3.88, "learning_rate": 2.796863272941285e-07, "loss": 0.0185, "step": 9856 }, { "epoch": 3.88, "learning_rate": 2.796629077355016e-07, "loss": 0.0157, "step": 9857 }, { "epoch": 3.89, "learning_rate": 2.796394868784904e-07, "loss": 0.02, "step": 9858 }, { "epoch": 3.89, "learning_rate": 2.7961606472347654e-07, "loss": 0.0213, "step": 9859 }, { "epoch": 3.89, "learning_rate": 2.7959264127084183e-07, "loss": 0.0186, "step": 9860 }, { "epoch": 3.89, "learning_rate": 2.7956921652096806e-07, "loss": 0.0182, "step": 9861 }, { "epoch": 3.89, "learning_rate": 2.7954579047423703e-07, "loss": 0.0207, "step": 9862 }, { "epoch": 3.89, "learning_rate": 2.795223631310305e-07, "loss": 0.0164, "step": 9863 }, { "epoch": 3.89, "learning_rate": 2.794989344917303e-07, "loss": 0.0155, "step": 9864 }, { "epoch": 3.89, "learning_rate": 2.7947550455671824e-07, "loss": 0.0196, "step": 9865 }, { "epoch": 3.89, "learning_rate": 2.794520733263763e-07, "loss": 0.0154, "step": 9866 }, { "epoch": 3.89, "learning_rate": 2.794286408010863e-07, "loss": 0.0173, "step": 9867 }, { "epoch": 3.89, "learning_rate": 2.794052069812302e-07, "loss": 0.0188, "step": 9868 }, { "epoch": 3.89, "learning_rate": 2.7938177186718987e-07, "loss": 0.0161, "step": 9869 }, { "epoch": 3.89, "learning_rate": 2.793583354593473e-07, "loss": 0.0199, "step": 9870 }, { "epoch": 3.89, "learning_rate": 2.793348977580845e-07, "loss": 0.0163, "step": 9871 }, { "epoch": 3.89, "learning_rate": 2.793114587637835e-07, "loss": 0.0166, "step": 9872 }, { "epoch": 3.89, "learning_rate": 2.7928801847682615e-07, "loss": 0.0182, "step": 9873 }, { "epoch": 3.89, "learning_rate": 2.792645768975947e-07, "loss": 0.016, "step": 9874 }, { "epoch": 3.89, "learning_rate": 2.7924113402647104e-07, "loss": 0.0184, "step": 9875 }, { "epoch": 3.89, "learning_rate": 2.7921768986383736e-07, "loss": 0.0162, "step": 9876 }, { "epoch": 3.89, "learning_rate": 2.791942444100757e-07, "loss": 0.0167, "step": 9877 }, { "epoch": 3.89, "learning_rate": 2.791707976655682e-07, "loss": 0.0226, "step": 9878 }, { "epoch": 3.89, "learning_rate": 2.791473496306971e-07, "loss": 0.0175, "step": 9879 }, { "epoch": 3.89, "learning_rate": 2.7912390030584447e-07, "loss": 0.017, "step": 9880 }, { "epoch": 3.89, "learning_rate": 2.791004496913925e-07, "loss": 0.0178, "step": 9881 }, { "epoch": 3.89, "learning_rate": 2.7907699778772343e-07, "loss": 0.0181, "step": 9882 }, { "epoch": 3.89, "learning_rate": 2.790535445952194e-07, "loss": 0.0165, "step": 9883 }, { "epoch": 3.9, "learning_rate": 2.7903009011426293e-07, "loss": 0.0243, "step": 9884 }, { "epoch": 3.9, "learning_rate": 2.7900663434523594e-07, "loss": 0.0161, "step": 9885 }, { "epoch": 3.9, "learning_rate": 2.789831772885209e-07, "loss": 0.0183, "step": 9886 }, { "epoch": 3.9, "learning_rate": 2.7895971894450017e-07, "loss": 0.0201, "step": 9887 }, { "epoch": 3.9, "learning_rate": 2.78936259313556e-07, "loss": 0.0161, "step": 9888 }, { "epoch": 3.9, "learning_rate": 2.7891279839607074e-07, "loss": 0.0159, "step": 9889 }, { "epoch": 3.9, "learning_rate": 2.788893361924269e-07, "loss": 0.0237, "step": 9890 }, { "epoch": 3.9, "learning_rate": 2.788658727030067e-07, "loss": 0.0153, "step": 9891 }, { "epoch": 3.9, "learning_rate": 2.7884240792819265e-07, "loss": 0.0161, "step": 9892 }, { "epoch": 3.9, "learning_rate": 2.7881894186836714e-07, "loss": 0.0174, "step": 9893 }, { "epoch": 3.9, "learning_rate": 2.787954745239128e-07, "loss": 0.016, "step": 9894 }, { "epoch": 3.9, "learning_rate": 2.7877200589521183e-07, "loss": 0.0161, "step": 9895 }, { "epoch": 3.9, "learning_rate": 2.787485359826469e-07, "loss": 0.0161, "step": 9896 }, { "epoch": 3.9, "learning_rate": 2.7872506478660064e-07, "loss": 0.016, "step": 9897 }, { "epoch": 3.9, "learning_rate": 2.787015923074553e-07, "loss": 0.0167, "step": 9898 }, { "epoch": 3.9, "learning_rate": 2.786781185455938e-07, "loss": 0.0169, "step": 9899 }, { "epoch": 3.9, "learning_rate": 2.786546435013984e-07, "loss": 0.0205, "step": 9900 }, { "epoch": 3.9, "learning_rate": 2.786311671752519e-07, "loss": 0.0185, "step": 9901 }, { "epoch": 3.9, "learning_rate": 2.7860768956753686e-07, "loss": 0.0163, "step": 9902 }, { "epoch": 3.9, "learning_rate": 2.78584210678636e-07, "loss": 0.0161, "step": 9903 }, { "epoch": 3.9, "learning_rate": 2.7856073050893194e-07, "loss": 0.0189, "step": 9904 }, { "epoch": 3.9, "learning_rate": 2.785372490588073e-07, "loss": 0.0155, "step": 9905 }, { "epoch": 3.9, "learning_rate": 2.7851376632864495e-07, "loss": 0.0165, "step": 9906 }, { "epoch": 3.9, "learning_rate": 2.7849028231882745e-07, "loss": 0.0162, "step": 9907 }, { "epoch": 3.9, "learning_rate": 2.7846679702973774e-07, "loss": 0.0185, "step": 9908 }, { "epoch": 3.91, "learning_rate": 2.7844331046175836e-07, "loss": 0.017, "step": 9909 }, { "epoch": 3.91, "learning_rate": 2.784198226152724e-07, "loss": 0.0187, "step": 9910 }, { "epoch": 3.91, "learning_rate": 2.783963334906624e-07, "loss": 0.0183, "step": 9911 }, { "epoch": 3.91, "learning_rate": 2.7837284308831135e-07, "loss": 0.0168, "step": 9912 }, { "epoch": 3.91, "learning_rate": 2.7834935140860205e-07, "loss": 0.016, "step": 9913 }, { "epoch": 3.91, "learning_rate": 2.783258584519175e-07, "loss": 0.0163, "step": 9914 }, { "epoch": 3.91, "learning_rate": 2.783023642186404e-07, "loss": 0.0169, "step": 9915 }, { "epoch": 3.91, "learning_rate": 2.782788687091537e-07, "loss": 0.0176, "step": 9916 }, { "epoch": 3.91, "learning_rate": 2.7825537192384057e-07, "loss": 0.0166, "step": 9917 }, { "epoch": 3.91, "learning_rate": 2.7823187386308373e-07, "loss": 0.0154, "step": 9918 }, { "epoch": 3.91, "learning_rate": 2.782083745272663e-07, "loss": 0.0129, "step": 9919 }, { "epoch": 3.91, "learning_rate": 2.781848739167711e-07, "loss": 0.0196, "step": 9920 }, { "epoch": 3.91, "learning_rate": 2.781613720319814e-07, "loss": 0.0191, "step": 9921 }, { "epoch": 3.91, "learning_rate": 2.781378688732801e-07, "loss": 0.0195, "step": 9922 }, { "epoch": 3.91, "learning_rate": 2.7811436444105026e-07, "loss": 0.0169, "step": 9923 }, { "epoch": 3.91, "learning_rate": 2.78090858735675e-07, "loss": 0.0169, "step": 9924 }, { "epoch": 3.91, "learning_rate": 2.780673517575375e-07, "loss": 0.0191, "step": 9925 }, { "epoch": 3.91, "learning_rate": 2.780438435070207e-07, "loss": 0.0175, "step": 9926 }, { "epoch": 3.91, "learning_rate": 2.780203339845079e-07, "loss": 0.0184, "step": 9927 }, { "epoch": 3.91, "learning_rate": 2.7799682319038235e-07, "loss": 0.0162, "step": 9928 }, { "epoch": 3.91, "learning_rate": 2.7797331112502697e-07, "loss": 0.0145, "step": 9929 }, { "epoch": 3.91, "learning_rate": 2.779497977888252e-07, "loss": 0.0171, "step": 9930 }, { "epoch": 3.91, "learning_rate": 2.779262831821602e-07, "loss": 0.0172, "step": 9931 }, { "epoch": 3.91, "learning_rate": 2.779027673054152e-07, "loss": 0.0196, "step": 9932 }, { "epoch": 3.91, "learning_rate": 2.778792501589735e-07, "loss": 0.0175, "step": 9933 }, { "epoch": 3.92, "learning_rate": 2.7785573174321837e-07, "loss": 0.0183, "step": 9934 }, { "epoch": 3.92, "learning_rate": 2.778322120585332e-07, "loss": 0.0148, "step": 9935 }, { "epoch": 3.92, "learning_rate": 2.778086911053012e-07, "loss": 0.0204, "step": 9936 }, { "epoch": 3.92, "learning_rate": 2.777851688839059e-07, "loss": 0.0156, "step": 9937 }, { "epoch": 3.92, "learning_rate": 2.777616453947304e-07, "loss": 0.0153, "step": 9938 }, { "epoch": 3.92, "learning_rate": 2.7773812063815834e-07, "loss": 0.0136, "step": 9939 }, { "epoch": 3.92, "learning_rate": 2.7771459461457307e-07, "loss": 0.016, "step": 9940 }, { "epoch": 3.92, "learning_rate": 2.77691067324358e-07, "loss": 0.0206, "step": 9941 }, { "epoch": 3.92, "learning_rate": 2.776675387678966e-07, "loss": 0.019, "step": 9942 }, { "epoch": 3.92, "learning_rate": 2.776440089455724e-07, "loss": 0.0139, "step": 9943 }, { "epoch": 3.92, "learning_rate": 2.7762047785776883e-07, "loss": 0.0211, "step": 9944 }, { "epoch": 3.92, "learning_rate": 2.7759694550486947e-07, "loss": 0.019, "step": 9945 }, { "epoch": 3.92, "learning_rate": 2.775734118872578e-07, "loss": 0.0171, "step": 9946 }, { "epoch": 3.92, "learning_rate": 2.775498770053174e-07, "loss": 0.0133, "step": 9947 }, { "epoch": 3.92, "learning_rate": 2.775263408594319e-07, "loss": 0.0185, "step": 9948 }, { "epoch": 3.92, "learning_rate": 2.7750280344998487e-07, "loss": 0.0161, "step": 9949 }, { "epoch": 3.92, "learning_rate": 2.7747926477735993e-07, "loss": 0.0188, "step": 9950 }, { "epoch": 3.92, "learning_rate": 2.7745572484194074e-07, "loss": 0.0155, "step": 9951 }, { "epoch": 3.92, "learning_rate": 2.774321836441109e-07, "loss": 0.0187, "step": 9952 }, { "epoch": 3.92, "learning_rate": 2.774086411842542e-07, "loss": 0.0158, "step": 9953 }, { "epoch": 3.92, "learning_rate": 2.773850974627543e-07, "loss": 0.016, "step": 9954 }, { "epoch": 3.92, "learning_rate": 2.773615524799949e-07, "loss": 0.0175, "step": 9955 }, { "epoch": 3.92, "learning_rate": 2.7733800623635977e-07, "loss": 0.015, "step": 9956 }, { "epoch": 3.92, "learning_rate": 2.773144587322327e-07, "loss": 0.0149, "step": 9957 }, { "epoch": 3.92, "learning_rate": 2.772909099679975e-07, "loss": 0.0185, "step": 9958 }, { "epoch": 3.92, "learning_rate": 2.7726735994403786e-07, "loss": 0.017, "step": 9959 }, { "epoch": 3.93, "learning_rate": 2.772438086607377e-07, "loss": 0.019, "step": 9960 }, { "epoch": 3.93, "learning_rate": 2.7722025611848087e-07, "loss": 0.016, "step": 9961 }, { "epoch": 3.93, "learning_rate": 2.7719670231765125e-07, "loss": 0.0187, "step": 9962 }, { "epoch": 3.93, "learning_rate": 2.7717314725863264e-07, "loss": 0.0206, "step": 9963 }, { "epoch": 3.93, "learning_rate": 2.7714959094180906e-07, "loss": 0.0181, "step": 9964 }, { "epoch": 3.93, "learning_rate": 2.771260333675644e-07, "loss": 0.0218, "step": 9965 }, { "epoch": 3.93, "learning_rate": 2.771024745362826e-07, "loss": 0.0161, "step": 9966 }, { "epoch": 3.93, "learning_rate": 2.7707891444834764e-07, "loss": 0.0191, "step": 9967 }, { "epoch": 3.93, "learning_rate": 2.7705535310414355e-07, "loss": 0.0167, "step": 9968 }, { "epoch": 3.93, "learning_rate": 2.770317905040543e-07, "loss": 0.013, "step": 9969 }, { "epoch": 3.93, "learning_rate": 2.77008226648464e-07, "loss": 0.0199, "step": 9970 }, { "epoch": 3.93, "learning_rate": 2.769846615377566e-07, "loss": 0.0177, "step": 9971 }, { "epoch": 3.93, "learning_rate": 2.7696109517231616e-07, "loss": 0.021, "step": 9972 }, { "epoch": 3.93, "learning_rate": 2.769375275525269e-07, "loss": 0.0138, "step": 9973 }, { "epoch": 3.93, "learning_rate": 2.7691395867877284e-07, "loss": 0.0151, "step": 9974 }, { "epoch": 3.93, "learning_rate": 2.768903885514382e-07, "loss": 0.0178, "step": 9975 }, { "epoch": 3.93, "learning_rate": 2.76866817170907e-07, "loss": 0.0166, "step": 9976 }, { "epoch": 3.93, "learning_rate": 2.768432445375636e-07, "loss": 0.0177, "step": 9977 }, { "epoch": 3.93, "learning_rate": 2.768196706517921e-07, "loss": 0.0212, "step": 9978 }, { "epoch": 3.93, "learning_rate": 2.7679609551397664e-07, "loss": 0.0204, "step": 9979 }, { "epoch": 3.93, "learning_rate": 2.7677251912450155e-07, "loss": 0.0172, "step": 9980 }, { "epoch": 3.93, "learning_rate": 2.7674894148375117e-07, "loss": 0.0162, "step": 9981 }, { "epoch": 3.93, "learning_rate": 2.767253625921096e-07, "loss": 0.0161, "step": 9982 }, { "epoch": 3.93, "learning_rate": 2.767017824499613e-07, "loss": 0.0233, "step": 9983 }, { "epoch": 3.93, "learning_rate": 2.7667820105769046e-07, "loss": 0.0153, "step": 9984 }, { "epoch": 3.94, "learning_rate": 2.766546184156815e-07, "loss": 0.0164, "step": 9985 }, { "epoch": 3.94, "learning_rate": 2.7663103452431874e-07, "loss": 0.0168, "step": 9986 }, { "epoch": 3.94, "learning_rate": 2.7660744938398666e-07, "loss": 0.0145, "step": 9987 }, { "epoch": 3.94, "learning_rate": 2.765838629950695e-07, "loss": 0.016, "step": 9988 }, { "epoch": 3.94, "learning_rate": 2.7656027535795176e-07, "loss": 0.013, "step": 9989 }, { "epoch": 3.94, "learning_rate": 2.7653668647301794e-07, "loss": 0.0153, "step": 9990 }, { "epoch": 3.94, "learning_rate": 2.765130963406525e-07, "loss": 0.0177, "step": 9991 }, { "epoch": 3.94, "learning_rate": 2.764895049612398e-07, "loss": 0.0147, "step": 9992 }, { "epoch": 3.94, "learning_rate": 2.764659123351644e-07, "loss": 0.0193, "step": 9993 }, { "epoch": 3.94, "learning_rate": 2.7644231846281086e-07, "loss": 0.017, "step": 9994 }, { "epoch": 3.94, "learning_rate": 2.7641872334456375e-07, "loss": 0.0151, "step": 9995 }, { "epoch": 3.94, "learning_rate": 2.763951269808075e-07, "loss": 0.0189, "step": 9996 }, { "epoch": 3.94, "learning_rate": 2.763715293719269e-07, "loss": 0.017, "step": 9997 }, { "epoch": 3.94, "learning_rate": 2.7634793051830636e-07, "loss": 0.019, "step": 9998 }, { "epoch": 3.94, "learning_rate": 2.7632433042033064e-07, "loss": 0.0202, "step": 9999 }, { "epoch": 3.94, "learning_rate": 2.763007290783844e-07, "loss": 0.0162, "step": 10000 }, { "epoch": 3.94, "learning_rate": 2.7627712649285214e-07, "loss": 0.0151, "step": 10001 }, { "epoch": 3.94, "learning_rate": 2.7625352266411864e-07, "loss": 0.018, "step": 10002 }, { "epoch": 3.94, "learning_rate": 2.7622991759256863e-07, "loss": 0.0197, "step": 10003 }, { "epoch": 3.94, "learning_rate": 2.762063112785869e-07, "loss": 0.0145, "step": 10004 }, { "epoch": 3.94, "learning_rate": 2.76182703722558e-07, "loss": 0.0183, "step": 10005 }, { "epoch": 3.94, "learning_rate": 2.761590949248669e-07, "loss": 0.0163, "step": 10006 }, { "epoch": 3.94, "learning_rate": 2.761354848858982e-07, "loss": 0.0188, "step": 10007 }, { "epoch": 3.94, "learning_rate": 2.7611187360603685e-07, "loss": 0.0195, "step": 10008 }, { "epoch": 3.94, "learning_rate": 2.7608826108566767e-07, "loss": 0.02, "step": 10009 }, { "epoch": 3.95, "learning_rate": 2.7606464732517544e-07, "loss": 0.0171, "step": 10010 }, { "epoch": 3.95, "learning_rate": 2.7604103232494515e-07, "loss": 0.0213, "step": 10011 }, { "epoch": 3.95, "learning_rate": 2.760174160853615e-07, "loss": 0.0151, "step": 10012 }, { "epoch": 3.95, "learning_rate": 2.759937986068095e-07, "loss": 0.0148, "step": 10013 }, { "epoch": 3.95, "learning_rate": 2.7597017988967414e-07, "loss": 0.015, "step": 10014 }, { "epoch": 3.95, "learning_rate": 2.759465599343403e-07, "loss": 0.0154, "step": 10015 }, { "epoch": 3.95, "learning_rate": 2.7592293874119297e-07, "loss": 0.0158, "step": 10016 }, { "epoch": 3.95, "learning_rate": 2.75899316310617e-07, "loss": 0.0185, "step": 10017 }, { "epoch": 3.95, "learning_rate": 2.758756926429977e-07, "loss": 0.0186, "step": 10018 }, { "epoch": 3.95, "learning_rate": 2.7585206773871983e-07, "loss": 0.0176, "step": 10019 }, { "epoch": 3.95, "learning_rate": 2.758284415981686e-07, "loss": 0.0184, "step": 10020 }, { "epoch": 3.95, "learning_rate": 2.75804814221729e-07, "loss": 0.0163, "step": 10021 }, { "epoch": 3.95, "learning_rate": 2.757811856097861e-07, "loss": 0.0159, "step": 10022 }, { "epoch": 3.95, "learning_rate": 2.757575557627251e-07, "loss": 0.0138, "step": 10023 }, { "epoch": 3.95, "learning_rate": 2.7573392468093105e-07, "loss": 0.017, "step": 10024 }, { "epoch": 3.95, "learning_rate": 2.757102923647891e-07, "loss": 0.0154, "step": 10025 }, { "epoch": 3.95, "learning_rate": 2.7568665881468455e-07, "loss": 0.0161, "step": 10026 }, { "epoch": 3.95, "learning_rate": 2.756630240310024e-07, "loss": 0.0178, "step": 10027 }, { "epoch": 3.95, "learning_rate": 2.7563938801412793e-07, "loss": 0.0184, "step": 10028 }, { "epoch": 3.95, "learning_rate": 2.7561575076444646e-07, "loss": 0.0151, "step": 10029 }, { "epoch": 3.95, "learning_rate": 2.7559211228234317e-07, "loss": 0.0178, "step": 10030 }, { "epoch": 3.95, "learning_rate": 2.755684725682033e-07, "loss": 0.018, "step": 10031 }, { "epoch": 3.95, "learning_rate": 2.755448316224122e-07, "loss": 0.0173, "step": 10032 }, { "epoch": 3.95, "learning_rate": 2.7552118944535515e-07, "loss": 0.0138, "step": 10033 }, { "epoch": 3.95, "learning_rate": 2.754975460374175e-07, "loss": 0.0168, "step": 10034 }, { "epoch": 3.95, "learning_rate": 2.754739013989846e-07, "loss": 0.0177, "step": 10035 }, { "epoch": 3.96, "learning_rate": 2.754502555304418e-07, "loss": 0.0175, "step": 10036 }, { "epoch": 3.96, "learning_rate": 2.7542660843217453e-07, "loss": 0.017, "step": 10037 }, { "epoch": 3.96, "learning_rate": 2.754029601045681e-07, "loss": 0.0169, "step": 10038 }, { "epoch": 3.96, "learning_rate": 2.7537931054800814e-07, "loss": 0.019, "step": 10039 }, { "epoch": 3.96, "learning_rate": 2.7535565976287993e-07, "loss": 0.0204, "step": 10040 }, { "epoch": 3.96, "learning_rate": 2.7533200774956893e-07, "loss": 0.0169, "step": 10041 }, { "epoch": 3.96, "learning_rate": 2.7530835450846076e-07, "loss": 0.0211, "step": 10042 }, { "epoch": 3.96, "learning_rate": 2.7528470003994085e-07, "loss": 0.0175, "step": 10043 }, { "epoch": 3.96, "learning_rate": 2.752610443443948e-07, "loss": 0.0157, "step": 10044 }, { "epoch": 3.96, "learning_rate": 2.752373874222081e-07, "loss": 0.0181, "step": 10045 }, { "epoch": 3.96, "learning_rate": 2.7521372927376625e-07, "loss": 0.0167, "step": 10046 }, { "epoch": 3.96, "learning_rate": 2.7519006989945497e-07, "loss": 0.0188, "step": 10047 }, { "epoch": 3.96, "learning_rate": 2.7516640929965987e-07, "loss": 0.0226, "step": 10048 }, { "epoch": 3.96, "learning_rate": 2.751427474747665e-07, "loss": 0.0155, "step": 10049 }, { "epoch": 3.96, "learning_rate": 2.751190844251605e-07, "loss": 0.0178, "step": 10050 }, { "epoch": 3.96, "learning_rate": 2.7509542015122765e-07, "loss": 0.017, "step": 10051 }, { "epoch": 3.96, "learning_rate": 2.7507175465335357e-07, "loss": 0.0196, "step": 10052 }, { "epoch": 3.96, "learning_rate": 2.75048087931924e-07, "loss": 0.0185, "step": 10053 }, { "epoch": 3.96, "learning_rate": 2.750244199873246e-07, "loss": 0.0203, "step": 10054 }, { "epoch": 3.96, "learning_rate": 2.7500075081994127e-07, "loss": 0.0163, "step": 10055 }, { "epoch": 3.96, "learning_rate": 2.7497708043015963e-07, "loss": 0.0152, "step": 10056 }, { "epoch": 3.96, "learning_rate": 2.749534088183655e-07, "loss": 0.0172, "step": 10057 }, { "epoch": 3.96, "learning_rate": 2.749297359849448e-07, "loss": 0.0159, "step": 10058 }, { "epoch": 3.96, "learning_rate": 2.7490606193028323e-07, "loss": 0.0136, "step": 10059 }, { "epoch": 3.96, "learning_rate": 2.7488238665476675e-07, "loss": 0.0151, "step": 10060 }, { "epoch": 3.97, "learning_rate": 2.7485871015878114e-07, "loss": 0.0175, "step": 10061 }, { "epoch": 3.97, "learning_rate": 2.748350324427124e-07, "loss": 0.0186, "step": 10062 }, { "epoch": 3.97, "learning_rate": 2.748113535069462e-07, "loss": 0.0188, "step": 10063 }, { "epoch": 3.97, "learning_rate": 2.747876733518688e-07, "loss": 0.0179, "step": 10064 }, { "epoch": 3.97, "learning_rate": 2.747639919778659e-07, "loss": 0.0189, "step": 10065 }, { "epoch": 3.97, "learning_rate": 2.7474030938532354e-07, "loss": 0.0159, "step": 10066 }, { "epoch": 3.97, "learning_rate": 2.747166255746278e-07, "loss": 0.0147, "step": 10067 }, { "epoch": 3.97, "learning_rate": 2.746929405461646e-07, "loss": 0.0188, "step": 10068 }, { "epoch": 3.97, "learning_rate": 2.7466925430031993e-07, "loss": 0.0171, "step": 10069 }, { "epoch": 3.97, "learning_rate": 2.7464556683748e-07, "loss": 0.017, "step": 10070 }, { "epoch": 3.97, "learning_rate": 2.746218781580307e-07, "loss": 0.0192, "step": 10071 }, { "epoch": 3.97, "learning_rate": 2.745981882623582e-07, "loss": 0.0187, "step": 10072 }, { "epoch": 3.97, "learning_rate": 2.7457449715084864e-07, "loss": 0.0173, "step": 10073 }, { "epoch": 3.97, "learning_rate": 2.7455080482388813e-07, "loss": 0.0195, "step": 10074 }, { "epoch": 3.97, "learning_rate": 2.7452711128186277e-07, "loss": 0.0171, "step": 10075 }, { "epoch": 3.97, "learning_rate": 2.7450341652515877e-07, "loss": 0.016, "step": 10076 }, { "epoch": 3.97, "learning_rate": 2.7447972055416236e-07, "loss": 0.0188, "step": 10077 }, { "epoch": 3.97, "learning_rate": 2.7445602336925963e-07, "loss": 0.0187, "step": 10078 }, { "epoch": 3.97, "learning_rate": 2.744323249708369e-07, "loss": 0.0163, "step": 10079 }, { "epoch": 3.97, "learning_rate": 2.7440862535928043e-07, "loss": 0.0139, "step": 10080 }, { "epoch": 3.97, "learning_rate": 2.7438492453497645e-07, "loss": 0.0153, "step": 10081 }, { "epoch": 3.97, "learning_rate": 2.743612224983112e-07, "loss": 0.0185, "step": 10082 }, { "epoch": 3.97, "learning_rate": 2.74337519249671e-07, "loss": 0.0157, "step": 10083 }, { "epoch": 3.97, "learning_rate": 2.743138147894424e-07, "loss": 0.0146, "step": 10084 }, { "epoch": 3.97, "learning_rate": 2.7429010911801136e-07, "loss": 0.0138, "step": 10085 }, { "epoch": 3.97, "learning_rate": 2.742664022357646e-07, "loss": 0.0192, "step": 10086 }, { "epoch": 3.98, "learning_rate": 2.742426941430883e-07, "loss": 0.0183, "step": 10087 }, { "epoch": 3.98, "learning_rate": 2.7421898484036897e-07, "loss": 0.016, "step": 10088 }, { "epoch": 3.98, "learning_rate": 2.741952743279929e-07, "loss": 0.0176, "step": 10089 }, { "epoch": 3.98, "learning_rate": 2.741715626063467e-07, "loss": 0.0179, "step": 10090 }, { "epoch": 3.98, "learning_rate": 2.7414784967581665e-07, "loss": 0.0143, "step": 10091 }, { "epoch": 3.98, "learning_rate": 2.7412413553678944e-07, "loss": 0.0167, "step": 10092 }, { "epoch": 3.98, "learning_rate": 2.7410042018965146e-07, "loss": 0.0189, "step": 10093 }, { "epoch": 3.98, "learning_rate": 2.740767036347892e-07, "loss": 0.0156, "step": 10094 }, { "epoch": 3.98, "learning_rate": 2.740529858725893e-07, "loss": 0.0175, "step": 10095 }, { "epoch": 3.98, "learning_rate": 2.7402926690343826e-07, "loss": 0.0164, "step": 10096 }, { "epoch": 3.98, "learning_rate": 2.740055467277227e-07, "loss": 0.0181, "step": 10097 }, { "epoch": 3.98, "learning_rate": 2.739818253458292e-07, "loss": 0.0189, "step": 10098 }, { "epoch": 3.98, "learning_rate": 2.7395810275814435e-07, "loss": 0.0178, "step": 10099 }, { "epoch": 3.98, "learning_rate": 2.739343789650549e-07, "loss": 0.0166, "step": 10100 }, { "epoch": 3.98, "learning_rate": 2.7391065396694733e-07, "loss": 0.0215, "step": 10101 }, { "epoch": 3.98, "learning_rate": 2.738869277642085e-07, "loss": 0.0148, "step": 10102 }, { "epoch": 3.98, "learning_rate": 2.7386320035722503e-07, "loss": 0.0183, "step": 10103 }, { "epoch": 3.98, "learning_rate": 2.738394717463837e-07, "loss": 0.0184, "step": 10104 }, { "epoch": 3.98, "learning_rate": 2.7381574193207115e-07, "loss": 0.0188, "step": 10105 }, { "epoch": 3.98, "learning_rate": 2.737920109146742e-07, "loss": 0.0172, "step": 10106 }, { "epoch": 3.98, "learning_rate": 2.737682786945796e-07, "loss": 0.018, "step": 10107 }, { "epoch": 3.98, "learning_rate": 2.737445452721743e-07, "loss": 0.0183, "step": 10108 }, { "epoch": 3.98, "learning_rate": 2.737208106478449e-07, "loss": 0.0173, "step": 10109 }, { "epoch": 3.98, "learning_rate": 2.7369707482197834e-07, "loss": 0.0146, "step": 10110 }, { "epoch": 3.98, "learning_rate": 2.7367333779496136e-07, "loss": 0.0204, "step": 10111 }, { "epoch": 3.99, "learning_rate": 2.736495995671811e-07, "loss": 0.0153, "step": 10112 }, { "epoch": 3.99, "learning_rate": 2.7362586013902427e-07, "loss": 0.0178, "step": 10113 }, { "epoch": 3.99, "learning_rate": 2.7360211951087773e-07, "loss": 0.0158, "step": 10114 }, { "epoch": 3.99, "learning_rate": 2.735783776831286e-07, "loss": 0.0159, "step": 10115 }, { "epoch": 3.99, "learning_rate": 2.7355463465616364e-07, "loss": 0.016, "step": 10116 }, { "epoch": 3.99, "learning_rate": 2.7353089043037e-07, "loss": 0.0165, "step": 10117 }, { "epoch": 3.99, "learning_rate": 2.7350714500613457e-07, "loss": 0.0139, "step": 10118 }, { "epoch": 3.99, "learning_rate": 2.7348339838384444e-07, "loss": 0.0167, "step": 10119 }, { "epoch": 3.99, "learning_rate": 2.734596505638866e-07, "loss": 0.0181, "step": 10120 }, { "epoch": 3.99, "learning_rate": 2.7343590154664806e-07, "loss": 0.0162, "step": 10121 }, { "epoch": 3.99, "learning_rate": 2.7341215133251594e-07, "loss": 0.0153, "step": 10122 }, { "epoch": 3.99, "learning_rate": 2.733883999218773e-07, "loss": 0.0159, "step": 10123 }, { "epoch": 3.99, "learning_rate": 2.7336464731511937e-07, "loss": 0.0165, "step": 10124 }, { "epoch": 3.99, "learning_rate": 2.733408935126291e-07, "loss": 0.0145, "step": 10125 }, { "epoch": 3.99, "learning_rate": 2.733171385147938e-07, "loss": 0.0193, "step": 10126 }, { "epoch": 3.99, "learning_rate": 2.732933823220005e-07, "loss": 0.0203, "step": 10127 }, { "epoch": 3.99, "learning_rate": 2.732696249346365e-07, "loss": 0.0166, "step": 10128 }, { "epoch": 3.99, "learning_rate": 2.73245866353089e-07, "loss": 0.0205, "step": 10129 }, { "epoch": 3.99, "learning_rate": 2.732221065777451e-07, "loss": 0.0174, "step": 10130 }, { "epoch": 3.99, "learning_rate": 2.7319834560899225e-07, "loss": 0.0143, "step": 10131 }, { "epoch": 3.99, "learning_rate": 2.7317458344721757e-07, "loss": 0.0154, "step": 10132 }, { "epoch": 3.99, "learning_rate": 2.7315082009280844e-07, "loss": 0.0149, "step": 10133 }, { "epoch": 3.99, "learning_rate": 2.7312705554615204e-07, "loss": 0.0156, "step": 10134 }, { "epoch": 3.99, "learning_rate": 2.7310328980763585e-07, "loss": 0.0147, "step": 10135 }, { "epoch": 3.99, "learning_rate": 2.7307952287764707e-07, "loss": 0.0198, "step": 10136 }, { "epoch": 4.0, "learning_rate": 2.730557547565732e-07, "loss": 0.019, "step": 10137 }, { "epoch": 4.0, "learning_rate": 2.730319854448015e-07, "loss": 0.017, "step": 10138 }, { "epoch": 4.0, "learning_rate": 2.7300821494271947e-07, "loss": 0.0155, "step": 10139 }, { "epoch": 4.0, "learning_rate": 2.729844432507145e-07, "loss": 0.0179, "step": 10140 }, { "epoch": 4.0, "learning_rate": 2.72960670369174e-07, "loss": 0.0178, "step": 10141 }, { "epoch": 4.0, "learning_rate": 2.729368962984855e-07, "loss": 0.016, "step": 10142 }, { "epoch": 4.0, "learning_rate": 2.729131210390364e-07, "loss": 0.0199, "step": 10143 }, { "epoch": 4.0, "learning_rate": 2.728893445912143e-07, "loss": 0.018, "step": 10144 }, { "epoch": 4.0, "learning_rate": 2.7286556695540657e-07, "loss": 0.0199, "step": 10145 }, { "epoch": 4.0, "learning_rate": 2.728417881320009e-07, "loss": 0.0177, "step": 10146 }, { "epoch": 4.0, "learning_rate": 2.7281800812138475e-07, "loss": 0.0151, "step": 10147 }, { "epoch": 4.0, "learning_rate": 2.727942269239458e-07, "loss": 0.014, "step": 10148 }, { "epoch": 4.0, "learning_rate": 2.7277044454007154e-07, "loss": 0.0155, "step": 10149 }, { "epoch": 4.0, "learning_rate": 2.727466609701497e-07, "loss": 0.0169, "step": 10150 }, { "epoch": 4.0, "learning_rate": 2.7272287621456777e-07, "loss": 0.0146, "step": 10151 }, { "epoch": 4.0, "learning_rate": 2.7269909027371354e-07, "loss": 0.0149, "step": 10152 }, { "epoch": 4.0, "learning_rate": 2.726753031479746e-07, "loss": 0.0151, "step": 10153 }, { "epoch": 4.0, "learning_rate": 2.7265151483773865e-07, "loss": 0.0153, "step": 10154 }, { "epoch": 4.0, "learning_rate": 2.726277253433935e-07, "loss": 0.0183, "step": 10155 }, { "epoch": 4.0, "learning_rate": 2.726039346653267e-07, "loss": 0.0147, "step": 10156 }, { "epoch": 4.0, "learning_rate": 2.7258014280392623e-07, "loss": 0.0143, "step": 10157 }, { "epoch": 4.0, "learning_rate": 2.7255634975957965e-07, "loss": 0.0141, "step": 10158 }, { "epoch": 4.0, "learning_rate": 2.7253255553267485e-07, "loss": 0.0166, "step": 10159 }, { "epoch": 4.0, "learning_rate": 2.7250876012359973e-07, "loss": 0.0166, "step": 10160 }, { "epoch": 4.0, "learning_rate": 2.7248496353274187e-07, "loss": 0.0149, "step": 10161 }, { "epoch": 4.0, "learning_rate": 2.724611657604894e-07, "loss": 0.0164, "step": 10162 }, { "epoch": 4.01, "learning_rate": 2.7243736680722997e-07, "loss": 0.0173, "step": 10163 }, { "epoch": 4.01, "learning_rate": 2.724135666733516e-07, "loss": 0.0146, "step": 10164 }, { "epoch": 4.01, "learning_rate": 2.723897653592421e-07, "loss": 0.0154, "step": 10165 }, { "epoch": 4.01, "learning_rate": 2.7236596286528943e-07, "loss": 0.0164, "step": 10166 }, { "epoch": 4.01, "learning_rate": 2.723421591918816e-07, "loss": 0.0168, "step": 10167 }, { "epoch": 4.01, "learning_rate": 2.7231835433940653e-07, "loss": 0.0139, "step": 10168 }, { "epoch": 4.01, "learning_rate": 2.7229454830825217e-07, "loss": 0.0132, "step": 10169 }, { "epoch": 4.01, "learning_rate": 2.722707410988066e-07, "loss": 0.0175, "step": 10170 }, { "epoch": 4.01, "learning_rate": 2.722469327114577e-07, "loss": 0.0152, "step": 10171 }, { "epoch": 4.01, "learning_rate": 2.722231231465937e-07, "loss": 0.0169, "step": 10172 }, { "epoch": 4.01, "learning_rate": 2.721993124046025e-07, "loss": 0.0181, "step": 10173 }, { "epoch": 4.01, "learning_rate": 2.721755004858723e-07, "loss": 0.0131, "step": 10174 }, { "epoch": 4.01, "learning_rate": 2.7215168739079103e-07, "loss": 0.0148, "step": 10175 }, { "epoch": 4.01, "learning_rate": 2.72127873119747e-07, "loss": 0.0159, "step": 10176 }, { "epoch": 4.01, "learning_rate": 2.7210405767312825e-07, "loss": 0.0141, "step": 10177 }, { "epoch": 4.01, "learning_rate": 2.7208024105132285e-07, "loss": 0.0126, "step": 10178 }, { "epoch": 4.01, "learning_rate": 2.720564232547192e-07, "loss": 0.0136, "step": 10179 }, { "epoch": 4.01, "learning_rate": 2.7203260428370527e-07, "loss": 0.0157, "step": 10180 }, { "epoch": 4.01, "learning_rate": 2.720087841386694e-07, "loss": 0.0139, "step": 10181 }, { "epoch": 4.01, "learning_rate": 2.719849628199998e-07, "loss": 0.0206, "step": 10182 }, { "epoch": 4.01, "learning_rate": 2.7196114032808473e-07, "loss": 0.0134, "step": 10183 }, { "epoch": 4.01, "learning_rate": 2.719373166633124e-07, "loss": 0.0147, "step": 10184 }, { "epoch": 4.01, "learning_rate": 2.719134918260711e-07, "loss": 0.0178, "step": 10185 }, { "epoch": 4.01, "learning_rate": 2.718896658167493e-07, "loss": 0.0173, "step": 10186 }, { "epoch": 4.01, "learning_rate": 2.718658386357351e-07, "loss": 0.0145, "step": 10187 }, { "epoch": 4.02, "learning_rate": 2.71842010283417e-07, "loss": 0.0181, "step": 10188 }, { "epoch": 4.02, "learning_rate": 2.718181807601833e-07, "loss": 0.0144, "step": 10189 }, { "epoch": 4.02, "learning_rate": 2.7179435006642245e-07, "loss": 0.0147, "step": 10190 }, { "epoch": 4.02, "learning_rate": 2.7177051820252275e-07, "loss": 0.0159, "step": 10191 }, { "epoch": 4.02, "learning_rate": 2.717466851688727e-07, "loss": 0.0125, "step": 10192 }, { "epoch": 4.02, "learning_rate": 2.717228509658608e-07, "loss": 0.014, "step": 10193 }, { "epoch": 4.02, "learning_rate": 2.716990155938753e-07, "loss": 0.0166, "step": 10194 }, { "epoch": 4.02, "learning_rate": 2.716751790533049e-07, "loss": 0.0155, "step": 10195 }, { "epoch": 4.02, "learning_rate": 2.71651341344538e-07, "loss": 0.0153, "step": 10196 }, { "epoch": 4.02, "learning_rate": 2.716275024679632e-07, "loss": 0.0159, "step": 10197 }, { "epoch": 4.02, "learning_rate": 2.7160366242396886e-07, "loss": 0.0142, "step": 10198 }, { "epoch": 4.02, "learning_rate": 2.715798212129437e-07, "loss": 0.0149, "step": 10199 }, { "epoch": 4.02, "learning_rate": 2.7155597883527624e-07, "loss": 0.0151, "step": 10200 }, { "epoch": 4.02, "learning_rate": 2.7153213529135505e-07, "loss": 0.0168, "step": 10201 }, { "epoch": 4.02, "learning_rate": 2.715082905815688e-07, "loss": 0.0134, "step": 10202 }, { "epoch": 4.02, "learning_rate": 2.7148444470630607e-07, "loss": 0.0168, "step": 10203 }, { "epoch": 4.02, "learning_rate": 2.714605976659556e-07, "loss": 0.0162, "step": 10204 }, { "epoch": 4.02, "learning_rate": 2.714367494609059e-07, "loss": 0.0185, "step": 10205 }, { "epoch": 4.02, "learning_rate": 2.7141290009154584e-07, "loss": 0.0144, "step": 10206 }, { "epoch": 4.02, "learning_rate": 2.71389049558264e-07, "loss": 0.0154, "step": 10207 }, { "epoch": 4.02, "learning_rate": 2.7136519786144915e-07, "loss": 0.0168, "step": 10208 }, { "epoch": 4.02, "learning_rate": 2.7134134500149e-07, "loss": 0.0155, "step": 10209 }, { "epoch": 4.02, "learning_rate": 2.7131749097877545e-07, "loss": 0.0177, "step": 10210 }, { "epoch": 4.02, "learning_rate": 2.712936357936941e-07, "loss": 0.0164, "step": 10211 }, { "epoch": 4.02, "learning_rate": 2.712697794466349e-07, "loss": 0.0176, "step": 10212 }, { "epoch": 4.03, "learning_rate": 2.7124592193798663e-07, "loss": 0.0187, "step": 10213 }, { "epoch": 4.03, "learning_rate": 2.71222063268138e-07, "loss": 0.0157, "step": 10214 }, { "epoch": 4.03, "learning_rate": 2.711982034374781e-07, "loss": 0.0151, "step": 10215 }, { "epoch": 4.03, "learning_rate": 2.7117434244639567e-07, "loss": 0.0214, "step": 10216 }, { "epoch": 4.03, "learning_rate": 2.7115048029527953e-07, "loss": 0.0134, "step": 10217 }, { "epoch": 4.03, "learning_rate": 2.711266169845188e-07, "loss": 0.0161, "step": 10218 }, { "epoch": 4.03, "learning_rate": 2.711027525145023e-07, "loss": 0.0161, "step": 10219 }, { "epoch": 4.03, "learning_rate": 2.7107888688561906e-07, "loss": 0.0124, "step": 10220 }, { "epoch": 4.03, "learning_rate": 2.7105502009825787e-07, "loss": 0.0162, "step": 10221 }, { "epoch": 4.03, "learning_rate": 2.7103115215280793e-07, "loss": 0.0206, "step": 10222 }, { "epoch": 4.03, "learning_rate": 2.710072830496581e-07, "loss": 0.0172, "step": 10223 }, { "epoch": 4.03, "learning_rate": 2.709834127891975e-07, "loss": 0.0141, "step": 10224 }, { "epoch": 4.03, "learning_rate": 2.7095954137181513e-07, "loss": 0.0165, "step": 10225 }, { "epoch": 4.03, "learning_rate": 2.7093566879790014e-07, "loss": 0.0143, "step": 10226 }, { "epoch": 4.03, "learning_rate": 2.709117950678415e-07, "loss": 0.0155, "step": 10227 }, { "epoch": 4.03, "learning_rate": 2.7088792018202837e-07, "loss": 0.0141, "step": 10228 }, { "epoch": 4.03, "learning_rate": 2.708640441408499e-07, "loss": 0.016, "step": 10229 }, { "epoch": 4.03, "learning_rate": 2.7084016694469524e-07, "loss": 0.0133, "step": 10230 }, { "epoch": 4.03, "learning_rate": 2.708162885939535e-07, "loss": 0.0164, "step": 10231 }, { "epoch": 4.03, "learning_rate": 2.707924090890139e-07, "loss": 0.0162, "step": 10232 }, { "epoch": 4.03, "learning_rate": 2.707685284302656e-07, "loss": 0.014, "step": 10233 }, { "epoch": 4.03, "learning_rate": 2.7074464661809785e-07, "loss": 0.0142, "step": 10234 }, { "epoch": 4.03, "learning_rate": 2.707207636528999e-07, "loss": 0.0142, "step": 10235 }, { "epoch": 4.03, "learning_rate": 2.7069687953506094e-07, "loss": 0.0139, "step": 10236 }, { "epoch": 4.03, "learning_rate": 2.706729942649703e-07, "loss": 0.0157, "step": 10237 }, { "epoch": 4.03, "learning_rate": 2.706491078430173e-07, "loss": 0.015, "step": 10238 }, { "epoch": 4.04, "learning_rate": 2.706252202695912e-07, "loss": 0.0141, "step": 10239 }, { "epoch": 4.04, "learning_rate": 2.706013315450814e-07, "loss": 0.0131, "step": 10240 }, { "epoch": 4.04, "learning_rate": 2.7057744166987713e-07, "loss": 0.0141, "step": 10241 }, { "epoch": 4.04, "learning_rate": 2.705535506443679e-07, "loss": 0.018, "step": 10242 }, { "epoch": 4.04, "learning_rate": 2.705296584689429e-07, "loss": 0.0171, "step": 10243 }, { "epoch": 4.04, "learning_rate": 2.7050576514399173e-07, "loss": 0.0166, "step": 10244 }, { "epoch": 4.04, "learning_rate": 2.704818706699037e-07, "loss": 0.0126, "step": 10245 }, { "epoch": 4.04, "learning_rate": 2.7045797504706836e-07, "loss": 0.0151, "step": 10246 }, { "epoch": 4.04, "learning_rate": 2.704340782758751e-07, "loss": 0.0162, "step": 10247 }, { "epoch": 4.04, "learning_rate": 2.7041018035671344e-07, "loss": 0.0131, "step": 10248 }, { "epoch": 4.04, "learning_rate": 2.703862812899727e-07, "loss": 0.0143, "step": 10249 }, { "epoch": 4.04, "learning_rate": 2.703623810760427e-07, "loss": 0.0156, "step": 10250 }, { "epoch": 4.04, "learning_rate": 2.703384797153128e-07, "loss": 0.0178, "step": 10251 }, { "epoch": 4.04, "learning_rate": 2.7031457720817256e-07, "loss": 0.0127, "step": 10252 }, { "epoch": 4.04, "learning_rate": 2.7029067355501164e-07, "loss": 0.012, "step": 10253 }, { "epoch": 4.04, "learning_rate": 2.7026676875621945e-07, "loss": 0.013, "step": 10254 }, { "epoch": 4.04, "learning_rate": 2.702428628121858e-07, "loss": 0.0161, "step": 10255 }, { "epoch": 4.04, "learning_rate": 2.7021895572330024e-07, "loss": 0.0126, "step": 10256 }, { "epoch": 4.04, "learning_rate": 2.701950474899524e-07, "loss": 0.0099, "step": 10257 }, { "epoch": 4.04, "learning_rate": 2.70171138112532e-07, "loss": 0.016, "step": 10258 }, { "epoch": 4.04, "learning_rate": 2.7014722759142865e-07, "loss": 0.0167, "step": 10259 }, { "epoch": 4.04, "learning_rate": 2.701233159270322e-07, "loss": 0.0157, "step": 10260 }, { "epoch": 4.04, "learning_rate": 2.700994031197322e-07, "loss": 0.0143, "step": 10261 }, { "epoch": 4.04, "learning_rate": 2.7007548916991845e-07, "loss": 0.0165, "step": 10262 }, { "epoch": 4.04, "learning_rate": 2.7005157407798077e-07, "loss": 0.0161, "step": 10263 }, { "epoch": 4.05, "learning_rate": 2.7002765784430896e-07, "loss": 0.0149, "step": 10264 }, { "epoch": 4.05, "learning_rate": 2.700037404692926e-07, "loss": 0.0138, "step": 10265 }, { "epoch": 4.05, "learning_rate": 2.6997982195332174e-07, "loss": 0.0129, "step": 10266 }, { "epoch": 4.05, "learning_rate": 2.6995590229678613e-07, "loss": 0.0141, "step": 10267 }, { "epoch": 4.05, "learning_rate": 2.699319815000757e-07, "loss": 0.0155, "step": 10268 }, { "epoch": 4.05, "learning_rate": 2.6990805956358016e-07, "loss": 0.0172, "step": 10269 }, { "epoch": 4.05, "learning_rate": 2.6988413648768954e-07, "loss": 0.0141, "step": 10270 }, { "epoch": 4.05, "learning_rate": 2.698602122727938e-07, "loss": 0.0133, "step": 10271 }, { "epoch": 4.05, "learning_rate": 2.6983628691928265e-07, "loss": 0.0133, "step": 10272 }, { "epoch": 4.05, "learning_rate": 2.6981236042754617e-07, "loss": 0.0159, "step": 10273 }, { "epoch": 4.05, "learning_rate": 2.6978843279797435e-07, "loss": 0.0163, "step": 10274 }, { "epoch": 4.05, "learning_rate": 2.6976450403095713e-07, "loss": 0.0167, "step": 10275 }, { "epoch": 4.05, "learning_rate": 2.697405741268845e-07, "loss": 0.0151, "step": 10276 }, { "epoch": 4.05, "learning_rate": 2.6971664308614654e-07, "loss": 0.0137, "step": 10277 }, { "epoch": 4.05, "learning_rate": 2.6969271090913323e-07, "loss": 0.0166, "step": 10278 }, { "epoch": 4.05, "learning_rate": 2.696687775962347e-07, "loss": 0.018, "step": 10279 }, { "epoch": 4.05, "learning_rate": 2.6964484314784094e-07, "loss": 0.0148, "step": 10280 }, { "epoch": 4.05, "learning_rate": 2.6962090756434206e-07, "loss": 0.0191, "step": 10281 }, { "epoch": 4.05, "learning_rate": 2.6959697084612826e-07, "loss": 0.0146, "step": 10282 }, { "epoch": 4.05, "learning_rate": 2.6957303299358964e-07, "loss": 0.015, "step": 10283 }, { "epoch": 4.05, "learning_rate": 2.6954909400711627e-07, "loss": 0.0115, "step": 10284 }, { "epoch": 4.05, "learning_rate": 2.6952515388709835e-07, "loss": 0.0176, "step": 10285 }, { "epoch": 4.05, "learning_rate": 2.695012126339261e-07, "loss": 0.0132, "step": 10286 }, { "epoch": 4.05, "learning_rate": 2.694772702479897e-07, "loss": 0.0127, "step": 10287 }, { "epoch": 4.05, "learning_rate": 2.694533267296794e-07, "loss": 0.0186, "step": 10288 }, { "epoch": 4.05, "learning_rate": 2.694293820793854e-07, "loss": 0.0129, "step": 10289 }, { "epoch": 4.06, "learning_rate": 2.694054362974981e-07, "loss": 0.0139, "step": 10290 }, { "epoch": 4.06, "learning_rate": 2.6938148938440763e-07, "loss": 0.0146, "step": 10291 }, { "epoch": 4.06, "learning_rate": 2.693575413405043e-07, "loss": 0.0181, "step": 10292 }, { "epoch": 4.06, "learning_rate": 2.6933359216617844e-07, "loss": 0.0168, "step": 10293 }, { "epoch": 4.06, "learning_rate": 2.693096418618204e-07, "loss": 0.0131, "step": 10294 }, { "epoch": 4.06, "learning_rate": 2.6928569042782055e-07, "loss": 0.0171, "step": 10295 }, { "epoch": 4.06, "learning_rate": 2.6926173786456925e-07, "loss": 0.019, "step": 10296 }, { "epoch": 4.06, "learning_rate": 2.692377841724569e-07, "loss": 0.0195, "step": 10297 }, { "epoch": 4.06, "learning_rate": 2.692138293518739e-07, "loss": 0.0209, "step": 10298 }, { "epoch": 4.06, "learning_rate": 2.6918987340321066e-07, "loss": 0.0177, "step": 10299 }, { "epoch": 4.06, "learning_rate": 2.6916591632685766e-07, "loss": 0.0146, "step": 10300 }, { "epoch": 4.06, "learning_rate": 2.6914195812320533e-07, "loss": 0.0169, "step": 10301 }, { "epoch": 4.06, "learning_rate": 2.691179987926442e-07, "loss": 0.0131, "step": 10302 }, { "epoch": 4.06, "learning_rate": 2.6909403833556467e-07, "loss": 0.0141, "step": 10303 }, { "epoch": 4.06, "learning_rate": 2.6907007675235735e-07, "loss": 0.0173, "step": 10304 }, { "epoch": 4.06, "learning_rate": 2.690461140434128e-07, "loss": 0.019, "step": 10305 }, { "epoch": 4.06, "learning_rate": 2.690221502091215e-07, "loss": 0.0229, "step": 10306 }, { "epoch": 4.06, "learning_rate": 2.6899818524987405e-07, "loss": 0.0145, "step": 10307 }, { "epoch": 4.06, "learning_rate": 2.6897421916606103e-07, "loss": 0.018, "step": 10308 }, { "epoch": 4.06, "learning_rate": 2.6895025195807303e-07, "loss": 0.0173, "step": 10309 }, { "epoch": 4.06, "learning_rate": 2.689262836263008e-07, "loss": 0.0132, "step": 10310 }, { "epoch": 4.06, "learning_rate": 2.689023141711349e-07, "loss": 0.0131, "step": 10311 }, { "epoch": 4.06, "learning_rate": 2.688783435929659e-07, "loss": 0.0179, "step": 10312 }, { "epoch": 4.06, "learning_rate": 2.688543718921847e-07, "loss": 0.0141, "step": 10313 }, { "epoch": 4.06, "learning_rate": 2.6883039906918176e-07, "loss": 0.0197, "step": 10314 }, { "epoch": 4.07, "learning_rate": 2.68806425124348e-07, "loss": 0.0136, "step": 10315 }, { "epoch": 4.07, "learning_rate": 2.687824500580741e-07, "loss": 0.0124, "step": 10316 }, { "epoch": 4.07, "learning_rate": 2.6875847387075077e-07, "loss": 0.0127, "step": 10317 }, { "epoch": 4.07, "learning_rate": 2.6873449656276885e-07, "loss": 0.0175, "step": 10318 }, { "epoch": 4.07, "learning_rate": 2.6871051813451905e-07, "loss": 0.0165, "step": 10319 }, { "epoch": 4.07, "learning_rate": 2.686865385863923e-07, "loss": 0.0147, "step": 10320 }, { "epoch": 4.07, "learning_rate": 2.686625579187793e-07, "loss": 0.0168, "step": 10321 }, { "epoch": 4.07, "learning_rate": 2.68638576132071e-07, "loss": 0.018, "step": 10322 }, { "epoch": 4.07, "learning_rate": 2.686145932266582e-07, "loss": 0.0133, "step": 10323 }, { "epoch": 4.07, "learning_rate": 2.685906092029318e-07, "loss": 0.0193, "step": 10324 }, { "epoch": 4.07, "learning_rate": 2.6856662406128274e-07, "loss": 0.0186, "step": 10325 }, { "epoch": 4.07, "learning_rate": 2.6854263780210196e-07, "loss": 0.0163, "step": 10326 }, { "epoch": 4.07, "learning_rate": 2.685186504257803e-07, "loss": 0.0154, "step": 10327 }, { "epoch": 4.07, "learning_rate": 2.684946619327088e-07, "loss": 0.0158, "step": 10328 }, { "epoch": 4.07, "learning_rate": 2.684706723232784e-07, "loss": 0.0166, "step": 10329 }, { "epoch": 4.07, "learning_rate": 2.6844668159788014e-07, "loss": 0.0168, "step": 10330 }, { "epoch": 4.07, "learning_rate": 2.6842268975690497e-07, "loss": 0.0149, "step": 10331 }, { "epoch": 4.07, "learning_rate": 2.683986968007439e-07, "loss": 0.0155, "step": 10332 }, { "epoch": 4.07, "learning_rate": 2.6837470272978805e-07, "loss": 0.0144, "step": 10333 }, { "epoch": 4.07, "learning_rate": 2.683507075444285e-07, "loss": 0.0183, "step": 10334 }, { "epoch": 4.07, "learning_rate": 2.6832671124505627e-07, "loss": 0.0166, "step": 10335 }, { "epoch": 4.07, "learning_rate": 2.683027138320625e-07, "loss": 0.0164, "step": 10336 }, { "epoch": 4.07, "learning_rate": 2.682787153058383e-07, "loss": 0.0177, "step": 10337 }, { "epoch": 4.07, "learning_rate": 2.682547156667748e-07, "loss": 0.0168, "step": 10338 }, { "epoch": 4.07, "learning_rate": 2.682307149152632e-07, "loss": 0.0134, "step": 10339 }, { "epoch": 4.08, "learning_rate": 2.6820671305169466e-07, "loss": 0.0166, "step": 10340 }, { "epoch": 4.08, "learning_rate": 2.681827100764603e-07, "loss": 0.0152, "step": 10341 }, { "epoch": 4.08, "learning_rate": 2.681587059899514e-07, "loss": 0.0167, "step": 10342 }, { "epoch": 4.08, "learning_rate": 2.6813470079255916e-07, "loss": 0.0168, "step": 10343 }, { "epoch": 4.08, "learning_rate": 2.681106944846749e-07, "loss": 0.0162, "step": 10344 }, { "epoch": 4.08, "learning_rate": 2.6808668706668983e-07, "loss": 0.0138, "step": 10345 }, { "epoch": 4.08, "learning_rate": 2.680626785389953e-07, "loss": 0.0157, "step": 10346 }, { "epoch": 4.08, "learning_rate": 2.6803866890198246e-07, "loss": 0.0134, "step": 10347 }, { "epoch": 4.08, "learning_rate": 2.680146581560428e-07, "loss": 0.0142, "step": 10348 }, { "epoch": 4.08, "learning_rate": 2.6799064630156747e-07, "loss": 0.0168, "step": 10349 }, { "epoch": 4.08, "learning_rate": 2.6796663333894807e-07, "loss": 0.0153, "step": 10350 }, { "epoch": 4.08, "learning_rate": 2.6794261926857575e-07, "loss": 0.0152, "step": 10351 }, { "epoch": 4.08, "learning_rate": 2.67918604090842e-07, "loss": 0.0156, "step": 10352 }, { "epoch": 4.08, "learning_rate": 2.678945878061383e-07, "loss": 0.0188, "step": 10353 }, { "epoch": 4.08, "learning_rate": 2.678705704148559e-07, "loss": 0.0121, "step": 10354 }, { "epoch": 4.08, "learning_rate": 2.678465519173865e-07, "loss": 0.0128, "step": 10355 }, { "epoch": 4.08, "learning_rate": 2.6782253231412127e-07, "loss": 0.0139, "step": 10356 }, { "epoch": 4.08, "learning_rate": 2.6779851160545195e-07, "loss": 0.0155, "step": 10357 }, { "epoch": 4.08, "learning_rate": 2.677744897917698e-07, "loss": 0.0171, "step": 10358 }, { "epoch": 4.08, "learning_rate": 2.677504668734666e-07, "loss": 0.0118, "step": 10359 }, { "epoch": 4.08, "learning_rate": 2.6772644285093373e-07, "loss": 0.0151, "step": 10360 }, { "epoch": 4.08, "learning_rate": 2.6770241772456276e-07, "loss": 0.0128, "step": 10361 }, { "epoch": 4.08, "learning_rate": 2.6767839149474534e-07, "loss": 0.0189, "step": 10362 }, { "epoch": 4.08, "learning_rate": 2.6765436416187286e-07, "loss": 0.016, "step": 10363 }, { "epoch": 4.08, "learning_rate": 2.676303357263372e-07, "loss": 0.015, "step": 10364 }, { "epoch": 4.08, "learning_rate": 2.6760630618852973e-07, "loss": 0.0139, "step": 10365 }, { "epoch": 4.09, "learning_rate": 2.675822755488423e-07, "loss": 0.017, "step": 10366 }, { "epoch": 4.09, "learning_rate": 2.6755824380766643e-07, "loss": 0.0157, "step": 10367 }, { "epoch": 4.09, "learning_rate": 2.67534210965394e-07, "loss": 0.0164, "step": 10368 }, { "epoch": 4.09, "learning_rate": 2.6751017702241646e-07, "loss": 0.0173, "step": 10369 }, { "epoch": 4.09, "learning_rate": 2.6748614197912567e-07, "loss": 0.0172, "step": 10370 }, { "epoch": 4.09, "learning_rate": 2.674621058359133e-07, "loss": 0.0149, "step": 10371 }, { "epoch": 4.09, "learning_rate": 2.674380685931712e-07, "loss": 0.0147, "step": 10372 }, { "epoch": 4.09, "learning_rate": 2.67414030251291e-07, "loss": 0.016, "step": 10373 }, { "epoch": 4.09, "learning_rate": 2.673899908106646e-07, "loss": 0.015, "step": 10374 }, { "epoch": 4.09, "learning_rate": 2.673659502716838e-07, "loss": 0.0159, "step": 10375 }, { "epoch": 4.09, "learning_rate": 2.6734190863474033e-07, "loss": 0.0171, "step": 10376 }, { "epoch": 4.09, "learning_rate": 2.6731786590022616e-07, "loss": 0.0174, "step": 10377 }, { "epoch": 4.09, "learning_rate": 2.67293822068533e-07, "loss": 0.0166, "step": 10378 }, { "epoch": 4.09, "learning_rate": 2.6726977714005295e-07, "loss": 0.0148, "step": 10379 }, { "epoch": 4.09, "learning_rate": 2.672457311151777e-07, "loss": 0.0121, "step": 10380 }, { "epoch": 4.09, "learning_rate": 2.672216839942992e-07, "loss": 0.017, "step": 10381 }, { "epoch": 4.09, "learning_rate": 2.6719763577780945e-07, "loss": 0.0155, "step": 10382 }, { "epoch": 4.09, "learning_rate": 2.6717358646610033e-07, "loss": 0.0141, "step": 10383 }, { "epoch": 4.09, "learning_rate": 2.6714953605956385e-07, "loss": 0.016, "step": 10384 }, { "epoch": 4.09, "learning_rate": 2.6712548455859204e-07, "loss": 0.0145, "step": 10385 }, { "epoch": 4.09, "learning_rate": 2.6710143196357685e-07, "loss": 0.0136, "step": 10386 }, { "epoch": 4.09, "learning_rate": 2.6707737827491026e-07, "loss": 0.0113, "step": 10387 }, { "epoch": 4.09, "learning_rate": 2.670533234929844e-07, "loss": 0.0122, "step": 10388 }, { "epoch": 4.09, "learning_rate": 2.670292676181912e-07, "loss": 0.0155, "step": 10389 }, { "epoch": 4.09, "learning_rate": 2.6700521065092285e-07, "loss": 0.0168, "step": 10390 }, { "epoch": 4.1, "learning_rate": 2.6698115259157143e-07, "loss": 0.016, "step": 10391 }, { "epoch": 4.1, "learning_rate": 2.66957093440529e-07, "loss": 0.0145, "step": 10392 }, { "epoch": 4.1, "learning_rate": 2.6693303319818767e-07, "loss": 0.0132, "step": 10393 }, { "epoch": 4.1, "learning_rate": 2.6690897186493967e-07, "loss": 0.0156, "step": 10394 }, { "epoch": 4.1, "learning_rate": 2.6688490944117715e-07, "loss": 0.0152, "step": 10395 }, { "epoch": 4.1, "learning_rate": 2.6686084592729217e-07, "loss": 0.0175, "step": 10396 }, { "epoch": 4.1, "learning_rate": 2.668367813236771e-07, "loss": 0.0128, "step": 10397 }, { "epoch": 4.1, "learning_rate": 2.66812715630724e-07, "loss": 0.0159, "step": 10398 }, { "epoch": 4.1, "learning_rate": 2.6678864884882526e-07, "loss": 0.0182, "step": 10399 }, { "epoch": 4.1, "learning_rate": 2.66764580978373e-07, "loss": 0.0151, "step": 10400 }, { "epoch": 4.1, "learning_rate": 2.6674051201975955e-07, "loss": 0.0161, "step": 10401 }, { "epoch": 4.1, "learning_rate": 2.6671644197337723e-07, "loss": 0.0184, "step": 10402 }, { "epoch": 4.1, "learning_rate": 2.666923708396182e-07, "loss": 0.0164, "step": 10403 }, { "epoch": 4.1, "learning_rate": 2.66668298618875e-07, "loss": 0.0155, "step": 10404 }, { "epoch": 4.1, "learning_rate": 2.6664422531153976e-07, "loss": 0.0161, "step": 10405 }, { "epoch": 4.1, "learning_rate": 2.66620150918005e-07, "loss": 0.0166, "step": 10406 }, { "epoch": 4.1, "learning_rate": 2.66596075438663e-07, "loss": 0.0151, "step": 10407 }, { "epoch": 4.1, "learning_rate": 2.665719988739062e-07, "loss": 0.0168, "step": 10408 }, { "epoch": 4.1, "learning_rate": 2.6654792122412697e-07, "loss": 0.0127, "step": 10409 }, { "epoch": 4.1, "learning_rate": 2.665238424897178e-07, "loss": 0.0165, "step": 10410 }, { "epoch": 4.1, "learning_rate": 2.664997626710711e-07, "loss": 0.0187, "step": 10411 }, { "epoch": 4.1, "learning_rate": 2.6647568176857937e-07, "loss": 0.0177, "step": 10412 }, { "epoch": 4.1, "learning_rate": 2.6645159978263503e-07, "loss": 0.0158, "step": 10413 }, { "epoch": 4.1, "learning_rate": 2.664275167136305e-07, "loss": 0.0158, "step": 10414 }, { "epoch": 4.1, "learning_rate": 2.6640343256195857e-07, "loss": 0.0125, "step": 10415 }, { "epoch": 4.11, "learning_rate": 2.663793473280115e-07, "loss": 0.0148, "step": 10416 }, { "epoch": 4.11, "learning_rate": 2.66355261012182e-07, "loss": 0.0143, "step": 10417 }, { "epoch": 4.11, "learning_rate": 2.663311736148626e-07, "loss": 0.0139, "step": 10418 }, { "epoch": 4.11, "learning_rate": 2.663070851364459e-07, "loss": 0.0217, "step": 10419 }, { "epoch": 4.11, "learning_rate": 2.6628299557732446e-07, "loss": 0.015, "step": 10420 }, { "epoch": 4.11, "learning_rate": 2.662589049378909e-07, "loss": 0.0179, "step": 10421 }, { "epoch": 4.11, "learning_rate": 2.6623481321853795e-07, "loss": 0.0147, "step": 10422 }, { "epoch": 4.11, "learning_rate": 2.6621072041965815e-07, "loss": 0.017, "step": 10423 }, { "epoch": 4.11, "learning_rate": 2.661866265416443e-07, "loss": 0.0117, "step": 10424 }, { "epoch": 4.11, "learning_rate": 2.66162531584889e-07, "loss": 0.0173, "step": 10425 }, { "epoch": 4.11, "learning_rate": 2.6613843554978496e-07, "loss": 0.0152, "step": 10426 }, { "epoch": 4.11, "learning_rate": 2.661143384367249e-07, "loss": 0.0156, "step": 10427 }, { "epoch": 4.11, "learning_rate": 2.660902402461017e-07, "loss": 0.0183, "step": 10428 }, { "epoch": 4.11, "learning_rate": 2.66066140978308e-07, "loss": 0.0151, "step": 10429 }, { "epoch": 4.11, "learning_rate": 2.660420406337366e-07, "loss": 0.0133, "step": 10430 }, { "epoch": 4.11, "learning_rate": 2.660179392127803e-07, "loss": 0.017, "step": 10431 }, { "epoch": 4.11, "learning_rate": 2.6599383671583194e-07, "loss": 0.0182, "step": 10432 }, { "epoch": 4.11, "learning_rate": 2.659697331432844e-07, "loss": 0.0153, "step": 10433 }, { "epoch": 4.11, "learning_rate": 2.659456284955304e-07, "loss": 0.016, "step": 10434 }, { "epoch": 4.11, "learning_rate": 2.659215227729629e-07, "loss": 0.0156, "step": 10435 }, { "epoch": 4.11, "learning_rate": 2.658974159759748e-07, "loss": 0.0148, "step": 10436 }, { "epoch": 4.11, "learning_rate": 2.6587330810495894e-07, "loss": 0.0178, "step": 10437 }, { "epoch": 4.11, "learning_rate": 2.6584919916030826e-07, "loss": 0.0129, "step": 10438 }, { "epoch": 4.11, "learning_rate": 2.6582508914241574e-07, "loss": 0.018, "step": 10439 }, { "epoch": 4.11, "learning_rate": 2.658009780516743e-07, "loss": 0.0147, "step": 10440 }, { "epoch": 4.11, "learning_rate": 2.6577686588847694e-07, "loss": 0.0174, "step": 10441 }, { "epoch": 4.12, "learning_rate": 2.657527526532166e-07, "loss": 0.0159, "step": 10442 }, { "epoch": 4.12, "learning_rate": 2.657286383462863e-07, "loss": 0.0173, "step": 10443 }, { "epoch": 4.12, "learning_rate": 2.657045229680792e-07, "loss": 0.0167, "step": 10444 }, { "epoch": 4.12, "learning_rate": 2.656804065189881e-07, "loss": 0.0161, "step": 10445 }, { "epoch": 4.12, "learning_rate": 2.656562889994063e-07, "loss": 0.0155, "step": 10446 }, { "epoch": 4.12, "learning_rate": 2.6563217040972674e-07, "loss": 0.0168, "step": 10447 }, { "epoch": 4.12, "learning_rate": 2.656080507503425e-07, "loss": 0.0154, "step": 10448 }, { "epoch": 4.12, "learning_rate": 2.6558393002164676e-07, "loss": 0.0179, "step": 10449 }, { "epoch": 4.12, "learning_rate": 2.6555980822403275e-07, "loss": 0.0129, "step": 10450 }, { "epoch": 4.12, "learning_rate": 2.6553568535789344e-07, "loss": 0.0141, "step": 10451 }, { "epoch": 4.12, "learning_rate": 2.6551156142362203e-07, "loss": 0.015, "step": 10452 }, { "epoch": 4.12, "learning_rate": 2.6548743642161175e-07, "loss": 0.0139, "step": 10453 }, { "epoch": 4.12, "learning_rate": 2.654633103522558e-07, "loss": 0.0148, "step": 10454 }, { "epoch": 4.12, "learning_rate": 2.6543918321594737e-07, "loss": 0.0171, "step": 10455 }, { "epoch": 4.12, "learning_rate": 2.6541505501307974e-07, "loss": 0.0145, "step": 10456 }, { "epoch": 4.12, "learning_rate": 2.653909257440461e-07, "loss": 0.0152, "step": 10457 }, { "epoch": 4.12, "learning_rate": 2.6536679540923975e-07, "loss": 0.015, "step": 10458 }, { "epoch": 4.12, "learning_rate": 2.65342664009054e-07, "loss": 0.0169, "step": 10459 }, { "epoch": 4.12, "learning_rate": 2.6531853154388215e-07, "loss": 0.016, "step": 10460 }, { "epoch": 4.12, "learning_rate": 2.652943980141175e-07, "loss": 0.0159, "step": 10461 }, { "epoch": 4.12, "learning_rate": 2.652702634201534e-07, "loss": 0.0141, "step": 10462 }, { "epoch": 4.12, "learning_rate": 2.652461277623831e-07, "loss": 0.0172, "step": 10463 }, { "epoch": 4.12, "learning_rate": 2.652219910412002e-07, "loss": 0.0163, "step": 10464 }, { "epoch": 4.12, "learning_rate": 2.651978532569979e-07, "loss": 0.0158, "step": 10465 }, { "epoch": 4.12, "learning_rate": 2.6517371441016973e-07, "loss": 0.0169, "step": 10466 }, { "epoch": 4.13, "learning_rate": 2.6514957450110904e-07, "loss": 0.0161, "step": 10467 }, { "epoch": 4.13, "learning_rate": 2.651254335302094e-07, "loss": 0.0201, "step": 10468 }, { "epoch": 4.13, "learning_rate": 2.6510129149786403e-07, "loss": 0.015, "step": 10469 }, { "epoch": 4.13, "learning_rate": 2.650771484044666e-07, "loss": 0.0166, "step": 10470 }, { "epoch": 4.13, "learning_rate": 2.650530042504106e-07, "loss": 0.0161, "step": 10471 }, { "epoch": 4.13, "learning_rate": 2.650288590360895e-07, "loss": 0.016, "step": 10472 }, { "epoch": 4.13, "learning_rate": 2.650047127618968e-07, "loss": 0.0132, "step": 10473 }, { "epoch": 4.13, "learning_rate": 2.6498056542822605e-07, "loss": 0.0169, "step": 10474 }, { "epoch": 4.13, "learning_rate": 2.6495641703547087e-07, "loss": 0.0156, "step": 10475 }, { "epoch": 4.13, "learning_rate": 2.6493226758402487e-07, "loss": 0.0177, "step": 10476 }, { "epoch": 4.13, "learning_rate": 2.649081170742815e-07, "loss": 0.0137, "step": 10477 }, { "epoch": 4.13, "learning_rate": 2.648839655066345e-07, "loss": 0.0163, "step": 10478 }, { "epoch": 4.13, "learning_rate": 2.6485981288147755e-07, "loss": 0.0116, "step": 10479 }, { "epoch": 4.13, "learning_rate": 2.6483565919920416e-07, "loss": 0.0141, "step": 10480 }, { "epoch": 4.13, "learning_rate": 2.6481150446020804e-07, "loss": 0.0142, "step": 10481 }, { "epoch": 4.13, "learning_rate": 2.6478734866488295e-07, "loss": 0.0142, "step": 10482 }, { "epoch": 4.13, "learning_rate": 2.647631918136225e-07, "loss": 0.0174, "step": 10483 }, { "epoch": 4.13, "learning_rate": 2.647390339068205e-07, "loss": 0.0177, "step": 10484 }, { "epoch": 4.13, "learning_rate": 2.647148749448706e-07, "loss": 0.0148, "step": 10485 }, { "epoch": 4.13, "learning_rate": 2.6469071492816655e-07, "loss": 0.0163, "step": 10486 }, { "epoch": 4.13, "learning_rate": 2.6466655385710225e-07, "loss": 0.0177, "step": 10487 }, { "epoch": 4.13, "learning_rate": 2.6464239173207137e-07, "loss": 0.0178, "step": 10488 }, { "epoch": 4.13, "learning_rate": 2.646182285534677e-07, "loss": 0.016, "step": 10489 }, { "epoch": 4.13, "learning_rate": 2.645940643216852e-07, "loss": 0.0155, "step": 10490 }, { "epoch": 4.13, "learning_rate": 2.645698990371176e-07, "loss": 0.0162, "step": 10491 }, { "epoch": 4.13, "learning_rate": 2.645457327001588e-07, "loss": 0.0127, "step": 10492 }, { "epoch": 4.14, "learning_rate": 2.645215653112026e-07, "loss": 0.015, "step": 10493 }, { "epoch": 4.14, "learning_rate": 2.6449739687064293e-07, "loss": 0.0152, "step": 10494 }, { "epoch": 4.14, "learning_rate": 2.644732273788738e-07, "loss": 0.0149, "step": 10495 }, { "epoch": 4.14, "learning_rate": 2.6444905683628895e-07, "loss": 0.0179, "step": 10496 }, { "epoch": 4.14, "learning_rate": 2.6442488524328255e-07, "loss": 0.0162, "step": 10497 }, { "epoch": 4.14, "learning_rate": 2.6440071260024836e-07, "loss": 0.0149, "step": 10498 }, { "epoch": 4.14, "learning_rate": 2.6437653890758043e-07, "loss": 0.0159, "step": 10499 }, { "epoch": 4.14, "learning_rate": 2.6435236416567277e-07, "loss": 0.0131, "step": 10500 }, { "epoch": 4.14, "learning_rate": 2.643281883749194e-07, "loss": 0.0149, "step": 10501 }, { "epoch": 4.14, "learning_rate": 2.6430401153571435e-07, "loss": 0.0165, "step": 10502 }, { "epoch": 4.14, "learning_rate": 2.642798336484516e-07, "loss": 0.0186, "step": 10503 }, { "epoch": 4.14, "learning_rate": 2.642556547135252e-07, "loss": 0.0131, "step": 10504 }, { "epoch": 4.14, "learning_rate": 2.6423147473132934e-07, "loss": 0.0176, "step": 10505 }, { "epoch": 4.14, "learning_rate": 2.6420729370225806e-07, "loss": 0.0122, "step": 10506 }, { "epoch": 4.14, "learning_rate": 2.641831116267054e-07, "loss": 0.0153, "step": 10507 }, { "epoch": 4.14, "learning_rate": 2.641589285050657e-07, "loss": 0.0139, "step": 10508 }, { "epoch": 4.14, "learning_rate": 2.641347443377329e-07, "loss": 0.0149, "step": 10509 }, { "epoch": 4.14, "learning_rate": 2.641105591251013e-07, "loss": 0.0138, "step": 10510 }, { "epoch": 4.14, "learning_rate": 2.6408637286756497e-07, "loss": 0.016, "step": 10511 }, { "epoch": 4.14, "learning_rate": 2.6406218556551814e-07, "loss": 0.015, "step": 10512 }, { "epoch": 4.14, "learning_rate": 2.6403799721935513e-07, "loss": 0.0126, "step": 10513 }, { "epoch": 4.14, "learning_rate": 2.6401380782947e-07, "loss": 0.0142, "step": 10514 }, { "epoch": 4.14, "learning_rate": 2.6398961739625716e-07, "loss": 0.0144, "step": 10515 }, { "epoch": 4.14, "learning_rate": 2.639654259201108e-07, "loss": 0.018, "step": 10516 }, { "epoch": 4.14, "learning_rate": 2.6394123340142516e-07, "loss": 0.0158, "step": 10517 }, { "epoch": 4.15, "learning_rate": 2.6391703984059465e-07, "loss": 0.0143, "step": 10518 }, { "epoch": 4.15, "learning_rate": 2.638928452380135e-07, "loss": 0.0155, "step": 10519 }, { "epoch": 4.15, "learning_rate": 2.6386864959407616e-07, "loss": 0.0134, "step": 10520 }, { "epoch": 4.15, "learning_rate": 2.638444529091768e-07, "loss": 0.0136, "step": 10521 }, { "epoch": 4.15, "learning_rate": 2.6382025518370997e-07, "loss": 0.018, "step": 10522 }, { "epoch": 4.15, "learning_rate": 2.6379605641806995e-07, "loss": 0.015, "step": 10523 }, { "epoch": 4.15, "learning_rate": 2.6377185661265114e-07, "loss": 0.0159, "step": 10524 }, { "epoch": 4.15, "learning_rate": 2.63747655767848e-07, "loss": 0.0116, "step": 10525 }, { "epoch": 4.15, "learning_rate": 2.6372345388405497e-07, "loss": 0.0143, "step": 10526 }, { "epoch": 4.15, "learning_rate": 2.6369925096166646e-07, "loss": 0.0145, "step": 10527 }, { "epoch": 4.15, "learning_rate": 2.6367504700107697e-07, "loss": 0.0176, "step": 10528 }, { "epoch": 4.15, "learning_rate": 2.63650842002681e-07, "loss": 0.0148, "step": 10529 }, { "epoch": 4.15, "learning_rate": 2.6362663596687314e-07, "loss": 0.014, "step": 10530 }, { "epoch": 4.15, "learning_rate": 2.636024288940477e-07, "loss": 0.016, "step": 10531 }, { "epoch": 4.15, "learning_rate": 2.6357822078459935e-07, "loss": 0.0151, "step": 10532 }, { "epoch": 4.15, "learning_rate": 2.635540116389226e-07, "loss": 0.0159, "step": 10533 }, { "epoch": 4.15, "learning_rate": 2.635298014574121e-07, "loss": 0.0132, "step": 10534 }, { "epoch": 4.15, "learning_rate": 2.6350559024046244e-07, "loss": 0.012, "step": 10535 }, { "epoch": 4.15, "learning_rate": 2.6348137798846807e-07, "loss": 0.0163, "step": 10536 }, { "epoch": 4.15, "learning_rate": 2.634571647018238e-07, "loss": 0.0168, "step": 10537 }, { "epoch": 4.15, "learning_rate": 2.634329503809241e-07, "loss": 0.0146, "step": 10538 }, { "epoch": 4.15, "learning_rate": 2.634087350261638e-07, "loss": 0.0138, "step": 10539 }, { "epoch": 4.15, "learning_rate": 2.633845186379375e-07, "loss": 0.013, "step": 10540 }, { "epoch": 4.15, "learning_rate": 2.6336030121663977e-07, "loss": 0.0145, "step": 10541 }, { "epoch": 4.15, "learning_rate": 2.6333608276266556e-07, "loss": 0.0156, "step": 10542 }, { "epoch": 4.16, "learning_rate": 2.633118632764094e-07, "loss": 0.0161, "step": 10543 }, { "epoch": 4.16, "learning_rate": 2.632876427582661e-07, "loss": 0.0142, "step": 10544 }, { "epoch": 4.16, "learning_rate": 2.6326342120863037e-07, "loss": 0.0151, "step": 10545 }, { "epoch": 4.16, "learning_rate": 2.632391986278971e-07, "loss": 0.0138, "step": 10546 }, { "epoch": 4.16, "learning_rate": 2.6321497501646095e-07, "loss": 0.014, "step": 10547 }, { "epoch": 4.16, "learning_rate": 2.6319075037471677e-07, "loss": 0.0119, "step": 10548 }, { "epoch": 4.16, "learning_rate": 2.631665247030594e-07, "loss": 0.02, "step": 10549 }, { "epoch": 4.16, "learning_rate": 2.6314229800188373e-07, "loss": 0.015, "step": 10550 }, { "epoch": 4.16, "learning_rate": 2.631180702715845e-07, "loss": 0.0189, "step": 10551 }, { "epoch": 4.16, "learning_rate": 2.630938415125567e-07, "loss": 0.0159, "step": 10552 }, { "epoch": 4.16, "learning_rate": 2.630696117251952e-07, "loss": 0.015, "step": 10553 }, { "epoch": 4.16, "learning_rate": 2.6304538090989485e-07, "loss": 0.0128, "step": 10554 }, { "epoch": 4.16, "learning_rate": 2.630211490670506e-07, "loss": 0.0179, "step": 10555 }, { "epoch": 4.16, "learning_rate": 2.629969161970574e-07, "loss": 0.0152, "step": 10556 }, { "epoch": 4.16, "learning_rate": 2.629726823003102e-07, "loss": 0.0144, "step": 10557 }, { "epoch": 4.16, "learning_rate": 2.62948447377204e-07, "loss": 0.0169, "step": 10558 }, { "epoch": 4.16, "learning_rate": 2.629242114281338e-07, "loss": 0.0161, "step": 10559 }, { "epoch": 4.16, "learning_rate": 2.628999744534946e-07, "loss": 0.0195, "step": 10560 }, { "epoch": 4.16, "learning_rate": 2.6287573645368135e-07, "loss": 0.0167, "step": 10561 }, { "epoch": 4.16, "learning_rate": 2.6285149742908924e-07, "loss": 0.014, "step": 10562 }, { "epoch": 4.16, "learning_rate": 2.628272573801131e-07, "loss": 0.0138, "step": 10563 }, { "epoch": 4.16, "learning_rate": 2.628030163071483e-07, "loss": 0.0157, "step": 10564 }, { "epoch": 4.16, "learning_rate": 2.627787742105898e-07, "loss": 0.0147, "step": 10565 }, { "epoch": 4.16, "learning_rate": 2.627545310908326e-07, "loss": 0.0159, "step": 10566 }, { "epoch": 4.16, "learning_rate": 2.6273028694827186e-07, "loss": 0.0151, "step": 10567 }, { "epoch": 4.16, "learning_rate": 2.6270604178330297e-07, "loss": 0.0183, "step": 10568 }, { "epoch": 4.17, "learning_rate": 2.626817955963207e-07, "loss": 0.0185, "step": 10569 }, { "epoch": 4.17, "learning_rate": 2.6265754838772054e-07, "loss": 0.0147, "step": 10570 }, { "epoch": 4.17, "learning_rate": 2.6263330015789755e-07, "loss": 0.0138, "step": 10571 }, { "epoch": 4.17, "learning_rate": 2.6260905090724697e-07, "loss": 0.0147, "step": 10572 }, { "epoch": 4.17, "learning_rate": 2.6258480063616405e-07, "loss": 0.016, "step": 10573 }, { "epoch": 4.17, "learning_rate": 2.625605493450439e-07, "loss": 0.0143, "step": 10574 }, { "epoch": 4.17, "learning_rate": 2.62536297034282e-07, "loss": 0.0146, "step": 10575 }, { "epoch": 4.17, "learning_rate": 2.6251204370427336e-07, "loss": 0.0151, "step": 10576 }, { "epoch": 4.17, "learning_rate": 2.6248778935541353e-07, "loss": 0.0196, "step": 10577 }, { "epoch": 4.17, "learning_rate": 2.6246353398809764e-07, "loss": 0.0151, "step": 10578 }, { "epoch": 4.17, "learning_rate": 2.624392776027211e-07, "loss": 0.0157, "step": 10579 }, { "epoch": 4.17, "learning_rate": 2.6241502019967926e-07, "loss": 0.0139, "step": 10580 }, { "epoch": 4.17, "learning_rate": 2.6239076177936737e-07, "loss": 0.0172, "step": 10581 }, { "epoch": 4.17, "learning_rate": 2.62366502342181e-07, "loss": 0.0134, "step": 10582 }, { "epoch": 4.17, "learning_rate": 2.623422418885153e-07, "loss": 0.0146, "step": 10583 }, { "epoch": 4.17, "learning_rate": 2.6231798041876596e-07, "loss": 0.0156, "step": 10584 }, { "epoch": 4.17, "learning_rate": 2.6229371793332817e-07, "loss": 0.0158, "step": 10585 }, { "epoch": 4.17, "learning_rate": 2.622694544325974e-07, "loss": 0.0199, "step": 10586 }, { "epoch": 4.17, "learning_rate": 2.622451899169692e-07, "loss": 0.0115, "step": 10587 }, { "epoch": 4.17, "learning_rate": 2.622209243868391e-07, "loss": 0.0116, "step": 10588 }, { "epoch": 4.17, "learning_rate": 2.6219665784260233e-07, "loss": 0.0132, "step": 10589 }, { "epoch": 4.17, "learning_rate": 2.621723902846547e-07, "loss": 0.0177, "step": 10590 }, { "epoch": 4.17, "learning_rate": 2.621481217133916e-07, "loss": 0.0123, "step": 10591 }, { "epoch": 4.17, "learning_rate": 2.6212385212920843e-07, "loss": 0.0142, "step": 10592 }, { "epoch": 4.17, "learning_rate": 2.6209958153250104e-07, "loss": 0.016, "step": 10593 }, { "epoch": 4.18, "learning_rate": 2.620753099236647e-07, "loss": 0.018, "step": 10594 }, { "epoch": 4.18, "learning_rate": 2.620510373030953e-07, "loss": 0.0155, "step": 10595 }, { "epoch": 4.18, "learning_rate": 2.6202676367118825e-07, "loss": 0.0195, "step": 10596 }, { "epoch": 4.18, "learning_rate": 2.620024890283392e-07, "loss": 0.0139, "step": 10597 }, { "epoch": 4.18, "learning_rate": 2.619782133749438e-07, "loss": 0.016, "step": 10598 }, { "epoch": 4.18, "learning_rate": 2.619539367113978e-07, "loss": 0.0149, "step": 10599 }, { "epoch": 4.18, "learning_rate": 2.619296590380967e-07, "loss": 0.0161, "step": 10600 }, { "epoch": 4.18, "learning_rate": 2.619053803554363e-07, "loss": 0.0161, "step": 10601 }, { "epoch": 4.18, "learning_rate": 2.618811006638123e-07, "loss": 0.0138, "step": 10602 }, { "epoch": 4.18, "learning_rate": 2.618568199636204e-07, "loss": 0.0155, "step": 10603 }, { "epoch": 4.18, "learning_rate": 2.6183253825525635e-07, "loss": 0.0122, "step": 10604 }, { "epoch": 4.18, "learning_rate": 2.618082555391159e-07, "loss": 0.0126, "step": 10605 }, { "epoch": 4.18, "learning_rate": 2.6178397181559485e-07, "loss": 0.0138, "step": 10606 }, { "epoch": 4.18, "learning_rate": 2.617596870850889e-07, "loss": 0.016, "step": 10607 }, { "epoch": 4.18, "learning_rate": 2.6173540134799396e-07, "loss": 0.0155, "step": 10608 }, { "epoch": 4.18, "learning_rate": 2.617111146047058e-07, "loss": 0.0202, "step": 10609 }, { "epoch": 4.18, "learning_rate": 2.6168682685562026e-07, "loss": 0.0142, "step": 10610 }, { "epoch": 4.18, "learning_rate": 2.616625381011332e-07, "loss": 0.014, "step": 10611 }, { "epoch": 4.18, "learning_rate": 2.6163824834164047e-07, "loss": 0.0158, "step": 10612 }, { "epoch": 4.18, "learning_rate": 2.61613957577538e-07, "loss": 0.0156, "step": 10613 }, { "epoch": 4.18, "learning_rate": 2.615896658092216e-07, "loss": 0.016, "step": 10614 }, { "epoch": 4.18, "learning_rate": 2.6156537303708737e-07, "loss": 0.014, "step": 10615 }, { "epoch": 4.18, "learning_rate": 2.61541079261531e-07, "loss": 0.0217, "step": 10616 }, { "epoch": 4.18, "learning_rate": 2.615167844829486e-07, "loss": 0.0171, "step": 10617 }, { "epoch": 4.18, "learning_rate": 2.6149248870173616e-07, "loss": 0.0175, "step": 10618 }, { "epoch": 4.19, "learning_rate": 2.6146819191828957e-07, "loss": 0.0138, "step": 10619 }, { "epoch": 4.19, "learning_rate": 2.614438941330049e-07, "loss": 0.0178, "step": 10620 }, { "epoch": 4.19, "learning_rate": 2.614195953462781e-07, "loss": 0.0168, "step": 10621 }, { "epoch": 4.19, "learning_rate": 2.613952955585053e-07, "loss": 0.0183, "step": 10622 }, { "epoch": 4.19, "learning_rate": 2.6137099477008246e-07, "loss": 0.0166, "step": 10623 }, { "epoch": 4.19, "learning_rate": 2.613466929814057e-07, "loss": 0.0144, "step": 10624 }, { "epoch": 4.19, "learning_rate": 2.61322390192871e-07, "loss": 0.0206, "step": 10625 }, { "epoch": 4.19, "learning_rate": 2.612980864048746e-07, "loss": 0.0123, "step": 10626 }, { "epoch": 4.19, "learning_rate": 2.6127378161781256e-07, "loss": 0.0148, "step": 10627 }, { "epoch": 4.19, "learning_rate": 2.61249475832081e-07, "loss": 0.0146, "step": 10628 }, { "epoch": 4.19, "learning_rate": 2.612251690480761e-07, "loss": 0.0153, "step": 10629 }, { "epoch": 4.19, "learning_rate": 2.6120086126619396e-07, "loss": 0.0159, "step": 10630 }, { "epoch": 4.19, "learning_rate": 2.6117655248683086e-07, "loss": 0.0133, "step": 10631 }, { "epoch": 4.19, "learning_rate": 2.6115224271038283e-07, "loss": 0.0115, "step": 10632 }, { "epoch": 4.19, "learning_rate": 2.611279319372462e-07, "loss": 0.0137, "step": 10633 }, { "epoch": 4.19, "learning_rate": 2.611036201678172e-07, "loss": 0.0169, "step": 10634 }, { "epoch": 4.19, "learning_rate": 2.6107930740249214e-07, "loss": 0.0145, "step": 10635 }, { "epoch": 4.19, "learning_rate": 2.610549936416671e-07, "loss": 0.0146, "step": 10636 }, { "epoch": 4.19, "learning_rate": 2.6103067888573846e-07, "loss": 0.0155, "step": 10637 }, { "epoch": 4.19, "learning_rate": 2.610063631351025e-07, "loss": 0.0131, "step": 10638 }, { "epoch": 4.19, "learning_rate": 2.609820463901556e-07, "loss": 0.0157, "step": 10639 }, { "epoch": 4.19, "learning_rate": 2.60957728651294e-07, "loss": 0.0179, "step": 10640 }, { "epoch": 4.19, "learning_rate": 2.609334099189141e-07, "loss": 0.0165, "step": 10641 }, { "epoch": 4.19, "learning_rate": 2.609090901934122e-07, "loss": 0.0157, "step": 10642 }, { "epoch": 4.19, "learning_rate": 2.6088476947518467e-07, "loss": 0.0163, "step": 10643 }, { "epoch": 4.19, "learning_rate": 2.608604477646279e-07, "loss": 0.0156, "step": 10644 }, { "epoch": 4.2, "learning_rate": 2.608361250621384e-07, "loss": 0.0168, "step": 10645 }, { "epoch": 4.2, "learning_rate": 2.6081180136811256e-07, "loss": 0.0131, "step": 10646 }, { "epoch": 4.2, "learning_rate": 2.6078747668294673e-07, "loss": 0.0137, "step": 10647 }, { "epoch": 4.2, "learning_rate": 2.6076315100703737e-07, "loss": 0.0179, "step": 10648 }, { "epoch": 4.2, "learning_rate": 2.607388243407811e-07, "loss": 0.0112, "step": 10649 }, { "epoch": 4.2, "learning_rate": 2.607144966845742e-07, "loss": 0.0152, "step": 10650 }, { "epoch": 4.2, "learning_rate": 2.606901680388134e-07, "loss": 0.0205, "step": 10651 }, { "epoch": 4.2, "learning_rate": 2.60665838403895e-07, "loss": 0.0137, "step": 10652 }, { "epoch": 4.2, "learning_rate": 2.6064150778021565e-07, "loss": 0.0154, "step": 10653 }, { "epoch": 4.2, "learning_rate": 2.606171761681719e-07, "loss": 0.0143, "step": 10654 }, { "epoch": 4.2, "learning_rate": 2.6059284356816034e-07, "loss": 0.015, "step": 10655 }, { "epoch": 4.2, "learning_rate": 2.605685099805775e-07, "loss": 0.0168, "step": 10656 }, { "epoch": 4.2, "learning_rate": 2.6054417540582e-07, "loss": 0.0148, "step": 10657 }, { "epoch": 4.2, "learning_rate": 2.605198398442844e-07, "loss": 0.0136, "step": 10658 }, { "epoch": 4.2, "learning_rate": 2.604955032963675e-07, "loss": 0.0194, "step": 10659 }, { "epoch": 4.2, "learning_rate": 2.6047116576246575e-07, "loss": 0.0185, "step": 10660 }, { "epoch": 4.2, "learning_rate": 2.604468272429759e-07, "loss": 0.0159, "step": 10661 }, { "epoch": 4.2, "learning_rate": 2.604224877382947e-07, "loss": 0.0149, "step": 10662 }, { "epoch": 4.2, "learning_rate": 2.603981472488187e-07, "loss": 0.015, "step": 10663 }, { "epoch": 4.2, "learning_rate": 2.6037380577494476e-07, "loss": 0.0135, "step": 10664 }, { "epoch": 4.2, "learning_rate": 2.603494633170695e-07, "loss": 0.0163, "step": 10665 }, { "epoch": 4.2, "learning_rate": 2.6032511987558974e-07, "loss": 0.0152, "step": 10666 }, { "epoch": 4.2, "learning_rate": 2.603007754509022e-07, "loss": 0.0148, "step": 10667 }, { "epoch": 4.2, "learning_rate": 2.602764300434037e-07, "loss": 0.015, "step": 10668 }, { "epoch": 4.2, "learning_rate": 2.602520836534909e-07, "loss": 0.0174, "step": 10669 }, { "epoch": 4.21, "learning_rate": 2.602277362815608e-07, "loss": 0.0151, "step": 10670 }, { "epoch": 4.21, "learning_rate": 2.6020338792801016e-07, "loss": 0.0173, "step": 10671 }, { "epoch": 4.21, "learning_rate": 2.601790385932357e-07, "loss": 0.0153, "step": 10672 }, { "epoch": 4.21, "learning_rate": 2.6015468827763445e-07, "loss": 0.0134, "step": 10673 }, { "epoch": 4.21, "learning_rate": 2.601303369816032e-07, "loss": 0.0157, "step": 10674 }, { "epoch": 4.21, "learning_rate": 2.601059847055388e-07, "loss": 0.0164, "step": 10675 }, { "epoch": 4.21, "learning_rate": 2.600816314498382e-07, "loss": 0.0149, "step": 10676 }, { "epoch": 4.21, "learning_rate": 2.6005727721489833e-07, "loss": 0.0168, "step": 10677 }, { "epoch": 4.21, "learning_rate": 2.6003292200111617e-07, "loss": 0.0163, "step": 10678 }, { "epoch": 4.21, "learning_rate": 2.6000856580888856e-07, "loss": 0.0196, "step": 10679 }, { "epoch": 4.21, "learning_rate": 2.5998420863861254e-07, "loss": 0.017, "step": 10680 }, { "epoch": 4.21, "learning_rate": 2.5995985049068507e-07, "loss": 0.0151, "step": 10681 }, { "epoch": 4.21, "learning_rate": 2.599354913655032e-07, "loss": 0.016, "step": 10682 }, { "epoch": 4.21, "learning_rate": 2.599111312634639e-07, "loss": 0.0165, "step": 10683 }, { "epoch": 4.21, "learning_rate": 2.5988677018496427e-07, "loss": 0.0128, "step": 10684 }, { "epoch": 4.21, "learning_rate": 2.598624081304012e-07, "loss": 0.017, "step": 10685 }, { "epoch": 4.21, "learning_rate": 2.598380451001719e-07, "loss": 0.0147, "step": 10686 }, { "epoch": 4.21, "learning_rate": 2.598136810946734e-07, "loss": 0.0172, "step": 10687 }, { "epoch": 4.21, "learning_rate": 2.5978931611430284e-07, "loss": 0.0142, "step": 10688 }, { "epoch": 4.21, "learning_rate": 2.5976495015945724e-07, "loss": 0.0146, "step": 10689 }, { "epoch": 4.21, "learning_rate": 2.5974058323053385e-07, "loss": 0.0165, "step": 10690 }, { "epoch": 4.21, "learning_rate": 2.5971621532792974e-07, "loss": 0.015, "step": 10691 }, { "epoch": 4.21, "learning_rate": 2.5969184645204205e-07, "loss": 0.0156, "step": 10692 }, { "epoch": 4.21, "learning_rate": 2.59667476603268e-07, "loss": 0.0147, "step": 10693 }, { "epoch": 4.21, "learning_rate": 2.5964310578200475e-07, "loss": 0.0128, "step": 10694 }, { "epoch": 4.21, "learning_rate": 2.5961873398864955e-07, "loss": 0.0146, "step": 10695 }, { "epoch": 4.22, "learning_rate": 2.5959436122359955e-07, "loss": 0.017, "step": 10696 }, { "epoch": 4.22, "learning_rate": 2.595699874872521e-07, "loss": 0.0148, "step": 10697 }, { "epoch": 4.22, "learning_rate": 2.5954561278000435e-07, "loss": 0.02, "step": 10698 }, { "epoch": 4.22, "learning_rate": 2.5952123710225363e-07, "loss": 0.0174, "step": 10699 }, { "epoch": 4.22, "learning_rate": 2.594968604543972e-07, "loss": 0.0143, "step": 10700 }, { "epoch": 4.22, "learning_rate": 2.594724828368323e-07, "loss": 0.0159, "step": 10701 }, { "epoch": 4.22, "learning_rate": 2.594481042499564e-07, "loss": 0.0151, "step": 10702 }, { "epoch": 4.22, "learning_rate": 2.594237246941667e-07, "loss": 0.0167, "step": 10703 }, { "epoch": 4.22, "learning_rate": 2.5939934416986066e-07, "loss": 0.0132, "step": 10704 }, { "epoch": 4.22, "learning_rate": 2.5937496267743554e-07, "loss": 0.0148, "step": 10705 }, { "epoch": 4.22, "learning_rate": 2.5935058021728877e-07, "loss": 0.0141, "step": 10706 }, { "epoch": 4.22, "learning_rate": 2.593261967898178e-07, "loss": 0.0154, "step": 10707 }, { "epoch": 4.22, "learning_rate": 2.593018123954199e-07, "loss": 0.014, "step": 10708 }, { "epoch": 4.22, "learning_rate": 2.592774270344926e-07, "loss": 0.0166, "step": 10709 }, { "epoch": 4.22, "learning_rate": 2.592530407074334e-07, "loss": 0.0155, "step": 10710 }, { "epoch": 4.22, "learning_rate": 2.592286534146397e-07, "loss": 0.0181, "step": 10711 }, { "epoch": 4.22, "learning_rate": 2.592042651565089e-07, "loss": 0.0144, "step": 10712 }, { "epoch": 4.22, "learning_rate": 2.5917987593343856e-07, "loss": 0.0141, "step": 10713 }, { "epoch": 4.22, "learning_rate": 2.591554857458262e-07, "loss": 0.0139, "step": 10714 }, { "epoch": 4.22, "learning_rate": 2.5913109459406933e-07, "loss": 0.0165, "step": 10715 }, { "epoch": 4.22, "learning_rate": 2.591067024785655e-07, "loss": 0.0156, "step": 10716 }, { "epoch": 4.22, "learning_rate": 2.590823093997123e-07, "loss": 0.0149, "step": 10717 }, { "epoch": 4.22, "learning_rate": 2.590579153579071e-07, "loss": 0.0196, "step": 10718 }, { "epoch": 4.22, "learning_rate": 2.5903352035354776e-07, "loss": 0.0129, "step": 10719 }, { "epoch": 4.22, "learning_rate": 2.590091243870318e-07, "loss": 0.0144, "step": 10720 }, { "epoch": 4.23, "learning_rate": 2.5898472745875665e-07, "loss": 0.0174, "step": 10721 }, { "epoch": 4.23, "learning_rate": 2.589603295691202e-07, "loss": 0.0179, "step": 10722 }, { "epoch": 4.23, "learning_rate": 2.589359307185199e-07, "loss": 0.016, "step": 10723 }, { "epoch": 4.23, "learning_rate": 2.5891153090735363e-07, "loss": 0.017, "step": 10724 }, { "epoch": 4.23, "learning_rate": 2.588871301360188e-07, "loss": 0.0138, "step": 10725 }, { "epoch": 4.23, "learning_rate": 2.5886272840491333e-07, "loss": 0.0141, "step": 10726 }, { "epoch": 4.23, "learning_rate": 2.588383257144348e-07, "loss": 0.0163, "step": 10727 }, { "epoch": 4.23, "learning_rate": 2.5881392206498104e-07, "loss": 0.0128, "step": 10728 }, { "epoch": 4.23, "learning_rate": 2.587895174569497e-07, "loss": 0.0159, "step": 10729 }, { "epoch": 4.23, "learning_rate": 2.5876511189073856e-07, "loss": 0.0153, "step": 10730 }, { "epoch": 4.23, "learning_rate": 2.587407053667455e-07, "loss": 0.0149, "step": 10731 }, { "epoch": 4.23, "learning_rate": 2.5871629788536806e-07, "loss": 0.0176, "step": 10732 }, { "epoch": 4.23, "learning_rate": 2.5869188944700435e-07, "loss": 0.0163, "step": 10733 }, { "epoch": 4.23, "learning_rate": 2.5866748005205196e-07, "loss": 0.0176, "step": 10734 }, { "epoch": 4.23, "learning_rate": 2.5864306970090887e-07, "loss": 0.0139, "step": 10735 }, { "epoch": 4.23, "learning_rate": 2.5861865839397284e-07, "loss": 0.0142, "step": 10736 }, { "epoch": 4.23, "learning_rate": 2.5859424613164177e-07, "loss": 0.0141, "step": 10737 }, { "epoch": 4.23, "learning_rate": 2.5856983291431356e-07, "loss": 0.0148, "step": 10738 }, { "epoch": 4.23, "learning_rate": 2.5854541874238605e-07, "loss": 0.0146, "step": 10739 }, { "epoch": 4.23, "learning_rate": 2.5852100361625723e-07, "loss": 0.0154, "step": 10740 }, { "epoch": 4.23, "learning_rate": 2.5849658753632496e-07, "loss": 0.0181, "step": 10741 }, { "epoch": 4.23, "learning_rate": 2.5847217050298727e-07, "loss": 0.0155, "step": 10742 }, { "epoch": 4.23, "learning_rate": 2.5844775251664196e-07, "loss": 0.0178, "step": 10743 }, { "epoch": 4.23, "learning_rate": 2.5842333357768724e-07, "loss": 0.0169, "step": 10744 }, { "epoch": 4.23, "learning_rate": 2.5839891368652085e-07, "loss": 0.0146, "step": 10745 }, { "epoch": 4.24, "learning_rate": 2.5837449284354105e-07, "loss": 0.0173, "step": 10746 }, { "epoch": 4.24, "learning_rate": 2.5835007104914567e-07, "loss": 0.0152, "step": 10747 }, { "epoch": 4.24, "learning_rate": 2.5832564830373283e-07, "loss": 0.0134, "step": 10748 }, { "epoch": 4.24, "learning_rate": 2.583012246077005e-07, "loss": 0.0163, "step": 10749 }, { "epoch": 4.24, "learning_rate": 2.5827679996144686e-07, "loss": 0.0162, "step": 10750 }, { "epoch": 4.24, "learning_rate": 2.5825237436537005e-07, "loss": 0.0175, "step": 10751 }, { "epoch": 4.24, "learning_rate": 2.5822794781986793e-07, "loss": 0.0153, "step": 10752 }, { "epoch": 4.24, "learning_rate": 2.582035203253389e-07, "loss": 0.0121, "step": 10753 }, { "epoch": 4.24, "learning_rate": 2.581790918821808e-07, "loss": 0.0148, "step": 10754 }, { "epoch": 4.24, "learning_rate": 2.58154662490792e-07, "loss": 0.0156, "step": 10755 }, { "epoch": 4.24, "learning_rate": 2.5813023215157055e-07, "loss": 0.0137, "step": 10756 }, { "epoch": 4.24, "learning_rate": 2.581058008649147e-07, "loss": 0.0156, "step": 10757 }, { "epoch": 4.24, "learning_rate": 2.580813686312226e-07, "loss": 0.0165, "step": 10758 }, { "epoch": 4.24, "learning_rate": 2.5805693545089244e-07, "loss": 0.0158, "step": 10759 }, { "epoch": 4.24, "learning_rate": 2.580325013243225e-07, "loss": 0.0119, "step": 10760 }, { "epoch": 4.24, "learning_rate": 2.5800806625191103e-07, "loss": 0.0151, "step": 10761 }, { "epoch": 4.24, "learning_rate": 2.5798363023405614e-07, "loss": 0.0145, "step": 10762 }, { "epoch": 4.24, "learning_rate": 2.5795919327115626e-07, "loss": 0.0148, "step": 10763 }, { "epoch": 4.24, "learning_rate": 2.579347553636096e-07, "loss": 0.0156, "step": 10764 }, { "epoch": 4.24, "learning_rate": 2.5791031651181446e-07, "loss": 0.015, "step": 10765 }, { "epoch": 4.24, "learning_rate": 2.578858767161693e-07, "loss": 0.0152, "step": 10766 }, { "epoch": 4.24, "learning_rate": 2.5786143597707217e-07, "loss": 0.0128, "step": 10767 }, { "epoch": 4.24, "learning_rate": 2.578369942949216e-07, "loss": 0.0153, "step": 10768 }, { "epoch": 4.24, "learning_rate": 2.57812551670116e-07, "loss": 0.0159, "step": 10769 }, { "epoch": 4.24, "learning_rate": 2.5778810810305363e-07, "loss": 0.0156, "step": 10770 }, { "epoch": 4.24, "learning_rate": 2.577636635941329e-07, "loss": 0.0183, "step": 10771 }, { "epoch": 4.25, "learning_rate": 2.5773921814375224e-07, "loss": 0.0132, "step": 10772 }, { "epoch": 4.25, "learning_rate": 2.5771477175231014e-07, "loss": 0.0152, "step": 10773 }, { "epoch": 4.25, "learning_rate": 2.5769032442020493e-07, "loss": 0.019, "step": 10774 }, { "epoch": 4.25, "learning_rate": 2.576658761478352e-07, "loss": 0.0135, "step": 10775 }, { "epoch": 4.25, "learning_rate": 2.5764142693559914e-07, "loss": 0.0155, "step": 10776 }, { "epoch": 4.25, "learning_rate": 2.576169767838956e-07, "loss": 0.0195, "step": 10777 }, { "epoch": 4.25, "learning_rate": 2.575925256931228e-07, "loss": 0.0123, "step": 10778 }, { "epoch": 4.25, "learning_rate": 2.575680736636794e-07, "loss": 0.017, "step": 10779 }, { "epoch": 4.25, "learning_rate": 2.5754362069596395e-07, "loss": 0.0121, "step": 10780 }, { "epoch": 4.25, "learning_rate": 2.575191667903749e-07, "loss": 0.0134, "step": 10781 }, { "epoch": 4.25, "learning_rate": 2.574947119473108e-07, "loss": 0.0146, "step": 10782 }, { "epoch": 4.25, "learning_rate": 2.5747025616717033e-07, "loss": 0.0165, "step": 10783 }, { "epoch": 4.25, "learning_rate": 2.57445799450352e-07, "loss": 0.0138, "step": 10784 }, { "epoch": 4.25, "learning_rate": 2.574213417972545e-07, "loss": 0.0155, "step": 10785 }, { "epoch": 4.25, "learning_rate": 2.5739688320827633e-07, "loss": 0.0156, "step": 10786 }, { "epoch": 4.25, "learning_rate": 2.5737242368381625e-07, "loss": 0.0137, "step": 10787 }, { "epoch": 4.25, "learning_rate": 2.573479632242728e-07, "loss": 0.018, "step": 10788 }, { "epoch": 4.25, "learning_rate": 2.5732350183004477e-07, "loss": 0.0162, "step": 10789 }, { "epoch": 4.25, "learning_rate": 2.572990395015307e-07, "loss": 0.0203, "step": 10790 }, { "epoch": 4.25, "learning_rate": 2.5727457623912953e-07, "loss": 0.0143, "step": 10791 }, { "epoch": 4.25, "learning_rate": 2.5725011204323964e-07, "loss": 0.0145, "step": 10792 }, { "epoch": 4.25, "learning_rate": 2.5722564691426006e-07, "loss": 0.0172, "step": 10793 }, { "epoch": 4.25, "learning_rate": 2.572011808525893e-07, "loss": 0.0144, "step": 10794 }, { "epoch": 4.25, "learning_rate": 2.571767138586263e-07, "loss": 0.016, "step": 10795 }, { "epoch": 4.25, "learning_rate": 2.571522459327697e-07, "loss": 0.0154, "step": 10796 }, { "epoch": 4.26, "learning_rate": 2.571277770754184e-07, "loss": 0.0141, "step": 10797 }, { "epoch": 4.26, "learning_rate": 2.571033072869712e-07, "loss": 0.0132, "step": 10798 }, { "epoch": 4.26, "learning_rate": 2.5707883656782684e-07, "loss": 0.0184, "step": 10799 }, { "epoch": 4.26, "learning_rate": 2.570543649183842e-07, "loss": 0.0167, "step": 10800 }, { "epoch": 4.26, "learning_rate": 2.570298923390421e-07, "loss": 0.015, "step": 10801 }, { "epoch": 4.26, "learning_rate": 2.5700541883019944e-07, "loss": 0.0171, "step": 10802 }, { "epoch": 4.26, "learning_rate": 2.5698094439225513e-07, "loss": 0.0168, "step": 10803 }, { "epoch": 4.26, "learning_rate": 2.5695646902560797e-07, "loss": 0.0138, "step": 10804 }, { "epoch": 4.26, "learning_rate": 2.5693199273065696e-07, "loss": 0.0136, "step": 10805 }, { "epoch": 4.26, "learning_rate": 2.56907515507801e-07, "loss": 0.0156, "step": 10806 }, { "epoch": 4.26, "learning_rate": 2.5688303735743897e-07, "loss": 0.0153, "step": 10807 }, { "epoch": 4.26, "learning_rate": 2.5685855827996995e-07, "loss": 0.0162, "step": 10808 }, { "epoch": 4.26, "learning_rate": 2.5683407827579284e-07, "loss": 0.0153, "step": 10809 }, { "epoch": 4.26, "learning_rate": 2.568095973453066e-07, "loss": 0.0167, "step": 10810 }, { "epoch": 4.26, "learning_rate": 2.567851154889103e-07, "loss": 0.0142, "step": 10811 }, { "epoch": 4.26, "learning_rate": 2.5676063270700283e-07, "loss": 0.0148, "step": 10812 }, { "epoch": 4.26, "learning_rate": 2.5673614899998344e-07, "loss": 0.018, "step": 10813 }, { "epoch": 4.26, "learning_rate": 2.5671166436825094e-07, "loss": 0.0148, "step": 10814 }, { "epoch": 4.26, "learning_rate": 2.566871788122046e-07, "loss": 0.017, "step": 10815 }, { "epoch": 4.26, "learning_rate": 2.5666269233224327e-07, "loss": 0.0165, "step": 10816 }, { "epoch": 4.26, "learning_rate": 2.5663820492876633e-07, "loss": 0.0164, "step": 10817 }, { "epoch": 4.26, "learning_rate": 2.5661371660217266e-07, "loss": 0.0139, "step": 10818 }, { "epoch": 4.26, "learning_rate": 2.5658922735286143e-07, "loss": 0.0156, "step": 10819 }, { "epoch": 4.26, "learning_rate": 2.565647371812318e-07, "loss": 0.016, "step": 10820 }, { "epoch": 4.26, "learning_rate": 2.5654024608768293e-07, "loss": 0.0188, "step": 10821 }, { "epoch": 4.27, "learning_rate": 2.5651575407261404e-07, "loss": 0.0149, "step": 10822 }, { "epoch": 4.27, "learning_rate": 2.5649126113642414e-07, "loss": 0.0152, "step": 10823 }, { "epoch": 4.27, "learning_rate": 2.564667672795126e-07, "loss": 0.017, "step": 10824 }, { "epoch": 4.27, "learning_rate": 2.564422725022786e-07, "loss": 0.019, "step": 10825 }, { "epoch": 4.27, "learning_rate": 2.564177768051214e-07, "loss": 0.0164, "step": 10826 }, { "epoch": 4.27, "learning_rate": 2.5639328018844e-07, "loss": 0.0163, "step": 10827 }, { "epoch": 4.27, "learning_rate": 2.56368782652634e-07, "loss": 0.0152, "step": 10828 }, { "epoch": 4.27, "learning_rate": 2.5634428419810243e-07, "loss": 0.0162, "step": 10829 }, { "epoch": 4.27, "learning_rate": 2.563197848252447e-07, "loss": 0.0155, "step": 10830 }, { "epoch": 4.27, "learning_rate": 2.5629528453446003e-07, "loss": 0.0124, "step": 10831 }, { "epoch": 4.27, "learning_rate": 2.5627078332614785e-07, "loss": 0.0162, "step": 10832 }, { "epoch": 4.27, "learning_rate": 2.562462812007074e-07, "loss": 0.0191, "step": 10833 }, { "epoch": 4.27, "learning_rate": 2.56221778158538e-07, "loss": 0.0139, "step": 10834 }, { "epoch": 4.27, "learning_rate": 2.561972742000391e-07, "loss": 0.0158, "step": 10835 }, { "epoch": 4.27, "learning_rate": 2.5617276932561004e-07, "loss": 0.0141, "step": 10836 }, { "epoch": 4.27, "learning_rate": 2.561482635356502e-07, "loss": 0.0122, "step": 10837 }, { "epoch": 4.27, "learning_rate": 2.5612375683055896e-07, "loss": 0.0153, "step": 10838 }, { "epoch": 4.27, "learning_rate": 2.5609924921073587e-07, "loss": 0.0161, "step": 10839 }, { "epoch": 4.27, "learning_rate": 2.560747406765803e-07, "loss": 0.0167, "step": 10840 }, { "epoch": 4.27, "learning_rate": 2.5605023122849154e-07, "loss": 0.0157, "step": 10841 }, { "epoch": 4.27, "learning_rate": 2.5602572086686933e-07, "loss": 0.0161, "step": 10842 }, { "epoch": 4.27, "learning_rate": 2.5600120959211294e-07, "loss": 0.0128, "step": 10843 }, { "epoch": 4.27, "learning_rate": 2.55976697404622e-07, "loss": 0.0165, "step": 10844 }, { "epoch": 4.27, "learning_rate": 2.5595218430479596e-07, "loss": 0.0133, "step": 10845 }, { "epoch": 4.27, "learning_rate": 2.559276702930343e-07, "loss": 0.0132, "step": 10846 }, { "epoch": 4.27, "learning_rate": 2.559031553697367e-07, "loss": 0.0155, "step": 10847 }, { "epoch": 4.28, "learning_rate": 2.5587863953530256e-07, "loss": 0.0135, "step": 10848 }, { "epoch": 4.28, "learning_rate": 2.558541227901315e-07, "loss": 0.017, "step": 10849 }, { "epoch": 4.28, "learning_rate": 2.558296051346232e-07, "loss": 0.016, "step": 10850 }, { "epoch": 4.28, "learning_rate": 2.558050865691772e-07, "loss": 0.0165, "step": 10851 }, { "epoch": 4.28, "learning_rate": 2.55780567094193e-07, "loss": 0.015, "step": 10852 }, { "epoch": 4.28, "learning_rate": 2.5575604671007043e-07, "loss": 0.0142, "step": 10853 }, { "epoch": 4.28, "learning_rate": 2.5573152541720894e-07, "loss": 0.0137, "step": 10854 }, { "epoch": 4.28, "learning_rate": 2.5570700321600837e-07, "loss": 0.0154, "step": 10855 }, { "epoch": 4.28, "learning_rate": 2.556824801068683e-07, "loss": 0.0175, "step": 10856 }, { "epoch": 4.28, "learning_rate": 2.556579560901884e-07, "loss": 0.0179, "step": 10857 }, { "epoch": 4.28, "learning_rate": 2.556334311663684e-07, "loss": 0.018, "step": 10858 }, { "epoch": 4.28, "learning_rate": 2.556089053358081e-07, "loss": 0.0167, "step": 10859 }, { "epoch": 4.28, "learning_rate": 2.5558437859890715e-07, "loss": 0.0128, "step": 10860 }, { "epoch": 4.28, "learning_rate": 2.555598509560652e-07, "loss": 0.0155, "step": 10861 }, { "epoch": 4.28, "learning_rate": 2.5553532240768225e-07, "loss": 0.0146, "step": 10862 }, { "epoch": 4.28, "learning_rate": 2.5551079295415784e-07, "loss": 0.0146, "step": 10863 }, { "epoch": 4.28, "learning_rate": 2.5548626259589193e-07, "loss": 0.0133, "step": 10864 }, { "epoch": 4.28, "learning_rate": 2.5546173133328424e-07, "loss": 0.0132, "step": 10865 }, { "epoch": 4.28, "learning_rate": 2.554371991667347e-07, "loss": 0.017, "step": 10866 }, { "epoch": 4.28, "learning_rate": 2.55412666096643e-07, "loss": 0.0152, "step": 10867 }, { "epoch": 4.28, "learning_rate": 2.5538813212340906e-07, "loss": 0.0137, "step": 10868 }, { "epoch": 4.28, "learning_rate": 2.5536359724743273e-07, "loss": 0.0168, "step": 10869 }, { "epoch": 4.28, "learning_rate": 2.55339061469114e-07, "loss": 0.0128, "step": 10870 }, { "epoch": 4.28, "learning_rate": 2.553145247888526e-07, "loss": 0.0178, "step": 10871 }, { "epoch": 4.28, "learning_rate": 2.5528998720704853e-07, "loss": 0.0148, "step": 10872 }, { "epoch": 4.29, "learning_rate": 2.5526544872410167e-07, "loss": 0.0171, "step": 10873 }, { "epoch": 4.29, "learning_rate": 2.5524090934041197e-07, "loss": 0.0182, "step": 10874 }, { "epoch": 4.29, "learning_rate": 2.5521636905637945e-07, "loss": 0.017, "step": 10875 }, { "epoch": 4.29, "learning_rate": 2.5519182787240403e-07, "loss": 0.0201, "step": 10876 }, { "epoch": 4.29, "learning_rate": 2.5516728578888575e-07, "loss": 0.0119, "step": 10877 }, { "epoch": 4.29, "learning_rate": 2.551427428062245e-07, "loss": 0.0124, "step": 10878 }, { "epoch": 4.29, "learning_rate": 2.551181989248203e-07, "loss": 0.0202, "step": 10879 }, { "epoch": 4.29, "learning_rate": 2.550936541450733e-07, "loss": 0.0182, "step": 10880 }, { "epoch": 4.29, "learning_rate": 2.550691084673835e-07, "loss": 0.0187, "step": 10881 }, { "epoch": 4.29, "learning_rate": 2.5504456189215086e-07, "loss": 0.0174, "step": 10882 }, { "epoch": 4.29, "learning_rate": 2.550200144197755e-07, "loss": 0.0152, "step": 10883 }, { "epoch": 4.29, "learning_rate": 2.549954660506576e-07, "loss": 0.0146, "step": 10884 }, { "epoch": 4.29, "learning_rate": 2.549709167851971e-07, "loss": 0.0159, "step": 10885 }, { "epoch": 4.29, "learning_rate": 2.549463666237943e-07, "loss": 0.0158, "step": 10886 }, { "epoch": 4.29, "learning_rate": 2.549218155668491e-07, "loss": 0.0142, "step": 10887 }, { "epoch": 4.29, "learning_rate": 2.548972636147619e-07, "loss": 0.0158, "step": 10888 }, { "epoch": 4.29, "learning_rate": 2.548727107679327e-07, "loss": 0.0132, "step": 10889 }, { "epoch": 4.29, "learning_rate": 2.5484815702676176e-07, "loss": 0.0199, "step": 10890 }, { "epoch": 4.29, "learning_rate": 2.548236023916492e-07, "loss": 0.0156, "step": 10891 }, { "epoch": 4.29, "learning_rate": 2.5479904686299514e-07, "loss": 0.0156, "step": 10892 }, { "epoch": 4.29, "learning_rate": 2.547744904412e-07, "loss": 0.0159, "step": 10893 }, { "epoch": 4.29, "learning_rate": 2.5474993312666393e-07, "loss": 0.0138, "step": 10894 }, { "epoch": 4.29, "learning_rate": 2.547253749197871e-07, "loss": 0.0169, "step": 10895 }, { "epoch": 4.29, "learning_rate": 2.5470081582096987e-07, "loss": 0.0193, "step": 10896 }, { "epoch": 4.29, "learning_rate": 2.546762558306124e-07, "loss": 0.0135, "step": 10897 }, { "epoch": 4.29, "learning_rate": 2.5465169494911516e-07, "loss": 0.0162, "step": 10898 }, { "epoch": 4.3, "learning_rate": 2.546271331768783e-07, "loss": 0.0222, "step": 10899 }, { "epoch": 4.3, "learning_rate": 2.5460257051430223e-07, "loss": 0.018, "step": 10900 }, { "epoch": 4.3, "learning_rate": 2.545780069617872e-07, "loss": 0.0181, "step": 10901 }, { "epoch": 4.3, "learning_rate": 2.5455344251973365e-07, "loss": 0.0136, "step": 10902 }, { "epoch": 4.3, "learning_rate": 2.5452887718854185e-07, "loss": 0.0139, "step": 10903 }, { "epoch": 4.3, "learning_rate": 2.5450431096861227e-07, "loss": 0.0172, "step": 10904 }, { "epoch": 4.3, "learning_rate": 2.544797438603452e-07, "loss": 0.0177, "step": 10905 }, { "epoch": 4.3, "learning_rate": 2.5445517586414114e-07, "loss": 0.0138, "step": 10906 }, { "epoch": 4.3, "learning_rate": 2.5443060698040047e-07, "loss": 0.0143, "step": 10907 }, { "epoch": 4.3, "learning_rate": 2.5440603720952364e-07, "loss": 0.0176, "step": 10908 }, { "epoch": 4.3, "learning_rate": 2.543814665519111e-07, "loss": 0.0126, "step": 10909 }, { "epoch": 4.3, "learning_rate": 2.543568950079633e-07, "loss": 0.0162, "step": 10910 }, { "epoch": 4.3, "learning_rate": 2.5433232257808075e-07, "loss": 0.0196, "step": 10911 }, { "epoch": 4.3, "learning_rate": 2.543077492626639e-07, "loss": 0.0175, "step": 10912 }, { "epoch": 4.3, "learning_rate": 2.542831750621133e-07, "loss": 0.0179, "step": 10913 }, { "epoch": 4.3, "learning_rate": 2.5425859997682937e-07, "loss": 0.0166, "step": 10914 }, { "epoch": 4.3, "learning_rate": 2.5423402400721276e-07, "loss": 0.0168, "step": 10915 }, { "epoch": 4.3, "learning_rate": 2.54209447153664e-07, "loss": 0.0166, "step": 10916 }, { "epoch": 4.3, "learning_rate": 2.541848694165836e-07, "loss": 0.0141, "step": 10917 }, { "epoch": 4.3, "learning_rate": 2.541602907963722e-07, "loss": 0.0165, "step": 10918 }, { "epoch": 4.3, "learning_rate": 2.541357112934304e-07, "loss": 0.0143, "step": 10919 }, { "epoch": 4.3, "learning_rate": 2.5411113090815883e-07, "loss": 0.0168, "step": 10920 }, { "epoch": 4.3, "learning_rate": 2.5408654964095793e-07, "loss": 0.0141, "step": 10921 }, { "epoch": 4.3, "learning_rate": 2.540619674922286e-07, "loss": 0.0143, "step": 10922 }, { "epoch": 4.3, "learning_rate": 2.540373844623713e-07, "loss": 0.0163, "step": 10923 }, { "epoch": 4.31, "learning_rate": 2.540128005517868e-07, "loss": 0.0177, "step": 10924 }, { "epoch": 4.31, "learning_rate": 2.5398821576087566e-07, "loss": 0.0136, "step": 10925 }, { "epoch": 4.31, "learning_rate": 2.539636300900387e-07, "loss": 0.0149, "step": 10926 }, { "epoch": 4.31, "learning_rate": 2.539390435396766e-07, "loss": 0.0176, "step": 10927 }, { "epoch": 4.31, "learning_rate": 2.539144561101901e-07, "loss": 0.0155, "step": 10928 }, { "epoch": 4.31, "learning_rate": 2.5388986780197985e-07, "loss": 0.018, "step": 10929 }, { "epoch": 4.31, "learning_rate": 2.538652786154467e-07, "loss": 0.0135, "step": 10930 }, { "epoch": 4.31, "learning_rate": 2.5384068855099134e-07, "loss": 0.013, "step": 10931 }, { "epoch": 4.31, "learning_rate": 2.538160976090146e-07, "loss": 0.016, "step": 10932 }, { "epoch": 4.31, "learning_rate": 2.5379150578991726e-07, "loss": 0.016, "step": 10933 }, { "epoch": 4.31, "learning_rate": 2.537669130941001e-07, "loss": 0.0163, "step": 10934 }, { "epoch": 4.31, "learning_rate": 2.5374231952196403e-07, "loss": 0.0124, "step": 10935 }, { "epoch": 4.31, "learning_rate": 2.5371772507390974e-07, "loss": 0.0157, "step": 10936 }, { "epoch": 4.31, "learning_rate": 2.5369312975033827e-07, "loss": 0.0154, "step": 10937 }, { "epoch": 4.31, "learning_rate": 2.536685335516503e-07, "loss": 0.0155, "step": 10938 }, { "epoch": 4.31, "learning_rate": 2.5364393647824684e-07, "loss": 0.0164, "step": 10939 }, { "epoch": 4.31, "learning_rate": 2.536193385305288e-07, "loss": 0.015, "step": 10940 }, { "epoch": 4.31, "learning_rate": 2.53594739708897e-07, "loss": 0.0158, "step": 10941 }, { "epoch": 4.31, "learning_rate": 2.5357014001375237e-07, "loss": 0.0142, "step": 10942 }, { "epoch": 4.31, "learning_rate": 2.5354553944549593e-07, "loss": 0.0145, "step": 10943 }, { "epoch": 4.31, "learning_rate": 2.535209380045285e-07, "loss": 0.0141, "step": 10944 }, { "epoch": 4.31, "learning_rate": 2.5349633569125113e-07, "loss": 0.0174, "step": 10945 }, { "epoch": 4.31, "learning_rate": 2.5347173250606485e-07, "loss": 0.0123, "step": 10946 }, { "epoch": 4.31, "learning_rate": 2.5344712844937056e-07, "loss": 0.0156, "step": 10947 }, { "epoch": 4.31, "learning_rate": 2.534225235215693e-07, "loss": 0.0136, "step": 10948 }, { "epoch": 4.32, "learning_rate": 2.5339791772306215e-07, "loss": 0.0166, "step": 10949 }, { "epoch": 4.32, "learning_rate": 2.5337331105425007e-07, "loss": 0.0176, "step": 10950 }, { "epoch": 4.32, "learning_rate": 2.5334870351553414e-07, "loss": 0.0133, "step": 10951 }, { "epoch": 4.32, "learning_rate": 2.533240951073154e-07, "loss": 0.0141, "step": 10952 }, { "epoch": 4.32, "learning_rate": 2.5329948582999494e-07, "loss": 0.0184, "step": 10953 }, { "epoch": 4.32, "learning_rate": 2.5327487568397396e-07, "loss": 0.0138, "step": 10954 }, { "epoch": 4.32, "learning_rate": 2.532502646696534e-07, "loss": 0.0151, "step": 10955 }, { "epoch": 4.32, "learning_rate": 2.5322565278743444e-07, "loss": 0.0118, "step": 10956 }, { "epoch": 4.32, "learning_rate": 2.532010400377183e-07, "loss": 0.0171, "step": 10957 }, { "epoch": 4.32, "learning_rate": 2.5317642642090597e-07, "loss": 0.0143, "step": 10958 }, { "epoch": 4.32, "learning_rate": 2.531518119373988e-07, "loss": 0.0166, "step": 10959 }, { "epoch": 4.32, "learning_rate": 2.5312719658759786e-07, "loss": 0.0147, "step": 10960 }, { "epoch": 4.32, "learning_rate": 2.531025803719043e-07, "loss": 0.0155, "step": 10961 }, { "epoch": 4.32, "learning_rate": 2.530779632907195e-07, "loss": 0.0103, "step": 10962 }, { "epoch": 4.32, "learning_rate": 2.5305334534444443e-07, "loss": 0.0148, "step": 10963 }, { "epoch": 4.32, "learning_rate": 2.5302872653348053e-07, "loss": 0.0131, "step": 10964 }, { "epoch": 4.32, "learning_rate": 2.53004106858229e-07, "loss": 0.0169, "step": 10965 }, { "epoch": 4.32, "learning_rate": 2.5297948631909104e-07, "loss": 0.0147, "step": 10966 }, { "epoch": 4.32, "learning_rate": 2.5295486491646796e-07, "loss": 0.0155, "step": 10967 }, { "epoch": 4.32, "learning_rate": 2.529302426507611e-07, "loss": 0.0138, "step": 10968 }, { "epoch": 4.32, "learning_rate": 2.529056195223718e-07, "loss": 0.0154, "step": 10969 }, { "epoch": 4.32, "learning_rate": 2.528809955317012e-07, "loss": 0.0123, "step": 10970 }, { "epoch": 4.32, "learning_rate": 2.528563706791508e-07, "loss": 0.0146, "step": 10971 }, { "epoch": 4.32, "learning_rate": 2.5283174496512183e-07, "loss": 0.0199, "step": 10972 }, { "epoch": 4.32, "learning_rate": 2.5280711839001576e-07, "loss": 0.0145, "step": 10973 }, { "epoch": 4.32, "learning_rate": 2.527824909542339e-07, "loss": 0.0183, "step": 10974 }, { "epoch": 4.33, "learning_rate": 2.5275786265817766e-07, "loss": 0.0182, "step": 10975 }, { "epoch": 4.33, "learning_rate": 2.527332335022485e-07, "loss": 0.0188, "step": 10976 }, { "epoch": 4.33, "learning_rate": 2.5270860348684776e-07, "loss": 0.0164, "step": 10977 }, { "epoch": 4.33, "learning_rate": 2.526839726123768e-07, "loss": 0.014, "step": 10978 }, { "epoch": 4.33, "learning_rate": 2.526593408792372e-07, "loss": 0.0133, "step": 10979 }, { "epoch": 4.33, "learning_rate": 2.5263470828783046e-07, "loss": 0.0131, "step": 10980 }, { "epoch": 4.33, "learning_rate": 2.526100748385579e-07, "loss": 0.0145, "step": 10981 }, { "epoch": 4.33, "learning_rate": 2.5258544053182106e-07, "loss": 0.0164, "step": 10982 }, { "epoch": 4.33, "learning_rate": 2.525608053680215e-07, "loss": 0.0168, "step": 10983 }, { "epoch": 4.33, "learning_rate": 2.525361693475607e-07, "loss": 0.0146, "step": 10984 }, { "epoch": 4.33, "learning_rate": 2.5251153247084007e-07, "loss": 0.0164, "step": 10985 }, { "epoch": 4.33, "learning_rate": 2.5248689473826144e-07, "loss": 0.0152, "step": 10986 }, { "epoch": 4.33, "learning_rate": 2.524622561502261e-07, "loss": 0.0176, "step": 10987 }, { "epoch": 4.33, "learning_rate": 2.5243761670713575e-07, "loss": 0.0168, "step": 10988 }, { "epoch": 4.33, "learning_rate": 2.5241297640939195e-07, "loss": 0.0144, "step": 10989 }, { "epoch": 4.33, "learning_rate": 2.5238833525739626e-07, "loss": 0.0142, "step": 10990 }, { "epoch": 4.33, "learning_rate": 2.523636932515504e-07, "loss": 0.0153, "step": 10991 }, { "epoch": 4.33, "learning_rate": 2.5233905039225576e-07, "loss": 0.017, "step": 10992 }, { "epoch": 4.33, "learning_rate": 2.523144066799143e-07, "loss": 0.017, "step": 10993 }, { "epoch": 4.33, "learning_rate": 2.522897621149274e-07, "loss": 0.0122, "step": 10994 }, { "epoch": 4.33, "learning_rate": 2.5226511669769695e-07, "loss": 0.0175, "step": 10995 }, { "epoch": 4.33, "learning_rate": 2.522404704286245e-07, "loss": 0.0174, "step": 10996 }, { "epoch": 4.33, "learning_rate": 2.522158233081118e-07, "loss": 0.0131, "step": 10997 }, { "epoch": 4.33, "learning_rate": 2.521911753365605e-07, "loss": 0.0157, "step": 10998 }, { "epoch": 4.33, "learning_rate": 2.521665265143724e-07, "loss": 0.0206, "step": 10999 }, { "epoch": 4.34, "learning_rate": 2.521418768419492e-07, "loss": 0.0137, "step": 11000 }, { "epoch": 4.34, "learning_rate": 2.5211722631969263e-07, "loss": 0.0138, "step": 11001 }, { "epoch": 4.34, "learning_rate": 2.520925749480045e-07, "loss": 0.0152, "step": 11002 }, { "epoch": 4.34, "learning_rate": 2.520679227272865e-07, "loss": 0.0152, "step": 11003 }, { "epoch": 4.34, "learning_rate": 2.520432696579406e-07, "loss": 0.0168, "step": 11004 }, { "epoch": 4.34, "learning_rate": 2.520186157403684e-07, "loss": 0.0162, "step": 11005 }, { "epoch": 4.34, "learning_rate": 2.519939609749719e-07, "loss": 0.0146, "step": 11006 }, { "epoch": 4.34, "learning_rate": 2.519693053621528e-07, "loss": 0.0133, "step": 11007 }, { "epoch": 4.34, "learning_rate": 2.5194464890231306e-07, "loss": 0.0161, "step": 11008 }, { "epoch": 4.34, "learning_rate": 2.519199915958545e-07, "loss": 0.0161, "step": 11009 }, { "epoch": 4.34, "learning_rate": 2.5189533344317895e-07, "loss": 0.0168, "step": 11010 }, { "epoch": 4.34, "learning_rate": 2.5187067444468833e-07, "loss": 0.0154, "step": 11011 }, { "epoch": 4.34, "learning_rate": 2.5184601460078455e-07, "loss": 0.0131, "step": 11012 }, { "epoch": 4.34, "learning_rate": 2.518213539118696e-07, "loss": 0.014, "step": 11013 }, { "epoch": 4.34, "learning_rate": 2.517966923783452e-07, "loss": 0.0179, "step": 11014 }, { "epoch": 4.34, "learning_rate": 2.517720300006135e-07, "loss": 0.0138, "step": 11015 }, { "epoch": 4.34, "learning_rate": 2.5174736677907644e-07, "loss": 0.0159, "step": 11016 }, { "epoch": 4.34, "learning_rate": 2.5172270271413595e-07, "loss": 0.0143, "step": 11017 }, { "epoch": 4.34, "learning_rate": 2.51698037806194e-07, "loss": 0.0178, "step": 11018 }, { "epoch": 4.34, "learning_rate": 2.516733720556526e-07, "loss": 0.0159, "step": 11019 }, { "epoch": 4.34, "learning_rate": 2.5164870546291374e-07, "loss": 0.0155, "step": 11020 }, { "epoch": 4.34, "learning_rate": 2.5162403802837953e-07, "loss": 0.0144, "step": 11021 }, { "epoch": 4.34, "learning_rate": 2.515993697524519e-07, "loss": 0.0177, "step": 11022 }, { "epoch": 4.34, "learning_rate": 2.51574700635533e-07, "loss": 0.0174, "step": 11023 }, { "epoch": 4.34, "learning_rate": 2.5155003067802484e-07, "loss": 0.0161, "step": 11024 }, { "epoch": 4.35, "learning_rate": 2.515253598803295e-07, "loss": 0.0173, "step": 11025 }, { "epoch": 4.35, "learning_rate": 2.515006882428492e-07, "loss": 0.0154, "step": 11026 }, { "epoch": 4.35, "learning_rate": 2.514760157659859e-07, "loss": 0.0151, "step": 11027 }, { "epoch": 4.35, "learning_rate": 2.514513424501418e-07, "loss": 0.0153, "step": 11028 }, { "epoch": 4.35, "learning_rate": 2.51426668295719e-07, "loss": 0.0146, "step": 11029 }, { "epoch": 4.35, "learning_rate": 2.5140199330311965e-07, "loss": 0.0142, "step": 11030 }, { "epoch": 4.35, "learning_rate": 2.5137731747274595e-07, "loss": 0.0152, "step": 11031 }, { "epoch": 4.35, "learning_rate": 2.5135264080500003e-07, "loss": 0.0183, "step": 11032 }, { "epoch": 4.35, "learning_rate": 2.513279633002841e-07, "loss": 0.0155, "step": 11033 }, { "epoch": 4.35, "learning_rate": 2.513032849590004e-07, "loss": 0.0205, "step": 11034 }, { "epoch": 4.35, "learning_rate": 2.512786057815512e-07, "loss": 0.0132, "step": 11035 }, { "epoch": 4.35, "learning_rate": 2.5125392576833853e-07, "loss": 0.0145, "step": 11036 }, { "epoch": 4.35, "learning_rate": 2.512292449197649e-07, "loss": 0.013, "step": 11037 }, { "epoch": 4.35, "learning_rate": 2.5120456323623236e-07, "loss": 0.0149, "step": 11038 }, { "epoch": 4.35, "learning_rate": 2.5117988071814327e-07, "loss": 0.0142, "step": 11039 }, { "epoch": 4.35, "learning_rate": 2.5115519736589994e-07, "loss": 0.0157, "step": 11040 }, { "epoch": 4.35, "learning_rate": 2.511305131799046e-07, "loss": 0.0144, "step": 11041 }, { "epoch": 4.35, "learning_rate": 2.5110582816055963e-07, "loss": 0.0148, "step": 11042 }, { "epoch": 4.35, "learning_rate": 2.5108114230826733e-07, "loss": 0.0187, "step": 11043 }, { "epoch": 4.35, "learning_rate": 2.510564556234301e-07, "loss": 0.0142, "step": 11044 }, { "epoch": 4.35, "learning_rate": 2.510317681064501e-07, "loss": 0.0169, "step": 11045 }, { "epoch": 4.35, "learning_rate": 2.5100707975772996e-07, "loss": 0.0155, "step": 11046 }, { "epoch": 4.35, "learning_rate": 2.5098239057767184e-07, "loss": 0.0158, "step": 11047 }, { "epoch": 4.35, "learning_rate": 2.509577005666783e-07, "loss": 0.0167, "step": 11048 }, { "epoch": 4.35, "learning_rate": 2.5093300972515173e-07, "loss": 0.017, "step": 11049 }, { "epoch": 4.35, "learning_rate": 2.509083180534944e-07, "loss": 0.0154, "step": 11050 }, { "epoch": 4.36, "learning_rate": 2.5088362555210894e-07, "loss": 0.0131, "step": 11051 }, { "epoch": 4.36, "learning_rate": 2.508589322213977e-07, "loss": 0.0138, "step": 11052 }, { "epoch": 4.36, "learning_rate": 2.508342380617632e-07, "loss": 0.0128, "step": 11053 }, { "epoch": 4.36, "learning_rate": 2.508095430736078e-07, "loss": 0.0139, "step": 11054 }, { "epoch": 4.36, "learning_rate": 2.5078484725733415e-07, "loss": 0.0158, "step": 11055 }, { "epoch": 4.36, "learning_rate": 2.5076015061334457e-07, "loss": 0.0165, "step": 11056 }, { "epoch": 4.36, "learning_rate": 2.5073545314204176e-07, "loss": 0.0154, "step": 11057 }, { "epoch": 4.36, "learning_rate": 2.5071075484382813e-07, "loss": 0.0199, "step": 11058 }, { "epoch": 4.36, "learning_rate": 2.506860557191063e-07, "loss": 0.017, "step": 11059 }, { "epoch": 4.36, "learning_rate": 2.5066135576827883e-07, "loss": 0.0144, "step": 11060 }, { "epoch": 4.36, "learning_rate": 2.506366549917482e-07, "loss": 0.0136, "step": 11061 }, { "epoch": 4.36, "learning_rate": 2.506119533899171e-07, "loss": 0.0153, "step": 11062 }, { "epoch": 4.36, "learning_rate": 2.50587250963188e-07, "loss": 0.0163, "step": 11063 }, { "epoch": 4.36, "learning_rate": 2.505625477119637e-07, "loss": 0.0171, "step": 11064 }, { "epoch": 4.36, "learning_rate": 2.505378436366466e-07, "loss": 0.0149, "step": 11065 }, { "epoch": 4.36, "learning_rate": 2.505131387376396e-07, "loss": 0.0175, "step": 11066 }, { "epoch": 4.36, "learning_rate": 2.5048843301534506e-07, "loss": 0.0159, "step": 11067 }, { "epoch": 4.36, "learning_rate": 2.504637264701659e-07, "loss": 0.015, "step": 11068 }, { "epoch": 4.36, "learning_rate": 2.5043901910250464e-07, "loss": 0.0132, "step": 11069 }, { "epoch": 4.36, "learning_rate": 2.504143109127641e-07, "loss": 0.0159, "step": 11070 }, { "epoch": 4.36, "learning_rate": 2.503896019013468e-07, "loss": 0.0115, "step": 11071 }, { "epoch": 4.36, "learning_rate": 2.503648920686556e-07, "loss": 0.0147, "step": 11072 }, { "epoch": 4.36, "learning_rate": 2.503401814150933e-07, "loss": 0.015, "step": 11073 }, { "epoch": 4.36, "learning_rate": 2.5031546994106245e-07, "loss": 0.0132, "step": 11074 }, { "epoch": 4.36, "learning_rate": 2.50290757646966e-07, "loss": 0.0128, "step": 11075 }, { "epoch": 4.37, "learning_rate": 2.502660445332065e-07, "loss": 0.016, "step": 11076 }, { "epoch": 4.37, "learning_rate": 2.50241330600187e-07, "loss": 0.0153, "step": 11077 }, { "epoch": 4.37, "learning_rate": 2.502166158483101e-07, "loss": 0.0126, "step": 11078 }, { "epoch": 4.37, "learning_rate": 2.501919002779787e-07, "loss": 0.0125, "step": 11079 }, { "epoch": 4.37, "learning_rate": 2.5016718388959563e-07, "loss": 0.0134, "step": 11080 }, { "epoch": 4.37, "learning_rate": 2.501424666835637e-07, "loss": 0.016, "step": 11081 }, { "epoch": 4.37, "learning_rate": 2.5011774866028574e-07, "loss": 0.0143, "step": 11082 }, { "epoch": 4.37, "learning_rate": 2.500930298201647e-07, "loss": 0.018, "step": 11083 }, { "epoch": 4.37, "learning_rate": 2.500683101636034e-07, "loss": 0.0132, "step": 11084 }, { "epoch": 4.37, "learning_rate": 2.500435896910047e-07, "loss": 0.0154, "step": 11085 }, { "epoch": 4.37, "learning_rate": 2.5001886840277156e-07, "loss": 0.016, "step": 11086 }, { "epoch": 4.37, "learning_rate": 2.4999414629930686e-07, "loss": 0.0164, "step": 11087 }, { "epoch": 4.37, "learning_rate": 2.499694233810136e-07, "loss": 0.015, "step": 11088 }, { "epoch": 4.37, "learning_rate": 2.499446996482947e-07, "loss": 0.015, "step": 11089 }, { "epoch": 4.37, "learning_rate": 2.49919975101553e-07, "loss": 0.0149, "step": 11090 }, { "epoch": 4.37, "learning_rate": 2.498952497411917e-07, "loss": 0.0149, "step": 11091 }, { "epoch": 4.37, "learning_rate": 2.4987052356761355e-07, "loss": 0.0149, "step": 11092 }, { "epoch": 4.37, "learning_rate": 2.4984579658122175e-07, "loss": 0.0157, "step": 11093 }, { "epoch": 4.37, "learning_rate": 2.4982106878241916e-07, "loss": 0.0195, "step": 11094 }, { "epoch": 4.37, "learning_rate": 2.497963401716089e-07, "loss": 0.0142, "step": 11095 }, { "epoch": 4.37, "learning_rate": 2.49771610749194e-07, "loss": 0.0117, "step": 11096 }, { "epoch": 4.37, "learning_rate": 2.4974688051557745e-07, "loss": 0.0158, "step": 11097 }, { "epoch": 4.37, "learning_rate": 2.4972214947116234e-07, "loss": 0.0144, "step": 11098 }, { "epoch": 4.37, "learning_rate": 2.496974176163518e-07, "loss": 0.0154, "step": 11099 }, { "epoch": 4.37, "learning_rate": 2.4967268495154885e-07, "loss": 0.0149, "step": 11100 }, { "epoch": 4.37, "learning_rate": 2.4964795147715663e-07, "loss": 0.0155, "step": 11101 }, { "epoch": 4.38, "learning_rate": 2.496232171935783e-07, "loss": 0.0146, "step": 11102 }, { "epoch": 4.38, "learning_rate": 2.495984821012169e-07, "loss": 0.0131, "step": 11103 }, { "epoch": 4.38, "learning_rate": 2.495737462004757e-07, "loss": 0.0166, "step": 11104 }, { "epoch": 4.38, "learning_rate": 2.4954900949175775e-07, "loss": 0.0141, "step": 11105 }, { "epoch": 4.38, "learning_rate": 2.4952427197546623e-07, "loss": 0.0161, "step": 11106 }, { "epoch": 4.38, "learning_rate": 2.4949953365200436e-07, "loss": 0.0133, "step": 11107 }, { "epoch": 4.38, "learning_rate": 2.494747945217753e-07, "loss": 0.0146, "step": 11108 }, { "epoch": 4.38, "learning_rate": 2.4945005458518233e-07, "loss": 0.017, "step": 11109 }, { "epoch": 4.38, "learning_rate": 2.4942531384262856e-07, "loss": 0.0173, "step": 11110 }, { "epoch": 4.38, "learning_rate": 2.4940057229451733e-07, "loss": 0.0165, "step": 11111 }, { "epoch": 4.38, "learning_rate": 2.493758299412518e-07, "loss": 0.0184, "step": 11112 }, { "epoch": 4.38, "learning_rate": 2.4935108678323534e-07, "loss": 0.013, "step": 11113 }, { "epoch": 4.38, "learning_rate": 2.493263428208711e-07, "loss": 0.0132, "step": 11114 }, { "epoch": 4.38, "learning_rate": 2.4930159805456254e-07, "loss": 0.0131, "step": 11115 }, { "epoch": 4.38, "learning_rate": 2.4927685248471284e-07, "loss": 0.0165, "step": 11116 }, { "epoch": 4.38, "learning_rate": 2.492521061117253e-07, "loss": 0.0152, "step": 11117 }, { "epoch": 4.38, "learning_rate": 2.4922735893600324e-07, "loss": 0.0131, "step": 11118 }, { "epoch": 4.38, "learning_rate": 2.4920261095795014e-07, "loss": 0.0152, "step": 11119 }, { "epoch": 4.38, "learning_rate": 2.4917786217796923e-07, "loss": 0.0158, "step": 11120 }, { "epoch": 4.38, "learning_rate": 2.4915311259646386e-07, "loss": 0.0177, "step": 11121 }, { "epoch": 4.38, "learning_rate": 2.4912836221383754e-07, "loss": 0.0179, "step": 11122 }, { "epoch": 4.38, "learning_rate": 2.4910361103049355e-07, "loss": 0.0137, "step": 11123 }, { "epoch": 4.38, "learning_rate": 2.4907885904683536e-07, "loss": 0.0165, "step": 11124 }, { "epoch": 4.38, "learning_rate": 2.490541062632663e-07, "loss": 0.0155, "step": 11125 }, { "epoch": 4.38, "learning_rate": 2.490293526801899e-07, "loss": 0.0189, "step": 11126 }, { "epoch": 4.39, "learning_rate": 2.490045982980095e-07, "loss": 0.0167, "step": 11127 }, { "epoch": 4.39, "learning_rate": 2.4897984311712873e-07, "loss": 0.0162, "step": 11128 }, { "epoch": 4.39, "learning_rate": 2.489550871379509e-07, "loss": 0.0144, "step": 11129 }, { "epoch": 4.39, "learning_rate": 2.489303303608796e-07, "loss": 0.0131, "step": 11130 }, { "epoch": 4.39, "learning_rate": 2.489055727863182e-07, "loss": 0.0155, "step": 11131 }, { "epoch": 4.39, "learning_rate": 2.488808144146703e-07, "loss": 0.0146, "step": 11132 }, { "epoch": 4.39, "learning_rate": 2.488560552463395e-07, "loss": 0.0185, "step": 11133 }, { "epoch": 4.39, "learning_rate": 2.4883129528172916e-07, "loss": 0.017, "step": 11134 }, { "epoch": 4.39, "learning_rate": 2.48806534521243e-07, "loss": 0.0148, "step": 11135 }, { "epoch": 4.39, "learning_rate": 2.487817729652844e-07, "loss": 0.0162, "step": 11136 }, { "epoch": 4.39, "learning_rate": 2.4875701061425714e-07, "loss": 0.015, "step": 11137 }, { "epoch": 4.39, "learning_rate": 2.4873224746856463e-07, "loss": 0.0148, "step": 11138 }, { "epoch": 4.39, "learning_rate": 2.4870748352861064e-07, "loss": 0.0182, "step": 11139 }, { "epoch": 4.39, "learning_rate": 2.486827187947987e-07, "loss": 0.0133, "step": 11140 }, { "epoch": 4.39, "learning_rate": 2.4865795326753233e-07, "loss": 0.0154, "step": 11141 }, { "epoch": 4.39, "learning_rate": 2.4863318694721533e-07, "loss": 0.0181, "step": 11142 }, { "epoch": 4.39, "learning_rate": 2.486084198342513e-07, "loss": 0.0166, "step": 11143 }, { "epoch": 4.39, "learning_rate": 2.4858365192904387e-07, "loss": 0.0182, "step": 11144 }, { "epoch": 4.39, "learning_rate": 2.4855888323199673e-07, "loss": 0.016, "step": 11145 }, { "epoch": 4.39, "learning_rate": 2.4853411374351367e-07, "loss": 0.0124, "step": 11146 }, { "epoch": 4.39, "learning_rate": 2.4850934346399834e-07, "loss": 0.0145, "step": 11147 }, { "epoch": 4.39, "learning_rate": 2.4848457239385436e-07, "loss": 0.014, "step": 11148 }, { "epoch": 4.39, "learning_rate": 2.4845980053348555e-07, "loss": 0.0177, "step": 11149 }, { "epoch": 4.39, "learning_rate": 2.4843502788329573e-07, "loss": 0.0132, "step": 11150 }, { "epoch": 4.39, "learning_rate": 2.4841025444368846e-07, "loss": 0.0154, "step": 11151 }, { "epoch": 4.4, "learning_rate": 2.483854802150676e-07, "loss": 0.016, "step": 11152 }, { "epoch": 4.4, "learning_rate": 2.483607051978371e-07, "loss": 0.0156, "step": 11153 }, { "epoch": 4.4, "learning_rate": 2.4833592939240047e-07, "loss": 0.0132, "step": 11154 }, { "epoch": 4.4, "learning_rate": 2.483111527991617e-07, "loss": 0.0163, "step": 11155 }, { "epoch": 4.4, "learning_rate": 2.482863754185245e-07, "loss": 0.0162, "step": 11156 }, { "epoch": 4.4, "learning_rate": 2.482615972508929e-07, "loss": 0.0139, "step": 11157 }, { "epoch": 4.4, "learning_rate": 2.482368182966705e-07, "loss": 0.0137, "step": 11158 }, { "epoch": 4.4, "learning_rate": 2.4821203855626136e-07, "loss": 0.0169, "step": 11159 }, { "epoch": 4.4, "learning_rate": 2.481872580300692e-07, "loss": 0.0134, "step": 11160 }, { "epoch": 4.4, "learning_rate": 2.481624767184979e-07, "loss": 0.0197, "step": 11161 }, { "epoch": 4.4, "learning_rate": 2.481376946219516e-07, "loss": 0.0185, "step": 11162 }, { "epoch": 4.4, "learning_rate": 2.481129117408339e-07, "loss": 0.0146, "step": 11163 }, { "epoch": 4.4, "learning_rate": 2.4808812807554897e-07, "loss": 0.0143, "step": 11164 }, { "epoch": 4.4, "learning_rate": 2.480633436265005e-07, "loss": 0.0119, "step": 11165 }, { "epoch": 4.4, "learning_rate": 2.480385583940927e-07, "loss": 0.0159, "step": 11166 }, { "epoch": 4.4, "learning_rate": 2.480137723787294e-07, "loss": 0.0152, "step": 11167 }, { "epoch": 4.4, "learning_rate": 2.479889855808145e-07, "loss": 0.016, "step": 11168 }, { "epoch": 4.4, "learning_rate": 2.4796419800075215e-07, "loss": 0.0129, "step": 11169 }, { "epoch": 4.4, "learning_rate": 2.4793940963894624e-07, "loss": 0.0152, "step": 11170 }, { "epoch": 4.4, "learning_rate": 2.479146204958008e-07, "loss": 0.0168, "step": 11171 }, { "epoch": 4.4, "learning_rate": 2.4788983057171997e-07, "loss": 0.0154, "step": 11172 }, { "epoch": 4.4, "learning_rate": 2.478650398671076e-07, "loss": 0.0132, "step": 11173 }, { "epoch": 4.4, "learning_rate": 2.478402483823678e-07, "loss": 0.0132, "step": 11174 }, { "epoch": 4.4, "learning_rate": 2.4781545611790464e-07, "loss": 0.012, "step": 11175 }, { "epoch": 4.4, "learning_rate": 2.4779066307412226e-07, "loss": 0.0131, "step": 11176 }, { "epoch": 4.4, "learning_rate": 2.4776586925142473e-07, "loss": 0.0133, "step": 11177 }, { "epoch": 4.41, "learning_rate": 2.4774107465021605e-07, "loss": 0.0165, "step": 11178 }, { "epoch": 4.41, "learning_rate": 2.477162792709005e-07, "loss": 0.013, "step": 11179 }, { "epoch": 4.41, "learning_rate": 2.476914831138821e-07, "loss": 0.0148, "step": 11180 }, { "epoch": 4.41, "learning_rate": 2.4766668617956494e-07, "loss": 0.0171, "step": 11181 }, { "epoch": 4.41, "learning_rate": 2.4764188846835335e-07, "loss": 0.0157, "step": 11182 }, { "epoch": 4.41, "learning_rate": 2.476170899806513e-07, "loss": 0.0161, "step": 11183 }, { "epoch": 4.41, "learning_rate": 2.475922907168631e-07, "loss": 0.0134, "step": 11184 }, { "epoch": 4.41, "learning_rate": 2.4756749067739283e-07, "loss": 0.0157, "step": 11185 }, { "epoch": 4.41, "learning_rate": 2.4754268986264484e-07, "loss": 0.0153, "step": 11186 }, { "epoch": 4.41, "learning_rate": 2.475178882730232e-07, "loss": 0.0144, "step": 11187 }, { "epoch": 4.41, "learning_rate": 2.4749308590893223e-07, "loss": 0.0163, "step": 11188 }, { "epoch": 4.41, "learning_rate": 2.4746828277077615e-07, "loss": 0.0138, "step": 11189 }, { "epoch": 4.41, "learning_rate": 2.4744347885895913e-07, "loss": 0.0176, "step": 11190 }, { "epoch": 4.41, "learning_rate": 2.474186741738856e-07, "loss": 0.0153, "step": 11191 }, { "epoch": 4.41, "learning_rate": 2.4739386871595973e-07, "loss": 0.0143, "step": 11192 }, { "epoch": 4.41, "learning_rate": 2.473690624855858e-07, "loss": 0.0159, "step": 11193 }, { "epoch": 4.41, "learning_rate": 2.4734425548316815e-07, "loss": 0.0124, "step": 11194 }, { "epoch": 4.41, "learning_rate": 2.4731944770911113e-07, "loss": 0.0163, "step": 11195 }, { "epoch": 4.41, "learning_rate": 2.4729463916381896e-07, "loss": 0.0141, "step": 11196 }, { "epoch": 4.41, "learning_rate": 2.472698298476961e-07, "loss": 0.0166, "step": 11197 }, { "epoch": 4.41, "learning_rate": 2.472450197611468e-07, "loss": 0.0157, "step": 11198 }, { "epoch": 4.41, "learning_rate": 2.472202089045756e-07, "loss": 0.014, "step": 11199 }, { "epoch": 4.41, "learning_rate": 2.4719539727838666e-07, "loss": 0.0164, "step": 11200 }, { "epoch": 4.41, "learning_rate": 2.471705848829845e-07, "loss": 0.0154, "step": 11201 }, { "epoch": 4.41, "learning_rate": 2.471457717187735e-07, "loss": 0.0153, "step": 11202 }, { "epoch": 4.42, "learning_rate": 2.4712095778615805e-07, "loss": 0.016, "step": 11203 }, { "epoch": 4.42, "learning_rate": 2.4709614308554267e-07, "loss": 0.0135, "step": 11204 }, { "epoch": 4.42, "learning_rate": 2.470713276173317e-07, "loss": 0.0141, "step": 11205 }, { "epoch": 4.42, "learning_rate": 2.470465113819296e-07, "loss": 0.0145, "step": 11206 }, { "epoch": 4.42, "learning_rate": 2.4702169437974086e-07, "loss": 0.0158, "step": 11207 }, { "epoch": 4.42, "learning_rate": 2.4699687661117003e-07, "loss": 0.0151, "step": 11208 }, { "epoch": 4.42, "learning_rate": 2.4697205807662154e-07, "loss": 0.0129, "step": 11209 }, { "epoch": 4.42, "learning_rate": 2.4694723877649984e-07, "loss": 0.0156, "step": 11210 }, { "epoch": 4.42, "learning_rate": 2.469224187112095e-07, "loss": 0.0148, "step": 11211 }, { "epoch": 4.42, "learning_rate": 2.46897597881155e-07, "loss": 0.0162, "step": 11212 }, { "epoch": 4.42, "learning_rate": 2.468727762867411e-07, "loss": 0.015, "step": 11213 }, { "epoch": 4.42, "learning_rate": 2.4684795392837204e-07, "loss": 0.0141, "step": 11214 }, { "epoch": 4.42, "learning_rate": 2.468231308064526e-07, "loss": 0.0139, "step": 11215 }, { "epoch": 4.42, "learning_rate": 2.4679830692138726e-07, "loss": 0.0177, "step": 11216 }, { "epoch": 4.42, "learning_rate": 2.4677348227358064e-07, "loss": 0.0165, "step": 11217 }, { "epoch": 4.42, "learning_rate": 2.4674865686343733e-07, "loss": 0.0142, "step": 11218 }, { "epoch": 4.42, "learning_rate": 2.46723830691362e-07, "loss": 0.0186, "step": 11219 }, { "epoch": 4.42, "learning_rate": 2.4669900375775926e-07, "loss": 0.0156, "step": 11220 }, { "epoch": 4.42, "learning_rate": 2.466741760630337e-07, "loss": 0.0153, "step": 11221 }, { "epoch": 4.42, "learning_rate": 2.4664934760759004e-07, "loss": 0.0127, "step": 11222 }, { "epoch": 4.42, "learning_rate": 2.4662451839183283e-07, "loss": 0.0192, "step": 11223 }, { "epoch": 4.42, "learning_rate": 2.4659968841616685e-07, "loss": 0.0142, "step": 11224 }, { "epoch": 4.42, "learning_rate": 2.465748576809968e-07, "loss": 0.0163, "step": 11225 }, { "epoch": 4.42, "learning_rate": 2.465500261867274e-07, "loss": 0.0138, "step": 11226 }, { "epoch": 4.42, "learning_rate": 2.4652519393376325e-07, "loss": 0.0138, "step": 11227 }, { "epoch": 4.43, "learning_rate": 2.4650036092250917e-07, "loss": 0.0184, "step": 11228 }, { "epoch": 4.43, "learning_rate": 2.464755271533699e-07, "loss": 0.0132, "step": 11229 }, { "epoch": 4.43, "learning_rate": 2.464506926267502e-07, "loss": 0.0171, "step": 11230 }, { "epoch": 4.43, "learning_rate": 2.464258573430547e-07, "loss": 0.0137, "step": 11231 }, { "epoch": 4.43, "learning_rate": 2.4640102130268837e-07, "loss": 0.0175, "step": 11232 }, { "epoch": 4.43, "learning_rate": 2.463761845060559e-07, "loss": 0.0134, "step": 11233 }, { "epoch": 4.43, "learning_rate": 2.463513469535621e-07, "loss": 0.014, "step": 11234 }, { "epoch": 4.43, "learning_rate": 2.463265086456118e-07, "loss": 0.0167, "step": 11235 }, { "epoch": 4.43, "learning_rate": 2.4630166958260985e-07, "loss": 0.0181, "step": 11236 }, { "epoch": 4.43, "learning_rate": 2.46276829764961e-07, "loss": 0.0172, "step": 11237 }, { "epoch": 4.43, "learning_rate": 2.4625198919307014e-07, "loss": 0.0134, "step": 11238 }, { "epoch": 4.43, "learning_rate": 2.462271478673422e-07, "loss": 0.0174, "step": 11239 }, { "epoch": 4.43, "learning_rate": 2.46202305788182e-07, "loss": 0.0157, "step": 11240 }, { "epoch": 4.43, "learning_rate": 2.461774629559944e-07, "loss": 0.016, "step": 11241 }, { "epoch": 4.43, "learning_rate": 2.4615261937118444e-07, "loss": 0.0155, "step": 11242 }, { "epoch": 4.43, "learning_rate": 2.4612777503415684e-07, "loss": 0.0149, "step": 11243 }, { "epoch": 4.43, "learning_rate": 2.461029299453166e-07, "loss": 0.0144, "step": 11244 }, { "epoch": 4.43, "learning_rate": 2.4607808410506864e-07, "loss": 0.0201, "step": 11245 }, { "epoch": 4.43, "learning_rate": 2.4605323751381807e-07, "loss": 0.0165, "step": 11246 }, { "epoch": 4.43, "learning_rate": 2.460283901719696e-07, "loss": 0.0133, "step": 11247 }, { "epoch": 4.43, "learning_rate": 2.4600354207992835e-07, "loss": 0.0134, "step": 11248 }, { "epoch": 4.43, "learning_rate": 2.4597869323809933e-07, "loss": 0.0175, "step": 11249 }, { "epoch": 4.43, "learning_rate": 2.4595384364688745e-07, "loss": 0.0146, "step": 11250 }, { "epoch": 4.43, "learning_rate": 2.459289933066978e-07, "loss": 0.0218, "step": 11251 }, { "epoch": 4.43, "learning_rate": 2.459041422179353e-07, "loss": 0.0139, "step": 11252 }, { "epoch": 4.43, "learning_rate": 2.458792903810051e-07, "loss": 0.0154, "step": 11253 }, { "epoch": 4.44, "learning_rate": 2.4585443779631213e-07, "loss": 0.0158, "step": 11254 }, { "epoch": 4.44, "learning_rate": 2.4582958446426156e-07, "loss": 0.0164, "step": 11255 }, { "epoch": 4.44, "learning_rate": 2.458047303852585e-07, "loss": 0.0168, "step": 11256 }, { "epoch": 4.44, "learning_rate": 2.457798755597079e-07, "loss": 0.0141, "step": 11257 }, { "epoch": 4.44, "learning_rate": 2.4575501998801484e-07, "loss": 0.0154, "step": 11258 }, { "epoch": 4.44, "learning_rate": 2.457301636705845e-07, "loss": 0.0172, "step": 11259 }, { "epoch": 4.44, "learning_rate": 2.4570530660782207e-07, "loss": 0.0161, "step": 11260 }, { "epoch": 4.44, "learning_rate": 2.456804488001325e-07, "loss": 0.0142, "step": 11261 }, { "epoch": 4.44, "learning_rate": 2.456555902479212e-07, "loss": 0.0133, "step": 11262 }, { "epoch": 4.44, "learning_rate": 2.45630730951593e-07, "loss": 0.0188, "step": 11263 }, { "epoch": 4.44, "learning_rate": 2.456058709115533e-07, "loss": 0.0137, "step": 11264 }, { "epoch": 4.44, "learning_rate": 2.455810101282072e-07, "loss": 0.0173, "step": 11265 }, { "epoch": 4.44, "learning_rate": 2.4555614860195994e-07, "loss": 0.0119, "step": 11266 }, { "epoch": 4.44, "learning_rate": 2.455312863332167e-07, "loss": 0.0135, "step": 11267 }, { "epoch": 4.44, "learning_rate": 2.4550642332238267e-07, "loss": 0.0156, "step": 11268 }, { "epoch": 4.44, "learning_rate": 2.454815595698631e-07, "loss": 0.0144, "step": 11269 }, { "epoch": 4.44, "learning_rate": 2.4545669507606325e-07, "loss": 0.0152, "step": 11270 }, { "epoch": 4.44, "learning_rate": 2.4543182984138833e-07, "loss": 0.0147, "step": 11271 }, { "epoch": 4.44, "learning_rate": 2.454069638662437e-07, "loss": 0.0164, "step": 11272 }, { "epoch": 4.44, "learning_rate": 2.4538209715103447e-07, "loss": 0.013, "step": 11273 }, { "epoch": 4.44, "learning_rate": 2.453572296961661e-07, "loss": 0.0169, "step": 11274 }, { "epoch": 4.44, "learning_rate": 2.453323615020438e-07, "loss": 0.017, "step": 11275 }, { "epoch": 4.44, "learning_rate": 2.4530749256907286e-07, "loss": 0.0143, "step": 11276 }, { "epoch": 4.44, "learning_rate": 2.4528262289765875e-07, "loss": 0.017, "step": 11277 }, { "epoch": 4.44, "learning_rate": 2.4525775248820667e-07, "loss": 0.0135, "step": 11278 }, { "epoch": 4.45, "learning_rate": 2.4523288134112203e-07, "loss": 0.0186, "step": 11279 }, { "epoch": 4.45, "learning_rate": 2.452080094568101e-07, "loss": 0.0192, "step": 11280 }, { "epoch": 4.45, "learning_rate": 2.451831368356764e-07, "loss": 0.0131, "step": 11281 }, { "epoch": 4.45, "learning_rate": 2.451582634781262e-07, "loss": 0.0129, "step": 11282 }, { "epoch": 4.45, "learning_rate": 2.451333893845649e-07, "loss": 0.0116, "step": 11283 }, { "epoch": 4.45, "learning_rate": 2.45108514555398e-07, "loss": 0.0129, "step": 11284 }, { "epoch": 4.45, "learning_rate": 2.450836389910308e-07, "loss": 0.0139, "step": 11285 }, { "epoch": 4.45, "learning_rate": 2.450587626918689e-07, "loss": 0.0168, "step": 11286 }, { "epoch": 4.45, "learning_rate": 2.450338856583176e-07, "loss": 0.0189, "step": 11287 }, { "epoch": 4.45, "learning_rate": 2.450090078907824e-07, "loss": 0.0145, "step": 11288 }, { "epoch": 4.45, "learning_rate": 2.449841293896688e-07, "loss": 0.0149, "step": 11289 }, { "epoch": 4.45, "learning_rate": 2.4495925015538224e-07, "loss": 0.0156, "step": 11290 }, { "epoch": 4.45, "learning_rate": 2.449343701883283e-07, "loss": 0.0151, "step": 11291 }, { "epoch": 4.45, "learning_rate": 2.449094894889123e-07, "loss": 0.0152, "step": 11292 }, { "epoch": 4.45, "learning_rate": 2.4488460805753995e-07, "loss": 0.0174, "step": 11293 }, { "epoch": 4.45, "learning_rate": 2.4485972589461666e-07, "loss": 0.0156, "step": 11294 }, { "epoch": 4.45, "learning_rate": 2.448348430005481e-07, "loss": 0.0171, "step": 11295 }, { "epoch": 4.45, "learning_rate": 2.448099593757397e-07, "loss": 0.0155, "step": 11296 }, { "epoch": 4.45, "learning_rate": 2.44785075020597e-07, "loss": 0.0121, "step": 11297 }, { "epoch": 4.45, "learning_rate": 2.4476018993552563e-07, "loss": 0.018, "step": 11298 }, { "epoch": 4.45, "learning_rate": 2.4473530412093134e-07, "loss": 0.0164, "step": 11299 }, { "epoch": 4.45, "learning_rate": 2.447104175772195e-07, "loss": 0.0128, "step": 11300 }, { "epoch": 4.45, "learning_rate": 2.4468553030479585e-07, "loss": 0.0159, "step": 11301 }, { "epoch": 4.45, "learning_rate": 2.446606423040659e-07, "loss": 0.0145, "step": 11302 }, { "epoch": 4.45, "learning_rate": 2.446357535754354e-07, "loss": 0.0149, "step": 11303 }, { "epoch": 4.45, "learning_rate": 2.4461086411930997e-07, "loss": 0.0177, "step": 11304 }, { "epoch": 4.46, "learning_rate": 2.445859739360951e-07, "loss": 0.0164, "step": 11305 }, { "epoch": 4.46, "learning_rate": 2.4456108302619685e-07, "loss": 0.0113, "step": 11306 }, { "epoch": 4.46, "learning_rate": 2.4453619139002055e-07, "loss": 0.0157, "step": 11307 }, { "epoch": 4.46, "learning_rate": 2.44511299027972e-07, "loss": 0.0127, "step": 11308 }, { "epoch": 4.46, "learning_rate": 2.44486405940457e-07, "loss": 0.0142, "step": 11309 }, { "epoch": 4.46, "learning_rate": 2.444615121278811e-07, "loss": 0.0163, "step": 11310 }, { "epoch": 4.46, "learning_rate": 2.444366175906502e-07, "loss": 0.0159, "step": 11311 }, { "epoch": 4.46, "learning_rate": 2.444117223291699e-07, "loss": 0.0135, "step": 11312 }, { "epoch": 4.46, "learning_rate": 2.4438682634384615e-07, "loss": 0.014, "step": 11313 }, { "epoch": 4.46, "learning_rate": 2.4436192963508445e-07, "loss": 0.0156, "step": 11314 }, { "epoch": 4.46, "learning_rate": 2.4433703220329083e-07, "loss": 0.0126, "step": 11315 }, { "epoch": 4.46, "learning_rate": 2.443121340488709e-07, "loss": 0.0177, "step": 11316 }, { "epoch": 4.46, "learning_rate": 2.4428723517223055e-07, "loss": 0.0116, "step": 11317 }, { "epoch": 4.46, "learning_rate": 2.4426233557377556e-07, "loss": 0.0131, "step": 11318 }, { "epoch": 4.46, "learning_rate": 2.4423743525391177e-07, "loss": 0.0138, "step": 11319 }, { "epoch": 4.46, "learning_rate": 2.4421253421304506e-07, "loss": 0.0175, "step": 11320 }, { "epoch": 4.46, "learning_rate": 2.441876324515812e-07, "loss": 0.014, "step": 11321 }, { "epoch": 4.46, "learning_rate": 2.4416272996992614e-07, "loss": 0.0163, "step": 11322 }, { "epoch": 4.46, "learning_rate": 2.441378267684857e-07, "loss": 0.0131, "step": 11323 }, { "epoch": 4.46, "learning_rate": 2.441129228476657e-07, "loss": 0.0142, "step": 11324 }, { "epoch": 4.46, "learning_rate": 2.440880182078721e-07, "loss": 0.0132, "step": 11325 }, { "epoch": 4.46, "learning_rate": 2.440631128495109e-07, "loss": 0.0141, "step": 11326 }, { "epoch": 4.46, "learning_rate": 2.440382067729879e-07, "loss": 0.0117, "step": 11327 }, { "epoch": 4.46, "learning_rate": 2.4401329997870896e-07, "loss": 0.0165, "step": 11328 }, { "epoch": 4.46, "learning_rate": 2.4398839246708027e-07, "loss": 0.0143, "step": 11329 }, { "epoch": 4.47, "learning_rate": 2.4396348423850754e-07, "loss": 0.0174, "step": 11330 }, { "epoch": 4.47, "learning_rate": 2.439385752933968e-07, "loss": 0.0133, "step": 11331 }, { "epoch": 4.47, "learning_rate": 2.4391366563215416e-07, "loss": 0.0132, "step": 11332 }, { "epoch": 4.47, "learning_rate": 2.438887552551855e-07, "loss": 0.0185, "step": 11333 }, { "epoch": 4.47, "learning_rate": 2.438638441628967e-07, "loss": 0.0171, "step": 11334 }, { "epoch": 4.47, "learning_rate": 2.4383893235569405e-07, "loss": 0.0163, "step": 11335 }, { "epoch": 4.47, "learning_rate": 2.438140198339834e-07, "loss": 0.0141, "step": 11336 }, { "epoch": 4.47, "learning_rate": 2.437891065981708e-07, "loss": 0.0152, "step": 11337 }, { "epoch": 4.47, "learning_rate": 2.4376419264866224e-07, "loss": 0.0158, "step": 11338 }, { "epoch": 4.47, "learning_rate": 2.43739277985864e-07, "loss": 0.0154, "step": 11339 }, { "epoch": 4.47, "learning_rate": 2.437143626101819e-07, "loss": 0.0172, "step": 11340 }, { "epoch": 4.47, "learning_rate": 2.4368944652202215e-07, "loss": 0.0145, "step": 11341 }, { "epoch": 4.47, "learning_rate": 2.436645297217909e-07, "loss": 0.0155, "step": 11342 }, { "epoch": 4.47, "learning_rate": 2.436396122098941e-07, "loss": 0.0128, "step": 11343 }, { "epoch": 4.47, "learning_rate": 2.4361469398673793e-07, "loss": 0.0168, "step": 11344 }, { "epoch": 4.47, "learning_rate": 2.435897750527286e-07, "loss": 0.0179, "step": 11345 }, { "epoch": 4.47, "learning_rate": 2.4356485540827213e-07, "loss": 0.0125, "step": 11346 }, { "epoch": 4.47, "learning_rate": 2.4353993505377473e-07, "loss": 0.0178, "step": 11347 }, { "epoch": 4.47, "learning_rate": 2.435150139896426e-07, "loss": 0.0152, "step": 11348 }, { "epoch": 4.47, "learning_rate": 2.4349009221628185e-07, "loss": 0.0166, "step": 11349 }, { "epoch": 4.47, "learning_rate": 2.434651697340987e-07, "loss": 0.0178, "step": 11350 }, { "epoch": 4.47, "learning_rate": 2.4344024654349933e-07, "loss": 0.0172, "step": 11351 }, { "epoch": 4.47, "learning_rate": 2.4341532264489e-07, "loss": 0.0157, "step": 11352 }, { "epoch": 4.47, "learning_rate": 2.4339039803867687e-07, "loss": 0.0137, "step": 11353 }, { "epoch": 4.47, "learning_rate": 2.433654727252661e-07, "loss": 0.0174, "step": 11354 }, { "epoch": 4.48, "learning_rate": 2.433405467050642e-07, "loss": 0.0152, "step": 11355 }, { "epoch": 4.48, "learning_rate": 2.4331561997847714e-07, "loss": 0.0157, "step": 11356 }, { "epoch": 4.48, "learning_rate": 2.432906925459114e-07, "loss": 0.0123, "step": 11357 }, { "epoch": 4.48, "learning_rate": 2.4326576440777306e-07, "loss": 0.0149, "step": 11358 }, { "epoch": 4.48, "learning_rate": 2.432408355644686e-07, "loss": 0.018, "step": 11359 }, { "epoch": 4.48, "learning_rate": 2.432159060164042e-07, "loss": 0.0171, "step": 11360 }, { "epoch": 4.48, "learning_rate": 2.431909757639862e-07, "loss": 0.0127, "step": 11361 }, { "epoch": 4.48, "learning_rate": 2.4316604480762106e-07, "loss": 0.0171, "step": 11362 }, { "epoch": 4.48, "learning_rate": 2.431411131477148e-07, "loss": 0.0168, "step": 11363 }, { "epoch": 4.48, "learning_rate": 2.4311618078467415e-07, "loss": 0.0135, "step": 11364 }, { "epoch": 4.48, "learning_rate": 2.430912477189052e-07, "loss": 0.0156, "step": 11365 }, { "epoch": 4.48, "learning_rate": 2.4306631395081437e-07, "loss": 0.0156, "step": 11366 }, { "epoch": 4.48, "learning_rate": 2.430413794808081e-07, "loss": 0.0132, "step": 11367 }, { "epoch": 4.48, "learning_rate": 2.430164443092928e-07, "loss": 0.0163, "step": 11368 }, { "epoch": 4.48, "learning_rate": 2.4299150843667486e-07, "loss": 0.0146, "step": 11369 }, { "epoch": 4.48, "learning_rate": 2.429665718633605e-07, "loss": 0.0125, "step": 11370 }, { "epoch": 4.48, "learning_rate": 2.429416345897565e-07, "loss": 0.0162, "step": 11371 }, { "epoch": 4.48, "learning_rate": 2.42916696616269e-07, "loss": 0.0114, "step": 11372 }, { "epoch": 4.48, "learning_rate": 2.4289175794330464e-07, "loss": 0.0142, "step": 11373 }, { "epoch": 4.48, "learning_rate": 2.428668185712698e-07, "loss": 0.0164, "step": 11374 }, { "epoch": 4.48, "learning_rate": 2.4284187850057096e-07, "loss": 0.0158, "step": 11375 }, { "epoch": 4.48, "learning_rate": 2.428169377316146e-07, "loss": 0.0188, "step": 11376 }, { "epoch": 4.48, "learning_rate": 2.427919962648073e-07, "loss": 0.0141, "step": 11377 }, { "epoch": 4.48, "learning_rate": 2.427670541005554e-07, "loss": 0.0131, "step": 11378 }, { "epoch": 4.48, "learning_rate": 2.427421112392656e-07, "loss": 0.0167, "step": 11379 }, { "epoch": 4.48, "learning_rate": 2.427171676813443e-07, "loss": 0.0151, "step": 11380 }, { "epoch": 4.49, "learning_rate": 2.426922234271981e-07, "loss": 0.0169, "step": 11381 }, { "epoch": 4.49, "learning_rate": 2.426672784772335e-07, "loss": 0.0206, "step": 11382 }, { "epoch": 4.49, "learning_rate": 2.4264233283185713e-07, "loss": 0.0168, "step": 11383 }, { "epoch": 4.49, "learning_rate": 2.426173864914756e-07, "loss": 0.0162, "step": 11384 }, { "epoch": 4.49, "learning_rate": 2.425924394564954e-07, "loss": 0.016, "step": 11385 }, { "epoch": 4.49, "learning_rate": 2.4256749172732314e-07, "loss": 0.0148, "step": 11386 }, { "epoch": 4.49, "learning_rate": 2.4254254330436553e-07, "loss": 0.0162, "step": 11387 }, { "epoch": 4.49, "learning_rate": 2.4251759418802907e-07, "loss": 0.0141, "step": 11388 }, { "epoch": 4.49, "learning_rate": 2.4249264437872045e-07, "loss": 0.0163, "step": 11389 }, { "epoch": 4.49, "learning_rate": 2.424676938768463e-07, "loss": 0.0187, "step": 11390 }, { "epoch": 4.49, "learning_rate": 2.424427426828133e-07, "loss": 0.0162, "step": 11391 }, { "epoch": 4.49, "learning_rate": 2.4241779079702815e-07, "loss": 0.0147, "step": 11392 }, { "epoch": 4.49, "learning_rate": 2.4239283821989744e-07, "loss": 0.0136, "step": 11393 }, { "epoch": 4.49, "learning_rate": 2.4236788495182783e-07, "loss": 0.0146, "step": 11394 }, { "epoch": 4.49, "learning_rate": 2.4234293099322615e-07, "loss": 0.0153, "step": 11395 }, { "epoch": 4.49, "learning_rate": 2.4231797634449897e-07, "loss": 0.0135, "step": 11396 }, { "epoch": 4.49, "learning_rate": 2.4229302100605325e-07, "loss": 0.0134, "step": 11397 }, { "epoch": 4.49, "learning_rate": 2.422680649782954e-07, "loss": 0.0174, "step": 11398 }, { "epoch": 4.49, "learning_rate": 2.422431082616324e-07, "loss": 0.0136, "step": 11399 }, { "epoch": 4.49, "learning_rate": 2.42218150856471e-07, "loss": 0.0154, "step": 11400 }, { "epoch": 4.49, "learning_rate": 2.4219319276321783e-07, "loss": 0.0156, "step": 11401 }, { "epoch": 4.49, "learning_rate": 2.421682339822798e-07, "loss": 0.0139, "step": 11402 }, { "epoch": 4.49, "learning_rate": 2.4214327451406354e-07, "loss": 0.0137, "step": 11403 }, { "epoch": 4.49, "learning_rate": 2.4211831435897603e-07, "loss": 0.013, "step": 11404 }, { "epoch": 4.49, "learning_rate": 2.42093353517424e-07, "loss": 0.0178, "step": 11405 }, { "epoch": 4.5, "learning_rate": 2.420683919898143e-07, "loss": 0.0179, "step": 11406 }, { "epoch": 4.5, "learning_rate": 2.420434297765538e-07, "loss": 0.0164, "step": 11407 }, { "epoch": 4.5, "learning_rate": 2.4201846687804916e-07, "loss": 0.0157, "step": 11408 }, { "epoch": 4.5, "learning_rate": 2.4199350329470745e-07, "loss": 0.0135, "step": 11409 }, { "epoch": 4.5, "learning_rate": 2.4196853902693547e-07, "loss": 0.0183, "step": 11410 }, { "epoch": 4.5, "learning_rate": 2.4194357407514003e-07, "loss": 0.0134, "step": 11411 }, { "epoch": 4.5, "learning_rate": 2.419186084397281e-07, "loss": 0.0183, "step": 11412 }, { "epoch": 4.5, "learning_rate": 2.4189364212110663e-07, "loss": 0.0179, "step": 11413 }, { "epoch": 4.5, "learning_rate": 2.4186867511968236e-07, "loss": 0.0157, "step": 11414 }, { "epoch": 4.5, "learning_rate": 2.418437074358624e-07, "loss": 0.0154, "step": 11415 }, { "epoch": 4.5, "learning_rate": 2.4181873907005355e-07, "loss": 0.0194, "step": 11416 }, { "epoch": 4.5, "learning_rate": 2.4179377002266286e-07, "loss": 0.0166, "step": 11417 }, { "epoch": 4.5, "learning_rate": 2.4176880029409716e-07, "loss": 0.0154, "step": 11418 }, { "epoch": 4.5, "learning_rate": 2.4174382988476357e-07, "loss": 0.0139, "step": 11419 }, { "epoch": 4.5, "learning_rate": 2.4171885879506897e-07, "loss": 0.0152, "step": 11420 }, { "epoch": 4.5, "learning_rate": 2.4169388702542037e-07, "loss": 0.0149, "step": 11421 }, { "epoch": 4.5, "learning_rate": 2.4166891457622476e-07, "loss": 0.0127, "step": 11422 }, { "epoch": 4.5, "learning_rate": 2.416439414478892e-07, "loss": 0.0178, "step": 11423 }, { "epoch": 4.5, "learning_rate": 2.416189676408207e-07, "loss": 0.0163, "step": 11424 }, { "epoch": 4.5, "learning_rate": 2.415939931554262e-07, "loss": 0.0116, "step": 11425 }, { "epoch": 4.5, "learning_rate": 2.4156901799211296e-07, "loss": 0.016, "step": 11426 }, { "epoch": 4.5, "learning_rate": 2.415440421512878e-07, "loss": 0.0153, "step": 11427 }, { "epoch": 4.5, "learning_rate": 2.4151906563335793e-07, "loss": 0.0162, "step": 11428 }, { "epoch": 4.5, "learning_rate": 2.4149408843873044e-07, "loss": 0.0151, "step": 11429 }, { "epoch": 4.5, "learning_rate": 2.414691105678123e-07, "loss": 0.0156, "step": 11430 }, { "epoch": 4.51, "learning_rate": 2.4144413202101074e-07, "loss": 0.0159, "step": 11431 }, { "epoch": 4.51, "learning_rate": 2.414191527987328e-07, "loss": 0.0139, "step": 11432 }, { "epoch": 4.51, "learning_rate": 2.413941729013856e-07, "loss": 0.0157, "step": 11433 }, { "epoch": 4.51, "learning_rate": 2.4136919232937626e-07, "loss": 0.0154, "step": 11434 }, { "epoch": 4.51, "learning_rate": 2.4134421108311203e-07, "loss": 0.0157, "step": 11435 }, { "epoch": 4.51, "learning_rate": 2.413192291629999e-07, "loss": 0.0144, "step": 11436 }, { "epoch": 4.51, "learning_rate": 2.4129424656944724e-07, "loss": 0.0135, "step": 11437 }, { "epoch": 4.51, "learning_rate": 2.4126926330286104e-07, "loss": 0.0176, "step": 11438 }, { "epoch": 4.51, "learning_rate": 2.4124427936364863e-07, "loss": 0.0162, "step": 11439 }, { "epoch": 4.51, "learning_rate": 2.4121929475221716e-07, "loss": 0.0151, "step": 11440 }, { "epoch": 4.51, "learning_rate": 2.411943094689738e-07, "loss": 0.015, "step": 11441 }, { "epoch": 4.51, "learning_rate": 2.411693235143258e-07, "loss": 0.0135, "step": 11442 }, { "epoch": 4.51, "learning_rate": 2.4114433688868035e-07, "loss": 0.0146, "step": 11443 }, { "epoch": 4.51, "learning_rate": 2.411193495924448e-07, "loss": 0.0167, "step": 11444 }, { "epoch": 4.51, "learning_rate": 2.4109436162602626e-07, "loss": 0.0133, "step": 11445 }, { "epoch": 4.51, "learning_rate": 2.4106937298983225e-07, "loss": 0.0139, "step": 11446 }, { "epoch": 4.51, "learning_rate": 2.410443836842697e-07, "loss": 0.0141, "step": 11447 }, { "epoch": 4.51, "learning_rate": 2.4101939370974616e-07, "loss": 0.0161, "step": 11448 }, { "epoch": 4.51, "learning_rate": 2.409944030666689e-07, "loss": 0.0134, "step": 11449 }, { "epoch": 4.51, "learning_rate": 2.409694117554451e-07, "loss": 0.0156, "step": 11450 }, { "epoch": 4.51, "learning_rate": 2.4094441977648217e-07, "loss": 0.0157, "step": 11451 }, { "epoch": 4.51, "learning_rate": 2.4091942713018743e-07, "loss": 0.0137, "step": 11452 }, { "epoch": 4.51, "learning_rate": 2.408944338169682e-07, "loss": 0.0143, "step": 11453 }, { "epoch": 4.51, "learning_rate": 2.408694398372319e-07, "loss": 0.0161, "step": 11454 }, { "epoch": 4.51, "learning_rate": 2.4084444519138583e-07, "loss": 0.0137, "step": 11455 }, { "epoch": 4.51, "learning_rate": 2.4081944987983734e-07, "loss": 0.0139, "step": 11456 }, { "epoch": 4.52, "learning_rate": 2.407944539029939e-07, "loss": 0.016, "step": 11457 }, { "epoch": 4.52, "learning_rate": 2.407694572612628e-07, "loss": 0.0155, "step": 11458 }, { "epoch": 4.52, "learning_rate": 2.4074445995505156e-07, "loss": 0.0136, "step": 11459 }, { "epoch": 4.52, "learning_rate": 2.407194619847676e-07, "loss": 0.013, "step": 11460 }, { "epoch": 4.52, "learning_rate": 2.406944633508182e-07, "loss": 0.0165, "step": 11461 }, { "epoch": 4.52, "learning_rate": 2.4066946405361095e-07, "loss": 0.0158, "step": 11462 }, { "epoch": 4.52, "learning_rate": 2.4064446409355316e-07, "loss": 0.0141, "step": 11463 }, { "epoch": 4.52, "learning_rate": 2.406194634710525e-07, "loss": 0.0141, "step": 11464 }, { "epoch": 4.52, "learning_rate": 2.4059446218651623e-07, "loss": 0.0151, "step": 11465 }, { "epoch": 4.52, "learning_rate": 2.40569460240352e-07, "loss": 0.0146, "step": 11466 }, { "epoch": 4.52, "learning_rate": 2.405444576329671e-07, "loss": 0.0129, "step": 11467 }, { "epoch": 4.52, "learning_rate": 2.405194543647693e-07, "loss": 0.017, "step": 11468 }, { "epoch": 4.52, "learning_rate": 2.4049445043616596e-07, "loss": 0.0142, "step": 11469 }, { "epoch": 4.52, "learning_rate": 2.404694458475645e-07, "loss": 0.0141, "step": 11470 }, { "epoch": 4.52, "learning_rate": 2.404444405993727e-07, "loss": 0.0169, "step": 11471 }, { "epoch": 4.52, "learning_rate": 2.404194346919979e-07, "loss": 0.0132, "step": 11472 }, { "epoch": 4.52, "learning_rate": 2.4039442812584783e-07, "loss": 0.0156, "step": 11473 }, { "epoch": 4.52, "learning_rate": 2.4036942090132997e-07, "loss": 0.0137, "step": 11474 }, { "epoch": 4.52, "learning_rate": 2.403444130188519e-07, "loss": 0.0143, "step": 11475 }, { "epoch": 4.52, "learning_rate": 2.4031940447882114e-07, "loss": 0.014, "step": 11476 }, { "epoch": 4.52, "learning_rate": 2.402943952816454e-07, "loss": 0.0132, "step": 11477 }, { "epoch": 4.52, "learning_rate": 2.402693854277323e-07, "loss": 0.0139, "step": 11478 }, { "epoch": 4.52, "learning_rate": 2.402443749174894e-07, "loss": 0.0141, "step": 11479 }, { "epoch": 4.52, "learning_rate": 2.4021936375132435e-07, "loss": 0.0137, "step": 11480 }, { "epoch": 4.52, "learning_rate": 2.4019435192964475e-07, "loss": 0.0169, "step": 11481 }, { "epoch": 4.53, "learning_rate": 2.401693394528584e-07, "loss": 0.0161, "step": 11482 }, { "epoch": 4.53, "learning_rate": 2.4014432632137277e-07, "loss": 0.0144, "step": 11483 }, { "epoch": 4.53, "learning_rate": 2.4011931253559566e-07, "loss": 0.0136, "step": 11484 }, { "epoch": 4.53, "learning_rate": 2.400942980959347e-07, "loss": 0.0177, "step": 11485 }, { "epoch": 4.53, "learning_rate": 2.400692830027977e-07, "loss": 0.0131, "step": 11486 }, { "epoch": 4.53, "learning_rate": 2.400442672565922e-07, "loss": 0.0163, "step": 11487 }, { "epoch": 4.53, "learning_rate": 2.400192508577261e-07, "loss": 0.0132, "step": 11488 }, { "epoch": 4.53, "learning_rate": 2.3999423380660695e-07, "loss": 0.0167, "step": 11489 }, { "epoch": 4.53, "learning_rate": 2.399692161036426e-07, "loss": 0.0165, "step": 11490 }, { "epoch": 4.53, "learning_rate": 2.399441977492408e-07, "loss": 0.0179, "step": 11491 }, { "epoch": 4.53, "learning_rate": 2.3991917874380925e-07, "loss": 0.0159, "step": 11492 }, { "epoch": 4.53, "learning_rate": 2.3989415908775575e-07, "loss": 0.0148, "step": 11493 }, { "epoch": 4.53, "learning_rate": 2.398691387814881e-07, "loss": 0.0131, "step": 11494 }, { "epoch": 4.53, "learning_rate": 2.3984411782541414e-07, "loss": 0.0128, "step": 11495 }, { "epoch": 4.53, "learning_rate": 2.3981909621994157e-07, "loss": 0.0147, "step": 11496 }, { "epoch": 4.53, "learning_rate": 2.3979407396547823e-07, "loss": 0.0158, "step": 11497 }, { "epoch": 4.53, "learning_rate": 2.39769051062432e-07, "loss": 0.0119, "step": 11498 }, { "epoch": 4.53, "learning_rate": 2.3974402751121064e-07, "loss": 0.0192, "step": 11499 }, { "epoch": 4.53, "learning_rate": 2.397190033122221e-07, "loss": 0.013, "step": 11500 }, { "epoch": 4.53, "learning_rate": 2.396939784658741e-07, "loss": 0.0209, "step": 11501 }, { "epoch": 4.53, "learning_rate": 2.396689529725746e-07, "loss": 0.0147, "step": 11502 }, { "epoch": 4.53, "learning_rate": 2.396439268327314e-07, "loss": 0.0152, "step": 11503 }, { "epoch": 4.53, "learning_rate": 2.3961890004675257e-07, "loss": 0.0127, "step": 11504 }, { "epoch": 4.53, "learning_rate": 2.395938726150458e-07, "loss": 0.015, "step": 11505 }, { "epoch": 4.53, "learning_rate": 2.395688445380191e-07, "loss": 0.0156, "step": 11506 }, { "epoch": 4.54, "learning_rate": 2.3954381581608036e-07, "loss": 0.0162, "step": 11507 }, { "epoch": 4.54, "learning_rate": 2.3951878644963755e-07, "loss": 0.0145, "step": 11508 }, { "epoch": 4.54, "learning_rate": 2.394937564390986e-07, "loss": 0.0161, "step": 11509 }, { "epoch": 4.54, "learning_rate": 2.3946872578487136e-07, "loss": 0.017, "step": 11510 }, { "epoch": 4.54, "learning_rate": 2.3944369448736395e-07, "loss": 0.0144, "step": 11511 }, { "epoch": 4.54, "learning_rate": 2.3941866254698426e-07, "loss": 0.0144, "step": 11512 }, { "epoch": 4.54, "learning_rate": 2.3939362996414024e-07, "loss": 0.0132, "step": 11513 }, { "epoch": 4.54, "learning_rate": 2.3936859673923995e-07, "loss": 0.0146, "step": 11514 }, { "epoch": 4.54, "learning_rate": 2.3934356287269135e-07, "loss": 0.0142, "step": 11515 }, { "epoch": 4.54, "learning_rate": 2.393185283649025e-07, "loss": 0.013, "step": 11516 }, { "epoch": 4.54, "learning_rate": 2.392934932162814e-07, "loss": 0.0152, "step": 11517 }, { "epoch": 4.54, "learning_rate": 2.39268457427236e-07, "loss": 0.02, "step": 11518 }, { "epoch": 4.54, "learning_rate": 2.392434209981745e-07, "loss": 0.0175, "step": 11519 }, { "epoch": 4.54, "learning_rate": 2.3921838392950495e-07, "loss": 0.017, "step": 11520 }, { "epoch": 4.54, "learning_rate": 2.391933462216352e-07, "loss": 0.0161, "step": 11521 }, { "epoch": 4.54, "learning_rate": 2.3916830787497354e-07, "loss": 0.0147, "step": 11522 }, { "epoch": 4.54, "learning_rate": 2.3914326888992797e-07, "loss": 0.014, "step": 11523 }, { "epoch": 4.54, "learning_rate": 2.3911822926690667e-07, "loss": 0.0137, "step": 11524 }, { "epoch": 4.54, "learning_rate": 2.3909318900631763e-07, "loss": 0.0151, "step": 11525 }, { "epoch": 4.54, "learning_rate": 2.39068148108569e-07, "loss": 0.015, "step": 11526 }, { "epoch": 4.54, "learning_rate": 2.3904310657406895e-07, "loss": 0.0186, "step": 11527 }, { "epoch": 4.54, "learning_rate": 2.390180644032257e-07, "loss": 0.0175, "step": 11528 }, { "epoch": 4.54, "learning_rate": 2.389930215964472e-07, "loss": 0.0162, "step": 11529 }, { "epoch": 4.54, "learning_rate": 2.389679781541417e-07, "loss": 0.0146, "step": 11530 }, { "epoch": 4.54, "learning_rate": 2.389429340767174e-07, "loss": 0.0127, "step": 11531 }, { "epoch": 4.54, "learning_rate": 2.3891788936458245e-07, "loss": 0.0136, "step": 11532 }, { "epoch": 4.55, "learning_rate": 2.388928440181451e-07, "loss": 0.0165, "step": 11533 }, { "epoch": 4.55, "learning_rate": 2.3886779803781343e-07, "loss": 0.0143, "step": 11534 }, { "epoch": 4.55, "learning_rate": 2.388427514239958e-07, "loss": 0.0164, "step": 11535 }, { "epoch": 4.55, "learning_rate": 2.388177041771003e-07, "loss": 0.0149, "step": 11536 }, { "epoch": 4.55, "learning_rate": 2.387926562975353e-07, "loss": 0.0158, "step": 11537 }, { "epoch": 4.55, "learning_rate": 2.387676077857089e-07, "loss": 0.0167, "step": 11538 }, { "epoch": 4.55, "learning_rate": 2.387425586420294e-07, "loss": 0.0155, "step": 11539 }, { "epoch": 4.55, "learning_rate": 2.387175088669051e-07, "loss": 0.0142, "step": 11540 }, { "epoch": 4.55, "learning_rate": 2.3869245846074426e-07, "loss": 0.0169, "step": 11541 }, { "epoch": 4.55, "learning_rate": 2.3866740742395515e-07, "loss": 0.0152, "step": 11542 }, { "epoch": 4.55, "learning_rate": 2.3864235575694606e-07, "loss": 0.0175, "step": 11543 }, { "epoch": 4.55, "learning_rate": 2.3861730346012525e-07, "loss": 0.0167, "step": 11544 }, { "epoch": 4.55, "learning_rate": 2.3859225053390115e-07, "loss": 0.0167, "step": 11545 }, { "epoch": 4.55, "learning_rate": 2.38567196978682e-07, "loss": 0.0131, "step": 11546 }, { "epoch": 4.55, "learning_rate": 2.385421427948761e-07, "loss": 0.0177, "step": 11547 }, { "epoch": 4.55, "learning_rate": 2.3851708798289197e-07, "loss": 0.012, "step": 11548 }, { "epoch": 4.55, "learning_rate": 2.384920325431378e-07, "loss": 0.0177, "step": 11549 }, { "epoch": 4.55, "learning_rate": 2.384669764760219e-07, "loss": 0.019, "step": 11550 }, { "epoch": 4.55, "learning_rate": 2.3844191978195285e-07, "loss": 0.0147, "step": 11551 }, { "epoch": 4.55, "learning_rate": 2.3841686246133892e-07, "loss": 0.0137, "step": 11552 }, { "epoch": 4.55, "learning_rate": 2.3839180451458851e-07, "loss": 0.0154, "step": 11553 }, { "epoch": 4.55, "learning_rate": 2.3836674594211002e-07, "loss": 0.0151, "step": 11554 }, { "epoch": 4.55, "learning_rate": 2.383416867443119e-07, "loss": 0.0145, "step": 11555 }, { "epoch": 4.55, "learning_rate": 2.383166269216025e-07, "loss": 0.0172, "step": 11556 }, { "epoch": 4.55, "learning_rate": 2.382915664743904e-07, "loss": 0.0138, "step": 11557 }, { "epoch": 4.56, "learning_rate": 2.3826650540308395e-07, "loss": 0.0149, "step": 11558 }, { "epoch": 4.56, "learning_rate": 2.3824144370809155e-07, "loss": 0.0153, "step": 11559 }, { "epoch": 4.56, "learning_rate": 2.3821638138982182e-07, "loss": 0.0164, "step": 11560 }, { "epoch": 4.56, "learning_rate": 2.3819131844868315e-07, "loss": 0.0173, "step": 11561 }, { "epoch": 4.56, "learning_rate": 2.38166254885084e-07, "loss": 0.0136, "step": 11562 }, { "epoch": 4.56, "learning_rate": 2.3814119069943292e-07, "loss": 0.0153, "step": 11563 }, { "epoch": 4.56, "learning_rate": 2.381161258921384e-07, "loss": 0.0142, "step": 11564 }, { "epoch": 4.56, "learning_rate": 2.3809106046360892e-07, "loss": 0.0147, "step": 11565 }, { "epoch": 4.56, "learning_rate": 2.3806599441425307e-07, "loss": 0.0138, "step": 11566 }, { "epoch": 4.56, "learning_rate": 2.3804092774447934e-07, "loss": 0.0167, "step": 11567 }, { "epoch": 4.56, "learning_rate": 2.3801586045469637e-07, "loss": 0.0158, "step": 11568 }, { "epoch": 4.56, "learning_rate": 2.3799079254531263e-07, "loss": 0.0138, "step": 11569 }, { "epoch": 4.56, "learning_rate": 2.3796572401673666e-07, "loss": 0.0176, "step": 11570 }, { "epoch": 4.56, "learning_rate": 2.3794065486937716e-07, "loss": 0.0141, "step": 11571 }, { "epoch": 4.56, "learning_rate": 2.3791558510364256e-07, "loss": 0.0146, "step": 11572 }, { "epoch": 4.56, "learning_rate": 2.3789051471994162e-07, "loss": 0.0137, "step": 11573 }, { "epoch": 4.56, "learning_rate": 2.3786544371868285e-07, "loss": 0.0133, "step": 11574 }, { "epoch": 4.56, "learning_rate": 2.378403721002749e-07, "loss": 0.0153, "step": 11575 }, { "epoch": 4.56, "learning_rate": 2.3781529986512638e-07, "loss": 0.0166, "step": 11576 }, { "epoch": 4.56, "learning_rate": 2.3779022701364598e-07, "loss": 0.0162, "step": 11577 }, { "epoch": 4.56, "learning_rate": 2.3776515354624235e-07, "loss": 0.0139, "step": 11578 }, { "epoch": 4.56, "learning_rate": 2.37740079463324e-07, "loss": 0.0156, "step": 11579 }, { "epoch": 4.56, "learning_rate": 2.3771500476529982e-07, "loss": 0.0143, "step": 11580 }, { "epoch": 4.56, "learning_rate": 2.3768992945257832e-07, "loss": 0.0149, "step": 11581 }, { "epoch": 4.56, "learning_rate": 2.3766485352556827e-07, "loss": 0.0143, "step": 11582 }, { "epoch": 4.56, "learning_rate": 2.3763977698467835e-07, "loss": 0.0161, "step": 11583 }, { "epoch": 4.57, "learning_rate": 2.3761469983031728e-07, "loss": 0.0176, "step": 11584 }, { "epoch": 4.57, "learning_rate": 2.3758962206289374e-07, "loss": 0.0157, "step": 11585 }, { "epoch": 4.57, "learning_rate": 2.3756454368281656e-07, "loss": 0.0162, "step": 11586 }, { "epoch": 4.57, "learning_rate": 2.3753946469049433e-07, "loss": 0.0121, "step": 11587 }, { "epoch": 4.57, "learning_rate": 2.375143850863359e-07, "loss": 0.0175, "step": 11588 }, { "epoch": 4.57, "learning_rate": 2.3748930487075004e-07, "loss": 0.0131, "step": 11589 }, { "epoch": 4.57, "learning_rate": 2.3746422404414543e-07, "loss": 0.0131, "step": 11590 }, { "epoch": 4.57, "learning_rate": 2.3743914260693097e-07, "loss": 0.0187, "step": 11591 }, { "epoch": 4.57, "learning_rate": 2.3741406055951536e-07, "loss": 0.0134, "step": 11592 }, { "epoch": 4.57, "learning_rate": 2.373889779023074e-07, "loss": 0.0149, "step": 11593 }, { "epoch": 4.57, "learning_rate": 2.3736389463571593e-07, "loss": 0.0157, "step": 11594 }, { "epoch": 4.57, "learning_rate": 2.3733881076014982e-07, "loss": 0.0139, "step": 11595 }, { "epoch": 4.57, "learning_rate": 2.3731372627601777e-07, "loss": 0.0147, "step": 11596 }, { "epoch": 4.57, "learning_rate": 2.3728864118372875e-07, "loss": 0.0142, "step": 11597 }, { "epoch": 4.57, "learning_rate": 2.3726355548369157e-07, "loss": 0.0178, "step": 11598 }, { "epoch": 4.57, "learning_rate": 2.37238469176315e-07, "loss": 0.0155, "step": 11599 }, { "epoch": 4.57, "learning_rate": 2.3721338226200802e-07, "loss": 0.0158, "step": 11600 }, { "epoch": 4.57, "learning_rate": 2.3718829474117945e-07, "loss": 0.0143, "step": 11601 }, { "epoch": 4.57, "learning_rate": 2.3716320661423822e-07, "loss": 0.0141, "step": 11602 }, { "epoch": 4.57, "learning_rate": 2.3713811788159318e-07, "loss": 0.0168, "step": 11603 }, { "epoch": 4.57, "learning_rate": 2.3711302854365326e-07, "loss": 0.0129, "step": 11604 }, { "epoch": 4.57, "learning_rate": 2.3708793860082741e-07, "loss": 0.0144, "step": 11605 }, { "epoch": 4.57, "learning_rate": 2.3706284805352453e-07, "loss": 0.0152, "step": 11606 }, { "epoch": 4.57, "learning_rate": 2.3703775690215345e-07, "loss": 0.016, "step": 11607 }, { "epoch": 4.57, "learning_rate": 2.3701266514712333e-07, "loss": 0.016, "step": 11608 }, { "epoch": 4.58, "learning_rate": 2.3698757278884301e-07, "loss": 0.0179, "step": 11609 }, { "epoch": 4.58, "learning_rate": 2.3696247982772143e-07, "loss": 0.0128, "step": 11610 }, { "epoch": 4.58, "learning_rate": 2.3693738626416766e-07, "loss": 0.0146, "step": 11611 }, { "epoch": 4.58, "learning_rate": 2.3691229209859056e-07, "loss": 0.0184, "step": 11612 }, { "epoch": 4.58, "learning_rate": 2.3688719733139928e-07, "loss": 0.0146, "step": 11613 }, { "epoch": 4.58, "learning_rate": 2.3686210196300268e-07, "loss": 0.018, "step": 11614 }, { "epoch": 4.58, "learning_rate": 2.368370059938099e-07, "loss": 0.0156, "step": 11615 }, { "epoch": 4.58, "learning_rate": 2.3681190942422984e-07, "loss": 0.0171, "step": 11616 }, { "epoch": 4.58, "learning_rate": 2.3678681225467162e-07, "loss": 0.0151, "step": 11617 }, { "epoch": 4.58, "learning_rate": 2.367617144855443e-07, "loss": 0.015, "step": 11618 }, { "epoch": 4.58, "learning_rate": 2.367366161172569e-07, "loss": 0.0158, "step": 11619 }, { "epoch": 4.58, "learning_rate": 2.367115171502185e-07, "loss": 0.0132, "step": 11620 }, { "epoch": 4.58, "learning_rate": 2.3668641758483812e-07, "loss": 0.0157, "step": 11621 }, { "epoch": 4.58, "learning_rate": 2.3666131742152493e-07, "loss": 0.0156, "step": 11622 }, { "epoch": 4.58, "learning_rate": 2.36636216660688e-07, "loss": 0.0167, "step": 11623 }, { "epoch": 4.58, "learning_rate": 2.3661111530273637e-07, "loss": 0.0162, "step": 11624 }, { "epoch": 4.58, "learning_rate": 2.3658601334807926e-07, "loss": 0.0139, "step": 11625 }, { "epoch": 4.58, "learning_rate": 2.3656091079712573e-07, "loss": 0.013, "step": 11626 }, { "epoch": 4.58, "learning_rate": 2.3653580765028486e-07, "loss": 0.0161, "step": 11627 }, { "epoch": 4.58, "learning_rate": 2.365107039079659e-07, "loss": 0.0174, "step": 11628 }, { "epoch": 4.58, "learning_rate": 2.3648559957057796e-07, "loss": 0.0126, "step": 11629 }, { "epoch": 4.58, "learning_rate": 2.3646049463853016e-07, "loss": 0.0149, "step": 11630 }, { "epoch": 4.58, "learning_rate": 2.3643538911223175e-07, "loss": 0.0135, "step": 11631 }, { "epoch": 4.58, "learning_rate": 2.3641028299209184e-07, "loss": 0.0147, "step": 11632 }, { "epoch": 4.58, "learning_rate": 2.363851762785197e-07, "loss": 0.0168, "step": 11633 }, { "epoch": 4.59, "learning_rate": 2.3636006897192442e-07, "loss": 0.0143, "step": 11634 }, { "epoch": 4.59, "learning_rate": 2.363349610727153e-07, "loss": 0.0143, "step": 11635 }, { "epoch": 4.59, "learning_rate": 2.3630985258130155e-07, "loss": 0.0182, "step": 11636 }, { "epoch": 4.59, "learning_rate": 2.362847434980924e-07, "loss": 0.0193, "step": 11637 }, { "epoch": 4.59, "learning_rate": 2.362596338234971e-07, "loss": 0.0139, "step": 11638 }, { "epoch": 4.59, "learning_rate": 2.362345235579248e-07, "loss": 0.0156, "step": 11639 }, { "epoch": 4.59, "learning_rate": 2.3620941270178488e-07, "loss": 0.0151, "step": 11640 }, { "epoch": 4.59, "learning_rate": 2.3618430125548658e-07, "loss": 0.0151, "step": 11641 }, { "epoch": 4.59, "learning_rate": 2.3615918921943915e-07, "loss": 0.0171, "step": 11642 }, { "epoch": 4.59, "learning_rate": 2.3613407659405187e-07, "loss": 0.015, "step": 11643 }, { "epoch": 4.59, "learning_rate": 2.361089633797341e-07, "loss": 0.0163, "step": 11644 }, { "epoch": 4.59, "learning_rate": 2.3608384957689505e-07, "loss": 0.0154, "step": 11645 }, { "epoch": 4.59, "learning_rate": 2.3605873518594411e-07, "loss": 0.0169, "step": 11646 }, { "epoch": 4.59, "learning_rate": 2.3603362020729065e-07, "loss": 0.0138, "step": 11647 }, { "epoch": 4.59, "learning_rate": 2.360085046413439e-07, "loss": 0.0151, "step": 11648 }, { "epoch": 4.59, "learning_rate": 2.3598338848851324e-07, "loss": 0.0163, "step": 11649 }, { "epoch": 4.59, "learning_rate": 2.3595827174920808e-07, "loss": 0.0145, "step": 11650 }, { "epoch": 4.59, "learning_rate": 2.3593315442383769e-07, "loss": 0.0128, "step": 11651 }, { "epoch": 4.59, "learning_rate": 2.3590803651281153e-07, "loss": 0.0141, "step": 11652 }, { "epoch": 4.59, "learning_rate": 2.3588291801653892e-07, "loss": 0.0181, "step": 11653 }, { "epoch": 4.59, "learning_rate": 2.358577989354293e-07, "loss": 0.0138, "step": 11654 }, { "epoch": 4.59, "learning_rate": 2.3583267926989206e-07, "loss": 0.0139, "step": 11655 }, { "epoch": 4.59, "learning_rate": 2.3580755902033655e-07, "loss": 0.0158, "step": 11656 }, { "epoch": 4.59, "learning_rate": 2.3578243818717235e-07, "loss": 0.0189, "step": 11657 }, { "epoch": 4.59, "learning_rate": 2.357573167708087e-07, "loss": 0.0139, "step": 11658 }, { "epoch": 4.59, "learning_rate": 2.357321947716552e-07, "loss": 0.016, "step": 11659 }, { "epoch": 4.6, "learning_rate": 2.3570707219012118e-07, "loss": 0.0143, "step": 11660 }, { "epoch": 4.6, "learning_rate": 2.3568194902661616e-07, "loss": 0.0163, "step": 11661 }, { "epoch": 4.6, "learning_rate": 2.3565682528154963e-07, "loss": 0.0126, "step": 11662 }, { "epoch": 4.6, "learning_rate": 2.3563170095533098e-07, "loss": 0.012, "step": 11663 }, { "epoch": 4.6, "learning_rate": 2.3560657604836981e-07, "loss": 0.0147, "step": 11664 }, { "epoch": 4.6, "learning_rate": 2.3558145056107554e-07, "loss": 0.0158, "step": 11665 }, { "epoch": 4.6, "learning_rate": 2.3555632449385777e-07, "loss": 0.0137, "step": 11666 }, { "epoch": 4.6, "learning_rate": 2.355311978471259e-07, "loss": 0.0142, "step": 11667 }, { "epoch": 4.6, "learning_rate": 2.3550607062128952e-07, "loss": 0.016, "step": 11668 }, { "epoch": 4.6, "learning_rate": 2.3548094281675813e-07, "loss": 0.0175, "step": 11669 }, { "epoch": 4.6, "learning_rate": 2.354558144339413e-07, "loss": 0.0174, "step": 11670 }, { "epoch": 4.6, "learning_rate": 2.3543068547324862e-07, "loss": 0.0135, "step": 11671 }, { "epoch": 4.6, "learning_rate": 2.354055559350896e-07, "loss": 0.0167, "step": 11672 }, { "epoch": 4.6, "learning_rate": 2.353804258198738e-07, "loss": 0.0143, "step": 11673 }, { "epoch": 4.6, "learning_rate": 2.353552951280108e-07, "loss": 0.0147, "step": 11674 }, { "epoch": 4.6, "learning_rate": 2.353301638599103e-07, "loss": 0.0167, "step": 11675 }, { "epoch": 4.6, "learning_rate": 2.3530503201598178e-07, "loss": 0.0148, "step": 11676 }, { "epoch": 4.6, "learning_rate": 2.3527989959663492e-07, "loss": 0.0155, "step": 11677 }, { "epoch": 4.6, "learning_rate": 2.3525476660227933e-07, "loss": 0.0141, "step": 11678 }, { "epoch": 4.6, "learning_rate": 2.3522963303332454e-07, "loss": 0.0166, "step": 11679 }, { "epoch": 4.6, "learning_rate": 2.3520449889018038e-07, "loss": 0.0135, "step": 11680 }, { "epoch": 4.6, "learning_rate": 2.3517936417325632e-07, "loss": 0.0144, "step": 11681 }, { "epoch": 4.6, "learning_rate": 2.3515422888296212e-07, "loss": 0.0182, "step": 11682 }, { "epoch": 4.6, "learning_rate": 2.351290930197074e-07, "loss": 0.0146, "step": 11683 }, { "epoch": 4.6, "learning_rate": 2.3510395658390188e-07, "loss": 0.0148, "step": 11684 }, { "epoch": 4.61, "learning_rate": 2.3507881957595516e-07, "loss": 0.0175, "step": 11685 }, { "epoch": 4.61, "learning_rate": 2.350536819962771e-07, "loss": 0.0181, "step": 11686 }, { "epoch": 4.61, "learning_rate": 2.3502854384527717e-07, "loss": 0.0182, "step": 11687 }, { "epoch": 4.61, "learning_rate": 2.350034051233653e-07, "loss": 0.0159, "step": 11688 }, { "epoch": 4.61, "learning_rate": 2.3497826583095116e-07, "loss": 0.0155, "step": 11689 }, { "epoch": 4.61, "learning_rate": 2.3495312596844438e-07, "loss": 0.0177, "step": 11690 }, { "epoch": 4.61, "learning_rate": 2.3492798553625478e-07, "loss": 0.0157, "step": 11691 }, { "epoch": 4.61, "learning_rate": 2.3490284453479213e-07, "loss": 0.0135, "step": 11692 }, { "epoch": 4.61, "learning_rate": 2.3487770296446615e-07, "loss": 0.0143, "step": 11693 }, { "epoch": 4.61, "learning_rate": 2.3485256082568656e-07, "loss": 0.0154, "step": 11694 }, { "epoch": 4.61, "learning_rate": 2.3482741811886328e-07, "loss": 0.0162, "step": 11695 }, { "epoch": 4.61, "learning_rate": 2.3480227484440593e-07, "loss": 0.0146, "step": 11696 }, { "epoch": 4.61, "learning_rate": 2.347771310027245e-07, "loss": 0.015, "step": 11697 }, { "epoch": 4.61, "learning_rate": 2.3475198659422864e-07, "loss": 0.0176, "step": 11698 }, { "epoch": 4.61, "learning_rate": 2.3472684161932816e-07, "loss": 0.0148, "step": 11699 }, { "epoch": 4.61, "learning_rate": 2.3470169607843302e-07, "loss": 0.0163, "step": 11700 }, { "epoch": 4.61, "learning_rate": 2.3467654997195293e-07, "loss": 0.0141, "step": 11701 }, { "epoch": 4.61, "learning_rate": 2.346514033002978e-07, "loss": 0.0136, "step": 11702 }, { "epoch": 4.61, "learning_rate": 2.3462625606387743e-07, "loss": 0.0137, "step": 11703 }, { "epoch": 4.61, "learning_rate": 2.3460110826310174e-07, "loss": 0.0185, "step": 11704 }, { "epoch": 4.61, "learning_rate": 2.3457595989838051e-07, "loss": 0.0165, "step": 11705 }, { "epoch": 4.61, "learning_rate": 2.345508109701238e-07, "loss": 0.0139, "step": 11706 }, { "epoch": 4.61, "learning_rate": 2.3452566147874128e-07, "loss": 0.012, "step": 11707 }, { "epoch": 4.61, "learning_rate": 2.34500511424643e-07, "loss": 0.0167, "step": 11708 }, { "epoch": 4.61, "learning_rate": 2.3447536080823886e-07, "loss": 0.014, "step": 11709 }, { "epoch": 4.62, "learning_rate": 2.3445020962993864e-07, "loss": 0.0131, "step": 11710 }, { "epoch": 4.62, "learning_rate": 2.344250578901524e-07, "loss": 0.0156, "step": 11711 }, { "epoch": 4.62, "learning_rate": 2.343999055892901e-07, "loss": 0.0147, "step": 11712 }, { "epoch": 4.62, "learning_rate": 2.3437475272776153e-07, "loss": 0.0165, "step": 11713 }, { "epoch": 4.62, "learning_rate": 2.343495993059768e-07, "loss": 0.0169, "step": 11714 }, { "epoch": 4.62, "learning_rate": 2.3432444532434578e-07, "loss": 0.0167, "step": 11715 }, { "epoch": 4.62, "learning_rate": 2.3429929078327846e-07, "loss": 0.0181, "step": 11716 }, { "epoch": 4.62, "learning_rate": 2.342741356831848e-07, "loss": 0.0139, "step": 11717 }, { "epoch": 4.62, "learning_rate": 2.3424898002447491e-07, "loss": 0.0115, "step": 11718 }, { "epoch": 4.62, "learning_rate": 2.3422382380755858e-07, "loss": 0.0175, "step": 11719 }, { "epoch": 4.62, "learning_rate": 2.3419866703284606e-07, "loss": 0.0164, "step": 11720 }, { "epoch": 4.62, "learning_rate": 2.341735097007471e-07, "loss": 0.0152, "step": 11721 }, { "epoch": 4.62, "learning_rate": 2.3414835181167198e-07, "loss": 0.016, "step": 11722 }, { "epoch": 4.62, "learning_rate": 2.341231933660306e-07, "loss": 0.0195, "step": 11723 }, { "epoch": 4.62, "learning_rate": 2.3409803436423306e-07, "loss": 0.0157, "step": 11724 }, { "epoch": 4.62, "learning_rate": 2.3407287480668932e-07, "loss": 0.016, "step": 11725 }, { "epoch": 4.62, "learning_rate": 2.3404771469380953e-07, "loss": 0.015, "step": 11726 }, { "epoch": 4.62, "learning_rate": 2.3402255402600367e-07, "loss": 0.0154, "step": 11727 }, { "epoch": 4.62, "learning_rate": 2.3399739280368203e-07, "loss": 0.0128, "step": 11728 }, { "epoch": 4.62, "learning_rate": 2.339722310272545e-07, "loss": 0.0143, "step": 11729 }, { "epoch": 4.62, "learning_rate": 2.3394706869713118e-07, "loss": 0.0103, "step": 11730 }, { "epoch": 4.62, "learning_rate": 2.3392190581372229e-07, "loss": 0.0124, "step": 11731 }, { "epoch": 4.62, "learning_rate": 2.3389674237743784e-07, "loss": 0.0135, "step": 11732 }, { "epoch": 4.62, "learning_rate": 2.3387157838868806e-07, "loss": 0.0138, "step": 11733 }, { "epoch": 4.62, "learning_rate": 2.3384641384788303e-07, "loss": 0.0152, "step": 11734 }, { "epoch": 4.62, "learning_rate": 2.338212487554329e-07, "loss": 0.0141, "step": 11735 }, { "epoch": 4.63, "learning_rate": 2.3379608311174775e-07, "loss": 0.0135, "step": 11736 }, { "epoch": 4.63, "learning_rate": 2.337709169172379e-07, "loss": 0.0192, "step": 11737 }, { "epoch": 4.63, "learning_rate": 2.3374575017231337e-07, "loss": 0.017, "step": 11738 }, { "epoch": 4.63, "learning_rate": 2.337205828773844e-07, "loss": 0.0147, "step": 11739 }, { "epoch": 4.63, "learning_rate": 2.336954150328612e-07, "loss": 0.0169, "step": 11740 }, { "epoch": 4.63, "learning_rate": 2.3367024663915393e-07, "loss": 0.013, "step": 11741 }, { "epoch": 4.63, "learning_rate": 2.3364507769667283e-07, "loss": 0.0138, "step": 11742 }, { "epoch": 4.63, "learning_rate": 2.3361990820582806e-07, "loss": 0.017, "step": 11743 }, { "epoch": 4.63, "learning_rate": 2.3359473816702997e-07, "loss": 0.0133, "step": 11744 }, { "epoch": 4.63, "learning_rate": 2.335695675806886e-07, "loss": 0.0124, "step": 11745 }, { "epoch": 4.63, "learning_rate": 2.3354439644721436e-07, "loss": 0.0147, "step": 11746 }, { "epoch": 4.63, "learning_rate": 2.3351922476701743e-07, "loss": 0.0149, "step": 11747 }, { "epoch": 4.63, "learning_rate": 2.3349405254050807e-07, "loss": 0.0207, "step": 11748 }, { "epoch": 4.63, "learning_rate": 2.334688797680966e-07, "loss": 0.0176, "step": 11749 }, { "epoch": 4.63, "learning_rate": 2.334437064501932e-07, "loss": 0.0133, "step": 11750 }, { "epoch": 4.63, "learning_rate": 2.3341853258720821e-07, "loss": 0.0171, "step": 11751 }, { "epoch": 4.63, "learning_rate": 2.3339335817955196e-07, "loss": 0.0181, "step": 11752 }, { "epoch": 4.63, "learning_rate": 2.333681832276347e-07, "loss": 0.0143, "step": 11753 }, { "epoch": 4.63, "learning_rate": 2.3334300773186683e-07, "loss": 0.0173, "step": 11754 }, { "epoch": 4.63, "learning_rate": 2.3331783169265854e-07, "loss": 0.0148, "step": 11755 }, { "epoch": 4.63, "learning_rate": 2.3329265511042026e-07, "loss": 0.0174, "step": 11756 }, { "epoch": 4.63, "learning_rate": 2.3326747798556236e-07, "loss": 0.0154, "step": 11757 }, { "epoch": 4.63, "learning_rate": 2.3324230031849502e-07, "loss": 0.0149, "step": 11758 }, { "epoch": 4.63, "learning_rate": 2.3321712210962876e-07, "loss": 0.0149, "step": 11759 }, { "epoch": 4.63, "learning_rate": 2.3319194335937392e-07, "loss": 0.0162, "step": 11760 }, { "epoch": 4.64, "learning_rate": 2.3316676406814079e-07, "loss": 0.0181, "step": 11761 }, { "epoch": 4.64, "learning_rate": 2.3314158423633983e-07, "loss": 0.0174, "step": 11762 }, { "epoch": 4.64, "learning_rate": 2.3311640386438141e-07, "loss": 0.0169, "step": 11763 }, { "epoch": 4.64, "learning_rate": 2.3309122295267597e-07, "loss": 0.0151, "step": 11764 }, { "epoch": 4.64, "learning_rate": 2.3306604150163384e-07, "loss": 0.0153, "step": 11765 }, { "epoch": 4.64, "learning_rate": 2.3304085951166552e-07, "loss": 0.0122, "step": 11766 }, { "epoch": 4.64, "learning_rate": 2.3301567698318139e-07, "loss": 0.0142, "step": 11767 }, { "epoch": 4.64, "learning_rate": 2.3299049391659197e-07, "loss": 0.0149, "step": 11768 }, { "epoch": 4.64, "learning_rate": 2.3296531031230757e-07, "loss": 0.0163, "step": 11769 }, { "epoch": 4.64, "learning_rate": 2.329401261707387e-07, "loss": 0.0196, "step": 11770 }, { "epoch": 4.64, "learning_rate": 2.329149414922959e-07, "loss": 0.0152, "step": 11771 }, { "epoch": 4.64, "learning_rate": 2.3288975627738953e-07, "loss": 0.0164, "step": 11772 }, { "epoch": 4.64, "learning_rate": 2.3286457052643015e-07, "loss": 0.0156, "step": 11773 }, { "epoch": 4.64, "learning_rate": 2.328393842398282e-07, "loss": 0.0141, "step": 11774 }, { "epoch": 4.64, "learning_rate": 2.328141974179942e-07, "loss": 0.0157, "step": 11775 }, { "epoch": 4.64, "learning_rate": 2.3278901006133868e-07, "loss": 0.0139, "step": 11776 }, { "epoch": 4.64, "learning_rate": 2.327638221702721e-07, "loss": 0.0174, "step": 11777 }, { "epoch": 4.64, "learning_rate": 2.32738633745205e-07, "loss": 0.0188, "step": 11778 }, { "epoch": 4.64, "learning_rate": 2.3271344478654796e-07, "loss": 0.0127, "step": 11779 }, { "epoch": 4.64, "learning_rate": 2.326882552947115e-07, "loss": 0.0144, "step": 11780 }, { "epoch": 4.64, "learning_rate": 2.3266306527010612e-07, "loss": 0.0183, "step": 11781 }, { "epoch": 4.64, "learning_rate": 2.326378747131424e-07, "loss": 0.0135, "step": 11782 }, { "epoch": 4.64, "learning_rate": 2.32612683624231e-07, "loss": 0.0197, "step": 11783 }, { "epoch": 4.64, "learning_rate": 2.3258749200378237e-07, "loss": 0.0175, "step": 11784 }, { "epoch": 4.64, "learning_rate": 2.3256229985220712e-07, "loss": 0.0163, "step": 11785 }, { "epoch": 4.64, "learning_rate": 2.3253710716991595e-07, "loss": 0.0175, "step": 11786 }, { "epoch": 4.65, "learning_rate": 2.3251191395731932e-07, "loss": 0.0145, "step": 11787 }, { "epoch": 4.65, "learning_rate": 2.3248672021482796e-07, "loss": 0.0152, "step": 11788 }, { "epoch": 4.65, "learning_rate": 2.324615259428524e-07, "loss": 0.0159, "step": 11789 }, { "epoch": 4.65, "learning_rate": 2.324363311418033e-07, "loss": 0.0164, "step": 11790 }, { "epoch": 4.65, "learning_rate": 2.3241113581209132e-07, "loss": 0.0158, "step": 11791 }, { "epoch": 4.65, "learning_rate": 2.3238593995412702e-07, "loss": 0.0174, "step": 11792 }, { "epoch": 4.65, "learning_rate": 2.323607435683212e-07, "loss": 0.0175, "step": 11793 }, { "epoch": 4.65, "learning_rate": 2.323355466550844e-07, "loss": 0.0169, "step": 11794 }, { "epoch": 4.65, "learning_rate": 2.3231034921482737e-07, "loss": 0.0164, "step": 11795 }, { "epoch": 4.65, "learning_rate": 2.322851512479607e-07, "loss": 0.016, "step": 11796 }, { "epoch": 4.65, "learning_rate": 2.322599527548952e-07, "loss": 0.0167, "step": 11797 }, { "epoch": 4.65, "learning_rate": 2.3223475373604145e-07, "loss": 0.0147, "step": 11798 }, { "epoch": 4.65, "learning_rate": 2.3220955419181022e-07, "loss": 0.013, "step": 11799 }, { "epoch": 4.65, "learning_rate": 2.321843541226122e-07, "loss": 0.0172, "step": 11800 }, { "epoch": 4.65, "learning_rate": 2.3215915352885816e-07, "loss": 0.0149, "step": 11801 }, { "epoch": 4.65, "learning_rate": 2.3213395241095878e-07, "loss": 0.0162, "step": 11802 }, { "epoch": 4.65, "learning_rate": 2.3210875076932477e-07, "loss": 0.0153, "step": 11803 }, { "epoch": 4.65, "learning_rate": 2.3208354860436698e-07, "loss": 0.0152, "step": 11804 }, { "epoch": 4.65, "learning_rate": 2.3205834591649603e-07, "loss": 0.0155, "step": 11805 }, { "epoch": 4.65, "learning_rate": 2.3203314270612283e-07, "loss": 0.016, "step": 11806 }, { "epoch": 4.65, "learning_rate": 2.3200793897365806e-07, "loss": 0.0148, "step": 11807 }, { "epoch": 4.65, "learning_rate": 2.3198273471951254e-07, "loss": 0.014, "step": 11808 }, { "epoch": 4.65, "learning_rate": 2.319575299440971e-07, "loss": 0.0187, "step": 11809 }, { "epoch": 4.65, "learning_rate": 2.3193232464782242e-07, "loss": 0.0141, "step": 11810 }, { "epoch": 4.65, "learning_rate": 2.3190711883109943e-07, "loss": 0.015, "step": 11811 }, { "epoch": 4.66, "learning_rate": 2.3188191249433884e-07, "loss": 0.0173, "step": 11812 }, { "epoch": 4.66, "learning_rate": 2.318567056379516e-07, "loss": 0.0158, "step": 11813 }, { "epoch": 4.66, "learning_rate": 2.3183149826234843e-07, "loss": 0.0124, "step": 11814 }, { "epoch": 4.66, "learning_rate": 2.318062903679402e-07, "loss": 0.0161, "step": 11815 }, { "epoch": 4.66, "learning_rate": 2.3178108195513778e-07, "loss": 0.016, "step": 11816 }, { "epoch": 4.66, "learning_rate": 2.3175587302435205e-07, "loss": 0.0156, "step": 11817 }, { "epoch": 4.66, "learning_rate": 2.317306635759939e-07, "loss": 0.013, "step": 11818 }, { "epoch": 4.66, "learning_rate": 2.317054536104741e-07, "loss": 0.0162, "step": 11819 }, { "epoch": 4.66, "learning_rate": 2.3168024312820363e-07, "loss": 0.0142, "step": 11820 }, { "epoch": 4.66, "learning_rate": 2.3165503212959332e-07, "loss": 0.0159, "step": 11821 }, { "epoch": 4.66, "learning_rate": 2.3162982061505413e-07, "loss": 0.0154, "step": 11822 }, { "epoch": 4.66, "learning_rate": 2.316046085849969e-07, "loss": 0.0153, "step": 11823 }, { "epoch": 4.66, "learning_rate": 2.315793960398326e-07, "loss": 0.013, "step": 11824 }, { "epoch": 4.66, "learning_rate": 2.315541829799721e-07, "loss": 0.0146, "step": 11825 }, { "epoch": 4.66, "learning_rate": 2.315289694058264e-07, "loss": 0.0141, "step": 11826 }, { "epoch": 4.66, "learning_rate": 2.3150375531780638e-07, "loss": 0.0154, "step": 11827 }, { "epoch": 4.66, "learning_rate": 2.3147854071632306e-07, "loss": 0.0196, "step": 11828 }, { "epoch": 4.66, "learning_rate": 2.3145332560178744e-07, "loss": 0.0138, "step": 11829 }, { "epoch": 4.66, "learning_rate": 2.3142810997461032e-07, "loss": 0.014, "step": 11830 }, { "epoch": 4.66, "learning_rate": 2.314028938352028e-07, "loss": 0.0128, "step": 11831 }, { "epoch": 4.66, "learning_rate": 2.313776771839758e-07, "loss": 0.0137, "step": 11832 }, { "epoch": 4.66, "learning_rate": 2.3135246002134043e-07, "loss": 0.0157, "step": 11833 }, { "epoch": 4.66, "learning_rate": 2.3132724234770755e-07, "loss": 0.0168, "step": 11834 }, { "epoch": 4.66, "learning_rate": 2.3130202416348826e-07, "loss": 0.0142, "step": 11835 }, { "epoch": 4.66, "learning_rate": 2.3127680546909354e-07, "loss": 0.0152, "step": 11836 }, { "epoch": 4.67, "learning_rate": 2.3125158626493443e-07, "loss": 0.0118, "step": 11837 }, { "epoch": 4.67, "learning_rate": 2.3122636655142198e-07, "loss": 0.0147, "step": 11838 }, { "epoch": 4.67, "learning_rate": 2.3120114632896714e-07, "loss": 0.0147, "step": 11839 }, { "epoch": 4.67, "learning_rate": 2.3117592559798114e-07, "loss": 0.0176, "step": 11840 }, { "epoch": 4.67, "learning_rate": 2.3115070435887482e-07, "loss": 0.0113, "step": 11841 }, { "epoch": 4.67, "learning_rate": 2.311254826120595e-07, "loss": 0.0162, "step": 11842 }, { "epoch": 4.67, "learning_rate": 2.3110026035794603e-07, "loss": 0.0151, "step": 11843 }, { "epoch": 4.67, "learning_rate": 2.310750375969456e-07, "loss": 0.0131, "step": 11844 }, { "epoch": 4.67, "learning_rate": 2.3104981432946928e-07, "loss": 0.0164, "step": 11845 }, { "epoch": 4.67, "learning_rate": 2.3102459055592817e-07, "loss": 0.0176, "step": 11846 }, { "epoch": 4.67, "learning_rate": 2.3099936627673339e-07, "loss": 0.0162, "step": 11847 }, { "epoch": 4.67, "learning_rate": 2.309741414922961e-07, "loss": 0.0148, "step": 11848 }, { "epoch": 4.67, "learning_rate": 2.3094891620302736e-07, "loss": 0.015, "step": 11849 }, { "epoch": 4.67, "learning_rate": 2.3092369040933825e-07, "loss": 0.0143, "step": 11850 }, { "epoch": 4.67, "learning_rate": 2.3089846411164005e-07, "loss": 0.0186, "step": 11851 }, { "epoch": 4.67, "learning_rate": 2.3087323731034384e-07, "loss": 0.0128, "step": 11852 }, { "epoch": 4.67, "learning_rate": 2.308480100058608e-07, "loss": 0.0122, "step": 11853 }, { "epoch": 4.67, "learning_rate": 2.30822782198602e-07, "loss": 0.0153, "step": 11854 }, { "epoch": 4.67, "learning_rate": 2.3079755388897878e-07, "loss": 0.0139, "step": 11855 }, { "epoch": 4.67, "learning_rate": 2.3077232507740217e-07, "loss": 0.0161, "step": 11856 }, { "epoch": 4.67, "learning_rate": 2.3074709576428351e-07, "loss": 0.0157, "step": 11857 }, { "epoch": 4.67, "learning_rate": 2.3072186595003388e-07, "loss": 0.0161, "step": 11858 }, { "epoch": 4.67, "learning_rate": 2.306966356350645e-07, "loss": 0.0162, "step": 11859 }, { "epoch": 4.67, "learning_rate": 2.3067140481978665e-07, "loss": 0.0189, "step": 11860 }, { "epoch": 4.67, "learning_rate": 2.3064617350461152e-07, "loss": 0.016, "step": 11861 }, { "epoch": 4.67, "learning_rate": 2.3062094168995033e-07, "loss": 0.0139, "step": 11862 }, { "epoch": 4.68, "learning_rate": 2.305957093762143e-07, "loss": 0.0154, "step": 11863 }, { "epoch": 4.68, "learning_rate": 2.3057047656381474e-07, "loss": 0.0107, "step": 11864 }, { "epoch": 4.68, "learning_rate": 2.305452432531629e-07, "loss": 0.0185, "step": 11865 }, { "epoch": 4.68, "learning_rate": 2.3052000944467e-07, "loss": 0.0164, "step": 11866 }, { "epoch": 4.68, "learning_rate": 2.3049477513874723e-07, "loss": 0.0155, "step": 11867 }, { "epoch": 4.68, "learning_rate": 2.3046954033580613e-07, "loss": 0.012, "step": 11868 }, { "epoch": 4.68, "learning_rate": 2.3044430503625776e-07, "loss": 0.0144, "step": 11869 }, { "epoch": 4.68, "learning_rate": 2.3041906924051342e-07, "loss": 0.0173, "step": 11870 }, { "epoch": 4.68, "learning_rate": 2.303938329489846e-07, "loss": 0.0167, "step": 11871 }, { "epoch": 4.68, "learning_rate": 2.303685961620824e-07, "loss": 0.0119, "step": 11872 }, { "epoch": 4.68, "learning_rate": 2.303433588802183e-07, "loss": 0.0171, "step": 11873 }, { "epoch": 4.68, "learning_rate": 2.3031812110380354e-07, "loss": 0.0142, "step": 11874 }, { "epoch": 4.68, "learning_rate": 2.302928828332495e-07, "loss": 0.0162, "step": 11875 }, { "epoch": 4.68, "learning_rate": 2.3026764406896747e-07, "loss": 0.0147, "step": 11876 }, { "epoch": 4.68, "learning_rate": 2.3024240481136893e-07, "loss": 0.0142, "step": 11877 }, { "epoch": 4.68, "learning_rate": 2.3021716506086508e-07, "loss": 0.0151, "step": 11878 }, { "epoch": 4.68, "learning_rate": 2.3019192481786737e-07, "loss": 0.0164, "step": 11879 }, { "epoch": 4.68, "learning_rate": 2.3016668408278722e-07, "loss": 0.0147, "step": 11880 }, { "epoch": 4.68, "learning_rate": 2.3014144285603589e-07, "loss": 0.0162, "step": 11881 }, { "epoch": 4.68, "learning_rate": 2.3011620113802493e-07, "loss": 0.0151, "step": 11882 }, { "epoch": 4.68, "learning_rate": 2.3009095892916558e-07, "loss": 0.0159, "step": 11883 }, { "epoch": 4.68, "learning_rate": 2.300657162298694e-07, "loss": 0.0191, "step": 11884 }, { "epoch": 4.68, "learning_rate": 2.3004047304054771e-07, "loss": 0.0147, "step": 11885 }, { "epoch": 4.68, "learning_rate": 2.3001522936161205e-07, "loss": 0.0139, "step": 11886 }, { "epoch": 4.68, "learning_rate": 2.2998998519347364e-07, "loss": 0.0158, "step": 11887 }, { "epoch": 4.69, "learning_rate": 2.2996474053654414e-07, "loss": 0.0174, "step": 11888 }, { "epoch": 4.69, "learning_rate": 2.2993949539123486e-07, "loss": 0.0179, "step": 11889 }, { "epoch": 4.69, "learning_rate": 2.2991424975795733e-07, "loss": 0.0162, "step": 11890 }, { "epoch": 4.69, "learning_rate": 2.29889003637123e-07, "loss": 0.0164, "step": 11891 }, { "epoch": 4.69, "learning_rate": 2.298637570291433e-07, "loss": 0.0138, "step": 11892 }, { "epoch": 4.69, "learning_rate": 2.2983850993442977e-07, "loss": 0.0142, "step": 11893 }, { "epoch": 4.69, "learning_rate": 2.2981326235339384e-07, "loss": 0.015, "step": 11894 }, { "epoch": 4.69, "learning_rate": 2.2978801428644713e-07, "loss": 0.0159, "step": 11895 }, { "epoch": 4.69, "learning_rate": 2.2976276573400097e-07, "loss": 0.0133, "step": 11896 }, { "epoch": 4.69, "learning_rate": 2.2973751669646702e-07, "loss": 0.0147, "step": 11897 }, { "epoch": 4.69, "learning_rate": 2.2971226717425675e-07, "loss": 0.0158, "step": 11898 }, { "epoch": 4.69, "learning_rate": 2.2968701716778166e-07, "loss": 0.0159, "step": 11899 }, { "epoch": 4.69, "learning_rate": 2.296617666774533e-07, "loss": 0.0156, "step": 11900 }, { "epoch": 4.69, "learning_rate": 2.296365157036832e-07, "loss": 0.0118, "step": 11901 }, { "epoch": 4.69, "learning_rate": 2.2961126424688303e-07, "loss": 0.0157, "step": 11902 }, { "epoch": 4.69, "learning_rate": 2.2958601230746416e-07, "loss": 0.0148, "step": 11903 }, { "epoch": 4.69, "learning_rate": 2.295607598858383e-07, "loss": 0.0148, "step": 11904 }, { "epoch": 4.69, "learning_rate": 2.2953550698241698e-07, "loss": 0.014, "step": 11905 }, { "epoch": 4.69, "learning_rate": 2.295102535976118e-07, "loss": 0.0153, "step": 11906 }, { "epoch": 4.69, "learning_rate": 2.294849997318343e-07, "loss": 0.0139, "step": 11907 }, { "epoch": 4.69, "learning_rate": 2.2945974538549615e-07, "loss": 0.0159, "step": 11908 }, { "epoch": 4.69, "learning_rate": 2.2943449055900892e-07, "loss": 0.0153, "step": 11909 }, { "epoch": 4.69, "learning_rate": 2.2940923525278424e-07, "loss": 0.016, "step": 11910 }, { "epoch": 4.69, "learning_rate": 2.2938397946723368e-07, "loss": 0.0118, "step": 11911 }, { "epoch": 4.69, "learning_rate": 2.293587232027689e-07, "loss": 0.0135, "step": 11912 }, { "epoch": 4.7, "learning_rate": 2.2933346645980162e-07, "loss": 0.0161, "step": 11913 }, { "epoch": 4.7, "learning_rate": 2.2930820923874338e-07, "loss": 0.0153, "step": 11914 }, { "epoch": 4.7, "learning_rate": 2.2928295154000588e-07, "loss": 0.0167, "step": 11915 }, { "epoch": 4.7, "learning_rate": 2.2925769336400073e-07, "loss": 0.0181, "step": 11916 }, { "epoch": 4.7, "learning_rate": 2.292324347111397e-07, "loss": 0.0155, "step": 11917 }, { "epoch": 4.7, "learning_rate": 2.2920717558183442e-07, "loss": 0.0144, "step": 11918 }, { "epoch": 4.7, "learning_rate": 2.291819159764965e-07, "loss": 0.0121, "step": 11919 }, { "epoch": 4.7, "learning_rate": 2.291566558955378e-07, "loss": 0.0118, "step": 11920 }, { "epoch": 4.7, "learning_rate": 2.2913139533936983e-07, "loss": 0.0151, "step": 11921 }, { "epoch": 4.7, "learning_rate": 2.2910613430840445e-07, "loss": 0.0144, "step": 11922 }, { "epoch": 4.7, "learning_rate": 2.2908087280305327e-07, "loss": 0.015, "step": 11923 }, { "epoch": 4.7, "learning_rate": 2.2905561082372808e-07, "loss": 0.0124, "step": 11924 }, { "epoch": 4.7, "learning_rate": 2.2903034837084058e-07, "loss": 0.016, "step": 11925 }, { "epoch": 4.7, "learning_rate": 2.2900508544480254e-07, "loss": 0.0169, "step": 11926 }, { "epoch": 4.7, "learning_rate": 2.2897982204602572e-07, "loss": 0.0125, "step": 11927 }, { "epoch": 4.7, "learning_rate": 2.289545581749218e-07, "loss": 0.0125, "step": 11928 }, { "epoch": 4.7, "learning_rate": 2.2892929383190264e-07, "loss": 0.0165, "step": 11929 }, { "epoch": 4.7, "learning_rate": 2.2890402901737995e-07, "loss": 0.0163, "step": 11930 }, { "epoch": 4.7, "learning_rate": 2.2887876373176546e-07, "loss": 0.0126, "step": 11931 }, { "epoch": 4.7, "learning_rate": 2.2885349797547104e-07, "loss": 0.0179, "step": 11932 }, { "epoch": 4.7, "learning_rate": 2.2882823174890848e-07, "loss": 0.0162, "step": 11933 }, { "epoch": 4.7, "learning_rate": 2.288029650524895e-07, "loss": 0.015, "step": 11934 }, { "epoch": 4.7, "learning_rate": 2.28777697886626e-07, "loss": 0.0162, "step": 11935 }, { "epoch": 4.7, "learning_rate": 2.2875243025172976e-07, "loss": 0.0154, "step": 11936 }, { "epoch": 4.7, "learning_rate": 2.2872716214821262e-07, "loss": 0.0121, "step": 11937 }, { "epoch": 4.7, "learning_rate": 2.287018935764864e-07, "loss": 0.0136, "step": 11938 }, { "epoch": 4.71, "learning_rate": 2.2867662453696287e-07, "loss": 0.0127, "step": 11939 }, { "epoch": 4.71, "learning_rate": 2.2865135503005404e-07, "loss": 0.0143, "step": 11940 }, { "epoch": 4.71, "learning_rate": 2.2862608505617156e-07, "loss": 0.0136, "step": 11941 }, { "epoch": 4.71, "learning_rate": 2.2860081461572748e-07, "loss": 0.0137, "step": 11942 }, { "epoch": 4.71, "learning_rate": 2.2857554370913355e-07, "loss": 0.0151, "step": 11943 }, { "epoch": 4.71, "learning_rate": 2.285502723368017e-07, "loss": 0.0158, "step": 11944 }, { "epoch": 4.71, "learning_rate": 2.2852500049914383e-07, "loss": 0.0162, "step": 11945 }, { "epoch": 4.71, "learning_rate": 2.284997281965718e-07, "loss": 0.0134, "step": 11946 }, { "epoch": 4.71, "learning_rate": 2.2847445542949745e-07, "loss": 0.0149, "step": 11947 }, { "epoch": 4.71, "learning_rate": 2.2844918219833285e-07, "loss": 0.0151, "step": 11948 }, { "epoch": 4.71, "learning_rate": 2.2842390850348977e-07, "loss": 0.0142, "step": 11949 }, { "epoch": 4.71, "learning_rate": 2.2839863434538017e-07, "loss": 0.0148, "step": 11950 }, { "epoch": 4.71, "learning_rate": 2.2837335972441602e-07, "loss": 0.0162, "step": 11951 }, { "epoch": 4.71, "learning_rate": 2.2834808464100922e-07, "loss": 0.0147, "step": 11952 }, { "epoch": 4.71, "learning_rate": 2.2832280909557173e-07, "loss": 0.0167, "step": 11953 }, { "epoch": 4.71, "learning_rate": 2.2829753308851542e-07, "loss": 0.0177, "step": 11954 }, { "epoch": 4.71, "learning_rate": 2.2827225662025245e-07, "loss": 0.0144, "step": 11955 }, { "epoch": 4.71, "learning_rate": 2.2824697969119455e-07, "loss": 0.0152, "step": 11956 }, { "epoch": 4.71, "learning_rate": 2.282217023017539e-07, "loss": 0.0152, "step": 11957 }, { "epoch": 4.71, "learning_rate": 2.281964244523424e-07, "loss": 0.0162, "step": 11958 }, { "epoch": 4.71, "learning_rate": 2.2817114614337196e-07, "loss": 0.0152, "step": 11959 }, { "epoch": 4.71, "learning_rate": 2.281458673752547e-07, "loss": 0.0147, "step": 11960 }, { "epoch": 4.71, "learning_rate": 2.2812058814840255e-07, "loss": 0.0139, "step": 11961 }, { "epoch": 4.71, "learning_rate": 2.2809530846322759e-07, "loss": 0.0139, "step": 11962 }, { "epoch": 4.71, "learning_rate": 2.2807002832014174e-07, "loss": 0.016, "step": 11963 }, { "epoch": 4.72, "learning_rate": 2.2804474771955715e-07, "loss": 0.0142, "step": 11964 }, { "epoch": 4.72, "learning_rate": 2.2801946666188575e-07, "loss": 0.0119, "step": 11965 }, { "epoch": 4.72, "learning_rate": 2.2799418514753967e-07, "loss": 0.015, "step": 11966 }, { "epoch": 4.72, "learning_rate": 2.279689031769309e-07, "loss": 0.0161, "step": 11967 }, { "epoch": 4.72, "learning_rate": 2.279436207504715e-07, "loss": 0.014, "step": 11968 }, { "epoch": 4.72, "learning_rate": 2.2791833786857363e-07, "loss": 0.0159, "step": 11969 }, { "epoch": 4.72, "learning_rate": 2.2789305453164917e-07, "loss": 0.015, "step": 11970 }, { "epoch": 4.72, "learning_rate": 2.278677707401104e-07, "loss": 0.0133, "step": 11971 }, { "epoch": 4.72, "learning_rate": 2.2784248649436932e-07, "loss": 0.0151, "step": 11972 }, { "epoch": 4.72, "learning_rate": 2.2781720179483798e-07, "loss": 0.0135, "step": 11973 }, { "epoch": 4.72, "learning_rate": 2.2779191664192857e-07, "loss": 0.0132, "step": 11974 }, { "epoch": 4.72, "learning_rate": 2.2776663103605313e-07, "loss": 0.0152, "step": 11975 }, { "epoch": 4.72, "learning_rate": 2.2774134497762375e-07, "loss": 0.0149, "step": 11976 }, { "epoch": 4.72, "learning_rate": 2.2771605846705272e-07, "loss": 0.0157, "step": 11977 }, { "epoch": 4.72, "learning_rate": 2.2769077150475203e-07, "loss": 0.0164, "step": 11978 }, { "epoch": 4.72, "learning_rate": 2.2766548409113379e-07, "loss": 0.0128, "step": 11979 }, { "epoch": 4.72, "learning_rate": 2.2764019622661026e-07, "loss": 0.0167, "step": 11980 }, { "epoch": 4.72, "learning_rate": 2.2761490791159348e-07, "loss": 0.0171, "step": 11981 }, { "epoch": 4.72, "learning_rate": 2.2758961914649576e-07, "loss": 0.0161, "step": 11982 }, { "epoch": 4.72, "learning_rate": 2.2756432993172908e-07, "loss": 0.0151, "step": 11983 }, { "epoch": 4.72, "learning_rate": 2.2753904026770582e-07, "loss": 0.0137, "step": 11984 }, { "epoch": 4.72, "learning_rate": 2.2751375015483797e-07, "loss": 0.0127, "step": 11985 }, { "epoch": 4.72, "learning_rate": 2.2748845959353785e-07, "loss": 0.015, "step": 11986 }, { "epoch": 4.72, "learning_rate": 2.2746316858421758e-07, "loss": 0.0191, "step": 11987 }, { "epoch": 4.72, "learning_rate": 2.2743787712728944e-07, "loss": 0.0184, "step": 11988 }, { "epoch": 4.72, "learning_rate": 2.274125852231656e-07, "loss": 0.0119, "step": 11989 }, { "epoch": 4.73, "learning_rate": 2.2738729287225827e-07, "loss": 0.0145, "step": 11990 }, { "epoch": 4.73, "learning_rate": 2.2736200007497972e-07, "loss": 0.0156, "step": 11991 }, { "epoch": 4.73, "learning_rate": 2.2733670683174208e-07, "loss": 0.0135, "step": 11992 }, { "epoch": 4.73, "learning_rate": 2.2731141314295775e-07, "loss": 0.014, "step": 11993 }, { "epoch": 4.73, "learning_rate": 2.272861190090389e-07, "loss": 0.0135, "step": 11994 }, { "epoch": 4.73, "learning_rate": 2.2726082443039773e-07, "loss": 0.015, "step": 11995 }, { "epoch": 4.73, "learning_rate": 2.2723552940744653e-07, "loss": 0.0136, "step": 11996 }, { "epoch": 4.73, "learning_rate": 2.2721023394059764e-07, "loss": 0.0158, "step": 11997 }, { "epoch": 4.73, "learning_rate": 2.2718493803026334e-07, "loss": 0.0162, "step": 11998 }, { "epoch": 4.73, "learning_rate": 2.2715964167685574e-07, "loss": 0.0135, "step": 11999 }, { "epoch": 4.73, "learning_rate": 2.2713434488078737e-07, "loss": 0.014, "step": 12000 }, { "epoch": 4.73, "learning_rate": 2.2710904764247032e-07, "loss": 0.0168, "step": 12001 }, { "epoch": 4.73, "learning_rate": 2.2708374996231708e-07, "loss": 0.0118, "step": 12002 }, { "epoch": 4.73, "learning_rate": 2.2705845184073983e-07, "loss": 0.0171, "step": 12003 }, { "epoch": 4.73, "learning_rate": 2.2703315327815095e-07, "loss": 0.0146, "step": 12004 }, { "epoch": 4.73, "learning_rate": 2.2700785427496277e-07, "loss": 0.0185, "step": 12005 }, { "epoch": 4.73, "learning_rate": 2.269825548315876e-07, "loss": 0.0151, "step": 12006 }, { "epoch": 4.73, "learning_rate": 2.269572549484378e-07, "loss": 0.0113, "step": 12007 }, { "epoch": 4.73, "learning_rate": 2.2693195462592572e-07, "loss": 0.0192, "step": 12008 }, { "epoch": 4.73, "learning_rate": 2.2690665386446371e-07, "loss": 0.0163, "step": 12009 }, { "epoch": 4.73, "learning_rate": 2.2688135266446414e-07, "loss": 0.0172, "step": 12010 }, { "epoch": 4.73, "learning_rate": 2.2685605102633937e-07, "loss": 0.0168, "step": 12011 }, { "epoch": 4.73, "learning_rate": 2.2683074895050174e-07, "loss": 0.0144, "step": 12012 }, { "epoch": 4.73, "learning_rate": 2.2680544643736374e-07, "loss": 0.0152, "step": 12013 }, { "epoch": 4.73, "learning_rate": 2.2678014348733772e-07, "loss": 0.0166, "step": 12014 }, { "epoch": 4.74, "learning_rate": 2.2675484010083599e-07, "loss": 0.0147, "step": 12015 }, { "epoch": 4.74, "learning_rate": 2.267295362782711e-07, "loss": 0.0162, "step": 12016 }, { "epoch": 4.74, "learning_rate": 2.2670423202005534e-07, "loss": 0.017, "step": 12017 }, { "epoch": 4.74, "learning_rate": 2.2667892732660125e-07, "loss": 0.0167, "step": 12018 }, { "epoch": 4.74, "learning_rate": 2.266536221983211e-07, "loss": 0.0127, "step": 12019 }, { "epoch": 4.74, "learning_rate": 2.266283166356275e-07, "loss": 0.014, "step": 12020 }, { "epoch": 4.74, "learning_rate": 2.2660301063893272e-07, "loss": 0.0159, "step": 12021 }, { "epoch": 4.74, "learning_rate": 2.265777042086494e-07, "loss": 0.0126, "step": 12022 }, { "epoch": 4.74, "learning_rate": 2.2655239734518982e-07, "loss": 0.0161, "step": 12023 }, { "epoch": 4.74, "learning_rate": 2.2652709004896656e-07, "loss": 0.0156, "step": 12024 }, { "epoch": 4.74, "learning_rate": 2.26501782320392e-07, "loss": 0.0157, "step": 12025 }, { "epoch": 4.74, "learning_rate": 2.264764741598787e-07, "loss": 0.0114, "step": 12026 }, { "epoch": 4.74, "learning_rate": 2.2645116556783912e-07, "loss": 0.0156, "step": 12027 }, { "epoch": 4.74, "learning_rate": 2.264258565446857e-07, "loss": 0.0143, "step": 12028 }, { "epoch": 4.74, "learning_rate": 2.2640054709083106e-07, "loss": 0.0161, "step": 12029 }, { "epoch": 4.74, "learning_rate": 2.2637523720668755e-07, "loss": 0.017, "step": 12030 }, { "epoch": 4.74, "learning_rate": 2.263499268926678e-07, "loss": 0.0142, "step": 12031 }, { "epoch": 4.74, "learning_rate": 2.2632461614918425e-07, "loss": 0.014, "step": 12032 }, { "epoch": 4.74, "learning_rate": 2.262993049766495e-07, "loss": 0.0167, "step": 12033 }, { "epoch": 4.74, "learning_rate": 2.2627399337547603e-07, "loss": 0.0154, "step": 12034 }, { "epoch": 4.74, "learning_rate": 2.2624868134607644e-07, "loss": 0.0143, "step": 12035 }, { "epoch": 4.74, "learning_rate": 2.2622336888886323e-07, "loss": 0.012, "step": 12036 }, { "epoch": 4.74, "learning_rate": 2.261980560042489e-07, "loss": 0.0118, "step": 12037 }, { "epoch": 4.74, "learning_rate": 2.2617274269264617e-07, "loss": 0.0164, "step": 12038 }, { "epoch": 4.74, "learning_rate": 2.2614742895446747e-07, "loss": 0.0173, "step": 12039 }, { "epoch": 4.75, "learning_rate": 2.2612211479012542e-07, "loss": 0.014, "step": 12040 }, { "epoch": 4.75, "learning_rate": 2.2609680020003253e-07, "loss": 0.0164, "step": 12041 }, { "epoch": 4.75, "learning_rate": 2.2607148518460154e-07, "loss": 0.0138, "step": 12042 }, { "epoch": 4.75, "learning_rate": 2.2604616974424493e-07, "loss": 0.0168, "step": 12043 }, { "epoch": 4.75, "learning_rate": 2.2602085387937534e-07, "loss": 0.0155, "step": 12044 }, { "epoch": 4.75, "learning_rate": 2.2599553759040536e-07, "loss": 0.0138, "step": 12045 }, { "epoch": 4.75, "learning_rate": 2.259702208777477e-07, "loss": 0.0129, "step": 12046 }, { "epoch": 4.75, "learning_rate": 2.259449037418148e-07, "loss": 0.0146, "step": 12047 }, { "epoch": 4.75, "learning_rate": 2.2591958618301953e-07, "loss": 0.0172, "step": 12048 }, { "epoch": 4.75, "learning_rate": 2.2589426820177433e-07, "loss": 0.0161, "step": 12049 }, { "epoch": 4.75, "learning_rate": 2.258689497984919e-07, "loss": 0.0121, "step": 12050 }, { "epoch": 4.75, "learning_rate": 2.2584363097358489e-07, "loss": 0.0148, "step": 12051 }, { "epoch": 4.75, "learning_rate": 2.25818311727466e-07, "loss": 0.0142, "step": 12052 }, { "epoch": 4.75, "learning_rate": 2.2579299206054791e-07, "loss": 0.0127, "step": 12053 }, { "epoch": 4.75, "learning_rate": 2.2576767197324317e-07, "loss": 0.017, "step": 12054 }, { "epoch": 4.75, "learning_rate": 2.2574235146596462e-07, "loss": 0.0184, "step": 12055 }, { "epoch": 4.75, "learning_rate": 2.2571703053912476e-07, "loss": 0.0166, "step": 12056 }, { "epoch": 4.75, "learning_rate": 2.2569170919313652e-07, "loss": 0.0154, "step": 12057 }, { "epoch": 4.75, "learning_rate": 2.2566638742841238e-07, "loss": 0.0147, "step": 12058 }, { "epoch": 4.75, "learning_rate": 2.2564106524536517e-07, "loss": 0.0196, "step": 12059 }, { "epoch": 4.75, "learning_rate": 2.2561574264440754e-07, "loss": 0.0127, "step": 12060 }, { "epoch": 4.75, "learning_rate": 2.2559041962595225e-07, "loss": 0.0181, "step": 12061 }, { "epoch": 4.75, "learning_rate": 2.2556509619041205e-07, "loss": 0.0153, "step": 12062 }, { "epoch": 4.75, "learning_rate": 2.2553977233819954e-07, "loss": 0.0144, "step": 12063 }, { "epoch": 4.75, "learning_rate": 2.2551444806972765e-07, "loss": 0.0115, "step": 12064 }, { "epoch": 4.75, "learning_rate": 2.2548912338540895e-07, "loss": 0.0168, "step": 12065 }, { "epoch": 4.76, "learning_rate": 2.2546379828565638e-07, "loss": 0.0158, "step": 12066 }, { "epoch": 4.76, "learning_rate": 2.2543847277088252e-07, "loss": 0.0098, "step": 12067 }, { "epoch": 4.76, "learning_rate": 2.2541314684150024e-07, "loss": 0.0156, "step": 12068 }, { "epoch": 4.76, "learning_rate": 2.253878204979223e-07, "loss": 0.014, "step": 12069 }, { "epoch": 4.76, "learning_rate": 2.2536249374056143e-07, "loss": 0.0153, "step": 12070 }, { "epoch": 4.76, "learning_rate": 2.2533716656983048e-07, "loss": 0.0135, "step": 12071 }, { "epoch": 4.76, "learning_rate": 2.2531183898614222e-07, "loss": 0.0165, "step": 12072 }, { "epoch": 4.76, "learning_rate": 2.252865109899095e-07, "loss": 0.0153, "step": 12073 }, { "epoch": 4.76, "learning_rate": 2.25261182581545e-07, "loss": 0.0143, "step": 12074 }, { "epoch": 4.76, "learning_rate": 2.2523585376146167e-07, "loss": 0.0116, "step": 12075 }, { "epoch": 4.76, "learning_rate": 2.2521052453007225e-07, "loss": 0.0191, "step": 12076 }, { "epoch": 4.76, "learning_rate": 2.2518519488778963e-07, "loss": 0.0148, "step": 12077 }, { "epoch": 4.76, "learning_rate": 2.2515986483502662e-07, "loss": 0.0137, "step": 12078 }, { "epoch": 4.76, "learning_rate": 2.2513453437219603e-07, "loss": 0.0147, "step": 12079 }, { "epoch": 4.76, "learning_rate": 2.2510920349971072e-07, "loss": 0.0172, "step": 12080 }, { "epoch": 4.76, "learning_rate": 2.2508387221798363e-07, "loss": 0.0157, "step": 12081 }, { "epoch": 4.76, "learning_rate": 2.2505854052742746e-07, "loss": 0.0173, "step": 12082 }, { "epoch": 4.76, "learning_rate": 2.250332084284552e-07, "loss": 0.016, "step": 12083 }, { "epoch": 4.76, "learning_rate": 2.2500787592147965e-07, "loss": 0.0162, "step": 12084 }, { "epoch": 4.76, "learning_rate": 2.2498254300691375e-07, "loss": 0.0147, "step": 12085 }, { "epoch": 4.76, "learning_rate": 2.2495720968517036e-07, "loss": 0.0125, "step": 12086 }, { "epoch": 4.76, "learning_rate": 2.249318759566624e-07, "loss": 0.0158, "step": 12087 }, { "epoch": 4.76, "learning_rate": 2.2490654182180274e-07, "loss": 0.013, "step": 12088 }, { "epoch": 4.76, "learning_rate": 2.2488120728100432e-07, "loss": 0.016, "step": 12089 }, { "epoch": 4.76, "learning_rate": 2.2485587233468003e-07, "loss": 0.0176, "step": 12090 }, { "epoch": 4.77, "learning_rate": 2.248305369832428e-07, "loss": 0.0161, "step": 12091 }, { "epoch": 4.77, "learning_rate": 2.2480520122710552e-07, "loss": 0.0166, "step": 12092 }, { "epoch": 4.77, "learning_rate": 2.2477986506668118e-07, "loss": 0.0121, "step": 12093 }, { "epoch": 4.77, "learning_rate": 2.2475452850238272e-07, "loss": 0.0166, "step": 12094 }, { "epoch": 4.77, "learning_rate": 2.2472919153462303e-07, "loss": 0.0133, "step": 12095 }, { "epoch": 4.77, "learning_rate": 2.2470385416381514e-07, "loss": 0.0157, "step": 12096 }, { "epoch": 4.77, "learning_rate": 2.2467851639037198e-07, "loss": 0.0141, "step": 12097 }, { "epoch": 4.77, "learning_rate": 2.246531782147065e-07, "loss": 0.0139, "step": 12098 }, { "epoch": 4.77, "learning_rate": 2.2462783963723164e-07, "loss": 0.0146, "step": 12099 }, { "epoch": 4.77, "learning_rate": 2.2460250065836052e-07, "loss": 0.0166, "step": 12100 }, { "epoch": 4.77, "learning_rate": 2.245771612785059e-07, "loss": 0.0134, "step": 12101 }, { "epoch": 4.77, "learning_rate": 2.2455182149808102e-07, "loss": 0.0158, "step": 12102 }, { "epoch": 4.77, "learning_rate": 2.2452648131749869e-07, "loss": 0.0172, "step": 12103 }, { "epoch": 4.77, "learning_rate": 2.2450114073717204e-07, "loss": 0.0163, "step": 12104 }, { "epoch": 4.77, "learning_rate": 2.24475799757514e-07, "loss": 0.0129, "step": 12105 }, { "epoch": 4.77, "learning_rate": 2.2445045837893768e-07, "loss": 0.0174, "step": 12106 }, { "epoch": 4.77, "learning_rate": 2.2442511660185595e-07, "loss": 0.0148, "step": 12107 }, { "epoch": 4.77, "learning_rate": 2.2439977442668202e-07, "loss": 0.0147, "step": 12108 }, { "epoch": 4.77, "learning_rate": 2.2437443185382885e-07, "loss": 0.0138, "step": 12109 }, { "epoch": 4.77, "learning_rate": 2.2434908888370942e-07, "loss": 0.0134, "step": 12110 }, { "epoch": 4.77, "learning_rate": 2.2432374551673692e-07, "loss": 0.014, "step": 12111 }, { "epoch": 4.77, "learning_rate": 2.2429840175332425e-07, "loss": 0.0137, "step": 12112 }, { "epoch": 4.77, "learning_rate": 2.2427305759388466e-07, "loss": 0.0162, "step": 12113 }, { "epoch": 4.77, "learning_rate": 2.2424771303883104e-07, "loss": 0.0151, "step": 12114 }, { "epoch": 4.77, "learning_rate": 2.2422236808857658e-07, "loss": 0.0192, "step": 12115 }, { "epoch": 4.78, "learning_rate": 2.2419702274353437e-07, "loss": 0.0143, "step": 12116 }, { "epoch": 4.78, "learning_rate": 2.241716770041174e-07, "loss": 0.0139, "step": 12117 }, { "epoch": 4.78, "learning_rate": 2.241463308707389e-07, "loss": 0.016, "step": 12118 }, { "epoch": 4.78, "learning_rate": 2.2412098434381183e-07, "loss": 0.0175, "step": 12119 }, { "epoch": 4.78, "learning_rate": 2.2409563742374944e-07, "loss": 0.0172, "step": 12120 }, { "epoch": 4.78, "learning_rate": 2.2407029011096474e-07, "loss": 0.0158, "step": 12121 }, { "epoch": 4.78, "learning_rate": 2.2404494240587092e-07, "loss": 0.0155, "step": 12122 }, { "epoch": 4.78, "learning_rate": 2.240195943088811e-07, "loss": 0.0139, "step": 12123 }, { "epoch": 4.78, "learning_rate": 2.2399424582040835e-07, "loss": 0.0132, "step": 12124 }, { "epoch": 4.78, "learning_rate": 2.239688969408659e-07, "loss": 0.0225, "step": 12125 }, { "epoch": 4.78, "learning_rate": 2.2394354767066685e-07, "loss": 0.0169, "step": 12126 }, { "epoch": 4.78, "learning_rate": 2.239181980102243e-07, "loss": 0.0116, "step": 12127 }, { "epoch": 4.78, "learning_rate": 2.238928479599516e-07, "loss": 0.0148, "step": 12128 }, { "epoch": 4.78, "learning_rate": 2.2386749752026172e-07, "loss": 0.016, "step": 12129 }, { "epoch": 4.78, "learning_rate": 2.2384214669156786e-07, "loss": 0.0134, "step": 12130 }, { "epoch": 4.78, "learning_rate": 2.2381679547428334e-07, "loss": 0.0161, "step": 12131 }, { "epoch": 4.78, "learning_rate": 2.2379144386882114e-07, "loss": 0.0166, "step": 12132 }, { "epoch": 4.78, "learning_rate": 2.2376609187559467e-07, "loss": 0.0193, "step": 12133 }, { "epoch": 4.78, "learning_rate": 2.2374073949501696e-07, "loss": 0.0156, "step": 12134 }, { "epoch": 4.78, "learning_rate": 2.2371538672750128e-07, "loss": 0.0182, "step": 12135 }, { "epoch": 4.78, "learning_rate": 2.2369003357346084e-07, "loss": 0.0122, "step": 12136 }, { "epoch": 4.78, "learning_rate": 2.236646800333089e-07, "loss": 0.0133, "step": 12137 }, { "epoch": 4.78, "learning_rate": 2.2363932610745865e-07, "loss": 0.0138, "step": 12138 }, { "epoch": 4.78, "learning_rate": 2.236139717963232e-07, "loss": 0.0156, "step": 12139 }, { "epoch": 4.78, "learning_rate": 2.2358861710031604e-07, "loss": 0.015, "step": 12140 }, { "epoch": 4.78, "learning_rate": 2.2356326201985016e-07, "loss": 0.0167, "step": 12141 }, { "epoch": 4.79, "learning_rate": 2.2353790655533897e-07, "loss": 0.0174, "step": 12142 }, { "epoch": 4.79, "learning_rate": 2.2351255070719562e-07, "loss": 0.0175, "step": 12143 }, { "epoch": 4.79, "learning_rate": 2.234871944758335e-07, "loss": 0.0117, "step": 12144 }, { "epoch": 4.79, "learning_rate": 2.2346183786166582e-07, "loss": 0.0141, "step": 12145 }, { "epoch": 4.79, "learning_rate": 2.2343648086510584e-07, "loss": 0.0152, "step": 12146 }, { "epoch": 4.79, "learning_rate": 2.2341112348656676e-07, "loss": 0.0173, "step": 12147 }, { "epoch": 4.79, "learning_rate": 2.2338576572646205e-07, "loss": 0.0146, "step": 12148 }, { "epoch": 4.79, "learning_rate": 2.2336040758520488e-07, "loss": 0.0146, "step": 12149 }, { "epoch": 4.79, "learning_rate": 2.2333504906320848e-07, "loss": 0.0175, "step": 12150 }, { "epoch": 4.79, "learning_rate": 2.2330969016088633e-07, "loss": 0.0178, "step": 12151 }, { "epoch": 4.79, "learning_rate": 2.232843308786516e-07, "loss": 0.0159, "step": 12152 }, { "epoch": 4.79, "learning_rate": 2.2325897121691773e-07, "loss": 0.016, "step": 12153 }, { "epoch": 4.79, "learning_rate": 2.232336111760979e-07, "loss": 0.0154, "step": 12154 }, { "epoch": 4.79, "learning_rate": 2.232082507566056e-07, "loss": 0.019, "step": 12155 }, { "epoch": 4.79, "learning_rate": 2.2318288995885402e-07, "loss": 0.0166, "step": 12156 }, { "epoch": 4.79, "learning_rate": 2.231575287832566e-07, "loss": 0.0138, "step": 12157 }, { "epoch": 4.79, "learning_rate": 2.231321672302267e-07, "loss": 0.0152, "step": 12158 }, { "epoch": 4.79, "learning_rate": 2.2310680530017754e-07, "loss": 0.0152, "step": 12159 }, { "epoch": 4.79, "learning_rate": 2.2308144299352266e-07, "loss": 0.0149, "step": 12160 }, { "epoch": 4.79, "learning_rate": 2.2305608031067523e-07, "loss": 0.0146, "step": 12161 }, { "epoch": 4.79, "learning_rate": 2.2303071725204883e-07, "loss": 0.0117, "step": 12162 }, { "epoch": 4.79, "learning_rate": 2.2300535381805665e-07, "loss": 0.0142, "step": 12163 }, { "epoch": 4.79, "learning_rate": 2.2297999000911227e-07, "loss": 0.0179, "step": 12164 }, { "epoch": 4.79, "learning_rate": 2.2295462582562892e-07, "loss": 0.0148, "step": 12165 }, { "epoch": 4.79, "learning_rate": 2.2292926126802006e-07, "loss": 0.0142, "step": 12166 }, { "epoch": 4.8, "learning_rate": 2.2290389633669916e-07, "loss": 0.0156, "step": 12167 }, { "epoch": 4.8, "learning_rate": 2.228785310320795e-07, "loss": 0.0165, "step": 12168 }, { "epoch": 4.8, "learning_rate": 2.2285316535457456e-07, "loss": 0.0134, "step": 12169 }, { "epoch": 4.8, "learning_rate": 2.228277993045977e-07, "loss": 0.0145, "step": 12170 }, { "epoch": 4.8, "learning_rate": 2.2280243288256248e-07, "loss": 0.0148, "step": 12171 }, { "epoch": 4.8, "learning_rate": 2.2277706608888222e-07, "loss": 0.0139, "step": 12172 }, { "epoch": 4.8, "learning_rate": 2.2275169892397043e-07, "loss": 0.0137, "step": 12173 }, { "epoch": 4.8, "learning_rate": 2.227263313882405e-07, "loss": 0.016, "step": 12174 }, { "epoch": 4.8, "learning_rate": 2.2270096348210593e-07, "loss": 0.015, "step": 12175 }, { "epoch": 4.8, "learning_rate": 2.2267559520598014e-07, "loss": 0.0133, "step": 12176 }, { "epoch": 4.8, "learning_rate": 2.2265022656027664e-07, "loss": 0.0176, "step": 12177 }, { "epoch": 4.8, "learning_rate": 2.2262485754540886e-07, "loss": 0.0152, "step": 12178 }, { "epoch": 4.8, "learning_rate": 2.2259948816179027e-07, "loss": 0.0164, "step": 12179 }, { "epoch": 4.8, "learning_rate": 2.2257411840983437e-07, "loss": 0.0167, "step": 12180 }, { "epoch": 4.8, "learning_rate": 2.2254874828995463e-07, "loss": 0.0144, "step": 12181 }, { "epoch": 4.8, "learning_rate": 2.2252337780256462e-07, "loss": 0.0156, "step": 12182 }, { "epoch": 4.8, "learning_rate": 2.2249800694807774e-07, "loss": 0.0139, "step": 12183 }, { "epoch": 4.8, "learning_rate": 2.2247263572690756e-07, "loss": 0.0156, "step": 12184 }, { "epoch": 4.8, "learning_rate": 2.2244726413946756e-07, "loss": 0.0134, "step": 12185 }, { "epoch": 4.8, "learning_rate": 2.2242189218617126e-07, "loss": 0.0181, "step": 12186 }, { "epoch": 4.8, "learning_rate": 2.2239651986743223e-07, "loss": 0.0126, "step": 12187 }, { "epoch": 4.8, "learning_rate": 2.2237114718366393e-07, "loss": 0.0153, "step": 12188 }, { "epoch": 4.8, "learning_rate": 2.2234577413527997e-07, "loss": 0.014, "step": 12189 }, { "epoch": 4.8, "learning_rate": 2.2232040072269385e-07, "loss": 0.017, "step": 12190 }, { "epoch": 4.8, "learning_rate": 2.2229502694631911e-07, "loss": 0.0151, "step": 12191 }, { "epoch": 4.8, "learning_rate": 2.2226965280656928e-07, "loss": 0.0146, "step": 12192 }, { "epoch": 4.81, "learning_rate": 2.2224427830385803e-07, "loss": 0.0126, "step": 12193 }, { "epoch": 4.81, "learning_rate": 2.2221890343859875e-07, "loss": 0.0155, "step": 12194 }, { "epoch": 4.81, "learning_rate": 2.221935282112052e-07, "loss": 0.0144, "step": 12195 }, { "epoch": 4.81, "learning_rate": 2.2216815262209078e-07, "loss": 0.0159, "step": 12196 }, { "epoch": 4.81, "learning_rate": 2.2214277667166924e-07, "loss": 0.0143, "step": 12197 }, { "epoch": 4.81, "learning_rate": 2.2211740036035412e-07, "loss": 0.0145, "step": 12198 }, { "epoch": 4.81, "learning_rate": 2.2209202368855894e-07, "loss": 0.011, "step": 12199 }, { "epoch": 4.81, "learning_rate": 2.2206664665669735e-07, "loss": 0.0127, "step": 12200 }, { "epoch": 4.81, "learning_rate": 2.2204126926518293e-07, "loss": 0.0133, "step": 12201 }, { "epoch": 4.81, "learning_rate": 2.2201589151442936e-07, "loss": 0.0179, "step": 12202 }, { "epoch": 4.81, "learning_rate": 2.2199051340485017e-07, "loss": 0.0164, "step": 12203 }, { "epoch": 4.81, "learning_rate": 2.2196513493685912e-07, "loss": 0.0136, "step": 12204 }, { "epoch": 4.81, "learning_rate": 2.2193975611086965e-07, "loss": 0.0148, "step": 12205 }, { "epoch": 4.81, "learning_rate": 2.2191437692729556e-07, "loss": 0.0171, "step": 12206 }, { "epoch": 4.81, "learning_rate": 2.2188899738655045e-07, "loss": 0.0121, "step": 12207 }, { "epoch": 4.81, "learning_rate": 2.2186361748904794e-07, "loss": 0.0144, "step": 12208 }, { "epoch": 4.81, "learning_rate": 2.2183823723520166e-07, "loss": 0.0165, "step": 12209 }, { "epoch": 4.81, "learning_rate": 2.2181285662542537e-07, "loss": 0.0159, "step": 12210 }, { "epoch": 4.81, "learning_rate": 2.2178747566013261e-07, "loss": 0.0162, "step": 12211 }, { "epoch": 4.81, "learning_rate": 2.2176209433973713e-07, "loss": 0.0182, "step": 12212 }, { "epoch": 4.81, "learning_rate": 2.217367126646526e-07, "loss": 0.016, "step": 12213 }, { "epoch": 4.81, "learning_rate": 2.2171133063529263e-07, "loss": 0.0144, "step": 12214 }, { "epoch": 4.81, "learning_rate": 2.2168594825207104e-07, "loss": 0.0128, "step": 12215 }, { "epoch": 4.81, "learning_rate": 2.216605655154014e-07, "loss": 0.0152, "step": 12216 }, { "epoch": 4.81, "learning_rate": 2.216351824256975e-07, "loss": 0.018, "step": 12217 }, { "epoch": 4.82, "learning_rate": 2.21609798983373e-07, "loss": 0.0161, "step": 12218 }, { "epoch": 4.82, "learning_rate": 2.2158441518884158e-07, "loss": 0.013, "step": 12219 }, { "epoch": 4.82, "learning_rate": 2.2155903104251704e-07, "loss": 0.0128, "step": 12220 }, { "epoch": 4.82, "learning_rate": 2.2153364654481304e-07, "loss": 0.0151, "step": 12221 }, { "epoch": 4.82, "learning_rate": 2.215082616961433e-07, "loss": 0.0164, "step": 12222 }, { "epoch": 4.82, "learning_rate": 2.2148287649692162e-07, "loss": 0.0163, "step": 12223 }, { "epoch": 4.82, "learning_rate": 2.2145749094756173e-07, "loss": 0.0149, "step": 12224 }, { "epoch": 4.82, "learning_rate": 2.2143210504847726e-07, "loss": 0.0112, "step": 12225 }, { "epoch": 4.82, "learning_rate": 2.2140671880008215e-07, "loss": 0.0167, "step": 12226 }, { "epoch": 4.82, "learning_rate": 2.2138133220279004e-07, "loss": 0.0123, "step": 12227 }, { "epoch": 4.82, "learning_rate": 2.2135594525701468e-07, "loss": 0.0143, "step": 12228 }, { "epoch": 4.82, "learning_rate": 2.2133055796316992e-07, "loss": 0.015, "step": 12229 }, { "epoch": 4.82, "learning_rate": 2.2130517032166938e-07, "loss": 0.0143, "step": 12230 }, { "epoch": 4.82, "learning_rate": 2.2127978233292705e-07, "loss": 0.0147, "step": 12231 }, { "epoch": 4.82, "learning_rate": 2.2125439399735657e-07, "loss": 0.0147, "step": 12232 }, { "epoch": 4.82, "learning_rate": 2.212290053153718e-07, "loss": 0.017, "step": 12233 }, { "epoch": 4.82, "learning_rate": 2.2120361628738643e-07, "loss": 0.0146, "step": 12234 }, { "epoch": 4.82, "learning_rate": 2.211782269138144e-07, "loss": 0.014, "step": 12235 }, { "epoch": 4.82, "learning_rate": 2.2115283719506945e-07, "loss": 0.0125, "step": 12236 }, { "epoch": 4.82, "learning_rate": 2.211274471315654e-07, "loss": 0.0161, "step": 12237 }, { "epoch": 4.82, "learning_rate": 2.2110205672371608e-07, "loss": 0.013, "step": 12238 }, { "epoch": 4.82, "learning_rate": 2.2107666597193528e-07, "loss": 0.0172, "step": 12239 }, { "epoch": 4.82, "learning_rate": 2.2105127487663688e-07, "loss": 0.0124, "step": 12240 }, { "epoch": 4.82, "learning_rate": 2.2102588343823466e-07, "loss": 0.0102, "step": 12241 }, { "epoch": 4.82, "learning_rate": 2.2100049165714255e-07, "loss": 0.017, "step": 12242 }, { "epoch": 4.83, "learning_rate": 2.209750995337743e-07, "loss": 0.0101, "step": 12243 }, { "epoch": 4.83, "learning_rate": 2.209497070685438e-07, "loss": 0.0121, "step": 12244 }, { "epoch": 4.83, "learning_rate": 2.2092431426186495e-07, "loss": 0.0182, "step": 12245 }, { "epoch": 4.83, "learning_rate": 2.2089892111415156e-07, "loss": 0.0147, "step": 12246 }, { "epoch": 4.83, "learning_rate": 2.2087352762581754e-07, "loss": 0.0142, "step": 12247 }, { "epoch": 4.83, "learning_rate": 2.2084813379727675e-07, "loss": 0.0125, "step": 12248 }, { "epoch": 4.83, "learning_rate": 2.2082273962894306e-07, "loss": 0.0183, "step": 12249 }, { "epoch": 4.83, "learning_rate": 2.207973451212303e-07, "loss": 0.0144, "step": 12250 }, { "epoch": 4.83, "learning_rate": 2.2077195027455246e-07, "loss": 0.0126, "step": 12251 }, { "epoch": 4.83, "learning_rate": 2.207465550893234e-07, "loss": 0.0147, "step": 12252 }, { "epoch": 4.83, "learning_rate": 2.2072115956595706e-07, "loss": 0.0152, "step": 12253 }, { "epoch": 4.83, "learning_rate": 2.2069576370486728e-07, "loss": 0.014, "step": 12254 }, { "epoch": 4.83, "learning_rate": 2.20670367506468e-07, "loss": 0.0152, "step": 12255 }, { "epoch": 4.83, "learning_rate": 2.2064497097117317e-07, "loss": 0.0109, "step": 12256 }, { "epoch": 4.83, "learning_rate": 2.2061957409939666e-07, "loss": 0.0141, "step": 12257 }, { "epoch": 4.83, "learning_rate": 2.2059417689155247e-07, "loss": 0.0114, "step": 12258 }, { "epoch": 4.83, "learning_rate": 2.2056877934805445e-07, "loss": 0.0172, "step": 12259 }, { "epoch": 4.83, "learning_rate": 2.2054338146931663e-07, "loss": 0.0134, "step": 12260 }, { "epoch": 4.83, "learning_rate": 2.2051798325575284e-07, "loss": 0.0166, "step": 12261 }, { "epoch": 4.83, "learning_rate": 2.2049258470777717e-07, "loss": 0.0135, "step": 12262 }, { "epoch": 4.83, "learning_rate": 2.2046718582580352e-07, "loss": 0.0154, "step": 12263 }, { "epoch": 4.83, "learning_rate": 2.204417866102458e-07, "loss": 0.0182, "step": 12264 }, { "epoch": 4.83, "learning_rate": 2.2041638706151805e-07, "loss": 0.0165, "step": 12265 }, { "epoch": 4.83, "learning_rate": 2.2039098718003424e-07, "loss": 0.0132, "step": 12266 }, { "epoch": 4.83, "learning_rate": 2.203655869662083e-07, "loss": 0.0129, "step": 12267 }, { "epoch": 4.83, "learning_rate": 2.2034018642045425e-07, "loss": 0.0141, "step": 12268 }, { "epoch": 4.84, "learning_rate": 2.2031478554318612e-07, "loss": 0.0149, "step": 12269 }, { "epoch": 4.84, "learning_rate": 2.202893843348178e-07, "loss": 0.0166, "step": 12270 }, { "epoch": 4.84, "learning_rate": 2.202639827957634e-07, "loss": 0.0148, "step": 12271 }, { "epoch": 4.84, "learning_rate": 2.2023858092643684e-07, "loss": 0.0157, "step": 12272 }, { "epoch": 4.84, "learning_rate": 2.202131787272522e-07, "loss": 0.0133, "step": 12273 }, { "epoch": 4.84, "learning_rate": 2.2018777619862345e-07, "loss": 0.0126, "step": 12274 }, { "epoch": 4.84, "learning_rate": 2.201623733409646e-07, "loss": 0.0165, "step": 12275 }, { "epoch": 4.84, "learning_rate": 2.2013697015468974e-07, "loss": 0.0119, "step": 12276 }, { "epoch": 4.84, "learning_rate": 2.201115666402129e-07, "loss": 0.0125, "step": 12277 }, { "epoch": 4.84, "learning_rate": 2.2008616279794808e-07, "loss": 0.0167, "step": 12278 }, { "epoch": 4.84, "learning_rate": 2.200607586283093e-07, "loss": 0.0122, "step": 12279 }, { "epoch": 4.84, "learning_rate": 2.2003535413171067e-07, "loss": 0.0155, "step": 12280 }, { "epoch": 4.84, "learning_rate": 2.2000994930856618e-07, "loss": 0.0122, "step": 12281 }, { "epoch": 4.84, "learning_rate": 2.1998454415928998e-07, "loss": 0.0137, "step": 12282 }, { "epoch": 4.84, "learning_rate": 2.1995913868429602e-07, "loss": 0.0186, "step": 12283 }, { "epoch": 4.84, "learning_rate": 2.1993373288399847e-07, "loss": 0.0158, "step": 12284 }, { "epoch": 4.84, "learning_rate": 2.1990832675881137e-07, "loss": 0.015, "step": 12285 }, { "epoch": 4.84, "learning_rate": 2.1988292030914883e-07, "loss": 0.0146, "step": 12286 }, { "epoch": 4.84, "learning_rate": 2.1985751353542484e-07, "loss": 0.0165, "step": 12287 }, { "epoch": 4.84, "learning_rate": 2.1983210643805364e-07, "loss": 0.0144, "step": 12288 }, { "epoch": 4.84, "learning_rate": 2.198066990174492e-07, "loss": 0.0148, "step": 12289 }, { "epoch": 4.84, "learning_rate": 2.1978129127402564e-07, "loss": 0.0149, "step": 12290 }, { "epoch": 4.84, "learning_rate": 2.1975588320819715e-07, "loss": 0.0145, "step": 12291 }, { "epoch": 4.84, "learning_rate": 2.1973047482037774e-07, "loss": 0.0123, "step": 12292 }, { "epoch": 4.84, "learning_rate": 2.1970506611098164e-07, "loss": 0.0151, "step": 12293 }, { "epoch": 4.85, "learning_rate": 2.1967965708042286e-07, "loss": 0.0176, "step": 12294 }, { "epoch": 4.85, "learning_rate": 2.1965424772911557e-07, "loss": 0.0165, "step": 12295 }, { "epoch": 4.85, "learning_rate": 2.1962883805747397e-07, "loss": 0.0124, "step": 12296 }, { "epoch": 4.85, "learning_rate": 2.1960342806591216e-07, "loss": 0.0164, "step": 12297 }, { "epoch": 4.85, "learning_rate": 2.1957801775484418e-07, "loss": 0.0134, "step": 12298 }, { "epoch": 4.85, "learning_rate": 2.1955260712468432e-07, "loss": 0.0127, "step": 12299 }, { "epoch": 4.85, "learning_rate": 2.195271961758467e-07, "loss": 0.0181, "step": 12300 }, { "epoch": 4.85, "learning_rate": 2.1950178490874535e-07, "loss": 0.0167, "step": 12301 }, { "epoch": 4.85, "learning_rate": 2.1947637332379468e-07, "loss": 0.0143, "step": 12302 }, { "epoch": 4.85, "learning_rate": 2.1945096142140862e-07, "loss": 0.0166, "step": 12303 }, { "epoch": 4.85, "learning_rate": 2.194255492020015e-07, "loss": 0.0151, "step": 12304 }, { "epoch": 4.85, "learning_rate": 2.1940013666598745e-07, "loss": 0.0158, "step": 12305 }, { "epoch": 4.85, "learning_rate": 2.1937472381378065e-07, "loss": 0.012, "step": 12306 }, { "epoch": 4.85, "learning_rate": 2.193493106457953e-07, "loss": 0.0123, "step": 12307 }, { "epoch": 4.85, "learning_rate": 2.193238971624456e-07, "loss": 0.0149, "step": 12308 }, { "epoch": 4.85, "learning_rate": 2.1929848336414575e-07, "loss": 0.0151, "step": 12309 }, { "epoch": 4.85, "learning_rate": 2.1927306925130993e-07, "loss": 0.0134, "step": 12310 }, { "epoch": 4.85, "learning_rate": 2.192476548243524e-07, "loss": 0.019, "step": 12311 }, { "epoch": 4.85, "learning_rate": 2.192222400836873e-07, "loss": 0.02, "step": 12312 }, { "epoch": 4.85, "learning_rate": 2.1919682502972895e-07, "loss": 0.0142, "step": 12313 }, { "epoch": 4.85, "learning_rate": 2.191714096628915e-07, "loss": 0.0159, "step": 12314 }, { "epoch": 4.85, "learning_rate": 2.1914599398358927e-07, "loss": 0.0142, "step": 12315 }, { "epoch": 4.85, "learning_rate": 2.191205779922363e-07, "loss": 0.0147, "step": 12316 }, { "epoch": 4.85, "learning_rate": 2.1909516168924714e-07, "loss": 0.0129, "step": 12317 }, { "epoch": 4.85, "learning_rate": 2.190697450750358e-07, "loss": 0.0155, "step": 12318 }, { "epoch": 4.86, "learning_rate": 2.190443281500166e-07, "loss": 0.0156, "step": 12319 }, { "epoch": 4.86, "learning_rate": 2.190189109146037e-07, "loss": 0.0179, "step": 12320 }, { "epoch": 4.86, "learning_rate": 2.1899349336921155e-07, "loss": 0.0141, "step": 12321 }, { "epoch": 4.86, "learning_rate": 2.1896807551425436e-07, "loss": 0.0156, "step": 12322 }, { "epoch": 4.86, "learning_rate": 2.1894265735014627e-07, "loss": 0.0137, "step": 12323 }, { "epoch": 4.86, "learning_rate": 2.1891723887730174e-07, "loss": 0.013, "step": 12324 }, { "epoch": 4.86, "learning_rate": 2.1889182009613489e-07, "loss": 0.013, "step": 12325 }, { "epoch": 4.86, "learning_rate": 2.1886640100706017e-07, "loss": 0.0157, "step": 12326 }, { "epoch": 4.86, "learning_rate": 2.1884098161049175e-07, "loss": 0.0126, "step": 12327 }, { "epoch": 4.86, "learning_rate": 2.1881556190684392e-07, "loss": 0.0134, "step": 12328 }, { "epoch": 4.86, "learning_rate": 2.187901418965311e-07, "loss": 0.0128, "step": 12329 }, { "epoch": 4.86, "learning_rate": 2.1876472157996746e-07, "loss": 0.017, "step": 12330 }, { "epoch": 4.86, "learning_rate": 2.187393009575674e-07, "loss": 0.0142, "step": 12331 }, { "epoch": 4.86, "learning_rate": 2.1871388002974523e-07, "loss": 0.013, "step": 12332 }, { "epoch": 4.86, "learning_rate": 2.1868845879691525e-07, "loss": 0.0176, "step": 12333 }, { "epoch": 4.86, "learning_rate": 2.1866303725949175e-07, "loss": 0.0138, "step": 12334 }, { "epoch": 4.86, "learning_rate": 2.186376154178892e-07, "loss": 0.0172, "step": 12335 }, { "epoch": 4.86, "learning_rate": 2.1861219327252174e-07, "loss": 0.0156, "step": 12336 }, { "epoch": 4.86, "learning_rate": 2.1858677082380387e-07, "loss": 0.016, "step": 12337 }, { "epoch": 4.86, "learning_rate": 2.1856134807214993e-07, "loss": 0.0159, "step": 12338 }, { "epoch": 4.86, "learning_rate": 2.1853592501797413e-07, "loss": 0.0119, "step": 12339 }, { "epoch": 4.86, "learning_rate": 2.18510501661691e-07, "loss": 0.0124, "step": 12340 }, { "epoch": 4.86, "learning_rate": 2.1848507800371485e-07, "loss": 0.0153, "step": 12341 }, { "epoch": 4.86, "learning_rate": 2.1845965404445997e-07, "loss": 0.0176, "step": 12342 }, { "epoch": 4.86, "learning_rate": 2.1843422978434075e-07, "loss": 0.0131, "step": 12343 }, { "epoch": 4.86, "learning_rate": 2.1840880522377167e-07, "loss": 0.0131, "step": 12344 }, { "epoch": 4.87, "learning_rate": 2.18383380363167e-07, "loss": 0.0129, "step": 12345 }, { "epoch": 4.87, "learning_rate": 2.1835795520294118e-07, "loss": 0.0139, "step": 12346 }, { "epoch": 4.87, "learning_rate": 2.1833252974350862e-07, "loss": 0.0149, "step": 12347 }, { "epoch": 4.87, "learning_rate": 2.1830710398528365e-07, "loss": 0.0133, "step": 12348 }, { "epoch": 4.87, "learning_rate": 2.1828167792868072e-07, "loss": 0.0177, "step": 12349 }, { "epoch": 4.87, "learning_rate": 2.182562515741142e-07, "loss": 0.0136, "step": 12350 }, { "epoch": 4.87, "learning_rate": 2.1823082492199858e-07, "loss": 0.0124, "step": 12351 }, { "epoch": 4.87, "learning_rate": 2.1820539797274815e-07, "loss": 0.0119, "step": 12352 }, { "epoch": 4.87, "learning_rate": 2.1817997072677747e-07, "loss": 0.0136, "step": 12353 }, { "epoch": 4.87, "learning_rate": 2.1815454318450084e-07, "loss": 0.0122, "step": 12354 }, { "epoch": 4.87, "learning_rate": 2.181291153463328e-07, "loss": 0.0131, "step": 12355 }, { "epoch": 4.87, "learning_rate": 2.181036872126877e-07, "loss": 0.0177, "step": 12356 }, { "epoch": 4.87, "learning_rate": 2.1807825878398e-07, "loss": 0.017, "step": 12357 }, { "epoch": 4.87, "learning_rate": 2.1805283006062423e-07, "loss": 0.0173, "step": 12358 }, { "epoch": 4.87, "learning_rate": 2.180274010430347e-07, "loss": 0.0152, "step": 12359 }, { "epoch": 4.87, "learning_rate": 2.1800197173162592e-07, "loss": 0.014, "step": 12360 }, { "epoch": 4.87, "learning_rate": 2.179765421268124e-07, "loss": 0.017, "step": 12361 }, { "epoch": 4.87, "learning_rate": 2.1795111222900857e-07, "loss": 0.0167, "step": 12362 }, { "epoch": 4.87, "learning_rate": 2.179256820386289e-07, "loss": 0.0164, "step": 12363 }, { "epoch": 4.87, "learning_rate": 2.1790025155608785e-07, "loss": 0.0162, "step": 12364 }, { "epoch": 4.87, "learning_rate": 2.178748207817999e-07, "loss": 0.0152, "step": 12365 }, { "epoch": 4.87, "learning_rate": 2.178493897161795e-07, "loss": 0.0143, "step": 12366 }, { "epoch": 4.87, "learning_rate": 2.1782395835964125e-07, "loss": 0.0157, "step": 12367 }, { "epoch": 4.87, "learning_rate": 2.1779852671259945e-07, "loss": 0.0181, "step": 12368 }, { "epoch": 4.87, "learning_rate": 2.1777309477546883e-07, "loss": 0.0188, "step": 12369 }, { "epoch": 4.88, "learning_rate": 2.177476625486637e-07, "loss": 0.0145, "step": 12370 }, { "epoch": 4.88, "learning_rate": 2.177222300325987e-07, "loss": 0.0152, "step": 12371 }, { "epoch": 4.88, "learning_rate": 2.1769679722768823e-07, "loss": 0.0166, "step": 12372 }, { "epoch": 4.88, "learning_rate": 2.1767136413434693e-07, "loss": 0.0116, "step": 12373 }, { "epoch": 4.88, "learning_rate": 2.176459307529892e-07, "loss": 0.0131, "step": 12374 }, { "epoch": 4.88, "learning_rate": 2.1762049708402963e-07, "loss": 0.0144, "step": 12375 }, { "epoch": 4.88, "learning_rate": 2.175950631278827e-07, "loss": 0.0143, "step": 12376 }, { "epoch": 4.88, "learning_rate": 2.1756962888496307e-07, "loss": 0.016, "step": 12377 }, { "epoch": 4.88, "learning_rate": 2.175441943556852e-07, "loss": 0.015, "step": 12378 }, { "epoch": 4.88, "learning_rate": 2.175187595404635e-07, "loss": 0.0178, "step": 12379 }, { "epoch": 4.88, "learning_rate": 2.1749332443971276e-07, "loss": 0.0163, "step": 12380 }, { "epoch": 4.88, "learning_rate": 2.174678890538473e-07, "loss": 0.0159, "step": 12381 }, { "epoch": 4.88, "learning_rate": 2.1744245338328193e-07, "loss": 0.015, "step": 12382 }, { "epoch": 4.88, "learning_rate": 2.1741701742843101e-07, "loss": 0.0158, "step": 12383 }, { "epoch": 4.88, "learning_rate": 2.1739158118970918e-07, "loss": 0.0182, "step": 12384 }, { "epoch": 4.88, "learning_rate": 2.1736614466753106e-07, "loss": 0.0139, "step": 12385 }, { "epoch": 4.88, "learning_rate": 2.173407078623111e-07, "loss": 0.016, "step": 12386 }, { "epoch": 4.88, "learning_rate": 2.1731527077446405e-07, "loss": 0.0143, "step": 12387 }, { "epoch": 4.88, "learning_rate": 2.172898334044043e-07, "loss": 0.0181, "step": 12388 }, { "epoch": 4.88, "learning_rate": 2.172643957525466e-07, "loss": 0.0148, "step": 12389 }, { "epoch": 4.88, "learning_rate": 2.1723895781930546e-07, "loss": 0.0143, "step": 12390 }, { "epoch": 4.88, "learning_rate": 2.1721351960509554e-07, "loss": 0.0143, "step": 12391 }, { "epoch": 4.88, "learning_rate": 2.1718808111033136e-07, "loss": 0.0155, "step": 12392 }, { "epoch": 4.88, "learning_rate": 2.1716264233542764e-07, "loss": 0.0153, "step": 12393 }, { "epoch": 4.88, "learning_rate": 2.1713720328079885e-07, "loss": 0.017, "step": 12394 }, { "epoch": 4.88, "learning_rate": 2.1711176394685978e-07, "loss": 0.014, "step": 12395 }, { "epoch": 4.89, "learning_rate": 2.170863243340249e-07, "loss": 0.0164, "step": 12396 }, { "epoch": 4.89, "learning_rate": 2.1706088444270896e-07, "loss": 0.0134, "step": 12397 }, { "epoch": 4.89, "learning_rate": 2.1703544427332653e-07, "loss": 0.0127, "step": 12398 }, { "epoch": 4.89, "learning_rate": 2.1701000382629223e-07, "loss": 0.0167, "step": 12399 }, { "epoch": 4.89, "learning_rate": 2.169845631020207e-07, "loss": 0.0113, "step": 12400 }, { "epoch": 4.89, "learning_rate": 2.169591221009266e-07, "loss": 0.0122, "step": 12401 }, { "epoch": 4.89, "learning_rate": 2.1693368082342464e-07, "loss": 0.0143, "step": 12402 }, { "epoch": 4.89, "learning_rate": 2.1690823926992933e-07, "loss": 0.0136, "step": 12403 }, { "epoch": 4.89, "learning_rate": 2.168827974408555e-07, "loss": 0.0122, "step": 12404 }, { "epoch": 4.89, "learning_rate": 2.1685735533661776e-07, "loss": 0.015, "step": 12405 }, { "epoch": 4.89, "learning_rate": 2.168319129576307e-07, "loss": 0.018, "step": 12406 }, { "epoch": 4.89, "learning_rate": 2.16806470304309e-07, "loss": 0.0163, "step": 12407 }, { "epoch": 4.89, "learning_rate": 2.1678102737706742e-07, "loss": 0.0145, "step": 12408 }, { "epoch": 4.89, "learning_rate": 2.167555841763206e-07, "loss": 0.0131, "step": 12409 }, { "epoch": 4.89, "learning_rate": 2.1673014070248316e-07, "loss": 0.0118, "step": 12410 }, { "epoch": 4.89, "learning_rate": 2.1670469695596995e-07, "loss": 0.0184, "step": 12411 }, { "epoch": 4.89, "learning_rate": 2.1667925293719552e-07, "loss": 0.0151, "step": 12412 }, { "epoch": 4.89, "learning_rate": 2.166538086465746e-07, "loss": 0.0143, "step": 12413 }, { "epoch": 4.89, "learning_rate": 2.166283640845219e-07, "loss": 0.0135, "step": 12414 }, { "epoch": 4.89, "learning_rate": 2.166029192514522e-07, "loss": 0.0151, "step": 12415 }, { "epoch": 4.89, "learning_rate": 2.1657747414778006e-07, "loss": 0.0168, "step": 12416 }, { "epoch": 4.89, "learning_rate": 2.1655202877392039e-07, "loss": 0.0133, "step": 12417 }, { "epoch": 4.89, "learning_rate": 2.165265831302878e-07, "loss": 0.0153, "step": 12418 }, { "epoch": 4.89, "learning_rate": 2.1650113721729694e-07, "loss": 0.0159, "step": 12419 }, { "epoch": 4.89, "learning_rate": 2.1647569103536268e-07, "loss": 0.0141, "step": 12420 }, { "epoch": 4.9, "learning_rate": 2.1645024458489964e-07, "loss": 0.0108, "step": 12421 }, { "epoch": 4.9, "learning_rate": 2.1642479786632268e-07, "loss": 0.0154, "step": 12422 }, { "epoch": 4.9, "learning_rate": 2.1639935088004643e-07, "loss": 0.0157, "step": 12423 }, { "epoch": 4.9, "learning_rate": 2.1637390362648573e-07, "loss": 0.0137, "step": 12424 }, { "epoch": 4.9, "learning_rate": 2.1634845610605526e-07, "loss": 0.0154, "step": 12425 }, { "epoch": 4.9, "learning_rate": 2.163230083191698e-07, "loss": 0.0177, "step": 12426 }, { "epoch": 4.9, "learning_rate": 2.1629756026624415e-07, "loss": 0.0171, "step": 12427 }, { "epoch": 4.9, "learning_rate": 2.16272111947693e-07, "loss": 0.014, "step": 12428 }, { "epoch": 4.9, "learning_rate": 2.1624666336393119e-07, "loss": 0.0148, "step": 12429 }, { "epoch": 4.9, "learning_rate": 2.162212145153734e-07, "loss": 0.0132, "step": 12430 }, { "epoch": 4.9, "learning_rate": 2.1619576540243455e-07, "loss": 0.011, "step": 12431 }, { "epoch": 4.9, "learning_rate": 2.1617031602552922e-07, "loss": 0.016, "step": 12432 }, { "epoch": 4.9, "learning_rate": 2.161448663850724e-07, "loss": 0.0118, "step": 12433 }, { "epoch": 4.9, "learning_rate": 2.1611941648147876e-07, "loss": 0.016, "step": 12434 }, { "epoch": 4.9, "learning_rate": 2.1609396631516316e-07, "loss": 0.0134, "step": 12435 }, { "epoch": 4.9, "learning_rate": 2.1606851588654033e-07, "loss": 0.0151, "step": 12436 }, { "epoch": 4.9, "learning_rate": 2.1604306519602518e-07, "loss": 0.0147, "step": 12437 }, { "epoch": 4.9, "learning_rate": 2.1601761424403241e-07, "loss": 0.0156, "step": 12438 }, { "epoch": 4.9, "learning_rate": 2.1599216303097684e-07, "loss": 0.0179, "step": 12439 }, { "epoch": 4.9, "learning_rate": 2.1596671155727338e-07, "loss": 0.0132, "step": 12440 }, { "epoch": 4.9, "learning_rate": 2.1594125982333673e-07, "loss": 0.0127, "step": 12441 }, { "epoch": 4.9, "learning_rate": 2.159158078295818e-07, "loss": 0.0195, "step": 12442 }, { "epoch": 4.9, "learning_rate": 2.1589035557642336e-07, "loss": 0.0157, "step": 12443 }, { "epoch": 4.9, "learning_rate": 2.1586490306427633e-07, "loss": 0.0144, "step": 12444 }, { "epoch": 4.9, "learning_rate": 2.1583945029355544e-07, "loss": 0.0169, "step": 12445 }, { "epoch": 4.91, "learning_rate": 2.158139972646756e-07, "loss": 0.0147, "step": 12446 }, { "epoch": 4.91, "learning_rate": 2.157885439780517e-07, "loss": 0.016, "step": 12447 }, { "epoch": 4.91, "learning_rate": 2.1576309043409844e-07, "loss": 0.0134, "step": 12448 }, { "epoch": 4.91, "learning_rate": 2.1573763663323085e-07, "loss": 0.0147, "step": 12449 }, { "epoch": 4.91, "learning_rate": 2.1571218257586365e-07, "loss": 0.0165, "step": 12450 }, { "epoch": 4.91, "learning_rate": 2.156867282624118e-07, "loss": 0.0193, "step": 12451 }, { "epoch": 4.91, "learning_rate": 2.1566127369329003e-07, "loss": 0.016, "step": 12452 }, { "epoch": 4.91, "learning_rate": 2.1563581886891336e-07, "loss": 0.0151, "step": 12453 }, { "epoch": 4.91, "learning_rate": 2.1561036378969658e-07, "loss": 0.0162, "step": 12454 }, { "epoch": 4.91, "learning_rate": 2.1558490845605463e-07, "loss": 0.0146, "step": 12455 }, { "epoch": 4.91, "learning_rate": 2.155594528684023e-07, "loss": 0.0131, "step": 12456 }, { "epoch": 4.91, "learning_rate": 2.1553399702715457e-07, "loss": 0.0217, "step": 12457 }, { "epoch": 4.91, "learning_rate": 2.1550854093272637e-07, "loss": 0.0124, "step": 12458 }, { "epoch": 4.91, "learning_rate": 2.154830845855324e-07, "loss": 0.0148, "step": 12459 }, { "epoch": 4.91, "learning_rate": 2.1545762798598774e-07, "loss": 0.0161, "step": 12460 }, { "epoch": 4.91, "learning_rate": 2.1543217113450725e-07, "loss": 0.0162, "step": 12461 }, { "epoch": 4.91, "learning_rate": 2.154067140315058e-07, "loss": 0.0132, "step": 12462 }, { "epoch": 4.91, "learning_rate": 2.153812566773983e-07, "loss": 0.016, "step": 12463 }, { "epoch": 4.91, "learning_rate": 2.1535579907259977e-07, "loss": 0.0172, "step": 12464 }, { "epoch": 4.91, "learning_rate": 2.1533034121752495e-07, "loss": 0.0167, "step": 12465 }, { "epoch": 4.91, "learning_rate": 2.1530488311258898e-07, "loss": 0.0163, "step": 12466 }, { "epoch": 4.91, "learning_rate": 2.1527942475820665e-07, "loss": 0.0165, "step": 12467 }, { "epoch": 4.91, "learning_rate": 2.1525396615479287e-07, "loss": 0.0124, "step": 12468 }, { "epoch": 4.91, "learning_rate": 2.1522850730276265e-07, "loss": 0.0155, "step": 12469 }, { "epoch": 4.91, "learning_rate": 2.152030482025309e-07, "loss": 0.0168, "step": 12470 }, { "epoch": 4.91, "learning_rate": 2.1517758885451255e-07, "loss": 0.0149, "step": 12471 }, { "epoch": 4.92, "learning_rate": 2.1515212925912266e-07, "loss": 0.0135, "step": 12472 }, { "epoch": 4.92, "learning_rate": 2.1512666941677603e-07, "loss": 0.0143, "step": 12473 }, { "epoch": 4.92, "learning_rate": 2.1510120932788762e-07, "loss": 0.0158, "step": 12474 }, { "epoch": 4.92, "learning_rate": 2.1507574899287256e-07, "loss": 0.0138, "step": 12475 }, { "epoch": 4.92, "learning_rate": 2.1505028841214558e-07, "loss": 0.0151, "step": 12476 }, { "epoch": 4.92, "learning_rate": 2.1502482758612185e-07, "loss": 0.0167, "step": 12477 }, { "epoch": 4.92, "learning_rate": 2.149993665152163e-07, "loss": 0.0182, "step": 12478 }, { "epoch": 4.92, "learning_rate": 2.1497390519984382e-07, "loss": 0.0143, "step": 12479 }, { "epoch": 4.92, "learning_rate": 2.1494844364041948e-07, "loss": 0.013, "step": 12480 }, { "epoch": 4.92, "learning_rate": 2.1492298183735812e-07, "loss": 0.0141, "step": 12481 }, { "epoch": 4.92, "learning_rate": 2.1489751979107496e-07, "loss": 0.0127, "step": 12482 }, { "epoch": 4.92, "learning_rate": 2.148720575019848e-07, "loss": 0.0165, "step": 12483 }, { "epoch": 4.92, "learning_rate": 2.1484659497050272e-07, "loss": 0.0138, "step": 12484 }, { "epoch": 4.92, "learning_rate": 2.1482113219704373e-07, "loss": 0.0131, "step": 12485 }, { "epoch": 4.92, "learning_rate": 2.1479566918202278e-07, "loss": 0.0159, "step": 12486 }, { "epoch": 4.92, "learning_rate": 2.14770205925855e-07, "loss": 0.0148, "step": 12487 }, { "epoch": 4.92, "learning_rate": 2.1474474242895514e-07, "loss": 0.0165, "step": 12488 }, { "epoch": 4.92, "learning_rate": 2.1471927869173854e-07, "loss": 0.0142, "step": 12489 }, { "epoch": 4.92, "learning_rate": 2.1469381471461996e-07, "loss": 0.0153, "step": 12490 }, { "epoch": 4.92, "learning_rate": 2.1466835049801464e-07, "loss": 0.0155, "step": 12491 }, { "epoch": 4.92, "learning_rate": 2.1464288604233744e-07, "loss": 0.0123, "step": 12492 }, { "epoch": 4.92, "learning_rate": 2.1461742134800347e-07, "loss": 0.017, "step": 12493 }, { "epoch": 4.92, "learning_rate": 2.145919564154277e-07, "loss": 0.0132, "step": 12494 }, { "epoch": 4.92, "learning_rate": 2.145664912450253e-07, "loss": 0.013, "step": 12495 }, { "epoch": 4.92, "learning_rate": 2.1454102583721118e-07, "loss": 0.0139, "step": 12496 }, { "epoch": 4.93, "learning_rate": 2.1451556019240042e-07, "loss": 0.0141, "step": 12497 }, { "epoch": 4.93, "learning_rate": 2.1449009431100814e-07, "loss": 0.0181, "step": 12498 }, { "epoch": 4.93, "learning_rate": 2.1446462819344928e-07, "loss": 0.0144, "step": 12499 }, { "epoch": 4.93, "learning_rate": 2.14439161840139e-07, "loss": 0.0136, "step": 12500 }, { "epoch": 4.93, "learning_rate": 2.1441369525149232e-07, "loss": 0.0125, "step": 12501 }, { "epoch": 4.93, "learning_rate": 2.1438822842792436e-07, "loss": 0.0206, "step": 12502 }, { "epoch": 4.93, "learning_rate": 2.1436276136985009e-07, "loss": 0.0138, "step": 12503 }, { "epoch": 4.93, "learning_rate": 2.1433729407768465e-07, "loss": 0.016, "step": 12504 }, { "epoch": 4.93, "learning_rate": 2.143118265518431e-07, "loss": 0.0133, "step": 12505 }, { "epoch": 4.93, "learning_rate": 2.1428635879274056e-07, "loss": 0.0135, "step": 12506 }, { "epoch": 4.93, "learning_rate": 2.1426089080079208e-07, "loss": 0.0126, "step": 12507 }, { "epoch": 4.93, "learning_rate": 2.142354225764127e-07, "loss": 0.0136, "step": 12508 }, { "epoch": 4.93, "learning_rate": 2.142099541200176e-07, "loss": 0.0135, "step": 12509 }, { "epoch": 4.93, "learning_rate": 2.1418448543202187e-07, "loss": 0.0161, "step": 12510 }, { "epoch": 4.93, "learning_rate": 2.1415901651284058e-07, "loss": 0.0148, "step": 12511 }, { "epoch": 4.93, "learning_rate": 2.1413354736288877e-07, "loss": 0.0121, "step": 12512 }, { "epoch": 4.93, "learning_rate": 2.141080779825817e-07, "loss": 0.015, "step": 12513 }, { "epoch": 4.93, "learning_rate": 2.1408260837233443e-07, "loss": 0.0155, "step": 12514 }, { "epoch": 4.93, "learning_rate": 2.1405713853256198e-07, "loss": 0.015, "step": 12515 }, { "epoch": 4.93, "learning_rate": 2.1403166846367955e-07, "loss": 0.0129, "step": 12516 }, { "epoch": 4.93, "learning_rate": 2.1400619816610223e-07, "loss": 0.0142, "step": 12517 }, { "epoch": 4.93, "learning_rate": 2.139807276402452e-07, "loss": 0.0172, "step": 12518 }, { "epoch": 4.93, "learning_rate": 2.1395525688652354e-07, "loss": 0.0131, "step": 12519 }, { "epoch": 4.93, "learning_rate": 2.1392978590535243e-07, "loss": 0.0178, "step": 12520 }, { "epoch": 4.93, "learning_rate": 2.139043146971469e-07, "loss": 0.012, "step": 12521 }, { "epoch": 4.94, "learning_rate": 2.1387884326232228e-07, "loss": 0.015, "step": 12522 }, { "epoch": 4.94, "learning_rate": 2.1385337160129352e-07, "loss": 0.0175, "step": 12523 }, { "epoch": 4.94, "learning_rate": 2.1382789971447592e-07, "loss": 0.0153, "step": 12524 }, { "epoch": 4.94, "learning_rate": 2.138024276022845e-07, "loss": 0.018, "step": 12525 }, { "epoch": 4.94, "learning_rate": 2.1377695526513454e-07, "loss": 0.0141, "step": 12526 }, { "epoch": 4.94, "learning_rate": 2.137514827034412e-07, "loss": 0.0153, "step": 12527 }, { "epoch": 4.94, "learning_rate": 2.1372600991761946e-07, "loss": 0.0151, "step": 12528 }, { "epoch": 4.94, "learning_rate": 2.1370053690808473e-07, "loss": 0.0178, "step": 12529 }, { "epoch": 4.94, "learning_rate": 2.1367506367525202e-07, "loss": 0.0128, "step": 12530 }, { "epoch": 4.94, "learning_rate": 2.1364959021953655e-07, "loss": 0.0173, "step": 12531 }, { "epoch": 4.94, "learning_rate": 2.1362411654135353e-07, "loss": 0.0139, "step": 12532 }, { "epoch": 4.94, "learning_rate": 2.1359864264111813e-07, "loss": 0.0142, "step": 12533 }, { "epoch": 4.94, "learning_rate": 2.135731685192455e-07, "loss": 0.014, "step": 12534 }, { "epoch": 4.94, "learning_rate": 2.1354769417615083e-07, "loss": 0.0133, "step": 12535 }, { "epoch": 4.94, "learning_rate": 2.135222196122494e-07, "loss": 0.0126, "step": 12536 }, { "epoch": 4.94, "learning_rate": 2.1349674482795628e-07, "loss": 0.0124, "step": 12537 }, { "epoch": 4.94, "learning_rate": 2.1347126982368678e-07, "loss": 0.0123, "step": 12538 }, { "epoch": 4.94, "learning_rate": 2.13445794599856e-07, "loss": 0.0161, "step": 12539 }, { "epoch": 4.94, "learning_rate": 2.1342031915687927e-07, "loss": 0.0118, "step": 12540 }, { "epoch": 4.94, "learning_rate": 2.1339484349517165e-07, "loss": 0.015, "step": 12541 }, { "epoch": 4.94, "learning_rate": 2.133693676151485e-07, "loss": 0.0148, "step": 12542 }, { "epoch": 4.94, "learning_rate": 2.1334389151722493e-07, "loss": 0.015, "step": 12543 }, { "epoch": 4.94, "learning_rate": 2.133184152018163e-07, "loss": 0.0158, "step": 12544 }, { "epoch": 4.94, "learning_rate": 2.1329293866933763e-07, "loss": 0.016, "step": 12545 }, { "epoch": 4.94, "learning_rate": 2.1326746192020432e-07, "loss": 0.0165, "step": 12546 }, { "epoch": 4.94, "learning_rate": 2.1324198495483157e-07, "loss": 0.0181, "step": 12547 }, { "epoch": 4.95, "learning_rate": 2.1321650777363453e-07, "loss": 0.0126, "step": 12548 }, { "epoch": 4.95, "learning_rate": 2.1319103037702858e-07, "loss": 0.0142, "step": 12549 }, { "epoch": 4.95, "learning_rate": 2.131655527654288e-07, "loss": 0.0152, "step": 12550 }, { "epoch": 4.95, "learning_rate": 2.1314007493925055e-07, "loss": 0.0168, "step": 12551 }, { "epoch": 4.95, "learning_rate": 2.1311459689890903e-07, "loss": 0.0129, "step": 12552 }, { "epoch": 4.95, "learning_rate": 2.1308911864481954e-07, "loss": 0.0173, "step": 12553 }, { "epoch": 4.95, "learning_rate": 2.130636401773973e-07, "loss": 0.0154, "step": 12554 }, { "epoch": 4.95, "learning_rate": 2.130381614970576e-07, "loss": 0.0168, "step": 12555 }, { "epoch": 4.95, "learning_rate": 2.130126826042157e-07, "loss": 0.015, "step": 12556 }, { "epoch": 4.95, "learning_rate": 2.1298720349928682e-07, "loss": 0.0143, "step": 12557 }, { "epoch": 4.95, "learning_rate": 2.1296172418268624e-07, "loss": 0.0145, "step": 12558 }, { "epoch": 4.95, "learning_rate": 2.129362446548293e-07, "loss": 0.0121, "step": 12559 }, { "epoch": 4.95, "learning_rate": 2.129107649161312e-07, "loss": 0.0151, "step": 12560 }, { "epoch": 4.95, "learning_rate": 2.1288528496700726e-07, "loss": 0.0134, "step": 12561 }, { "epoch": 4.95, "learning_rate": 2.1285980480787278e-07, "loss": 0.0167, "step": 12562 }, { "epoch": 4.95, "learning_rate": 2.1283432443914298e-07, "loss": 0.0149, "step": 12563 }, { "epoch": 4.95, "learning_rate": 2.128088438612332e-07, "loss": 0.014, "step": 12564 }, { "epoch": 4.95, "learning_rate": 2.1278336307455875e-07, "loss": 0.0142, "step": 12565 }, { "epoch": 4.95, "learning_rate": 2.1275788207953496e-07, "loss": 0.0133, "step": 12566 }, { "epoch": 4.95, "learning_rate": 2.1273240087657706e-07, "loss": 0.0146, "step": 12567 }, { "epoch": 4.95, "learning_rate": 2.1270691946610034e-07, "loss": 0.0155, "step": 12568 }, { "epoch": 4.95, "learning_rate": 2.1268143784852017e-07, "loss": 0.0157, "step": 12569 }, { "epoch": 4.95, "learning_rate": 2.126559560242518e-07, "loss": 0.0146, "step": 12570 }, { "epoch": 4.95, "learning_rate": 2.1263047399371059e-07, "loss": 0.0149, "step": 12571 }, { "epoch": 4.95, "learning_rate": 2.126049917573119e-07, "loss": 0.0159, "step": 12572 }, { "epoch": 4.96, "learning_rate": 2.1257950931547092e-07, "loss": 0.0151, "step": 12573 }, { "epoch": 4.96, "learning_rate": 2.125540266686031e-07, "loss": 0.0148, "step": 12574 }, { "epoch": 4.96, "learning_rate": 2.1252854381712378e-07, "loss": 0.0161, "step": 12575 }, { "epoch": 4.96, "learning_rate": 2.1250306076144814e-07, "loss": 0.0174, "step": 12576 }, { "epoch": 4.96, "learning_rate": 2.1247757750199167e-07, "loss": 0.0138, "step": 12577 }, { "epoch": 4.96, "learning_rate": 2.124520940391697e-07, "loss": 0.0156, "step": 12578 }, { "epoch": 4.96, "learning_rate": 2.1242661037339747e-07, "loss": 0.0145, "step": 12579 }, { "epoch": 4.96, "learning_rate": 2.1240112650509033e-07, "loss": 0.0151, "step": 12580 }, { "epoch": 4.96, "learning_rate": 2.1237564243466374e-07, "loss": 0.0155, "step": 12581 }, { "epoch": 4.96, "learning_rate": 2.1235015816253295e-07, "loss": 0.0137, "step": 12582 }, { "epoch": 4.96, "learning_rate": 2.1232467368911336e-07, "loss": 0.0141, "step": 12583 }, { "epoch": 4.96, "learning_rate": 2.1229918901482034e-07, "loss": 0.0188, "step": 12584 }, { "epoch": 4.96, "learning_rate": 2.122737041400692e-07, "loss": 0.0137, "step": 12585 }, { "epoch": 4.96, "learning_rate": 2.1224821906527536e-07, "loss": 0.0157, "step": 12586 }, { "epoch": 4.96, "learning_rate": 2.122227337908542e-07, "loss": 0.0165, "step": 12587 }, { "epoch": 4.96, "learning_rate": 2.1219724831722096e-07, "loss": 0.012, "step": 12588 }, { "epoch": 4.96, "learning_rate": 2.121717626447912e-07, "loss": 0.0146, "step": 12589 }, { "epoch": 4.96, "learning_rate": 2.1214627677398016e-07, "loss": 0.0157, "step": 12590 }, { "epoch": 4.96, "learning_rate": 2.121207907052033e-07, "loss": 0.0145, "step": 12591 }, { "epoch": 4.96, "learning_rate": 2.1209530443887595e-07, "loss": 0.0114, "step": 12592 }, { "epoch": 4.96, "learning_rate": 2.120698179754135e-07, "loss": 0.016, "step": 12593 }, { "epoch": 4.96, "learning_rate": 2.1204433131523136e-07, "loss": 0.0119, "step": 12594 }, { "epoch": 4.96, "learning_rate": 2.1201884445874495e-07, "loss": 0.0113, "step": 12595 }, { "epoch": 4.96, "learning_rate": 2.1199335740636965e-07, "loss": 0.0124, "step": 12596 }, { "epoch": 4.96, "learning_rate": 2.1196787015852085e-07, "loss": 0.0174, "step": 12597 }, { "epoch": 4.96, "learning_rate": 2.1194238271561397e-07, "loss": 0.0139, "step": 12598 }, { "epoch": 4.97, "learning_rate": 2.1191689507806435e-07, "loss": 0.0136, "step": 12599 }, { "epoch": 4.97, "learning_rate": 2.1189140724628748e-07, "loss": 0.0164, "step": 12600 }, { "epoch": 4.97, "learning_rate": 2.118659192206988e-07, "loss": 0.0187, "step": 12601 }, { "epoch": 4.97, "learning_rate": 2.1184043100171363e-07, "loss": 0.014, "step": 12602 }, { "epoch": 4.97, "learning_rate": 2.118149425897474e-07, "loss": 0.0136, "step": 12603 }, { "epoch": 4.97, "learning_rate": 2.117894539852156e-07, "loss": 0.0155, "step": 12604 }, { "epoch": 4.97, "learning_rate": 2.117639651885336e-07, "loss": 0.0162, "step": 12605 }, { "epoch": 4.97, "learning_rate": 2.1173847620011682e-07, "loss": 0.0134, "step": 12606 }, { "epoch": 4.97, "learning_rate": 2.117129870203808e-07, "loss": 0.0142, "step": 12607 }, { "epoch": 4.97, "learning_rate": 2.116874976497408e-07, "loss": 0.0147, "step": 12608 }, { "epoch": 4.97, "learning_rate": 2.1166200808861241e-07, "loss": 0.0151, "step": 12609 }, { "epoch": 4.97, "learning_rate": 2.1163651833741097e-07, "loss": 0.0171, "step": 12610 }, { "epoch": 4.97, "learning_rate": 2.1161102839655197e-07, "loss": 0.0145, "step": 12611 }, { "epoch": 4.97, "learning_rate": 2.1158553826645088e-07, "loss": 0.0128, "step": 12612 }, { "epoch": 4.97, "learning_rate": 2.1156004794752312e-07, "loss": 0.0147, "step": 12613 }, { "epoch": 4.97, "learning_rate": 2.1153455744018413e-07, "loss": 0.0155, "step": 12614 }, { "epoch": 4.97, "learning_rate": 2.115090667448494e-07, "loss": 0.0187, "step": 12615 }, { "epoch": 4.97, "learning_rate": 2.1148357586193436e-07, "loss": 0.0122, "step": 12616 }, { "epoch": 4.97, "learning_rate": 2.1145808479185452e-07, "loss": 0.0143, "step": 12617 }, { "epoch": 4.97, "learning_rate": 2.1143259353502528e-07, "loss": 0.0137, "step": 12618 }, { "epoch": 4.97, "learning_rate": 2.1140710209186213e-07, "loss": 0.0146, "step": 12619 }, { "epoch": 4.97, "learning_rate": 2.1138161046278057e-07, "loss": 0.0131, "step": 12620 }, { "epoch": 4.97, "learning_rate": 2.1135611864819613e-07, "loss": 0.0134, "step": 12621 }, { "epoch": 4.97, "learning_rate": 2.113306266485241e-07, "loss": 0.0156, "step": 12622 }, { "epoch": 4.97, "learning_rate": 2.1130513446418012e-07, "loss": 0.0162, "step": 12623 }, { "epoch": 4.98, "learning_rate": 2.1127964209557967e-07, "loss": 0.0153, "step": 12624 }, { "epoch": 4.98, "learning_rate": 2.1125414954313815e-07, "loss": 0.0153, "step": 12625 }, { "epoch": 4.98, "learning_rate": 2.1122865680727112e-07, "loss": 0.0144, "step": 12626 }, { "epoch": 4.98, "learning_rate": 2.1120316388839404e-07, "loss": 0.0144, "step": 12627 }, { "epoch": 4.98, "learning_rate": 2.111776707869224e-07, "loss": 0.0142, "step": 12628 }, { "epoch": 4.98, "learning_rate": 2.111521775032717e-07, "loss": 0.0156, "step": 12629 }, { "epoch": 4.98, "learning_rate": 2.1112668403785748e-07, "loss": 0.0173, "step": 12630 }, { "epoch": 4.98, "learning_rate": 2.1110119039109526e-07, "loss": 0.015, "step": 12631 }, { "epoch": 4.98, "learning_rate": 2.1107569656340046e-07, "loss": 0.0158, "step": 12632 }, { "epoch": 4.98, "learning_rate": 2.1105020255518866e-07, "loss": 0.0127, "step": 12633 }, { "epoch": 4.98, "learning_rate": 2.110247083668753e-07, "loss": 0.0178, "step": 12634 }, { "epoch": 4.98, "learning_rate": 2.1099921399887602e-07, "loss": 0.0134, "step": 12635 }, { "epoch": 4.98, "learning_rate": 2.1097371945160624e-07, "loss": 0.0172, "step": 12636 }, { "epoch": 4.98, "learning_rate": 2.109482247254815e-07, "loss": 0.0153, "step": 12637 }, { "epoch": 4.98, "learning_rate": 2.1092272982091735e-07, "loss": 0.0158, "step": 12638 }, { "epoch": 4.98, "learning_rate": 2.1089723473832928e-07, "loss": 0.0147, "step": 12639 }, { "epoch": 4.98, "learning_rate": 2.1087173947813288e-07, "loss": 0.0153, "step": 12640 }, { "epoch": 4.98, "learning_rate": 2.1084624404074365e-07, "loss": 0.0141, "step": 12641 }, { "epoch": 4.98, "learning_rate": 2.108207484265771e-07, "loss": 0.0127, "step": 12642 }, { "epoch": 4.98, "learning_rate": 2.1079525263604884e-07, "loss": 0.0191, "step": 12643 }, { "epoch": 4.98, "learning_rate": 2.107697566695743e-07, "loss": 0.0124, "step": 12644 }, { "epoch": 4.98, "learning_rate": 2.1074426052756914e-07, "loss": 0.0193, "step": 12645 }, { "epoch": 4.98, "learning_rate": 2.1071876421044892e-07, "loss": 0.0137, "step": 12646 }, { "epoch": 4.98, "learning_rate": 2.106932677186291e-07, "loss": 0.0177, "step": 12647 }, { "epoch": 4.98, "learning_rate": 2.1066777105252518e-07, "loss": 0.0141, "step": 12648 }, { "epoch": 4.99, "learning_rate": 2.1064227421255286e-07, "loss": 0.0112, "step": 12649 }, { "epoch": 4.99, "learning_rate": 2.1061677719912768e-07, "loss": 0.0161, "step": 12650 }, { "epoch": 4.99, "learning_rate": 2.1059128001266515e-07, "loss": 0.0156, "step": 12651 }, { "epoch": 4.99, "learning_rate": 2.105657826535808e-07, "loss": 0.0144, "step": 12652 }, { "epoch": 4.99, "learning_rate": 2.105402851222903e-07, "loss": 0.0141, "step": 12653 }, { "epoch": 4.99, "learning_rate": 2.1051478741920915e-07, "loss": 0.0155, "step": 12654 }, { "epoch": 4.99, "learning_rate": 2.1048928954475298e-07, "loss": 0.0134, "step": 12655 }, { "epoch": 4.99, "learning_rate": 2.1046379149933729e-07, "loss": 0.013, "step": 12656 }, { "epoch": 4.99, "learning_rate": 2.104382932833777e-07, "loss": 0.0166, "step": 12657 }, { "epoch": 4.99, "learning_rate": 2.1041279489728983e-07, "loss": 0.0153, "step": 12658 }, { "epoch": 4.99, "learning_rate": 2.1038729634148923e-07, "loss": 0.0156, "step": 12659 }, { "epoch": 4.99, "learning_rate": 2.1036179761639148e-07, "loss": 0.0149, "step": 12660 }, { "epoch": 4.99, "learning_rate": 2.1033629872241212e-07, "loss": 0.0172, "step": 12661 }, { "epoch": 4.99, "learning_rate": 2.103107996599669e-07, "loss": 0.0149, "step": 12662 }, { "epoch": 4.99, "learning_rate": 2.102853004294712e-07, "loss": 0.0154, "step": 12663 }, { "epoch": 4.99, "learning_rate": 2.1025980103134083e-07, "loss": 0.0125, "step": 12664 }, { "epoch": 4.99, "learning_rate": 2.102343014659913e-07, "loss": 0.0175, "step": 12665 }, { "epoch": 4.99, "learning_rate": 2.1020880173383818e-07, "loss": 0.0143, "step": 12666 }, { "epoch": 4.99, "learning_rate": 2.1018330183529707e-07, "loss": 0.0122, "step": 12667 }, { "epoch": 4.99, "learning_rate": 2.101578017707837e-07, "loss": 0.0165, "step": 12668 }, { "epoch": 4.99, "learning_rate": 2.1013230154071354e-07, "loss": 0.0127, "step": 12669 }, { "epoch": 4.99, "learning_rate": 2.1010680114550224e-07, "loss": 0.0163, "step": 12670 }, { "epoch": 4.99, "learning_rate": 2.100813005855655e-07, "loss": 0.0158, "step": 12671 }, { "epoch": 4.99, "learning_rate": 2.1005579986131882e-07, "loss": 0.0138, "step": 12672 }, { "epoch": 4.99, "learning_rate": 2.1003029897317797e-07, "loss": 0.0148, "step": 12673 }, { "epoch": 4.99, "learning_rate": 2.100047979215584e-07, "loss": 0.016, "step": 12674 }, { "epoch": 5.0, "learning_rate": 2.099792967068759e-07, "loss": 0.0124, "step": 12675 }, { "epoch": 5.0, "learning_rate": 2.0995379532954595e-07, "loss": 0.0163, "step": 12676 }, { "epoch": 5.0, "learning_rate": 2.0992829378998435e-07, "loss": 0.0144, "step": 12677 }, { "epoch": 5.0, "learning_rate": 2.0990279208860664e-07, "loss": 0.014, "step": 12678 }, { "epoch": 5.0, "learning_rate": 2.0987729022582842e-07, "loss": 0.0173, "step": 12679 }, { "epoch": 5.0, "learning_rate": 2.0985178820206544e-07, "loss": 0.0141, "step": 12680 }, { "epoch": 5.0, "learning_rate": 2.0982628601773324e-07, "loss": 0.0134, "step": 12681 }, { "epoch": 5.0, "learning_rate": 2.098007836732475e-07, "loss": 0.0148, "step": 12682 }, { "epoch": 5.0, "learning_rate": 2.0977528116902391e-07, "loss": 0.014, "step": 12683 }, { "epoch": 5.0, "learning_rate": 2.0974977850547813e-07, "loss": 0.0165, "step": 12684 }, { "epoch": 5.0, "learning_rate": 2.0972427568302574e-07, "loss": 0.0174, "step": 12685 }, { "epoch": 5.0, "learning_rate": 2.0969877270208242e-07, "loss": 0.0136, "step": 12686 }, { "epoch": 5.0, "learning_rate": 2.0967326956306385e-07, "loss": 0.0149, "step": 12687 }, { "epoch": 5.0, "learning_rate": 2.0964776626638577e-07, "loss": 0.0129, "step": 12688 }, { "epoch": 5.0, "learning_rate": 2.0962226281246367e-07, "loss": 0.016, "step": 12689 }, { "epoch": 5.0, "learning_rate": 2.0959675920171335e-07, "loss": 0.015, "step": 12690 }, { "epoch": 5.0, "learning_rate": 2.095712554345504e-07, "loss": 0.0136, "step": 12691 }, { "epoch": 5.0, "learning_rate": 2.0954575151139054e-07, "loss": 0.0137, "step": 12692 }, { "epoch": 5.0, "learning_rate": 2.0952024743264948e-07, "loss": 0.0134, "step": 12693 }, { "epoch": 5.0, "learning_rate": 2.0949474319874279e-07, "loss": 0.0115, "step": 12694 }, { "epoch": 5.0, "learning_rate": 2.0946923881008627e-07, "loss": 0.0152, "step": 12695 }, { "epoch": 5.0, "learning_rate": 2.0944373426709553e-07, "loss": 0.0106, "step": 12696 }, { "epoch": 5.0, "learning_rate": 2.0941822957018628e-07, "loss": 0.013, "step": 12697 }, { "epoch": 5.0, "learning_rate": 2.093927247197742e-07, "loss": 0.0146, "step": 12698 }, { "epoch": 5.0, "learning_rate": 2.09367219716275e-07, "loss": 0.017, "step": 12699 }, { "epoch": 5.01, "learning_rate": 2.093417145601043e-07, "loss": 0.0141, "step": 12700 }, { "epoch": 5.01, "learning_rate": 2.093162092516779e-07, "loss": 0.0149, "step": 12701 }, { "epoch": 5.01, "learning_rate": 2.0929070379141147e-07, "loss": 0.0125, "step": 12702 }, { "epoch": 5.01, "learning_rate": 2.092651981797206e-07, "loss": 0.0145, "step": 12703 }, { "epoch": 5.01, "learning_rate": 2.0923969241702117e-07, "loss": 0.0149, "step": 12704 }, { "epoch": 5.01, "learning_rate": 2.0921418650372872e-07, "loss": 0.0166, "step": 12705 }, { "epoch": 5.01, "learning_rate": 2.0918868044025912e-07, "loss": 0.0142, "step": 12706 }, { "epoch": 5.01, "learning_rate": 2.0916317422702794e-07, "loss": 0.0157, "step": 12707 }, { "epoch": 5.01, "learning_rate": 2.0913766786445097e-07, "loss": 0.0118, "step": 12708 }, { "epoch": 5.01, "learning_rate": 2.091121613529439e-07, "loss": 0.0123, "step": 12709 }, { "epoch": 5.01, "learning_rate": 2.0908665469292246e-07, "loss": 0.0153, "step": 12710 }, { "epoch": 5.01, "learning_rate": 2.0906114788480233e-07, "loss": 0.0126, "step": 12711 }, { "epoch": 5.01, "learning_rate": 2.0903564092899922e-07, "loss": 0.0137, "step": 12712 }, { "epoch": 5.01, "learning_rate": 2.09010133825929e-07, "loss": 0.0148, "step": 12713 }, { "epoch": 5.01, "learning_rate": 2.0898462657600721e-07, "loss": 0.0132, "step": 12714 }, { "epoch": 5.01, "learning_rate": 2.0895911917964967e-07, "loss": 0.0154, "step": 12715 }, { "epoch": 5.01, "learning_rate": 2.0893361163727215e-07, "loss": 0.0126, "step": 12716 }, { "epoch": 5.01, "learning_rate": 2.089081039492903e-07, "loss": 0.0132, "step": 12717 }, { "epoch": 5.01, "learning_rate": 2.0888259611611993e-07, "loss": 0.0148, "step": 12718 }, { "epoch": 5.01, "learning_rate": 2.0885708813817667e-07, "loss": 0.0119, "step": 12719 }, { "epoch": 5.01, "learning_rate": 2.0883158001587643e-07, "loss": 0.0146, "step": 12720 }, { "epoch": 5.01, "learning_rate": 2.0880607174963476e-07, "loss": 0.0146, "step": 12721 }, { "epoch": 5.01, "learning_rate": 2.0878056333986755e-07, "loss": 0.015, "step": 12722 }, { "epoch": 5.01, "learning_rate": 2.0875505478699052e-07, "loss": 0.011, "step": 12723 }, { "epoch": 5.01, "learning_rate": 2.087295460914194e-07, "loss": 0.0138, "step": 12724 }, { "epoch": 5.02, "learning_rate": 2.0870403725356992e-07, "loss": 0.0111, "step": 12725 }, { "epoch": 5.02, "learning_rate": 2.0867852827385787e-07, "loss": 0.0125, "step": 12726 }, { "epoch": 5.02, "learning_rate": 2.0865301915269905e-07, "loss": 0.0147, "step": 12727 }, { "epoch": 5.02, "learning_rate": 2.0862750989050908e-07, "loss": 0.0119, "step": 12728 }, { "epoch": 5.02, "learning_rate": 2.0860200048770387e-07, "loss": 0.0135, "step": 12729 }, { "epoch": 5.02, "learning_rate": 2.085764909446991e-07, "loss": 0.0146, "step": 12730 }, { "epoch": 5.02, "learning_rate": 2.0855098126191056e-07, "loss": 0.0116, "step": 12731 }, { "epoch": 5.02, "learning_rate": 2.0852547143975405e-07, "loss": 0.0159, "step": 12732 }, { "epoch": 5.02, "learning_rate": 2.0849996147864536e-07, "loss": 0.0124, "step": 12733 }, { "epoch": 5.02, "learning_rate": 2.0847445137900011e-07, "loss": 0.0114, "step": 12734 }, { "epoch": 5.02, "learning_rate": 2.0844894114123423e-07, "loss": 0.0141, "step": 12735 }, { "epoch": 5.02, "learning_rate": 2.0842343076576343e-07, "loss": 0.0166, "step": 12736 }, { "epoch": 5.02, "learning_rate": 2.0839792025300357e-07, "loss": 0.0132, "step": 12737 }, { "epoch": 5.02, "learning_rate": 2.0837240960337032e-07, "loss": 0.0121, "step": 12738 }, { "epoch": 5.02, "learning_rate": 2.083468988172795e-07, "loss": 0.0144, "step": 12739 }, { "epoch": 5.02, "learning_rate": 2.0832138789514694e-07, "loss": 0.0099, "step": 12740 }, { "epoch": 5.02, "learning_rate": 2.0829587683738842e-07, "loss": 0.0137, "step": 12741 }, { "epoch": 5.02, "learning_rate": 2.082703656444197e-07, "loss": 0.0146, "step": 12742 }, { "epoch": 5.02, "learning_rate": 2.0824485431665657e-07, "loss": 0.0132, "step": 12743 }, { "epoch": 5.02, "learning_rate": 2.0821934285451488e-07, "loss": 0.0127, "step": 12744 }, { "epoch": 5.02, "learning_rate": 2.081938312584104e-07, "loss": 0.016, "step": 12745 }, { "epoch": 5.02, "learning_rate": 2.081683195287589e-07, "loss": 0.0121, "step": 12746 }, { "epoch": 5.02, "learning_rate": 2.0814280766597621e-07, "loss": 0.0119, "step": 12747 }, { "epoch": 5.02, "learning_rate": 2.081172956704781e-07, "loss": 0.0126, "step": 12748 }, { "epoch": 5.02, "learning_rate": 2.080917835426805e-07, "loss": 0.0156, "step": 12749 }, { "epoch": 5.02, "learning_rate": 2.0806627128299904e-07, "loss": 0.0138, "step": 12750 }, { "epoch": 5.03, "learning_rate": 2.0804075889184962e-07, "loss": 0.013, "step": 12751 }, { "epoch": 5.03, "learning_rate": 2.080152463696481e-07, "loss": 0.0137, "step": 12752 }, { "epoch": 5.03, "learning_rate": 2.0798973371681024e-07, "loss": 0.0126, "step": 12753 }, { "epoch": 5.03, "learning_rate": 2.079642209337518e-07, "loss": 0.0153, "step": 12754 }, { "epoch": 5.03, "learning_rate": 2.0793870802088875e-07, "loss": 0.0112, "step": 12755 }, { "epoch": 5.03, "learning_rate": 2.0791319497863675e-07, "loss": 0.0134, "step": 12756 }, { "epoch": 5.03, "learning_rate": 2.0788768180741173e-07, "loss": 0.0108, "step": 12757 }, { "epoch": 5.03, "learning_rate": 2.0786216850762953e-07, "loss": 0.0089, "step": 12758 }, { "epoch": 5.03, "learning_rate": 2.0783665507970586e-07, "loss": 0.0144, "step": 12759 }, { "epoch": 5.03, "learning_rate": 2.0781114152405665e-07, "loss": 0.0167, "step": 12760 }, { "epoch": 5.03, "learning_rate": 2.077856278410977e-07, "loss": 0.0153, "step": 12761 }, { "epoch": 5.03, "learning_rate": 2.0776011403124487e-07, "loss": 0.0123, "step": 12762 }, { "epoch": 5.03, "learning_rate": 2.0773460009491394e-07, "loss": 0.0123, "step": 12763 }, { "epoch": 5.03, "learning_rate": 2.077090860325208e-07, "loss": 0.0113, "step": 12764 }, { "epoch": 5.03, "learning_rate": 2.0768357184448123e-07, "loss": 0.0125, "step": 12765 }, { "epoch": 5.03, "learning_rate": 2.076580575312112e-07, "loss": 0.0117, "step": 12766 }, { "epoch": 5.03, "learning_rate": 2.0763254309312646e-07, "loss": 0.0142, "step": 12767 }, { "epoch": 5.03, "learning_rate": 2.0760702853064283e-07, "loss": 0.0149, "step": 12768 }, { "epoch": 5.03, "learning_rate": 2.075815138441762e-07, "loss": 0.016, "step": 12769 }, { "epoch": 5.03, "learning_rate": 2.075559990341424e-07, "loss": 0.0154, "step": 12770 }, { "epoch": 5.03, "learning_rate": 2.0753048410095734e-07, "loss": 0.0141, "step": 12771 }, { "epoch": 5.03, "learning_rate": 2.0750496904503682e-07, "loss": 0.0139, "step": 12772 }, { "epoch": 5.03, "learning_rate": 2.0747945386679672e-07, "loss": 0.0103, "step": 12773 }, { "epoch": 5.03, "learning_rate": 2.074539385666529e-07, "loss": 0.0171, "step": 12774 }, { "epoch": 5.03, "learning_rate": 2.0742842314502117e-07, "loss": 0.0133, "step": 12775 }, { "epoch": 5.04, "learning_rate": 2.0740290760231748e-07, "loss": 0.0124, "step": 12776 }, { "epoch": 5.04, "learning_rate": 2.073773919389576e-07, "loss": 0.0127, "step": 12777 }, { "epoch": 5.04, "learning_rate": 2.0735187615535748e-07, "loss": 0.0171, "step": 12778 }, { "epoch": 5.04, "learning_rate": 2.0732636025193293e-07, "loss": 0.0177, "step": 12779 }, { "epoch": 5.04, "learning_rate": 2.0730084422909986e-07, "loss": 0.0155, "step": 12780 }, { "epoch": 5.04, "learning_rate": 2.0727532808727405e-07, "loss": 0.0141, "step": 12781 }, { "epoch": 5.04, "learning_rate": 2.0724981182687156e-07, "loss": 0.0135, "step": 12782 }, { "epoch": 5.04, "learning_rate": 2.0722429544830806e-07, "loss": 0.0121, "step": 12783 }, { "epoch": 5.04, "learning_rate": 2.071987789519996e-07, "loss": 0.0133, "step": 12784 }, { "epoch": 5.04, "learning_rate": 2.0717326233836192e-07, "loss": 0.0126, "step": 12785 }, { "epoch": 5.04, "learning_rate": 2.0714774560781104e-07, "loss": 0.0134, "step": 12786 }, { "epoch": 5.04, "learning_rate": 2.0712222876076275e-07, "loss": 0.0141, "step": 12787 }, { "epoch": 5.04, "learning_rate": 2.070967117976329e-07, "loss": 0.0156, "step": 12788 }, { "epoch": 5.04, "learning_rate": 2.070711947188375e-07, "loss": 0.0143, "step": 12789 }, { "epoch": 5.04, "learning_rate": 2.0704567752479232e-07, "loss": 0.0137, "step": 12790 }, { "epoch": 5.04, "learning_rate": 2.0702016021591337e-07, "loss": 0.0139, "step": 12791 }, { "epoch": 5.04, "learning_rate": 2.0699464279261635e-07, "loss": 0.0159, "step": 12792 }, { "epoch": 5.04, "learning_rate": 2.069691252553174e-07, "loss": 0.0136, "step": 12793 }, { "epoch": 5.04, "learning_rate": 2.0694360760443232e-07, "loss": 0.0145, "step": 12794 }, { "epoch": 5.04, "learning_rate": 2.069180898403769e-07, "loss": 0.0154, "step": 12795 }, { "epoch": 5.04, "learning_rate": 2.068925719635672e-07, "loss": 0.0138, "step": 12796 }, { "epoch": 5.04, "learning_rate": 2.06867053974419e-07, "loss": 0.0126, "step": 12797 }, { "epoch": 5.04, "learning_rate": 2.068415358733483e-07, "loss": 0.0136, "step": 12798 }, { "epoch": 5.04, "learning_rate": 2.0681601766077096e-07, "loss": 0.0133, "step": 12799 }, { "epoch": 5.04, "learning_rate": 2.0679049933710283e-07, "loss": 0.0157, "step": 12800 }, { "epoch": 5.04, "learning_rate": 2.0676498090275993e-07, "loss": 0.011, "step": 12801 }, { "epoch": 5.05, "learning_rate": 2.067394623581581e-07, "loss": 0.012, "step": 12802 }, { "epoch": 5.05, "learning_rate": 2.0671394370371332e-07, "loss": 0.0116, "step": 12803 }, { "epoch": 5.05, "learning_rate": 2.0668842493984143e-07, "loss": 0.0129, "step": 12804 }, { "epoch": 5.05, "learning_rate": 2.0666290606695838e-07, "loss": 0.0141, "step": 12805 }, { "epoch": 5.05, "learning_rate": 2.066373870854801e-07, "loss": 0.0142, "step": 12806 }, { "epoch": 5.05, "learning_rate": 2.066118679958225e-07, "loss": 0.0142, "step": 12807 }, { "epoch": 5.05, "learning_rate": 2.0658634879840146e-07, "loss": 0.0121, "step": 12808 }, { "epoch": 5.05, "learning_rate": 2.06560829493633e-07, "loss": 0.0147, "step": 12809 }, { "epoch": 5.05, "learning_rate": 2.0653531008193292e-07, "loss": 0.0127, "step": 12810 }, { "epoch": 5.05, "learning_rate": 2.0650979056371728e-07, "loss": 0.0138, "step": 12811 }, { "epoch": 5.05, "learning_rate": 2.064842709394019e-07, "loss": 0.0153, "step": 12812 }, { "epoch": 5.05, "learning_rate": 2.0645875120940283e-07, "loss": 0.0135, "step": 12813 }, { "epoch": 5.05, "learning_rate": 2.0643323137413585e-07, "loss": 0.0123, "step": 12814 }, { "epoch": 5.05, "learning_rate": 2.0640771143401702e-07, "loss": 0.0127, "step": 12815 }, { "epoch": 5.05, "learning_rate": 2.0638219138946223e-07, "loss": 0.0153, "step": 12816 }, { "epoch": 5.05, "learning_rate": 2.0635667124088746e-07, "loss": 0.0153, "step": 12817 }, { "epoch": 5.05, "learning_rate": 2.0633115098870854e-07, "loss": 0.0133, "step": 12818 }, { "epoch": 5.05, "learning_rate": 2.0630563063334155e-07, "loss": 0.0109, "step": 12819 }, { "epoch": 5.05, "learning_rate": 2.0628011017520235e-07, "loss": 0.0111, "step": 12820 }, { "epoch": 5.05, "learning_rate": 2.062545896147068e-07, "loss": 0.0128, "step": 12821 }, { "epoch": 5.05, "learning_rate": 2.0622906895227107e-07, "loss": 0.0129, "step": 12822 }, { "epoch": 5.05, "learning_rate": 2.0620354818831092e-07, "loss": 0.0119, "step": 12823 }, { "epoch": 5.05, "learning_rate": 2.0617802732324238e-07, "loss": 0.0127, "step": 12824 }, { "epoch": 5.05, "learning_rate": 2.0615250635748135e-07, "loss": 0.0125, "step": 12825 }, { "epoch": 5.05, "learning_rate": 2.0612698529144387e-07, "loss": 0.0114, "step": 12826 }, { "epoch": 5.06, "learning_rate": 2.061014641255458e-07, "loss": 0.0114, "step": 12827 }, { "epoch": 5.06, "learning_rate": 2.0607594286020314e-07, "loss": 0.0102, "step": 12828 }, { "epoch": 5.06, "learning_rate": 2.060504214958319e-07, "loss": 0.0138, "step": 12829 }, { "epoch": 5.06, "learning_rate": 2.060249000328479e-07, "loss": 0.0154, "step": 12830 }, { "epoch": 5.06, "learning_rate": 2.0599937847166724e-07, "loss": 0.0119, "step": 12831 }, { "epoch": 5.06, "learning_rate": 2.059738568127058e-07, "loss": 0.0114, "step": 12832 }, { "epoch": 5.06, "learning_rate": 2.0594833505637963e-07, "loss": 0.0142, "step": 12833 }, { "epoch": 5.06, "learning_rate": 2.0592281320310457e-07, "loss": 0.0131, "step": 12834 }, { "epoch": 5.06, "learning_rate": 2.058972912532967e-07, "loss": 0.0134, "step": 12835 }, { "epoch": 5.06, "learning_rate": 2.0587176920737192e-07, "loss": 0.0171, "step": 12836 }, { "epoch": 5.06, "learning_rate": 2.0584624706574627e-07, "loss": 0.0108, "step": 12837 }, { "epoch": 5.06, "learning_rate": 2.0582072482883564e-07, "loss": 0.0142, "step": 12838 }, { "epoch": 5.06, "learning_rate": 2.0579520249705607e-07, "loss": 0.0149, "step": 12839 }, { "epoch": 5.06, "learning_rate": 2.0576968007082347e-07, "loss": 0.0124, "step": 12840 }, { "epoch": 5.06, "learning_rate": 2.057441575505539e-07, "loss": 0.0144, "step": 12841 }, { "epoch": 5.06, "learning_rate": 2.0571863493666326e-07, "loss": 0.0141, "step": 12842 }, { "epoch": 5.06, "learning_rate": 2.0569311222956753e-07, "loss": 0.0136, "step": 12843 }, { "epoch": 5.06, "learning_rate": 2.056675894296828e-07, "loss": 0.0121, "step": 12844 }, { "epoch": 5.06, "learning_rate": 2.0564206653742492e-07, "loss": 0.0117, "step": 12845 }, { "epoch": 5.06, "learning_rate": 2.0561654355320997e-07, "loss": 0.0145, "step": 12846 }, { "epoch": 5.06, "learning_rate": 2.0559102047745384e-07, "loss": 0.013, "step": 12847 }, { "epoch": 5.06, "learning_rate": 2.0556549731057263e-07, "loss": 0.0125, "step": 12848 }, { "epoch": 5.06, "learning_rate": 2.055399740529822e-07, "loss": 0.0124, "step": 12849 }, { "epoch": 5.06, "learning_rate": 2.0551445070509867e-07, "loss": 0.0101, "step": 12850 }, { "epoch": 5.06, "learning_rate": 2.05488927267338e-07, "loss": 0.0117, "step": 12851 }, { "epoch": 5.07, "learning_rate": 2.0546340374011608e-07, "loss": 0.0141, "step": 12852 }, { "epoch": 5.07, "learning_rate": 2.0543788012384903e-07, "loss": 0.0128, "step": 12853 }, { "epoch": 5.07, "learning_rate": 2.054123564189528e-07, "loss": 0.0129, "step": 12854 }, { "epoch": 5.07, "learning_rate": 2.053868326258434e-07, "loss": 0.0145, "step": 12855 }, { "epoch": 5.07, "learning_rate": 2.0536130874493674e-07, "loss": 0.0158, "step": 12856 }, { "epoch": 5.07, "learning_rate": 2.05335784776649e-07, "loss": 0.0163, "step": 12857 }, { "epoch": 5.07, "learning_rate": 2.0531026072139604e-07, "loss": 0.013, "step": 12858 }, { "epoch": 5.07, "learning_rate": 2.0528473657959386e-07, "loss": 0.0151, "step": 12859 }, { "epoch": 5.07, "learning_rate": 2.0525921235165854e-07, "loss": 0.0137, "step": 12860 }, { "epoch": 5.07, "learning_rate": 2.0523368803800608e-07, "loss": 0.0116, "step": 12861 }, { "epoch": 5.07, "learning_rate": 2.0520816363905245e-07, "loss": 0.0154, "step": 12862 }, { "epoch": 5.07, "learning_rate": 2.051826391552136e-07, "loss": 0.0156, "step": 12863 }, { "epoch": 5.07, "learning_rate": 2.0515711458690567e-07, "loss": 0.0142, "step": 12864 }, { "epoch": 5.07, "learning_rate": 2.0513158993454458e-07, "loss": 0.0137, "step": 12865 }, { "epoch": 5.07, "learning_rate": 2.0510606519854638e-07, "loss": 0.0144, "step": 12866 }, { "epoch": 5.07, "learning_rate": 2.0508054037932712e-07, "loss": 0.0163, "step": 12867 }, { "epoch": 5.07, "learning_rate": 2.050550154773027e-07, "loss": 0.0156, "step": 12868 }, { "epoch": 5.07, "learning_rate": 2.0502949049288927e-07, "loss": 0.014, "step": 12869 }, { "epoch": 5.07, "learning_rate": 2.0500396542650275e-07, "loss": 0.0115, "step": 12870 }, { "epoch": 5.07, "learning_rate": 2.049784402785592e-07, "loss": 0.0117, "step": 12871 }, { "epoch": 5.07, "learning_rate": 2.0495291504947464e-07, "loss": 0.0119, "step": 12872 }, { "epoch": 5.07, "learning_rate": 2.049273897396651e-07, "loss": 0.0127, "step": 12873 }, { "epoch": 5.07, "learning_rate": 2.049018643495466e-07, "loss": 0.0138, "step": 12874 }, { "epoch": 5.07, "learning_rate": 2.0487633887953519e-07, "loss": 0.0147, "step": 12875 }, { "epoch": 5.07, "learning_rate": 2.0485081333004678e-07, "loss": 0.0111, "step": 12876 }, { "epoch": 5.07, "learning_rate": 2.048252877014976e-07, "loss": 0.0117, "step": 12877 }, { "epoch": 5.08, "learning_rate": 2.047997619943035e-07, "loss": 0.0141, "step": 12878 }, { "epoch": 5.08, "learning_rate": 2.047742362088806e-07, "loss": 0.0138, "step": 12879 }, { "epoch": 5.08, "learning_rate": 2.0474871034564488e-07, "loss": 0.0121, "step": 12880 }, { "epoch": 5.08, "learning_rate": 2.0472318440501244e-07, "loss": 0.0154, "step": 12881 }, { "epoch": 5.08, "learning_rate": 2.046976583873992e-07, "loss": 0.0129, "step": 12882 }, { "epoch": 5.08, "learning_rate": 2.0467213229322136e-07, "loss": 0.0134, "step": 12883 }, { "epoch": 5.08, "learning_rate": 2.046466061228948e-07, "loss": 0.0133, "step": 12884 }, { "epoch": 5.08, "learning_rate": 2.046210798768356e-07, "loss": 0.0139, "step": 12885 }, { "epoch": 5.08, "learning_rate": 2.045955535554599e-07, "loss": 0.0123, "step": 12886 }, { "epoch": 5.08, "learning_rate": 2.0457002715918362e-07, "loss": 0.0126, "step": 12887 }, { "epoch": 5.08, "learning_rate": 2.0454450068842282e-07, "loss": 0.016, "step": 12888 }, { "epoch": 5.08, "learning_rate": 2.0451897414359362e-07, "loss": 0.0104, "step": 12889 }, { "epoch": 5.08, "learning_rate": 2.0449344752511192e-07, "loss": 0.0125, "step": 12890 }, { "epoch": 5.08, "learning_rate": 2.0446792083339397e-07, "loss": 0.0124, "step": 12891 }, { "epoch": 5.08, "learning_rate": 2.044423940688556e-07, "loss": 0.014, "step": 12892 }, { "epoch": 5.08, "learning_rate": 2.04416867231913e-07, "loss": 0.0138, "step": 12893 }, { "epoch": 5.08, "learning_rate": 2.0439134032298217e-07, "loss": 0.0135, "step": 12894 }, { "epoch": 5.08, "learning_rate": 2.0436581334247917e-07, "loss": 0.0138, "step": 12895 }, { "epoch": 5.08, "learning_rate": 2.0434028629082002e-07, "loss": 0.0114, "step": 12896 }, { "epoch": 5.08, "learning_rate": 2.043147591684209e-07, "loss": 0.0148, "step": 12897 }, { "epoch": 5.08, "learning_rate": 2.042892319756977e-07, "loss": 0.0137, "step": 12898 }, { "epoch": 5.08, "learning_rate": 2.042637047130665e-07, "loss": 0.0151, "step": 12899 }, { "epoch": 5.08, "learning_rate": 2.0423817738094345e-07, "loss": 0.017, "step": 12900 }, { "epoch": 5.08, "learning_rate": 2.0421264997974448e-07, "loss": 0.0147, "step": 12901 }, { "epoch": 5.08, "learning_rate": 2.0418712250988582e-07, "loss": 0.014, "step": 12902 }, { "epoch": 5.09, "learning_rate": 2.0416159497178339e-07, "loss": 0.0134, "step": 12903 }, { "epoch": 5.09, "learning_rate": 2.041360673658532e-07, "loss": 0.0139, "step": 12904 }, { "epoch": 5.09, "learning_rate": 2.041105396925115e-07, "loss": 0.0136, "step": 12905 }, { "epoch": 5.09, "learning_rate": 2.0408501195217425e-07, "loss": 0.0128, "step": 12906 }, { "epoch": 5.09, "learning_rate": 2.0405948414525746e-07, "loss": 0.0128, "step": 12907 }, { "epoch": 5.09, "learning_rate": 2.0403395627217727e-07, "loss": 0.0122, "step": 12908 }, { "epoch": 5.09, "learning_rate": 2.0400842833334974e-07, "loss": 0.0154, "step": 12909 }, { "epoch": 5.09, "learning_rate": 2.0398290032919085e-07, "loss": 0.0154, "step": 12910 }, { "epoch": 5.09, "learning_rate": 2.039573722601168e-07, "loss": 0.0151, "step": 12911 }, { "epoch": 5.09, "learning_rate": 2.039318441265436e-07, "loss": 0.0131, "step": 12912 }, { "epoch": 5.09, "learning_rate": 2.0390631592888733e-07, "loss": 0.0118, "step": 12913 }, { "epoch": 5.09, "learning_rate": 2.0388078766756397e-07, "loss": 0.0127, "step": 12914 }, { "epoch": 5.09, "learning_rate": 2.0385525934298972e-07, "loss": 0.015, "step": 12915 }, { "epoch": 5.09, "learning_rate": 2.0382973095558058e-07, "loss": 0.012, "step": 12916 }, { "epoch": 5.09, "learning_rate": 2.038042025057527e-07, "loss": 0.0129, "step": 12917 }, { "epoch": 5.09, "learning_rate": 2.037786739939221e-07, "loss": 0.0154, "step": 12918 }, { "epoch": 5.09, "learning_rate": 2.0375314542050478e-07, "loss": 0.0141, "step": 12919 }, { "epoch": 5.09, "learning_rate": 2.0372761678591695e-07, "loss": 0.0149, "step": 12920 }, { "epoch": 5.09, "learning_rate": 2.0370208809057462e-07, "loss": 0.0138, "step": 12921 }, { "epoch": 5.09, "learning_rate": 2.0367655933489387e-07, "loss": 0.0116, "step": 12922 }, { "epoch": 5.09, "learning_rate": 2.036510305192908e-07, "loss": 0.0148, "step": 12923 }, { "epoch": 5.09, "learning_rate": 2.0362550164418152e-07, "loss": 0.0129, "step": 12924 }, { "epoch": 5.09, "learning_rate": 2.03599972709982e-07, "loss": 0.0102, "step": 12925 }, { "epoch": 5.09, "learning_rate": 2.0357444371710848e-07, "loss": 0.0101, "step": 12926 }, { "epoch": 5.09, "learning_rate": 2.0354891466597687e-07, "loss": 0.0138, "step": 12927 }, { "epoch": 5.1, "learning_rate": 2.0352338555700342e-07, "loss": 0.0136, "step": 12928 }, { "epoch": 5.1, "learning_rate": 2.034978563906041e-07, "loss": 0.0142, "step": 12929 }, { "epoch": 5.1, "learning_rate": 2.0347232716719502e-07, "loss": 0.0136, "step": 12930 }, { "epoch": 5.1, "learning_rate": 2.0344679788719235e-07, "loss": 0.0153, "step": 12931 }, { "epoch": 5.1, "learning_rate": 2.0342126855101203e-07, "loss": 0.0122, "step": 12932 }, { "epoch": 5.1, "learning_rate": 2.0339573915907028e-07, "loss": 0.0153, "step": 12933 }, { "epoch": 5.1, "learning_rate": 2.0337020971178314e-07, "loss": 0.0115, "step": 12934 }, { "epoch": 5.1, "learning_rate": 2.0334468020956673e-07, "loss": 0.0162, "step": 12935 }, { "epoch": 5.1, "learning_rate": 2.0331915065283702e-07, "loss": 0.0091, "step": 12936 }, { "epoch": 5.1, "learning_rate": 2.032936210420103e-07, "loss": 0.0158, "step": 12937 }, { "epoch": 5.1, "learning_rate": 2.0326809137750255e-07, "loss": 0.0143, "step": 12938 }, { "epoch": 5.1, "learning_rate": 2.0324256165972983e-07, "loss": 0.0149, "step": 12939 }, { "epoch": 5.1, "learning_rate": 2.0321703188910835e-07, "loss": 0.0143, "step": 12940 }, { "epoch": 5.1, "learning_rate": 2.0319150206605404e-07, "loss": 0.015, "step": 12941 }, { "epoch": 5.1, "learning_rate": 2.0316597219098318e-07, "loss": 0.0133, "step": 12942 }, { "epoch": 5.1, "learning_rate": 2.0314044226431175e-07, "loss": 0.0107, "step": 12943 }, { "epoch": 5.1, "learning_rate": 2.031149122864559e-07, "loss": 0.0132, "step": 12944 }, { "epoch": 5.1, "learning_rate": 2.0308938225783173e-07, "loss": 0.0139, "step": 12945 }, { "epoch": 5.1, "learning_rate": 2.0306385217885529e-07, "loss": 0.0143, "step": 12946 }, { "epoch": 5.1, "learning_rate": 2.030383220499427e-07, "loss": 0.0112, "step": 12947 }, { "epoch": 5.1, "learning_rate": 2.0301279187151014e-07, "loss": 0.0115, "step": 12948 }, { "epoch": 5.1, "learning_rate": 2.0298726164397367e-07, "loss": 0.0133, "step": 12949 }, { "epoch": 5.1, "learning_rate": 2.0296173136774927e-07, "loss": 0.0118, "step": 12950 }, { "epoch": 5.1, "learning_rate": 2.0293620104325323e-07, "loss": 0.013, "step": 12951 }, { "epoch": 5.1, "learning_rate": 2.0291067067090153e-07, "loss": 0.0154, "step": 12952 }, { "epoch": 5.1, "learning_rate": 2.0288514025111034e-07, "loss": 0.0148, "step": 12953 }, { "epoch": 5.11, "learning_rate": 2.0285960978429575e-07, "loss": 0.0088, "step": 12954 }, { "epoch": 5.11, "learning_rate": 2.0283407927087385e-07, "loss": 0.0121, "step": 12955 }, { "epoch": 5.11, "learning_rate": 2.028085487112608e-07, "loss": 0.0171, "step": 12956 }, { "epoch": 5.11, "learning_rate": 2.0278301810587264e-07, "loss": 0.0142, "step": 12957 }, { "epoch": 5.11, "learning_rate": 2.0275748745512559e-07, "loss": 0.0157, "step": 12958 }, { "epoch": 5.11, "learning_rate": 2.0273195675943557e-07, "loss": 0.0111, "step": 12959 }, { "epoch": 5.11, "learning_rate": 2.0270642601921889e-07, "loss": 0.0117, "step": 12960 }, { "epoch": 5.11, "learning_rate": 2.026808952348915e-07, "loss": 0.0168, "step": 12961 }, { "epoch": 5.11, "learning_rate": 2.026553644068697e-07, "loss": 0.0141, "step": 12962 }, { "epoch": 5.11, "learning_rate": 2.026298335355694e-07, "loss": 0.0169, "step": 12963 }, { "epoch": 5.11, "learning_rate": 2.0260430262140686e-07, "loss": 0.0131, "step": 12964 }, { "epoch": 5.11, "learning_rate": 2.0257877166479812e-07, "loss": 0.011, "step": 12965 }, { "epoch": 5.11, "learning_rate": 2.0255324066615933e-07, "loss": 0.0131, "step": 12966 }, { "epoch": 5.11, "learning_rate": 2.0252770962590663e-07, "loss": 0.0114, "step": 12967 }, { "epoch": 5.11, "learning_rate": 2.0250217854445606e-07, "loss": 0.0128, "step": 12968 }, { "epoch": 5.11, "learning_rate": 2.0247664742222383e-07, "loss": 0.0136, "step": 12969 }, { "epoch": 5.11, "learning_rate": 2.02451116259626e-07, "loss": 0.0126, "step": 12970 }, { "epoch": 5.11, "learning_rate": 2.024255850570787e-07, "loss": 0.0118, "step": 12971 }, { "epoch": 5.11, "learning_rate": 2.02400053814998e-07, "loss": 0.0123, "step": 12972 }, { "epoch": 5.11, "learning_rate": 2.0237452253380013e-07, "loss": 0.0142, "step": 12973 }, { "epoch": 5.11, "learning_rate": 2.023489912139011e-07, "loss": 0.0137, "step": 12974 }, { "epoch": 5.11, "learning_rate": 2.0232345985571713e-07, "loss": 0.0115, "step": 12975 }, { "epoch": 5.11, "learning_rate": 2.0229792845966425e-07, "loss": 0.0113, "step": 12976 }, { "epoch": 5.11, "learning_rate": 2.0227239702615866e-07, "loss": 0.013, "step": 12977 }, { "epoch": 5.11, "learning_rate": 2.022468655556165e-07, "loss": 0.0143, "step": 12978 }, { "epoch": 5.12, "learning_rate": 2.0222133404845371e-07, "loss": 0.0119, "step": 12979 }, { "epoch": 5.12, "learning_rate": 2.021958025050867e-07, "loss": 0.0166, "step": 12980 }, { "epoch": 5.12, "learning_rate": 2.0217027092593133e-07, "loss": 0.0132, "step": 12981 }, { "epoch": 5.12, "learning_rate": 2.0214473931140393e-07, "loss": 0.014, "step": 12982 }, { "epoch": 5.12, "learning_rate": 2.0211920766192043e-07, "loss": 0.0176, "step": 12983 }, { "epoch": 5.12, "learning_rate": 2.0209367597789715e-07, "loss": 0.0163, "step": 12984 }, { "epoch": 5.12, "learning_rate": 2.020681442597501e-07, "loss": 0.0147, "step": 12985 }, { "epoch": 5.12, "learning_rate": 2.0204261250789546e-07, "loss": 0.0111, "step": 12986 }, { "epoch": 5.12, "learning_rate": 2.020170807227494e-07, "loss": 0.0152, "step": 12987 }, { "epoch": 5.12, "learning_rate": 2.019915489047279e-07, "loss": 0.0105, "step": 12988 }, { "epoch": 5.12, "learning_rate": 2.019660170542472e-07, "loss": 0.0111, "step": 12989 }, { "epoch": 5.12, "learning_rate": 2.0194048517172347e-07, "loss": 0.0117, "step": 12990 }, { "epoch": 5.12, "learning_rate": 2.019149532575727e-07, "loss": 0.0138, "step": 12991 }, { "epoch": 5.12, "learning_rate": 2.0188942131221114e-07, "loss": 0.0159, "step": 12992 }, { "epoch": 5.12, "learning_rate": 2.018638893360549e-07, "loss": 0.0131, "step": 12993 }, { "epoch": 5.12, "learning_rate": 2.0183835732952007e-07, "loss": 0.011, "step": 12994 }, { "epoch": 5.12, "learning_rate": 2.0181282529302283e-07, "loss": 0.013, "step": 12995 }, { "epoch": 5.12, "learning_rate": 2.0178729322697924e-07, "loss": 0.0128, "step": 12996 }, { "epoch": 5.12, "learning_rate": 2.0176176113180556e-07, "loss": 0.0128, "step": 12997 }, { "epoch": 5.12, "learning_rate": 2.0173622900791784e-07, "loss": 0.0111, "step": 12998 }, { "epoch": 5.12, "learning_rate": 2.0171069685573214e-07, "loss": 0.0131, "step": 12999 }, { "epoch": 5.12, "learning_rate": 2.0168516467566477e-07, "loss": 0.0145, "step": 13000 }, { "epoch": 5.12, "learning_rate": 2.0165963246813172e-07, "loss": 0.0124, "step": 13001 }, { "epoch": 5.12, "learning_rate": 2.0163410023354923e-07, "loss": 0.0129, "step": 13002 }, { "epoch": 5.12, "learning_rate": 2.0160856797233335e-07, "loss": 0.0117, "step": 13003 }, { "epoch": 5.12, "learning_rate": 2.015830356849003e-07, "loss": 0.012, "step": 13004 }, { "epoch": 5.13, "learning_rate": 2.0155750337166615e-07, "loss": 0.0144, "step": 13005 }, { "epoch": 5.13, "learning_rate": 2.0153197103304708e-07, "loss": 0.0118, "step": 13006 }, { "epoch": 5.13, "learning_rate": 2.015064386694592e-07, "loss": 0.0137, "step": 13007 }, { "epoch": 5.13, "learning_rate": 2.0148090628131863e-07, "loss": 0.0112, "step": 13008 }, { "epoch": 5.13, "learning_rate": 2.0145537386904157e-07, "loss": 0.012, "step": 13009 }, { "epoch": 5.13, "learning_rate": 2.0142984143304412e-07, "loss": 0.0133, "step": 13010 }, { "epoch": 5.13, "learning_rate": 2.0140430897374246e-07, "loss": 0.0115, "step": 13011 }, { "epoch": 5.13, "learning_rate": 2.0137877649155265e-07, "loss": 0.0124, "step": 13012 }, { "epoch": 5.13, "learning_rate": 2.0135324398689092e-07, "loss": 0.0156, "step": 13013 }, { "epoch": 5.13, "learning_rate": 2.0132771146017333e-07, "loss": 0.0128, "step": 13014 }, { "epoch": 5.13, "learning_rate": 2.0130217891181606e-07, "loss": 0.0136, "step": 13015 }, { "epoch": 5.13, "learning_rate": 2.012766463422353e-07, "loss": 0.0111, "step": 13016 }, { "epoch": 5.13, "learning_rate": 2.012511137518471e-07, "loss": 0.016, "step": 13017 }, { "epoch": 5.13, "learning_rate": 2.012255811410677e-07, "loss": 0.0139, "step": 13018 }, { "epoch": 5.13, "learning_rate": 2.012000485103131e-07, "loss": 0.0125, "step": 13019 }, { "epoch": 5.13, "learning_rate": 2.011745158599996e-07, "loss": 0.0138, "step": 13020 }, { "epoch": 5.13, "learning_rate": 2.0114898319054325e-07, "loss": 0.0111, "step": 13021 }, { "epoch": 5.13, "learning_rate": 2.0112345050236024e-07, "loss": 0.0157, "step": 13022 }, { "epoch": 5.13, "learning_rate": 2.0109791779586666e-07, "loss": 0.0126, "step": 13023 }, { "epoch": 5.13, "learning_rate": 2.0107238507147874e-07, "loss": 0.0175, "step": 13024 }, { "epoch": 5.13, "learning_rate": 2.0104685232961253e-07, "loss": 0.0142, "step": 13025 }, { "epoch": 5.13, "learning_rate": 2.0102131957068424e-07, "loss": 0.0119, "step": 13026 }, { "epoch": 5.13, "learning_rate": 2.0099578679511e-07, "loss": 0.0142, "step": 13027 }, { "epoch": 5.13, "learning_rate": 2.0097025400330594e-07, "loss": 0.0125, "step": 13028 }, { "epoch": 5.13, "learning_rate": 2.0094472119568825e-07, "loss": 0.0107, "step": 13029 }, { "epoch": 5.14, "learning_rate": 2.0091918837267294e-07, "loss": 0.0148, "step": 13030 }, { "epoch": 5.14, "learning_rate": 2.0089365553467633e-07, "loss": 0.0113, "step": 13031 }, { "epoch": 5.14, "learning_rate": 2.0086812268211446e-07, "loss": 0.0119, "step": 13032 }, { "epoch": 5.14, "learning_rate": 2.0084258981540353e-07, "loss": 0.0135, "step": 13033 }, { "epoch": 5.14, "learning_rate": 2.008170569349597e-07, "loss": 0.0126, "step": 13034 }, { "epoch": 5.14, "learning_rate": 2.0079152404119908e-07, "loss": 0.0118, "step": 13035 }, { "epoch": 5.14, "learning_rate": 2.0076599113453777e-07, "loss": 0.0146, "step": 13036 }, { "epoch": 5.14, "learning_rate": 2.0074045821539204e-07, "loss": 0.0114, "step": 13037 }, { "epoch": 5.14, "learning_rate": 2.007149252841779e-07, "loss": 0.0108, "step": 13038 }, { "epoch": 5.14, "learning_rate": 2.0068939234131155e-07, "loss": 0.0131, "step": 13039 }, { "epoch": 5.14, "learning_rate": 2.006638593872092e-07, "loss": 0.011, "step": 13040 }, { "epoch": 5.14, "learning_rate": 2.0063832642228692e-07, "loss": 0.0136, "step": 13041 }, { "epoch": 5.14, "learning_rate": 2.0061279344696093e-07, "loss": 0.0128, "step": 13042 }, { "epoch": 5.14, "learning_rate": 2.005872604616473e-07, "loss": 0.0139, "step": 13043 }, { "epoch": 5.14, "learning_rate": 2.0056172746676223e-07, "loss": 0.014, "step": 13044 }, { "epoch": 5.14, "learning_rate": 2.0053619446272182e-07, "loss": 0.0144, "step": 13045 }, { "epoch": 5.14, "learning_rate": 2.0051066144994234e-07, "loss": 0.0119, "step": 13046 }, { "epoch": 5.14, "learning_rate": 2.004851284288398e-07, "loss": 0.0117, "step": 13047 }, { "epoch": 5.14, "learning_rate": 2.0045959539983038e-07, "loss": 0.0119, "step": 13048 }, { "epoch": 5.14, "learning_rate": 2.004340623633303e-07, "loss": 0.0147, "step": 13049 }, { "epoch": 5.14, "learning_rate": 2.0040852931975562e-07, "loss": 0.0145, "step": 13050 }, { "epoch": 5.14, "learning_rate": 2.0038299626952256e-07, "loss": 0.013, "step": 13051 }, { "epoch": 5.14, "learning_rate": 2.0035746321304727e-07, "loss": 0.0145, "step": 13052 }, { "epoch": 5.14, "learning_rate": 2.0033193015074583e-07, "loss": 0.0135, "step": 13053 }, { "epoch": 5.14, "learning_rate": 2.0030639708303449e-07, "loss": 0.0139, "step": 13054 }, { "epoch": 5.15, "learning_rate": 2.0028086401032928e-07, "loss": 0.0137, "step": 13055 }, { "epoch": 5.15, "learning_rate": 2.0025533093304646e-07, "loss": 0.0113, "step": 13056 }, { "epoch": 5.15, "learning_rate": 2.002297978516021e-07, "loss": 0.0103, "step": 13057 }, { "epoch": 5.15, "learning_rate": 2.0020426476641245e-07, "loss": 0.0159, "step": 13058 }, { "epoch": 5.15, "learning_rate": 2.0017873167789348e-07, "loss": 0.013, "step": 13059 }, { "epoch": 5.15, "learning_rate": 2.0015319858646156e-07, "loss": 0.0112, "step": 13060 }, { "epoch": 5.15, "learning_rate": 2.0012766549253267e-07, "loss": 0.0129, "step": 13061 }, { "epoch": 5.15, "learning_rate": 2.0010213239652302e-07, "loss": 0.0124, "step": 13062 }, { "epoch": 5.15, "learning_rate": 2.0007659929884876e-07, "loss": 0.0157, "step": 13063 }, { "epoch": 5.15, "learning_rate": 2.0005106619992614e-07, "loss": 0.0151, "step": 13064 }, { "epoch": 5.15, "learning_rate": 2.000255331001711e-07, "loss": 0.0135, "step": 13065 }, { "epoch": 5.15, "learning_rate": 2e-07, "loss": 0.013, "step": 13066 }, { "epoch": 5.15, "learning_rate": 1.999744668998289e-07, "loss": 0.0102, "step": 13067 }, { "epoch": 5.15, "learning_rate": 1.9994893380007392e-07, "loss": 0.0138, "step": 13068 }, { "epoch": 5.15, "learning_rate": 1.9992340070115117e-07, "loss": 0.0117, "step": 13069 }, { "epoch": 5.15, "learning_rate": 1.99897867603477e-07, "loss": 0.0152, "step": 13070 }, { "epoch": 5.15, "learning_rate": 1.9987233450746736e-07, "loss": 0.0132, "step": 13071 }, { "epoch": 5.15, "learning_rate": 1.998468014135385e-07, "loss": 0.0144, "step": 13072 }, { "epoch": 5.15, "learning_rate": 1.9982126832210653e-07, "loss": 0.0178, "step": 13073 }, { "epoch": 5.15, "learning_rate": 1.997957352335876e-07, "loss": 0.0095, "step": 13074 }, { "epoch": 5.15, "learning_rate": 1.997702021483979e-07, "loss": 0.0147, "step": 13075 }, { "epoch": 5.15, "learning_rate": 1.9974466906695354e-07, "loss": 0.0134, "step": 13076 }, { "epoch": 5.15, "learning_rate": 1.9971913598967073e-07, "loss": 0.0149, "step": 13077 }, { "epoch": 5.15, "learning_rate": 1.9969360291696552e-07, "loss": 0.0141, "step": 13078 }, { "epoch": 5.15, "learning_rate": 1.9966806984925416e-07, "loss": 0.0129, "step": 13079 }, { "epoch": 5.15, "learning_rate": 1.9964253678695276e-07, "loss": 0.0143, "step": 13080 }, { "epoch": 5.16, "learning_rate": 1.9961700373047745e-07, "loss": 0.0143, "step": 13081 }, { "epoch": 5.16, "learning_rate": 1.9959147068024434e-07, "loss": 0.0126, "step": 13082 }, { "epoch": 5.16, "learning_rate": 1.9956593763666968e-07, "loss": 0.0139, "step": 13083 }, { "epoch": 5.16, "learning_rate": 1.995404046001696e-07, "loss": 0.015, "step": 13084 }, { "epoch": 5.16, "learning_rate": 1.9951487157116025e-07, "loss": 0.014, "step": 13085 }, { "epoch": 5.16, "learning_rate": 1.9948933855005772e-07, "loss": 0.0165, "step": 13086 }, { "epoch": 5.16, "learning_rate": 1.9946380553727813e-07, "loss": 0.0153, "step": 13087 }, { "epoch": 5.16, "learning_rate": 1.9943827253323778e-07, "loss": 0.0126, "step": 13088 }, { "epoch": 5.16, "learning_rate": 1.9941273953835272e-07, "loss": 0.016, "step": 13089 }, { "epoch": 5.16, "learning_rate": 1.993872065530391e-07, "loss": 0.0128, "step": 13090 }, { "epoch": 5.16, "learning_rate": 1.9936167357771304e-07, "loss": 0.0108, "step": 13091 }, { "epoch": 5.16, "learning_rate": 1.993361406127908e-07, "loss": 0.0134, "step": 13092 }, { "epoch": 5.16, "learning_rate": 1.9931060765868843e-07, "loss": 0.0161, "step": 13093 }, { "epoch": 5.16, "learning_rate": 1.9928507471582215e-07, "loss": 0.0136, "step": 13094 }, { "epoch": 5.16, "learning_rate": 1.99259541784608e-07, "loss": 0.0152, "step": 13095 }, { "epoch": 5.16, "learning_rate": 1.992340088654622e-07, "loss": 0.0147, "step": 13096 }, { "epoch": 5.16, "learning_rate": 1.9920847595880093e-07, "loss": 0.0155, "step": 13097 }, { "epoch": 5.16, "learning_rate": 1.991829430650403e-07, "loss": 0.0138, "step": 13098 }, { "epoch": 5.16, "learning_rate": 1.9915741018459645e-07, "loss": 0.0138, "step": 13099 }, { "epoch": 5.16, "learning_rate": 1.991318773178855e-07, "loss": 0.0122, "step": 13100 }, { "epoch": 5.16, "learning_rate": 1.9910634446532365e-07, "loss": 0.015, "step": 13101 }, { "epoch": 5.16, "learning_rate": 1.9908081162732706e-07, "loss": 0.0183, "step": 13102 }, { "epoch": 5.16, "learning_rate": 1.9905527880431184e-07, "loss": 0.0159, "step": 13103 }, { "epoch": 5.16, "learning_rate": 1.9902974599669404e-07, "loss": 0.012, "step": 13104 }, { "epoch": 5.16, "learning_rate": 1.9900421320488997e-07, "loss": 0.0119, "step": 13105 }, { "epoch": 5.17, "learning_rate": 1.9897868042931577e-07, "loss": 0.012, "step": 13106 }, { "epoch": 5.17, "learning_rate": 1.9895314767038748e-07, "loss": 0.0131, "step": 13107 }, { "epoch": 5.17, "learning_rate": 1.989276149285213e-07, "loss": 0.0155, "step": 13108 }, { "epoch": 5.17, "learning_rate": 1.989020822041333e-07, "loss": 0.0114, "step": 13109 }, { "epoch": 5.17, "learning_rate": 1.9887654949763976e-07, "loss": 0.0132, "step": 13110 }, { "epoch": 5.17, "learning_rate": 1.9885101680945676e-07, "loss": 0.0127, "step": 13111 }, { "epoch": 5.17, "learning_rate": 1.9882548414000046e-07, "loss": 0.0121, "step": 13112 }, { "epoch": 5.17, "learning_rate": 1.9879995148968683e-07, "loss": 0.0135, "step": 13113 }, { "epoch": 5.17, "learning_rate": 1.9877441885893232e-07, "loss": 0.013, "step": 13114 }, { "epoch": 5.17, "learning_rate": 1.987488862481529e-07, "loss": 0.0172, "step": 13115 }, { "epoch": 5.17, "learning_rate": 1.9872335365776474e-07, "loss": 0.0116, "step": 13116 }, { "epoch": 5.17, "learning_rate": 1.9869782108818395e-07, "loss": 0.0116, "step": 13117 }, { "epoch": 5.17, "learning_rate": 1.986722885398267e-07, "loss": 0.0128, "step": 13118 }, { "epoch": 5.17, "learning_rate": 1.9864675601310909e-07, "loss": 0.0142, "step": 13119 }, { "epoch": 5.17, "learning_rate": 1.9862122350844733e-07, "loss": 0.0123, "step": 13120 }, { "epoch": 5.17, "learning_rate": 1.9859569102625758e-07, "loss": 0.0139, "step": 13121 }, { "epoch": 5.17, "learning_rate": 1.9857015856695584e-07, "loss": 0.0153, "step": 13122 }, { "epoch": 5.17, "learning_rate": 1.9854462613095842e-07, "loss": 0.0116, "step": 13123 }, { "epoch": 5.17, "learning_rate": 1.9851909371868138e-07, "loss": 0.0115, "step": 13124 }, { "epoch": 5.17, "learning_rate": 1.9849356133054084e-07, "loss": 0.0139, "step": 13125 }, { "epoch": 5.17, "learning_rate": 1.9846802896695298e-07, "loss": 0.0175, "step": 13126 }, { "epoch": 5.17, "learning_rate": 1.9844249662833383e-07, "loss": 0.0125, "step": 13127 }, { "epoch": 5.17, "learning_rate": 1.9841696431509972e-07, "loss": 0.0118, "step": 13128 }, { "epoch": 5.17, "learning_rate": 1.9839143202766664e-07, "loss": 0.0135, "step": 13129 }, { "epoch": 5.17, "learning_rate": 1.983658997664508e-07, "loss": 0.0145, "step": 13130 }, { "epoch": 5.18, "learning_rate": 1.9834036753186824e-07, "loss": 0.0171, "step": 13131 }, { "epoch": 5.18, "learning_rate": 1.9831483532433524e-07, "loss": 0.0129, "step": 13132 }, { "epoch": 5.18, "learning_rate": 1.9828930314426784e-07, "loss": 0.0151, "step": 13133 }, { "epoch": 5.18, "learning_rate": 1.9826377099208222e-07, "loss": 0.0122, "step": 13134 }, { "epoch": 5.18, "learning_rate": 1.9823823886819442e-07, "loss": 0.0131, "step": 13135 }, { "epoch": 5.18, "learning_rate": 1.982127067730207e-07, "loss": 0.0136, "step": 13136 }, { "epoch": 5.18, "learning_rate": 1.9818717470697718e-07, "loss": 0.0117, "step": 13137 }, { "epoch": 5.18, "learning_rate": 1.9816164267047994e-07, "loss": 0.0116, "step": 13138 }, { "epoch": 5.18, "learning_rate": 1.9813611066394513e-07, "loss": 0.0156, "step": 13139 }, { "epoch": 5.18, "learning_rate": 1.9811057868778884e-07, "loss": 0.0113, "step": 13140 }, { "epoch": 5.18, "learning_rate": 1.980850467424273e-07, "loss": 0.0184, "step": 13141 }, { "epoch": 5.18, "learning_rate": 1.9805951482827654e-07, "loss": 0.0138, "step": 13142 }, { "epoch": 5.18, "learning_rate": 1.9803398294575278e-07, "loss": 0.0108, "step": 13143 }, { "epoch": 5.18, "learning_rate": 1.9800845109527207e-07, "loss": 0.0133, "step": 13144 }, { "epoch": 5.18, "learning_rate": 1.9798291927725062e-07, "loss": 0.0132, "step": 13145 }, { "epoch": 5.18, "learning_rate": 1.9795738749210452e-07, "loss": 0.0161, "step": 13146 }, { "epoch": 5.18, "learning_rate": 1.979318557402499e-07, "loss": 0.0123, "step": 13147 }, { "epoch": 5.18, "learning_rate": 1.9790632402210288e-07, "loss": 0.016, "step": 13148 }, { "epoch": 5.18, "learning_rate": 1.978807923380795e-07, "loss": 0.0148, "step": 13149 }, { "epoch": 5.18, "learning_rate": 1.978552606885961e-07, "loss": 0.0131, "step": 13150 }, { "epoch": 5.18, "learning_rate": 1.9782972907406868e-07, "loss": 0.0198, "step": 13151 }, { "epoch": 5.18, "learning_rate": 1.9780419749491337e-07, "loss": 0.0139, "step": 13152 }, { "epoch": 5.18, "learning_rate": 1.9777866595154621e-07, "loss": 0.0128, "step": 13153 }, { "epoch": 5.18, "learning_rate": 1.9775313444438355e-07, "loss": 0.0119, "step": 13154 }, { "epoch": 5.18, "learning_rate": 1.9772760297384134e-07, "loss": 0.0149, "step": 13155 }, { "epoch": 5.18, "learning_rate": 1.9770207154033576e-07, "loss": 0.0131, "step": 13156 }, { "epoch": 5.19, "learning_rate": 1.9767654014428293e-07, "loss": 0.013, "step": 13157 }, { "epoch": 5.19, "learning_rate": 1.9765100878609886e-07, "loss": 0.0123, "step": 13158 }, { "epoch": 5.19, "learning_rate": 1.9762547746619987e-07, "loss": 0.0116, "step": 13159 }, { "epoch": 5.19, "learning_rate": 1.9759994618500199e-07, "loss": 0.0135, "step": 13160 }, { "epoch": 5.19, "learning_rate": 1.9757441494292135e-07, "loss": 0.0143, "step": 13161 }, { "epoch": 5.19, "learning_rate": 1.9754888374037402e-07, "loss": 0.0098, "step": 13162 }, { "epoch": 5.19, "learning_rate": 1.9752335257777618e-07, "loss": 0.0135, "step": 13163 }, { "epoch": 5.19, "learning_rate": 1.9749782145554395e-07, "loss": 0.0107, "step": 13164 }, { "epoch": 5.19, "learning_rate": 1.9747229037409337e-07, "loss": 0.015, "step": 13165 }, { "epoch": 5.19, "learning_rate": 1.9744675933384065e-07, "loss": 0.0136, "step": 13166 }, { "epoch": 5.19, "learning_rate": 1.9742122833520184e-07, "loss": 0.013, "step": 13167 }, { "epoch": 5.19, "learning_rate": 1.9739569737859314e-07, "loss": 0.0129, "step": 13168 }, { "epoch": 5.19, "learning_rate": 1.9737016646443062e-07, "loss": 0.0134, "step": 13169 }, { "epoch": 5.19, "learning_rate": 1.9734463559313037e-07, "loss": 0.0145, "step": 13170 }, { "epoch": 5.19, "learning_rate": 1.9731910476510845e-07, "loss": 0.0134, "step": 13171 }, { "epoch": 5.19, "learning_rate": 1.9729357398078112e-07, "loss": 0.0175, "step": 13172 }, { "epoch": 5.19, "learning_rate": 1.9726804324056444e-07, "loss": 0.0173, "step": 13173 }, { "epoch": 5.19, "learning_rate": 1.972425125448745e-07, "loss": 0.0104, "step": 13174 }, { "epoch": 5.19, "learning_rate": 1.9721698189412729e-07, "loss": 0.0137, "step": 13175 }, { "epoch": 5.19, "learning_rate": 1.9719145128873916e-07, "loss": 0.0135, "step": 13176 }, { "epoch": 5.19, "learning_rate": 1.9716592072912613e-07, "loss": 0.014, "step": 13177 }, { "epoch": 5.19, "learning_rate": 1.971403902157043e-07, "loss": 0.0164, "step": 13178 }, { "epoch": 5.19, "learning_rate": 1.971148597488897e-07, "loss": 0.0139, "step": 13179 }, { "epoch": 5.19, "learning_rate": 1.9708932932909843e-07, "loss": 0.0122, "step": 13180 }, { "epoch": 5.19, "learning_rate": 1.9706379895674677e-07, "loss": 0.0123, "step": 13181 }, { "epoch": 5.2, "learning_rate": 1.9703826863225074e-07, "loss": 0.0154, "step": 13182 }, { "epoch": 5.2, "learning_rate": 1.970127383560264e-07, "loss": 0.0122, "step": 13183 }, { "epoch": 5.2, "learning_rate": 1.9698720812848984e-07, "loss": 0.0163, "step": 13184 }, { "epoch": 5.2, "learning_rate": 1.9696167795005728e-07, "loss": 0.0118, "step": 13185 }, { "epoch": 5.2, "learning_rate": 1.9693614782114475e-07, "loss": 0.0142, "step": 13186 }, { "epoch": 5.2, "learning_rate": 1.9691061774216828e-07, "loss": 0.0141, "step": 13187 }, { "epoch": 5.2, "learning_rate": 1.968850877135441e-07, "loss": 0.0153, "step": 13188 }, { "epoch": 5.2, "learning_rate": 1.9685955773568823e-07, "loss": 0.0125, "step": 13189 }, { "epoch": 5.2, "learning_rate": 1.968340278090168e-07, "loss": 0.0143, "step": 13190 }, { "epoch": 5.2, "learning_rate": 1.9680849793394597e-07, "loss": 0.0104, "step": 13191 }, { "epoch": 5.2, "learning_rate": 1.967829681108917e-07, "loss": 0.0114, "step": 13192 }, { "epoch": 5.2, "learning_rate": 1.9675743834027012e-07, "loss": 0.0147, "step": 13193 }, { "epoch": 5.2, "learning_rate": 1.9673190862249743e-07, "loss": 0.0139, "step": 13194 }, { "epoch": 5.2, "learning_rate": 1.967063789579897e-07, "loss": 0.0195, "step": 13195 }, { "epoch": 5.2, "learning_rate": 1.9668084934716296e-07, "loss": 0.01, "step": 13196 }, { "epoch": 5.2, "learning_rate": 1.9665531979043333e-07, "loss": 0.0122, "step": 13197 }, { "epoch": 5.2, "learning_rate": 1.9662979028821682e-07, "loss": 0.012, "step": 13198 }, { "epoch": 5.2, "learning_rate": 1.966042608409297e-07, "loss": 0.0143, "step": 13199 }, { "epoch": 5.2, "learning_rate": 1.9657873144898795e-07, "loss": 0.0118, "step": 13200 }, { "epoch": 5.2, "learning_rate": 1.965532021128077e-07, "loss": 0.0139, "step": 13201 }, { "epoch": 5.2, "learning_rate": 1.965276728328049e-07, "loss": 0.0149, "step": 13202 }, { "epoch": 5.2, "learning_rate": 1.9650214360939588e-07, "loss": 0.0136, "step": 13203 }, { "epoch": 5.2, "learning_rate": 1.9647661444299659e-07, "loss": 0.0155, "step": 13204 }, { "epoch": 5.2, "learning_rate": 1.9645108533402314e-07, "loss": 0.0123, "step": 13205 }, { "epoch": 5.2, "learning_rate": 1.9642555628289158e-07, "loss": 0.0133, "step": 13206 }, { "epoch": 5.2, "learning_rate": 1.96400027290018e-07, "loss": 0.0164, "step": 13207 }, { "epoch": 5.21, "learning_rate": 1.9637449835581852e-07, "loss": 0.0138, "step": 13208 }, { "epoch": 5.21, "learning_rate": 1.9634896948070922e-07, "loss": 0.0141, "step": 13209 }, { "epoch": 5.21, "learning_rate": 1.9632344066510614e-07, "loss": 0.0141, "step": 13210 }, { "epoch": 5.21, "learning_rate": 1.9629791190942534e-07, "loss": 0.0105, "step": 13211 }, { "epoch": 5.21, "learning_rate": 1.9627238321408306e-07, "loss": 0.014, "step": 13212 }, { "epoch": 5.21, "learning_rate": 1.9624685457949523e-07, "loss": 0.0137, "step": 13213 }, { "epoch": 5.21, "learning_rate": 1.9622132600607798e-07, "loss": 0.0125, "step": 13214 }, { "epoch": 5.21, "learning_rate": 1.9619579749424727e-07, "loss": 0.0131, "step": 13215 }, { "epoch": 5.21, "learning_rate": 1.9617026904441937e-07, "loss": 0.0159, "step": 13216 }, { "epoch": 5.21, "learning_rate": 1.961447406570103e-07, "loss": 0.0134, "step": 13217 }, { "epoch": 5.21, "learning_rate": 1.9611921233243604e-07, "loss": 0.0167, "step": 13218 }, { "epoch": 5.21, "learning_rate": 1.9609368407111276e-07, "loss": 0.0111, "step": 13219 }, { "epoch": 5.21, "learning_rate": 1.9606815587345636e-07, "loss": 0.0115, "step": 13220 }, { "epoch": 5.21, "learning_rate": 1.9604262773988317e-07, "loss": 0.0119, "step": 13221 }, { "epoch": 5.21, "learning_rate": 1.9601709967080916e-07, "loss": 0.0127, "step": 13222 }, { "epoch": 5.21, "learning_rate": 1.9599157166665032e-07, "loss": 0.0141, "step": 13223 }, { "epoch": 5.21, "learning_rate": 1.9596604372782271e-07, "loss": 0.0129, "step": 13224 }, { "epoch": 5.21, "learning_rate": 1.9594051585474254e-07, "loss": 0.0106, "step": 13225 }, { "epoch": 5.21, "learning_rate": 1.9591498804782578e-07, "loss": 0.0121, "step": 13226 }, { "epoch": 5.21, "learning_rate": 1.958894603074885e-07, "loss": 0.0121, "step": 13227 }, { "epoch": 5.21, "learning_rate": 1.958639326341468e-07, "loss": 0.014, "step": 13228 }, { "epoch": 5.21, "learning_rate": 1.9583840502821665e-07, "loss": 0.0138, "step": 13229 }, { "epoch": 5.21, "learning_rate": 1.958128774901142e-07, "loss": 0.0137, "step": 13230 }, { "epoch": 5.21, "learning_rate": 1.9578735002025553e-07, "loss": 0.0149, "step": 13231 }, { "epoch": 5.21, "learning_rate": 1.9576182261905662e-07, "loss": 0.0131, "step": 13232 }, { "epoch": 5.22, "learning_rate": 1.9573629528693348e-07, "loss": 0.0133, "step": 13233 }, { "epoch": 5.22, "learning_rate": 1.9571076802430233e-07, "loss": 0.0105, "step": 13234 }, { "epoch": 5.22, "learning_rate": 1.9568524083157914e-07, "loss": 0.0126, "step": 13235 }, { "epoch": 5.22, "learning_rate": 1.9565971370917996e-07, "loss": 0.0131, "step": 13236 }, { "epoch": 5.22, "learning_rate": 1.9563418665752087e-07, "loss": 0.0141, "step": 13237 }, { "epoch": 5.22, "learning_rate": 1.956086596770178e-07, "loss": 0.0114, "step": 13238 }, { "epoch": 5.22, "learning_rate": 1.95583132768087e-07, "loss": 0.0112, "step": 13239 }, { "epoch": 5.22, "learning_rate": 1.955576059311444e-07, "loss": 0.0114, "step": 13240 }, { "epoch": 5.22, "learning_rate": 1.955320791666061e-07, "loss": 0.0128, "step": 13241 }, { "epoch": 5.22, "learning_rate": 1.9550655247488804e-07, "loss": 0.0123, "step": 13242 }, { "epoch": 5.22, "learning_rate": 1.954810258564064e-07, "loss": 0.0167, "step": 13243 }, { "epoch": 5.22, "learning_rate": 1.954554993115772e-07, "loss": 0.0132, "step": 13244 }, { "epoch": 5.22, "learning_rate": 1.9542997284081641e-07, "loss": 0.0138, "step": 13245 }, { "epoch": 5.22, "learning_rate": 1.9540444644454015e-07, "loss": 0.0133, "step": 13246 }, { "epoch": 5.22, "learning_rate": 1.9537892012316435e-07, "loss": 0.0148, "step": 13247 }, { "epoch": 5.22, "learning_rate": 1.9535339387710522e-07, "loss": 0.0123, "step": 13248 }, { "epoch": 5.22, "learning_rate": 1.9532786770677867e-07, "loss": 0.0121, "step": 13249 }, { "epoch": 5.22, "learning_rate": 1.953023416126008e-07, "loss": 0.0098, "step": 13250 }, { "epoch": 5.22, "learning_rate": 1.9527681559498757e-07, "loss": 0.0116, "step": 13251 }, { "epoch": 5.22, "learning_rate": 1.952512896543551e-07, "loss": 0.0125, "step": 13252 }, { "epoch": 5.22, "learning_rate": 1.9522576379111945e-07, "loss": 0.0101, "step": 13253 }, { "epoch": 5.22, "learning_rate": 1.9520023800569653e-07, "loss": 0.0147, "step": 13254 }, { "epoch": 5.22, "learning_rate": 1.9517471229850236e-07, "loss": 0.0157, "step": 13255 }, { "epoch": 5.22, "learning_rate": 1.9514918666995315e-07, "loss": 0.0114, "step": 13256 }, { "epoch": 5.22, "learning_rate": 1.9512366112046482e-07, "loss": 0.0107, "step": 13257 }, { "epoch": 5.23, "learning_rate": 1.950981356504534e-07, "loss": 0.0176, "step": 13258 }, { "epoch": 5.23, "learning_rate": 1.9507261026033492e-07, "loss": 0.0119, "step": 13259 }, { "epoch": 5.23, "learning_rate": 1.9504708495052534e-07, "loss": 0.0131, "step": 13260 }, { "epoch": 5.23, "learning_rate": 1.9502155972144078e-07, "loss": 0.0154, "step": 13261 }, { "epoch": 5.23, "learning_rate": 1.9499603457349726e-07, "loss": 0.0101, "step": 13262 }, { "epoch": 5.23, "learning_rate": 1.9497050950711082e-07, "loss": 0.0155, "step": 13263 }, { "epoch": 5.23, "learning_rate": 1.9494498452269728e-07, "loss": 0.0108, "step": 13264 }, { "epoch": 5.23, "learning_rate": 1.9491945962067292e-07, "loss": 0.0191, "step": 13265 }, { "epoch": 5.23, "learning_rate": 1.9489393480145362e-07, "loss": 0.0159, "step": 13266 }, { "epoch": 5.23, "learning_rate": 1.9486841006545545e-07, "loss": 0.0118, "step": 13267 }, { "epoch": 5.23, "learning_rate": 1.948428854130944e-07, "loss": 0.0132, "step": 13268 }, { "epoch": 5.23, "learning_rate": 1.948173608447864e-07, "loss": 0.0143, "step": 13269 }, { "epoch": 5.23, "learning_rate": 1.9479183636094758e-07, "loss": 0.0136, "step": 13270 }, { "epoch": 5.23, "learning_rate": 1.9476631196199393e-07, "loss": 0.0153, "step": 13271 }, { "epoch": 5.23, "learning_rate": 1.9474078764834147e-07, "loss": 0.0122, "step": 13272 }, { "epoch": 5.23, "learning_rate": 1.9471526342040612e-07, "loss": 0.0118, "step": 13273 }, { "epoch": 5.23, "learning_rate": 1.94689739278604e-07, "loss": 0.0155, "step": 13274 }, { "epoch": 5.23, "learning_rate": 1.9466421522335104e-07, "loss": 0.0156, "step": 13275 }, { "epoch": 5.23, "learning_rate": 1.9463869125506326e-07, "loss": 0.0116, "step": 13276 }, { "epoch": 5.23, "learning_rate": 1.9461316737415668e-07, "loss": 0.0133, "step": 13277 }, { "epoch": 5.23, "learning_rate": 1.9458764358104718e-07, "loss": 0.0149, "step": 13278 }, { "epoch": 5.23, "learning_rate": 1.9456211987615098e-07, "loss": 0.0126, "step": 13279 }, { "epoch": 5.23, "learning_rate": 1.9453659625988393e-07, "loss": 0.0134, "step": 13280 }, { "epoch": 5.23, "learning_rate": 1.9451107273266207e-07, "loss": 0.0143, "step": 13281 }, { "epoch": 5.23, "learning_rate": 1.944855492949013e-07, "loss": 0.0129, "step": 13282 }, { "epoch": 5.23, "learning_rate": 1.9446002594701777e-07, "loss": 0.0114, "step": 13283 }, { "epoch": 5.24, "learning_rate": 1.944345026894274e-07, "loss": 0.011, "step": 13284 }, { "epoch": 5.24, "learning_rate": 1.9440897952254617e-07, "loss": 0.012, "step": 13285 }, { "epoch": 5.24, "learning_rate": 1.943834564467901e-07, "loss": 0.0124, "step": 13286 }, { "epoch": 5.24, "learning_rate": 1.9435793346257506e-07, "loss": 0.0164, "step": 13287 }, { "epoch": 5.24, "learning_rate": 1.943324105703172e-07, "loss": 0.0158, "step": 13288 }, { "epoch": 5.24, "learning_rate": 1.9430688777043246e-07, "loss": 0.0153, "step": 13289 }, { "epoch": 5.24, "learning_rate": 1.942813650633368e-07, "loss": 0.014, "step": 13290 }, { "epoch": 5.24, "learning_rate": 1.942558424494461e-07, "loss": 0.0117, "step": 13291 }, { "epoch": 5.24, "learning_rate": 1.9423031992917654e-07, "loss": 0.0136, "step": 13292 }, { "epoch": 5.24, "learning_rate": 1.942047975029439e-07, "loss": 0.0111, "step": 13293 }, { "epoch": 5.24, "learning_rate": 1.9417927517116434e-07, "loss": 0.0117, "step": 13294 }, { "epoch": 5.24, "learning_rate": 1.941537529342537e-07, "loss": 0.014, "step": 13295 }, { "epoch": 5.24, "learning_rate": 1.9412823079262806e-07, "loss": 0.0145, "step": 13296 }, { "epoch": 5.24, "learning_rate": 1.941027087467033e-07, "loss": 0.0156, "step": 13297 }, { "epoch": 5.24, "learning_rate": 1.9407718679689544e-07, "loss": 0.0108, "step": 13298 }, { "epoch": 5.24, "learning_rate": 1.9405166494362043e-07, "loss": 0.0127, "step": 13299 }, { "epoch": 5.24, "learning_rate": 1.9402614318729416e-07, "loss": 0.0124, "step": 13300 }, { "epoch": 5.24, "learning_rate": 1.9400062152833274e-07, "loss": 0.0137, "step": 13301 }, { "epoch": 5.24, "learning_rate": 1.939750999671521e-07, "loss": 0.0131, "step": 13302 }, { "epoch": 5.24, "learning_rate": 1.9394957850416816e-07, "loss": 0.0116, "step": 13303 }, { "epoch": 5.24, "learning_rate": 1.939240571397968e-07, "loss": 0.0104, "step": 13304 }, { "epoch": 5.24, "learning_rate": 1.938985358744542e-07, "loss": 0.0145, "step": 13305 }, { "epoch": 5.24, "learning_rate": 1.9387301470855614e-07, "loss": 0.0143, "step": 13306 }, { "epoch": 5.24, "learning_rate": 1.9384749364251866e-07, "loss": 0.0133, "step": 13307 }, { "epoch": 5.24, "learning_rate": 1.9382197267675768e-07, "loss": 0.0135, "step": 13308 }, { "epoch": 5.25, "learning_rate": 1.9379645181168906e-07, "loss": 0.0133, "step": 13309 }, { "epoch": 5.25, "learning_rate": 1.9377093104772894e-07, "loss": 0.0164, "step": 13310 }, { "epoch": 5.25, "learning_rate": 1.9374541038529318e-07, "loss": 0.0168, "step": 13311 }, { "epoch": 5.25, "learning_rate": 1.9371988982479773e-07, "loss": 0.0139, "step": 13312 }, { "epoch": 5.25, "learning_rate": 1.9369436936665846e-07, "loss": 0.0135, "step": 13313 }, { "epoch": 5.25, "learning_rate": 1.9366884901129145e-07, "loss": 0.0147, "step": 13314 }, { "epoch": 5.25, "learning_rate": 1.9364332875911252e-07, "loss": 0.0119, "step": 13315 }, { "epoch": 5.25, "learning_rate": 1.9361780861053775e-07, "loss": 0.0127, "step": 13316 }, { "epoch": 5.25, "learning_rate": 1.9359228856598298e-07, "loss": 0.015, "step": 13317 }, { "epoch": 5.25, "learning_rate": 1.9356676862586413e-07, "loss": 0.0139, "step": 13318 }, { "epoch": 5.25, "learning_rate": 1.9354124879059718e-07, "loss": 0.0112, "step": 13319 }, { "epoch": 5.25, "learning_rate": 1.935157290605981e-07, "loss": 0.0114, "step": 13320 }, { "epoch": 5.25, "learning_rate": 1.9349020943628275e-07, "loss": 0.0136, "step": 13321 }, { "epoch": 5.25, "learning_rate": 1.9346468991806704e-07, "loss": 0.0143, "step": 13322 }, { "epoch": 5.25, "learning_rate": 1.9343917050636701e-07, "loss": 0.0126, "step": 13323 }, { "epoch": 5.25, "learning_rate": 1.9341365120159857e-07, "loss": 0.0134, "step": 13324 }, { "epoch": 5.25, "learning_rate": 1.9338813200417755e-07, "loss": 0.0122, "step": 13325 }, { "epoch": 5.25, "learning_rate": 1.9336261291451996e-07, "loss": 0.0144, "step": 13326 }, { "epoch": 5.25, "learning_rate": 1.933370939330416e-07, "loss": 0.0124, "step": 13327 }, { "epoch": 5.25, "learning_rate": 1.9331157506015858e-07, "loss": 0.0148, "step": 13328 }, { "epoch": 5.25, "learning_rate": 1.9328605629628672e-07, "loss": 0.0138, "step": 13329 }, { "epoch": 5.25, "learning_rate": 1.932605376418419e-07, "loss": 0.0157, "step": 13330 }, { "epoch": 5.25, "learning_rate": 1.9323501909724003e-07, "loss": 0.0143, "step": 13331 }, { "epoch": 5.25, "learning_rate": 1.9320950066289715e-07, "loss": 0.0119, "step": 13332 }, { "epoch": 5.25, "learning_rate": 1.9318398233922908e-07, "loss": 0.0142, "step": 13333 }, { "epoch": 5.26, "learning_rate": 1.9315846412665173e-07, "loss": 0.014, "step": 13334 }, { "epoch": 5.26, "learning_rate": 1.9313294602558098e-07, "loss": 0.0154, "step": 13335 }, { "epoch": 5.26, "learning_rate": 1.9310742803643282e-07, "loss": 0.0091, "step": 13336 }, { "epoch": 5.26, "learning_rate": 1.9308191015962312e-07, "loss": 0.0134, "step": 13337 }, { "epoch": 5.26, "learning_rate": 1.930563923955677e-07, "loss": 0.0128, "step": 13338 }, { "epoch": 5.26, "learning_rate": 1.930308747446826e-07, "loss": 0.0121, "step": 13339 }, { "epoch": 5.26, "learning_rate": 1.9300535720738357e-07, "loss": 0.013, "step": 13340 }, { "epoch": 5.26, "learning_rate": 1.9297983978408667e-07, "loss": 0.0124, "step": 13341 }, { "epoch": 5.26, "learning_rate": 1.929543224752077e-07, "loss": 0.0154, "step": 13342 }, { "epoch": 5.26, "learning_rate": 1.9292880528116256e-07, "loss": 0.0135, "step": 13343 }, { "epoch": 5.26, "learning_rate": 1.9290328820236705e-07, "loss": 0.0145, "step": 13344 }, { "epoch": 5.26, "learning_rate": 1.9287777123923726e-07, "loss": 0.0146, "step": 13345 }, { "epoch": 5.26, "learning_rate": 1.9285225439218897e-07, "loss": 0.013, "step": 13346 }, { "epoch": 5.26, "learning_rate": 1.9282673766163806e-07, "loss": 0.0135, "step": 13347 }, { "epoch": 5.26, "learning_rate": 1.9280122104800042e-07, "loss": 0.0116, "step": 13348 }, { "epoch": 5.26, "learning_rate": 1.9277570455169187e-07, "loss": 0.0144, "step": 13349 }, { "epoch": 5.26, "learning_rate": 1.9275018817312845e-07, "loss": 0.0143, "step": 13350 }, { "epoch": 5.26, "learning_rate": 1.9272467191272594e-07, "loss": 0.0104, "step": 13351 }, { "epoch": 5.26, "learning_rate": 1.926991557709002e-07, "loss": 0.017, "step": 13352 }, { "epoch": 5.26, "learning_rate": 1.9267363974806706e-07, "loss": 0.0167, "step": 13353 }, { "epoch": 5.26, "learning_rate": 1.926481238446425e-07, "loss": 0.0128, "step": 13354 }, { "epoch": 5.26, "learning_rate": 1.926226080610424e-07, "loss": 0.0111, "step": 13355 }, { "epoch": 5.26, "learning_rate": 1.9259709239768255e-07, "loss": 0.0128, "step": 13356 }, { "epoch": 5.26, "learning_rate": 1.9257157685497884e-07, "loss": 0.0115, "step": 13357 }, { "epoch": 5.26, "learning_rate": 1.925460614333471e-07, "loss": 0.0146, "step": 13358 }, { "epoch": 5.26, "learning_rate": 1.925205461332033e-07, "loss": 0.0109, "step": 13359 }, { "epoch": 5.27, "learning_rate": 1.924950309549632e-07, "loss": 0.0114, "step": 13360 }, { "epoch": 5.27, "learning_rate": 1.9246951589904267e-07, "loss": 0.0129, "step": 13361 }, { "epoch": 5.27, "learning_rate": 1.9244400096585756e-07, "loss": 0.0104, "step": 13362 }, { "epoch": 5.27, "learning_rate": 1.924184861558238e-07, "loss": 0.01, "step": 13363 }, { "epoch": 5.27, "learning_rate": 1.923929714693572e-07, "loss": 0.0123, "step": 13364 }, { "epoch": 5.27, "learning_rate": 1.923674569068736e-07, "loss": 0.0123, "step": 13365 }, { "epoch": 5.27, "learning_rate": 1.9234194246878884e-07, "loss": 0.0146, "step": 13366 }, { "epoch": 5.27, "learning_rate": 1.923164281555187e-07, "loss": 0.0135, "step": 13367 }, { "epoch": 5.27, "learning_rate": 1.9229091396747922e-07, "loss": 0.0152, "step": 13368 }, { "epoch": 5.27, "learning_rate": 1.9226539990508607e-07, "loss": 0.0127, "step": 13369 }, { "epoch": 5.27, "learning_rate": 1.9223988596875517e-07, "loss": 0.0161, "step": 13370 }, { "epoch": 5.27, "learning_rate": 1.9221437215890227e-07, "loss": 0.0139, "step": 13371 }, { "epoch": 5.27, "learning_rate": 1.9218885847594336e-07, "loss": 0.0133, "step": 13372 }, { "epoch": 5.27, "learning_rate": 1.9216334492029415e-07, "loss": 0.0149, "step": 13373 }, { "epoch": 5.27, "learning_rate": 1.9213783149237053e-07, "loss": 0.0128, "step": 13374 }, { "epoch": 5.27, "learning_rate": 1.921123181925882e-07, "loss": 0.014, "step": 13375 }, { "epoch": 5.27, "learning_rate": 1.920868050213632e-07, "loss": 0.013, "step": 13376 }, { "epoch": 5.27, "learning_rate": 1.9206129197911126e-07, "loss": 0.0126, "step": 13377 }, { "epoch": 5.27, "learning_rate": 1.9203577906624817e-07, "loss": 0.0123, "step": 13378 }, { "epoch": 5.27, "learning_rate": 1.920102662831898e-07, "loss": 0.0152, "step": 13379 }, { "epoch": 5.27, "learning_rate": 1.919847536303519e-07, "loss": 0.0137, "step": 13380 }, { "epoch": 5.27, "learning_rate": 1.9195924110815036e-07, "loss": 0.0117, "step": 13381 }, { "epoch": 5.27, "learning_rate": 1.91933728717001e-07, "loss": 0.0131, "step": 13382 }, { "epoch": 5.27, "learning_rate": 1.9190821645731954e-07, "loss": 0.0125, "step": 13383 }, { "epoch": 5.27, "learning_rate": 1.9188270432952185e-07, "loss": 0.0134, "step": 13384 }, { "epoch": 5.28, "learning_rate": 1.9185719233402377e-07, "loss": 0.0138, "step": 13385 }, { "epoch": 5.28, "learning_rate": 1.918316804712411e-07, "loss": 0.0146, "step": 13386 }, { "epoch": 5.28, "learning_rate": 1.9180616874158964e-07, "loss": 0.016, "step": 13387 }, { "epoch": 5.28, "learning_rate": 1.9178065714548515e-07, "loss": 0.0128, "step": 13388 }, { "epoch": 5.28, "learning_rate": 1.917551456833434e-07, "loss": 0.0111, "step": 13389 }, { "epoch": 5.28, "learning_rate": 1.9172963435558028e-07, "loss": 0.0117, "step": 13390 }, { "epoch": 5.28, "learning_rate": 1.917041231626116e-07, "loss": 0.0139, "step": 13391 }, { "epoch": 5.28, "learning_rate": 1.9167861210485307e-07, "loss": 0.0141, "step": 13392 }, { "epoch": 5.28, "learning_rate": 1.9165310118272046e-07, "loss": 0.0131, "step": 13393 }, { "epoch": 5.28, "learning_rate": 1.9162759039662968e-07, "loss": 0.014, "step": 13394 }, { "epoch": 5.28, "learning_rate": 1.9160207974699647e-07, "loss": 0.0103, "step": 13395 }, { "epoch": 5.28, "learning_rate": 1.9157656923423658e-07, "loss": 0.0124, "step": 13396 }, { "epoch": 5.28, "learning_rate": 1.915510588587658e-07, "loss": 0.0141, "step": 13397 }, { "epoch": 5.28, "learning_rate": 1.9152554862099987e-07, "loss": 0.0123, "step": 13398 }, { "epoch": 5.28, "learning_rate": 1.9150003852135467e-07, "loss": 0.0163, "step": 13399 }, { "epoch": 5.28, "learning_rate": 1.9147452856024593e-07, "loss": 0.0126, "step": 13400 }, { "epoch": 5.28, "learning_rate": 1.9144901873808942e-07, "loss": 0.0123, "step": 13401 }, { "epoch": 5.28, "learning_rate": 1.9142350905530088e-07, "loss": 0.0137, "step": 13402 }, { "epoch": 5.28, "learning_rate": 1.9139799951229614e-07, "loss": 0.0134, "step": 13403 }, { "epoch": 5.28, "learning_rate": 1.9137249010949093e-07, "loss": 0.0105, "step": 13404 }, { "epoch": 5.28, "learning_rate": 1.9134698084730104e-07, "loss": 0.0141, "step": 13405 }, { "epoch": 5.28, "learning_rate": 1.913214717261421e-07, "loss": 0.0139, "step": 13406 }, { "epoch": 5.28, "learning_rate": 1.9129596274643004e-07, "loss": 0.0154, "step": 13407 }, { "epoch": 5.28, "learning_rate": 1.912704539085806e-07, "loss": 0.0128, "step": 13408 }, { "epoch": 5.28, "learning_rate": 1.9124494521300952e-07, "loss": 0.0161, "step": 13409 }, { "epoch": 5.28, "learning_rate": 1.9121943666013246e-07, "loss": 0.0121, "step": 13410 }, { "epoch": 5.29, "learning_rate": 1.911939282503652e-07, "loss": 0.0102, "step": 13411 }, { "epoch": 5.29, "learning_rate": 1.9116841998412358e-07, "loss": 0.012, "step": 13412 }, { "epoch": 5.29, "learning_rate": 1.911429118618233e-07, "loss": 0.0164, "step": 13413 }, { "epoch": 5.29, "learning_rate": 1.9111740388388013e-07, "loss": 0.011, "step": 13414 }, { "epoch": 5.29, "learning_rate": 1.9109189605070966e-07, "loss": 0.0168, "step": 13415 }, { "epoch": 5.29, "learning_rate": 1.9106638836272783e-07, "loss": 0.0112, "step": 13416 }, { "epoch": 5.29, "learning_rate": 1.910408808203503e-07, "loss": 0.0141, "step": 13417 }, { "epoch": 5.29, "learning_rate": 1.9101537342399282e-07, "loss": 0.0126, "step": 13418 }, { "epoch": 5.29, "learning_rate": 1.9098986617407108e-07, "loss": 0.0125, "step": 13419 }, { "epoch": 5.29, "learning_rate": 1.909643590710007e-07, "loss": 0.0153, "step": 13420 }, { "epoch": 5.29, "learning_rate": 1.9093885211519768e-07, "loss": 0.013, "step": 13421 }, { "epoch": 5.29, "learning_rate": 1.9091334530707757e-07, "loss": 0.0117, "step": 13422 }, { "epoch": 5.29, "learning_rate": 1.9088783864705612e-07, "loss": 0.0131, "step": 13423 }, { "epoch": 5.29, "learning_rate": 1.90862332135549e-07, "loss": 0.0096, "step": 13424 }, { "epoch": 5.29, "learning_rate": 1.9083682577297204e-07, "loss": 0.0151, "step": 13425 }, { "epoch": 5.29, "learning_rate": 1.908113195597409e-07, "loss": 0.0139, "step": 13426 }, { "epoch": 5.29, "learning_rate": 1.9078581349627127e-07, "loss": 0.015, "step": 13427 }, { "epoch": 5.29, "learning_rate": 1.9076030758297887e-07, "loss": 0.0113, "step": 13428 }, { "epoch": 5.29, "learning_rate": 1.9073480182027935e-07, "loss": 0.0145, "step": 13429 }, { "epoch": 5.29, "learning_rate": 1.9070929620858857e-07, "loss": 0.0155, "step": 13430 }, { "epoch": 5.29, "learning_rate": 1.906837907483221e-07, "loss": 0.0118, "step": 13431 }, { "epoch": 5.29, "learning_rate": 1.906582854398957e-07, "loss": 0.0151, "step": 13432 }, { "epoch": 5.29, "learning_rate": 1.9063278028372497e-07, "loss": 0.0122, "step": 13433 }, { "epoch": 5.29, "learning_rate": 1.906072752802258e-07, "loss": 0.0125, "step": 13434 }, { "epoch": 5.29, "learning_rate": 1.905817704298137e-07, "loss": 0.0156, "step": 13435 }, { "epoch": 5.3, "learning_rate": 1.9055626573290448e-07, "loss": 0.0136, "step": 13436 }, { "epoch": 5.3, "learning_rate": 1.9053076118991377e-07, "loss": 0.0146, "step": 13437 }, { "epoch": 5.3, "learning_rate": 1.9050525680125717e-07, "loss": 0.0145, "step": 13438 }, { "epoch": 5.3, "learning_rate": 1.9047975256735053e-07, "loss": 0.0135, "step": 13439 }, { "epoch": 5.3, "learning_rate": 1.9045424848860947e-07, "loss": 0.0131, "step": 13440 }, { "epoch": 5.3, "learning_rate": 1.9042874456544964e-07, "loss": 0.0131, "step": 13441 }, { "epoch": 5.3, "learning_rate": 1.904032407982867e-07, "loss": 0.0124, "step": 13442 }, { "epoch": 5.3, "learning_rate": 1.9037773718753634e-07, "loss": 0.0126, "step": 13443 }, { "epoch": 5.3, "learning_rate": 1.9035223373361426e-07, "loss": 0.012, "step": 13444 }, { "epoch": 5.3, "learning_rate": 1.9032673043693613e-07, "loss": 0.0104, "step": 13445 }, { "epoch": 5.3, "learning_rate": 1.903012272979176e-07, "loss": 0.0176, "step": 13446 }, { "epoch": 5.3, "learning_rate": 1.9027572431697427e-07, "loss": 0.0125, "step": 13447 }, { "epoch": 5.3, "learning_rate": 1.902502214945219e-07, "loss": 0.0138, "step": 13448 }, { "epoch": 5.3, "learning_rate": 1.902247188309761e-07, "loss": 0.0162, "step": 13449 }, { "epoch": 5.3, "learning_rate": 1.901992163267525e-07, "loss": 0.0095, "step": 13450 }, { "epoch": 5.3, "learning_rate": 1.9017371398226674e-07, "loss": 0.0116, "step": 13451 }, { "epoch": 5.3, "learning_rate": 1.9014821179793456e-07, "loss": 0.0157, "step": 13452 }, { "epoch": 5.3, "learning_rate": 1.901227097741716e-07, "loss": 0.0148, "step": 13453 }, { "epoch": 5.3, "learning_rate": 1.9009720791139342e-07, "loss": 0.0113, "step": 13454 }, { "epoch": 5.3, "learning_rate": 1.900717062100156e-07, "loss": 0.012, "step": 13455 }, { "epoch": 5.3, "learning_rate": 1.9004620467045398e-07, "loss": 0.0152, "step": 13456 }, { "epoch": 5.3, "learning_rate": 1.900207032931241e-07, "loss": 0.0145, "step": 13457 }, { "epoch": 5.3, "learning_rate": 1.899952020784416e-07, "loss": 0.0174, "step": 13458 }, { "epoch": 5.3, "learning_rate": 1.899697010268221e-07, "loss": 0.0151, "step": 13459 }, { "epoch": 5.3, "learning_rate": 1.8994420013868114e-07, "loss": 0.0125, "step": 13460 }, { "epoch": 5.31, "learning_rate": 1.899186994144345e-07, "loss": 0.0135, "step": 13461 }, { "epoch": 5.31, "learning_rate": 1.8989319885449774e-07, "loss": 0.0119, "step": 13462 }, { "epoch": 5.31, "learning_rate": 1.898676984592865e-07, "loss": 0.0098, "step": 13463 }, { "epoch": 5.31, "learning_rate": 1.8984219822921632e-07, "loss": 0.0109, "step": 13464 }, { "epoch": 5.31, "learning_rate": 1.898166981647029e-07, "loss": 0.0126, "step": 13465 }, { "epoch": 5.31, "learning_rate": 1.8979119826616183e-07, "loss": 0.009, "step": 13466 }, { "epoch": 5.31, "learning_rate": 1.8976569853400872e-07, "loss": 0.0133, "step": 13467 }, { "epoch": 5.31, "learning_rate": 1.8974019896865916e-07, "loss": 0.0132, "step": 13468 }, { "epoch": 5.31, "learning_rate": 1.8971469957052875e-07, "loss": 0.0125, "step": 13469 }, { "epoch": 5.31, "learning_rate": 1.8968920034003313e-07, "loss": 0.0128, "step": 13470 }, { "epoch": 5.31, "learning_rate": 1.8966370127758786e-07, "loss": 0.0165, "step": 13471 }, { "epoch": 5.31, "learning_rate": 1.8963820238360858e-07, "loss": 0.0122, "step": 13472 }, { "epoch": 5.31, "learning_rate": 1.8961270365851073e-07, "loss": 0.0131, "step": 13473 }, { "epoch": 5.31, "learning_rate": 1.8958720510271015e-07, "loss": 0.0136, "step": 13474 }, { "epoch": 5.31, "learning_rate": 1.8956170671662228e-07, "loss": 0.0137, "step": 13475 }, { "epoch": 5.31, "learning_rate": 1.8953620850066275e-07, "loss": 0.0113, "step": 13476 }, { "epoch": 5.31, "learning_rate": 1.8951071045524708e-07, "loss": 0.0164, "step": 13477 }, { "epoch": 5.31, "learning_rate": 1.8948521258079083e-07, "loss": 0.0131, "step": 13478 }, { "epoch": 5.31, "learning_rate": 1.894597148777097e-07, "loss": 0.0106, "step": 13479 }, { "epoch": 5.31, "learning_rate": 1.894342173464192e-07, "loss": 0.0126, "step": 13480 }, { "epoch": 5.31, "learning_rate": 1.8940871998733491e-07, "loss": 0.0117, "step": 13481 }, { "epoch": 5.31, "learning_rate": 1.893832228008723e-07, "loss": 0.0145, "step": 13482 }, { "epoch": 5.31, "learning_rate": 1.8935772578744712e-07, "loss": 0.0125, "step": 13483 }, { "epoch": 5.31, "learning_rate": 1.893322289474748e-07, "loss": 0.0121, "step": 13484 }, { "epoch": 5.31, "learning_rate": 1.8930673228137097e-07, "loss": 0.0152, "step": 13485 }, { "epoch": 5.31, "learning_rate": 1.8928123578955117e-07, "loss": 0.0169, "step": 13486 }, { "epoch": 5.32, "learning_rate": 1.8925573947243084e-07, "loss": 0.0124, "step": 13487 }, { "epoch": 5.32, "learning_rate": 1.892302433304257e-07, "loss": 0.0114, "step": 13488 }, { "epoch": 5.32, "learning_rate": 1.8920474736395115e-07, "loss": 0.0131, "step": 13489 }, { "epoch": 5.32, "learning_rate": 1.891792515734229e-07, "loss": 0.014, "step": 13490 }, { "epoch": 5.32, "learning_rate": 1.8915375595925633e-07, "loss": 0.0132, "step": 13491 }, { "epoch": 5.32, "learning_rate": 1.891282605218671e-07, "loss": 0.0134, "step": 13492 }, { "epoch": 5.32, "learning_rate": 1.891027652616707e-07, "loss": 0.0134, "step": 13493 }, { "epoch": 5.32, "learning_rate": 1.890772701790827e-07, "loss": 0.0143, "step": 13494 }, { "epoch": 5.32, "learning_rate": 1.8905177527451845e-07, "loss": 0.0115, "step": 13495 }, { "epoch": 5.32, "learning_rate": 1.8902628054839374e-07, "loss": 0.0131, "step": 13496 }, { "epoch": 5.32, "learning_rate": 1.89000786001124e-07, "loss": 0.0139, "step": 13497 }, { "epoch": 5.32, "learning_rate": 1.889752916331247e-07, "loss": 0.0108, "step": 13498 }, { "epoch": 5.32, "learning_rate": 1.889497974448114e-07, "loss": 0.0093, "step": 13499 }, { "epoch": 5.32, "learning_rate": 1.8892430343659952e-07, "loss": 0.0094, "step": 13500 }, { "epoch": 5.32, "learning_rate": 1.8889880960890477e-07, "loss": 0.0182, "step": 13501 }, { "epoch": 5.32, "learning_rate": 1.888733159621425e-07, "loss": 0.0118, "step": 13502 }, { "epoch": 5.32, "learning_rate": 1.8884782249672832e-07, "loss": 0.0156, "step": 13503 }, { "epoch": 5.32, "learning_rate": 1.8882232921307757e-07, "loss": 0.014, "step": 13504 }, { "epoch": 5.32, "learning_rate": 1.8879683611160597e-07, "loss": 0.0115, "step": 13505 }, { "epoch": 5.32, "learning_rate": 1.887713431927289e-07, "loss": 0.0112, "step": 13506 }, { "epoch": 5.32, "learning_rate": 1.8874585045686186e-07, "loss": 0.0157, "step": 13507 }, { "epoch": 5.32, "learning_rate": 1.887203579044204e-07, "loss": 0.0138, "step": 13508 }, { "epoch": 5.32, "learning_rate": 1.8869486553581986e-07, "loss": 0.0123, "step": 13509 }, { "epoch": 5.32, "learning_rate": 1.886693733514759e-07, "loss": 0.0101, "step": 13510 }, { "epoch": 5.32, "learning_rate": 1.8864388135180388e-07, "loss": 0.0116, "step": 13511 }, { "epoch": 5.33, "learning_rate": 1.886183895372194e-07, "loss": 0.0136, "step": 13512 }, { "epoch": 5.33, "learning_rate": 1.8859289790813785e-07, "loss": 0.0143, "step": 13513 }, { "epoch": 5.33, "learning_rate": 1.885674064649747e-07, "loss": 0.0128, "step": 13514 }, { "epoch": 5.33, "learning_rate": 1.8854191520814551e-07, "loss": 0.0116, "step": 13515 }, { "epoch": 5.33, "learning_rate": 1.8851642413806565e-07, "loss": 0.012, "step": 13516 }, { "epoch": 5.33, "learning_rate": 1.8849093325515066e-07, "loss": 0.012, "step": 13517 }, { "epoch": 5.33, "learning_rate": 1.8846544255981586e-07, "loss": 0.0135, "step": 13518 }, { "epoch": 5.33, "learning_rate": 1.884399520524769e-07, "loss": 0.0102, "step": 13519 }, { "epoch": 5.33, "learning_rate": 1.8841446173354915e-07, "loss": 0.0135, "step": 13520 }, { "epoch": 5.33, "learning_rate": 1.8838897160344806e-07, "loss": 0.0151, "step": 13521 }, { "epoch": 5.33, "learning_rate": 1.88363481662589e-07, "loss": 0.0114, "step": 13522 }, { "epoch": 5.33, "learning_rate": 1.8833799191138762e-07, "loss": 0.0143, "step": 13523 }, { "epoch": 5.33, "learning_rate": 1.883125023502592e-07, "loss": 0.0142, "step": 13524 }, { "epoch": 5.33, "learning_rate": 1.882870129796193e-07, "loss": 0.0142, "step": 13525 }, { "epoch": 5.33, "learning_rate": 1.8826152379988321e-07, "loss": 0.015, "step": 13526 }, { "epoch": 5.33, "learning_rate": 1.882360348114664e-07, "loss": 0.0148, "step": 13527 }, { "epoch": 5.33, "learning_rate": 1.8821054601478442e-07, "loss": 0.0153, "step": 13528 }, { "epoch": 5.33, "learning_rate": 1.881850574102526e-07, "loss": 0.0122, "step": 13529 }, { "epoch": 5.33, "learning_rate": 1.881595689982864e-07, "loss": 0.013, "step": 13530 }, { "epoch": 5.33, "learning_rate": 1.881340807793012e-07, "loss": 0.0136, "step": 13531 }, { "epoch": 5.33, "learning_rate": 1.881085927537125e-07, "loss": 0.0121, "step": 13532 }, { "epoch": 5.33, "learning_rate": 1.8808310492193563e-07, "loss": 0.018, "step": 13533 }, { "epoch": 5.33, "learning_rate": 1.8805761728438603e-07, "loss": 0.0146, "step": 13534 }, { "epoch": 5.33, "learning_rate": 1.880321298414791e-07, "loss": 0.0135, "step": 13535 }, { "epoch": 5.33, "learning_rate": 1.8800664259363033e-07, "loss": 0.0108, "step": 13536 }, { "epoch": 5.34, "learning_rate": 1.8798115554125503e-07, "loss": 0.015, "step": 13537 }, { "epoch": 5.34, "learning_rate": 1.8795566868476865e-07, "loss": 0.013, "step": 13538 }, { "epoch": 5.34, "learning_rate": 1.8793018202458655e-07, "loss": 0.0144, "step": 13539 }, { "epoch": 5.34, "learning_rate": 1.8790469556112404e-07, "loss": 0.0178, "step": 13540 }, { "epoch": 5.34, "learning_rate": 1.8787920929479672e-07, "loss": 0.0155, "step": 13541 }, { "epoch": 5.34, "learning_rate": 1.8785372322601987e-07, "loss": 0.0151, "step": 13542 }, { "epoch": 5.34, "learning_rate": 1.8782823735520884e-07, "loss": 0.0129, "step": 13543 }, { "epoch": 5.34, "learning_rate": 1.8780275168277897e-07, "loss": 0.013, "step": 13544 }, { "epoch": 5.34, "learning_rate": 1.8777726620914582e-07, "loss": 0.0128, "step": 13545 }, { "epoch": 5.34, "learning_rate": 1.8775178093472462e-07, "loss": 0.017, "step": 13546 }, { "epoch": 5.34, "learning_rate": 1.877262958599308e-07, "loss": 0.0155, "step": 13547 }, { "epoch": 5.34, "learning_rate": 1.877008109851797e-07, "loss": 0.0141, "step": 13548 }, { "epoch": 5.34, "learning_rate": 1.8767532631088662e-07, "loss": 0.0119, "step": 13549 }, { "epoch": 5.34, "learning_rate": 1.8764984183746703e-07, "loss": 0.0113, "step": 13550 }, { "epoch": 5.34, "learning_rate": 1.8762435756533627e-07, "loss": 0.012, "step": 13551 }, { "epoch": 5.34, "learning_rate": 1.8759887349490968e-07, "loss": 0.0124, "step": 13552 }, { "epoch": 5.34, "learning_rate": 1.8757338962660254e-07, "loss": 0.0102, "step": 13553 }, { "epoch": 5.34, "learning_rate": 1.8754790596083035e-07, "loss": 0.0114, "step": 13554 }, { "epoch": 5.34, "learning_rate": 1.875224224980083e-07, "loss": 0.0147, "step": 13555 }, { "epoch": 5.34, "learning_rate": 1.8749693923855187e-07, "loss": 0.0137, "step": 13556 }, { "epoch": 5.34, "learning_rate": 1.8747145618287625e-07, "loss": 0.012, "step": 13557 }, { "epoch": 5.34, "learning_rate": 1.8744597333139685e-07, "loss": 0.0135, "step": 13558 }, { "epoch": 5.34, "learning_rate": 1.8742049068452906e-07, "loss": 0.0146, "step": 13559 }, { "epoch": 5.34, "learning_rate": 1.8739500824268814e-07, "loss": 0.0131, "step": 13560 }, { "epoch": 5.34, "learning_rate": 1.8736952600628945e-07, "loss": 0.0147, "step": 13561 }, { "epoch": 5.34, "learning_rate": 1.873440439757482e-07, "loss": 0.0124, "step": 13562 }, { "epoch": 5.35, "learning_rate": 1.8731856215147986e-07, "loss": 0.0136, "step": 13563 }, { "epoch": 5.35, "learning_rate": 1.872930805338997e-07, "loss": 0.0153, "step": 13564 }, { "epoch": 5.35, "learning_rate": 1.87267599123423e-07, "loss": 0.0159, "step": 13565 }, { "epoch": 5.35, "learning_rate": 1.8724211792046512e-07, "loss": 0.0131, "step": 13566 }, { "epoch": 5.35, "learning_rate": 1.872166369254412e-07, "loss": 0.0127, "step": 13567 }, { "epoch": 5.35, "learning_rate": 1.8719115613876677e-07, "loss": 0.0159, "step": 13568 }, { "epoch": 5.35, "learning_rate": 1.8716567556085705e-07, "loss": 0.0126, "step": 13569 }, { "epoch": 5.35, "learning_rate": 1.8714019519212728e-07, "loss": 0.0107, "step": 13570 }, { "epoch": 5.35, "learning_rate": 1.8711471503299272e-07, "loss": 0.0129, "step": 13571 }, { "epoch": 5.35, "learning_rate": 1.870892350838688e-07, "loss": 0.0127, "step": 13572 }, { "epoch": 5.35, "learning_rate": 1.8706375534517072e-07, "loss": 0.0115, "step": 13573 }, { "epoch": 5.35, "learning_rate": 1.8703827581731377e-07, "loss": 0.0122, "step": 13574 }, { "epoch": 5.35, "learning_rate": 1.8701279650071319e-07, "loss": 0.012, "step": 13575 }, { "epoch": 5.35, "learning_rate": 1.8698731739578431e-07, "loss": 0.0138, "step": 13576 }, { "epoch": 5.35, "learning_rate": 1.869618385029424e-07, "loss": 0.0186, "step": 13577 }, { "epoch": 5.35, "learning_rate": 1.8693635982260272e-07, "loss": 0.013, "step": 13578 }, { "epoch": 5.35, "learning_rate": 1.869108813551805e-07, "loss": 0.0145, "step": 13579 }, { "epoch": 5.35, "learning_rate": 1.8688540310109092e-07, "loss": 0.011, "step": 13580 }, { "epoch": 5.35, "learning_rate": 1.8685992506074945e-07, "loss": 0.0153, "step": 13581 }, { "epoch": 5.35, "learning_rate": 1.8683444723457121e-07, "loss": 0.0121, "step": 13582 }, { "epoch": 5.35, "learning_rate": 1.868089696229715e-07, "loss": 0.0141, "step": 13583 }, { "epoch": 5.35, "learning_rate": 1.8678349222636542e-07, "loss": 0.0134, "step": 13584 }, { "epoch": 5.35, "learning_rate": 1.8675801504516844e-07, "loss": 0.0157, "step": 13585 }, { "epoch": 5.35, "learning_rate": 1.8673253807979567e-07, "loss": 0.0145, "step": 13586 }, { "epoch": 5.35, "learning_rate": 1.8670706133066238e-07, "loss": 0.0128, "step": 13587 }, { "epoch": 5.36, "learning_rate": 1.866815847981838e-07, "loss": 0.0125, "step": 13588 }, { "epoch": 5.36, "learning_rate": 1.8665610848277503e-07, "loss": 0.014, "step": 13589 }, { "epoch": 5.36, "learning_rate": 1.866306323848515e-07, "loss": 0.0137, "step": 13590 }, { "epoch": 5.36, "learning_rate": 1.8660515650482836e-07, "loss": 0.0144, "step": 13591 }, { "epoch": 5.36, "learning_rate": 1.865796808431208e-07, "loss": 0.0144, "step": 13592 }, { "epoch": 5.36, "learning_rate": 1.86554205400144e-07, "loss": 0.0128, "step": 13593 }, { "epoch": 5.36, "learning_rate": 1.8652873017631323e-07, "loss": 0.0129, "step": 13594 }, { "epoch": 5.36, "learning_rate": 1.865032551720437e-07, "loss": 0.0135, "step": 13595 }, { "epoch": 5.36, "learning_rate": 1.8647778038775065e-07, "loss": 0.0131, "step": 13596 }, { "epoch": 5.36, "learning_rate": 1.8645230582384918e-07, "loss": 0.015, "step": 13597 }, { "epoch": 5.36, "learning_rate": 1.8642683148075451e-07, "loss": 0.0129, "step": 13598 }, { "epoch": 5.36, "learning_rate": 1.864013573588819e-07, "loss": 0.0134, "step": 13599 }, { "epoch": 5.36, "learning_rate": 1.863758834586465e-07, "loss": 0.0144, "step": 13600 }, { "epoch": 5.36, "learning_rate": 1.863504097804635e-07, "loss": 0.0134, "step": 13601 }, { "epoch": 5.36, "learning_rate": 1.8632493632474796e-07, "loss": 0.0128, "step": 13602 }, { "epoch": 5.36, "learning_rate": 1.8629946309191528e-07, "loss": 0.013, "step": 13603 }, { "epoch": 5.36, "learning_rate": 1.8627399008238053e-07, "loss": 0.0152, "step": 13604 }, { "epoch": 5.36, "learning_rate": 1.8624851729655887e-07, "loss": 0.0098, "step": 13605 }, { "epoch": 5.36, "learning_rate": 1.862230447348655e-07, "loss": 0.0114, "step": 13606 }, { "epoch": 5.36, "learning_rate": 1.8619757239771545e-07, "loss": 0.0134, "step": 13607 }, { "epoch": 5.36, "learning_rate": 1.861721002855241e-07, "loss": 0.0151, "step": 13608 }, { "epoch": 5.36, "learning_rate": 1.861466283987065e-07, "loss": 0.013, "step": 13609 }, { "epoch": 5.36, "learning_rate": 1.8612115673767778e-07, "loss": 0.0151, "step": 13610 }, { "epoch": 5.36, "learning_rate": 1.8609568530285305e-07, "loss": 0.0138, "step": 13611 }, { "epoch": 5.36, "learning_rate": 1.8607021409464757e-07, "loss": 0.0158, "step": 13612 }, { "epoch": 5.36, "learning_rate": 1.8604474311347646e-07, "loss": 0.0138, "step": 13613 }, { "epoch": 5.37, "learning_rate": 1.8601927235975482e-07, "loss": 0.0121, "step": 13614 }, { "epoch": 5.37, "learning_rate": 1.8599380183389778e-07, "loss": 0.0111, "step": 13615 }, { "epoch": 5.37, "learning_rate": 1.8596833153632048e-07, "loss": 0.0176, "step": 13616 }, { "epoch": 5.37, "learning_rate": 1.8594286146743803e-07, "loss": 0.016, "step": 13617 }, { "epoch": 5.37, "learning_rate": 1.859173916276656e-07, "loss": 0.0123, "step": 13618 }, { "epoch": 5.37, "learning_rate": 1.858919220174183e-07, "loss": 0.012, "step": 13619 }, { "epoch": 5.37, "learning_rate": 1.858664526371112e-07, "loss": 0.0124, "step": 13620 }, { "epoch": 5.37, "learning_rate": 1.8584098348715945e-07, "loss": 0.0109, "step": 13621 }, { "epoch": 5.37, "learning_rate": 1.8581551456797814e-07, "loss": 0.0142, "step": 13622 }, { "epoch": 5.37, "learning_rate": 1.8579004587998243e-07, "loss": 0.0135, "step": 13623 }, { "epoch": 5.37, "learning_rate": 1.8576457742358728e-07, "loss": 0.0179, "step": 13624 }, { "epoch": 5.37, "learning_rate": 1.8573910919920795e-07, "loss": 0.0107, "step": 13625 }, { "epoch": 5.37, "learning_rate": 1.8571364120725945e-07, "loss": 0.013, "step": 13626 }, { "epoch": 5.37, "learning_rate": 1.8568817344815693e-07, "loss": 0.0109, "step": 13627 }, { "epoch": 5.37, "learning_rate": 1.8566270592231538e-07, "loss": 0.0138, "step": 13628 }, { "epoch": 5.37, "learning_rate": 1.856372386301499e-07, "loss": 0.0124, "step": 13629 }, { "epoch": 5.37, "learning_rate": 1.8561177157207565e-07, "loss": 0.011, "step": 13630 }, { "epoch": 5.37, "learning_rate": 1.8558630474850766e-07, "loss": 0.013, "step": 13631 }, { "epoch": 5.37, "learning_rate": 1.85560838159861e-07, "loss": 0.0146, "step": 13632 }, { "epoch": 5.37, "learning_rate": 1.8553537180655067e-07, "loss": 0.0124, "step": 13633 }, { "epoch": 5.37, "learning_rate": 1.8550990568899187e-07, "loss": 0.014, "step": 13634 }, { "epoch": 5.37, "learning_rate": 1.854844398075996e-07, "loss": 0.0127, "step": 13635 }, { "epoch": 5.37, "learning_rate": 1.8545897416278886e-07, "loss": 0.012, "step": 13636 }, { "epoch": 5.37, "learning_rate": 1.8543350875497477e-07, "loss": 0.0115, "step": 13637 }, { "epoch": 5.37, "learning_rate": 1.8540804358457228e-07, "loss": 0.0123, "step": 13638 }, { "epoch": 5.38, "learning_rate": 1.8538257865199653e-07, "loss": 0.0139, "step": 13639 }, { "epoch": 5.38, "learning_rate": 1.8535711395766257e-07, "loss": 0.0153, "step": 13640 }, { "epoch": 5.38, "learning_rate": 1.853316495019854e-07, "loss": 0.0111, "step": 13641 }, { "epoch": 5.38, "learning_rate": 1.8530618528537997e-07, "loss": 0.0131, "step": 13642 }, { "epoch": 5.38, "learning_rate": 1.8528072130826147e-07, "loss": 0.0144, "step": 13643 }, { "epoch": 5.38, "learning_rate": 1.8525525757104484e-07, "loss": 0.0137, "step": 13644 }, { "epoch": 5.38, "learning_rate": 1.852297940741451e-07, "loss": 0.0165, "step": 13645 }, { "epoch": 5.38, "learning_rate": 1.8520433081797726e-07, "loss": 0.0124, "step": 13646 }, { "epoch": 5.38, "learning_rate": 1.8517886780295625e-07, "loss": 0.0128, "step": 13647 }, { "epoch": 5.38, "learning_rate": 1.8515340502949726e-07, "loss": 0.0129, "step": 13648 }, { "epoch": 5.38, "learning_rate": 1.8512794249801523e-07, "loss": 0.015, "step": 13649 }, { "epoch": 5.38, "learning_rate": 1.851024802089251e-07, "loss": 0.0119, "step": 13650 }, { "epoch": 5.38, "learning_rate": 1.850770181626418e-07, "loss": 0.0148, "step": 13651 }, { "epoch": 5.38, "learning_rate": 1.8505155635958056e-07, "loss": 0.0137, "step": 13652 }, { "epoch": 5.38, "learning_rate": 1.8502609480015621e-07, "loss": 0.0154, "step": 13653 }, { "epoch": 5.38, "learning_rate": 1.8500063348478374e-07, "loss": 0.0128, "step": 13654 }, { "epoch": 5.38, "learning_rate": 1.8497517241387807e-07, "loss": 0.0122, "step": 13655 }, { "epoch": 5.38, "learning_rate": 1.8494971158785435e-07, "loss": 0.0143, "step": 13656 }, { "epoch": 5.38, "learning_rate": 1.8492425100712747e-07, "loss": 0.0171, "step": 13657 }, { "epoch": 5.38, "learning_rate": 1.8489879067211236e-07, "loss": 0.0109, "step": 13658 }, { "epoch": 5.38, "learning_rate": 1.8487333058322403e-07, "loss": 0.0144, "step": 13659 }, { "epoch": 5.38, "learning_rate": 1.8484787074087737e-07, "loss": 0.013, "step": 13660 }, { "epoch": 5.38, "learning_rate": 1.8482241114548743e-07, "loss": 0.0114, "step": 13661 }, { "epoch": 5.38, "learning_rate": 1.8479695179746907e-07, "loss": 0.0118, "step": 13662 }, { "epoch": 5.38, "learning_rate": 1.8477149269723736e-07, "loss": 0.0142, "step": 13663 }, { "epoch": 5.39, "learning_rate": 1.847460338452071e-07, "loss": 0.0099, "step": 13664 }, { "epoch": 5.39, "learning_rate": 1.8472057524179336e-07, "loss": 0.0118, "step": 13665 }, { "epoch": 5.39, "learning_rate": 1.8469511688741103e-07, "loss": 0.0124, "step": 13666 }, { "epoch": 5.39, "learning_rate": 1.8466965878247504e-07, "loss": 0.0139, "step": 13667 }, { "epoch": 5.39, "learning_rate": 1.8464420092740029e-07, "loss": 0.0141, "step": 13668 }, { "epoch": 5.39, "learning_rate": 1.8461874332260164e-07, "loss": 0.013, "step": 13669 }, { "epoch": 5.39, "learning_rate": 1.8459328596849418e-07, "loss": 0.0131, "step": 13670 }, { "epoch": 5.39, "learning_rate": 1.8456782886549279e-07, "loss": 0.0159, "step": 13671 }, { "epoch": 5.39, "learning_rate": 1.845423720140123e-07, "loss": 0.0139, "step": 13672 }, { "epoch": 5.39, "learning_rate": 1.8451691541446753e-07, "loss": 0.0118, "step": 13673 }, { "epoch": 5.39, "learning_rate": 1.8449145906727366e-07, "loss": 0.0144, "step": 13674 }, { "epoch": 5.39, "learning_rate": 1.844660029728454e-07, "loss": 0.0148, "step": 13675 }, { "epoch": 5.39, "learning_rate": 1.844405471315977e-07, "loss": 0.0135, "step": 13676 }, { "epoch": 5.39, "learning_rate": 1.8441509154394543e-07, "loss": 0.0115, "step": 13677 }, { "epoch": 5.39, "learning_rate": 1.8438963621030338e-07, "loss": 0.0159, "step": 13678 }, { "epoch": 5.39, "learning_rate": 1.8436418113108662e-07, "loss": 0.012, "step": 13679 }, { "epoch": 5.39, "learning_rate": 1.8433872630670997e-07, "loss": 0.0125, "step": 13680 }, { "epoch": 5.39, "learning_rate": 1.843132717375883e-07, "loss": 0.0112, "step": 13681 }, { "epoch": 5.39, "learning_rate": 1.8428781742413636e-07, "loss": 0.0136, "step": 13682 }, { "epoch": 5.39, "learning_rate": 1.8426236336676916e-07, "loss": 0.0126, "step": 13683 }, { "epoch": 5.39, "learning_rate": 1.8423690956590155e-07, "loss": 0.0142, "step": 13684 }, { "epoch": 5.39, "learning_rate": 1.8421145602194829e-07, "loss": 0.0156, "step": 13685 }, { "epoch": 5.39, "learning_rate": 1.8418600273532438e-07, "loss": 0.014, "step": 13686 }, { "epoch": 5.39, "learning_rate": 1.8416054970644452e-07, "loss": 0.0135, "step": 13687 }, { "epoch": 5.39, "learning_rate": 1.8413509693572365e-07, "loss": 0.0131, "step": 13688 }, { "epoch": 5.39, "learning_rate": 1.8410964442357664e-07, "loss": 0.0155, "step": 13689 }, { "epoch": 5.4, "learning_rate": 1.8408419217041824e-07, "loss": 0.0125, "step": 13690 }, { "epoch": 5.4, "learning_rate": 1.8405874017666326e-07, "loss": 0.0135, "step": 13691 }, { "epoch": 5.4, "learning_rate": 1.8403328844272663e-07, "loss": 0.0128, "step": 13692 }, { "epoch": 5.4, "learning_rate": 1.8400783696902316e-07, "loss": 0.012, "step": 13693 }, { "epoch": 5.4, "learning_rate": 1.8398238575596765e-07, "loss": 0.0131, "step": 13694 }, { "epoch": 5.4, "learning_rate": 1.839569348039748e-07, "loss": 0.0129, "step": 13695 }, { "epoch": 5.4, "learning_rate": 1.8393148411345963e-07, "loss": 0.0131, "step": 13696 }, { "epoch": 5.4, "learning_rate": 1.8390603368483682e-07, "loss": 0.0133, "step": 13697 }, { "epoch": 5.4, "learning_rate": 1.8388058351852125e-07, "loss": 0.0139, "step": 13698 }, { "epoch": 5.4, "learning_rate": 1.8385513361492764e-07, "loss": 0.0147, "step": 13699 }, { "epoch": 5.4, "learning_rate": 1.8382968397447074e-07, "loss": 0.0127, "step": 13700 }, { "epoch": 5.4, "learning_rate": 1.838042345975655e-07, "loss": 0.0147, "step": 13701 }, { "epoch": 5.4, "learning_rate": 1.837787854846266e-07, "loss": 0.0145, "step": 13702 }, { "epoch": 5.4, "learning_rate": 1.8375333663606888e-07, "loss": 0.0134, "step": 13703 }, { "epoch": 5.4, "learning_rate": 1.83727888052307e-07, "loss": 0.0131, "step": 13704 }, { "epoch": 5.4, "learning_rate": 1.8370243973375589e-07, "loss": 0.0113, "step": 13705 }, { "epoch": 5.4, "learning_rate": 1.8367699168083019e-07, "loss": 0.0124, "step": 13706 }, { "epoch": 5.4, "learning_rate": 1.8365154389394477e-07, "loss": 0.0151, "step": 13707 }, { "epoch": 5.4, "learning_rate": 1.8362609637351428e-07, "loss": 0.0116, "step": 13708 }, { "epoch": 5.4, "learning_rate": 1.8360064911995353e-07, "loss": 0.0131, "step": 13709 }, { "epoch": 5.4, "learning_rate": 1.8357520213367733e-07, "loss": 0.0123, "step": 13710 }, { "epoch": 5.4, "learning_rate": 1.8354975541510036e-07, "loss": 0.0103, "step": 13711 }, { "epoch": 5.4, "learning_rate": 1.8352430896463739e-07, "loss": 0.0139, "step": 13712 }, { "epoch": 5.4, "learning_rate": 1.8349886278270304e-07, "loss": 0.0107, "step": 13713 }, { "epoch": 5.4, "learning_rate": 1.8347341686971222e-07, "loss": 0.0124, "step": 13714 }, { "epoch": 5.41, "learning_rate": 1.8344797122607962e-07, "loss": 0.0145, "step": 13715 }, { "epoch": 5.41, "learning_rate": 1.8342252585221992e-07, "loss": 0.0123, "step": 13716 }, { "epoch": 5.41, "learning_rate": 1.8339708074854785e-07, "loss": 0.0153, "step": 13717 }, { "epoch": 5.41, "learning_rate": 1.8337163591547804e-07, "loss": 0.0135, "step": 13718 }, { "epoch": 5.41, "learning_rate": 1.8334619135342538e-07, "loss": 0.013, "step": 13719 }, { "epoch": 5.41, "learning_rate": 1.8332074706280451e-07, "loss": 0.0127, "step": 13720 }, { "epoch": 5.41, "learning_rate": 1.8329530304403008e-07, "loss": 0.0192, "step": 13721 }, { "epoch": 5.41, "learning_rate": 1.8326985929751677e-07, "loss": 0.0168, "step": 13722 }, { "epoch": 5.41, "learning_rate": 1.832444158236794e-07, "loss": 0.0122, "step": 13723 }, { "epoch": 5.41, "learning_rate": 1.832189726229326e-07, "loss": 0.0133, "step": 13724 }, { "epoch": 5.41, "learning_rate": 1.83193529695691e-07, "loss": 0.014, "step": 13725 }, { "epoch": 5.41, "learning_rate": 1.8316808704236937e-07, "loss": 0.0135, "step": 13726 }, { "epoch": 5.41, "learning_rate": 1.8314264466338227e-07, "loss": 0.0129, "step": 13727 }, { "epoch": 5.41, "learning_rate": 1.8311720255914448e-07, "loss": 0.017, "step": 13728 }, { "epoch": 5.41, "learning_rate": 1.8309176073007065e-07, "loss": 0.0122, "step": 13729 }, { "epoch": 5.41, "learning_rate": 1.8306631917657537e-07, "loss": 0.0125, "step": 13730 }, { "epoch": 5.41, "learning_rate": 1.8304087789907336e-07, "loss": 0.0145, "step": 13731 }, { "epoch": 5.41, "learning_rate": 1.830154368979793e-07, "loss": 0.0149, "step": 13732 }, { "epoch": 5.41, "learning_rate": 1.829899961737078e-07, "loss": 0.0139, "step": 13733 }, { "epoch": 5.41, "learning_rate": 1.829645557266735e-07, "loss": 0.0158, "step": 13734 }, { "epoch": 5.41, "learning_rate": 1.82939115557291e-07, "loss": 0.0167, "step": 13735 }, { "epoch": 5.41, "learning_rate": 1.8291367566597505e-07, "loss": 0.0146, "step": 13736 }, { "epoch": 5.41, "learning_rate": 1.828882360531402e-07, "loss": 0.0134, "step": 13737 }, { "epoch": 5.41, "learning_rate": 1.8286279671920113e-07, "loss": 0.0116, "step": 13738 }, { "epoch": 5.41, "learning_rate": 1.8283735766457242e-07, "loss": 0.0108, "step": 13739 }, { "epoch": 5.42, "learning_rate": 1.828119188896686e-07, "loss": 0.0119, "step": 13740 }, { "epoch": 5.42, "learning_rate": 1.8278648039490447e-07, "loss": 0.0156, "step": 13741 }, { "epoch": 5.42, "learning_rate": 1.8276104218069455e-07, "loss": 0.0153, "step": 13742 }, { "epoch": 5.42, "learning_rate": 1.8273560424745344e-07, "loss": 0.0126, "step": 13743 }, { "epoch": 5.42, "learning_rate": 1.827101665955957e-07, "loss": 0.0156, "step": 13744 }, { "epoch": 5.42, "learning_rate": 1.8268472922553596e-07, "loss": 0.0134, "step": 13745 }, { "epoch": 5.42, "learning_rate": 1.8265929213768888e-07, "loss": 0.0125, "step": 13746 }, { "epoch": 5.42, "learning_rate": 1.8263385533246898e-07, "loss": 0.0139, "step": 13747 }, { "epoch": 5.42, "learning_rate": 1.826084188102908e-07, "loss": 0.0104, "step": 13748 }, { "epoch": 5.42, "learning_rate": 1.8258298257156897e-07, "loss": 0.014, "step": 13749 }, { "epoch": 5.42, "learning_rate": 1.825575466167181e-07, "loss": 0.0107, "step": 13750 }, { "epoch": 5.42, "learning_rate": 1.8253211094615267e-07, "loss": 0.011, "step": 13751 }, { "epoch": 5.42, "learning_rate": 1.8250667556028733e-07, "loss": 0.0105, "step": 13752 }, { "epoch": 5.42, "learning_rate": 1.8248124045953645e-07, "loss": 0.0131, "step": 13753 }, { "epoch": 5.42, "learning_rate": 1.8245580564431484e-07, "loss": 0.0123, "step": 13754 }, { "epoch": 5.42, "learning_rate": 1.8243037111503694e-07, "loss": 0.0162, "step": 13755 }, { "epoch": 5.42, "learning_rate": 1.8240493687211728e-07, "loss": 0.0131, "step": 13756 }, { "epoch": 5.42, "learning_rate": 1.823795029159704e-07, "loss": 0.0136, "step": 13757 }, { "epoch": 5.42, "learning_rate": 1.8235406924701077e-07, "loss": 0.0114, "step": 13758 }, { "epoch": 5.42, "learning_rate": 1.8232863586565308e-07, "loss": 0.0146, "step": 13759 }, { "epoch": 5.42, "learning_rate": 1.8230320277231176e-07, "loss": 0.016, "step": 13760 }, { "epoch": 5.42, "learning_rate": 1.8227776996740133e-07, "loss": 0.0143, "step": 13761 }, { "epoch": 5.42, "learning_rate": 1.8225233745133623e-07, "loss": 0.0141, "step": 13762 }, { "epoch": 5.42, "learning_rate": 1.8222690522453116e-07, "loss": 0.0167, "step": 13763 }, { "epoch": 5.42, "learning_rate": 1.8220147328740053e-07, "loss": 0.0137, "step": 13764 }, { "epoch": 5.42, "learning_rate": 1.8217604164035883e-07, "loss": 0.016, "step": 13765 }, { "epoch": 5.43, "learning_rate": 1.8215061028382053e-07, "loss": 0.0158, "step": 13766 }, { "epoch": 5.43, "learning_rate": 1.821251792182001e-07, "loss": 0.0139, "step": 13767 }, { "epoch": 5.43, "learning_rate": 1.8209974844391216e-07, "loss": 0.0118, "step": 13768 }, { "epoch": 5.43, "learning_rate": 1.8207431796137112e-07, "loss": 0.0134, "step": 13769 }, { "epoch": 5.43, "learning_rate": 1.8204888777099144e-07, "loss": 0.0147, "step": 13770 }, { "epoch": 5.43, "learning_rate": 1.8202345787318756e-07, "loss": 0.0134, "step": 13771 }, { "epoch": 5.43, "learning_rate": 1.8199802826837406e-07, "loss": 0.0103, "step": 13772 }, { "epoch": 5.43, "learning_rate": 1.8197259895696534e-07, "loss": 0.0144, "step": 13773 }, { "epoch": 5.43, "learning_rate": 1.8194716993937583e-07, "loss": 0.0161, "step": 13774 }, { "epoch": 5.43, "learning_rate": 1.8192174121601994e-07, "loss": 0.0106, "step": 13775 }, { "epoch": 5.43, "learning_rate": 1.8189631278731228e-07, "loss": 0.0146, "step": 13776 }, { "epoch": 5.43, "learning_rate": 1.818708846536672e-07, "loss": 0.0134, "step": 13777 }, { "epoch": 5.43, "learning_rate": 1.8184545681549917e-07, "loss": 0.0161, "step": 13778 }, { "epoch": 5.43, "learning_rate": 1.8182002927322257e-07, "loss": 0.0148, "step": 13779 }, { "epoch": 5.43, "learning_rate": 1.817946020272518e-07, "loss": 0.0121, "step": 13780 }, { "epoch": 5.43, "learning_rate": 1.8176917507800146e-07, "loss": 0.0118, "step": 13781 }, { "epoch": 5.43, "learning_rate": 1.817437484258858e-07, "loss": 0.0147, "step": 13782 }, { "epoch": 5.43, "learning_rate": 1.8171832207131932e-07, "loss": 0.0169, "step": 13783 }, { "epoch": 5.43, "learning_rate": 1.8169289601471633e-07, "loss": 0.016, "step": 13784 }, { "epoch": 5.43, "learning_rate": 1.816674702564914e-07, "loss": 0.0122, "step": 13785 }, { "epoch": 5.43, "learning_rate": 1.816420447970588e-07, "loss": 0.0144, "step": 13786 }, { "epoch": 5.43, "learning_rate": 1.81616619636833e-07, "loss": 0.0137, "step": 13787 }, { "epoch": 5.43, "learning_rate": 1.8159119477622837e-07, "loss": 0.0145, "step": 13788 }, { "epoch": 5.43, "learning_rate": 1.8156577021565923e-07, "loss": 0.0137, "step": 13789 }, { "epoch": 5.43, "learning_rate": 1.8154034595554002e-07, "loss": 0.0117, "step": 13790 }, { "epoch": 5.44, "learning_rate": 1.8151492199628518e-07, "loss": 0.0163, "step": 13791 }, { "epoch": 5.44, "learning_rate": 1.81489498338309e-07, "loss": 0.0138, "step": 13792 }, { "epoch": 5.44, "learning_rate": 1.814640749820258e-07, "loss": 0.0125, "step": 13793 }, { "epoch": 5.44, "learning_rate": 1.8143865192785007e-07, "loss": 0.014, "step": 13794 }, { "epoch": 5.44, "learning_rate": 1.814132291761961e-07, "loss": 0.0132, "step": 13795 }, { "epoch": 5.44, "learning_rate": 1.8138780672747826e-07, "loss": 0.0143, "step": 13796 }, { "epoch": 5.44, "learning_rate": 1.813623845821109e-07, "loss": 0.0102, "step": 13797 }, { "epoch": 5.44, "learning_rate": 1.813369627405082e-07, "loss": 0.0122, "step": 13798 }, { "epoch": 5.44, "learning_rate": 1.8131154120308476e-07, "loss": 0.0138, "step": 13799 }, { "epoch": 5.44, "learning_rate": 1.812861199702548e-07, "loss": 0.013, "step": 13800 }, { "epoch": 5.44, "learning_rate": 1.8126069904243264e-07, "loss": 0.0131, "step": 13801 }, { "epoch": 5.44, "learning_rate": 1.8123527842003252e-07, "loss": 0.0132, "step": 13802 }, { "epoch": 5.44, "learning_rate": 1.8120985810346892e-07, "loss": 0.0142, "step": 13803 }, { "epoch": 5.44, "learning_rate": 1.811844380931561e-07, "loss": 0.0132, "step": 13804 }, { "epoch": 5.44, "learning_rate": 1.811590183895083e-07, "loss": 0.0136, "step": 13805 }, { "epoch": 5.44, "learning_rate": 1.811335989929399e-07, "loss": 0.0209, "step": 13806 }, { "epoch": 5.44, "learning_rate": 1.8110817990386507e-07, "loss": 0.0144, "step": 13807 }, { "epoch": 5.44, "learning_rate": 1.8108276112269827e-07, "loss": 0.0135, "step": 13808 }, { "epoch": 5.44, "learning_rate": 1.810573426498537e-07, "loss": 0.0134, "step": 13809 }, { "epoch": 5.44, "learning_rate": 1.810319244857457e-07, "loss": 0.0149, "step": 13810 }, { "epoch": 5.44, "learning_rate": 1.8100650663078844e-07, "loss": 0.0146, "step": 13811 }, { "epoch": 5.44, "learning_rate": 1.8098108908539628e-07, "loss": 0.0137, "step": 13812 }, { "epoch": 5.44, "learning_rate": 1.8095567184998345e-07, "loss": 0.013, "step": 13813 }, { "epoch": 5.44, "learning_rate": 1.8093025492496423e-07, "loss": 0.013, "step": 13814 }, { "epoch": 5.44, "learning_rate": 1.8090483831075284e-07, "loss": 0.0089, "step": 13815 }, { "epoch": 5.44, "learning_rate": 1.8087942200776362e-07, "loss": 0.0128, "step": 13816 }, { "epoch": 5.45, "learning_rate": 1.8085400601641077e-07, "loss": 0.0123, "step": 13817 }, { "epoch": 5.45, "learning_rate": 1.808285903371085e-07, "loss": 0.0115, "step": 13818 }, { "epoch": 5.45, "learning_rate": 1.8080317497027108e-07, "loss": 0.0106, "step": 13819 }, { "epoch": 5.45, "learning_rate": 1.8077775991631265e-07, "loss": 0.015, "step": 13820 }, { "epoch": 5.45, "learning_rate": 1.8075234517564762e-07, "loss": 0.0106, "step": 13821 }, { "epoch": 5.45, "learning_rate": 1.8072693074869008e-07, "loss": 0.0115, "step": 13822 }, { "epoch": 5.45, "learning_rate": 1.8070151663585428e-07, "loss": 0.0154, "step": 13823 }, { "epoch": 5.45, "learning_rate": 1.8067610283755438e-07, "loss": 0.0125, "step": 13824 }, { "epoch": 5.45, "learning_rate": 1.806506893542047e-07, "loss": 0.012, "step": 13825 }, { "epoch": 5.45, "learning_rate": 1.8062527618621933e-07, "loss": 0.013, "step": 13826 }, { "epoch": 5.45, "learning_rate": 1.805998633340126e-07, "loss": 0.0134, "step": 13827 }, { "epoch": 5.45, "learning_rate": 1.8057445079799853e-07, "loss": 0.0139, "step": 13828 }, { "epoch": 5.45, "learning_rate": 1.8054903857859133e-07, "loss": 0.0135, "step": 13829 }, { "epoch": 5.45, "learning_rate": 1.8052362667620536e-07, "loss": 0.0113, "step": 13830 }, { "epoch": 5.45, "learning_rate": 1.8049821509125464e-07, "loss": 0.0152, "step": 13831 }, { "epoch": 5.45, "learning_rate": 1.8047280382415337e-07, "loss": 0.0127, "step": 13832 }, { "epoch": 5.45, "learning_rate": 1.804473928753157e-07, "loss": 0.015, "step": 13833 }, { "epoch": 5.45, "learning_rate": 1.8042198224515583e-07, "loss": 0.0106, "step": 13834 }, { "epoch": 5.45, "learning_rate": 1.8039657193408787e-07, "loss": 0.0141, "step": 13835 }, { "epoch": 5.45, "learning_rate": 1.80371161942526e-07, "loss": 0.0114, "step": 13836 }, { "epoch": 5.45, "learning_rate": 1.803457522708844e-07, "loss": 0.0101, "step": 13837 }, { "epoch": 5.45, "learning_rate": 1.8032034291957712e-07, "loss": 0.0141, "step": 13838 }, { "epoch": 5.45, "learning_rate": 1.8029493388901837e-07, "loss": 0.0134, "step": 13839 }, { "epoch": 5.45, "learning_rate": 1.8026952517962227e-07, "loss": 0.0188, "step": 13840 }, { "epoch": 5.45, "learning_rate": 1.8024411679180289e-07, "loss": 0.0129, "step": 13841 }, { "epoch": 5.46, "learning_rate": 1.8021870872597432e-07, "loss": 0.0125, "step": 13842 }, { "epoch": 5.46, "learning_rate": 1.8019330098255081e-07, "loss": 0.0136, "step": 13843 }, { "epoch": 5.46, "learning_rate": 1.801678935619464e-07, "loss": 0.0133, "step": 13844 }, { "epoch": 5.46, "learning_rate": 1.8014248646457517e-07, "loss": 0.016, "step": 13845 }, { "epoch": 5.46, "learning_rate": 1.8011707969085123e-07, "loss": 0.0142, "step": 13846 }, { "epoch": 5.46, "learning_rate": 1.8009167324118858e-07, "loss": 0.0123, "step": 13847 }, { "epoch": 5.46, "learning_rate": 1.8006626711600151e-07, "loss": 0.0146, "step": 13848 }, { "epoch": 5.46, "learning_rate": 1.8004086131570397e-07, "loss": 0.0139, "step": 13849 }, { "epoch": 5.46, "learning_rate": 1.8001545584071008e-07, "loss": 0.0138, "step": 13850 }, { "epoch": 5.46, "learning_rate": 1.7999005069143377e-07, "loss": 0.0108, "step": 13851 }, { "epoch": 5.46, "learning_rate": 1.799646458682893e-07, "loss": 0.0145, "step": 13852 }, { "epoch": 5.46, "learning_rate": 1.799392413716907e-07, "loss": 0.0151, "step": 13853 }, { "epoch": 5.46, "learning_rate": 1.7991383720205195e-07, "loss": 0.0113, "step": 13854 }, { "epoch": 5.46, "learning_rate": 1.7988843335978713e-07, "loss": 0.0128, "step": 13855 }, { "epoch": 5.46, "learning_rate": 1.7986302984531024e-07, "loss": 0.0143, "step": 13856 }, { "epoch": 5.46, "learning_rate": 1.7983762665903538e-07, "loss": 0.0138, "step": 13857 }, { "epoch": 5.46, "learning_rate": 1.7981222380137656e-07, "loss": 0.0114, "step": 13858 }, { "epoch": 5.46, "learning_rate": 1.7978682127274782e-07, "loss": 0.0146, "step": 13859 }, { "epoch": 5.46, "learning_rate": 1.7976141907356315e-07, "loss": 0.0128, "step": 13860 }, { "epoch": 5.46, "learning_rate": 1.797360172042366e-07, "loss": 0.0118, "step": 13861 }, { "epoch": 5.46, "learning_rate": 1.7971061566518222e-07, "loss": 0.0134, "step": 13862 }, { "epoch": 5.46, "learning_rate": 1.7968521445681394e-07, "loss": 0.0151, "step": 13863 }, { "epoch": 5.46, "learning_rate": 1.796598135795457e-07, "loss": 0.0147, "step": 13864 }, { "epoch": 5.46, "learning_rate": 1.7963441303379169e-07, "loss": 0.0128, "step": 13865 }, { "epoch": 5.46, "learning_rate": 1.7960901281996577e-07, "loss": 0.0143, "step": 13866 }, { "epoch": 5.47, "learning_rate": 1.7958361293848196e-07, "loss": 0.0108, "step": 13867 }, { "epoch": 5.47, "learning_rate": 1.7955821338975422e-07, "loss": 0.0114, "step": 13868 }, { "epoch": 5.47, "learning_rate": 1.7953281417419647e-07, "loss": 0.0148, "step": 13869 }, { "epoch": 5.47, "learning_rate": 1.795074152922228e-07, "loss": 0.0148, "step": 13870 }, { "epoch": 5.47, "learning_rate": 1.7948201674424716e-07, "loss": 0.0122, "step": 13871 }, { "epoch": 5.47, "learning_rate": 1.7945661853068343e-07, "loss": 0.0162, "step": 13872 }, { "epoch": 5.47, "learning_rate": 1.794312206519455e-07, "loss": 0.0132, "step": 13873 }, { "epoch": 5.47, "learning_rate": 1.7940582310844754e-07, "loss": 0.0126, "step": 13874 }, { "epoch": 5.47, "learning_rate": 1.7938042590060335e-07, "loss": 0.0127, "step": 13875 }, { "epoch": 5.47, "learning_rate": 1.7935502902882687e-07, "loss": 0.0124, "step": 13876 }, { "epoch": 5.47, "learning_rate": 1.7932963249353203e-07, "loss": 0.0163, "step": 13877 }, { "epoch": 5.47, "learning_rate": 1.7930423629513273e-07, "loss": 0.0132, "step": 13878 }, { "epoch": 5.47, "learning_rate": 1.7927884043404298e-07, "loss": 0.0109, "step": 13879 }, { "epoch": 5.47, "learning_rate": 1.7925344491067661e-07, "loss": 0.0125, "step": 13880 }, { "epoch": 5.47, "learning_rate": 1.7922804972544755e-07, "loss": 0.0154, "step": 13881 }, { "epoch": 5.47, "learning_rate": 1.792026548787697e-07, "loss": 0.0162, "step": 13882 }, { "epoch": 5.47, "learning_rate": 1.7917726037105698e-07, "loss": 0.0128, "step": 13883 }, { "epoch": 5.47, "learning_rate": 1.7915186620272329e-07, "loss": 0.013, "step": 13884 }, { "epoch": 5.47, "learning_rate": 1.791264723741825e-07, "loss": 0.0122, "step": 13885 }, { "epoch": 5.47, "learning_rate": 1.7910107888584847e-07, "loss": 0.0143, "step": 13886 }, { "epoch": 5.47, "learning_rate": 1.7907568573813503e-07, "loss": 0.016, "step": 13887 }, { "epoch": 5.47, "learning_rate": 1.7905029293145617e-07, "loss": 0.0122, "step": 13888 }, { "epoch": 5.47, "learning_rate": 1.790249004662257e-07, "loss": 0.0147, "step": 13889 }, { "epoch": 5.47, "learning_rate": 1.789995083428575e-07, "loss": 0.0137, "step": 13890 }, { "epoch": 5.47, "learning_rate": 1.789741165617653e-07, "loss": 0.0146, "step": 13891 }, { "epoch": 5.47, "learning_rate": 1.7894872512336313e-07, "loss": 0.0144, "step": 13892 }, { "epoch": 5.48, "learning_rate": 1.7892333402806473e-07, "loss": 0.0133, "step": 13893 }, { "epoch": 5.48, "learning_rate": 1.7889794327628396e-07, "loss": 0.0129, "step": 13894 }, { "epoch": 5.48, "learning_rate": 1.7887255286843464e-07, "loss": 0.0139, "step": 13895 }, { "epoch": 5.48, "learning_rate": 1.7884716280493053e-07, "loss": 0.0145, "step": 13896 }, { "epoch": 5.48, "learning_rate": 1.788217730861856e-07, "loss": 0.0119, "step": 13897 }, { "epoch": 5.48, "learning_rate": 1.7879638371261355e-07, "loss": 0.0139, "step": 13898 }, { "epoch": 5.48, "learning_rate": 1.7877099468462826e-07, "loss": 0.0151, "step": 13899 }, { "epoch": 5.48, "learning_rate": 1.7874560600264344e-07, "loss": 0.0149, "step": 13900 }, { "epoch": 5.48, "learning_rate": 1.7872021766707296e-07, "loss": 0.013, "step": 13901 }, { "epoch": 5.48, "learning_rate": 1.786948296783306e-07, "loss": 0.0129, "step": 13902 }, { "epoch": 5.48, "learning_rate": 1.7866944203683017e-07, "loss": 0.015, "step": 13903 }, { "epoch": 5.48, "learning_rate": 1.786440547429853e-07, "loss": 0.0124, "step": 13904 }, { "epoch": 5.48, "learning_rate": 1.7861866779720994e-07, "loss": 0.0156, "step": 13905 }, { "epoch": 5.48, "learning_rate": 1.7859328119991783e-07, "loss": 0.0118, "step": 13906 }, { "epoch": 5.48, "learning_rate": 1.7856789495152272e-07, "loss": 0.0127, "step": 13907 }, { "epoch": 5.48, "learning_rate": 1.7854250905243833e-07, "loss": 0.011, "step": 13908 }, { "epoch": 5.48, "learning_rate": 1.7851712350307834e-07, "loss": 0.0132, "step": 13909 }, { "epoch": 5.48, "learning_rate": 1.7849173830385667e-07, "loss": 0.0161, "step": 13910 }, { "epoch": 5.48, "learning_rate": 1.78466353455187e-07, "loss": 0.012, "step": 13911 }, { "epoch": 5.48, "learning_rate": 1.7844096895748302e-07, "loss": 0.0147, "step": 13912 }, { "epoch": 5.48, "learning_rate": 1.784155848111584e-07, "loss": 0.013, "step": 13913 }, { "epoch": 5.48, "learning_rate": 1.78390201016627e-07, "loss": 0.0137, "step": 13914 }, { "epoch": 5.48, "learning_rate": 1.7836481757430252e-07, "loss": 0.0132, "step": 13915 }, { "epoch": 5.48, "learning_rate": 1.7833943448459862e-07, "loss": 0.0118, "step": 13916 }, { "epoch": 5.48, "learning_rate": 1.7831405174792902e-07, "loss": 0.0131, "step": 13917 }, { "epoch": 5.49, "learning_rate": 1.7828866936470733e-07, "loss": 0.0139, "step": 13918 }, { "epoch": 5.49, "learning_rate": 1.782632873353474e-07, "loss": 0.0127, "step": 13919 }, { "epoch": 5.49, "learning_rate": 1.7823790566026288e-07, "loss": 0.0107, "step": 13920 }, { "epoch": 5.49, "learning_rate": 1.7821252433986742e-07, "loss": 0.0134, "step": 13921 }, { "epoch": 5.49, "learning_rate": 1.7818714337457464e-07, "loss": 0.0121, "step": 13922 }, { "epoch": 5.49, "learning_rate": 1.7816176276479832e-07, "loss": 0.0147, "step": 13923 }, { "epoch": 5.49, "learning_rate": 1.781363825109521e-07, "loss": 0.0142, "step": 13924 }, { "epoch": 5.49, "learning_rate": 1.781110026134496e-07, "loss": 0.0118, "step": 13925 }, { "epoch": 5.49, "learning_rate": 1.7808562307270442e-07, "loss": 0.0136, "step": 13926 }, { "epoch": 5.49, "learning_rate": 1.780602438891303e-07, "loss": 0.013, "step": 13927 }, { "epoch": 5.49, "learning_rate": 1.7803486506314089e-07, "loss": 0.0147, "step": 13928 }, { "epoch": 5.49, "learning_rate": 1.7800948659514984e-07, "loss": 0.015, "step": 13929 }, { "epoch": 5.49, "learning_rate": 1.7798410848557067e-07, "loss": 0.0138, "step": 13930 }, { "epoch": 5.49, "learning_rate": 1.7795873073481706e-07, "loss": 0.0148, "step": 13931 }, { "epoch": 5.49, "learning_rate": 1.7793335334330266e-07, "loss": 0.0142, "step": 13932 }, { "epoch": 5.49, "learning_rate": 1.779079763114411e-07, "loss": 0.0125, "step": 13933 }, { "epoch": 5.49, "learning_rate": 1.7788259963964594e-07, "loss": 0.0122, "step": 13934 }, { "epoch": 5.49, "learning_rate": 1.7785722332833077e-07, "loss": 0.0128, "step": 13935 }, { "epoch": 5.49, "learning_rate": 1.7783184737790915e-07, "loss": 0.0155, "step": 13936 }, { "epoch": 5.49, "learning_rate": 1.778064717887948e-07, "loss": 0.0135, "step": 13937 }, { "epoch": 5.49, "learning_rate": 1.7778109656140123e-07, "loss": 0.0118, "step": 13938 }, { "epoch": 5.49, "learning_rate": 1.7775572169614204e-07, "loss": 0.0136, "step": 13939 }, { "epoch": 5.49, "learning_rate": 1.7773034719343073e-07, "loss": 0.0139, "step": 13940 }, { "epoch": 5.49, "learning_rate": 1.777049730536809e-07, "loss": 0.0128, "step": 13941 }, { "epoch": 5.49, "learning_rate": 1.7767959927730616e-07, "loss": 0.0148, "step": 13942 }, { "epoch": 5.5, "learning_rate": 1.7765422586472003e-07, "loss": 0.0154, "step": 13943 }, { "epoch": 5.5, "learning_rate": 1.7762885281633602e-07, "loss": 0.0122, "step": 13944 }, { "epoch": 5.5, "learning_rate": 1.7760348013256775e-07, "loss": 0.0117, "step": 13945 }, { "epoch": 5.5, "learning_rate": 1.7757810781382875e-07, "loss": 0.0126, "step": 13946 }, { "epoch": 5.5, "learning_rate": 1.7755273586053245e-07, "loss": 0.0132, "step": 13947 }, { "epoch": 5.5, "learning_rate": 1.775273642730925e-07, "loss": 0.0129, "step": 13948 }, { "epoch": 5.5, "learning_rate": 1.7750199305192224e-07, "loss": 0.0131, "step": 13949 }, { "epoch": 5.5, "learning_rate": 1.7747662219743539e-07, "loss": 0.0125, "step": 13950 }, { "epoch": 5.5, "learning_rate": 1.7745125171004536e-07, "loss": 0.0114, "step": 13951 }, { "epoch": 5.5, "learning_rate": 1.7742588159016566e-07, "loss": 0.0123, "step": 13952 }, { "epoch": 5.5, "learning_rate": 1.774005118382097e-07, "loss": 0.0144, "step": 13953 }, { "epoch": 5.5, "learning_rate": 1.7737514245459115e-07, "loss": 0.0143, "step": 13954 }, { "epoch": 5.5, "learning_rate": 1.7734977343972337e-07, "loss": 0.0158, "step": 13955 }, { "epoch": 5.5, "learning_rate": 1.7732440479401987e-07, "loss": 0.0124, "step": 13956 }, { "epoch": 5.5, "learning_rate": 1.772990365178941e-07, "loss": 0.0151, "step": 13957 }, { "epoch": 5.5, "learning_rate": 1.7727366861175945e-07, "loss": 0.0162, "step": 13958 }, { "epoch": 5.5, "learning_rate": 1.7724830107602955e-07, "loss": 0.0096, "step": 13959 }, { "epoch": 5.5, "learning_rate": 1.7722293391111779e-07, "loss": 0.0151, "step": 13960 }, { "epoch": 5.5, "learning_rate": 1.7719756711743753e-07, "loss": 0.0121, "step": 13961 }, { "epoch": 5.5, "learning_rate": 1.7717220069540227e-07, "loss": 0.0145, "step": 13962 }, { "epoch": 5.5, "learning_rate": 1.7714683464542545e-07, "loss": 0.0132, "step": 13963 }, { "epoch": 5.5, "learning_rate": 1.7712146896792053e-07, "loss": 0.0165, "step": 13964 }, { "epoch": 5.5, "learning_rate": 1.7709610366330088e-07, "loss": 0.0136, "step": 13965 }, { "epoch": 5.5, "learning_rate": 1.7707073873197992e-07, "loss": 0.0132, "step": 13966 }, { "epoch": 5.5, "learning_rate": 1.7704537417437106e-07, "loss": 0.0145, "step": 13967 }, { "epoch": 5.5, "learning_rate": 1.7702000999088774e-07, "loss": 0.0135, "step": 13968 }, { "epoch": 5.51, "learning_rate": 1.7699464618194333e-07, "loss": 0.0127, "step": 13969 }, { "epoch": 5.51, "learning_rate": 1.7696928274795123e-07, "loss": 0.0148, "step": 13970 }, { "epoch": 5.51, "learning_rate": 1.7694391968932473e-07, "loss": 0.0163, "step": 13971 }, { "epoch": 5.51, "learning_rate": 1.7691855700647737e-07, "loss": 0.0129, "step": 13972 }, { "epoch": 5.51, "learning_rate": 1.7689319469982247e-07, "loss": 0.0133, "step": 13973 }, { "epoch": 5.51, "learning_rate": 1.7686783276977337e-07, "loss": 0.0128, "step": 13974 }, { "epoch": 5.51, "learning_rate": 1.7684247121674345e-07, "loss": 0.0125, "step": 13975 }, { "epoch": 5.51, "learning_rate": 1.7681711004114594e-07, "loss": 0.0134, "step": 13976 }, { "epoch": 5.51, "learning_rate": 1.7679174924339442e-07, "loss": 0.0149, "step": 13977 }, { "epoch": 5.51, "learning_rate": 1.7676638882390208e-07, "loss": 0.012, "step": 13978 }, { "epoch": 5.51, "learning_rate": 1.7674102878308233e-07, "loss": 0.0157, "step": 13979 }, { "epoch": 5.51, "learning_rate": 1.7671566912134837e-07, "loss": 0.015, "step": 13980 }, { "epoch": 5.51, "learning_rate": 1.7669030983911368e-07, "loss": 0.0145, "step": 13981 }, { "epoch": 5.51, "learning_rate": 1.7666495093679152e-07, "loss": 0.0123, "step": 13982 }, { "epoch": 5.51, "learning_rate": 1.766395924147952e-07, "loss": 0.0101, "step": 13983 }, { "epoch": 5.51, "learning_rate": 1.7661423427353796e-07, "loss": 0.0113, "step": 13984 }, { "epoch": 5.51, "learning_rate": 1.7658887651343322e-07, "loss": 0.0165, "step": 13985 }, { "epoch": 5.51, "learning_rate": 1.7656351913489417e-07, "loss": 0.0131, "step": 13986 }, { "epoch": 5.51, "learning_rate": 1.7653816213833416e-07, "loss": 0.0133, "step": 13987 }, { "epoch": 5.51, "learning_rate": 1.7651280552416648e-07, "loss": 0.0134, "step": 13988 }, { "epoch": 5.51, "learning_rate": 1.7648744929280433e-07, "loss": 0.015, "step": 13989 }, { "epoch": 5.51, "learning_rate": 1.7646209344466104e-07, "loss": 0.0142, "step": 13990 }, { "epoch": 5.51, "learning_rate": 1.7643673798014987e-07, "loss": 0.0164, "step": 13991 }, { "epoch": 5.51, "learning_rate": 1.7641138289968405e-07, "loss": 0.0149, "step": 13992 }, { "epoch": 5.51, "learning_rate": 1.7638602820367675e-07, "loss": 0.0134, "step": 13993 }, { "epoch": 5.52, "learning_rate": 1.7636067389254139e-07, "loss": 0.0134, "step": 13994 }, { "epoch": 5.52, "learning_rate": 1.7633531996669114e-07, "loss": 0.0121, "step": 13995 }, { "epoch": 5.52, "learning_rate": 1.7630996642653917e-07, "loss": 0.0118, "step": 13996 }, { "epoch": 5.52, "learning_rate": 1.7628461327249878e-07, "loss": 0.0113, "step": 13997 }, { "epoch": 5.52, "learning_rate": 1.7625926050498302e-07, "loss": 0.0139, "step": 13998 }, { "epoch": 5.52, "learning_rate": 1.7623390812440534e-07, "loss": 0.0164, "step": 13999 }, { "epoch": 5.52, "learning_rate": 1.7620855613117884e-07, "loss": 0.0149, "step": 14000 }, { "epoch": 5.52, "learning_rate": 1.7618320452571673e-07, "loss": 0.0167, "step": 14001 }, { "epoch": 5.52, "learning_rate": 1.761578533084321e-07, "loss": 0.0135, "step": 14002 }, { "epoch": 5.52, "learning_rate": 1.761325024797383e-07, "loss": 0.0138, "step": 14003 }, { "epoch": 5.52, "learning_rate": 1.7610715204004843e-07, "loss": 0.0149, "step": 14004 }, { "epoch": 5.52, "learning_rate": 1.7608180198977567e-07, "loss": 0.0137, "step": 14005 }, { "epoch": 5.52, "learning_rate": 1.7605645232933318e-07, "loss": 0.0131, "step": 14006 }, { "epoch": 5.52, "learning_rate": 1.7603110305913412e-07, "loss": 0.0113, "step": 14007 }, { "epoch": 5.52, "learning_rate": 1.7600575417959166e-07, "loss": 0.0128, "step": 14008 }, { "epoch": 5.52, "learning_rate": 1.7598040569111888e-07, "loss": 0.0105, "step": 14009 }, { "epoch": 5.52, "learning_rate": 1.7595505759412906e-07, "loss": 0.0157, "step": 14010 }, { "epoch": 5.52, "learning_rate": 1.759297098890352e-07, "loss": 0.0126, "step": 14011 }, { "epoch": 5.52, "learning_rate": 1.7590436257625057e-07, "loss": 0.013, "step": 14012 }, { "epoch": 5.52, "learning_rate": 1.7587901565618818e-07, "loss": 0.0133, "step": 14013 }, { "epoch": 5.52, "learning_rate": 1.7585366912926114e-07, "loss": 0.0156, "step": 14014 }, { "epoch": 5.52, "learning_rate": 1.7582832299588264e-07, "loss": 0.0107, "step": 14015 }, { "epoch": 5.52, "learning_rate": 1.7580297725646561e-07, "loss": 0.0146, "step": 14016 }, { "epoch": 5.52, "learning_rate": 1.757776319114234e-07, "loss": 0.0156, "step": 14017 }, { "epoch": 5.52, "learning_rate": 1.7575228696116897e-07, "loss": 0.0129, "step": 14018 }, { "epoch": 5.53, "learning_rate": 1.757269424061154e-07, "loss": 0.0109, "step": 14019 }, { "epoch": 5.53, "learning_rate": 1.757015982466757e-07, "loss": 0.0114, "step": 14020 }, { "epoch": 5.53, "learning_rate": 1.7567625448326312e-07, "loss": 0.0136, "step": 14021 }, { "epoch": 5.53, "learning_rate": 1.7565091111629057e-07, "loss": 0.0145, "step": 14022 }, { "epoch": 5.53, "learning_rate": 1.7562556814617121e-07, "loss": 0.0125, "step": 14023 }, { "epoch": 5.53, "learning_rate": 1.7560022557331796e-07, "loss": 0.0129, "step": 14024 }, { "epoch": 5.53, "learning_rate": 1.75574883398144e-07, "loss": 0.0114, "step": 14025 }, { "epoch": 5.53, "learning_rate": 1.7554954162106235e-07, "loss": 0.0122, "step": 14026 }, { "epoch": 5.53, "learning_rate": 1.7552420024248599e-07, "loss": 0.0134, "step": 14027 }, { "epoch": 5.53, "learning_rate": 1.75498859262828e-07, "loss": 0.0161, "step": 14028 }, { "epoch": 5.53, "learning_rate": 1.754735186825013e-07, "loss": 0.0154, "step": 14029 }, { "epoch": 5.53, "learning_rate": 1.7544817850191902e-07, "loss": 0.0144, "step": 14030 }, { "epoch": 5.53, "learning_rate": 1.754228387214941e-07, "loss": 0.0147, "step": 14031 }, { "epoch": 5.53, "learning_rate": 1.7539749934163952e-07, "loss": 0.0132, "step": 14032 }, { "epoch": 5.53, "learning_rate": 1.7537216036276832e-07, "loss": 0.012, "step": 14033 }, { "epoch": 5.53, "learning_rate": 1.7534682178529348e-07, "loss": 0.0118, "step": 14034 }, { "epoch": 5.53, "learning_rate": 1.7532148360962805e-07, "loss": 0.0128, "step": 14035 }, { "epoch": 5.53, "learning_rate": 1.7529614583618487e-07, "loss": 0.0144, "step": 14036 }, { "epoch": 5.53, "learning_rate": 1.7527080846537697e-07, "loss": 0.0149, "step": 14037 }, { "epoch": 5.53, "learning_rate": 1.7524547149761726e-07, "loss": 0.0133, "step": 14038 }, { "epoch": 5.53, "learning_rate": 1.752201349333188e-07, "loss": 0.0113, "step": 14039 }, { "epoch": 5.53, "learning_rate": 1.751947987728945e-07, "loss": 0.011, "step": 14040 }, { "epoch": 5.53, "learning_rate": 1.7516946301675726e-07, "loss": 0.0125, "step": 14041 }, { "epoch": 5.53, "learning_rate": 1.7514412766531993e-07, "loss": 0.0143, "step": 14042 }, { "epoch": 5.53, "learning_rate": 1.7511879271899567e-07, "loss": 0.0142, "step": 14043 }, { "epoch": 5.53, "learning_rate": 1.7509345817819724e-07, "loss": 0.0145, "step": 14044 }, { "epoch": 5.54, "learning_rate": 1.750681240433376e-07, "loss": 0.0161, "step": 14045 }, { "epoch": 5.54, "learning_rate": 1.7504279031482967e-07, "loss": 0.0146, "step": 14046 }, { "epoch": 5.54, "learning_rate": 1.7501745699308623e-07, "loss": 0.0148, "step": 14047 }, { "epoch": 5.54, "learning_rate": 1.7499212407852036e-07, "loss": 0.016, "step": 14048 }, { "epoch": 5.54, "learning_rate": 1.7496679157154484e-07, "loss": 0.0114, "step": 14049 }, { "epoch": 5.54, "learning_rate": 1.7494145947257257e-07, "loss": 0.0112, "step": 14050 }, { "epoch": 5.54, "learning_rate": 1.749161277820164e-07, "loss": 0.0103, "step": 14051 }, { "epoch": 5.54, "learning_rate": 1.7489079650028927e-07, "loss": 0.0124, "step": 14052 }, { "epoch": 5.54, "learning_rate": 1.7486546562780398e-07, "loss": 0.0167, "step": 14053 }, { "epoch": 5.54, "learning_rate": 1.7484013516497336e-07, "loss": 0.0124, "step": 14054 }, { "epoch": 5.54, "learning_rate": 1.7481480511221035e-07, "loss": 0.013, "step": 14055 }, { "epoch": 5.54, "learning_rate": 1.7478947546992768e-07, "loss": 0.0143, "step": 14056 }, { "epoch": 5.54, "learning_rate": 1.747641462385383e-07, "loss": 0.0112, "step": 14057 }, { "epoch": 5.54, "learning_rate": 1.74738817418455e-07, "loss": 0.0147, "step": 14058 }, { "epoch": 5.54, "learning_rate": 1.7471348901009056e-07, "loss": 0.0148, "step": 14059 }, { "epoch": 5.54, "learning_rate": 1.7468816101385773e-07, "loss": 0.0125, "step": 14060 }, { "epoch": 5.54, "learning_rate": 1.746628334301695e-07, "loss": 0.0114, "step": 14061 }, { "epoch": 5.54, "learning_rate": 1.7463750625943858e-07, "loss": 0.0128, "step": 14062 }, { "epoch": 5.54, "learning_rate": 1.7461217950207776e-07, "loss": 0.0138, "step": 14063 }, { "epoch": 5.54, "learning_rate": 1.745868531584997e-07, "loss": 0.013, "step": 14064 }, { "epoch": 5.54, "learning_rate": 1.7456152722911747e-07, "loss": 0.0155, "step": 14065 }, { "epoch": 5.54, "learning_rate": 1.7453620171434366e-07, "loss": 0.013, "step": 14066 }, { "epoch": 5.54, "learning_rate": 1.7451087661459103e-07, "loss": 0.0145, "step": 14067 }, { "epoch": 5.54, "learning_rate": 1.7448555193027241e-07, "loss": 0.012, "step": 14068 }, { "epoch": 5.54, "learning_rate": 1.7446022766180039e-07, "loss": 0.0125, "step": 14069 }, { "epoch": 5.55, "learning_rate": 1.7443490380958798e-07, "loss": 0.0115, "step": 14070 }, { "epoch": 5.55, "learning_rate": 1.7440958037404773e-07, "loss": 0.0111, "step": 14071 }, { "epoch": 5.55, "learning_rate": 1.743842573555925e-07, "loss": 0.0146, "step": 14072 }, { "epoch": 5.55, "learning_rate": 1.7435893475463484e-07, "loss": 0.015, "step": 14073 }, { "epoch": 5.55, "learning_rate": 1.743336125715876e-07, "loss": 0.0153, "step": 14074 }, { "epoch": 5.55, "learning_rate": 1.7430829080686352e-07, "loss": 0.0117, "step": 14075 }, { "epoch": 5.55, "learning_rate": 1.7428296946087523e-07, "loss": 0.013, "step": 14076 }, { "epoch": 5.55, "learning_rate": 1.7425764853403542e-07, "loss": 0.0112, "step": 14077 }, { "epoch": 5.55, "learning_rate": 1.742323280267568e-07, "loss": 0.0155, "step": 14078 }, { "epoch": 5.55, "learning_rate": 1.7420700793945212e-07, "loss": 0.0144, "step": 14079 }, { "epoch": 5.55, "learning_rate": 1.7418168827253399e-07, "loss": 0.0133, "step": 14080 }, { "epoch": 5.55, "learning_rate": 1.7415636902641512e-07, "loss": 0.0132, "step": 14081 }, { "epoch": 5.55, "learning_rate": 1.741310502015081e-07, "loss": 0.0119, "step": 14082 }, { "epoch": 5.55, "learning_rate": 1.7410573179822568e-07, "loss": 0.0158, "step": 14083 }, { "epoch": 5.55, "learning_rate": 1.740804138169805e-07, "loss": 0.0134, "step": 14084 }, { "epoch": 5.55, "learning_rate": 1.7405509625818518e-07, "loss": 0.0115, "step": 14085 }, { "epoch": 5.55, "learning_rate": 1.7402977912225237e-07, "loss": 0.0144, "step": 14086 }, { "epoch": 5.55, "learning_rate": 1.740044624095946e-07, "loss": 0.012, "step": 14087 }, { "epoch": 5.55, "learning_rate": 1.7397914612062466e-07, "loss": 0.0127, "step": 14088 }, { "epoch": 5.55, "learning_rate": 1.739538302557551e-07, "loss": 0.0113, "step": 14089 }, { "epoch": 5.55, "learning_rate": 1.7392851481539852e-07, "loss": 0.0126, "step": 14090 }, { "epoch": 5.55, "learning_rate": 1.739031997999674e-07, "loss": 0.0126, "step": 14091 }, { "epoch": 5.55, "learning_rate": 1.7387788520987462e-07, "loss": 0.0139, "step": 14092 }, { "epoch": 5.55, "learning_rate": 1.7385257104553257e-07, "loss": 0.0166, "step": 14093 }, { "epoch": 5.55, "learning_rate": 1.7382725730735386e-07, "loss": 0.0135, "step": 14094 }, { "epoch": 5.55, "learning_rate": 1.738019439957511e-07, "loss": 0.014, "step": 14095 }, { "epoch": 5.56, "learning_rate": 1.7377663111113678e-07, "loss": 0.0127, "step": 14096 }, { "epoch": 5.56, "learning_rate": 1.7375131865392356e-07, "loss": 0.0133, "step": 14097 }, { "epoch": 5.56, "learning_rate": 1.7372600662452398e-07, "loss": 0.0152, "step": 14098 }, { "epoch": 5.56, "learning_rate": 1.737006950233505e-07, "loss": 0.0128, "step": 14099 }, { "epoch": 5.56, "learning_rate": 1.736753838508157e-07, "loss": 0.015, "step": 14100 }, { "epoch": 5.56, "learning_rate": 1.736500731073322e-07, "loss": 0.0111, "step": 14101 }, { "epoch": 5.56, "learning_rate": 1.7362476279331245e-07, "loss": 0.0121, "step": 14102 }, { "epoch": 5.56, "learning_rate": 1.73599452909169e-07, "loss": 0.0167, "step": 14103 }, { "epoch": 5.56, "learning_rate": 1.7357414345531423e-07, "loss": 0.0133, "step": 14104 }, { "epoch": 5.56, "learning_rate": 1.7354883443216086e-07, "loss": 0.014, "step": 14105 }, { "epoch": 5.56, "learning_rate": 1.735235258401213e-07, "loss": 0.0104, "step": 14106 }, { "epoch": 5.56, "learning_rate": 1.73498217679608e-07, "loss": 0.0141, "step": 14107 }, { "epoch": 5.56, "learning_rate": 1.734729099510335e-07, "loss": 0.0135, "step": 14108 }, { "epoch": 5.56, "learning_rate": 1.7344760265481017e-07, "loss": 0.0119, "step": 14109 }, { "epoch": 5.56, "learning_rate": 1.734222957913506e-07, "loss": 0.0133, "step": 14110 }, { "epoch": 5.56, "learning_rate": 1.7339698936106726e-07, "loss": 0.0128, "step": 14111 }, { "epoch": 5.56, "learning_rate": 1.7337168336437254e-07, "loss": 0.0113, "step": 14112 }, { "epoch": 5.56, "learning_rate": 1.733463778016789e-07, "loss": 0.012, "step": 14113 }, { "epoch": 5.56, "learning_rate": 1.7332107267339875e-07, "loss": 0.0147, "step": 14114 }, { "epoch": 5.56, "learning_rate": 1.7329576797994464e-07, "loss": 0.0126, "step": 14115 }, { "epoch": 5.56, "learning_rate": 1.732704637217289e-07, "loss": 0.0121, "step": 14116 }, { "epoch": 5.56, "learning_rate": 1.73245159899164e-07, "loss": 0.0134, "step": 14117 }, { "epoch": 5.56, "learning_rate": 1.732198565126623e-07, "loss": 0.0137, "step": 14118 }, { "epoch": 5.56, "learning_rate": 1.7319455356263627e-07, "loss": 0.0119, "step": 14119 }, { "epoch": 5.56, "learning_rate": 1.7316925104949824e-07, "loss": 0.0137, "step": 14120 }, { "epoch": 5.57, "learning_rate": 1.731439489736607e-07, "loss": 0.0152, "step": 14121 }, { "epoch": 5.57, "learning_rate": 1.7311864733553584e-07, "loss": 0.0112, "step": 14122 }, { "epoch": 5.57, "learning_rate": 1.7309334613553627e-07, "loss": 0.0152, "step": 14123 }, { "epoch": 5.57, "learning_rate": 1.730680453740743e-07, "loss": 0.0151, "step": 14124 }, { "epoch": 5.57, "learning_rate": 1.7304274505156223e-07, "loss": 0.0144, "step": 14125 }, { "epoch": 5.57, "learning_rate": 1.7301744516841243e-07, "loss": 0.0141, "step": 14126 }, { "epoch": 5.57, "learning_rate": 1.729921457250372e-07, "loss": 0.0145, "step": 14127 }, { "epoch": 5.57, "learning_rate": 1.7296684672184903e-07, "loss": 0.0135, "step": 14128 }, { "epoch": 5.57, "learning_rate": 1.7294154815926016e-07, "loss": 0.0152, "step": 14129 }, { "epoch": 5.57, "learning_rate": 1.7291625003768295e-07, "loss": 0.0135, "step": 14130 }, { "epoch": 5.57, "learning_rate": 1.728909523575296e-07, "loss": 0.0091, "step": 14131 }, { "epoch": 5.57, "learning_rate": 1.7286565511921266e-07, "loss": 0.0127, "step": 14132 }, { "epoch": 5.57, "learning_rate": 1.7284035832314424e-07, "loss": 0.0111, "step": 14133 }, { "epoch": 5.57, "learning_rate": 1.7281506196973675e-07, "loss": 0.0132, "step": 14134 }, { "epoch": 5.57, "learning_rate": 1.727897660594024e-07, "loss": 0.0157, "step": 14135 }, { "epoch": 5.57, "learning_rate": 1.7276447059255345e-07, "loss": 0.0144, "step": 14136 }, { "epoch": 5.57, "learning_rate": 1.7273917556960226e-07, "loss": 0.0114, "step": 14137 }, { "epoch": 5.57, "learning_rate": 1.7271388099096114e-07, "loss": 0.012, "step": 14138 }, { "epoch": 5.57, "learning_rate": 1.7268858685704226e-07, "loss": 0.011, "step": 14139 }, { "epoch": 5.57, "learning_rate": 1.7266329316825787e-07, "loss": 0.0107, "step": 14140 }, { "epoch": 5.57, "learning_rate": 1.726379999250203e-07, "loss": 0.0143, "step": 14141 }, { "epoch": 5.57, "learning_rate": 1.7261270712774173e-07, "loss": 0.013, "step": 14142 }, { "epoch": 5.57, "learning_rate": 1.7258741477683443e-07, "loss": 0.0162, "step": 14143 }, { "epoch": 5.57, "learning_rate": 1.725621228727105e-07, "loss": 0.0143, "step": 14144 }, { "epoch": 5.57, "learning_rate": 1.725368314157824e-07, "loss": 0.0134, "step": 14145 }, { "epoch": 5.58, "learning_rate": 1.7251154040646216e-07, "loss": 0.0137, "step": 14146 }, { "epoch": 5.58, "learning_rate": 1.7248624984516207e-07, "loss": 0.015, "step": 14147 }, { "epoch": 5.58, "learning_rate": 1.7246095973229427e-07, "loss": 0.0127, "step": 14148 }, { "epoch": 5.58, "learning_rate": 1.724356700682709e-07, "loss": 0.0143, "step": 14149 }, { "epoch": 5.58, "learning_rate": 1.7241038085350427e-07, "loss": 0.0152, "step": 14150 }, { "epoch": 5.58, "learning_rate": 1.723850920884065e-07, "loss": 0.0111, "step": 14151 }, { "epoch": 5.58, "learning_rate": 1.723598037733898e-07, "loss": 0.0159, "step": 14152 }, { "epoch": 5.58, "learning_rate": 1.723345159088662e-07, "loss": 0.0127, "step": 14153 }, { "epoch": 5.58, "learning_rate": 1.7230922849524798e-07, "loss": 0.0132, "step": 14154 }, { "epoch": 5.58, "learning_rate": 1.722839415329473e-07, "loss": 0.0131, "step": 14155 }, { "epoch": 5.58, "learning_rate": 1.7225865502237623e-07, "loss": 0.0132, "step": 14156 }, { "epoch": 5.58, "learning_rate": 1.722333689639469e-07, "loss": 0.0116, "step": 14157 }, { "epoch": 5.58, "learning_rate": 1.7220808335807146e-07, "loss": 0.0107, "step": 14158 }, { "epoch": 5.58, "learning_rate": 1.7218279820516203e-07, "loss": 0.0127, "step": 14159 }, { "epoch": 5.58, "learning_rate": 1.7215751350563069e-07, "loss": 0.0132, "step": 14160 }, { "epoch": 5.58, "learning_rate": 1.721322292598896e-07, "loss": 0.0089, "step": 14161 }, { "epoch": 5.58, "learning_rate": 1.7210694546835075e-07, "loss": 0.0151, "step": 14162 }, { "epoch": 5.58, "learning_rate": 1.7208166213142638e-07, "loss": 0.0116, "step": 14163 }, { "epoch": 5.58, "learning_rate": 1.7205637924952848e-07, "loss": 0.0139, "step": 14164 }, { "epoch": 5.58, "learning_rate": 1.720310968230691e-07, "loss": 0.0142, "step": 14165 }, { "epoch": 5.58, "learning_rate": 1.7200581485246036e-07, "loss": 0.0142, "step": 14166 }, { "epoch": 5.58, "learning_rate": 1.719805333381142e-07, "loss": 0.0133, "step": 14167 }, { "epoch": 5.58, "learning_rate": 1.7195525228044284e-07, "loss": 0.0135, "step": 14168 }, { "epoch": 5.58, "learning_rate": 1.7192997167985824e-07, "loss": 0.0131, "step": 14169 }, { "epoch": 5.58, "learning_rate": 1.7190469153677248e-07, "loss": 0.0143, "step": 14170 }, { "epoch": 5.58, "learning_rate": 1.7187941185159743e-07, "loss": 0.0134, "step": 14171 }, { "epoch": 5.59, "learning_rate": 1.718541326247453e-07, "loss": 0.0098, "step": 14172 }, { "epoch": 5.59, "learning_rate": 1.7182885385662808e-07, "loss": 0.0116, "step": 14173 }, { "epoch": 5.59, "learning_rate": 1.7180357554765766e-07, "loss": 0.0139, "step": 14174 }, { "epoch": 5.59, "learning_rate": 1.7177829769824614e-07, "loss": 0.0147, "step": 14175 }, { "epoch": 5.59, "learning_rate": 1.7175302030880538e-07, "loss": 0.0133, "step": 14176 }, { "epoch": 5.59, "learning_rate": 1.717277433797476e-07, "loss": 0.0129, "step": 14177 }, { "epoch": 5.59, "learning_rate": 1.7170246691148457e-07, "loss": 0.0164, "step": 14178 }, { "epoch": 5.59, "learning_rate": 1.7167719090442833e-07, "loss": 0.0157, "step": 14179 }, { "epoch": 5.59, "learning_rate": 1.716519153589908e-07, "loss": 0.0126, "step": 14180 }, { "epoch": 5.59, "learning_rate": 1.71626640275584e-07, "loss": 0.0135, "step": 14181 }, { "epoch": 5.59, "learning_rate": 1.716013656546198e-07, "loss": 0.0121, "step": 14182 }, { "epoch": 5.59, "learning_rate": 1.7157609149651024e-07, "loss": 0.0113, "step": 14183 }, { "epoch": 5.59, "learning_rate": 1.7155081780166713e-07, "loss": 0.0133, "step": 14184 }, { "epoch": 5.59, "learning_rate": 1.715255445705025e-07, "loss": 0.0153, "step": 14185 }, { "epoch": 5.59, "learning_rate": 1.7150027180342824e-07, "loss": 0.0122, "step": 14186 }, { "epoch": 5.59, "learning_rate": 1.714749995008562e-07, "loss": 0.0134, "step": 14187 }, { "epoch": 5.59, "learning_rate": 1.7144972766319832e-07, "loss": 0.0137, "step": 14188 }, { "epoch": 5.59, "learning_rate": 1.714244562908664e-07, "loss": 0.0162, "step": 14189 }, { "epoch": 5.59, "learning_rate": 1.713991853842725e-07, "loss": 0.0139, "step": 14190 }, { "epoch": 5.59, "learning_rate": 1.7137391494382842e-07, "loss": 0.014, "step": 14191 }, { "epoch": 5.59, "learning_rate": 1.7134864496994604e-07, "loss": 0.0137, "step": 14192 }, { "epoch": 5.59, "learning_rate": 1.713233754630371e-07, "loss": 0.0148, "step": 14193 }, { "epoch": 5.59, "learning_rate": 1.712981064235136e-07, "loss": 0.0151, "step": 14194 }, { "epoch": 5.59, "learning_rate": 1.7127283785178741e-07, "loss": 0.014, "step": 14195 }, { "epoch": 5.59, "learning_rate": 1.7124756974827025e-07, "loss": 0.014, "step": 14196 }, { "epoch": 5.6, "learning_rate": 1.7122230211337402e-07, "loss": 0.0138, "step": 14197 }, { "epoch": 5.6, "learning_rate": 1.7119703494751047e-07, "loss": 0.012, "step": 14198 }, { "epoch": 5.6, "learning_rate": 1.7117176825109153e-07, "loss": 0.0132, "step": 14199 }, { "epoch": 5.6, "learning_rate": 1.7114650202452897e-07, "loss": 0.0134, "step": 14200 }, { "epoch": 5.6, "learning_rate": 1.7112123626823455e-07, "loss": 0.0144, "step": 14201 }, { "epoch": 5.6, "learning_rate": 1.710959709826201e-07, "loss": 0.014, "step": 14202 }, { "epoch": 5.6, "learning_rate": 1.7107070616809737e-07, "loss": 0.0118, "step": 14203 }, { "epoch": 5.6, "learning_rate": 1.7104544182507822e-07, "loss": 0.0162, "step": 14204 }, { "epoch": 5.6, "learning_rate": 1.7102017795397428e-07, "loss": 0.012, "step": 14205 }, { "epoch": 5.6, "learning_rate": 1.7099491455519744e-07, "loss": 0.0135, "step": 14206 }, { "epoch": 5.6, "learning_rate": 1.7096965162915938e-07, "loss": 0.0128, "step": 14207 }, { "epoch": 5.6, "learning_rate": 1.7094438917627193e-07, "loss": 0.0129, "step": 14208 }, { "epoch": 5.6, "learning_rate": 1.7091912719694674e-07, "loss": 0.0121, "step": 14209 }, { "epoch": 5.6, "learning_rate": 1.708938656915956e-07, "loss": 0.0156, "step": 14210 }, { "epoch": 5.6, "learning_rate": 1.7086860466063012e-07, "loss": 0.0161, "step": 14211 }, { "epoch": 5.6, "learning_rate": 1.708433441044622e-07, "loss": 0.0106, "step": 14212 }, { "epoch": 5.6, "learning_rate": 1.7081808402350347e-07, "loss": 0.0118, "step": 14213 }, { "epoch": 5.6, "learning_rate": 1.7079282441816562e-07, "loss": 0.0117, "step": 14214 }, { "epoch": 5.6, "learning_rate": 1.7076756528886033e-07, "loss": 0.0112, "step": 14215 }, { "epoch": 5.6, "learning_rate": 1.7074230663599922e-07, "loss": 0.0118, "step": 14216 }, { "epoch": 5.6, "learning_rate": 1.7071704845999415e-07, "loss": 0.0137, "step": 14217 }, { "epoch": 5.6, "learning_rate": 1.7069179076125666e-07, "loss": 0.0165, "step": 14218 }, { "epoch": 5.6, "learning_rate": 1.7066653354019844e-07, "loss": 0.0135, "step": 14219 }, { "epoch": 5.6, "learning_rate": 1.7064127679723103e-07, "loss": 0.0123, "step": 14220 }, { "epoch": 5.6, "learning_rate": 1.7061602053276632e-07, "loss": 0.0117, "step": 14221 }, { "epoch": 5.61, "learning_rate": 1.705907647472158e-07, "loss": 0.014, "step": 14222 }, { "epoch": 5.61, "learning_rate": 1.7056550944099112e-07, "loss": 0.0144, "step": 14223 }, { "epoch": 5.61, "learning_rate": 1.7054025461450383e-07, "loss": 0.0127, "step": 14224 }, { "epoch": 5.61, "learning_rate": 1.7051500026816567e-07, "loss": 0.0128, "step": 14225 }, { "epoch": 5.61, "learning_rate": 1.7048974640238823e-07, "loss": 0.0125, "step": 14226 }, { "epoch": 5.61, "learning_rate": 1.7046449301758303e-07, "loss": 0.0151, "step": 14227 }, { "epoch": 5.61, "learning_rate": 1.7043924011416168e-07, "loss": 0.0136, "step": 14228 }, { "epoch": 5.61, "learning_rate": 1.7041398769253582e-07, "loss": 0.0159, "step": 14229 }, { "epoch": 5.61, "learning_rate": 1.7038873575311697e-07, "loss": 0.0156, "step": 14230 }, { "epoch": 5.61, "learning_rate": 1.7036348429631678e-07, "loss": 0.0141, "step": 14231 }, { "epoch": 5.61, "learning_rate": 1.7033823332254673e-07, "loss": 0.0178, "step": 14232 }, { "epoch": 5.61, "learning_rate": 1.703129828322183e-07, "loss": 0.013, "step": 14233 }, { "epoch": 5.61, "learning_rate": 1.7028773282574325e-07, "loss": 0.0121, "step": 14234 }, { "epoch": 5.61, "learning_rate": 1.7026248330353296e-07, "loss": 0.0139, "step": 14235 }, { "epoch": 5.61, "learning_rate": 1.7023723426599904e-07, "loss": 0.0137, "step": 14236 }, { "epoch": 5.61, "learning_rate": 1.7021198571355293e-07, "loss": 0.0113, "step": 14237 }, { "epoch": 5.61, "learning_rate": 1.701867376466061e-07, "loss": 0.0127, "step": 14238 }, { "epoch": 5.61, "learning_rate": 1.7016149006557024e-07, "loss": 0.0159, "step": 14239 }, { "epoch": 5.61, "learning_rate": 1.701362429708567e-07, "loss": 0.0143, "step": 14240 }, { "epoch": 5.61, "learning_rate": 1.7011099636287706e-07, "loss": 0.0136, "step": 14241 }, { "epoch": 5.61, "learning_rate": 1.7008575024204265e-07, "loss": 0.0149, "step": 14242 }, { "epoch": 5.61, "learning_rate": 1.7006050460876512e-07, "loss": 0.0172, "step": 14243 }, { "epoch": 5.61, "learning_rate": 1.700352594634559e-07, "loss": 0.0157, "step": 14244 }, { "epoch": 5.61, "learning_rate": 1.7001001480652637e-07, "loss": 0.0126, "step": 14245 }, { "epoch": 5.61, "learning_rate": 1.6998477063838804e-07, "loss": 0.0126, "step": 14246 }, { "epoch": 5.61, "learning_rate": 1.699595269594523e-07, "loss": 0.0115, "step": 14247 }, { "epoch": 5.62, "learning_rate": 1.6993428377013063e-07, "loss": 0.0123, "step": 14248 }, { "epoch": 5.62, "learning_rate": 1.6990904107083443e-07, "loss": 0.0122, "step": 14249 }, { "epoch": 5.62, "learning_rate": 1.6988379886197513e-07, "loss": 0.0156, "step": 14250 }, { "epoch": 5.62, "learning_rate": 1.698585571439641e-07, "loss": 0.013, "step": 14251 }, { "epoch": 5.62, "learning_rate": 1.6983331591721282e-07, "loss": 0.0124, "step": 14252 }, { "epoch": 5.62, "learning_rate": 1.6980807518213264e-07, "loss": 0.0127, "step": 14253 }, { "epoch": 5.62, "learning_rate": 1.6978283493913495e-07, "loss": 0.0126, "step": 14254 }, { "epoch": 5.62, "learning_rate": 1.6975759518863116e-07, "loss": 0.0132, "step": 14255 }, { "epoch": 5.62, "learning_rate": 1.697323559310325e-07, "loss": 0.013, "step": 14256 }, { "epoch": 5.62, "learning_rate": 1.697071171667505e-07, "loss": 0.0148, "step": 14257 }, { "epoch": 5.62, "learning_rate": 1.6968187889619646e-07, "loss": 0.0152, "step": 14258 }, { "epoch": 5.62, "learning_rate": 1.6965664111978175e-07, "loss": 0.0142, "step": 14259 }, { "epoch": 5.62, "learning_rate": 1.6963140383791757e-07, "loss": 0.0127, "step": 14260 }, { "epoch": 5.62, "learning_rate": 1.6960616705101543e-07, "loss": 0.0129, "step": 14261 }, { "epoch": 5.62, "learning_rate": 1.6958093075948656e-07, "loss": 0.0115, "step": 14262 }, { "epoch": 5.62, "learning_rate": 1.6955569496374232e-07, "loss": 0.0119, "step": 14263 }, { "epoch": 5.62, "learning_rate": 1.695304596641939e-07, "loss": 0.0138, "step": 14264 }, { "epoch": 5.62, "learning_rate": 1.695052248612527e-07, "loss": 0.0125, "step": 14265 }, { "epoch": 5.62, "learning_rate": 1.6947999055533004e-07, "loss": 0.0141, "step": 14266 }, { "epoch": 5.62, "learning_rate": 1.6945475674683715e-07, "loss": 0.0113, "step": 14267 }, { "epoch": 5.62, "learning_rate": 1.6942952343618527e-07, "loss": 0.0101, "step": 14268 }, { "epoch": 5.62, "learning_rate": 1.6940429062378568e-07, "loss": 0.0137, "step": 14269 }, { "epoch": 5.62, "learning_rate": 1.693790583100497e-07, "loss": 0.01, "step": 14270 }, { "epoch": 5.62, "learning_rate": 1.6935382649538852e-07, "loss": 0.0157, "step": 14271 }, { "epoch": 5.62, "learning_rate": 1.693285951802134e-07, "loss": 0.0108, "step": 14272 }, { "epoch": 5.63, "learning_rate": 1.6930336436493545e-07, "loss": 0.0162, "step": 14273 }, { "epoch": 5.63, "learning_rate": 1.6927813404996613e-07, "loss": 0.0142, "step": 14274 }, { "epoch": 5.63, "learning_rate": 1.6925290423571652e-07, "loss": 0.0153, "step": 14275 }, { "epoch": 5.63, "learning_rate": 1.692276749225978e-07, "loss": 0.0144, "step": 14276 }, { "epoch": 5.63, "learning_rate": 1.6920244611102128e-07, "loss": 0.0112, "step": 14277 }, { "epoch": 5.63, "learning_rate": 1.6917721780139796e-07, "loss": 0.0152, "step": 14278 }, { "epoch": 5.63, "learning_rate": 1.6915198999413924e-07, "loss": 0.0172, "step": 14279 }, { "epoch": 5.63, "learning_rate": 1.691267626896562e-07, "loss": 0.0141, "step": 14280 }, { "epoch": 5.63, "learning_rate": 1.6910153588835998e-07, "loss": 0.011, "step": 14281 }, { "epoch": 5.63, "learning_rate": 1.690763095906617e-07, "loss": 0.0137, "step": 14282 }, { "epoch": 5.63, "learning_rate": 1.6905108379697267e-07, "loss": 0.0163, "step": 14283 }, { "epoch": 5.63, "learning_rate": 1.6902585850770393e-07, "loss": 0.0112, "step": 14284 }, { "epoch": 5.63, "learning_rate": 1.690006337232666e-07, "loss": 0.0112, "step": 14285 }, { "epoch": 5.63, "learning_rate": 1.6897540944407184e-07, "loss": 0.0168, "step": 14286 }, { "epoch": 5.63, "learning_rate": 1.6895018567053073e-07, "loss": 0.0118, "step": 14287 }, { "epoch": 5.63, "learning_rate": 1.689249624030544e-07, "loss": 0.0139, "step": 14288 }, { "epoch": 5.63, "learning_rate": 1.6889973964205398e-07, "loss": 0.0159, "step": 14289 }, { "epoch": 5.63, "learning_rate": 1.6887451738794054e-07, "loss": 0.0132, "step": 14290 }, { "epoch": 5.63, "learning_rate": 1.688492956411251e-07, "loss": 0.0129, "step": 14291 }, { "epoch": 5.63, "learning_rate": 1.688240744020189e-07, "loss": 0.0164, "step": 14292 }, { "epoch": 5.63, "learning_rate": 1.6879885367103284e-07, "loss": 0.0113, "step": 14293 }, { "epoch": 5.63, "learning_rate": 1.6877363344857808e-07, "loss": 0.0097, "step": 14294 }, { "epoch": 5.63, "learning_rate": 1.6874841373506563e-07, "loss": 0.0137, "step": 14295 }, { "epoch": 5.63, "learning_rate": 1.6872319453090644e-07, "loss": 0.0113, "step": 14296 }, { "epoch": 5.63, "learning_rate": 1.6869797583651175e-07, "loss": 0.0146, "step": 14297 }, { "epoch": 5.63, "learning_rate": 1.6867275765229249e-07, "loss": 0.0121, "step": 14298 }, { "epoch": 5.64, "learning_rate": 1.6864753997865963e-07, "loss": 0.014, "step": 14299 }, { "epoch": 5.64, "learning_rate": 1.6862232281602415e-07, "loss": 0.0134, "step": 14300 }, { "epoch": 5.64, "learning_rate": 1.6859710616479722e-07, "loss": 0.0154, "step": 14301 }, { "epoch": 5.64, "learning_rate": 1.6857189002538971e-07, "loss": 0.015, "step": 14302 }, { "epoch": 5.64, "learning_rate": 1.6854667439821265e-07, "loss": 0.0135, "step": 14303 }, { "epoch": 5.64, "learning_rate": 1.6852145928367687e-07, "loss": 0.0138, "step": 14304 }, { "epoch": 5.64, "learning_rate": 1.684962446821936e-07, "loss": 0.014, "step": 14305 }, { "epoch": 5.64, "learning_rate": 1.6847103059417363e-07, "loss": 0.0118, "step": 14306 }, { "epoch": 5.64, "learning_rate": 1.6844581702002793e-07, "loss": 0.0137, "step": 14307 }, { "epoch": 5.64, "learning_rate": 1.684206039601675e-07, "loss": 0.0138, "step": 14308 }, { "epoch": 5.64, "learning_rate": 1.6839539141500314e-07, "loss": 0.0133, "step": 14309 }, { "epoch": 5.64, "learning_rate": 1.6837017938494588e-07, "loss": 0.0124, "step": 14310 }, { "epoch": 5.64, "learning_rate": 1.683449678704067e-07, "loss": 0.013, "step": 14311 }, { "epoch": 5.64, "learning_rate": 1.6831975687179638e-07, "loss": 0.0113, "step": 14312 }, { "epoch": 5.64, "learning_rate": 1.6829454638952587e-07, "loss": 0.0111, "step": 14313 }, { "epoch": 5.64, "learning_rate": 1.682693364240061e-07, "loss": 0.0139, "step": 14314 }, { "epoch": 5.64, "learning_rate": 1.6824412697564793e-07, "loss": 0.014, "step": 14315 }, { "epoch": 5.64, "learning_rate": 1.682189180448622e-07, "loss": 0.0135, "step": 14316 }, { "epoch": 5.64, "learning_rate": 1.6819370963205983e-07, "loss": 0.0143, "step": 14317 }, { "epoch": 5.64, "learning_rate": 1.6816850173765155e-07, "loss": 0.0156, "step": 14318 }, { "epoch": 5.64, "learning_rate": 1.681432943620484e-07, "loss": 0.0128, "step": 14319 }, { "epoch": 5.64, "learning_rate": 1.6811808750566114e-07, "loss": 0.0143, "step": 14320 }, { "epoch": 5.64, "learning_rate": 1.680928811689006e-07, "loss": 0.0153, "step": 14321 }, { "epoch": 5.64, "learning_rate": 1.6806767535217754e-07, "loss": 0.0115, "step": 14322 }, { "epoch": 5.64, "learning_rate": 1.680424700559029e-07, "loss": 0.014, "step": 14323 }, { "epoch": 5.65, "learning_rate": 1.6801726528048744e-07, "loss": 0.0144, "step": 14324 }, { "epoch": 5.65, "learning_rate": 1.6799206102634193e-07, "loss": 0.0133, "step": 14325 }, { "epoch": 5.65, "learning_rate": 1.679668572938772e-07, "loss": 0.0147, "step": 14326 }, { "epoch": 5.65, "learning_rate": 1.6794165408350392e-07, "loss": 0.011, "step": 14327 }, { "epoch": 5.65, "learning_rate": 1.6791645139563305e-07, "loss": 0.0145, "step": 14328 }, { "epoch": 5.65, "learning_rate": 1.6789124923067524e-07, "loss": 0.0183, "step": 14329 }, { "epoch": 5.65, "learning_rate": 1.6786604758904126e-07, "loss": 0.0119, "step": 14330 }, { "epoch": 5.65, "learning_rate": 1.6784084647114185e-07, "loss": 0.0135, "step": 14331 }, { "epoch": 5.65, "learning_rate": 1.6781564587738778e-07, "loss": 0.0132, "step": 14332 }, { "epoch": 5.65, "learning_rate": 1.6779044580818976e-07, "loss": 0.0117, "step": 14333 }, { "epoch": 5.65, "learning_rate": 1.6776524626395853e-07, "loss": 0.0136, "step": 14334 }, { "epoch": 5.65, "learning_rate": 1.6774004724510482e-07, "loss": 0.0115, "step": 14335 }, { "epoch": 5.65, "learning_rate": 1.6771484875203928e-07, "loss": 0.0132, "step": 14336 }, { "epoch": 5.65, "learning_rate": 1.6768965078517266e-07, "loss": 0.0109, "step": 14337 }, { "epoch": 5.65, "learning_rate": 1.6766445334491562e-07, "loss": 0.011, "step": 14338 }, { "epoch": 5.65, "learning_rate": 1.6763925643167884e-07, "loss": 0.0127, "step": 14339 }, { "epoch": 5.65, "learning_rate": 1.6761406004587294e-07, "loss": 0.0126, "step": 14340 }, { "epoch": 5.65, "learning_rate": 1.675888641879087e-07, "loss": 0.0152, "step": 14341 }, { "epoch": 5.65, "learning_rate": 1.6756366885819673e-07, "loss": 0.0126, "step": 14342 }, { "epoch": 5.65, "learning_rate": 1.6753847405714765e-07, "loss": 0.0112, "step": 14343 }, { "epoch": 5.65, "learning_rate": 1.6751327978517202e-07, "loss": 0.0126, "step": 14344 }, { "epoch": 5.65, "learning_rate": 1.6748808604268066e-07, "loss": 0.0153, "step": 14345 }, { "epoch": 5.65, "learning_rate": 1.6746289283008406e-07, "loss": 0.0123, "step": 14346 }, { "epoch": 5.65, "learning_rate": 1.6743770014779287e-07, "loss": 0.0131, "step": 14347 }, { "epoch": 5.65, "learning_rate": 1.6741250799621766e-07, "loss": 0.0157, "step": 14348 }, { "epoch": 5.66, "learning_rate": 1.6738731637576898e-07, "loss": 0.0118, "step": 14349 }, { "epoch": 5.66, "learning_rate": 1.6736212528685757e-07, "loss": 0.0129, "step": 14350 }, { "epoch": 5.66, "learning_rate": 1.6733693472989389e-07, "loss": 0.0116, "step": 14351 }, { "epoch": 5.66, "learning_rate": 1.6731174470528856e-07, "loss": 0.0143, "step": 14352 }, { "epoch": 5.66, "learning_rate": 1.6728655521345202e-07, "loss": 0.0128, "step": 14353 }, { "epoch": 5.66, "learning_rate": 1.6726136625479498e-07, "loss": 0.0155, "step": 14354 }, { "epoch": 5.66, "learning_rate": 1.6723617782972794e-07, "loss": 0.0133, "step": 14355 }, { "epoch": 5.66, "learning_rate": 1.6721098993866133e-07, "loss": 0.014, "step": 14356 }, { "epoch": 5.66, "learning_rate": 1.671858025820058e-07, "loss": 0.0121, "step": 14357 }, { "epoch": 5.66, "learning_rate": 1.6716061576017179e-07, "loss": 0.0112, "step": 14358 }, { "epoch": 5.66, "learning_rate": 1.6713542947356984e-07, "loss": 0.0128, "step": 14359 }, { "epoch": 5.66, "learning_rate": 1.6711024372261048e-07, "loss": 0.0126, "step": 14360 }, { "epoch": 5.66, "learning_rate": 1.6708505850770415e-07, "loss": 0.0111, "step": 14361 }, { "epoch": 5.66, "learning_rate": 1.6705987382926122e-07, "loss": 0.015, "step": 14362 }, { "epoch": 5.66, "learning_rate": 1.6703468968769244e-07, "loss": 0.0165, "step": 14363 }, { "epoch": 5.66, "learning_rate": 1.6700950608340807e-07, "loss": 0.0134, "step": 14364 }, { "epoch": 5.66, "learning_rate": 1.669843230168186e-07, "loss": 0.0113, "step": 14365 }, { "epoch": 5.66, "learning_rate": 1.6695914048833452e-07, "loss": 0.0132, "step": 14366 }, { "epoch": 5.66, "learning_rate": 1.6693395849836614e-07, "loss": 0.0149, "step": 14367 }, { "epoch": 5.66, "learning_rate": 1.6690877704732404e-07, "loss": 0.015, "step": 14368 }, { "epoch": 5.66, "learning_rate": 1.668835961356186e-07, "loss": 0.0115, "step": 14369 }, { "epoch": 5.66, "learning_rate": 1.6685841576366023e-07, "loss": 0.0101, "step": 14370 }, { "epoch": 5.66, "learning_rate": 1.668332359318592e-07, "loss": 0.0133, "step": 14371 }, { "epoch": 5.66, "learning_rate": 1.6680805664062612e-07, "loss": 0.0139, "step": 14372 }, { "epoch": 5.66, "learning_rate": 1.6678287789037128e-07, "loss": 0.0153, "step": 14373 }, { "epoch": 5.66, "learning_rate": 1.6675769968150498e-07, "loss": 0.0169, "step": 14374 }, { "epoch": 5.67, "learning_rate": 1.6673252201443773e-07, "loss": 0.0114, "step": 14375 }, { "epoch": 5.67, "learning_rate": 1.6670734488957975e-07, "loss": 0.0132, "step": 14376 }, { "epoch": 5.67, "learning_rate": 1.6668216830734147e-07, "loss": 0.0156, "step": 14377 }, { "epoch": 5.67, "learning_rate": 1.6665699226813316e-07, "loss": 0.0123, "step": 14378 }, { "epoch": 5.67, "learning_rate": 1.6663181677236527e-07, "loss": 0.0106, "step": 14379 }, { "epoch": 5.67, "learning_rate": 1.6660664182044802e-07, "loss": 0.0125, "step": 14380 }, { "epoch": 5.67, "learning_rate": 1.6658146741279177e-07, "loss": 0.0123, "step": 14381 }, { "epoch": 5.67, "learning_rate": 1.665562935498068e-07, "loss": 0.0133, "step": 14382 }, { "epoch": 5.67, "learning_rate": 1.6653112023190347e-07, "loss": 0.0153, "step": 14383 }, { "epoch": 5.67, "learning_rate": 1.6650594745949188e-07, "loss": 0.0102, "step": 14384 }, { "epoch": 5.67, "learning_rate": 1.6648077523298255e-07, "loss": 0.0099, "step": 14385 }, { "epoch": 5.67, "learning_rate": 1.6645560355278562e-07, "loss": 0.0111, "step": 14386 }, { "epoch": 5.67, "learning_rate": 1.664304324193114e-07, "loss": 0.0112, "step": 14387 }, { "epoch": 5.67, "learning_rate": 1.664052618329701e-07, "loss": 0.0097, "step": 14388 }, { "epoch": 5.67, "learning_rate": 1.6638009179417187e-07, "loss": 0.0157, "step": 14389 }, { "epoch": 5.67, "learning_rate": 1.6635492230332715e-07, "loss": 0.0123, "step": 14390 }, { "epoch": 5.67, "learning_rate": 1.6632975336084608e-07, "loss": 0.0124, "step": 14391 }, { "epoch": 5.67, "learning_rate": 1.663045849671388e-07, "loss": 0.0133, "step": 14392 }, { "epoch": 5.67, "learning_rate": 1.6627941712261554e-07, "loss": 0.0137, "step": 14393 }, { "epoch": 5.67, "learning_rate": 1.662542498276866e-07, "loss": 0.0122, "step": 14394 }, { "epoch": 5.67, "learning_rate": 1.6622908308276212e-07, "loss": 0.0129, "step": 14395 }, { "epoch": 5.67, "learning_rate": 1.6620391688825223e-07, "loss": 0.012, "step": 14396 }, { "epoch": 5.67, "learning_rate": 1.6617875124456715e-07, "loss": 0.0118, "step": 14397 }, { "epoch": 5.67, "learning_rate": 1.6615358615211695e-07, "loss": 0.0142, "step": 14398 }, { "epoch": 5.67, "learning_rate": 1.6612842161131192e-07, "loss": 0.0111, "step": 14399 }, { "epoch": 5.68, "learning_rate": 1.6610325762256217e-07, "loss": 0.0176, "step": 14400 }, { "epoch": 5.68, "learning_rate": 1.6607809418627772e-07, "loss": 0.0118, "step": 14401 }, { "epoch": 5.68, "learning_rate": 1.660529313028688e-07, "loss": 0.0127, "step": 14402 }, { "epoch": 5.68, "learning_rate": 1.6602776897274554e-07, "loss": 0.0123, "step": 14403 }, { "epoch": 5.68, "learning_rate": 1.66002607196318e-07, "loss": 0.0148, "step": 14404 }, { "epoch": 5.68, "learning_rate": 1.659774459739963e-07, "loss": 0.0118, "step": 14405 }, { "epoch": 5.68, "learning_rate": 1.6595228530619053e-07, "loss": 0.0132, "step": 14406 }, { "epoch": 5.68, "learning_rate": 1.6592712519331066e-07, "loss": 0.015, "step": 14407 }, { "epoch": 5.68, "learning_rate": 1.6590196563576697e-07, "loss": 0.0141, "step": 14408 }, { "epoch": 5.68, "learning_rate": 1.6587680663396941e-07, "loss": 0.012, "step": 14409 }, { "epoch": 5.68, "learning_rate": 1.6585164818832805e-07, "loss": 0.0139, "step": 14410 }, { "epoch": 5.68, "learning_rate": 1.6582649029925282e-07, "loss": 0.0126, "step": 14411 }, { "epoch": 5.68, "learning_rate": 1.6580133296715398e-07, "loss": 0.0133, "step": 14412 }, { "epoch": 5.68, "learning_rate": 1.657761761924414e-07, "loss": 0.0109, "step": 14413 }, { "epoch": 5.68, "learning_rate": 1.6575101997552515e-07, "loss": 0.0116, "step": 14414 }, { "epoch": 5.68, "learning_rate": 1.6572586431681522e-07, "loss": 0.013, "step": 14415 }, { "epoch": 5.68, "learning_rate": 1.6570070921672152e-07, "loss": 0.0135, "step": 14416 }, { "epoch": 5.68, "learning_rate": 1.656755546756542e-07, "loss": 0.0155, "step": 14417 }, { "epoch": 5.68, "learning_rate": 1.6565040069402322e-07, "loss": 0.0107, "step": 14418 }, { "epoch": 5.68, "learning_rate": 1.6562524727223848e-07, "loss": 0.0127, "step": 14419 }, { "epoch": 5.68, "learning_rate": 1.6560009441070992e-07, "loss": 0.0132, "step": 14420 }, { "epoch": 5.68, "learning_rate": 1.6557494210984757e-07, "loss": 0.0118, "step": 14421 }, { "epoch": 5.68, "learning_rate": 1.6554979037006137e-07, "loss": 0.0159, "step": 14422 }, { "epoch": 5.68, "learning_rate": 1.655246391917612e-07, "loss": 0.0112, "step": 14423 }, { "epoch": 5.68, "learning_rate": 1.6549948857535694e-07, "loss": 0.0145, "step": 14424 }, { "epoch": 5.69, "learning_rate": 1.6547433852125867e-07, "loss": 0.0141, "step": 14425 }, { "epoch": 5.69, "learning_rate": 1.6544918902987621e-07, "loss": 0.016, "step": 14426 }, { "epoch": 5.69, "learning_rate": 1.6542404010161947e-07, "loss": 0.0139, "step": 14427 }, { "epoch": 5.69, "learning_rate": 1.6539889173689832e-07, "loss": 0.0114, "step": 14428 }, { "epoch": 5.69, "learning_rate": 1.6537374393612255e-07, "loss": 0.0156, "step": 14429 }, { "epoch": 5.69, "learning_rate": 1.6534859669970223e-07, "loss": 0.0128, "step": 14430 }, { "epoch": 5.69, "learning_rate": 1.6532345002804708e-07, "loss": 0.0098, "step": 14431 }, { "epoch": 5.69, "learning_rate": 1.6529830392156702e-07, "loss": 0.0136, "step": 14432 }, { "epoch": 5.69, "learning_rate": 1.652731583806718e-07, "loss": 0.0115, "step": 14433 }, { "epoch": 5.69, "learning_rate": 1.652480134057714e-07, "loss": 0.0103, "step": 14434 }, { "epoch": 5.69, "learning_rate": 1.6522286899727554e-07, "loss": 0.011, "step": 14435 }, { "epoch": 5.69, "learning_rate": 1.6519772515559405e-07, "loss": 0.0144, "step": 14436 }, { "epoch": 5.69, "learning_rate": 1.6517258188113678e-07, "loss": 0.0152, "step": 14437 }, { "epoch": 5.69, "learning_rate": 1.651474391743134e-07, "loss": 0.0153, "step": 14438 }, { "epoch": 5.69, "learning_rate": 1.6512229703553386e-07, "loss": 0.0116, "step": 14439 }, { "epoch": 5.69, "learning_rate": 1.6509715546520788e-07, "loss": 0.014, "step": 14440 }, { "epoch": 5.69, "learning_rate": 1.6507201446374523e-07, "loss": 0.0122, "step": 14441 }, { "epoch": 5.69, "learning_rate": 1.650468740315556e-07, "loss": 0.0112, "step": 14442 }, { "epoch": 5.69, "learning_rate": 1.6502173416904887e-07, "loss": 0.0135, "step": 14443 }, { "epoch": 5.69, "learning_rate": 1.649965948766347e-07, "loss": 0.0107, "step": 14444 }, { "epoch": 5.69, "learning_rate": 1.649714561547228e-07, "loss": 0.0129, "step": 14445 }, { "epoch": 5.69, "learning_rate": 1.64946318003723e-07, "loss": 0.0127, "step": 14446 }, { "epoch": 5.69, "learning_rate": 1.649211804240448e-07, "loss": 0.0112, "step": 14447 }, { "epoch": 5.69, "learning_rate": 1.6489604341609813e-07, "loss": 0.0123, "step": 14448 }, { "epoch": 5.69, "learning_rate": 1.6487090698029262e-07, "loss": 0.0142, "step": 14449 }, { "epoch": 5.69, "learning_rate": 1.648457711170379e-07, "loss": 0.0154, "step": 14450 }, { "epoch": 5.7, "learning_rate": 1.6482063582674367e-07, "loss": 0.0109, "step": 14451 }, { "epoch": 5.7, "learning_rate": 1.6479550110981963e-07, "loss": 0.013, "step": 14452 }, { "epoch": 5.7, "learning_rate": 1.6477036696667547e-07, "loss": 0.0095, "step": 14453 }, { "epoch": 5.7, "learning_rate": 1.6474523339772073e-07, "loss": 0.015, "step": 14454 }, { "epoch": 5.7, "learning_rate": 1.6472010040336514e-07, "loss": 0.0154, "step": 14455 }, { "epoch": 5.7, "learning_rate": 1.646949679840182e-07, "loss": 0.0117, "step": 14456 }, { "epoch": 5.7, "learning_rate": 1.6466983614008969e-07, "loss": 0.0126, "step": 14457 }, { "epoch": 5.7, "learning_rate": 1.6464470487198917e-07, "loss": 0.0116, "step": 14458 }, { "epoch": 5.7, "learning_rate": 1.6461957418012623e-07, "loss": 0.0151, "step": 14459 }, { "epoch": 5.7, "learning_rate": 1.6459444406491043e-07, "loss": 0.0111, "step": 14460 }, { "epoch": 5.7, "learning_rate": 1.645693145267514e-07, "loss": 0.0133, "step": 14461 }, { "epoch": 5.7, "learning_rate": 1.6454418556605868e-07, "loss": 0.0151, "step": 14462 }, { "epoch": 5.7, "learning_rate": 1.6451905718324185e-07, "loss": 0.0122, "step": 14463 }, { "epoch": 5.7, "learning_rate": 1.6449392937871047e-07, "loss": 0.0148, "step": 14464 }, { "epoch": 5.7, "learning_rate": 1.6446880215287408e-07, "loss": 0.0141, "step": 14465 }, { "epoch": 5.7, "learning_rate": 1.6444367550614223e-07, "loss": 0.0124, "step": 14466 }, { "epoch": 5.7, "learning_rate": 1.6441854943892444e-07, "loss": 0.0157, "step": 14467 }, { "epoch": 5.7, "learning_rate": 1.643934239516302e-07, "loss": 0.0133, "step": 14468 }, { "epoch": 5.7, "learning_rate": 1.6436829904466895e-07, "loss": 0.0134, "step": 14469 }, { "epoch": 5.7, "learning_rate": 1.6434317471845036e-07, "loss": 0.0144, "step": 14470 }, { "epoch": 5.7, "learning_rate": 1.6431805097338384e-07, "loss": 0.0145, "step": 14471 }, { "epoch": 5.7, "learning_rate": 1.6429292780987885e-07, "loss": 0.0101, "step": 14472 }, { "epoch": 5.7, "learning_rate": 1.642678052283448e-07, "loss": 0.0118, "step": 14473 }, { "epoch": 5.7, "learning_rate": 1.6424268322919127e-07, "loss": 0.0127, "step": 14474 }, { "epoch": 5.7, "learning_rate": 1.6421756181282768e-07, "loss": 0.0128, "step": 14475 }, { "epoch": 5.71, "learning_rate": 1.6419244097966343e-07, "loss": 0.0136, "step": 14476 }, { "epoch": 5.71, "learning_rate": 1.64167320730108e-07, "loss": 0.0161, "step": 14477 }, { "epoch": 5.71, "learning_rate": 1.6414220106457067e-07, "loss": 0.0142, "step": 14478 }, { "epoch": 5.71, "learning_rate": 1.6411708198346107e-07, "loss": 0.0121, "step": 14479 }, { "epoch": 5.71, "learning_rate": 1.640919634871885e-07, "loss": 0.0107, "step": 14480 }, { "epoch": 5.71, "learning_rate": 1.6406684557616232e-07, "loss": 0.0139, "step": 14481 }, { "epoch": 5.71, "learning_rate": 1.6404172825079195e-07, "loss": 0.0128, "step": 14482 }, { "epoch": 5.71, "learning_rate": 1.6401661151148675e-07, "loss": 0.0129, "step": 14483 }, { "epoch": 5.71, "learning_rate": 1.639914953586561e-07, "loss": 0.0166, "step": 14484 }, { "epoch": 5.71, "learning_rate": 1.6396637979270936e-07, "loss": 0.013, "step": 14485 }, { "epoch": 5.71, "learning_rate": 1.6394126481405587e-07, "loss": 0.0126, "step": 14486 }, { "epoch": 5.71, "learning_rate": 1.639161504231049e-07, "loss": 0.0153, "step": 14487 }, { "epoch": 5.71, "learning_rate": 1.6389103662026594e-07, "loss": 0.014, "step": 14488 }, { "epoch": 5.71, "learning_rate": 1.6386592340594817e-07, "loss": 0.0143, "step": 14489 }, { "epoch": 5.71, "learning_rate": 1.6384081078056092e-07, "loss": 0.0123, "step": 14490 }, { "epoch": 5.71, "learning_rate": 1.638156987445134e-07, "loss": 0.0152, "step": 14491 }, { "epoch": 5.71, "learning_rate": 1.637905872982151e-07, "loss": 0.019, "step": 14492 }, { "epoch": 5.71, "learning_rate": 1.637654764420752e-07, "loss": 0.0134, "step": 14493 }, { "epoch": 5.71, "learning_rate": 1.6374036617650297e-07, "loss": 0.0131, "step": 14494 }, { "epoch": 5.71, "learning_rate": 1.6371525650190766e-07, "loss": 0.0139, "step": 14495 }, { "epoch": 5.71, "learning_rate": 1.6369014741869843e-07, "loss": 0.0146, "step": 14496 }, { "epoch": 5.71, "learning_rate": 1.6366503892728467e-07, "loss": 0.0147, "step": 14497 }, { "epoch": 5.71, "learning_rate": 1.636399310280756e-07, "loss": 0.0117, "step": 14498 }, { "epoch": 5.71, "learning_rate": 1.6361482372148037e-07, "loss": 0.0164, "step": 14499 }, { "epoch": 5.71, "learning_rate": 1.6358971700790814e-07, "loss": 0.0136, "step": 14500 }, { "epoch": 5.71, "learning_rate": 1.6356461088776825e-07, "loss": 0.0116, "step": 14501 }, { "epoch": 5.72, "learning_rate": 1.6353950536146983e-07, "loss": 0.0116, "step": 14502 }, { "epoch": 5.72, "learning_rate": 1.6351440042942208e-07, "loss": 0.0134, "step": 14503 }, { "epoch": 5.72, "learning_rate": 1.634892960920341e-07, "loss": 0.0142, "step": 14504 }, { "epoch": 5.72, "learning_rate": 1.6346419234971515e-07, "loss": 0.0104, "step": 14505 }, { "epoch": 5.72, "learning_rate": 1.6343908920287428e-07, "loss": 0.0101, "step": 14506 }, { "epoch": 5.72, "learning_rate": 1.6341398665192075e-07, "loss": 0.0138, "step": 14507 }, { "epoch": 5.72, "learning_rate": 1.633888846972636e-07, "loss": 0.0165, "step": 14508 }, { "epoch": 5.72, "learning_rate": 1.63363783339312e-07, "loss": 0.0123, "step": 14509 }, { "epoch": 5.72, "learning_rate": 1.6333868257847505e-07, "loss": 0.016, "step": 14510 }, { "epoch": 5.72, "learning_rate": 1.6331358241516186e-07, "loss": 0.0107, "step": 14511 }, { "epoch": 5.72, "learning_rate": 1.6328848284978154e-07, "loss": 0.0126, "step": 14512 }, { "epoch": 5.72, "learning_rate": 1.6326338388274305e-07, "loss": 0.0116, "step": 14513 }, { "epoch": 5.72, "learning_rate": 1.6323828551445568e-07, "loss": 0.0166, "step": 14514 }, { "epoch": 5.72, "learning_rate": 1.6321318774532836e-07, "loss": 0.0137, "step": 14515 }, { "epoch": 5.72, "learning_rate": 1.631880905757702e-07, "loss": 0.0144, "step": 14516 }, { "epoch": 5.72, "learning_rate": 1.631629940061902e-07, "loss": 0.0098, "step": 14517 }, { "epoch": 5.72, "learning_rate": 1.631378980369973e-07, "loss": 0.015, "step": 14518 }, { "epoch": 5.72, "learning_rate": 1.6311280266860073e-07, "loss": 0.0132, "step": 14519 }, { "epoch": 5.72, "learning_rate": 1.6308770790140943e-07, "loss": 0.0157, "step": 14520 }, { "epoch": 5.72, "learning_rate": 1.6306261373583238e-07, "loss": 0.0129, "step": 14521 }, { "epoch": 5.72, "learning_rate": 1.6303752017227852e-07, "loss": 0.0115, "step": 14522 }, { "epoch": 5.72, "learning_rate": 1.6301242721115697e-07, "loss": 0.014, "step": 14523 }, { "epoch": 5.72, "learning_rate": 1.6298733485287665e-07, "loss": 0.0143, "step": 14524 }, { "epoch": 5.72, "learning_rate": 1.6296224309784653e-07, "loss": 0.0168, "step": 14525 }, { "epoch": 5.72, "learning_rate": 1.6293715194647556e-07, "loss": 0.0147, "step": 14526 }, { "epoch": 5.73, "learning_rate": 1.6291206139917262e-07, "loss": 0.0135, "step": 14527 }, { "epoch": 5.73, "learning_rate": 1.6288697145634675e-07, "loss": 0.0137, "step": 14528 }, { "epoch": 5.73, "learning_rate": 1.628618821184068e-07, "loss": 0.0133, "step": 14529 }, { "epoch": 5.73, "learning_rate": 1.6283679338576181e-07, "loss": 0.0115, "step": 14530 }, { "epoch": 5.73, "learning_rate": 1.6281170525882053e-07, "loss": 0.0141, "step": 14531 }, { "epoch": 5.73, "learning_rate": 1.62786617737992e-07, "loss": 0.0162, "step": 14532 }, { "epoch": 5.73, "learning_rate": 1.6276153082368501e-07, "loss": 0.0135, "step": 14533 }, { "epoch": 5.73, "learning_rate": 1.6273644451630851e-07, "loss": 0.0166, "step": 14534 }, { "epoch": 5.73, "learning_rate": 1.6271135881627129e-07, "loss": 0.0124, "step": 14535 }, { "epoch": 5.73, "learning_rate": 1.6268627372398218e-07, "loss": 0.0141, "step": 14536 }, { "epoch": 5.73, "learning_rate": 1.6266118923985016e-07, "loss": 0.0129, "step": 14537 }, { "epoch": 5.73, "learning_rate": 1.6263610536428405e-07, "loss": 0.0102, "step": 14538 }, { "epoch": 5.73, "learning_rate": 1.6261102209769262e-07, "loss": 0.0094, "step": 14539 }, { "epoch": 5.73, "learning_rate": 1.6258593944048462e-07, "loss": 0.0124, "step": 14540 }, { "epoch": 5.73, "learning_rate": 1.62560857393069e-07, "loss": 0.0115, "step": 14541 }, { "epoch": 5.73, "learning_rate": 1.6253577595585455e-07, "loss": 0.0137, "step": 14542 }, { "epoch": 5.73, "learning_rate": 1.6251069512925e-07, "loss": 0.0158, "step": 14543 }, { "epoch": 5.73, "learning_rate": 1.6248561491366407e-07, "loss": 0.0129, "step": 14544 }, { "epoch": 5.73, "learning_rate": 1.6246053530950565e-07, "loss": 0.0131, "step": 14545 }, { "epoch": 5.73, "learning_rate": 1.6243545631718348e-07, "loss": 0.0126, "step": 14546 }, { "epoch": 5.73, "learning_rate": 1.6241037793710625e-07, "loss": 0.0112, "step": 14547 }, { "epoch": 5.73, "learning_rate": 1.6238530016968276e-07, "loss": 0.0138, "step": 14548 }, { "epoch": 5.73, "learning_rate": 1.6236022301532166e-07, "loss": 0.0124, "step": 14549 }, { "epoch": 5.73, "learning_rate": 1.6233514647443176e-07, "loss": 0.013, "step": 14550 }, { "epoch": 5.73, "learning_rate": 1.6231007054742172e-07, "loss": 0.0125, "step": 14551 }, { "epoch": 5.74, "learning_rate": 1.622849952347002e-07, "loss": 0.0121, "step": 14552 }, { "epoch": 5.74, "learning_rate": 1.6225992053667597e-07, "loss": 0.0149, "step": 14553 }, { "epoch": 5.74, "learning_rate": 1.622348464537577e-07, "loss": 0.0166, "step": 14554 }, { "epoch": 5.74, "learning_rate": 1.6220977298635403e-07, "loss": 0.013, "step": 14555 }, { "epoch": 5.74, "learning_rate": 1.6218470013487363e-07, "loss": 0.0102, "step": 14556 }, { "epoch": 5.74, "learning_rate": 1.6215962789972513e-07, "loss": 0.0124, "step": 14557 }, { "epoch": 5.74, "learning_rate": 1.6213455628131713e-07, "loss": 0.0123, "step": 14558 }, { "epoch": 5.74, "learning_rate": 1.6210948528005837e-07, "loss": 0.0136, "step": 14559 }, { "epoch": 5.74, "learning_rate": 1.6208441489635743e-07, "loss": 0.0127, "step": 14560 }, { "epoch": 5.74, "learning_rate": 1.6205934513062293e-07, "loss": 0.0126, "step": 14561 }, { "epoch": 5.74, "learning_rate": 1.620342759832633e-07, "loss": 0.0119, "step": 14562 }, { "epoch": 5.74, "learning_rate": 1.6200920745468738e-07, "loss": 0.0155, "step": 14563 }, { "epoch": 5.74, "learning_rate": 1.6198413954530364e-07, "loss": 0.0129, "step": 14564 }, { "epoch": 5.74, "learning_rate": 1.6195907225552067e-07, "loss": 0.0129, "step": 14565 }, { "epoch": 5.74, "learning_rate": 1.6193400558574696e-07, "loss": 0.0156, "step": 14566 }, { "epoch": 5.74, "learning_rate": 1.6190893953639104e-07, "loss": 0.013, "step": 14567 }, { "epoch": 5.74, "learning_rate": 1.6188387410786161e-07, "loss": 0.0125, "step": 14568 }, { "epoch": 5.74, "learning_rate": 1.618588093005671e-07, "loss": 0.0127, "step": 14569 }, { "epoch": 5.74, "learning_rate": 1.6183374511491603e-07, "loss": 0.013, "step": 14570 }, { "epoch": 5.74, "learning_rate": 1.6180868155131686e-07, "loss": 0.0125, "step": 14571 }, { "epoch": 5.74, "learning_rate": 1.6178361861017816e-07, "loss": 0.0114, "step": 14572 }, { "epoch": 5.74, "learning_rate": 1.6175855629190843e-07, "loss": 0.0113, "step": 14573 }, { "epoch": 5.74, "learning_rate": 1.617334945969161e-07, "loss": 0.0146, "step": 14574 }, { "epoch": 5.74, "learning_rate": 1.617084335256096e-07, "loss": 0.0129, "step": 14575 }, { "epoch": 5.74, "learning_rate": 1.6168337307839746e-07, "loss": 0.0119, "step": 14576 }, { "epoch": 5.74, "learning_rate": 1.6165831325568812e-07, "loss": 0.0127, "step": 14577 }, { "epoch": 5.75, "learning_rate": 1.6163325405789001e-07, "loss": 0.015, "step": 14578 }, { "epoch": 5.75, "learning_rate": 1.6160819548541155e-07, "loss": 0.0128, "step": 14579 }, { "epoch": 5.75, "learning_rate": 1.6158313753866107e-07, "loss": 0.0132, "step": 14580 }, { "epoch": 5.75, "learning_rate": 1.6155808021804713e-07, "loss": 0.0123, "step": 14581 }, { "epoch": 5.75, "learning_rate": 1.6153302352397807e-07, "loss": 0.0122, "step": 14582 }, { "epoch": 5.75, "learning_rate": 1.615079674568623e-07, "loss": 0.0125, "step": 14583 }, { "epoch": 5.75, "learning_rate": 1.6148291201710802e-07, "loss": 0.0145, "step": 14584 }, { "epoch": 5.75, "learning_rate": 1.6145785720512382e-07, "loss": 0.0108, "step": 14585 }, { "epoch": 5.75, "learning_rate": 1.61432803021318e-07, "loss": 0.011, "step": 14586 }, { "epoch": 5.75, "learning_rate": 1.6140774946609889e-07, "loss": 0.0115, "step": 14587 }, { "epoch": 5.75, "learning_rate": 1.6138269653987478e-07, "loss": 0.0144, "step": 14588 }, { "epoch": 5.75, "learning_rate": 1.6135764424305395e-07, "loss": 0.0158, "step": 14589 }, { "epoch": 5.75, "learning_rate": 1.6133259257604485e-07, "loss": 0.012, "step": 14590 }, { "epoch": 5.75, "learning_rate": 1.6130754153925575e-07, "loss": 0.0123, "step": 14591 }, { "epoch": 5.75, "learning_rate": 1.612824911330949e-07, "loss": 0.0135, "step": 14592 }, { "epoch": 5.75, "learning_rate": 1.6125744135797056e-07, "loss": 0.0118, "step": 14593 }, { "epoch": 5.75, "learning_rate": 1.612323922142911e-07, "loss": 0.0122, "step": 14594 }, { "epoch": 5.75, "learning_rate": 1.6120734370246472e-07, "loss": 0.0193, "step": 14595 }, { "epoch": 5.75, "learning_rate": 1.6118229582289969e-07, "loss": 0.0109, "step": 14596 }, { "epoch": 5.75, "learning_rate": 1.6115724857600418e-07, "loss": 0.011, "step": 14597 }, { "epoch": 5.75, "learning_rate": 1.611322019621865e-07, "loss": 0.0124, "step": 14598 }, { "epoch": 5.75, "learning_rate": 1.6110715598185488e-07, "loss": 0.0168, "step": 14599 }, { "epoch": 5.75, "learning_rate": 1.6108211063541756e-07, "loss": 0.0132, "step": 14600 }, { "epoch": 5.75, "learning_rate": 1.6105706592328262e-07, "loss": 0.0155, "step": 14601 }, { "epoch": 5.75, "learning_rate": 1.6103202184585828e-07, "loss": 0.0142, "step": 14602 }, { "epoch": 5.76, "learning_rate": 1.6100697840355281e-07, "loss": 0.0136, "step": 14603 }, { "epoch": 5.76, "learning_rate": 1.6098193559677434e-07, "loss": 0.0133, "step": 14604 }, { "epoch": 5.76, "learning_rate": 1.6095689342593103e-07, "loss": 0.0134, "step": 14605 }, { "epoch": 5.76, "learning_rate": 1.6093185189143102e-07, "loss": 0.0127, "step": 14606 }, { "epoch": 5.76, "learning_rate": 1.6090681099368235e-07, "loss": 0.0148, "step": 14607 }, { "epoch": 5.76, "learning_rate": 1.6088177073309334e-07, "loss": 0.0167, "step": 14608 }, { "epoch": 5.76, "learning_rate": 1.60856731110072e-07, "loss": 0.0122, "step": 14609 }, { "epoch": 5.76, "learning_rate": 1.6083169212502647e-07, "loss": 0.0125, "step": 14610 }, { "epoch": 5.76, "learning_rate": 1.608066537783648e-07, "loss": 0.0097, "step": 14611 }, { "epoch": 5.76, "learning_rate": 1.6078161607049508e-07, "loss": 0.0135, "step": 14612 }, { "epoch": 5.76, "learning_rate": 1.6075657900182546e-07, "loss": 0.0134, "step": 14613 }, { "epoch": 5.76, "learning_rate": 1.6073154257276396e-07, "loss": 0.0141, "step": 14614 }, { "epoch": 5.76, "learning_rate": 1.6070650678371863e-07, "loss": 0.0138, "step": 14615 }, { "epoch": 5.76, "learning_rate": 1.606814716350975e-07, "loss": 0.0126, "step": 14616 }, { "epoch": 5.76, "learning_rate": 1.6065643712730866e-07, "loss": 0.0112, "step": 14617 }, { "epoch": 5.76, "learning_rate": 1.6063140326076008e-07, "loss": 0.0155, "step": 14618 }, { "epoch": 5.76, "learning_rate": 1.606063700358598e-07, "loss": 0.0169, "step": 14619 }, { "epoch": 5.76, "learning_rate": 1.6058133745301575e-07, "loss": 0.0149, "step": 14620 }, { "epoch": 5.76, "learning_rate": 1.6055630551263603e-07, "loss": 0.0124, "step": 14621 }, { "epoch": 5.76, "learning_rate": 1.6053127421512862e-07, "loss": 0.0133, "step": 14622 }, { "epoch": 5.76, "learning_rate": 1.6050624356090146e-07, "loss": 0.0131, "step": 14623 }, { "epoch": 5.76, "learning_rate": 1.6048121355036251e-07, "loss": 0.0121, "step": 14624 }, { "epoch": 5.76, "learning_rate": 1.604561841839196e-07, "loss": 0.0135, "step": 14625 }, { "epoch": 5.76, "learning_rate": 1.604311554619809e-07, "loss": 0.0135, "step": 14626 }, { "epoch": 5.76, "learning_rate": 1.6040612738495424e-07, "loss": 0.0155, "step": 14627 }, { "epoch": 5.77, "learning_rate": 1.603810999532475e-07, "loss": 0.0145, "step": 14628 }, { "epoch": 5.77, "learning_rate": 1.603560731672685e-07, "loss": 0.0146, "step": 14629 }, { "epoch": 5.77, "learning_rate": 1.6033104702742538e-07, "loss": 0.0123, "step": 14630 }, { "epoch": 5.77, "learning_rate": 1.603060215341259e-07, "loss": 0.0143, "step": 14631 }, { "epoch": 5.77, "learning_rate": 1.6028099668777796e-07, "loss": 0.0132, "step": 14632 }, { "epoch": 5.77, "learning_rate": 1.6025597248878934e-07, "loss": 0.013, "step": 14633 }, { "epoch": 5.77, "learning_rate": 1.6023094893756798e-07, "loss": 0.0088, "step": 14634 }, { "epoch": 5.77, "learning_rate": 1.6020592603452178e-07, "loss": 0.0109, "step": 14635 }, { "epoch": 5.77, "learning_rate": 1.6018090378005846e-07, "loss": 0.0129, "step": 14636 }, { "epoch": 5.77, "learning_rate": 1.6015588217458592e-07, "loss": 0.0154, "step": 14637 }, { "epoch": 5.77, "learning_rate": 1.6013086121851186e-07, "loss": 0.014, "step": 14638 }, { "epoch": 5.77, "learning_rate": 1.6010584091224423e-07, "loss": 0.0153, "step": 14639 }, { "epoch": 5.77, "learning_rate": 1.6008082125619078e-07, "loss": 0.0128, "step": 14640 }, { "epoch": 5.77, "learning_rate": 1.6005580225075924e-07, "loss": 0.0122, "step": 14641 }, { "epoch": 5.77, "learning_rate": 1.6003078389635736e-07, "loss": 0.0164, "step": 14642 }, { "epoch": 5.77, "learning_rate": 1.6000576619339304e-07, "loss": 0.0116, "step": 14643 }, { "epoch": 5.77, "learning_rate": 1.5998074914227392e-07, "loss": 0.0118, "step": 14644 }, { "epoch": 5.77, "learning_rate": 1.599557327434078e-07, "loss": 0.0124, "step": 14645 }, { "epoch": 5.77, "learning_rate": 1.5993071699720234e-07, "loss": 0.0129, "step": 14646 }, { "epoch": 5.77, "learning_rate": 1.5990570190406524e-07, "loss": 0.0124, "step": 14647 }, { "epoch": 5.77, "learning_rate": 1.5988068746440435e-07, "loss": 0.0133, "step": 14648 }, { "epoch": 5.77, "learning_rate": 1.5985567367862726e-07, "loss": 0.0082, "step": 14649 }, { "epoch": 5.77, "learning_rate": 1.5983066054714168e-07, "loss": 0.0113, "step": 14650 }, { "epoch": 5.77, "learning_rate": 1.598056480703552e-07, "loss": 0.0153, "step": 14651 }, { "epoch": 5.77, "learning_rate": 1.597806362486757e-07, "loss": 0.0149, "step": 14652 }, { "epoch": 5.77, "learning_rate": 1.5975562508251063e-07, "loss": 0.0125, "step": 14653 }, { "epoch": 5.78, "learning_rate": 1.5973061457226773e-07, "loss": 0.0136, "step": 14654 }, { "epoch": 5.78, "learning_rate": 1.5970560471835462e-07, "loss": 0.0142, "step": 14655 }, { "epoch": 5.78, "learning_rate": 1.5968059552117884e-07, "loss": 0.011, "step": 14656 }, { "epoch": 5.78, "learning_rate": 1.596555869811481e-07, "loss": 0.0143, "step": 14657 }, { "epoch": 5.78, "learning_rate": 1.5963057909867004e-07, "loss": 0.0129, "step": 14658 }, { "epoch": 5.78, "learning_rate": 1.5960557187415215e-07, "loss": 0.013, "step": 14659 }, { "epoch": 5.78, "learning_rate": 1.5958056530800203e-07, "loss": 0.0136, "step": 14660 }, { "epoch": 5.78, "learning_rate": 1.595555594006273e-07, "loss": 0.013, "step": 14661 }, { "epoch": 5.78, "learning_rate": 1.595305541524355e-07, "loss": 0.0112, "step": 14662 }, { "epoch": 5.78, "learning_rate": 1.5950554956383413e-07, "loss": 0.0132, "step": 14663 }, { "epoch": 5.78, "learning_rate": 1.5948054563523076e-07, "loss": 0.0156, "step": 14664 }, { "epoch": 5.78, "learning_rate": 1.5945554236703282e-07, "loss": 0.0138, "step": 14665 }, { "epoch": 5.78, "learning_rate": 1.5943053975964804e-07, "loss": 0.0124, "step": 14666 }, { "epoch": 5.78, "learning_rate": 1.5940553781348378e-07, "loss": 0.0158, "step": 14667 }, { "epoch": 5.78, "learning_rate": 1.5938053652894757e-07, "loss": 0.0117, "step": 14668 }, { "epoch": 5.78, "learning_rate": 1.5935553590644677e-07, "loss": 0.0137, "step": 14669 }, { "epoch": 5.78, "learning_rate": 1.5933053594638906e-07, "loss": 0.0146, "step": 14670 }, { "epoch": 5.78, "learning_rate": 1.5930553664918182e-07, "loss": 0.0122, "step": 14671 }, { "epoch": 5.78, "learning_rate": 1.5928053801523246e-07, "loss": 0.0152, "step": 14672 }, { "epoch": 5.78, "learning_rate": 1.5925554004494836e-07, "loss": 0.0136, "step": 14673 }, { "epoch": 5.78, "learning_rate": 1.5923054273873712e-07, "loss": 0.0123, "step": 14674 }, { "epoch": 5.78, "learning_rate": 1.592055460970061e-07, "loss": 0.0167, "step": 14675 }, { "epoch": 5.78, "learning_rate": 1.5918055012016264e-07, "loss": 0.013, "step": 14676 }, { "epoch": 5.78, "learning_rate": 1.591555548086142e-07, "loss": 0.0128, "step": 14677 }, { "epoch": 5.78, "learning_rate": 1.591305601627681e-07, "loss": 0.0116, "step": 14678 }, { "epoch": 5.79, "learning_rate": 1.5910556618303179e-07, "loss": 0.0123, "step": 14679 }, { "epoch": 5.79, "learning_rate": 1.5908057286981252e-07, "loss": 0.0155, "step": 14680 }, { "epoch": 5.79, "learning_rate": 1.590555802235178e-07, "loss": 0.0114, "step": 14681 }, { "epoch": 5.79, "learning_rate": 1.5903058824455485e-07, "loss": 0.0143, "step": 14682 }, { "epoch": 5.79, "learning_rate": 1.590055969333311e-07, "loss": 0.0141, "step": 14683 }, { "epoch": 5.79, "learning_rate": 1.589806062902538e-07, "loss": 0.0102, "step": 14684 }, { "epoch": 5.79, "learning_rate": 1.5895561631573028e-07, "loss": 0.0138, "step": 14685 }, { "epoch": 5.79, "learning_rate": 1.5893062701016784e-07, "loss": 0.0147, "step": 14686 }, { "epoch": 5.79, "learning_rate": 1.5890563837397367e-07, "loss": 0.0108, "step": 14687 }, { "epoch": 5.79, "learning_rate": 1.5888065040755523e-07, "loss": 0.0114, "step": 14688 }, { "epoch": 5.79, "learning_rate": 1.588556631113197e-07, "loss": 0.0141, "step": 14689 }, { "epoch": 5.79, "learning_rate": 1.588306764856743e-07, "loss": 0.0167, "step": 14690 }, { "epoch": 5.79, "learning_rate": 1.5880569053102622e-07, "loss": 0.013, "step": 14691 }, { "epoch": 5.79, "learning_rate": 1.5878070524778285e-07, "loss": 0.0115, "step": 14692 }, { "epoch": 5.79, "learning_rate": 1.5875572063635138e-07, "loss": 0.0183, "step": 14693 }, { "epoch": 5.79, "learning_rate": 1.5873073669713897e-07, "loss": 0.0115, "step": 14694 }, { "epoch": 5.79, "learning_rate": 1.5870575343055283e-07, "loss": 0.0135, "step": 14695 }, { "epoch": 5.79, "learning_rate": 1.5868077083700006e-07, "loss": 0.0145, "step": 14696 }, { "epoch": 5.79, "learning_rate": 1.5865578891688798e-07, "loss": 0.0158, "step": 14697 }, { "epoch": 5.79, "learning_rate": 1.5863080767062373e-07, "loss": 0.0121, "step": 14698 }, { "epoch": 5.79, "learning_rate": 1.5860582709861445e-07, "loss": 0.0139, "step": 14699 }, { "epoch": 5.79, "learning_rate": 1.5858084720126724e-07, "loss": 0.0149, "step": 14700 }, { "epoch": 5.79, "learning_rate": 1.585558679789893e-07, "loss": 0.012, "step": 14701 }, { "epoch": 5.79, "learning_rate": 1.585308894321877e-07, "loss": 0.0147, "step": 14702 }, { "epoch": 5.79, "learning_rate": 1.5850591156126957e-07, "loss": 0.0131, "step": 14703 }, { "epoch": 5.79, "learning_rate": 1.5848093436664205e-07, "loss": 0.0133, "step": 14704 }, { "epoch": 5.8, "learning_rate": 1.5845595784871214e-07, "loss": 0.0124, "step": 14705 }, { "epoch": 5.8, "learning_rate": 1.5843098200788705e-07, "loss": 0.0143, "step": 14706 }, { "epoch": 5.8, "learning_rate": 1.5840600684457374e-07, "loss": 0.0146, "step": 14707 }, { "epoch": 5.8, "learning_rate": 1.5838103235917935e-07, "loss": 0.0115, "step": 14708 }, { "epoch": 5.8, "learning_rate": 1.5835605855211076e-07, "loss": 0.0136, "step": 14709 }, { "epoch": 5.8, "learning_rate": 1.583310854237752e-07, "loss": 0.0118, "step": 14710 }, { "epoch": 5.8, "learning_rate": 1.5830611297457964e-07, "loss": 0.0119, "step": 14711 }, { "epoch": 5.8, "learning_rate": 1.5828114120493107e-07, "loss": 0.0133, "step": 14712 }, { "epoch": 5.8, "learning_rate": 1.5825617011523642e-07, "loss": 0.0164, "step": 14713 }, { "epoch": 5.8, "learning_rate": 1.582311997059028e-07, "loss": 0.011, "step": 14714 }, { "epoch": 5.8, "learning_rate": 1.5820622997733718e-07, "loss": 0.0158, "step": 14715 }, { "epoch": 5.8, "learning_rate": 1.5818126092994648e-07, "loss": 0.0144, "step": 14716 }, { "epoch": 5.8, "learning_rate": 1.5815629256413768e-07, "loss": 0.0125, "step": 14717 }, { "epoch": 5.8, "learning_rate": 1.581313248803176e-07, "loss": 0.0127, "step": 14718 }, { "epoch": 5.8, "learning_rate": 1.581063578788934e-07, "loss": 0.0117, "step": 14719 }, { "epoch": 5.8, "learning_rate": 1.580813915602719e-07, "loss": 0.0133, "step": 14720 }, { "epoch": 5.8, "learning_rate": 1.5805642592485998e-07, "loss": 0.0137, "step": 14721 }, { "epoch": 5.8, "learning_rate": 1.5803146097306454e-07, "loss": 0.012, "step": 14722 }, { "epoch": 5.8, "learning_rate": 1.5800649670529256e-07, "loss": 0.0152, "step": 14723 }, { "epoch": 5.8, "learning_rate": 1.5798153312195087e-07, "loss": 0.0121, "step": 14724 }, { "epoch": 5.8, "learning_rate": 1.5795657022344625e-07, "loss": 0.0144, "step": 14725 }, { "epoch": 5.8, "learning_rate": 1.579316080101857e-07, "loss": 0.0129, "step": 14726 }, { "epoch": 5.8, "learning_rate": 1.5790664648257598e-07, "loss": 0.0122, "step": 14727 }, { "epoch": 5.8, "learning_rate": 1.5788168564102393e-07, "loss": 0.0143, "step": 14728 }, { "epoch": 5.8, "learning_rate": 1.5785672548593644e-07, "loss": 0.0133, "step": 14729 }, { "epoch": 5.81, "learning_rate": 1.5783176601772028e-07, "loss": 0.0125, "step": 14730 }, { "epoch": 5.81, "learning_rate": 1.5780680723678215e-07, "loss": 0.0123, "step": 14731 }, { "epoch": 5.81, "learning_rate": 1.5778184914352898e-07, "loss": 0.0119, "step": 14732 }, { "epoch": 5.81, "learning_rate": 1.5775689173836754e-07, "loss": 0.0116, "step": 14733 }, { "epoch": 5.81, "learning_rate": 1.5773193502170459e-07, "loss": 0.0144, "step": 14734 }, { "epoch": 5.81, "learning_rate": 1.5770697899394684e-07, "loss": 0.0136, "step": 14735 }, { "epoch": 5.81, "learning_rate": 1.5768202365550093e-07, "loss": 0.0142, "step": 14736 }, { "epoch": 5.81, "learning_rate": 1.5765706900677383e-07, "loss": 0.0107, "step": 14737 }, { "epoch": 5.81, "learning_rate": 1.5763211504817218e-07, "loss": 0.0116, "step": 14738 }, { "epoch": 5.81, "learning_rate": 1.5760716178010265e-07, "loss": 0.0116, "step": 14739 }, { "epoch": 5.81, "learning_rate": 1.5758220920297184e-07, "loss": 0.0113, "step": 14740 }, { "epoch": 5.81, "learning_rate": 1.5755725731718665e-07, "loss": 0.0122, "step": 14741 }, { "epoch": 5.81, "learning_rate": 1.5753230612315368e-07, "loss": 0.0153, "step": 14742 }, { "epoch": 5.81, "learning_rate": 1.5750735562127956e-07, "loss": 0.0125, "step": 14743 }, { "epoch": 5.81, "learning_rate": 1.5748240581197097e-07, "loss": 0.0146, "step": 14744 }, { "epoch": 5.81, "learning_rate": 1.5745745669563448e-07, "loss": 0.015, "step": 14745 }, { "epoch": 5.81, "learning_rate": 1.5743250827267684e-07, "loss": 0.0159, "step": 14746 }, { "epoch": 5.81, "learning_rate": 1.5740756054350467e-07, "loss": 0.0164, "step": 14747 }, { "epoch": 5.81, "learning_rate": 1.573826135085244e-07, "loss": 0.0125, "step": 14748 }, { "epoch": 5.81, "learning_rate": 1.5735766716814282e-07, "loss": 0.013, "step": 14749 }, { "epoch": 5.81, "learning_rate": 1.573327215227665e-07, "loss": 0.0147, "step": 14750 }, { "epoch": 5.81, "learning_rate": 1.5730777657280193e-07, "loss": 0.0112, "step": 14751 }, { "epoch": 5.81, "learning_rate": 1.5728283231865575e-07, "loss": 0.0103, "step": 14752 }, { "epoch": 5.81, "learning_rate": 1.572578887607344e-07, "loss": 0.0156, "step": 14753 }, { "epoch": 5.81, "learning_rate": 1.5723294589944457e-07, "loss": 0.0139, "step": 14754 }, { "epoch": 5.82, "learning_rate": 1.5720800373519275e-07, "loss": 0.013, "step": 14755 }, { "epoch": 5.82, "learning_rate": 1.571830622683854e-07, "loss": 0.0143, "step": 14756 }, { "epoch": 5.82, "learning_rate": 1.5715812149942907e-07, "loss": 0.0127, "step": 14757 }, { "epoch": 5.82, "learning_rate": 1.5713318142873018e-07, "loss": 0.0119, "step": 14758 }, { "epoch": 5.82, "learning_rate": 1.5710824205669534e-07, "loss": 0.0135, "step": 14759 }, { "epoch": 5.82, "learning_rate": 1.5708330338373096e-07, "loss": 0.0137, "step": 14760 }, { "epoch": 5.82, "learning_rate": 1.5705836541024355e-07, "loss": 0.0167, "step": 14761 }, { "epoch": 5.82, "learning_rate": 1.570334281366394e-07, "loss": 0.0146, "step": 14762 }, { "epoch": 5.82, "learning_rate": 1.570084915633252e-07, "loss": 0.0134, "step": 14763 }, { "epoch": 5.82, "learning_rate": 1.569835556907072e-07, "loss": 0.0142, "step": 14764 }, { "epoch": 5.82, "learning_rate": 1.5695862051919188e-07, "loss": 0.0122, "step": 14765 }, { "epoch": 5.82, "learning_rate": 1.569336860491856e-07, "loss": 0.0141, "step": 14766 }, { "epoch": 5.82, "learning_rate": 1.569087522810948e-07, "loss": 0.0141, "step": 14767 }, { "epoch": 5.82, "learning_rate": 1.5688381921532588e-07, "loss": 0.0117, "step": 14768 }, { "epoch": 5.82, "learning_rate": 1.5685888685228513e-07, "loss": 0.0122, "step": 14769 }, { "epoch": 5.82, "learning_rate": 1.56833955192379e-07, "loss": 0.0113, "step": 14770 }, { "epoch": 5.82, "learning_rate": 1.568090242360137e-07, "loss": 0.0143, "step": 14771 }, { "epoch": 5.82, "learning_rate": 1.5678409398359576e-07, "loss": 0.0144, "step": 14772 }, { "epoch": 5.82, "learning_rate": 1.567591644355314e-07, "loss": 0.0123, "step": 14773 }, { "epoch": 5.82, "learning_rate": 1.5673423559222692e-07, "loss": 0.0124, "step": 14774 }, { "epoch": 5.82, "learning_rate": 1.5670930745408865e-07, "loss": 0.0189, "step": 14775 }, { "epoch": 5.82, "learning_rate": 1.566843800215228e-07, "loss": 0.0164, "step": 14776 }, { "epoch": 5.82, "learning_rate": 1.5665945329493582e-07, "loss": 0.0148, "step": 14777 }, { "epoch": 5.82, "learning_rate": 1.5663452727473384e-07, "loss": 0.0131, "step": 14778 }, { "epoch": 5.82, "learning_rate": 1.566096019613232e-07, "loss": 0.0122, "step": 14779 }, { "epoch": 5.82, "learning_rate": 1.5658467735511e-07, "loss": 0.0121, "step": 14780 }, { "epoch": 5.83, "learning_rate": 1.5655975345650065e-07, "loss": 0.0119, "step": 14781 }, { "epoch": 5.83, "learning_rate": 1.565348302659013e-07, "loss": 0.0162, "step": 14782 }, { "epoch": 5.83, "learning_rate": 1.565099077837182e-07, "loss": 0.0146, "step": 14783 }, { "epoch": 5.83, "learning_rate": 1.5648498601035746e-07, "loss": 0.0148, "step": 14784 }, { "epoch": 5.83, "learning_rate": 1.5646006494622522e-07, "loss": 0.0127, "step": 14785 }, { "epoch": 5.83, "learning_rate": 1.5643514459172785e-07, "loss": 0.0114, "step": 14786 }, { "epoch": 5.83, "learning_rate": 1.5641022494727144e-07, "loss": 0.0148, "step": 14787 }, { "epoch": 5.83, "learning_rate": 1.563853060132621e-07, "loss": 0.0134, "step": 14788 }, { "epoch": 5.83, "learning_rate": 1.563603877901059e-07, "loss": 0.0141, "step": 14789 }, { "epoch": 5.83, "learning_rate": 1.5633547027820915e-07, "loss": 0.0142, "step": 14790 }, { "epoch": 5.83, "learning_rate": 1.5631055347797783e-07, "loss": 0.0115, "step": 14791 }, { "epoch": 5.83, "learning_rate": 1.5628563738981812e-07, "loss": 0.014, "step": 14792 }, { "epoch": 5.83, "learning_rate": 1.56260722014136e-07, "loss": 0.0108, "step": 14793 }, { "epoch": 5.83, "learning_rate": 1.562358073513377e-07, "loss": 0.0135, "step": 14794 }, { "epoch": 5.83, "learning_rate": 1.5621089340182923e-07, "loss": 0.011, "step": 14795 }, { "epoch": 5.83, "learning_rate": 1.5618598016601666e-07, "loss": 0.0136, "step": 14796 }, { "epoch": 5.83, "learning_rate": 1.56161067644306e-07, "loss": 0.0127, "step": 14797 }, { "epoch": 5.83, "learning_rate": 1.5613615583710326e-07, "loss": 0.0137, "step": 14798 }, { "epoch": 5.83, "learning_rate": 1.5611124474481457e-07, "loss": 0.0112, "step": 14799 }, { "epoch": 5.83, "learning_rate": 1.5608633436784587e-07, "loss": 0.0128, "step": 14800 }, { "epoch": 5.83, "learning_rate": 1.560614247066032e-07, "loss": 0.0142, "step": 14801 }, { "epoch": 5.83, "learning_rate": 1.5603651576149246e-07, "loss": 0.0117, "step": 14802 }, { "epoch": 5.83, "learning_rate": 1.5601160753291976e-07, "loss": 0.0167, "step": 14803 }, { "epoch": 5.83, "learning_rate": 1.5598670002129102e-07, "loss": 0.0185, "step": 14804 }, { "epoch": 5.83, "learning_rate": 1.5596179322701218e-07, "loss": 0.0149, "step": 14805 }, { "epoch": 5.84, "learning_rate": 1.5593688715048916e-07, "loss": 0.0125, "step": 14806 }, { "epoch": 5.84, "learning_rate": 1.5591198179212789e-07, "loss": 0.015, "step": 14807 }, { "epoch": 5.84, "learning_rate": 1.5588707715233426e-07, "loss": 0.0141, "step": 14808 }, { "epoch": 5.84, "learning_rate": 1.5586217323151432e-07, "loss": 0.0151, "step": 14809 }, { "epoch": 5.84, "learning_rate": 1.5583727003007387e-07, "loss": 0.014, "step": 14810 }, { "epoch": 5.84, "learning_rate": 1.5581236754841876e-07, "loss": 0.0119, "step": 14811 }, { "epoch": 5.84, "learning_rate": 1.5578746578695492e-07, "loss": 0.0123, "step": 14812 }, { "epoch": 5.84, "learning_rate": 1.557625647460882e-07, "loss": 0.0103, "step": 14813 }, { "epoch": 5.84, "learning_rate": 1.5573766442622445e-07, "loss": 0.0127, "step": 14814 }, { "epoch": 5.84, "learning_rate": 1.557127648277695e-07, "loss": 0.0141, "step": 14815 }, { "epoch": 5.84, "learning_rate": 1.5568786595112909e-07, "loss": 0.0125, "step": 14816 }, { "epoch": 5.84, "learning_rate": 1.556629677967092e-07, "loss": 0.0154, "step": 14817 }, { "epoch": 5.84, "learning_rate": 1.5563807036491555e-07, "loss": 0.0155, "step": 14818 }, { "epoch": 5.84, "learning_rate": 1.556131736561539e-07, "loss": 0.015, "step": 14819 }, { "epoch": 5.84, "learning_rate": 1.5558827767083003e-07, "loss": 0.0133, "step": 14820 }, { "epoch": 5.84, "learning_rate": 1.5556338240934978e-07, "loss": 0.0151, "step": 14821 }, { "epoch": 5.84, "learning_rate": 1.555384878721189e-07, "loss": 0.0135, "step": 14822 }, { "epoch": 5.84, "learning_rate": 1.5551359405954306e-07, "loss": 0.0158, "step": 14823 }, { "epoch": 5.84, "learning_rate": 1.5548870097202803e-07, "loss": 0.0148, "step": 14824 }, { "epoch": 5.84, "learning_rate": 1.5546380860997946e-07, "loss": 0.0133, "step": 14825 }, { "epoch": 5.84, "learning_rate": 1.5543891697380319e-07, "loss": 0.0161, "step": 14826 }, { "epoch": 5.84, "learning_rate": 1.5541402606390483e-07, "loss": 0.0133, "step": 14827 }, { "epoch": 5.84, "learning_rate": 1.5538913588069012e-07, "loss": 0.0127, "step": 14828 }, { "epoch": 5.84, "learning_rate": 1.5536424642456463e-07, "loss": 0.0114, "step": 14829 }, { "epoch": 5.84, "learning_rate": 1.5533935769593414e-07, "loss": 0.0149, "step": 14830 }, { "epoch": 5.85, "learning_rate": 1.553144696952042e-07, "loss": 0.0131, "step": 14831 }, { "epoch": 5.85, "learning_rate": 1.5528958242278048e-07, "loss": 0.0166, "step": 14832 }, { "epoch": 5.85, "learning_rate": 1.5526469587906865e-07, "loss": 0.0138, "step": 14833 }, { "epoch": 5.85, "learning_rate": 1.5523981006447427e-07, "loss": 0.0127, "step": 14834 }, { "epoch": 5.85, "learning_rate": 1.5521492497940301e-07, "loss": 0.0122, "step": 14835 }, { "epoch": 5.85, "learning_rate": 1.5519004062426039e-07, "loss": 0.0152, "step": 14836 }, { "epoch": 5.85, "learning_rate": 1.5516515699945198e-07, "loss": 0.0172, "step": 14837 }, { "epoch": 5.85, "learning_rate": 1.551402741053833e-07, "loss": 0.0159, "step": 14838 }, { "epoch": 5.85, "learning_rate": 1.5511539194246006e-07, "loss": 0.0109, "step": 14839 }, { "epoch": 5.85, "learning_rate": 1.5509051051108773e-07, "loss": 0.0102, "step": 14840 }, { "epoch": 5.85, "learning_rate": 1.5506562981167178e-07, "loss": 0.0124, "step": 14841 }, { "epoch": 5.85, "learning_rate": 1.5504074984461772e-07, "loss": 0.0114, "step": 14842 }, { "epoch": 5.85, "learning_rate": 1.550158706103312e-07, "loss": 0.0151, "step": 14843 }, { "epoch": 5.85, "learning_rate": 1.549909921092176e-07, "loss": 0.0154, "step": 14844 }, { "epoch": 5.85, "learning_rate": 1.5496611434168243e-07, "loss": 0.0176, "step": 14845 }, { "epoch": 5.85, "learning_rate": 1.5494123730813115e-07, "loss": 0.0145, "step": 14846 }, { "epoch": 5.85, "learning_rate": 1.5491636100896915e-07, "loss": 0.0132, "step": 14847 }, { "epoch": 5.85, "learning_rate": 1.5489148544460203e-07, "loss": 0.0106, "step": 14848 }, { "epoch": 5.85, "learning_rate": 1.5486661061543513e-07, "loss": 0.013, "step": 14849 }, { "epoch": 5.85, "learning_rate": 1.5484173652187389e-07, "loss": 0.0131, "step": 14850 }, { "epoch": 5.85, "learning_rate": 1.5481686316432366e-07, "loss": 0.0124, "step": 14851 }, { "epoch": 5.85, "learning_rate": 1.547919905431899e-07, "loss": 0.0133, "step": 14852 }, { "epoch": 5.85, "learning_rate": 1.5476711865887803e-07, "loss": 0.0143, "step": 14853 }, { "epoch": 5.85, "learning_rate": 1.5474224751179334e-07, "loss": 0.0126, "step": 14854 }, { "epoch": 5.85, "learning_rate": 1.5471737710234129e-07, "loss": 0.0121, "step": 14855 }, { "epoch": 5.85, "learning_rate": 1.5469250743092706e-07, "loss": 0.0143, "step": 14856 }, { "epoch": 5.86, "learning_rate": 1.5466763849795622e-07, "loss": 0.0117, "step": 14857 }, { "epoch": 5.86, "learning_rate": 1.5464277030383392e-07, "loss": 0.0152, "step": 14858 }, { "epoch": 5.86, "learning_rate": 1.5461790284896554e-07, "loss": 0.0151, "step": 14859 }, { "epoch": 5.86, "learning_rate": 1.5459303613375628e-07, "loss": 0.0131, "step": 14860 }, { "epoch": 5.86, "learning_rate": 1.5456817015861162e-07, "loss": 0.0168, "step": 14861 }, { "epoch": 5.86, "learning_rate": 1.5454330492393676e-07, "loss": 0.0121, "step": 14862 }, { "epoch": 5.86, "learning_rate": 1.545184404301369e-07, "loss": 0.0124, "step": 14863 }, { "epoch": 5.86, "learning_rate": 1.544935766776174e-07, "loss": 0.0148, "step": 14864 }, { "epoch": 5.86, "learning_rate": 1.544687136667833e-07, "loss": 0.0125, "step": 14865 }, { "epoch": 5.86, "learning_rate": 1.5444385139804004e-07, "loss": 0.0141, "step": 14866 }, { "epoch": 5.86, "learning_rate": 1.544189898717928e-07, "loss": 0.0144, "step": 14867 }, { "epoch": 5.86, "learning_rate": 1.5439412908844672e-07, "loss": 0.0135, "step": 14868 }, { "epoch": 5.86, "learning_rate": 1.5436926904840697e-07, "loss": 0.0137, "step": 14869 }, { "epoch": 5.86, "learning_rate": 1.5434440975207886e-07, "loss": 0.0131, "step": 14870 }, { "epoch": 5.86, "learning_rate": 1.543195511998675e-07, "loss": 0.0106, "step": 14871 }, { "epoch": 5.86, "learning_rate": 1.5429469339217797e-07, "loss": 0.0121, "step": 14872 }, { "epoch": 5.86, "learning_rate": 1.5426983632941549e-07, "loss": 0.0126, "step": 14873 }, { "epoch": 5.86, "learning_rate": 1.542449800119852e-07, "loss": 0.0133, "step": 14874 }, { "epoch": 5.86, "learning_rate": 1.5422012444029219e-07, "loss": 0.0128, "step": 14875 }, { "epoch": 5.86, "learning_rate": 1.5419526961474153e-07, "loss": 0.0169, "step": 14876 }, { "epoch": 5.86, "learning_rate": 1.541704155357384e-07, "loss": 0.0151, "step": 14877 }, { "epoch": 5.86, "learning_rate": 1.541455622036878e-07, "loss": 0.0121, "step": 14878 }, { "epoch": 5.86, "learning_rate": 1.541207096189949e-07, "loss": 0.0131, "step": 14879 }, { "epoch": 5.86, "learning_rate": 1.5409585778206472e-07, "loss": 0.0134, "step": 14880 }, { "epoch": 5.86, "learning_rate": 1.5407100669330224e-07, "loss": 0.0125, "step": 14881 }, { "epoch": 5.87, "learning_rate": 1.540461563531125e-07, "loss": 0.0163, "step": 14882 }, { "epoch": 5.87, "learning_rate": 1.5402130676190065e-07, "loss": 0.0115, "step": 14883 }, { "epoch": 5.87, "learning_rate": 1.5399645792007163e-07, "loss": 0.012, "step": 14884 }, { "epoch": 5.87, "learning_rate": 1.539716098280304e-07, "loss": 0.0099, "step": 14885 }, { "epoch": 5.87, "learning_rate": 1.5394676248618201e-07, "loss": 0.0114, "step": 14886 }, { "epoch": 5.87, "learning_rate": 1.539219158949313e-07, "loss": 0.0134, "step": 14887 }, { "epoch": 5.87, "learning_rate": 1.538970700546834e-07, "loss": 0.0123, "step": 14888 }, { "epoch": 5.87, "learning_rate": 1.538722249658432e-07, "loss": 0.0119, "step": 14889 }, { "epoch": 5.87, "learning_rate": 1.5384738062881562e-07, "loss": 0.0133, "step": 14890 }, { "epoch": 5.87, "learning_rate": 1.538225370440055e-07, "loss": 0.016, "step": 14891 }, { "epoch": 5.87, "learning_rate": 1.5379769421181797e-07, "loss": 0.013, "step": 14892 }, { "epoch": 5.87, "learning_rate": 1.5377285213265778e-07, "loss": 0.0144, "step": 14893 }, { "epoch": 5.87, "learning_rate": 1.5374801080692985e-07, "loss": 0.0138, "step": 14894 }, { "epoch": 5.87, "learning_rate": 1.5372317023503906e-07, "loss": 0.0116, "step": 14895 }, { "epoch": 5.87, "learning_rate": 1.5369833041739019e-07, "loss": 0.0097, "step": 14896 }, { "epoch": 5.87, "learning_rate": 1.5367349135438822e-07, "loss": 0.0118, "step": 14897 }, { "epoch": 5.87, "learning_rate": 1.536486530464379e-07, "loss": 0.0116, "step": 14898 }, { "epoch": 5.87, "learning_rate": 1.536238154939441e-07, "loss": 0.0136, "step": 14899 }, { "epoch": 5.87, "learning_rate": 1.5359897869731159e-07, "loss": 0.0169, "step": 14900 }, { "epoch": 5.87, "learning_rate": 1.5357414265694526e-07, "loss": 0.0113, "step": 14901 }, { "epoch": 5.87, "learning_rate": 1.5354930737324986e-07, "loss": 0.0129, "step": 14902 }, { "epoch": 5.87, "learning_rate": 1.5352447284663011e-07, "loss": 0.012, "step": 14903 }, { "epoch": 5.87, "learning_rate": 1.5349963907749087e-07, "loss": 0.0129, "step": 14904 }, { "epoch": 5.87, "learning_rate": 1.5347480606623673e-07, "loss": 0.0161, "step": 14905 }, { "epoch": 5.87, "learning_rate": 1.5344997381327262e-07, "loss": 0.0136, "step": 14906 }, { "epoch": 5.87, "learning_rate": 1.5342514231900322e-07, "loss": 0.0137, "step": 14907 }, { "epoch": 5.88, "learning_rate": 1.5340031158383318e-07, "loss": 0.0119, "step": 14908 }, { "epoch": 5.88, "learning_rate": 1.5337548160816715e-07, "loss": 0.0144, "step": 14909 }, { "epoch": 5.88, "learning_rate": 1.5335065239241e-07, "loss": 0.0145, "step": 14910 }, { "epoch": 5.88, "learning_rate": 1.5332582393696633e-07, "loss": 0.0131, "step": 14911 }, { "epoch": 5.88, "learning_rate": 1.5330099624224077e-07, "loss": 0.0147, "step": 14912 }, { "epoch": 5.88, "learning_rate": 1.5327616930863795e-07, "loss": 0.0099, "step": 14913 }, { "epoch": 5.88, "learning_rate": 1.5325134313656262e-07, "loss": 0.0128, "step": 14914 }, { "epoch": 5.88, "learning_rate": 1.5322651772641934e-07, "loss": 0.0128, "step": 14915 }, { "epoch": 5.88, "learning_rate": 1.5320169307861278e-07, "loss": 0.011, "step": 14916 }, { "epoch": 5.88, "learning_rate": 1.5317686919354745e-07, "loss": 0.0126, "step": 14917 }, { "epoch": 5.88, "learning_rate": 1.5315204607162794e-07, "loss": 0.0111, "step": 14918 }, { "epoch": 5.88, "learning_rate": 1.5312722371325895e-07, "loss": 0.0134, "step": 14919 }, { "epoch": 5.88, "learning_rate": 1.5310240211884496e-07, "loss": 0.012, "step": 14920 }, { "epoch": 5.88, "learning_rate": 1.530775812887905e-07, "loss": 0.0127, "step": 14921 }, { "epoch": 5.88, "learning_rate": 1.5305276122350015e-07, "loss": 0.0138, "step": 14922 }, { "epoch": 5.88, "learning_rate": 1.530279419233785e-07, "loss": 0.0124, "step": 14923 }, { "epoch": 5.88, "learning_rate": 1.5300312338882998e-07, "loss": 0.0118, "step": 14924 }, { "epoch": 5.88, "learning_rate": 1.5297830562025912e-07, "loss": 0.0146, "step": 14925 }, { "epoch": 5.88, "learning_rate": 1.5295348861807044e-07, "loss": 0.0126, "step": 14926 }, { "epoch": 5.88, "learning_rate": 1.529286723826683e-07, "loss": 0.0164, "step": 14927 }, { "epoch": 5.88, "learning_rate": 1.5290385691445734e-07, "loss": 0.013, "step": 14928 }, { "epoch": 5.88, "learning_rate": 1.5287904221384193e-07, "loss": 0.0107, "step": 14929 }, { "epoch": 5.88, "learning_rate": 1.528542282812265e-07, "loss": 0.0119, "step": 14930 }, { "epoch": 5.88, "learning_rate": 1.5282941511701545e-07, "loss": 0.0137, "step": 14931 }, { "epoch": 5.88, "learning_rate": 1.5280460272161332e-07, "loss": 0.0106, "step": 14932 }, { "epoch": 5.89, "learning_rate": 1.5277979109542443e-07, "loss": 0.013, "step": 14933 }, { "epoch": 5.89, "learning_rate": 1.5275498023885315e-07, "loss": 0.0173, "step": 14934 }, { "epoch": 5.89, "learning_rate": 1.527301701523039e-07, "loss": 0.0137, "step": 14935 }, { "epoch": 5.89, "learning_rate": 1.52705360836181e-07, "loss": 0.0109, "step": 14936 }, { "epoch": 5.89, "learning_rate": 1.5268055229088888e-07, "loss": 0.0113, "step": 14937 }, { "epoch": 5.89, "learning_rate": 1.5265574451683186e-07, "loss": 0.0156, "step": 14938 }, { "epoch": 5.89, "learning_rate": 1.5263093751441422e-07, "loss": 0.0124, "step": 14939 }, { "epoch": 5.89, "learning_rate": 1.5260613128404028e-07, "loss": 0.0111, "step": 14940 }, { "epoch": 5.89, "learning_rate": 1.525813258261144e-07, "loss": 0.0179, "step": 14941 }, { "epoch": 5.89, "learning_rate": 1.5255652114104088e-07, "loss": 0.0146, "step": 14942 }, { "epoch": 5.89, "learning_rate": 1.525317172292239e-07, "loss": 0.0149, "step": 14943 }, { "epoch": 5.89, "learning_rate": 1.525069140910678e-07, "loss": 0.0148, "step": 14944 }, { "epoch": 5.89, "learning_rate": 1.5248211172697678e-07, "loss": 0.0114, "step": 14945 }, { "epoch": 5.89, "learning_rate": 1.524573101373552e-07, "loss": 0.0124, "step": 14946 }, { "epoch": 5.89, "learning_rate": 1.5243250932260718e-07, "loss": 0.0112, "step": 14947 }, { "epoch": 5.89, "learning_rate": 1.5240770928313696e-07, "loss": 0.0139, "step": 14948 }, { "epoch": 5.89, "learning_rate": 1.5238291001934865e-07, "loss": 0.0111, "step": 14949 }, { "epoch": 5.89, "learning_rate": 1.5235811153164666e-07, "loss": 0.0125, "step": 14950 }, { "epoch": 5.89, "learning_rate": 1.5233331382043504e-07, "loss": 0.0176, "step": 14951 }, { "epoch": 5.89, "learning_rate": 1.5230851688611794e-07, "loss": 0.013, "step": 14952 }, { "epoch": 5.89, "learning_rate": 1.5228372072909944e-07, "loss": 0.0171, "step": 14953 }, { "epoch": 5.89, "learning_rate": 1.5225892534978388e-07, "loss": 0.012, "step": 14954 }, { "epoch": 5.89, "learning_rate": 1.5223413074857527e-07, "loss": 0.0144, "step": 14955 }, { "epoch": 5.89, "learning_rate": 1.5220933692587775e-07, "loss": 0.0134, "step": 14956 }, { "epoch": 5.89, "learning_rate": 1.5218454388209537e-07, "loss": 0.0147, "step": 14957 }, { "epoch": 5.9, "learning_rate": 1.5215975161763218e-07, "loss": 0.0156, "step": 14958 }, { "epoch": 5.9, "learning_rate": 1.5213496013289242e-07, "loss": 0.0144, "step": 14959 }, { "epoch": 5.9, "learning_rate": 1.521101694282801e-07, "loss": 0.0144, "step": 14960 }, { "epoch": 5.9, "learning_rate": 1.5208537950419917e-07, "loss": 0.0156, "step": 14961 }, { "epoch": 5.9, "learning_rate": 1.5206059036105372e-07, "loss": 0.0155, "step": 14962 }, { "epoch": 5.9, "learning_rate": 1.5203580199924783e-07, "loss": 0.0132, "step": 14963 }, { "epoch": 5.9, "learning_rate": 1.5201101441918546e-07, "loss": 0.014, "step": 14964 }, { "epoch": 5.9, "learning_rate": 1.5198622762127063e-07, "loss": 0.0146, "step": 14965 }, { "epoch": 5.9, "learning_rate": 1.5196144160590732e-07, "loss": 0.0145, "step": 14966 }, { "epoch": 5.9, "learning_rate": 1.519366563734994e-07, "loss": 0.014, "step": 14967 }, { "epoch": 5.9, "learning_rate": 1.5191187192445104e-07, "loss": 0.0131, "step": 14968 }, { "epoch": 5.9, "learning_rate": 1.5188708825916608e-07, "loss": 0.0145, "step": 14969 }, { "epoch": 5.9, "learning_rate": 1.5186230537804844e-07, "loss": 0.0144, "step": 14970 }, { "epoch": 5.9, "learning_rate": 1.5183752328150198e-07, "loss": 0.015, "step": 14971 }, { "epoch": 5.9, "learning_rate": 1.518127419699308e-07, "loss": 0.0123, "step": 14972 }, { "epoch": 5.9, "learning_rate": 1.517879614437387e-07, "loss": 0.0142, "step": 14973 }, { "epoch": 5.9, "learning_rate": 1.5176318170332952e-07, "loss": 0.0116, "step": 14974 }, { "epoch": 5.9, "learning_rate": 1.5173840274910716e-07, "loss": 0.0114, "step": 14975 }, { "epoch": 5.9, "learning_rate": 1.517136245814754e-07, "loss": 0.014, "step": 14976 }, { "epoch": 5.9, "learning_rate": 1.516888472008383e-07, "loss": 0.0146, "step": 14977 }, { "epoch": 5.9, "learning_rate": 1.5166407060759954e-07, "loss": 0.0113, "step": 14978 }, { "epoch": 5.9, "learning_rate": 1.5163929480216297e-07, "loss": 0.0109, "step": 14979 }, { "epoch": 5.9, "learning_rate": 1.5161451978493233e-07, "loss": 0.0133, "step": 14980 }, { "epoch": 5.9, "learning_rate": 1.515897455563115e-07, "loss": 0.0127, "step": 14981 }, { "epoch": 5.9, "learning_rate": 1.515649721167043e-07, "loss": 0.0125, "step": 14982 }, { "epoch": 5.9, "learning_rate": 1.515401994665144e-07, "loss": 0.0121, "step": 14983 }, { "epoch": 5.91, "learning_rate": 1.5151542760614565e-07, "loss": 0.0121, "step": 14984 }, { "epoch": 5.91, "learning_rate": 1.514906565360017e-07, "loss": 0.0156, "step": 14985 }, { "epoch": 5.91, "learning_rate": 1.514658862564863e-07, "loss": 0.01, "step": 14986 }, { "epoch": 5.91, "learning_rate": 1.5144111676800325e-07, "loss": 0.0112, "step": 14987 }, { "epoch": 5.91, "learning_rate": 1.514163480709562e-07, "loss": 0.0131, "step": 14988 }, { "epoch": 5.91, "learning_rate": 1.513915801657487e-07, "loss": 0.0127, "step": 14989 }, { "epoch": 5.91, "learning_rate": 1.5136681305278468e-07, "loss": 0.0141, "step": 14990 }, { "epoch": 5.91, "learning_rate": 1.5134204673246768e-07, "loss": 0.0108, "step": 14991 }, { "epoch": 5.91, "learning_rate": 1.513172812052014e-07, "loss": 0.0122, "step": 14992 }, { "epoch": 5.91, "learning_rate": 1.5129251647138931e-07, "loss": 0.011, "step": 14993 }, { "epoch": 5.91, "learning_rate": 1.512677525314353e-07, "loss": 0.0161, "step": 14994 }, { "epoch": 5.91, "learning_rate": 1.5124298938574287e-07, "loss": 0.0126, "step": 14995 }, { "epoch": 5.91, "learning_rate": 1.512182270347156e-07, "loss": 0.0137, "step": 14996 }, { "epoch": 5.91, "learning_rate": 1.5119346547875707e-07, "loss": 0.0132, "step": 14997 }, { "epoch": 5.91, "learning_rate": 1.5116870471827082e-07, "loss": 0.0113, "step": 14998 }, { "epoch": 5.91, "learning_rate": 1.5114394475366054e-07, "loss": 0.0126, "step": 14999 }, { "epoch": 5.91, "learning_rate": 1.5111918558532968e-07, "loss": 0.0129, "step": 15000 }, { "epoch": 5.91, "learning_rate": 1.5109442721368183e-07, "loss": 0.015, "step": 15001 }, { "epoch": 5.91, "learning_rate": 1.5106966963912046e-07, "loss": 0.0144, "step": 15002 }, { "epoch": 5.91, "learning_rate": 1.5104491286204913e-07, "loss": 0.0115, "step": 15003 }, { "epoch": 5.91, "learning_rate": 1.5102015688287125e-07, "loss": 0.0126, "step": 15004 }, { "epoch": 5.91, "learning_rate": 1.5099540170199045e-07, "loss": 0.0147, "step": 15005 }, { "epoch": 5.91, "learning_rate": 1.5097064731981013e-07, "loss": 0.0123, "step": 15006 }, { "epoch": 5.91, "learning_rate": 1.509458937367337e-07, "loss": 0.0135, "step": 15007 }, { "epoch": 5.91, "learning_rate": 1.5092114095316468e-07, "loss": 0.0142, "step": 15008 }, { "epoch": 5.92, "learning_rate": 1.5089638896950648e-07, "loss": 0.0124, "step": 15009 }, { "epoch": 5.92, "learning_rate": 1.5087163778616247e-07, "loss": 0.0128, "step": 15010 }, { "epoch": 5.92, "learning_rate": 1.5084688740353607e-07, "loss": 0.0138, "step": 15011 }, { "epoch": 5.92, "learning_rate": 1.5082213782203075e-07, "loss": 0.0163, "step": 15012 }, { "epoch": 5.92, "learning_rate": 1.5079738904204987e-07, "loss": 0.0124, "step": 15013 }, { "epoch": 5.92, "learning_rate": 1.5077264106399674e-07, "loss": 0.0123, "step": 15014 }, { "epoch": 5.92, "learning_rate": 1.5074789388827475e-07, "loss": 0.0141, "step": 15015 }, { "epoch": 5.92, "learning_rate": 1.5072314751528717e-07, "loss": 0.0125, "step": 15016 }, { "epoch": 5.92, "learning_rate": 1.5069840194543747e-07, "loss": 0.0138, "step": 15017 }, { "epoch": 5.92, "learning_rate": 1.5067365717912885e-07, "loss": 0.0122, "step": 15018 }, { "epoch": 5.92, "learning_rate": 1.506489132167647e-07, "loss": 0.0134, "step": 15019 }, { "epoch": 5.92, "learning_rate": 1.5062417005874813e-07, "loss": 0.0136, "step": 15020 }, { "epoch": 5.92, "learning_rate": 1.5059942770548268e-07, "loss": 0.0127, "step": 15021 }, { "epoch": 5.92, "learning_rate": 1.5057468615737145e-07, "loss": 0.015, "step": 15022 }, { "epoch": 5.92, "learning_rate": 1.5054994541481773e-07, "loss": 0.0121, "step": 15023 }, { "epoch": 5.92, "learning_rate": 1.5052520547822473e-07, "loss": 0.0122, "step": 15024 }, { "epoch": 5.92, "learning_rate": 1.5050046634799568e-07, "loss": 0.0129, "step": 15025 }, { "epoch": 5.92, "learning_rate": 1.504757280245338e-07, "loss": 0.0145, "step": 15026 }, { "epoch": 5.92, "learning_rate": 1.5045099050824226e-07, "loss": 0.0121, "step": 15027 }, { "epoch": 5.92, "learning_rate": 1.5042625379952432e-07, "loss": 0.0145, "step": 15028 }, { "epoch": 5.92, "learning_rate": 1.50401517898783e-07, "loss": 0.0138, "step": 15029 }, { "epoch": 5.92, "learning_rate": 1.503767828064217e-07, "loss": 0.0147, "step": 15030 }, { "epoch": 5.92, "learning_rate": 1.5035204852284333e-07, "loss": 0.0123, "step": 15031 }, { "epoch": 5.92, "learning_rate": 1.5032731504845116e-07, "loss": 0.0118, "step": 15032 }, { "epoch": 5.92, "learning_rate": 1.5030258238364818e-07, "loss": 0.0101, "step": 15033 }, { "epoch": 5.93, "learning_rate": 1.5027785052883764e-07, "loss": 0.0117, "step": 15034 }, { "epoch": 5.93, "learning_rate": 1.5025311948442256e-07, "loss": 0.0173, "step": 15035 }, { "epoch": 5.93, "learning_rate": 1.5022838925080605e-07, "loss": 0.0122, "step": 15036 }, { "epoch": 5.93, "learning_rate": 1.5020365982839115e-07, "loss": 0.0137, "step": 15037 }, { "epoch": 5.93, "learning_rate": 1.5017893121758082e-07, "loss": 0.0146, "step": 15038 }, { "epoch": 5.93, "learning_rate": 1.5015420341877825e-07, "loss": 0.0117, "step": 15039 }, { "epoch": 5.93, "learning_rate": 1.5012947643238643e-07, "loss": 0.0177, "step": 15040 }, { "epoch": 5.93, "learning_rate": 1.5010475025880835e-07, "loss": 0.0154, "step": 15041 }, { "epoch": 5.93, "learning_rate": 1.5008002489844693e-07, "loss": 0.0144, "step": 15042 }, { "epoch": 5.93, "learning_rate": 1.5005530035170532e-07, "loss": 0.0127, "step": 15043 }, { "epoch": 5.93, "learning_rate": 1.500305766189864e-07, "loss": 0.0134, "step": 15044 }, { "epoch": 5.93, "learning_rate": 1.5000585370069315e-07, "loss": 0.0138, "step": 15045 }, { "epoch": 5.93, "learning_rate": 1.499811315972285e-07, "loss": 0.014, "step": 15046 }, { "epoch": 5.93, "learning_rate": 1.4995641030899532e-07, "loss": 0.0118, "step": 15047 }, { "epoch": 5.93, "learning_rate": 1.4993168983639665e-07, "loss": 0.0153, "step": 15048 }, { "epoch": 5.93, "learning_rate": 1.499069701798353e-07, "loss": 0.0149, "step": 15049 }, { "epoch": 5.93, "learning_rate": 1.4988225133971424e-07, "loss": 0.0133, "step": 15050 }, { "epoch": 5.93, "learning_rate": 1.4985753331643628e-07, "loss": 0.0125, "step": 15051 }, { "epoch": 5.93, "learning_rate": 1.4983281611040438e-07, "loss": 0.0153, "step": 15052 }, { "epoch": 5.93, "learning_rate": 1.498080997220213e-07, "loss": 0.0156, "step": 15053 }, { "epoch": 5.93, "learning_rate": 1.4978338415168992e-07, "loss": 0.0117, "step": 15054 }, { "epoch": 5.93, "learning_rate": 1.4975866939981305e-07, "loss": 0.0149, "step": 15055 }, { "epoch": 5.93, "learning_rate": 1.4973395546679343e-07, "loss": 0.0134, "step": 15056 }, { "epoch": 5.93, "learning_rate": 1.4970924235303403e-07, "loss": 0.0138, "step": 15057 }, { "epoch": 5.93, "learning_rate": 1.4968453005893755e-07, "loss": 0.0143, "step": 15058 }, { "epoch": 5.93, "learning_rate": 1.4965981858490675e-07, "loss": 0.0127, "step": 15059 }, { "epoch": 5.94, "learning_rate": 1.496351079313443e-07, "loss": 0.0099, "step": 15060 }, { "epoch": 5.94, "learning_rate": 1.4961039809865318e-07, "loss": 0.0129, "step": 15061 }, { "epoch": 5.94, "learning_rate": 1.4958568908723593e-07, "loss": 0.0167, "step": 15062 }, { "epoch": 5.94, "learning_rate": 1.4956098089749536e-07, "loss": 0.015, "step": 15063 }, { "epoch": 5.94, "learning_rate": 1.4953627352983412e-07, "loss": 0.0148, "step": 15064 }, { "epoch": 5.94, "learning_rate": 1.4951156698465487e-07, "loss": 0.0134, "step": 15065 }, { "epoch": 5.94, "learning_rate": 1.4948686126236043e-07, "loss": 0.0153, "step": 15066 }, { "epoch": 5.94, "learning_rate": 1.4946215636335337e-07, "loss": 0.0113, "step": 15067 }, { "epoch": 5.94, "learning_rate": 1.4943745228803635e-07, "loss": 0.0122, "step": 15068 }, { "epoch": 5.94, "learning_rate": 1.4941274903681195e-07, "loss": 0.0127, "step": 15069 }, { "epoch": 5.94, "learning_rate": 1.4938804661008294e-07, "loss": 0.011, "step": 15070 }, { "epoch": 5.94, "learning_rate": 1.4936334500825184e-07, "loss": 0.0156, "step": 15071 }, { "epoch": 5.94, "learning_rate": 1.4933864423172118e-07, "loss": 0.013, "step": 15072 }, { "epoch": 5.94, "learning_rate": 1.4931394428089364e-07, "loss": 0.0144, "step": 15073 }, { "epoch": 5.94, "learning_rate": 1.4928924515617183e-07, "loss": 0.0146, "step": 15074 }, { "epoch": 5.94, "learning_rate": 1.4926454685795824e-07, "loss": 0.0136, "step": 15075 }, { "epoch": 5.94, "learning_rate": 1.4923984938665544e-07, "loss": 0.0143, "step": 15076 }, { "epoch": 5.94, "learning_rate": 1.4921515274266594e-07, "loss": 0.0134, "step": 15077 }, { "epoch": 5.94, "learning_rate": 1.4919045692639217e-07, "loss": 0.0136, "step": 15078 }, { "epoch": 5.94, "learning_rate": 1.4916576193823686e-07, "loss": 0.0117, "step": 15079 }, { "epoch": 5.94, "learning_rate": 1.491410677786023e-07, "loss": 0.0159, "step": 15080 }, { "epoch": 5.94, "learning_rate": 1.4911637444789107e-07, "loss": 0.0154, "step": 15081 }, { "epoch": 5.94, "learning_rate": 1.490916819465055e-07, "loss": 0.0113, "step": 15082 }, { "epoch": 5.94, "learning_rate": 1.490669902748483e-07, "loss": 0.013, "step": 15083 }, { "epoch": 5.94, "learning_rate": 1.4904229943332167e-07, "loss": 0.0139, "step": 15084 }, { "epoch": 5.95, "learning_rate": 1.4901760942232814e-07, "loss": 0.0146, "step": 15085 }, { "epoch": 5.95, "learning_rate": 1.4899292024227013e-07, "loss": 0.0111, "step": 15086 }, { "epoch": 5.95, "learning_rate": 1.4896823189354986e-07, "loss": 0.0117, "step": 15087 }, { "epoch": 5.95, "learning_rate": 1.4894354437656996e-07, "loss": 0.0142, "step": 15088 }, { "epoch": 5.95, "learning_rate": 1.4891885769173268e-07, "loss": 0.0123, "step": 15089 }, { "epoch": 5.95, "learning_rate": 1.488941718394404e-07, "loss": 0.0136, "step": 15090 }, { "epoch": 5.95, "learning_rate": 1.488694868200954e-07, "loss": 0.0125, "step": 15091 }, { "epoch": 5.95, "learning_rate": 1.488448026341001e-07, "loss": 0.0101, "step": 15092 }, { "epoch": 5.95, "learning_rate": 1.4882011928185674e-07, "loss": 0.0117, "step": 15093 }, { "epoch": 5.95, "learning_rate": 1.4879543676376767e-07, "loss": 0.0153, "step": 15094 }, { "epoch": 5.95, "learning_rate": 1.487707550802351e-07, "loss": 0.0124, "step": 15095 }, { "epoch": 5.95, "learning_rate": 1.487460742316614e-07, "loss": 0.0133, "step": 15096 }, { "epoch": 5.95, "learning_rate": 1.4872139421844882e-07, "loss": 0.0134, "step": 15097 }, { "epoch": 5.95, "learning_rate": 1.486967150409996e-07, "loss": 0.0134, "step": 15098 }, { "epoch": 5.95, "learning_rate": 1.486720366997159e-07, "loss": 0.012, "step": 15099 }, { "epoch": 5.95, "learning_rate": 1.4864735919499993e-07, "loss": 0.0127, "step": 15100 }, { "epoch": 5.95, "learning_rate": 1.4862268252725409e-07, "loss": 0.0146, "step": 15101 }, { "epoch": 5.95, "learning_rate": 1.4859800669688038e-07, "loss": 0.0135, "step": 15102 }, { "epoch": 5.95, "learning_rate": 1.4857333170428103e-07, "loss": 0.0133, "step": 15103 }, { "epoch": 5.95, "learning_rate": 1.4854865754985826e-07, "loss": 0.0104, "step": 15104 }, { "epoch": 5.95, "learning_rate": 1.4852398423401406e-07, "loss": 0.0113, "step": 15105 }, { "epoch": 5.95, "learning_rate": 1.484993117571508e-07, "loss": 0.0131, "step": 15106 }, { "epoch": 5.95, "learning_rate": 1.4847464011967046e-07, "loss": 0.0129, "step": 15107 }, { "epoch": 5.95, "learning_rate": 1.4844996932197517e-07, "loss": 0.0121, "step": 15108 }, { "epoch": 5.95, "learning_rate": 1.4842529936446698e-07, "loss": 0.0132, "step": 15109 }, { "epoch": 5.95, "learning_rate": 1.484006302475481e-07, "loss": 0.0144, "step": 15110 }, { "epoch": 5.96, "learning_rate": 1.483759619716205e-07, "loss": 0.0127, "step": 15111 }, { "epoch": 5.96, "learning_rate": 1.483512945370863e-07, "loss": 0.0138, "step": 15112 }, { "epoch": 5.96, "learning_rate": 1.483266279443474e-07, "loss": 0.0128, "step": 15113 }, { "epoch": 5.96, "learning_rate": 1.4830196219380603e-07, "loss": 0.0158, "step": 15114 }, { "epoch": 5.96, "learning_rate": 1.4827729728586408e-07, "loss": 0.0146, "step": 15115 }, { "epoch": 5.96, "learning_rate": 1.4825263322092357e-07, "loss": 0.0116, "step": 15116 }, { "epoch": 5.96, "learning_rate": 1.482279699993865e-07, "loss": 0.013, "step": 15117 }, { "epoch": 5.96, "learning_rate": 1.4820330762165476e-07, "loss": 0.0116, "step": 15118 }, { "epoch": 5.96, "learning_rate": 1.4817864608813048e-07, "loss": 0.0112, "step": 15119 }, { "epoch": 5.96, "learning_rate": 1.4815398539921546e-07, "loss": 0.0154, "step": 15120 }, { "epoch": 5.96, "learning_rate": 1.4812932555531173e-07, "loss": 0.0102, "step": 15121 }, { "epoch": 5.96, "learning_rate": 1.4810466655682103e-07, "loss": 0.0134, "step": 15122 }, { "epoch": 5.96, "learning_rate": 1.480800084041455e-07, "loss": 0.0115, "step": 15123 }, { "epoch": 5.96, "learning_rate": 1.4805535109768695e-07, "loss": 0.012, "step": 15124 }, { "epoch": 5.96, "learning_rate": 1.4803069463784721e-07, "loss": 0.016, "step": 15125 }, { "epoch": 5.96, "learning_rate": 1.4800603902502812e-07, "loss": 0.0142, "step": 15126 }, { "epoch": 5.96, "learning_rate": 1.4798138425963153e-07, "loss": 0.0109, "step": 15127 }, { "epoch": 5.96, "learning_rate": 1.479567303420594e-07, "loss": 0.0123, "step": 15128 }, { "epoch": 5.96, "learning_rate": 1.4793207727271347e-07, "loss": 0.0114, "step": 15129 }, { "epoch": 5.96, "learning_rate": 1.4790742505199555e-07, "loss": 0.0146, "step": 15130 }, { "epoch": 5.96, "learning_rate": 1.478827736803074e-07, "loss": 0.0118, "step": 15131 }, { "epoch": 5.96, "learning_rate": 1.4785812315805079e-07, "loss": 0.0131, "step": 15132 }, { "epoch": 5.96, "learning_rate": 1.478334734856276e-07, "loss": 0.0155, "step": 15133 }, { "epoch": 5.96, "learning_rate": 1.478088246634395e-07, "loss": 0.0137, "step": 15134 }, { "epoch": 5.96, "learning_rate": 1.4778417669188822e-07, "loss": 0.0111, "step": 15135 }, { "epoch": 5.97, "learning_rate": 1.4775952957137548e-07, "loss": 0.0114, "step": 15136 }, { "epoch": 5.97, "learning_rate": 1.4773488330230306e-07, "loss": 0.0129, "step": 15137 }, { "epoch": 5.97, "learning_rate": 1.4771023788507256e-07, "loss": 0.014, "step": 15138 }, { "epoch": 5.97, "learning_rate": 1.4768559332008578e-07, "loss": 0.0115, "step": 15139 }, { "epoch": 5.97, "learning_rate": 1.476609496077442e-07, "loss": 0.014, "step": 15140 }, { "epoch": 5.97, "learning_rate": 1.4763630674844968e-07, "loss": 0.0128, "step": 15141 }, { "epoch": 5.97, "learning_rate": 1.4761166474260377e-07, "loss": 0.0132, "step": 15142 }, { "epoch": 5.97, "learning_rate": 1.4758702359060814e-07, "loss": 0.0151, "step": 15143 }, { "epoch": 5.97, "learning_rate": 1.475623832928643e-07, "loss": 0.0121, "step": 15144 }, { "epoch": 5.97, "learning_rate": 1.475377438497739e-07, "loss": 0.0155, "step": 15145 }, { "epoch": 5.97, "learning_rate": 1.4751310526173857e-07, "loss": 0.0109, "step": 15146 }, { "epoch": 5.97, "learning_rate": 1.4748846752915988e-07, "loss": 0.013, "step": 15147 }, { "epoch": 5.97, "learning_rate": 1.4746383065243938e-07, "loss": 0.0146, "step": 15148 }, { "epoch": 5.97, "learning_rate": 1.474391946319785e-07, "loss": 0.0114, "step": 15149 }, { "epoch": 5.97, "learning_rate": 1.4741455946817892e-07, "loss": 0.0144, "step": 15150 }, { "epoch": 5.97, "learning_rate": 1.4738992516144211e-07, "loss": 0.0119, "step": 15151 }, { "epoch": 5.97, "learning_rate": 1.4736529171216957e-07, "loss": 0.0137, "step": 15152 }, { "epoch": 5.97, "learning_rate": 1.4734065912076273e-07, "loss": 0.0131, "step": 15153 }, { "epoch": 5.97, "learning_rate": 1.473160273876232e-07, "loss": 0.0129, "step": 15154 }, { "epoch": 5.97, "learning_rate": 1.4729139651315225e-07, "loss": 0.0128, "step": 15155 }, { "epoch": 5.97, "learning_rate": 1.4726676649775152e-07, "loss": 0.0141, "step": 15156 }, { "epoch": 5.97, "learning_rate": 1.4724213734182233e-07, "loss": 0.0158, "step": 15157 }, { "epoch": 5.97, "learning_rate": 1.4721750904576604e-07, "loss": 0.0109, "step": 15158 }, { "epoch": 5.97, "learning_rate": 1.4719288160998422e-07, "loss": 0.0128, "step": 15159 }, { "epoch": 5.97, "learning_rate": 1.4716825503487815e-07, "loss": 0.0133, "step": 15160 }, { "epoch": 5.98, "learning_rate": 1.4714362932084927e-07, "loss": 0.0112, "step": 15161 }, { "epoch": 5.98, "learning_rate": 1.4711900446829876e-07, "loss": 0.0154, "step": 15162 }, { "epoch": 5.98, "learning_rate": 1.470943804776282e-07, "loss": 0.0122, "step": 15163 }, { "epoch": 5.98, "learning_rate": 1.4706975734923888e-07, "loss": 0.0161, "step": 15164 }, { "epoch": 5.98, "learning_rate": 1.4704513508353202e-07, "loss": 0.0135, "step": 15165 }, { "epoch": 5.98, "learning_rate": 1.47020513680909e-07, "loss": 0.015, "step": 15166 }, { "epoch": 5.98, "learning_rate": 1.46995893141771e-07, "loss": 0.0111, "step": 15167 }, { "epoch": 5.98, "learning_rate": 1.4697127346651947e-07, "loss": 0.0144, "step": 15168 }, { "epoch": 5.98, "learning_rate": 1.4694665465555558e-07, "loss": 0.0121, "step": 15169 }, { "epoch": 5.98, "learning_rate": 1.4692203670928058e-07, "loss": 0.0105, "step": 15170 }, { "epoch": 5.98, "learning_rate": 1.4689741962809566e-07, "loss": 0.0148, "step": 15171 }, { "epoch": 5.98, "learning_rate": 1.4687280341240213e-07, "loss": 0.0152, "step": 15172 }, { "epoch": 5.98, "learning_rate": 1.468481880626012e-07, "loss": 0.0113, "step": 15173 }, { "epoch": 5.98, "learning_rate": 1.46823573579094e-07, "loss": 0.0117, "step": 15174 }, { "epoch": 5.98, "learning_rate": 1.4679895996228176e-07, "loss": 0.0119, "step": 15175 }, { "epoch": 5.98, "learning_rate": 1.4677434721256554e-07, "loss": 0.014, "step": 15176 }, { "epoch": 5.98, "learning_rate": 1.467497353303466e-07, "loss": 0.0115, "step": 15177 }, { "epoch": 5.98, "learning_rate": 1.4672512431602607e-07, "loss": 0.0092, "step": 15178 }, { "epoch": 5.98, "learning_rate": 1.4670051417000501e-07, "loss": 0.0129, "step": 15179 }, { "epoch": 5.98, "learning_rate": 1.466759048926846e-07, "loss": 0.0133, "step": 15180 }, { "epoch": 5.98, "learning_rate": 1.4665129648446586e-07, "loss": 0.0119, "step": 15181 }, { "epoch": 5.98, "learning_rate": 1.4662668894574994e-07, "loss": 0.0118, "step": 15182 }, { "epoch": 5.98, "learning_rate": 1.4660208227693786e-07, "loss": 0.013, "step": 15183 }, { "epoch": 5.98, "learning_rate": 1.4657747647843071e-07, "loss": 0.0097, "step": 15184 }, { "epoch": 5.98, "learning_rate": 1.4655287155062942e-07, "loss": 0.0134, "step": 15185 }, { "epoch": 5.98, "learning_rate": 1.4652826749393513e-07, "loss": 0.0131, "step": 15186 }, { "epoch": 5.99, "learning_rate": 1.4650366430874885e-07, "loss": 0.0161, "step": 15187 }, { "epoch": 5.99, "learning_rate": 1.4647906199547153e-07, "loss": 0.0134, "step": 15188 }, { "epoch": 5.99, "learning_rate": 1.4645446055450408e-07, "loss": 0.0142, "step": 15189 }, { "epoch": 5.99, "learning_rate": 1.4642985998624762e-07, "loss": 0.0143, "step": 15190 }, { "epoch": 5.99, "learning_rate": 1.46405260291103e-07, "loss": 0.0118, "step": 15191 }, { "epoch": 5.99, "learning_rate": 1.4638066146947125e-07, "loss": 0.0116, "step": 15192 }, { "epoch": 5.99, "learning_rate": 1.4635606352175306e-07, "loss": 0.0133, "step": 15193 }, { "epoch": 5.99, "learning_rate": 1.4633146644834964e-07, "loss": 0.0105, "step": 15194 }, { "epoch": 5.99, "learning_rate": 1.4630687024966176e-07, "loss": 0.0127, "step": 15195 }, { "epoch": 5.99, "learning_rate": 1.4628227492609024e-07, "loss": 0.0134, "step": 15196 }, { "epoch": 5.99, "learning_rate": 1.4625768047803603e-07, "loss": 0.0124, "step": 15197 }, { "epoch": 5.99, "learning_rate": 1.4623308690589991e-07, "loss": 0.0116, "step": 15198 }, { "epoch": 5.99, "learning_rate": 1.4620849421008278e-07, "loss": 0.0135, "step": 15199 }, { "epoch": 5.99, "learning_rate": 1.461839023909854e-07, "loss": 0.0146, "step": 15200 }, { "epoch": 5.99, "learning_rate": 1.461593114490087e-07, "loss": 0.0127, "step": 15201 }, { "epoch": 5.99, "learning_rate": 1.4613472138455333e-07, "loss": 0.0169, "step": 15202 }, { "epoch": 5.99, "learning_rate": 1.4611013219802016e-07, "loss": 0.0129, "step": 15203 }, { "epoch": 5.99, "learning_rate": 1.4608554388980995e-07, "loss": 0.0146, "step": 15204 }, { "epoch": 5.99, "learning_rate": 1.4606095646032342e-07, "loss": 0.0132, "step": 15205 }, { "epoch": 5.99, "learning_rate": 1.4603636990996134e-07, "loss": 0.0103, "step": 15206 }, { "epoch": 5.99, "learning_rate": 1.4601178423912432e-07, "loss": 0.0136, "step": 15207 }, { "epoch": 5.99, "learning_rate": 1.4598719944821323e-07, "loss": 0.012, "step": 15208 }, { "epoch": 5.99, "learning_rate": 1.4596261553762875e-07, "loss": 0.0117, "step": 15209 }, { "epoch": 5.99, "learning_rate": 1.4593803250777145e-07, "loss": 0.0127, "step": 15210 }, { "epoch": 5.99, "learning_rate": 1.45913450359042e-07, "loss": 0.0144, "step": 15211 }, { "epoch": 6.0, "learning_rate": 1.458888690918412e-07, "loss": 0.0134, "step": 15212 }, { "epoch": 6.0, "learning_rate": 1.4586428870656958e-07, "loss": 0.0125, "step": 15213 }, { "epoch": 6.0, "learning_rate": 1.4583970920362777e-07, "loss": 0.0133, "step": 15214 }, { "epoch": 6.0, "learning_rate": 1.4581513058341642e-07, "loss": 0.0137, "step": 15215 }, { "epoch": 6.0, "learning_rate": 1.45790552846336e-07, "loss": 0.0113, "step": 15216 }, { "epoch": 6.0, "learning_rate": 1.4576597599278723e-07, "loss": 0.0103, "step": 15217 }, { "epoch": 6.0, "learning_rate": 1.4574140002317067e-07, "loss": 0.0152, "step": 15218 }, { "epoch": 6.0, "learning_rate": 1.4571682493788676e-07, "loss": 0.0173, "step": 15219 }, { "epoch": 6.0, "learning_rate": 1.456922507373361e-07, "loss": 0.0126, "step": 15220 }, { "epoch": 6.0, "learning_rate": 1.4566767742191926e-07, "loss": 0.0116, "step": 15221 }, { "epoch": 6.0, "learning_rate": 1.456431049920367e-07, "loss": 0.0143, "step": 15222 }, { "epoch": 6.0, "learning_rate": 1.4561853344808887e-07, "loss": 0.0142, "step": 15223 }, { "epoch": 6.0, "learning_rate": 1.4559396279047637e-07, "loss": 0.0124, "step": 15224 }, { "epoch": 6.0, "learning_rate": 1.455693930195995e-07, "loss": 0.0111, "step": 15225 }, { "epoch": 6.0, "learning_rate": 1.4554482413585886e-07, "loss": 0.0125, "step": 15226 }, { "epoch": 6.0, "learning_rate": 1.4552025613965482e-07, "loss": 0.0105, "step": 15227 }, { "epoch": 6.0, "learning_rate": 1.454956890313878e-07, "loss": 0.012, "step": 15228 }, { "epoch": 6.0, "learning_rate": 1.4547112281145816e-07, "loss": 0.0135, "step": 15229 }, { "epoch": 6.0, "learning_rate": 1.4544655748026638e-07, "loss": 0.0099, "step": 15230 }, { "epoch": 6.0, "learning_rate": 1.4542199303821282e-07, "loss": 0.0113, "step": 15231 }, { "epoch": 6.0, "learning_rate": 1.4539742948569783e-07, "loss": 0.0112, "step": 15232 }, { "epoch": 6.0, "learning_rate": 1.4537286682312166e-07, "loss": 0.0101, "step": 15233 }, { "epoch": 6.0, "learning_rate": 1.4534830505088485e-07, "loss": 0.0162, "step": 15234 }, { "epoch": 6.0, "learning_rate": 1.4532374416938757e-07, "loss": 0.0141, "step": 15235 }, { "epoch": 6.0, "learning_rate": 1.4529918417903017e-07, "loss": 0.0132, "step": 15236 }, { "epoch": 6.01, "learning_rate": 1.4527462508021296e-07, "loss": 0.0169, "step": 15237 }, { "epoch": 6.01, "learning_rate": 1.4525006687333608e-07, "loss": 0.0149, "step": 15238 }, { "epoch": 6.01, "learning_rate": 1.4522550955879997e-07, "loss": 0.0135, "step": 15239 }, { "epoch": 6.01, "learning_rate": 1.4520095313700482e-07, "loss": 0.0114, "step": 15240 }, { "epoch": 6.01, "learning_rate": 1.4517639760835088e-07, "loss": 0.0131, "step": 15241 }, { "epoch": 6.01, "learning_rate": 1.4515184297323825e-07, "loss": 0.0152, "step": 15242 }, { "epoch": 6.01, "learning_rate": 1.4512728923206727e-07, "loss": 0.0121, "step": 15243 }, { "epoch": 6.01, "learning_rate": 1.451027363852381e-07, "loss": 0.0135, "step": 15244 }, { "epoch": 6.01, "learning_rate": 1.4507818443315087e-07, "loss": 0.0153, "step": 15245 }, { "epoch": 6.01, "learning_rate": 1.4505363337620573e-07, "loss": 0.0117, "step": 15246 }, { "epoch": 6.01, "learning_rate": 1.4502908321480285e-07, "loss": 0.0102, "step": 15247 }, { "epoch": 6.01, "learning_rate": 1.4500453394934242e-07, "loss": 0.0136, "step": 15248 }, { "epoch": 6.01, "learning_rate": 1.449799855802245e-07, "loss": 0.0096, "step": 15249 }, { "epoch": 6.01, "learning_rate": 1.449554381078492e-07, "loss": 0.0096, "step": 15250 }, { "epoch": 6.01, "learning_rate": 1.4493089153261653e-07, "loss": 0.0121, "step": 15251 }, { "epoch": 6.01, "learning_rate": 1.4490634585492668e-07, "loss": 0.0116, "step": 15252 }, { "epoch": 6.01, "learning_rate": 1.4488180107517969e-07, "loss": 0.0107, "step": 15253 }, { "epoch": 6.01, "learning_rate": 1.4485725719377556e-07, "loss": 0.0117, "step": 15254 }, { "epoch": 6.01, "learning_rate": 1.448327142111143e-07, "loss": 0.0108, "step": 15255 }, { "epoch": 6.01, "learning_rate": 1.448081721275959e-07, "loss": 0.0125, "step": 15256 }, { "epoch": 6.01, "learning_rate": 1.4478363094362053e-07, "loss": 0.0125, "step": 15257 }, { "epoch": 6.01, "learning_rate": 1.44759090659588e-07, "loss": 0.0082, "step": 15258 }, { "epoch": 6.01, "learning_rate": 1.4473455127589834e-07, "loss": 0.012, "step": 15259 }, { "epoch": 6.01, "learning_rate": 1.4471001279295148e-07, "loss": 0.0115, "step": 15260 }, { "epoch": 6.01, "learning_rate": 1.446854752111474e-07, "loss": 0.0118, "step": 15261 }, { "epoch": 6.01, "learning_rate": 1.4466093853088604e-07, "loss": 0.011, "step": 15262 }, { "epoch": 6.02, "learning_rate": 1.4463640275256723e-07, "loss": 0.0103, "step": 15263 }, { "epoch": 6.02, "learning_rate": 1.4461186787659097e-07, "loss": 0.0125, "step": 15264 }, { "epoch": 6.02, "learning_rate": 1.4458733390335702e-07, "loss": 0.015, "step": 15265 }, { "epoch": 6.02, "learning_rate": 1.4456280083326534e-07, "loss": 0.0106, "step": 15266 }, { "epoch": 6.02, "learning_rate": 1.4453826866671577e-07, "loss": 0.0139, "step": 15267 }, { "epoch": 6.02, "learning_rate": 1.4451373740410806e-07, "loss": 0.0108, "step": 15268 }, { "epoch": 6.02, "learning_rate": 1.444892070458421e-07, "loss": 0.0124, "step": 15269 }, { "epoch": 6.02, "learning_rate": 1.4446467759231776e-07, "loss": 0.0132, "step": 15270 }, { "epoch": 6.02, "learning_rate": 1.4444014904393478e-07, "loss": 0.0142, "step": 15271 }, { "epoch": 6.02, "learning_rate": 1.444156214010929e-07, "loss": 0.0119, "step": 15272 }, { "epoch": 6.02, "learning_rate": 1.4439109466419187e-07, "loss": 0.0139, "step": 15273 }, { "epoch": 6.02, "learning_rate": 1.4436656883363152e-07, "loss": 0.0122, "step": 15274 }, { "epoch": 6.02, "learning_rate": 1.4434204390981157e-07, "loss": 0.0107, "step": 15275 }, { "epoch": 6.02, "learning_rate": 1.4431751989313172e-07, "loss": 0.013, "step": 15276 }, { "epoch": 6.02, "learning_rate": 1.4429299678399164e-07, "loss": 0.0142, "step": 15277 }, { "epoch": 6.02, "learning_rate": 1.44268474582791e-07, "loss": 0.0126, "step": 15278 }, { "epoch": 6.02, "learning_rate": 1.4424395328992958e-07, "loss": 0.0135, "step": 15279 }, { "epoch": 6.02, "learning_rate": 1.4421943290580698e-07, "loss": 0.0106, "step": 15280 }, { "epoch": 6.02, "learning_rate": 1.4419491343082285e-07, "loss": 0.0114, "step": 15281 }, { "epoch": 6.02, "learning_rate": 1.441703948653768e-07, "loss": 0.0115, "step": 15282 }, { "epoch": 6.02, "learning_rate": 1.4414587720986843e-07, "loss": 0.0116, "step": 15283 }, { "epoch": 6.02, "learning_rate": 1.4412136046469745e-07, "loss": 0.0115, "step": 15284 }, { "epoch": 6.02, "learning_rate": 1.4409684463026333e-07, "loss": 0.0162, "step": 15285 }, { "epoch": 6.02, "learning_rate": 1.440723297069657e-07, "loss": 0.0098, "step": 15286 }, { "epoch": 6.02, "learning_rate": 1.4404781569520405e-07, "loss": 0.0107, "step": 15287 }, { "epoch": 6.03, "learning_rate": 1.44023302595378e-07, "loss": 0.0128, "step": 15288 }, { "epoch": 6.03, "learning_rate": 1.4399879040788707e-07, "loss": 0.0133, "step": 15289 }, { "epoch": 6.03, "learning_rate": 1.4397427913313073e-07, "loss": 0.0115, "step": 15290 }, { "epoch": 6.03, "learning_rate": 1.439497687715084e-07, "loss": 0.0108, "step": 15291 }, { "epoch": 6.03, "learning_rate": 1.4392525932341976e-07, "loss": 0.0108, "step": 15292 }, { "epoch": 6.03, "learning_rate": 1.4390075078926414e-07, "loss": 0.0123, "step": 15293 }, { "epoch": 6.03, "learning_rate": 1.43876243169441e-07, "loss": 0.015, "step": 15294 }, { "epoch": 6.03, "learning_rate": 1.4385173646434985e-07, "loss": 0.0114, "step": 15295 }, { "epoch": 6.03, "learning_rate": 1.4382723067438995e-07, "loss": 0.0134, "step": 15296 }, { "epoch": 6.03, "learning_rate": 1.438027257999609e-07, "loss": 0.0127, "step": 15297 }, { "epoch": 6.03, "learning_rate": 1.4377822184146202e-07, "loss": 0.0138, "step": 15298 }, { "epoch": 6.03, "learning_rate": 1.4375371879929268e-07, "loss": 0.0126, "step": 15299 }, { "epoch": 6.03, "learning_rate": 1.4372921667385213e-07, "loss": 0.0122, "step": 15300 }, { "epoch": 6.03, "learning_rate": 1.4370471546553992e-07, "loss": 0.011, "step": 15301 }, { "epoch": 6.03, "learning_rate": 1.436802151747553e-07, "loss": 0.0124, "step": 15302 }, { "epoch": 6.03, "learning_rate": 1.4365571580189758e-07, "loss": 0.0121, "step": 15303 }, { "epoch": 6.03, "learning_rate": 1.4363121734736606e-07, "loss": 0.0146, "step": 15304 }, { "epoch": 6.03, "learning_rate": 1.4360671981155993e-07, "loss": 0.0122, "step": 15305 }, { "epoch": 6.03, "learning_rate": 1.4358222319487865e-07, "loss": 0.0172, "step": 15306 }, { "epoch": 6.03, "learning_rate": 1.4355772749772136e-07, "loss": 0.013, "step": 15307 }, { "epoch": 6.03, "learning_rate": 1.4353323272048735e-07, "loss": 0.0114, "step": 15308 }, { "epoch": 6.03, "learning_rate": 1.4350873886357579e-07, "loss": 0.0114, "step": 15309 }, { "epoch": 6.03, "learning_rate": 1.43484245927386e-07, "loss": 0.0119, "step": 15310 }, { "epoch": 6.03, "learning_rate": 1.4345975391231705e-07, "loss": 0.0093, "step": 15311 }, { "epoch": 6.03, "learning_rate": 1.4343526281876824e-07, "loss": 0.0111, "step": 15312 }, { "epoch": 6.03, "learning_rate": 1.4341077264713855e-07, "loss": 0.0101, "step": 15313 }, { "epoch": 6.04, "learning_rate": 1.4338628339782735e-07, "loss": 0.0104, "step": 15314 }, { "epoch": 6.04, "learning_rate": 1.433617950712337e-07, "loss": 0.0135, "step": 15315 }, { "epoch": 6.04, "learning_rate": 1.433373076677567e-07, "loss": 0.0084, "step": 15316 }, { "epoch": 6.04, "learning_rate": 1.4331282118779546e-07, "loss": 0.0099, "step": 15317 }, { "epoch": 6.04, "learning_rate": 1.43288335631749e-07, "loss": 0.0119, "step": 15318 }, { "epoch": 6.04, "learning_rate": 1.432638510000166e-07, "loss": 0.012, "step": 15319 }, { "epoch": 6.04, "learning_rate": 1.4323936729299715e-07, "loss": 0.0126, "step": 15320 }, { "epoch": 6.04, "learning_rate": 1.4321488451108979e-07, "loss": 0.0095, "step": 15321 }, { "epoch": 6.04, "learning_rate": 1.4319040265469337e-07, "loss": 0.0136, "step": 15322 }, { "epoch": 6.04, "learning_rate": 1.4316592172420717e-07, "loss": 0.0149, "step": 15323 }, { "epoch": 6.04, "learning_rate": 1.4314144172003005e-07, "loss": 0.0101, "step": 15324 }, { "epoch": 6.04, "learning_rate": 1.43116962642561e-07, "loss": 0.012, "step": 15325 }, { "epoch": 6.04, "learning_rate": 1.4309248449219907e-07, "loss": 0.0117, "step": 15326 }, { "epoch": 6.04, "learning_rate": 1.4306800726934305e-07, "loss": 0.0116, "step": 15327 }, { "epoch": 6.04, "learning_rate": 1.43043530974392e-07, "loss": 0.0141, "step": 15328 }, { "epoch": 6.04, "learning_rate": 1.430190556077449e-07, "loss": 0.014, "step": 15329 }, { "epoch": 6.04, "learning_rate": 1.4299458116980057e-07, "loss": 0.0113, "step": 15330 }, { "epoch": 6.04, "learning_rate": 1.4297010766095787e-07, "loss": 0.0132, "step": 15331 }, { "epoch": 6.04, "learning_rate": 1.4294563508161582e-07, "loss": 0.0111, "step": 15332 }, { "epoch": 6.04, "learning_rate": 1.4292116343217317e-07, "loss": 0.015, "step": 15333 }, { "epoch": 6.04, "learning_rate": 1.4289669271302885e-07, "loss": 0.0115, "step": 15334 }, { "epoch": 6.04, "learning_rate": 1.4287222292458161e-07, "loss": 0.0134, "step": 15335 }, { "epoch": 6.04, "learning_rate": 1.4284775406723024e-07, "loss": 0.0119, "step": 15336 }, { "epoch": 6.04, "learning_rate": 1.428232861413737e-07, "loss": 0.0124, "step": 15337 }, { "epoch": 6.04, "learning_rate": 1.427988191474107e-07, "loss": 0.012, "step": 15338 }, { "epoch": 6.05, "learning_rate": 1.4277435308574003e-07, "loss": 0.0119, "step": 15339 }, { "epoch": 6.05, "learning_rate": 1.427498879567603e-07, "loss": 0.0126, "step": 15340 }, { "epoch": 6.05, "learning_rate": 1.427254237608705e-07, "loss": 0.0096, "step": 15341 }, { "epoch": 6.05, "learning_rate": 1.4270096049846924e-07, "loss": 0.0105, "step": 15342 }, { "epoch": 6.05, "learning_rate": 1.4267649816995526e-07, "loss": 0.0084, "step": 15343 }, { "epoch": 6.05, "learning_rate": 1.4265203677572722e-07, "loss": 0.0104, "step": 15344 }, { "epoch": 6.05, "learning_rate": 1.4262757631618374e-07, "loss": 0.0112, "step": 15345 }, { "epoch": 6.05, "learning_rate": 1.4260311679172365e-07, "loss": 0.0088, "step": 15346 }, { "epoch": 6.05, "learning_rate": 1.4257865820274554e-07, "loss": 0.0129, "step": 15347 }, { "epoch": 6.05, "learning_rate": 1.4255420054964802e-07, "loss": 0.0121, "step": 15348 }, { "epoch": 6.05, "learning_rate": 1.4252974383282967e-07, "loss": 0.0106, "step": 15349 }, { "epoch": 6.05, "learning_rate": 1.425052880526892e-07, "loss": 0.0123, "step": 15350 }, { "epoch": 6.05, "learning_rate": 1.4248083320962514e-07, "loss": 0.0129, "step": 15351 }, { "epoch": 6.05, "learning_rate": 1.4245637930403608e-07, "loss": 0.0121, "step": 15352 }, { "epoch": 6.05, "learning_rate": 1.4243192633632056e-07, "loss": 0.014, "step": 15353 }, { "epoch": 6.05, "learning_rate": 1.424074743068772e-07, "loss": 0.0118, "step": 15354 }, { "epoch": 6.05, "learning_rate": 1.4238302321610443e-07, "loss": 0.0089, "step": 15355 }, { "epoch": 6.05, "learning_rate": 1.4235857306440084e-07, "loss": 0.0126, "step": 15356 }, { "epoch": 6.05, "learning_rate": 1.4233412385216491e-07, "loss": 0.0135, "step": 15357 }, { "epoch": 6.05, "learning_rate": 1.4230967557979505e-07, "loss": 0.0114, "step": 15358 }, { "epoch": 6.05, "learning_rate": 1.4228522824768986e-07, "loss": 0.0122, "step": 15359 }, { "epoch": 6.05, "learning_rate": 1.4226078185624774e-07, "loss": 0.0115, "step": 15360 }, { "epoch": 6.05, "learning_rate": 1.4223633640586714e-07, "loss": 0.0131, "step": 15361 }, { "epoch": 6.05, "learning_rate": 1.4221189189694635e-07, "loss": 0.0136, "step": 15362 }, { "epoch": 6.05, "learning_rate": 1.42187448329884e-07, "loss": 0.0138, "step": 15363 }, { "epoch": 6.06, "learning_rate": 1.4216300570507837e-07, "loss": 0.0124, "step": 15364 }, { "epoch": 6.06, "learning_rate": 1.4213856402292786e-07, "loss": 0.01, "step": 15365 }, { "epoch": 6.06, "learning_rate": 1.421141232838308e-07, "loss": 0.0106, "step": 15366 }, { "epoch": 6.06, "learning_rate": 1.4208968348818547e-07, "loss": 0.011, "step": 15367 }, { "epoch": 6.06, "learning_rate": 1.4206524463639037e-07, "loss": 0.0156, "step": 15368 }, { "epoch": 6.06, "learning_rate": 1.4204080672884372e-07, "loss": 0.0144, "step": 15369 }, { "epoch": 6.06, "learning_rate": 1.4201636976594387e-07, "loss": 0.0112, "step": 15370 }, { "epoch": 6.06, "learning_rate": 1.41991933748089e-07, "loss": 0.0106, "step": 15371 }, { "epoch": 6.06, "learning_rate": 1.419674986756775e-07, "loss": 0.0113, "step": 15372 }, { "epoch": 6.06, "learning_rate": 1.4194306454910757e-07, "loss": 0.011, "step": 15373 }, { "epoch": 6.06, "learning_rate": 1.4191863136877737e-07, "loss": 0.0138, "step": 15374 }, { "epoch": 6.06, "learning_rate": 1.418941991350853e-07, "loss": 0.0105, "step": 15375 }, { "epoch": 6.06, "learning_rate": 1.418697678484294e-07, "loss": 0.0116, "step": 15376 }, { "epoch": 6.06, "learning_rate": 1.41845337509208e-07, "loss": 0.0107, "step": 15377 }, { "epoch": 6.06, "learning_rate": 1.418209081178192e-07, "loss": 0.0119, "step": 15378 }, { "epoch": 6.06, "learning_rate": 1.4179647967466119e-07, "loss": 0.0095, "step": 15379 }, { "epoch": 6.06, "learning_rate": 1.41772052180132e-07, "loss": 0.0119, "step": 15380 }, { "epoch": 6.06, "learning_rate": 1.4174762563462996e-07, "loss": 0.0083, "step": 15381 }, { "epoch": 6.06, "learning_rate": 1.417232000385531e-07, "loss": 0.0097, "step": 15382 }, { "epoch": 6.06, "learning_rate": 1.4169877539229948e-07, "loss": 0.0114, "step": 15383 }, { "epoch": 6.06, "learning_rate": 1.4167435169626725e-07, "loss": 0.0111, "step": 15384 }, { "epoch": 6.06, "learning_rate": 1.4164992895085434e-07, "loss": 0.0139, "step": 15385 }, { "epoch": 6.06, "learning_rate": 1.41625507156459e-07, "loss": 0.0122, "step": 15386 }, { "epoch": 6.06, "learning_rate": 1.416010863134791e-07, "loss": 0.0113, "step": 15387 }, { "epoch": 6.06, "learning_rate": 1.4157666642231282e-07, "loss": 0.0106, "step": 15388 }, { "epoch": 6.06, "learning_rate": 1.4155224748335797e-07, "loss": 0.0112, "step": 15389 }, { "epoch": 6.07, "learning_rate": 1.4152782949701276e-07, "loss": 0.0094, "step": 15390 }, { "epoch": 6.07, "learning_rate": 1.4150341246367505e-07, "loss": 0.011, "step": 15391 }, { "epoch": 6.07, "learning_rate": 1.414789963837428e-07, "loss": 0.0124, "step": 15392 }, { "epoch": 6.07, "learning_rate": 1.4145458125761401e-07, "loss": 0.0151, "step": 15393 }, { "epoch": 6.07, "learning_rate": 1.4143016708568648e-07, "loss": 0.0125, "step": 15394 }, { "epoch": 6.07, "learning_rate": 1.4140575386835827e-07, "loss": 0.0108, "step": 15395 }, { "epoch": 6.07, "learning_rate": 1.4138134160602717e-07, "loss": 0.0161, "step": 15396 }, { "epoch": 6.07, "learning_rate": 1.4135693029909114e-07, "loss": 0.0149, "step": 15397 }, { "epoch": 6.07, "learning_rate": 1.41332519947948e-07, "loss": 0.01, "step": 15398 }, { "epoch": 6.07, "learning_rate": 1.4130811055299566e-07, "loss": 0.0129, "step": 15399 }, { "epoch": 6.07, "learning_rate": 1.412837021146319e-07, "loss": 0.0139, "step": 15400 }, { "epoch": 6.07, "learning_rate": 1.4125929463325458e-07, "loss": 0.0123, "step": 15401 }, { "epoch": 6.07, "learning_rate": 1.412348881092614e-07, "loss": 0.014, "step": 15402 }, { "epoch": 6.07, "learning_rate": 1.4121048254305027e-07, "loss": 0.0117, "step": 15403 }, { "epoch": 6.07, "learning_rate": 1.4118607793501896e-07, "loss": 0.0145, "step": 15404 }, { "epoch": 6.07, "learning_rate": 1.411616742855652e-07, "loss": 0.0129, "step": 15405 }, { "epoch": 6.07, "learning_rate": 1.4113727159508673e-07, "loss": 0.0113, "step": 15406 }, { "epoch": 6.07, "learning_rate": 1.4111286986398115e-07, "loss": 0.0121, "step": 15407 }, { "epoch": 6.07, "learning_rate": 1.4108846909264643e-07, "loss": 0.0104, "step": 15408 }, { "epoch": 6.07, "learning_rate": 1.410640692814801e-07, "loss": 0.0136, "step": 15409 }, { "epoch": 6.07, "learning_rate": 1.4103967043087986e-07, "loss": 0.0131, "step": 15410 }, { "epoch": 6.07, "learning_rate": 1.410152725412433e-07, "loss": 0.0108, "step": 15411 }, { "epoch": 6.07, "learning_rate": 1.4099087561296825e-07, "loss": 0.0122, "step": 15412 }, { "epoch": 6.07, "learning_rate": 1.4096647964645222e-07, "loss": 0.0115, "step": 15413 }, { "epoch": 6.07, "learning_rate": 1.4094208464209288e-07, "loss": 0.0143, "step": 15414 }, { "epoch": 6.08, "learning_rate": 1.4091769060028778e-07, "loss": 0.0121, "step": 15415 }, { "epoch": 6.08, "learning_rate": 1.408932975214345e-07, "loss": 0.0141, "step": 15416 }, { "epoch": 6.08, "learning_rate": 1.4086890540593066e-07, "loss": 0.0094, "step": 15417 }, { "epoch": 6.08, "learning_rate": 1.4084451425417383e-07, "loss": 0.0113, "step": 15418 }, { "epoch": 6.08, "learning_rate": 1.4082012406656145e-07, "loss": 0.0125, "step": 15419 }, { "epoch": 6.08, "learning_rate": 1.4079573484349108e-07, "loss": 0.0105, "step": 15420 }, { "epoch": 6.08, "learning_rate": 1.4077134658536035e-07, "loss": 0.0106, "step": 15421 }, { "epoch": 6.08, "learning_rate": 1.407469592925666e-07, "loss": 0.0132, "step": 15422 }, { "epoch": 6.08, "learning_rate": 1.4072257296550738e-07, "loss": 0.0116, "step": 15423 }, { "epoch": 6.08, "learning_rate": 1.406981876045801e-07, "loss": 0.0138, "step": 15424 }, { "epoch": 6.08, "learning_rate": 1.406738032101822e-07, "loss": 0.0121, "step": 15425 }, { "epoch": 6.08, "learning_rate": 1.406494197827112e-07, "loss": 0.014, "step": 15426 }, { "epoch": 6.08, "learning_rate": 1.4062503732256447e-07, "loss": 0.0112, "step": 15427 }, { "epoch": 6.08, "learning_rate": 1.406006558301394e-07, "loss": 0.0119, "step": 15428 }, { "epoch": 6.08, "learning_rate": 1.4057627530583326e-07, "loss": 0.0117, "step": 15429 }, { "epoch": 6.08, "learning_rate": 1.4055189575004361e-07, "loss": 0.0128, "step": 15430 }, { "epoch": 6.08, "learning_rate": 1.4052751716316768e-07, "loss": 0.0104, "step": 15431 }, { "epoch": 6.08, "learning_rate": 1.4050313954560288e-07, "loss": 0.0126, "step": 15432 }, { "epoch": 6.08, "learning_rate": 1.4047876289774646e-07, "loss": 0.0134, "step": 15433 }, { "epoch": 6.08, "learning_rate": 1.4045438721999563e-07, "loss": 0.0138, "step": 15434 }, { "epoch": 6.08, "learning_rate": 1.404300125127479e-07, "loss": 0.0115, "step": 15435 }, { "epoch": 6.08, "learning_rate": 1.4040563877640044e-07, "loss": 0.0095, "step": 15436 }, { "epoch": 6.08, "learning_rate": 1.4038126601135048e-07, "loss": 0.0103, "step": 15437 }, { "epoch": 6.08, "learning_rate": 1.403568942179952e-07, "loss": 0.0138, "step": 15438 }, { "epoch": 6.08, "learning_rate": 1.40332523396732e-07, "loss": 0.0115, "step": 15439 }, { "epoch": 6.09, "learning_rate": 1.4030815354795796e-07, "loss": 0.0092, "step": 15440 }, { "epoch": 6.09, "learning_rate": 1.4028378467207032e-07, "loss": 0.0147, "step": 15441 }, { "epoch": 6.09, "learning_rate": 1.4025941676946608e-07, "loss": 0.0114, "step": 15442 }, { "epoch": 6.09, "learning_rate": 1.4023504984054272e-07, "loss": 0.0136, "step": 15443 }, { "epoch": 6.09, "learning_rate": 1.4021068388569717e-07, "loss": 0.0123, "step": 15444 }, { "epoch": 6.09, "learning_rate": 1.401863189053266e-07, "loss": 0.0132, "step": 15445 }, { "epoch": 6.09, "learning_rate": 1.4016195489982811e-07, "loss": 0.0119, "step": 15446 }, { "epoch": 6.09, "learning_rate": 1.4013759186959876e-07, "loss": 0.0138, "step": 15447 }, { "epoch": 6.09, "learning_rate": 1.4011322981503576e-07, "loss": 0.0112, "step": 15448 }, { "epoch": 6.09, "learning_rate": 1.400888687365361e-07, "loss": 0.0114, "step": 15449 }, { "epoch": 6.09, "learning_rate": 1.400645086344968e-07, "loss": 0.0134, "step": 15450 }, { "epoch": 6.09, "learning_rate": 1.4004014950931486e-07, "loss": 0.0131, "step": 15451 }, { "epoch": 6.09, "learning_rate": 1.4001579136138744e-07, "loss": 0.0115, "step": 15452 }, { "epoch": 6.09, "learning_rate": 1.3999143419111142e-07, "loss": 0.0106, "step": 15453 }, { "epoch": 6.09, "learning_rate": 1.3996707799888387e-07, "loss": 0.0133, "step": 15454 }, { "epoch": 6.09, "learning_rate": 1.3994272278510168e-07, "loss": 0.0146, "step": 15455 }, { "epoch": 6.09, "learning_rate": 1.3991836855016175e-07, "loss": 0.0125, "step": 15456 }, { "epoch": 6.09, "learning_rate": 1.398940152944612e-07, "loss": 0.0143, "step": 15457 }, { "epoch": 6.09, "learning_rate": 1.3986966301839683e-07, "loss": 0.0132, "step": 15458 }, { "epoch": 6.09, "learning_rate": 1.398453117223656e-07, "loss": 0.0144, "step": 15459 }, { "epoch": 6.09, "learning_rate": 1.3982096140676426e-07, "loss": 0.013, "step": 15460 }, { "epoch": 6.09, "learning_rate": 1.3979661207198987e-07, "loss": 0.0117, "step": 15461 }, { "epoch": 6.09, "learning_rate": 1.397722637184392e-07, "loss": 0.0106, "step": 15462 }, { "epoch": 6.09, "learning_rate": 1.3974791634650908e-07, "loss": 0.0106, "step": 15463 }, { "epoch": 6.09, "learning_rate": 1.3972356995659633e-07, "loss": 0.0136, "step": 15464 }, { "epoch": 6.09, "learning_rate": 1.3969922454909777e-07, "loss": 0.0144, "step": 15465 }, { "epoch": 6.1, "learning_rate": 1.3967488012441027e-07, "loss": 0.0106, "step": 15466 }, { "epoch": 6.1, "learning_rate": 1.396505366829305e-07, "loss": 0.0146, "step": 15467 }, { "epoch": 6.1, "learning_rate": 1.3962619422505528e-07, "loss": 0.0134, "step": 15468 }, { "epoch": 6.1, "learning_rate": 1.3960185275118127e-07, "loss": 0.0126, "step": 15469 }, { "epoch": 6.1, "learning_rate": 1.3957751226170535e-07, "loss": 0.0118, "step": 15470 }, { "epoch": 6.1, "learning_rate": 1.395531727570241e-07, "loss": 0.0109, "step": 15471 }, { "epoch": 6.1, "learning_rate": 1.395288342375343e-07, "loss": 0.0116, "step": 15472 }, { "epoch": 6.1, "learning_rate": 1.395044967036326e-07, "loss": 0.0129, "step": 15473 }, { "epoch": 6.1, "learning_rate": 1.3948016015571557e-07, "loss": 0.0117, "step": 15474 }, { "epoch": 6.1, "learning_rate": 1.3945582459418e-07, "loss": 0.0123, "step": 15475 }, { "epoch": 6.1, "learning_rate": 1.3943149001942255e-07, "loss": 0.0132, "step": 15476 }, { "epoch": 6.1, "learning_rate": 1.3940715643183972e-07, "loss": 0.0115, "step": 15477 }, { "epoch": 6.1, "learning_rate": 1.3938282383182806e-07, "loss": 0.0109, "step": 15478 }, { "epoch": 6.1, "learning_rate": 1.393584922197843e-07, "loss": 0.0089, "step": 15479 }, { "epoch": 6.1, "learning_rate": 1.39334161596105e-07, "loss": 0.0125, "step": 15480 }, { "epoch": 6.1, "learning_rate": 1.3930983196118664e-07, "loss": 0.0158, "step": 15481 }, { "epoch": 6.1, "learning_rate": 1.3928550331542574e-07, "loss": 0.0114, "step": 15482 }, { "epoch": 6.1, "learning_rate": 1.392611756592189e-07, "loss": 0.0115, "step": 15483 }, { "epoch": 6.1, "learning_rate": 1.3923684899296259e-07, "loss": 0.009, "step": 15484 }, { "epoch": 6.1, "learning_rate": 1.392125233170533e-07, "loss": 0.0116, "step": 15485 }, { "epoch": 6.1, "learning_rate": 1.391881986318875e-07, "loss": 0.0134, "step": 15486 }, { "epoch": 6.1, "learning_rate": 1.3916387493786152e-07, "loss": 0.0132, "step": 15487 }, { "epoch": 6.1, "learning_rate": 1.3913955223537204e-07, "loss": 0.0102, "step": 15488 }, { "epoch": 6.1, "learning_rate": 1.3911523052481534e-07, "loss": 0.0106, "step": 15489 }, { "epoch": 6.1, "learning_rate": 1.3909090980658786e-07, "loss": 0.014, "step": 15490 }, { "epoch": 6.11, "learning_rate": 1.390665900810859e-07, "loss": 0.0098, "step": 15491 }, { "epoch": 6.11, "learning_rate": 1.3904227134870595e-07, "loss": 0.0129, "step": 15492 }, { "epoch": 6.11, "learning_rate": 1.390179536098444e-07, "loss": 0.0121, "step": 15493 }, { "epoch": 6.11, "learning_rate": 1.389936368648975e-07, "loss": 0.0133, "step": 15494 }, { "epoch": 6.11, "learning_rate": 1.3896932111426157e-07, "loss": 0.0118, "step": 15495 }, { "epoch": 6.11, "learning_rate": 1.389450063583329e-07, "loss": 0.012, "step": 15496 }, { "epoch": 6.11, "learning_rate": 1.389206925975079e-07, "loss": 0.0106, "step": 15497 }, { "epoch": 6.11, "learning_rate": 1.388963798321828e-07, "loss": 0.0096, "step": 15498 }, { "epoch": 6.11, "learning_rate": 1.388720680627538e-07, "loss": 0.0126, "step": 15499 }, { "epoch": 6.11, "learning_rate": 1.3884775728961718e-07, "loss": 0.0094, "step": 15500 }, { "epoch": 6.11, "learning_rate": 1.388234475131692e-07, "loss": 0.0109, "step": 15501 }, { "epoch": 6.11, "learning_rate": 1.3879913873380603e-07, "loss": 0.0115, "step": 15502 }, { "epoch": 6.11, "learning_rate": 1.3877483095192394e-07, "loss": 0.0143, "step": 15503 }, { "epoch": 6.11, "learning_rate": 1.38750524167919e-07, "loss": 0.0104, "step": 15504 }, { "epoch": 6.11, "learning_rate": 1.3872621838218742e-07, "loss": 0.0142, "step": 15505 }, { "epoch": 6.11, "learning_rate": 1.3870191359512537e-07, "loss": 0.0097, "step": 15506 }, { "epoch": 6.11, "learning_rate": 1.38677609807129e-07, "loss": 0.0112, "step": 15507 }, { "epoch": 6.11, "learning_rate": 1.3865330701859438e-07, "loss": 0.0118, "step": 15508 }, { "epoch": 6.11, "learning_rate": 1.3862900522991755e-07, "loss": 0.0137, "step": 15509 }, { "epoch": 6.11, "learning_rate": 1.3860470444149472e-07, "loss": 0.0122, "step": 15510 }, { "epoch": 6.11, "learning_rate": 1.385804046537219e-07, "loss": 0.0135, "step": 15511 }, { "epoch": 6.11, "learning_rate": 1.3855610586699513e-07, "loss": 0.0113, "step": 15512 }, { "epoch": 6.11, "learning_rate": 1.3853180808171047e-07, "loss": 0.0113, "step": 15513 }, { "epoch": 6.11, "learning_rate": 1.3850751129826384e-07, "loss": 0.0113, "step": 15514 }, { "epoch": 6.11, "learning_rate": 1.3848321551705135e-07, "loss": 0.0124, "step": 15515 }, { "epoch": 6.11, "learning_rate": 1.38458920738469e-07, "loss": 0.0089, "step": 15516 }, { "epoch": 6.12, "learning_rate": 1.3843462696291272e-07, "loss": 0.0103, "step": 15517 }, { "epoch": 6.12, "learning_rate": 1.3841033419077834e-07, "loss": 0.0102, "step": 15518 }, { "epoch": 6.12, "learning_rate": 1.3838604242246202e-07, "loss": 0.0122, "step": 15519 }, { "epoch": 6.12, "learning_rate": 1.3836175165835954e-07, "loss": 0.0107, "step": 15520 }, { "epoch": 6.12, "learning_rate": 1.3833746189886682e-07, "loss": 0.0116, "step": 15521 }, { "epoch": 6.12, "learning_rate": 1.3831317314437973e-07, "loss": 0.0116, "step": 15522 }, { "epoch": 6.12, "learning_rate": 1.382888853952942e-07, "loss": 0.0108, "step": 15523 }, { "epoch": 6.12, "learning_rate": 1.3826459865200602e-07, "loss": 0.0087, "step": 15524 }, { "epoch": 6.12, "learning_rate": 1.3824031291491106e-07, "loss": 0.0109, "step": 15525 }, { "epoch": 6.12, "learning_rate": 1.3821602818440518e-07, "loss": 0.0098, "step": 15526 }, { "epoch": 6.12, "learning_rate": 1.381917444608841e-07, "loss": 0.0162, "step": 15527 }, { "epoch": 6.12, "learning_rate": 1.3816746174474366e-07, "loss": 0.0104, "step": 15528 }, { "epoch": 6.12, "learning_rate": 1.381431800363796e-07, "loss": 0.0116, "step": 15529 }, { "epoch": 6.12, "learning_rate": 1.3811889933618775e-07, "loss": 0.0109, "step": 15530 }, { "epoch": 6.12, "learning_rate": 1.3809461964456369e-07, "loss": 0.0109, "step": 15531 }, { "epoch": 6.12, "learning_rate": 1.3807034096190327e-07, "loss": 0.0121, "step": 15532 }, { "epoch": 6.12, "learning_rate": 1.3804606328860224e-07, "loss": 0.011, "step": 15533 }, { "epoch": 6.12, "learning_rate": 1.3802178662505622e-07, "loss": 0.0103, "step": 15534 }, { "epoch": 6.12, "learning_rate": 1.3799751097166086e-07, "loss": 0.0133, "step": 15535 }, { "epoch": 6.12, "learning_rate": 1.3797323632881175e-07, "loss": 0.0108, "step": 15536 }, { "epoch": 6.12, "learning_rate": 1.379489626969047e-07, "loss": 0.0119, "step": 15537 }, { "epoch": 6.12, "learning_rate": 1.3792469007633524e-07, "loss": 0.0113, "step": 15538 }, { "epoch": 6.12, "learning_rate": 1.3790041846749902e-07, "loss": 0.0167, "step": 15539 }, { "epoch": 6.12, "learning_rate": 1.378761478707915e-07, "loss": 0.0128, "step": 15540 }, { "epoch": 6.12, "learning_rate": 1.3785187828660844e-07, "loss": 0.011, "step": 15541 }, { "epoch": 6.13, "learning_rate": 1.378276097153453e-07, "loss": 0.0147, "step": 15542 }, { "epoch": 6.13, "learning_rate": 1.3780334215739765e-07, "loss": 0.0109, "step": 15543 }, { "epoch": 6.13, "learning_rate": 1.3777907561316097e-07, "loss": 0.0105, "step": 15544 }, { "epoch": 6.13, "learning_rate": 1.3775481008303075e-07, "loss": 0.0112, "step": 15545 }, { "epoch": 6.13, "learning_rate": 1.377305455674026e-07, "loss": 0.0126, "step": 15546 }, { "epoch": 6.13, "learning_rate": 1.3770628206667186e-07, "loss": 0.0131, "step": 15547 }, { "epoch": 6.13, "learning_rate": 1.3768201958123408e-07, "loss": 0.0109, "step": 15548 }, { "epoch": 6.13, "learning_rate": 1.3765775811148465e-07, "loss": 0.0113, "step": 15549 }, { "epoch": 6.13, "learning_rate": 1.3763349765781902e-07, "loss": 0.0123, "step": 15550 }, { "epoch": 6.13, "learning_rate": 1.376092382206326e-07, "loss": 0.0102, "step": 15551 }, { "epoch": 6.13, "learning_rate": 1.375849798003208e-07, "loss": 0.0129, "step": 15552 }, { "epoch": 6.13, "learning_rate": 1.3756072239727893e-07, "loss": 0.0112, "step": 15553 }, { "epoch": 6.13, "learning_rate": 1.375364660119023e-07, "loss": 0.0122, "step": 15554 }, { "epoch": 6.13, "learning_rate": 1.3751221064458648e-07, "loss": 0.0099, "step": 15555 }, { "epoch": 6.13, "learning_rate": 1.3748795629572662e-07, "loss": 0.0132, "step": 15556 }, { "epoch": 6.13, "learning_rate": 1.3746370296571806e-07, "loss": 0.0115, "step": 15557 }, { "epoch": 6.13, "learning_rate": 1.3743945065495606e-07, "loss": 0.0125, "step": 15558 }, { "epoch": 6.13, "learning_rate": 1.3741519936383596e-07, "loss": 0.0118, "step": 15559 }, { "epoch": 6.13, "learning_rate": 1.37390949092753e-07, "loss": 0.0109, "step": 15560 }, { "epoch": 6.13, "learning_rate": 1.3736669984210246e-07, "loss": 0.012, "step": 15561 }, { "epoch": 6.13, "learning_rate": 1.3734245161227936e-07, "loss": 0.0127, "step": 15562 }, { "epoch": 6.13, "learning_rate": 1.3731820440367924e-07, "loss": 0.0135, "step": 15563 }, { "epoch": 6.13, "learning_rate": 1.3729395821669707e-07, "loss": 0.0115, "step": 15564 }, { "epoch": 6.13, "learning_rate": 1.372697130517281e-07, "loss": 0.0111, "step": 15565 }, { "epoch": 6.13, "learning_rate": 1.3724546890916746e-07, "loss": 0.0128, "step": 15566 }, { "epoch": 6.14, "learning_rate": 1.3722122578941028e-07, "loss": 0.0121, "step": 15567 }, { "epoch": 6.14, "learning_rate": 1.3719698369285172e-07, "loss": 0.0122, "step": 15568 }, { "epoch": 6.14, "learning_rate": 1.3717274261988685e-07, "loss": 0.0138, "step": 15569 }, { "epoch": 6.14, "learning_rate": 1.371485025709108e-07, "loss": 0.0128, "step": 15570 }, { "epoch": 6.14, "learning_rate": 1.371242635463186e-07, "loss": 0.013, "step": 15571 }, { "epoch": 6.14, "learning_rate": 1.371000255465054e-07, "loss": 0.0125, "step": 15572 }, { "epoch": 6.14, "learning_rate": 1.3707578857186622e-07, "loss": 0.0109, "step": 15573 }, { "epoch": 6.14, "learning_rate": 1.37051552622796e-07, "loss": 0.0122, "step": 15574 }, { "epoch": 6.14, "learning_rate": 1.3702731769968983e-07, "loss": 0.014, "step": 15575 }, { "epoch": 6.14, "learning_rate": 1.3700308380294256e-07, "loss": 0.0128, "step": 15576 }, { "epoch": 6.14, "learning_rate": 1.3697885093294942e-07, "loss": 0.0127, "step": 15577 }, { "epoch": 6.14, "learning_rate": 1.3695461909010519e-07, "loss": 0.0118, "step": 15578 }, { "epoch": 6.14, "learning_rate": 1.3693038827480483e-07, "loss": 0.0128, "step": 15579 }, { "epoch": 6.14, "learning_rate": 1.3690615848744325e-07, "loss": 0.0145, "step": 15580 }, { "epoch": 6.14, "learning_rate": 1.3688192972841543e-07, "loss": 0.0118, "step": 15581 }, { "epoch": 6.14, "learning_rate": 1.3685770199811625e-07, "loss": 0.012, "step": 15582 }, { "epoch": 6.14, "learning_rate": 1.3683347529694058e-07, "loss": 0.0116, "step": 15583 }, { "epoch": 6.14, "learning_rate": 1.3680924962528326e-07, "loss": 0.0179, "step": 15584 }, { "epoch": 6.14, "learning_rate": 1.3678502498353903e-07, "loss": 0.0106, "step": 15585 }, { "epoch": 6.14, "learning_rate": 1.367608013721029e-07, "loss": 0.0154, "step": 15586 }, { "epoch": 6.14, "learning_rate": 1.367365787913696e-07, "loss": 0.0122, "step": 15587 }, { "epoch": 6.14, "learning_rate": 1.3671235724173392e-07, "loss": 0.0179, "step": 15588 }, { "epoch": 6.14, "learning_rate": 1.3668813672359058e-07, "loss": 0.0118, "step": 15589 }, { "epoch": 6.14, "learning_rate": 1.3666391723733445e-07, "loss": 0.0151, "step": 15590 }, { "epoch": 6.14, "learning_rate": 1.3663969878336018e-07, "loss": 0.0125, "step": 15591 }, { "epoch": 6.14, "learning_rate": 1.3661548136206252e-07, "loss": 0.0122, "step": 15592 }, { "epoch": 6.15, "learning_rate": 1.365912649738362e-07, "loss": 0.0118, "step": 15593 }, { "epoch": 6.15, "learning_rate": 1.3656704961907584e-07, "loss": 0.0112, "step": 15594 }, { "epoch": 6.15, "learning_rate": 1.365428352981762e-07, "loss": 0.0118, "step": 15595 }, { "epoch": 6.15, "learning_rate": 1.3651862201153192e-07, "loss": 0.011, "step": 15596 }, { "epoch": 6.15, "learning_rate": 1.3649440975953763e-07, "loss": 0.0125, "step": 15597 }, { "epoch": 6.15, "learning_rate": 1.3647019854258785e-07, "loss": 0.0103, "step": 15598 }, { "epoch": 6.15, "learning_rate": 1.3644598836107735e-07, "loss": 0.0136, "step": 15599 }, { "epoch": 6.15, "learning_rate": 1.3642177921540068e-07, "loss": 0.0117, "step": 15600 }, { "epoch": 6.15, "learning_rate": 1.3639757110595237e-07, "loss": 0.0125, "step": 15601 }, { "epoch": 6.15, "learning_rate": 1.3637336403312687e-07, "loss": 0.0114, "step": 15602 }, { "epoch": 6.15, "learning_rate": 1.3634915799731895e-07, "loss": 0.0108, "step": 15603 }, { "epoch": 6.15, "learning_rate": 1.3632495299892298e-07, "loss": 0.012, "step": 15604 }, { "epoch": 6.15, "learning_rate": 1.3630074903833352e-07, "loss": 0.0119, "step": 15605 }, { "epoch": 6.15, "learning_rate": 1.3627654611594506e-07, "loss": 0.0113, "step": 15606 }, { "epoch": 6.15, "learning_rate": 1.3625234423215198e-07, "loss": 0.0135, "step": 15607 }, { "epoch": 6.15, "learning_rate": 1.3622814338734884e-07, "loss": 0.012, "step": 15608 }, { "epoch": 6.15, "learning_rate": 1.3620394358193008e-07, "loss": 0.0115, "step": 15609 }, { "epoch": 6.15, "learning_rate": 1.3617974481629006e-07, "loss": 0.0099, "step": 15610 }, { "epoch": 6.15, "learning_rate": 1.3615554709082315e-07, "loss": 0.0106, "step": 15611 }, { "epoch": 6.15, "learning_rate": 1.3613135040592385e-07, "loss": 0.0134, "step": 15612 }, { "epoch": 6.15, "learning_rate": 1.3610715476198646e-07, "loss": 0.0109, "step": 15613 }, { "epoch": 6.15, "learning_rate": 1.3608296015940536e-07, "loss": 0.0139, "step": 15614 }, { "epoch": 6.15, "learning_rate": 1.360587665985748e-07, "loss": 0.0107, "step": 15615 }, { "epoch": 6.15, "learning_rate": 1.3603457407988919e-07, "loss": 0.0112, "step": 15616 }, { "epoch": 6.15, "learning_rate": 1.3601038260374282e-07, "loss": 0.0127, "step": 15617 }, { "epoch": 6.16, "learning_rate": 1.3598619217053e-07, "loss": 0.0111, "step": 15618 }, { "epoch": 6.16, "learning_rate": 1.359620027806449e-07, "loss": 0.0112, "step": 15619 }, { "epoch": 6.16, "learning_rate": 1.3593781443448181e-07, "loss": 0.0138, "step": 15620 }, { "epoch": 6.16, "learning_rate": 1.35913627132435e-07, "loss": 0.0143, "step": 15621 }, { "epoch": 6.16, "learning_rate": 1.3588944087489873e-07, "loss": 0.012, "step": 15622 }, { "epoch": 6.16, "learning_rate": 1.358652556622671e-07, "loss": 0.0118, "step": 15623 }, { "epoch": 6.16, "learning_rate": 1.3584107149493435e-07, "loss": 0.0129, "step": 15624 }, { "epoch": 6.16, "learning_rate": 1.358168883732945e-07, "loss": 0.0117, "step": 15625 }, { "epoch": 6.16, "learning_rate": 1.3579270629774192e-07, "loss": 0.0111, "step": 15626 }, { "epoch": 6.16, "learning_rate": 1.357685252686707e-07, "loss": 0.0114, "step": 15627 }, { "epoch": 6.16, "learning_rate": 1.3574434528647484e-07, "loss": 0.0119, "step": 15628 }, { "epoch": 6.16, "learning_rate": 1.3572016635154842e-07, "loss": 0.0134, "step": 15629 }, { "epoch": 6.16, "learning_rate": 1.356959884642857e-07, "loss": 0.0137, "step": 15630 }, { "epoch": 6.16, "learning_rate": 1.3567181162508062e-07, "loss": 0.012, "step": 15631 }, { "epoch": 6.16, "learning_rate": 1.3564763583432724e-07, "loss": 0.0135, "step": 15632 }, { "epoch": 6.16, "learning_rate": 1.3562346109241958e-07, "loss": 0.0123, "step": 15633 }, { "epoch": 6.16, "learning_rate": 1.3559928739975165e-07, "loss": 0.0116, "step": 15634 }, { "epoch": 6.16, "learning_rate": 1.355751147567175e-07, "loss": 0.0112, "step": 15635 }, { "epoch": 6.16, "learning_rate": 1.3555094316371103e-07, "loss": 0.0134, "step": 15636 }, { "epoch": 6.16, "learning_rate": 1.3552677262112626e-07, "loss": 0.0124, "step": 15637 }, { "epoch": 6.16, "learning_rate": 1.3550260312935703e-07, "loss": 0.0114, "step": 15638 }, { "epoch": 6.16, "learning_rate": 1.354784346887974e-07, "loss": 0.0118, "step": 15639 }, { "epoch": 6.16, "learning_rate": 1.3545426729984125e-07, "loss": 0.011, "step": 15640 }, { "epoch": 6.16, "learning_rate": 1.3543010096288242e-07, "loss": 0.0111, "step": 15641 }, { "epoch": 6.16, "learning_rate": 1.3540593567831475e-07, "loss": 0.0119, "step": 15642 }, { "epoch": 6.17, "learning_rate": 1.3538177144653222e-07, "loss": 0.0108, "step": 15643 }, { "epoch": 6.17, "learning_rate": 1.3535760826792864e-07, "loss": 0.0102, "step": 15644 }, { "epoch": 6.17, "learning_rate": 1.3533344614289776e-07, "loss": 0.0095, "step": 15645 }, { "epoch": 6.17, "learning_rate": 1.3530928507183346e-07, "loss": 0.0124, "step": 15646 }, { "epoch": 6.17, "learning_rate": 1.352851250551294e-07, "loss": 0.0163, "step": 15647 }, { "epoch": 6.17, "learning_rate": 1.3526096609317952e-07, "loss": 0.0122, "step": 15648 }, { "epoch": 6.17, "learning_rate": 1.352368081863775e-07, "loss": 0.0114, "step": 15649 }, { "epoch": 6.17, "learning_rate": 1.352126513351171e-07, "loss": 0.0126, "step": 15650 }, { "epoch": 6.17, "learning_rate": 1.3518849553979194e-07, "loss": 0.0112, "step": 15651 }, { "epoch": 6.17, "learning_rate": 1.3516434080079585e-07, "loss": 0.0103, "step": 15652 }, { "epoch": 6.17, "learning_rate": 1.3514018711852246e-07, "loss": 0.0125, "step": 15653 }, { "epoch": 6.17, "learning_rate": 1.3511603449336545e-07, "loss": 0.0129, "step": 15654 }, { "epoch": 6.17, "learning_rate": 1.3509188292571852e-07, "loss": 0.0112, "step": 15655 }, { "epoch": 6.17, "learning_rate": 1.3506773241597516e-07, "loss": 0.0136, "step": 15656 }, { "epoch": 6.17, "learning_rate": 1.3504358296452912e-07, "loss": 0.0122, "step": 15657 }, { "epoch": 6.17, "learning_rate": 1.3501943457177393e-07, "loss": 0.0126, "step": 15658 }, { "epoch": 6.17, "learning_rate": 1.3499528723810325e-07, "loss": 0.011, "step": 15659 }, { "epoch": 6.17, "learning_rate": 1.349711409639105e-07, "loss": 0.0136, "step": 15660 }, { "epoch": 6.17, "learning_rate": 1.3494699574958936e-07, "loss": 0.0109, "step": 15661 }, { "epoch": 6.17, "learning_rate": 1.3492285159553337e-07, "loss": 0.0126, "step": 15662 }, { "epoch": 6.17, "learning_rate": 1.3489870850213598e-07, "loss": 0.0118, "step": 15663 }, { "epoch": 6.17, "learning_rate": 1.3487456646979069e-07, "loss": 0.0111, "step": 15664 }, { "epoch": 6.17, "learning_rate": 1.348504254988909e-07, "loss": 0.0155, "step": 15665 }, { "epoch": 6.17, "learning_rate": 1.3482628558983025e-07, "loss": 0.0127, "step": 15666 }, { "epoch": 6.17, "learning_rate": 1.3480214674300207e-07, "loss": 0.0128, "step": 15667 }, { "epoch": 6.17, "learning_rate": 1.3477800895879984e-07, "loss": 0.0113, "step": 15668 }, { "epoch": 6.18, "learning_rate": 1.3475387223761683e-07, "loss": 0.0118, "step": 15669 }, { "epoch": 6.18, "learning_rate": 1.3472973657984666e-07, "loss": 0.0108, "step": 15670 }, { "epoch": 6.18, "learning_rate": 1.3470560198588256e-07, "loss": 0.0116, "step": 15671 }, { "epoch": 6.18, "learning_rate": 1.346814684561179e-07, "loss": 0.0118, "step": 15672 }, { "epoch": 6.18, "learning_rate": 1.3465733599094605e-07, "loss": 0.0106, "step": 15673 }, { "epoch": 6.18, "learning_rate": 1.3463320459076025e-07, "loss": 0.0156, "step": 15674 }, { "epoch": 6.18, "learning_rate": 1.346090742559539e-07, "loss": 0.0124, "step": 15675 }, { "epoch": 6.18, "learning_rate": 1.3458494498692027e-07, "loss": 0.0112, "step": 15676 }, { "epoch": 6.18, "learning_rate": 1.3456081678405264e-07, "loss": 0.0106, "step": 15677 }, { "epoch": 6.18, "learning_rate": 1.345366896477442e-07, "loss": 0.0097, "step": 15678 }, { "epoch": 6.18, "learning_rate": 1.3451256357838825e-07, "loss": 0.01, "step": 15679 }, { "epoch": 6.18, "learning_rate": 1.34488438576378e-07, "loss": 0.0094, "step": 15680 }, { "epoch": 6.18, "learning_rate": 1.3446431464210662e-07, "loss": 0.0132, "step": 15681 }, { "epoch": 6.18, "learning_rate": 1.344401917759672e-07, "loss": 0.0143, "step": 15682 }, { "epoch": 6.18, "learning_rate": 1.3441606997835314e-07, "loss": 0.0106, "step": 15683 }, { "epoch": 6.18, "learning_rate": 1.3439194924965746e-07, "loss": 0.0109, "step": 15684 }, { "epoch": 6.18, "learning_rate": 1.343678295902733e-07, "loss": 0.0108, "step": 15685 }, { "epoch": 6.18, "learning_rate": 1.3434371100059373e-07, "loss": 0.0122, "step": 15686 }, { "epoch": 6.18, "learning_rate": 1.3431959348101186e-07, "loss": 0.0107, "step": 15687 }, { "epoch": 6.18, "learning_rate": 1.3429547703192084e-07, "loss": 0.0124, "step": 15688 }, { "epoch": 6.18, "learning_rate": 1.3427136165371368e-07, "loss": 0.0099, "step": 15689 }, { "epoch": 6.18, "learning_rate": 1.3424724734678346e-07, "loss": 0.0129, "step": 15690 }, { "epoch": 6.18, "learning_rate": 1.3422313411152306e-07, "loss": 0.0119, "step": 15691 }, { "epoch": 6.18, "learning_rate": 1.341990219483257e-07, "loss": 0.0129, "step": 15692 }, { "epoch": 6.18, "learning_rate": 1.3417491085758426e-07, "loss": 0.0114, "step": 15693 }, { "epoch": 6.19, "learning_rate": 1.3415080083969175e-07, "loss": 0.0115, "step": 15694 }, { "epoch": 6.19, "learning_rate": 1.341266918950411e-07, "loss": 0.0147, "step": 15695 }, { "epoch": 6.19, "learning_rate": 1.341025840240252e-07, "loss": 0.0139, "step": 15696 }, { "epoch": 6.19, "learning_rate": 1.340784772270371e-07, "loss": 0.011, "step": 15697 }, { "epoch": 6.19, "learning_rate": 1.3405437150446957e-07, "loss": 0.012, "step": 15698 }, { "epoch": 6.19, "learning_rate": 1.3403026685671562e-07, "loss": 0.0107, "step": 15699 }, { "epoch": 6.19, "learning_rate": 1.3400616328416798e-07, "loss": 0.0123, "step": 15700 }, { "epoch": 6.19, "learning_rate": 1.3398206078721966e-07, "loss": 0.0119, "step": 15701 }, { "epoch": 6.19, "learning_rate": 1.339579593662634e-07, "loss": 0.0115, "step": 15702 }, { "epoch": 6.19, "learning_rate": 1.3393385902169204e-07, "loss": 0.0115, "step": 15703 }, { "epoch": 6.19, "learning_rate": 1.3390975975389832e-07, "loss": 0.0115, "step": 15704 }, { "epoch": 6.19, "learning_rate": 1.3388566156327503e-07, "loss": 0.0124, "step": 15705 }, { "epoch": 6.19, "learning_rate": 1.3386156445021505e-07, "loss": 0.0118, "step": 15706 }, { "epoch": 6.19, "learning_rate": 1.338374684151111e-07, "loss": 0.0126, "step": 15707 }, { "epoch": 6.19, "learning_rate": 1.338133734583558e-07, "loss": 0.0135, "step": 15708 }, { "epoch": 6.19, "learning_rate": 1.3378927958034183e-07, "loss": 0.0122, "step": 15709 }, { "epoch": 6.19, "learning_rate": 1.3376518678146208e-07, "loss": 0.0118, "step": 15710 }, { "epoch": 6.19, "learning_rate": 1.3374109506210912e-07, "loss": 0.0123, "step": 15711 }, { "epoch": 6.19, "learning_rate": 1.337170044226756e-07, "loss": 0.0107, "step": 15712 }, { "epoch": 6.19, "learning_rate": 1.3369291486355415e-07, "loss": 0.0101, "step": 15713 }, { "epoch": 6.19, "learning_rate": 1.3366882638513737e-07, "loss": 0.011, "step": 15714 }, { "epoch": 6.19, "learning_rate": 1.3364473898781796e-07, "loss": 0.0126, "step": 15715 }, { "epoch": 6.19, "learning_rate": 1.3362065267198846e-07, "loss": 0.0124, "step": 15716 }, { "epoch": 6.19, "learning_rate": 1.335965674380415e-07, "loss": 0.0094, "step": 15717 }, { "epoch": 6.19, "learning_rate": 1.3357248328636943e-07, "loss": 0.011, "step": 15718 }, { "epoch": 6.19, "learning_rate": 1.3354840021736503e-07, "loss": 0.014, "step": 15719 }, { "epoch": 6.2, "learning_rate": 1.3352431823142066e-07, "loss": 0.0113, "step": 15720 }, { "epoch": 6.2, "learning_rate": 1.335002373289289e-07, "loss": 0.0106, "step": 15721 }, { "epoch": 6.2, "learning_rate": 1.3347615751028217e-07, "loss": 0.0119, "step": 15722 }, { "epoch": 6.2, "learning_rate": 1.3345207877587298e-07, "loss": 0.0109, "step": 15723 }, { "epoch": 6.2, "learning_rate": 1.334280011260938e-07, "loss": 0.0121, "step": 15724 }, { "epoch": 6.2, "learning_rate": 1.3340392456133703e-07, "loss": 0.014, "step": 15725 }, { "epoch": 6.2, "learning_rate": 1.3337984908199504e-07, "loss": 0.014, "step": 15726 }, { "epoch": 6.2, "learning_rate": 1.333557746884602e-07, "loss": 0.0105, "step": 15727 }, { "epoch": 6.2, "learning_rate": 1.33331701381125e-07, "loss": 0.0119, "step": 15728 }, { "epoch": 6.2, "learning_rate": 1.333076291603818e-07, "loss": 0.0111, "step": 15729 }, { "epoch": 6.2, "learning_rate": 1.3328355802662283e-07, "loss": 0.013, "step": 15730 }, { "epoch": 6.2, "learning_rate": 1.332594879802404e-07, "loss": 0.0131, "step": 15731 }, { "epoch": 6.2, "learning_rate": 1.3323541902162697e-07, "loss": 0.0094, "step": 15732 }, { "epoch": 6.2, "learning_rate": 1.3321135115117475e-07, "loss": 0.0139, "step": 15733 }, { "epoch": 6.2, "learning_rate": 1.33187284369276e-07, "loss": 0.0116, "step": 15734 }, { "epoch": 6.2, "learning_rate": 1.3316321867632294e-07, "loss": 0.0122, "step": 15735 }, { "epoch": 6.2, "learning_rate": 1.3313915407270776e-07, "loss": 0.0092, "step": 15736 }, { "epoch": 6.2, "learning_rate": 1.3311509055882286e-07, "loss": 0.0128, "step": 15737 }, { "epoch": 6.2, "learning_rate": 1.3309102813506034e-07, "loss": 0.0119, "step": 15738 }, { "epoch": 6.2, "learning_rate": 1.3306696680181234e-07, "loss": 0.0091, "step": 15739 }, { "epoch": 6.2, "learning_rate": 1.3304290655947102e-07, "loss": 0.0128, "step": 15740 }, { "epoch": 6.2, "learning_rate": 1.330188474084286e-07, "loss": 0.0115, "step": 15741 }, { "epoch": 6.2, "learning_rate": 1.3299478934907716e-07, "loss": 0.0138, "step": 15742 }, { "epoch": 6.2, "learning_rate": 1.3297073238180877e-07, "loss": 0.0141, "step": 15743 }, { "epoch": 6.2, "learning_rate": 1.3294667650701565e-07, "loss": 0.0126, "step": 15744 }, { "epoch": 6.21, "learning_rate": 1.3292262172508972e-07, "loss": 0.01, "step": 15745 }, { "epoch": 6.21, "learning_rate": 1.3289856803642316e-07, "loss": 0.0135, "step": 15746 }, { "epoch": 6.21, "learning_rate": 1.3287451544140797e-07, "loss": 0.0111, "step": 15747 }, { "epoch": 6.21, "learning_rate": 1.3285046394043616e-07, "loss": 0.0153, "step": 15748 }, { "epoch": 6.21, "learning_rate": 1.3282641353389963e-07, "loss": 0.0148, "step": 15749 }, { "epoch": 6.21, "learning_rate": 1.3280236422219056e-07, "loss": 0.0088, "step": 15750 }, { "epoch": 6.21, "learning_rate": 1.3277831600570083e-07, "loss": 0.0086, "step": 15751 }, { "epoch": 6.21, "learning_rate": 1.3275426888482236e-07, "loss": 0.0109, "step": 15752 }, { "epoch": 6.21, "learning_rate": 1.3273022285994714e-07, "loss": 0.0129, "step": 15753 }, { "epoch": 6.21, "learning_rate": 1.3270617793146692e-07, "loss": 0.0137, "step": 15754 }, { "epoch": 6.21, "learning_rate": 1.3268213409977384e-07, "loss": 0.0096, "step": 15755 }, { "epoch": 6.21, "learning_rate": 1.3265809136525967e-07, "loss": 0.0129, "step": 15756 }, { "epoch": 6.21, "learning_rate": 1.3263404972831626e-07, "loss": 0.0118, "step": 15757 }, { "epoch": 6.21, "learning_rate": 1.3261000918933537e-07, "loss": 0.0123, "step": 15758 }, { "epoch": 6.21, "learning_rate": 1.3258596974870898e-07, "loss": 0.0133, "step": 15759 }, { "epoch": 6.21, "learning_rate": 1.3256193140682882e-07, "loss": 0.0113, "step": 15760 }, { "epoch": 6.21, "learning_rate": 1.325378941640867e-07, "loss": 0.0153, "step": 15761 }, { "epoch": 6.21, "learning_rate": 1.325138580208743e-07, "loss": 0.0114, "step": 15762 }, { "epoch": 6.21, "learning_rate": 1.3248982297758352e-07, "loss": 0.015, "step": 15763 }, { "epoch": 6.21, "learning_rate": 1.3246578903460605e-07, "loss": 0.015, "step": 15764 }, { "epoch": 6.21, "learning_rate": 1.3244175619233353e-07, "loss": 0.0135, "step": 15765 }, { "epoch": 6.21, "learning_rate": 1.324177244511577e-07, "loss": 0.0129, "step": 15766 }, { "epoch": 6.21, "learning_rate": 1.3239369381147023e-07, "loss": 0.014, "step": 15767 }, { "epoch": 6.21, "learning_rate": 1.3236966427366283e-07, "loss": 0.0101, "step": 15768 }, { "epoch": 6.21, "learning_rate": 1.3234563583812715e-07, "loss": 0.0123, "step": 15769 }, { "epoch": 6.22, "learning_rate": 1.3232160850525475e-07, "loss": 0.0124, "step": 15770 }, { "epoch": 6.22, "learning_rate": 1.3229758227543722e-07, "loss": 0.0105, "step": 15771 }, { "epoch": 6.22, "learning_rate": 1.3227355714906625e-07, "loss": 0.0123, "step": 15772 }, { "epoch": 6.22, "learning_rate": 1.322495331265334e-07, "loss": 0.0109, "step": 15773 }, { "epoch": 6.22, "learning_rate": 1.322255102082302e-07, "loss": 0.0124, "step": 15774 }, { "epoch": 6.22, "learning_rate": 1.322014883945481e-07, "loss": 0.0116, "step": 15775 }, { "epoch": 6.22, "learning_rate": 1.321774676858787e-07, "loss": 0.0124, "step": 15776 }, { "epoch": 6.22, "learning_rate": 1.3215344808261355e-07, "loss": 0.0108, "step": 15777 }, { "epoch": 6.22, "learning_rate": 1.3212942958514407e-07, "loss": 0.0106, "step": 15778 }, { "epoch": 6.22, "learning_rate": 1.3210541219386174e-07, "loss": 0.0134, "step": 15779 }, { "epoch": 6.22, "learning_rate": 1.3208139590915795e-07, "loss": 0.0103, "step": 15780 }, { "epoch": 6.22, "learning_rate": 1.3205738073142423e-07, "loss": 0.0109, "step": 15781 }, { "epoch": 6.22, "learning_rate": 1.3203336666105197e-07, "loss": 0.0101, "step": 15782 }, { "epoch": 6.22, "learning_rate": 1.3200935369843252e-07, "loss": 0.0121, "step": 15783 }, { "epoch": 6.22, "learning_rate": 1.3198534184395726e-07, "loss": 0.0133, "step": 15784 }, { "epoch": 6.22, "learning_rate": 1.3196133109801752e-07, "loss": 0.0138, "step": 15785 }, { "epoch": 6.22, "learning_rate": 1.3193732146100475e-07, "loss": 0.0145, "step": 15786 }, { "epoch": 6.22, "learning_rate": 1.3191331293331018e-07, "loss": 0.0125, "step": 15787 }, { "epoch": 6.22, "learning_rate": 1.3188930551532512e-07, "loss": 0.0133, "step": 15788 }, { "epoch": 6.22, "learning_rate": 1.3186529920744077e-07, "loss": 0.0136, "step": 15789 }, { "epoch": 6.22, "learning_rate": 1.318412940100486e-07, "loss": 0.0132, "step": 15790 }, { "epoch": 6.22, "learning_rate": 1.3181728992353973e-07, "loss": 0.0132, "step": 15791 }, { "epoch": 6.22, "learning_rate": 1.3179328694830542e-07, "loss": 0.0112, "step": 15792 }, { "epoch": 6.22, "learning_rate": 1.3176928508473685e-07, "loss": 0.0126, "step": 15793 }, { "epoch": 6.22, "learning_rate": 1.317452843332252e-07, "loss": 0.0142, "step": 15794 }, { "epoch": 6.22, "learning_rate": 1.317212846941617e-07, "loss": 0.0142, "step": 15795 }, { "epoch": 6.23, "learning_rate": 1.316972861679375e-07, "loss": 0.0115, "step": 15796 }, { "epoch": 6.23, "learning_rate": 1.3167328875494377e-07, "loss": 0.0112, "step": 15797 }, { "epoch": 6.23, "learning_rate": 1.3164929245557147e-07, "loss": 0.0109, "step": 15798 }, { "epoch": 6.23, "learning_rate": 1.316252972702119e-07, "loss": 0.0132, "step": 15799 }, { "epoch": 6.23, "learning_rate": 1.3160130319925609e-07, "loss": 0.0123, "step": 15800 }, { "epoch": 6.23, "learning_rate": 1.315773102430951e-07, "loss": 0.013, "step": 15801 }, { "epoch": 6.23, "learning_rate": 1.315533184021199e-07, "loss": 0.0154, "step": 15802 }, { "epoch": 6.23, "learning_rate": 1.3152932767672158e-07, "loss": 0.0136, "step": 15803 }, { "epoch": 6.23, "learning_rate": 1.3150533806729118e-07, "loss": 0.0118, "step": 15804 }, { "epoch": 6.23, "learning_rate": 1.314813495742197e-07, "loss": 0.011, "step": 15805 }, { "epoch": 6.23, "learning_rate": 1.3145736219789805e-07, "loss": 0.0106, "step": 15806 }, { "epoch": 6.23, "learning_rate": 1.3143337593871722e-07, "loss": 0.0127, "step": 15807 }, { "epoch": 6.23, "learning_rate": 1.314093907970682e-07, "loss": 0.0137, "step": 15808 }, { "epoch": 6.23, "learning_rate": 1.313854067733418e-07, "loss": 0.0133, "step": 15809 }, { "epoch": 6.23, "learning_rate": 1.3136142386792904e-07, "loss": 0.0096, "step": 15810 }, { "epoch": 6.23, "learning_rate": 1.3133744208122066e-07, "loss": 0.0127, "step": 15811 }, { "epoch": 6.23, "learning_rate": 1.313134614136077e-07, "loss": 0.0122, "step": 15812 }, { "epoch": 6.23, "learning_rate": 1.3128948186548093e-07, "loss": 0.0106, "step": 15813 }, { "epoch": 6.23, "learning_rate": 1.3126550343723118e-07, "loss": 0.0133, "step": 15814 }, { "epoch": 6.23, "learning_rate": 1.3124152612924927e-07, "loss": 0.0111, "step": 15815 }, { "epoch": 6.23, "learning_rate": 1.3121754994192587e-07, "loss": 0.0109, "step": 15816 }, { "epoch": 6.23, "learning_rate": 1.3119357487565196e-07, "loss": 0.0117, "step": 15817 }, { "epoch": 6.23, "learning_rate": 1.3116960093081822e-07, "loss": 0.0134, "step": 15818 }, { "epoch": 6.23, "learning_rate": 1.3114562810781539e-07, "loss": 0.0114, "step": 15819 }, { "epoch": 6.23, "learning_rate": 1.3112165640703406e-07, "loss": 0.0112, "step": 15820 }, { "epoch": 6.24, "learning_rate": 1.3109768582886515e-07, "loss": 0.0117, "step": 15821 }, { "epoch": 6.24, "learning_rate": 1.310737163736992e-07, "loss": 0.0126, "step": 15822 }, { "epoch": 6.24, "learning_rate": 1.3104974804192698e-07, "loss": 0.0101, "step": 15823 }, { "epoch": 6.24, "learning_rate": 1.3102578083393903e-07, "loss": 0.0111, "step": 15824 }, { "epoch": 6.24, "learning_rate": 1.3100181475012598e-07, "loss": 0.0111, "step": 15825 }, { "epoch": 6.24, "learning_rate": 1.309778497908785e-07, "loss": 0.0121, "step": 15826 }, { "epoch": 6.24, "learning_rate": 1.3095388595658724e-07, "loss": 0.0123, "step": 15827 }, { "epoch": 6.24, "learning_rate": 1.3092992324764266e-07, "loss": 0.0125, "step": 15828 }, { "epoch": 6.24, "learning_rate": 1.3090596166443534e-07, "loss": 0.0121, "step": 15829 }, { "epoch": 6.24, "learning_rate": 1.3088200120735583e-07, "loss": 0.0105, "step": 15830 }, { "epoch": 6.24, "learning_rate": 1.308580418767947e-07, "loss": 0.0106, "step": 15831 }, { "epoch": 6.24, "learning_rate": 1.3083408367314238e-07, "loss": 0.0099, "step": 15832 }, { "epoch": 6.24, "learning_rate": 1.308101265967894e-07, "loss": 0.0134, "step": 15833 }, { "epoch": 6.24, "learning_rate": 1.307861706481261e-07, "loss": 0.0119, "step": 15834 }, { "epoch": 6.24, "learning_rate": 1.307622158275431e-07, "loss": 0.0133, "step": 15835 }, { "epoch": 6.24, "learning_rate": 1.3073826213543076e-07, "loss": 0.015, "step": 15836 }, { "epoch": 6.24, "learning_rate": 1.307143095721795e-07, "loss": 0.0117, "step": 15837 }, { "epoch": 6.24, "learning_rate": 1.3069035813817957e-07, "loss": 0.0102, "step": 15838 }, { "epoch": 6.24, "learning_rate": 1.306664078338216e-07, "loss": 0.0132, "step": 15839 }, { "epoch": 6.24, "learning_rate": 1.3064245865949576e-07, "loss": 0.01, "step": 15840 }, { "epoch": 6.24, "learning_rate": 1.3061851061559245e-07, "loss": 0.0114, "step": 15841 }, { "epoch": 6.24, "learning_rate": 1.3059456370250188e-07, "loss": 0.01, "step": 15842 }, { "epoch": 6.24, "learning_rate": 1.305706179206145e-07, "loss": 0.0124, "step": 15843 }, { "epoch": 6.24, "learning_rate": 1.3054667327032058e-07, "loss": 0.0095, "step": 15844 }, { "epoch": 6.24, "learning_rate": 1.305227297520103e-07, "loss": 0.0122, "step": 15845 }, { "epoch": 6.25, "learning_rate": 1.3049878736607393e-07, "loss": 0.0129, "step": 15846 }, { "epoch": 6.25, "learning_rate": 1.3047484611290166e-07, "loss": 0.0107, "step": 15847 }, { "epoch": 6.25, "learning_rate": 1.3045090599288374e-07, "loss": 0.0104, "step": 15848 }, { "epoch": 6.25, "learning_rate": 1.304269670064104e-07, "loss": 0.0125, "step": 15849 }, { "epoch": 6.25, "learning_rate": 1.3040302915387172e-07, "loss": 0.0144, "step": 15850 }, { "epoch": 6.25, "learning_rate": 1.3037909243565787e-07, "loss": 0.0138, "step": 15851 }, { "epoch": 6.25, "learning_rate": 1.3035515685215902e-07, "loss": 0.0127, "step": 15852 }, { "epoch": 6.25, "learning_rate": 1.303312224037653e-07, "loss": 0.0079, "step": 15853 }, { "epoch": 6.25, "learning_rate": 1.3030728909086678e-07, "loss": 0.0099, "step": 15854 }, { "epoch": 6.25, "learning_rate": 1.302833569138535e-07, "loss": 0.0107, "step": 15855 }, { "epoch": 6.25, "learning_rate": 1.3025942587311548e-07, "loss": 0.009, "step": 15856 }, { "epoch": 6.25, "learning_rate": 1.3023549596904285e-07, "loss": 0.0121, "step": 15857 }, { "epoch": 6.25, "learning_rate": 1.302115672020257e-07, "loss": 0.0132, "step": 15858 }, { "epoch": 6.25, "learning_rate": 1.3018763957245386e-07, "loss": 0.0122, "step": 15859 }, { "epoch": 6.25, "learning_rate": 1.3016371308071733e-07, "loss": 0.0119, "step": 15860 }, { "epoch": 6.25, "learning_rate": 1.3013978772720626e-07, "loss": 0.0136, "step": 15861 }, { "epoch": 6.25, "learning_rate": 1.3011586351231044e-07, "loss": 0.0118, "step": 15862 }, { "epoch": 6.25, "learning_rate": 1.3009194043641985e-07, "loss": 0.0121, "step": 15863 }, { "epoch": 6.25, "learning_rate": 1.3006801849992436e-07, "loss": 0.0132, "step": 15864 }, { "epoch": 6.25, "learning_rate": 1.300440977032138e-07, "loss": 0.011, "step": 15865 }, { "epoch": 6.25, "learning_rate": 1.3002017804667822e-07, "loss": 0.0116, "step": 15866 }, { "epoch": 6.25, "learning_rate": 1.299962595307074e-07, "loss": 0.0133, "step": 15867 }, { "epoch": 6.25, "learning_rate": 1.2997234215569113e-07, "loss": 0.0153, "step": 15868 }, { "epoch": 6.25, "learning_rate": 1.299484259220192e-07, "loss": 0.0134, "step": 15869 }, { "epoch": 6.25, "learning_rate": 1.2992451083008153e-07, "loss": 0.0116, "step": 15870 }, { "epoch": 6.25, "learning_rate": 1.299005968802678e-07, "loss": 0.0118, "step": 15871 }, { "epoch": 6.26, "learning_rate": 1.2987668407296782e-07, "loss": 0.0118, "step": 15872 }, { "epoch": 6.26, "learning_rate": 1.298527724085713e-07, "loss": 0.0108, "step": 15873 }, { "epoch": 6.26, "learning_rate": 1.29828861887468e-07, "loss": 0.0112, "step": 15874 }, { "epoch": 6.26, "learning_rate": 1.2980495251004756e-07, "loss": 0.0119, "step": 15875 }, { "epoch": 6.26, "learning_rate": 1.297810442766998e-07, "loss": 0.0114, "step": 15876 }, { "epoch": 6.26, "learning_rate": 1.297571371878142e-07, "loss": 0.0115, "step": 15877 }, { "epoch": 6.26, "learning_rate": 1.2973323124378048e-07, "loss": 0.0134, "step": 15878 }, { "epoch": 6.26, "learning_rate": 1.297093264449884e-07, "loss": 0.0136, "step": 15879 }, { "epoch": 6.26, "learning_rate": 1.2968542279182742e-07, "loss": 0.0106, "step": 15880 }, { "epoch": 6.26, "learning_rate": 1.296615202846872e-07, "loss": 0.0114, "step": 15881 }, { "epoch": 6.26, "learning_rate": 1.2963761892395723e-07, "loss": 0.0121, "step": 15882 }, { "epoch": 6.26, "learning_rate": 1.2961371871002722e-07, "loss": 0.0096, "step": 15883 }, { "epoch": 6.26, "learning_rate": 1.2958981964328663e-07, "loss": 0.0118, "step": 15884 }, { "epoch": 6.26, "learning_rate": 1.2956592172412493e-07, "loss": 0.0112, "step": 15885 }, { "epoch": 6.26, "learning_rate": 1.2954202495293167e-07, "loss": 0.0108, "step": 15886 }, { "epoch": 6.26, "learning_rate": 1.2951812933009626e-07, "loss": 0.0137, "step": 15887 }, { "epoch": 6.26, "learning_rate": 1.2949423485600828e-07, "loss": 0.0109, "step": 15888 }, { "epoch": 6.26, "learning_rate": 1.294703415310571e-07, "loss": 0.0144, "step": 15889 }, { "epoch": 6.26, "learning_rate": 1.2944644935563218e-07, "loss": 0.0105, "step": 15890 }, { "epoch": 6.26, "learning_rate": 1.2942255833012285e-07, "loss": 0.0111, "step": 15891 }, { "epoch": 6.26, "learning_rate": 1.2939866845491865e-07, "loss": 0.0115, "step": 15892 }, { "epoch": 6.26, "learning_rate": 1.293747797304088e-07, "loss": 0.0125, "step": 15893 }, { "epoch": 6.26, "learning_rate": 1.2935089215698266e-07, "loss": 0.0135, "step": 15894 }, { "epoch": 6.26, "learning_rate": 1.2932700573502967e-07, "loss": 0.0099, "step": 15895 }, { "epoch": 6.26, "learning_rate": 1.2930312046493902e-07, "loss": 0.0129, "step": 15896 }, { "epoch": 6.27, "learning_rate": 1.292792363471001e-07, "loss": 0.012, "step": 15897 }, { "epoch": 6.27, "learning_rate": 1.2925535338190215e-07, "loss": 0.0105, "step": 15898 }, { "epoch": 6.27, "learning_rate": 1.2923147156973441e-07, "loss": 0.0112, "step": 15899 }, { "epoch": 6.27, "learning_rate": 1.2920759091098608e-07, "loss": 0.0102, "step": 15900 }, { "epoch": 6.27, "learning_rate": 1.2918371140604646e-07, "loss": 0.0116, "step": 15901 }, { "epoch": 6.27, "learning_rate": 1.2915983305530477e-07, "loss": 0.0128, "step": 15902 }, { "epoch": 6.27, "learning_rate": 1.291359558591501e-07, "loss": 0.012, "step": 15903 }, { "epoch": 6.27, "learning_rate": 1.2911207981797166e-07, "loss": 0.011, "step": 15904 }, { "epoch": 6.27, "learning_rate": 1.2908820493215848e-07, "loss": 0.0115, "step": 15905 }, { "epoch": 6.27, "learning_rate": 1.290643312020999e-07, "loss": 0.0124, "step": 15906 }, { "epoch": 6.27, "learning_rate": 1.290404586281849e-07, "loss": 0.0131, "step": 15907 }, { "epoch": 6.27, "learning_rate": 1.2901658721080255e-07, "loss": 0.0118, "step": 15908 }, { "epoch": 6.27, "learning_rate": 1.2899271695034187e-07, "loss": 0.0103, "step": 15909 }, { "epoch": 6.27, "learning_rate": 1.289688478471921e-07, "loss": 0.012, "step": 15910 }, { "epoch": 6.27, "learning_rate": 1.2894497990174214e-07, "loss": 0.0104, "step": 15911 }, { "epoch": 6.27, "learning_rate": 1.2892111311438103e-07, "loss": 0.0163, "step": 15912 }, { "epoch": 6.27, "learning_rate": 1.2889724748549773e-07, "loss": 0.01, "step": 15913 }, { "epoch": 6.27, "learning_rate": 1.2887338301548117e-07, "loss": 0.0115, "step": 15914 }, { "epoch": 6.27, "learning_rate": 1.2884951970472045e-07, "loss": 0.0159, "step": 15915 }, { "epoch": 6.27, "learning_rate": 1.2882565755360437e-07, "loss": 0.0131, "step": 15916 }, { "epoch": 6.27, "learning_rate": 1.2880179656252193e-07, "loss": 0.0131, "step": 15917 }, { "epoch": 6.27, "learning_rate": 1.2877793673186195e-07, "loss": 0.0107, "step": 15918 }, { "epoch": 6.27, "learning_rate": 1.287540780620134e-07, "loss": 0.0117, "step": 15919 }, { "epoch": 6.27, "learning_rate": 1.2873022055336513e-07, "loss": 0.0097, "step": 15920 }, { "epoch": 6.27, "learning_rate": 1.2870636420630592e-07, "loss": 0.0148, "step": 15921 }, { "epoch": 6.27, "learning_rate": 1.2868250902122456e-07, "loss": 0.0119, "step": 15922 }, { "epoch": 6.28, "learning_rate": 1.2865865499850994e-07, "loss": 0.0112, "step": 15923 }, { "epoch": 6.28, "learning_rate": 1.2863480213855086e-07, "loss": 0.0126, "step": 15924 }, { "epoch": 6.28, "learning_rate": 1.2861095044173604e-07, "loss": 0.0142, "step": 15925 }, { "epoch": 6.28, "learning_rate": 1.2858709990845422e-07, "loss": 0.0124, "step": 15926 }, { "epoch": 6.28, "learning_rate": 1.2856325053909406e-07, "loss": 0.0139, "step": 15927 }, { "epoch": 6.28, "learning_rate": 1.2853940233404442e-07, "loss": 0.0126, "step": 15928 }, { "epoch": 6.28, "learning_rate": 1.2851555529369394e-07, "loss": 0.0107, "step": 15929 }, { "epoch": 6.28, "learning_rate": 1.2849170941843125e-07, "loss": 0.0124, "step": 15930 }, { "epoch": 6.28, "learning_rate": 1.284678647086449e-07, "loss": 0.013, "step": 15931 }, { "epoch": 6.28, "learning_rate": 1.2844402116472375e-07, "loss": 0.0125, "step": 15932 }, { "epoch": 6.28, "learning_rate": 1.2842017878705628e-07, "loss": 0.0127, "step": 15933 }, { "epoch": 6.28, "learning_rate": 1.2839633757603117e-07, "loss": 0.0131, "step": 15934 }, { "epoch": 6.28, "learning_rate": 1.2837249753203686e-07, "loss": 0.0125, "step": 15935 }, { "epoch": 6.28, "learning_rate": 1.2834865865546198e-07, "loss": 0.0134, "step": 15936 }, { "epoch": 6.28, "learning_rate": 1.283248209466951e-07, "loss": 0.014, "step": 15937 }, { "epoch": 6.28, "learning_rate": 1.283009844061247e-07, "loss": 0.0097, "step": 15938 }, { "epoch": 6.28, "learning_rate": 1.2827714903413925e-07, "loss": 0.0135, "step": 15939 }, { "epoch": 6.28, "learning_rate": 1.2825331483112723e-07, "loss": 0.0105, "step": 15940 }, { "epoch": 6.28, "learning_rate": 1.282294817974772e-07, "loss": 0.0139, "step": 15941 }, { "epoch": 6.28, "learning_rate": 1.2820564993357756e-07, "loss": 0.0167, "step": 15942 }, { "epoch": 6.28, "learning_rate": 1.2818181923981671e-07, "loss": 0.01, "step": 15943 }, { "epoch": 6.28, "learning_rate": 1.2815798971658303e-07, "loss": 0.0182, "step": 15944 }, { "epoch": 6.28, "learning_rate": 1.2813416136426485e-07, "loss": 0.0102, "step": 15945 }, { "epoch": 6.28, "learning_rate": 1.2811033418325073e-07, "loss": 0.0135, "step": 15946 }, { "epoch": 6.28, "learning_rate": 1.280865081739289e-07, "loss": 0.0111, "step": 15947 }, { "epoch": 6.29, "learning_rate": 1.2806268333668764e-07, "loss": 0.0107, "step": 15948 }, { "epoch": 6.29, "learning_rate": 1.2803885967191526e-07, "loss": 0.014, "step": 15949 }, { "epoch": 6.29, "learning_rate": 1.2801503718000017e-07, "loss": 0.0131, "step": 15950 }, { "epoch": 6.29, "learning_rate": 1.2799121586133058e-07, "loss": 0.0121, "step": 15951 }, { "epoch": 6.29, "learning_rate": 1.2796739571629474e-07, "loss": 0.0118, "step": 15952 }, { "epoch": 6.29, "learning_rate": 1.2794357674528086e-07, "loss": 0.0154, "step": 15953 }, { "epoch": 6.29, "learning_rate": 1.2791975894867708e-07, "loss": 0.0123, "step": 15954 }, { "epoch": 6.29, "learning_rate": 1.2789594232687179e-07, "loss": 0.0112, "step": 15955 }, { "epoch": 6.29, "learning_rate": 1.2787212688025303e-07, "loss": 0.0137, "step": 15956 }, { "epoch": 6.29, "learning_rate": 1.2784831260920895e-07, "loss": 0.0116, "step": 15957 }, { "epoch": 6.29, "learning_rate": 1.2782449951412772e-07, "loss": 0.0119, "step": 15958 }, { "epoch": 6.29, "learning_rate": 1.278006875953975e-07, "loss": 0.0124, "step": 15959 }, { "epoch": 6.29, "learning_rate": 1.2777687685340633e-07, "loss": 0.0113, "step": 15960 }, { "epoch": 6.29, "learning_rate": 1.277530672885423e-07, "loss": 0.0108, "step": 15961 }, { "epoch": 6.29, "learning_rate": 1.2772925890119337e-07, "loss": 0.0109, "step": 15962 }, { "epoch": 6.29, "learning_rate": 1.2770545169174779e-07, "loss": 0.0123, "step": 15963 }, { "epoch": 6.29, "learning_rate": 1.2768164566059345e-07, "loss": 0.0136, "step": 15964 }, { "epoch": 6.29, "learning_rate": 1.2765784080811837e-07, "loss": 0.0132, "step": 15965 }, { "epoch": 6.29, "learning_rate": 1.2763403713471058e-07, "loss": 0.0119, "step": 15966 }, { "epoch": 6.29, "learning_rate": 1.276102346407579e-07, "loss": 0.0115, "step": 15967 }, { "epoch": 6.29, "learning_rate": 1.2758643332664843e-07, "loss": 0.0126, "step": 15968 }, { "epoch": 6.29, "learning_rate": 1.2756263319277004e-07, "loss": 0.0116, "step": 15969 }, { "epoch": 6.29, "learning_rate": 1.2753883423951068e-07, "loss": 0.0129, "step": 15970 }, { "epoch": 6.29, "learning_rate": 1.275150364672581e-07, "loss": 0.0097, "step": 15971 }, { "epoch": 6.29, "learning_rate": 1.274912398764003e-07, "loss": 0.008, "step": 15972 }, { "epoch": 6.3, "learning_rate": 1.274674444673251e-07, "loss": 0.0129, "step": 15973 }, { "epoch": 6.3, "learning_rate": 1.2744365024042036e-07, "loss": 0.014, "step": 15974 }, { "epoch": 6.3, "learning_rate": 1.2741985719607386e-07, "loss": 0.0155, "step": 15975 }, { "epoch": 6.3, "learning_rate": 1.2739606533467323e-07, "loss": 0.0116, "step": 15976 }, { "epoch": 6.3, "learning_rate": 1.2737227465660652e-07, "loss": 0.011, "step": 15977 }, { "epoch": 6.3, "learning_rate": 1.2734848516226133e-07, "loss": 0.0115, "step": 15978 }, { "epoch": 6.3, "learning_rate": 1.273246968520254e-07, "loss": 0.0119, "step": 15979 }, { "epoch": 6.3, "learning_rate": 1.2730090972628644e-07, "loss": 0.0125, "step": 15980 }, { "epoch": 6.3, "learning_rate": 1.2727712378543222e-07, "loss": 0.0106, "step": 15981 }, { "epoch": 6.3, "learning_rate": 1.2725333902985033e-07, "loss": 0.0102, "step": 15982 }, { "epoch": 6.3, "learning_rate": 1.2722955545992847e-07, "loss": 0.0116, "step": 15983 }, { "epoch": 6.3, "learning_rate": 1.2720577307605425e-07, "loss": 0.0108, "step": 15984 }, { "epoch": 6.3, "learning_rate": 1.271819918786152e-07, "loss": 0.0142, "step": 15985 }, { "epoch": 6.3, "learning_rate": 1.271582118679991e-07, "loss": 0.0145, "step": 15986 }, { "epoch": 6.3, "learning_rate": 1.2713443304459344e-07, "loss": 0.0132, "step": 15987 }, { "epoch": 6.3, "learning_rate": 1.2711065540878577e-07, "loss": 0.0087, "step": 15988 }, { "epoch": 6.3, "learning_rate": 1.2708687896096358e-07, "loss": 0.0135, "step": 15989 }, { "epoch": 6.3, "learning_rate": 1.2706310370151452e-07, "loss": 0.0114, "step": 15990 }, { "epoch": 6.3, "learning_rate": 1.27039329630826e-07, "loss": 0.0124, "step": 15991 }, { "epoch": 6.3, "learning_rate": 1.2701555674928554e-07, "loss": 0.0132, "step": 15992 }, { "epoch": 6.3, "learning_rate": 1.269917850572806e-07, "loss": 0.0124, "step": 15993 }, { "epoch": 6.3, "learning_rate": 1.2696801455519847e-07, "loss": 0.0089, "step": 15994 }, { "epoch": 6.3, "learning_rate": 1.2694424524342683e-07, "loss": 0.0094, "step": 15995 }, { "epoch": 6.3, "learning_rate": 1.2692047712235291e-07, "loss": 0.0125, "step": 15996 }, { "epoch": 6.3, "learning_rate": 1.268967101923642e-07, "loss": 0.0119, "step": 15997 }, { "epoch": 6.3, "learning_rate": 1.2687294445384797e-07, "loss": 0.0122, "step": 15998 }, { "epoch": 6.31, "learning_rate": 1.2684917990719157e-07, "loss": 0.0109, "step": 15999 }, { "epoch": 6.31, "learning_rate": 1.268254165527824e-07, "loss": 0.0133, "step": 16000 }, { "epoch": 6.31, "learning_rate": 1.2680165439100776e-07, "loss": 0.0108, "step": 16001 }, { "epoch": 6.31, "learning_rate": 1.2677789342225485e-07, "loss": 0.0112, "step": 16002 }, { "epoch": 6.31, "learning_rate": 1.2675413364691104e-07, "loss": 0.0146, "step": 16003 }, { "epoch": 6.31, "learning_rate": 1.2673037506536353e-07, "loss": 0.0121, "step": 16004 }, { "epoch": 6.31, "learning_rate": 1.2670661767799954e-07, "loss": 0.0108, "step": 16005 }, { "epoch": 6.31, "learning_rate": 1.2668286148520628e-07, "loss": 0.0103, "step": 16006 }, { "epoch": 6.31, "learning_rate": 1.2665910648737088e-07, "loss": 0.0116, "step": 16007 }, { "epoch": 6.31, "learning_rate": 1.2663535268488067e-07, "loss": 0.0119, "step": 16008 }, { "epoch": 6.31, "learning_rate": 1.2661160007812267e-07, "loss": 0.0115, "step": 16009 }, { "epoch": 6.31, "learning_rate": 1.2658784866748412e-07, "loss": 0.0105, "step": 16010 }, { "epoch": 6.31, "learning_rate": 1.2656409845335192e-07, "loss": 0.0106, "step": 16011 }, { "epoch": 6.31, "learning_rate": 1.265403494361134e-07, "loss": 0.0084, "step": 16012 }, { "epoch": 6.31, "learning_rate": 1.2651660161615554e-07, "loss": 0.0111, "step": 16013 }, { "epoch": 6.31, "learning_rate": 1.264928549938654e-07, "loss": 0.0106, "step": 16014 }, { "epoch": 6.31, "learning_rate": 1.2646910956963002e-07, "loss": 0.0094, "step": 16015 }, { "epoch": 6.31, "learning_rate": 1.264453653438363e-07, "loss": 0.0105, "step": 16016 }, { "epoch": 6.31, "learning_rate": 1.2642162231687142e-07, "loss": 0.0105, "step": 16017 }, { "epoch": 6.31, "learning_rate": 1.2639788048912228e-07, "loss": 0.0112, "step": 16018 }, { "epoch": 6.31, "learning_rate": 1.2637413986097582e-07, "loss": 0.011, "step": 16019 }, { "epoch": 6.31, "learning_rate": 1.263504004328189e-07, "loss": 0.0124, "step": 16020 }, { "epoch": 6.31, "learning_rate": 1.263266622050386e-07, "loss": 0.0121, "step": 16021 }, { "epoch": 6.31, "learning_rate": 1.263029251780217e-07, "loss": 0.0107, "step": 16022 }, { "epoch": 6.31, "learning_rate": 1.2627918935215513e-07, "loss": 0.0101, "step": 16023 }, { "epoch": 6.32, "learning_rate": 1.2625545472782576e-07, "loss": 0.0148, "step": 16024 }, { "epoch": 6.32, "learning_rate": 1.2623172130542034e-07, "loss": 0.0129, "step": 16025 }, { "epoch": 6.32, "learning_rate": 1.2620798908532579e-07, "loss": 0.0126, "step": 16026 }, { "epoch": 6.32, "learning_rate": 1.2618425806792886e-07, "loss": 0.0101, "step": 16027 }, { "epoch": 6.32, "learning_rate": 1.2616052825361635e-07, "loss": 0.0114, "step": 16028 }, { "epoch": 6.32, "learning_rate": 1.261367996427749e-07, "loss": 0.0133, "step": 16029 }, { "epoch": 6.32, "learning_rate": 1.2611307223579148e-07, "loss": 0.013, "step": 16030 }, { "epoch": 6.32, "learning_rate": 1.2608934603305265e-07, "loss": 0.0114, "step": 16031 }, { "epoch": 6.32, "learning_rate": 1.2606562103494515e-07, "loss": 0.011, "step": 16032 }, { "epoch": 6.32, "learning_rate": 1.260418972418557e-07, "loss": 0.0118, "step": 16033 }, { "epoch": 6.32, "learning_rate": 1.260181746541708e-07, "loss": 0.013, "step": 16034 }, { "epoch": 6.32, "learning_rate": 1.2599445327227732e-07, "loss": 0.0127, "step": 16035 }, { "epoch": 6.32, "learning_rate": 1.2597073309656178e-07, "loss": 0.0108, "step": 16036 }, { "epoch": 6.32, "learning_rate": 1.2594701412741073e-07, "loss": 0.0141, "step": 16037 }, { "epoch": 6.32, "learning_rate": 1.2592329636521077e-07, "loss": 0.0117, "step": 16038 }, { "epoch": 6.32, "learning_rate": 1.2589957981034855e-07, "loss": 0.015, "step": 16039 }, { "epoch": 6.32, "learning_rate": 1.2587586446321057e-07, "loss": 0.0124, "step": 16040 }, { "epoch": 6.32, "learning_rate": 1.2585215032418335e-07, "loss": 0.0096, "step": 16041 }, { "epoch": 6.32, "learning_rate": 1.2582843739365335e-07, "loss": 0.0112, "step": 16042 }, { "epoch": 6.32, "learning_rate": 1.2580472567200713e-07, "loss": 0.0116, "step": 16043 }, { "epoch": 6.32, "learning_rate": 1.2578101515963107e-07, "loss": 0.0126, "step": 16044 }, { "epoch": 6.32, "learning_rate": 1.2575730585691168e-07, "loss": 0.0136, "step": 16045 }, { "epoch": 6.32, "learning_rate": 1.257335977642354e-07, "loss": 0.0101, "step": 16046 }, { "epoch": 6.32, "learning_rate": 1.2570989088198854e-07, "loss": 0.014, "step": 16047 }, { "epoch": 6.32, "learning_rate": 1.2568618521055766e-07, "loss": 0.0162, "step": 16048 }, { "epoch": 6.33, "learning_rate": 1.2566248075032894e-07, "loss": 0.0119, "step": 16049 }, { "epoch": 6.33, "learning_rate": 1.2563877750168884e-07, "loss": 0.0119, "step": 16050 }, { "epoch": 6.33, "learning_rate": 1.2561507546502353e-07, "loss": 0.0147, "step": 16051 }, { "epoch": 6.33, "learning_rate": 1.2559137464071955e-07, "loss": 0.0124, "step": 16052 }, { "epoch": 6.33, "learning_rate": 1.255676750291631e-07, "loss": 0.0101, "step": 16053 }, { "epoch": 6.33, "learning_rate": 1.2554397663074038e-07, "loss": 0.0101, "step": 16054 }, { "epoch": 6.33, "learning_rate": 1.255202794458377e-07, "loss": 0.0123, "step": 16055 }, { "epoch": 6.33, "learning_rate": 1.254965834748412e-07, "loss": 0.0117, "step": 16056 }, { "epoch": 6.33, "learning_rate": 1.254728887181372e-07, "loss": 0.0108, "step": 16057 }, { "epoch": 6.33, "learning_rate": 1.254491951761119e-07, "loss": 0.0119, "step": 16058 }, { "epoch": 6.33, "learning_rate": 1.2542550284915137e-07, "loss": 0.0119, "step": 16059 }, { "epoch": 6.33, "learning_rate": 1.2540181173764172e-07, "loss": 0.0121, "step": 16060 }, { "epoch": 6.33, "learning_rate": 1.2537812184196927e-07, "loss": 0.0106, "step": 16061 }, { "epoch": 6.33, "learning_rate": 1.2535443316252003e-07, "loss": 0.0124, "step": 16062 }, { "epoch": 6.33, "learning_rate": 1.2533074569968005e-07, "loss": 0.0119, "step": 16063 }, { "epoch": 6.33, "learning_rate": 1.2530705945383546e-07, "loss": 0.0114, "step": 16064 }, { "epoch": 6.33, "learning_rate": 1.252833744253722e-07, "loss": 0.0106, "step": 16065 }, { "epoch": 6.33, "learning_rate": 1.2525969061467647e-07, "loss": 0.0124, "step": 16066 }, { "epoch": 6.33, "learning_rate": 1.252360080221341e-07, "loss": 0.0088, "step": 16067 }, { "epoch": 6.33, "learning_rate": 1.2521232664813126e-07, "loss": 0.0103, "step": 16068 }, { "epoch": 6.33, "learning_rate": 1.2518864649305376e-07, "loss": 0.011, "step": 16069 }, { "epoch": 6.33, "learning_rate": 1.2516496755728768e-07, "loss": 0.011, "step": 16070 }, { "epoch": 6.33, "learning_rate": 1.2514128984121887e-07, "loss": 0.0117, "step": 16071 }, { "epoch": 6.33, "learning_rate": 1.2511761334523329e-07, "loss": 0.0107, "step": 16072 }, { "epoch": 6.33, "learning_rate": 1.2509393806971678e-07, "loss": 0.0124, "step": 16073 }, { "epoch": 6.33, "learning_rate": 1.2507026401505516e-07, "loss": 0.0104, "step": 16074 }, { "epoch": 6.34, "learning_rate": 1.2504659118163446e-07, "loss": 0.0128, "step": 16075 }, { "epoch": 6.34, "learning_rate": 1.2502291956984037e-07, "loss": 0.0122, "step": 16076 }, { "epoch": 6.34, "learning_rate": 1.2499924918005877e-07, "loss": 0.0114, "step": 16077 }, { "epoch": 6.34, "learning_rate": 1.2497558001267532e-07, "loss": 0.0134, "step": 16078 }, { "epoch": 6.34, "learning_rate": 1.2495191206807598e-07, "loss": 0.0133, "step": 16079 }, { "epoch": 6.34, "learning_rate": 1.2492824534664644e-07, "loss": 0.0117, "step": 16080 }, { "epoch": 6.34, "learning_rate": 1.2490457984877236e-07, "loss": 0.0111, "step": 16081 }, { "epoch": 6.34, "learning_rate": 1.2488091557483944e-07, "loss": 0.0119, "step": 16082 }, { "epoch": 6.34, "learning_rate": 1.2485725252523352e-07, "loss": 0.0151, "step": 16083 }, { "epoch": 6.34, "learning_rate": 1.2483359070034017e-07, "loss": 0.0119, "step": 16084 }, { "epoch": 6.34, "learning_rate": 1.24809930100545e-07, "loss": 0.0132, "step": 16085 }, { "epoch": 6.34, "learning_rate": 1.2478627072623376e-07, "loss": 0.0088, "step": 16086 }, { "epoch": 6.34, "learning_rate": 1.2476261257779194e-07, "loss": 0.014, "step": 16087 }, { "epoch": 6.34, "learning_rate": 1.2473895565560523e-07, "loss": 0.0143, "step": 16088 }, { "epoch": 6.34, "learning_rate": 1.2471529996005913e-07, "loss": 0.0149, "step": 16089 }, { "epoch": 6.34, "learning_rate": 1.246916454915392e-07, "loss": 0.009, "step": 16090 }, { "epoch": 6.34, "learning_rate": 1.2466799225043102e-07, "loss": 0.0096, "step": 16091 }, { "epoch": 6.34, "learning_rate": 1.246443402371201e-07, "loss": 0.0146, "step": 16092 }, { "epoch": 6.34, "learning_rate": 1.246206894519919e-07, "loss": 0.0124, "step": 16093 }, { "epoch": 6.34, "learning_rate": 1.2459703989543188e-07, "loss": 0.0126, "step": 16094 }, { "epoch": 6.34, "learning_rate": 1.2457339156782553e-07, "loss": 0.0145, "step": 16095 }, { "epoch": 6.34, "learning_rate": 1.245497444695582e-07, "loss": 0.013, "step": 16096 }, { "epoch": 6.34, "learning_rate": 1.245260986010154e-07, "loss": 0.0093, "step": 16097 }, { "epoch": 6.34, "learning_rate": 1.2450245396258252e-07, "loss": 0.0084, "step": 16098 }, { "epoch": 6.34, "learning_rate": 1.2447881055464486e-07, "loss": 0.0113, "step": 16099 }, { "epoch": 6.35, "learning_rate": 1.2445516837758777e-07, "loss": 0.0126, "step": 16100 }, { "epoch": 6.35, "learning_rate": 1.244315274317967e-07, "loss": 0.0113, "step": 16101 }, { "epoch": 6.35, "learning_rate": 1.2440788771765687e-07, "loss": 0.0148, "step": 16102 }, { "epoch": 6.35, "learning_rate": 1.2438424923555358e-07, "loss": 0.0136, "step": 16103 }, { "epoch": 6.35, "learning_rate": 1.2436061198587208e-07, "loss": 0.0086, "step": 16104 }, { "epoch": 6.35, "learning_rate": 1.2433697596899762e-07, "loss": 0.0127, "step": 16105 }, { "epoch": 6.35, "learning_rate": 1.2431334118531549e-07, "loss": 0.0122, "step": 16106 }, { "epoch": 6.35, "learning_rate": 1.242897076352109e-07, "loss": 0.0115, "step": 16107 }, { "epoch": 6.35, "learning_rate": 1.24266075319069e-07, "loss": 0.0142, "step": 16108 }, { "epoch": 6.35, "learning_rate": 1.2424244423727492e-07, "loss": 0.0133, "step": 16109 }, { "epoch": 6.35, "learning_rate": 1.242188143902139e-07, "loss": 0.0109, "step": 16110 }, { "epoch": 6.35, "learning_rate": 1.2419518577827104e-07, "loss": 0.009, "step": 16111 }, { "epoch": 6.35, "learning_rate": 1.2417155840183142e-07, "loss": 0.013, "step": 16112 }, { "epoch": 6.35, "learning_rate": 1.2414793226128015e-07, "loss": 0.0131, "step": 16113 }, { "epoch": 6.35, "learning_rate": 1.2412430735700227e-07, "loss": 0.0129, "step": 16114 }, { "epoch": 6.35, "learning_rate": 1.2410068368938293e-07, "loss": 0.0131, "step": 16115 }, { "epoch": 6.35, "learning_rate": 1.2407706125880707e-07, "loss": 0.01, "step": 16116 }, { "epoch": 6.35, "learning_rate": 1.2405344006565973e-07, "loss": 0.0141, "step": 16117 }, { "epoch": 6.35, "learning_rate": 1.2402982011032582e-07, "loss": 0.0107, "step": 16118 }, { "epoch": 6.35, "learning_rate": 1.2400620139319044e-07, "loss": 0.0131, "step": 16119 }, { "epoch": 6.35, "learning_rate": 1.239825839146385e-07, "loss": 0.0152, "step": 16120 }, { "epoch": 6.35, "learning_rate": 1.2395896767505492e-07, "loss": 0.0123, "step": 16121 }, { "epoch": 6.35, "learning_rate": 1.239353526748245e-07, "loss": 0.0125, "step": 16122 }, { "epoch": 6.35, "learning_rate": 1.2391173891433232e-07, "loss": 0.0143, "step": 16123 }, { "epoch": 6.35, "learning_rate": 1.238881263939631e-07, "loss": 0.0139, "step": 16124 }, { "epoch": 6.35, "learning_rate": 1.2386451511410182e-07, "loss": 0.0106, "step": 16125 }, { "epoch": 6.36, "learning_rate": 1.2384090507513318e-07, "loss": 0.0131, "step": 16126 }, { "epoch": 6.36, "learning_rate": 1.2381729627744198e-07, "loss": 0.0112, "step": 16127 }, { "epoch": 6.36, "learning_rate": 1.2379368872141313e-07, "loss": 0.0125, "step": 16128 }, { "epoch": 6.36, "learning_rate": 1.2377008240743135e-07, "loss": 0.0139, "step": 16129 }, { "epoch": 6.36, "learning_rate": 1.2374647733588134e-07, "loss": 0.0112, "step": 16130 }, { "epoch": 6.36, "learning_rate": 1.2372287350714787e-07, "loss": 0.0114, "step": 16131 }, { "epoch": 6.36, "learning_rate": 1.2369927092161563e-07, "loss": 0.0161, "step": 16132 }, { "epoch": 6.36, "learning_rate": 1.2367566957966932e-07, "loss": 0.016, "step": 16133 }, { "epoch": 6.36, "learning_rate": 1.2365206948169362e-07, "loss": 0.0114, "step": 16134 }, { "epoch": 6.36, "learning_rate": 1.2362847062807312e-07, "loss": 0.0122, "step": 16135 }, { "epoch": 6.36, "learning_rate": 1.2360487301919243e-07, "loss": 0.0116, "step": 16136 }, { "epoch": 6.36, "learning_rate": 1.2358127665543626e-07, "loss": 0.012, "step": 16137 }, { "epoch": 6.36, "learning_rate": 1.2355768153718912e-07, "loss": 0.0115, "step": 16138 }, { "epoch": 6.36, "learning_rate": 1.2353408766483563e-07, "loss": 0.0142, "step": 16139 }, { "epoch": 6.36, "learning_rate": 1.2351049503876022e-07, "loss": 0.0114, "step": 16140 }, { "epoch": 6.36, "learning_rate": 1.2348690365934756e-07, "loss": 0.0097, "step": 16141 }, { "epoch": 6.36, "learning_rate": 1.2346331352698205e-07, "loss": 0.0113, "step": 16142 }, { "epoch": 6.36, "learning_rate": 1.2343972464204822e-07, "loss": 0.0137, "step": 16143 }, { "epoch": 6.36, "learning_rate": 1.2341613700493054e-07, "loss": 0.0107, "step": 16144 }, { "epoch": 6.36, "learning_rate": 1.2339255061601335e-07, "loss": 0.0123, "step": 16145 }, { "epoch": 6.36, "learning_rate": 1.2336896547568124e-07, "loss": 0.0126, "step": 16146 }, { "epoch": 6.36, "learning_rate": 1.233453815843185e-07, "loss": 0.0119, "step": 16147 }, { "epoch": 6.36, "learning_rate": 1.2332179894230957e-07, "loss": 0.0135, "step": 16148 }, { "epoch": 6.36, "learning_rate": 1.2329821755003873e-07, "loss": 0.0119, "step": 16149 }, { "epoch": 6.36, "learning_rate": 1.2327463740789033e-07, "loss": 0.0124, "step": 16150 }, { "epoch": 6.37, "learning_rate": 1.2325105851624882e-07, "loss": 0.0139, "step": 16151 }, { "epoch": 6.37, "learning_rate": 1.232274808754984e-07, "loss": 0.0124, "step": 16152 }, { "epoch": 6.37, "learning_rate": 1.2320390448602337e-07, "loss": 0.0152, "step": 16153 }, { "epoch": 6.37, "learning_rate": 1.231803293482079e-07, "loss": 0.0139, "step": 16154 }, { "epoch": 6.37, "learning_rate": 1.231567554624364e-07, "loss": 0.0123, "step": 16155 }, { "epoch": 6.37, "learning_rate": 1.23133182829093e-07, "loss": 0.0164, "step": 16156 }, { "epoch": 6.37, "learning_rate": 1.2310961144856188e-07, "loss": 0.0114, "step": 16157 }, { "epoch": 6.37, "learning_rate": 1.2308604132122712e-07, "loss": 0.0094, "step": 16158 }, { "epoch": 6.37, "learning_rate": 1.230624724474731e-07, "loss": 0.0104, "step": 16159 }, { "epoch": 6.37, "learning_rate": 1.2303890482768385e-07, "loss": 0.0138, "step": 16160 }, { "epoch": 6.37, "learning_rate": 1.2301533846224348e-07, "loss": 0.012, "step": 16161 }, { "epoch": 6.37, "learning_rate": 1.2299177335153608e-07, "loss": 0.0111, "step": 16162 }, { "epoch": 6.37, "learning_rate": 1.2296820949594566e-07, "loss": 0.0114, "step": 16163 }, { "epoch": 6.37, "learning_rate": 1.229446468958564e-07, "loss": 0.0118, "step": 16164 }, { "epoch": 6.37, "learning_rate": 1.2292108555165235e-07, "loss": 0.0108, "step": 16165 }, { "epoch": 6.37, "learning_rate": 1.2289752546371742e-07, "loss": 0.0147, "step": 16166 }, { "epoch": 6.37, "learning_rate": 1.2287396663243556e-07, "loss": 0.0107, "step": 16167 }, { "epoch": 6.37, "learning_rate": 1.2285040905819092e-07, "loss": 0.0122, "step": 16168 }, { "epoch": 6.37, "learning_rate": 1.2282685274136735e-07, "loss": 0.0108, "step": 16169 }, { "epoch": 6.37, "learning_rate": 1.2280329768234881e-07, "loss": 0.0129, "step": 16170 }, { "epoch": 6.37, "learning_rate": 1.2277974388151914e-07, "loss": 0.0116, "step": 16171 }, { "epoch": 6.37, "learning_rate": 1.2275619133926226e-07, "loss": 0.0128, "step": 16172 }, { "epoch": 6.37, "learning_rate": 1.2273264005596213e-07, "loss": 0.0115, "step": 16173 }, { "epoch": 6.37, "learning_rate": 1.2270909003200255e-07, "loss": 0.0116, "step": 16174 }, { "epoch": 6.37, "learning_rate": 1.2268554126776728e-07, "loss": 0.0106, "step": 16175 }, { "epoch": 6.38, "learning_rate": 1.226619937636402e-07, "loss": 0.0111, "step": 16176 }, { "epoch": 6.38, "learning_rate": 1.226384475200051e-07, "loss": 0.012, "step": 16177 }, { "epoch": 6.38, "learning_rate": 1.2261490253724574e-07, "loss": 0.016, "step": 16178 }, { "epoch": 6.38, "learning_rate": 1.2259135881574582e-07, "loss": 0.0161, "step": 16179 }, { "epoch": 6.38, "learning_rate": 1.2256781635588906e-07, "loss": 0.0128, "step": 16180 }, { "epoch": 6.38, "learning_rate": 1.2254427515805924e-07, "loss": 0.0155, "step": 16181 }, { "epoch": 6.38, "learning_rate": 1.2252073522264008e-07, "loss": 0.0122, "step": 16182 }, { "epoch": 6.38, "learning_rate": 1.2249719655001516e-07, "loss": 0.0124, "step": 16183 }, { "epoch": 6.38, "learning_rate": 1.2247365914056813e-07, "loss": 0.013, "step": 16184 }, { "epoch": 6.38, "learning_rate": 1.2245012299468255e-07, "loss": 0.0126, "step": 16185 }, { "epoch": 6.38, "learning_rate": 1.224265881127422e-07, "loss": 0.0114, "step": 16186 }, { "epoch": 6.38, "learning_rate": 1.2240305449513057e-07, "loss": 0.0128, "step": 16187 }, { "epoch": 6.38, "learning_rate": 1.223795221422312e-07, "loss": 0.0104, "step": 16188 }, { "epoch": 6.38, "learning_rate": 1.2235599105442757e-07, "loss": 0.0137, "step": 16189 }, { "epoch": 6.38, "learning_rate": 1.2233246123210338e-07, "loss": 0.0139, "step": 16190 }, { "epoch": 6.38, "learning_rate": 1.22308932675642e-07, "loss": 0.0122, "step": 16191 }, { "epoch": 6.38, "learning_rate": 1.2228540538542696e-07, "loss": 0.0119, "step": 16192 }, { "epoch": 6.38, "learning_rate": 1.222618793618417e-07, "loss": 0.0125, "step": 16193 }, { "epoch": 6.38, "learning_rate": 1.222383546052696e-07, "loss": 0.0134, "step": 16194 }, { "epoch": 6.38, "learning_rate": 1.2221483111609416e-07, "loss": 0.0107, "step": 16195 }, { "epoch": 6.38, "learning_rate": 1.2219130889469876e-07, "loss": 0.0102, "step": 16196 }, { "epoch": 6.38, "learning_rate": 1.2216778794146683e-07, "loss": 0.0127, "step": 16197 }, { "epoch": 6.38, "learning_rate": 1.2214426825678156e-07, "loss": 0.011, "step": 16198 }, { "epoch": 6.38, "learning_rate": 1.2212074984102648e-07, "loss": 0.0137, "step": 16199 }, { "epoch": 6.38, "learning_rate": 1.2209723269458478e-07, "loss": 0.0136, "step": 16200 }, { "epoch": 6.38, "learning_rate": 1.220737168178398e-07, "loss": 0.0122, "step": 16201 }, { "epoch": 6.39, "learning_rate": 1.2205020221117482e-07, "loss": 0.0119, "step": 16202 }, { "epoch": 6.39, "learning_rate": 1.2202668887497298e-07, "loss": 0.011, "step": 16203 }, { "epoch": 6.39, "learning_rate": 1.2200317680961766e-07, "loss": 0.012, "step": 16204 }, { "epoch": 6.39, "learning_rate": 1.2197966601549206e-07, "loss": 0.0128, "step": 16205 }, { "epoch": 6.39, "learning_rate": 1.219561564929793e-07, "loss": 0.0127, "step": 16206 }, { "epoch": 6.39, "learning_rate": 1.219326482424625e-07, "loss": 0.0094, "step": 16207 }, { "epoch": 6.39, "learning_rate": 1.2190914126432496e-07, "loss": 0.013, "step": 16208 }, { "epoch": 6.39, "learning_rate": 1.2188563555894975e-07, "loss": 0.0127, "step": 16209 }, { "epoch": 6.39, "learning_rate": 1.2186213112671995e-07, "loss": 0.0109, "step": 16210 }, { "epoch": 6.39, "learning_rate": 1.218386279680186e-07, "loss": 0.0117, "step": 16211 }, { "epoch": 6.39, "learning_rate": 1.2181512608322886e-07, "loss": 0.0133, "step": 16212 }, { "epoch": 6.39, "learning_rate": 1.2179162547273377e-07, "loss": 0.0102, "step": 16213 }, { "epoch": 6.39, "learning_rate": 1.217681261369163e-07, "loss": 0.0125, "step": 16214 }, { "epoch": 6.39, "learning_rate": 1.2174462807615947e-07, "loss": 0.011, "step": 16215 }, { "epoch": 6.39, "learning_rate": 1.2172113129084624e-07, "loss": 0.0119, "step": 16216 }, { "epoch": 6.39, "learning_rate": 1.2169763578135966e-07, "loss": 0.0122, "step": 16217 }, { "epoch": 6.39, "learning_rate": 1.2167414154808256e-07, "loss": 0.0145, "step": 16218 }, { "epoch": 6.39, "learning_rate": 1.2165064859139793e-07, "loss": 0.0123, "step": 16219 }, { "epoch": 6.39, "learning_rate": 1.2162715691168863e-07, "loss": 0.0115, "step": 16220 }, { "epoch": 6.39, "learning_rate": 1.2160366650933757e-07, "loss": 0.0111, "step": 16221 }, { "epoch": 6.39, "learning_rate": 1.2158017738472765e-07, "loss": 0.0118, "step": 16222 }, { "epoch": 6.39, "learning_rate": 1.215566895382416e-07, "loss": 0.0111, "step": 16223 }, { "epoch": 6.39, "learning_rate": 1.2153320297026232e-07, "loss": 0.0125, "step": 16224 }, { "epoch": 6.39, "learning_rate": 1.2150971768117247e-07, "loss": 0.0144, "step": 16225 }, { "epoch": 6.39, "learning_rate": 1.2148623367135506e-07, "loss": 0.0128, "step": 16226 }, { "epoch": 6.4, "learning_rate": 1.214627509411927e-07, "loss": 0.0103, "step": 16227 }, { "epoch": 6.4, "learning_rate": 1.214392694910681e-07, "loss": 0.0126, "step": 16228 }, { "epoch": 6.4, "learning_rate": 1.2141578932136397e-07, "loss": 0.0129, "step": 16229 }, { "epoch": 6.4, "learning_rate": 1.213923104324631e-07, "loss": 0.0103, "step": 16230 }, { "epoch": 6.4, "learning_rate": 1.213688328247481e-07, "loss": 0.0107, "step": 16231 }, { "epoch": 6.4, "learning_rate": 1.213453564986016e-07, "loss": 0.0148, "step": 16232 }, { "epoch": 6.4, "learning_rate": 1.2132188145440628e-07, "loss": 0.0119, "step": 16233 }, { "epoch": 6.4, "learning_rate": 1.2129840769254462e-07, "loss": 0.0118, "step": 16234 }, { "epoch": 6.4, "learning_rate": 1.212749352133994e-07, "loss": 0.0134, "step": 16235 }, { "epoch": 6.4, "learning_rate": 1.2125146401735308e-07, "loss": 0.0127, "step": 16236 }, { "epoch": 6.4, "learning_rate": 1.2122799410478818e-07, "loss": 0.0119, "step": 16237 }, { "epoch": 6.4, "learning_rate": 1.2120452547608726e-07, "loss": 0.0117, "step": 16238 }, { "epoch": 6.4, "learning_rate": 1.2118105813163284e-07, "loss": 0.0145, "step": 16239 }, { "epoch": 6.4, "learning_rate": 1.2115759207180738e-07, "loss": 0.0113, "step": 16240 }, { "epoch": 6.4, "learning_rate": 1.211341272969933e-07, "loss": 0.0113, "step": 16241 }, { "epoch": 6.4, "learning_rate": 1.2111066380757312e-07, "loss": 0.0116, "step": 16242 }, { "epoch": 6.4, "learning_rate": 1.210872016039292e-07, "loss": 0.012, "step": 16243 }, { "epoch": 6.4, "learning_rate": 1.21063740686444e-07, "loss": 0.0149, "step": 16244 }, { "epoch": 6.4, "learning_rate": 1.2104028105549984e-07, "loss": 0.0116, "step": 16245 }, { "epoch": 6.4, "learning_rate": 1.210168227114791e-07, "loss": 0.0125, "step": 16246 }, { "epoch": 6.4, "learning_rate": 1.2099336565476404e-07, "loss": 0.012, "step": 16247 }, { "epoch": 6.4, "learning_rate": 1.2096990988573713e-07, "loss": 0.0084, "step": 16248 }, { "epoch": 6.4, "learning_rate": 1.2094645540478054e-07, "loss": 0.0114, "step": 16249 }, { "epoch": 6.4, "learning_rate": 1.209230022122766e-07, "loss": 0.0154, "step": 16250 }, { "epoch": 6.4, "learning_rate": 1.2089955030860747e-07, "loss": 0.0126, "step": 16251 }, { "epoch": 6.41, "learning_rate": 1.2087609969415554e-07, "loss": 0.0128, "step": 16252 }, { "epoch": 6.41, "learning_rate": 1.208526503693029e-07, "loss": 0.0169, "step": 16253 }, { "epoch": 6.41, "learning_rate": 1.2082920233443176e-07, "loss": 0.0121, "step": 16254 }, { "epoch": 6.41, "learning_rate": 1.2080575558992434e-07, "loss": 0.0135, "step": 16255 }, { "epoch": 6.41, "learning_rate": 1.2078231013616265e-07, "loss": 0.013, "step": 16256 }, { "epoch": 6.41, "learning_rate": 1.2075886597352897e-07, "loss": 0.0113, "step": 16257 }, { "epoch": 6.41, "learning_rate": 1.2073542310240536e-07, "loss": 0.0115, "step": 16258 }, { "epoch": 6.41, "learning_rate": 1.2071198152317386e-07, "loss": 0.0109, "step": 16259 }, { "epoch": 6.41, "learning_rate": 1.2068854123621656e-07, "loss": 0.0142, "step": 16260 }, { "epoch": 6.41, "learning_rate": 1.206651022419155e-07, "loss": 0.0148, "step": 16261 }, { "epoch": 6.41, "learning_rate": 1.206416645406527e-07, "loss": 0.0117, "step": 16262 }, { "epoch": 6.41, "learning_rate": 1.2061822813281014e-07, "loss": 0.0118, "step": 16263 }, { "epoch": 6.41, "learning_rate": 1.2059479301876982e-07, "loss": 0.0105, "step": 16264 }, { "epoch": 6.41, "learning_rate": 1.2057135919891368e-07, "loss": 0.0117, "step": 16265 }, { "epoch": 6.41, "learning_rate": 1.205479266736237e-07, "loss": 0.0109, "step": 16266 }, { "epoch": 6.41, "learning_rate": 1.2052449544328177e-07, "loss": 0.0109, "step": 16267 }, { "epoch": 6.41, "learning_rate": 1.2050106550826977e-07, "loss": 0.0124, "step": 16268 }, { "epoch": 6.41, "learning_rate": 1.204776368689695e-07, "loss": 0.0114, "step": 16269 }, { "epoch": 6.41, "learning_rate": 1.2045420952576298e-07, "loss": 0.0126, "step": 16270 }, { "epoch": 6.41, "learning_rate": 1.204307834790319e-07, "loss": 0.0105, "step": 16271 }, { "epoch": 6.41, "learning_rate": 1.2040735872915818e-07, "loss": 0.0105, "step": 16272 }, { "epoch": 6.41, "learning_rate": 1.2038393527652352e-07, "loss": 0.0157, "step": 16273 }, { "epoch": 6.41, "learning_rate": 1.203605131215096e-07, "loss": 0.0143, "step": 16274 }, { "epoch": 6.41, "learning_rate": 1.203370922644984e-07, "loss": 0.01, "step": 16275 }, { "epoch": 6.41, "learning_rate": 1.2031367270587152e-07, "loss": 0.0094, "step": 16276 }, { "epoch": 6.41, "learning_rate": 1.2029025444601067e-07, "loss": 0.0122, "step": 16277 }, { "epoch": 6.42, "learning_rate": 1.2026683748529743e-07, "loss": 0.0123, "step": 16278 }, { "epoch": 6.42, "learning_rate": 1.2024342182411366e-07, "loss": 0.0124, "step": 16279 }, { "epoch": 6.42, "learning_rate": 1.2022000746284088e-07, "loss": 0.0123, "step": 16280 }, { "epoch": 6.42, "learning_rate": 1.201965944018607e-07, "loss": 0.0131, "step": 16281 }, { "epoch": 6.42, "learning_rate": 1.2017318264155477e-07, "loss": 0.0088, "step": 16282 }, { "epoch": 6.42, "learning_rate": 1.201497721823046e-07, "loss": 0.0139, "step": 16283 }, { "epoch": 6.42, "learning_rate": 1.2012636302449184e-07, "loss": 0.0137, "step": 16284 }, { "epoch": 6.42, "learning_rate": 1.2010295516849796e-07, "loss": 0.0155, "step": 16285 }, { "epoch": 6.42, "learning_rate": 1.200795486147044e-07, "loss": 0.0093, "step": 16286 }, { "epoch": 6.42, "learning_rate": 1.200561433634928e-07, "loss": 0.0107, "step": 16287 }, { "epoch": 6.42, "learning_rate": 1.200327394152446e-07, "loss": 0.012, "step": 16288 }, { "epoch": 6.42, "learning_rate": 1.2000933677034116e-07, "loss": 0.011, "step": 16289 }, { "epoch": 6.42, "learning_rate": 1.19985935429164e-07, "loss": 0.008, "step": 16290 }, { "epoch": 6.42, "learning_rate": 1.1996253539209436e-07, "loss": 0.0122, "step": 16291 }, { "epoch": 6.42, "learning_rate": 1.1993913665951387e-07, "loss": 0.012, "step": 16292 }, { "epoch": 6.42, "learning_rate": 1.1991573923180375e-07, "loss": 0.012, "step": 16293 }, { "epoch": 6.42, "learning_rate": 1.198923431093454e-07, "loss": 0.0116, "step": 16294 }, { "epoch": 6.42, "learning_rate": 1.1986894829252007e-07, "loss": 0.0111, "step": 16295 }, { "epoch": 6.42, "learning_rate": 1.1984555478170902e-07, "loss": 0.0112, "step": 16296 }, { "epoch": 6.42, "learning_rate": 1.1982216257729372e-07, "loss": 0.0173, "step": 16297 }, { "epoch": 6.42, "learning_rate": 1.197987716796553e-07, "loss": 0.0134, "step": 16298 }, { "epoch": 6.42, "learning_rate": 1.19775382089175e-07, "loss": 0.0121, "step": 16299 }, { "epoch": 6.42, "learning_rate": 1.1975199380623396e-07, "loss": 0.0129, "step": 16300 }, { "epoch": 6.42, "learning_rate": 1.1972860683121355e-07, "loss": 0.0149, "step": 16301 }, { "epoch": 6.42, "learning_rate": 1.1970522116449484e-07, "loss": 0.012, "step": 16302 }, { "epoch": 6.43, "learning_rate": 1.19681836806459e-07, "loss": 0.0144, "step": 16303 }, { "epoch": 6.43, "learning_rate": 1.1965845375748717e-07, "loss": 0.0122, "step": 16304 }, { "epoch": 6.43, "learning_rate": 1.1963507201796038e-07, "loss": 0.015, "step": 16305 }, { "epoch": 6.43, "learning_rate": 1.1961169158825982e-07, "loss": 0.013, "step": 16306 }, { "epoch": 6.43, "learning_rate": 1.1958831246876647e-07, "loss": 0.0114, "step": 16307 }, { "epoch": 6.43, "learning_rate": 1.1956493465986145e-07, "loss": 0.015, "step": 16308 }, { "epoch": 6.43, "learning_rate": 1.1954155816192568e-07, "loss": 0.0105, "step": 16309 }, { "epoch": 6.43, "learning_rate": 1.1951818297534028e-07, "loss": 0.0154, "step": 16310 }, { "epoch": 6.43, "learning_rate": 1.1949480910048619e-07, "loss": 0.0119, "step": 16311 }, { "epoch": 6.43, "learning_rate": 1.1947143653774436e-07, "loss": 0.0099, "step": 16312 }, { "epoch": 6.43, "learning_rate": 1.1944806528749576e-07, "loss": 0.0113, "step": 16313 }, { "epoch": 6.43, "learning_rate": 1.1942469535012115e-07, "loss": 0.0086, "step": 16314 }, { "epoch": 6.43, "learning_rate": 1.1940132672600162e-07, "loss": 0.0105, "step": 16315 }, { "epoch": 6.43, "learning_rate": 1.1937795941551798e-07, "loss": 0.011, "step": 16316 }, { "epoch": 6.43, "learning_rate": 1.193545934190511e-07, "loss": 0.0108, "step": 16317 }, { "epoch": 6.43, "learning_rate": 1.1933122873698164e-07, "loss": 0.0107, "step": 16318 }, { "epoch": 6.43, "learning_rate": 1.1930786536969067e-07, "loss": 0.0108, "step": 16319 }, { "epoch": 6.43, "learning_rate": 1.1928450331755887e-07, "loss": 0.0121, "step": 16320 }, { "epoch": 6.43, "learning_rate": 1.1926114258096697e-07, "loss": 0.0132, "step": 16321 }, { "epoch": 6.43, "learning_rate": 1.1923778316029572e-07, "loss": 0.0106, "step": 16322 }, { "epoch": 6.43, "learning_rate": 1.192144250559258e-07, "loss": 0.0108, "step": 16323 }, { "epoch": 6.43, "learning_rate": 1.1919106826823804e-07, "loss": 0.0142, "step": 16324 }, { "epoch": 6.43, "learning_rate": 1.1916771279761309e-07, "loss": 0.0113, "step": 16325 }, { "epoch": 6.43, "learning_rate": 1.1914435864443153e-07, "loss": 0.012, "step": 16326 }, { "epoch": 6.43, "learning_rate": 1.19121005809074e-07, "loss": 0.0144, "step": 16327 }, { "epoch": 6.43, "learning_rate": 1.1909765429192122e-07, "loss": 0.0118, "step": 16328 }, { "epoch": 6.44, "learning_rate": 1.1907430409335369e-07, "loss": 0.0096, "step": 16329 }, { "epoch": 6.44, "learning_rate": 1.1905095521375201e-07, "loss": 0.0112, "step": 16330 }, { "epoch": 6.44, "learning_rate": 1.1902760765349667e-07, "loss": 0.0131, "step": 16331 }, { "epoch": 6.44, "learning_rate": 1.1900426141296835e-07, "loss": 0.0115, "step": 16332 }, { "epoch": 6.44, "learning_rate": 1.1898091649254745e-07, "loss": 0.0118, "step": 16333 }, { "epoch": 6.44, "learning_rate": 1.1895757289261447e-07, "loss": 0.0109, "step": 16334 }, { "epoch": 6.44, "learning_rate": 1.189342306135499e-07, "loss": 0.0116, "step": 16335 }, { "epoch": 6.44, "learning_rate": 1.1891088965573408e-07, "loss": 0.0114, "step": 16336 }, { "epoch": 6.44, "learning_rate": 1.188875500195476e-07, "loss": 0.0137, "step": 16337 }, { "epoch": 6.44, "learning_rate": 1.1886421170537077e-07, "loss": 0.0105, "step": 16338 }, { "epoch": 6.44, "learning_rate": 1.1884087471358397e-07, "loss": 0.011, "step": 16339 }, { "epoch": 6.44, "learning_rate": 1.188175390445675e-07, "loss": 0.0103, "step": 16340 }, { "epoch": 6.44, "learning_rate": 1.1879420469870183e-07, "loss": 0.0124, "step": 16341 }, { "epoch": 6.44, "learning_rate": 1.1877087167636721e-07, "loss": 0.0161, "step": 16342 }, { "epoch": 6.44, "learning_rate": 1.1874753997794393e-07, "loss": 0.0125, "step": 16343 }, { "epoch": 6.44, "learning_rate": 1.1872420960381227e-07, "loss": 0.0158, "step": 16344 }, { "epoch": 6.44, "learning_rate": 1.1870088055435239e-07, "loss": 0.0112, "step": 16345 }, { "epoch": 6.44, "learning_rate": 1.1867755282994463e-07, "loss": 0.0156, "step": 16346 }, { "epoch": 6.44, "learning_rate": 1.1865422643096921e-07, "loss": 0.0125, "step": 16347 }, { "epoch": 6.44, "learning_rate": 1.1863090135780622e-07, "loss": 0.014, "step": 16348 }, { "epoch": 6.44, "learning_rate": 1.1860757761083586e-07, "loss": 0.0092, "step": 16349 }, { "epoch": 6.44, "learning_rate": 1.1858425519043835e-07, "loss": 0.0121, "step": 16350 }, { "epoch": 6.44, "learning_rate": 1.185609340969937e-07, "loss": 0.0122, "step": 16351 }, { "epoch": 6.44, "learning_rate": 1.185376143308821e-07, "loss": 0.0137, "step": 16352 }, { "epoch": 6.44, "learning_rate": 1.1851429589248354e-07, "loss": 0.0145, "step": 16353 }, { "epoch": 6.45, "learning_rate": 1.1849097878217804e-07, "loss": 0.0117, "step": 16354 }, { "epoch": 6.45, "learning_rate": 1.1846766300034578e-07, "loss": 0.0105, "step": 16355 }, { "epoch": 6.45, "learning_rate": 1.184443485473667e-07, "loss": 0.0115, "step": 16356 }, { "epoch": 6.45, "learning_rate": 1.1842103542362081e-07, "loss": 0.0118, "step": 16357 }, { "epoch": 6.45, "learning_rate": 1.1839772362948794e-07, "loss": 0.0157, "step": 16358 }, { "epoch": 6.45, "learning_rate": 1.1837441316534826e-07, "loss": 0.0109, "step": 16359 }, { "epoch": 6.45, "learning_rate": 1.1835110403158158e-07, "loss": 0.0129, "step": 16360 }, { "epoch": 6.45, "learning_rate": 1.183277962285678e-07, "loss": 0.0117, "step": 16361 }, { "epoch": 6.45, "learning_rate": 1.1830448975668685e-07, "loss": 0.0099, "step": 16362 }, { "epoch": 6.45, "learning_rate": 1.1828118461631844e-07, "loss": 0.0119, "step": 16363 }, { "epoch": 6.45, "learning_rate": 1.1825788080784263e-07, "loss": 0.0113, "step": 16364 }, { "epoch": 6.45, "learning_rate": 1.182345783316391e-07, "loss": 0.0126, "step": 16365 }, { "epoch": 6.45, "learning_rate": 1.182112771880877e-07, "loss": 0.011, "step": 16366 }, { "epoch": 6.45, "learning_rate": 1.1818797737756812e-07, "loss": 0.0143, "step": 16367 }, { "epoch": 6.45, "learning_rate": 1.1816467890046022e-07, "loss": 0.0151, "step": 16368 }, { "epoch": 6.45, "learning_rate": 1.1814138175714363e-07, "loss": 0.0118, "step": 16369 }, { "epoch": 6.45, "learning_rate": 1.1811808594799815e-07, "loss": 0.0114, "step": 16370 }, { "epoch": 6.45, "learning_rate": 1.180947914734034e-07, "loss": 0.0171, "step": 16371 }, { "epoch": 6.45, "learning_rate": 1.1807149833373906e-07, "loss": 0.0144, "step": 16372 }, { "epoch": 6.45, "learning_rate": 1.1804820652938481e-07, "loss": 0.011, "step": 16373 }, { "epoch": 6.45, "learning_rate": 1.1802491606072024e-07, "loss": 0.0126, "step": 16374 }, { "epoch": 6.45, "learning_rate": 1.1800162692812496e-07, "loss": 0.0112, "step": 16375 }, { "epoch": 6.45, "learning_rate": 1.1797833913197845e-07, "loss": 0.0106, "step": 16376 }, { "epoch": 6.45, "learning_rate": 1.1795505267266043e-07, "loss": 0.0127, "step": 16377 }, { "epoch": 6.45, "learning_rate": 1.1793176755055036e-07, "loss": 0.0118, "step": 16378 }, { "epoch": 6.46, "learning_rate": 1.1790848376602775e-07, "loss": 0.0119, "step": 16379 }, { "epoch": 6.46, "learning_rate": 1.1788520131947204e-07, "loss": 0.011, "step": 16380 }, { "epoch": 6.46, "learning_rate": 1.1786192021126278e-07, "loss": 0.0108, "step": 16381 }, { "epoch": 6.46, "learning_rate": 1.178386404417794e-07, "loss": 0.0153, "step": 16382 }, { "epoch": 6.46, "learning_rate": 1.1781536201140132e-07, "loss": 0.0128, "step": 16383 }, { "epoch": 6.46, "learning_rate": 1.1779208492050793e-07, "loss": 0.0142, "step": 16384 }, { "epoch": 6.46, "learning_rate": 1.1776880916947853e-07, "loss": 0.0106, "step": 16385 }, { "epoch": 6.46, "learning_rate": 1.1774553475869264e-07, "loss": 0.011, "step": 16386 }, { "epoch": 6.46, "learning_rate": 1.1772226168852952e-07, "loss": 0.0131, "step": 16387 }, { "epoch": 6.46, "learning_rate": 1.1769898995936851e-07, "loss": 0.012, "step": 16388 }, { "epoch": 6.46, "learning_rate": 1.1767571957158882e-07, "loss": 0.0123, "step": 16389 }, { "epoch": 6.46, "learning_rate": 1.1765245052556983e-07, "loss": 0.013, "step": 16390 }, { "epoch": 6.46, "learning_rate": 1.176291828216907e-07, "loss": 0.012, "step": 16391 }, { "epoch": 6.46, "learning_rate": 1.1760591646033074e-07, "loss": 0.0119, "step": 16392 }, { "epoch": 6.46, "learning_rate": 1.175826514418691e-07, "loss": 0.0088, "step": 16393 }, { "epoch": 6.46, "learning_rate": 1.1755938776668495e-07, "loss": 0.0116, "step": 16394 }, { "epoch": 6.46, "learning_rate": 1.1753612543515754e-07, "loss": 0.012, "step": 16395 }, { "epoch": 6.46, "learning_rate": 1.1751286444766595e-07, "loss": 0.0114, "step": 16396 }, { "epoch": 6.46, "learning_rate": 1.174896048045893e-07, "loss": 0.01, "step": 16397 }, { "epoch": 6.46, "learning_rate": 1.174663465063066e-07, "loss": 0.0112, "step": 16398 }, { "epoch": 6.46, "learning_rate": 1.174430895531971e-07, "loss": 0.0111, "step": 16399 }, { "epoch": 6.46, "learning_rate": 1.1741983394563976e-07, "loss": 0.0124, "step": 16400 }, { "epoch": 6.46, "learning_rate": 1.1739657968401365e-07, "loss": 0.0208, "step": 16401 }, { "epoch": 6.46, "learning_rate": 1.1737332676869773e-07, "loss": 0.0124, "step": 16402 }, { "epoch": 6.46, "learning_rate": 1.1735007520007094e-07, "loss": 0.0103, "step": 16403 }, { "epoch": 6.46, "learning_rate": 1.1732682497851238e-07, "loss": 0.0115, "step": 16404 }, { "epoch": 6.47, "learning_rate": 1.1730357610440091e-07, "loss": 0.0113, "step": 16405 }, { "epoch": 6.47, "learning_rate": 1.1728032857811547e-07, "loss": 0.013, "step": 16406 }, { "epoch": 6.47, "learning_rate": 1.1725708240003489e-07, "loss": 0.0123, "step": 16407 }, { "epoch": 6.47, "learning_rate": 1.1723383757053817e-07, "loss": 0.0136, "step": 16408 }, { "epoch": 6.47, "learning_rate": 1.1721059409000413e-07, "loss": 0.0143, "step": 16409 }, { "epoch": 6.47, "learning_rate": 1.1718735195881158e-07, "loss": 0.0096, "step": 16410 }, { "epoch": 6.47, "learning_rate": 1.1716411117733929e-07, "loss": 0.0148, "step": 16411 }, { "epoch": 6.47, "learning_rate": 1.1714087174596611e-07, "loss": 0.0126, "step": 16412 }, { "epoch": 6.47, "learning_rate": 1.1711763366507081e-07, "loss": 0.0134, "step": 16413 }, { "epoch": 6.47, "learning_rate": 1.1709439693503207e-07, "loss": 0.0102, "step": 16414 }, { "epoch": 6.47, "learning_rate": 1.1707116155622871e-07, "loss": 0.0113, "step": 16415 }, { "epoch": 6.47, "learning_rate": 1.1704792752903934e-07, "loss": 0.0112, "step": 16416 }, { "epoch": 6.47, "learning_rate": 1.170246948538427e-07, "loss": 0.0117, "step": 16417 }, { "epoch": 6.47, "learning_rate": 1.1700146353101742e-07, "loss": 0.0158, "step": 16418 }, { "epoch": 6.47, "learning_rate": 1.1697823356094216e-07, "loss": 0.0122, "step": 16419 }, { "epoch": 6.47, "learning_rate": 1.1695500494399543e-07, "loss": 0.016, "step": 16420 }, { "epoch": 6.47, "learning_rate": 1.1693177768055598e-07, "loss": 0.012, "step": 16421 }, { "epoch": 6.47, "learning_rate": 1.1690855177100232e-07, "loss": 0.0122, "step": 16422 }, { "epoch": 6.47, "learning_rate": 1.1688532721571295e-07, "loss": 0.0117, "step": 16423 }, { "epoch": 6.47, "learning_rate": 1.1686210401506644e-07, "loss": 0.0109, "step": 16424 }, { "epoch": 6.47, "learning_rate": 1.1683888216944123e-07, "loss": 0.0141, "step": 16425 }, { "epoch": 6.47, "learning_rate": 1.1681566167921589e-07, "loss": 0.0136, "step": 16426 }, { "epoch": 6.47, "learning_rate": 1.1679244254476888e-07, "loss": 0.0096, "step": 16427 }, { "epoch": 6.47, "learning_rate": 1.1676922476647859e-07, "loss": 0.01, "step": 16428 }, { "epoch": 6.47, "learning_rate": 1.167460083447233e-07, "loss": 0.0105, "step": 16429 }, { "epoch": 6.48, "learning_rate": 1.167227932798817e-07, "loss": 0.0089, "step": 16430 }, { "epoch": 6.48, "learning_rate": 1.1669957957233192e-07, "loss": 0.0113, "step": 16431 }, { "epoch": 6.48, "learning_rate": 1.1667636722245249e-07, "loss": 0.0133, "step": 16432 }, { "epoch": 6.48, "learning_rate": 1.1665315623062153e-07, "loss": 0.0142, "step": 16433 }, { "epoch": 6.48, "learning_rate": 1.1662994659721748e-07, "loss": 0.0134, "step": 16434 }, { "epoch": 6.48, "learning_rate": 1.1660673832261858e-07, "loss": 0.011, "step": 16435 }, { "epoch": 6.48, "learning_rate": 1.1658353140720319e-07, "loss": 0.0139, "step": 16436 }, { "epoch": 6.48, "learning_rate": 1.1656032585134935e-07, "loss": 0.0136, "step": 16437 }, { "epoch": 6.48, "learning_rate": 1.1653712165543541e-07, "loss": 0.0153, "step": 16438 }, { "epoch": 6.48, "learning_rate": 1.1651391881983954e-07, "loss": 0.0117, "step": 16439 }, { "epoch": 6.48, "learning_rate": 1.1649071734493995e-07, "loss": 0.0122, "step": 16440 }, { "epoch": 6.48, "learning_rate": 1.1646751723111472e-07, "loss": 0.0142, "step": 16441 }, { "epoch": 6.48, "learning_rate": 1.1644431847874208e-07, "loss": 0.0098, "step": 16442 }, { "epoch": 6.48, "learning_rate": 1.1642112108819995e-07, "loss": 0.0115, "step": 16443 }, { "epoch": 6.48, "learning_rate": 1.1639792505986652e-07, "loss": 0.0119, "step": 16444 }, { "epoch": 6.48, "learning_rate": 1.1637473039411987e-07, "loss": 0.0082, "step": 16445 }, { "epoch": 6.48, "learning_rate": 1.1635153709133811e-07, "loss": 0.0108, "step": 16446 }, { "epoch": 6.48, "learning_rate": 1.1632834515189904e-07, "loss": 0.0121, "step": 16447 }, { "epoch": 6.48, "learning_rate": 1.1630515457618082e-07, "loss": 0.011, "step": 16448 }, { "epoch": 6.48, "learning_rate": 1.1628196536456145e-07, "loss": 0.0112, "step": 16449 }, { "epoch": 6.48, "learning_rate": 1.1625877751741873e-07, "loss": 0.0126, "step": 16450 }, { "epoch": 6.48, "learning_rate": 1.1623559103513066e-07, "loss": 0.0136, "step": 16451 }, { "epoch": 6.48, "learning_rate": 1.1621240591807513e-07, "loss": 0.0093, "step": 16452 }, { "epoch": 6.48, "learning_rate": 1.1618922216663014e-07, "loss": 0.0127, "step": 16453 }, { "epoch": 6.48, "learning_rate": 1.1616603978117336e-07, "loss": 0.0118, "step": 16454 }, { "epoch": 6.49, "learning_rate": 1.161428587620828e-07, "loss": 0.0111, "step": 16455 }, { "epoch": 6.49, "learning_rate": 1.1611967910973604e-07, "loss": 0.0093, "step": 16456 }, { "epoch": 6.49, "learning_rate": 1.1609650082451121e-07, "loss": 0.0131, "step": 16457 }, { "epoch": 6.49, "learning_rate": 1.1607332390678579e-07, "loss": 0.0113, "step": 16458 }, { "epoch": 6.49, "learning_rate": 1.1605014835693772e-07, "loss": 0.0118, "step": 16459 }, { "epoch": 6.49, "learning_rate": 1.1602697417534448e-07, "loss": 0.0097, "step": 16460 }, { "epoch": 6.49, "learning_rate": 1.160038013623841e-07, "loss": 0.015, "step": 16461 }, { "epoch": 6.49, "learning_rate": 1.1598062991843403e-07, "loss": 0.0159, "step": 16462 }, { "epoch": 6.49, "learning_rate": 1.1595745984387207e-07, "loss": 0.0109, "step": 16463 }, { "epoch": 6.49, "learning_rate": 1.1593429113907568e-07, "loss": 0.0107, "step": 16464 }, { "epoch": 6.49, "learning_rate": 1.159111238044226e-07, "loss": 0.0096, "step": 16465 }, { "epoch": 6.49, "learning_rate": 1.1588795784029049e-07, "loss": 0.0124, "step": 16466 }, { "epoch": 6.49, "learning_rate": 1.1586479324705676e-07, "loss": 0.0119, "step": 16467 }, { "epoch": 6.49, "learning_rate": 1.1584163002509901e-07, "loss": 0.0124, "step": 16468 }, { "epoch": 6.49, "learning_rate": 1.1581846817479477e-07, "loss": 0.0111, "step": 16469 }, { "epoch": 6.49, "learning_rate": 1.1579530769652167e-07, "loss": 0.0089, "step": 16470 }, { "epoch": 6.49, "learning_rate": 1.1577214859065698e-07, "loss": 0.0114, "step": 16471 }, { "epoch": 6.49, "learning_rate": 1.1574899085757834e-07, "loss": 0.0101, "step": 16472 }, { "epoch": 6.49, "learning_rate": 1.1572583449766302e-07, "loss": 0.0109, "step": 16473 }, { "epoch": 6.49, "learning_rate": 1.1570267951128852e-07, "loss": 0.0087, "step": 16474 }, { "epoch": 6.49, "learning_rate": 1.1567952589883221e-07, "loss": 0.015, "step": 16475 }, { "epoch": 6.49, "learning_rate": 1.1565637366067158e-07, "loss": 0.0098, "step": 16476 }, { "epoch": 6.49, "learning_rate": 1.1563322279718377e-07, "loss": 0.012, "step": 16477 }, { "epoch": 6.49, "learning_rate": 1.1561007330874624e-07, "loss": 0.0117, "step": 16478 }, { "epoch": 6.49, "learning_rate": 1.1558692519573623e-07, "loss": 0.0115, "step": 16479 }, { "epoch": 6.49, "learning_rate": 1.1556377845853114e-07, "loss": 0.011, "step": 16480 }, { "epoch": 6.5, "learning_rate": 1.1554063309750803e-07, "loss": 0.0099, "step": 16481 }, { "epoch": 6.5, "learning_rate": 1.1551748911304433e-07, "loss": 0.0115, "step": 16482 }, { "epoch": 6.5, "learning_rate": 1.1549434650551702e-07, "loss": 0.0125, "step": 16483 }, { "epoch": 6.5, "learning_rate": 1.1547120527530357e-07, "loss": 0.0138, "step": 16484 }, { "epoch": 6.5, "learning_rate": 1.1544806542278092e-07, "loss": 0.0111, "step": 16485 }, { "epoch": 6.5, "learning_rate": 1.1542492694832637e-07, "loss": 0.0112, "step": 16486 }, { "epoch": 6.5, "learning_rate": 1.154017898523169e-07, "loss": 0.0126, "step": 16487 }, { "epoch": 6.5, "learning_rate": 1.1537865413512969e-07, "loss": 0.0138, "step": 16488 }, { "epoch": 6.5, "learning_rate": 1.153555197971418e-07, "loss": 0.011, "step": 16489 }, { "epoch": 6.5, "learning_rate": 1.1533238683873037e-07, "loss": 0.0105, "step": 16490 }, { "epoch": 6.5, "learning_rate": 1.1530925526027229e-07, "loss": 0.0113, "step": 16491 }, { "epoch": 6.5, "learning_rate": 1.1528612506214462e-07, "loss": 0.0124, "step": 16492 }, { "epoch": 6.5, "learning_rate": 1.1526299624472444e-07, "loss": 0.0132, "step": 16493 }, { "epoch": 6.5, "learning_rate": 1.1523986880838857e-07, "loss": 0.0117, "step": 16494 }, { "epoch": 6.5, "learning_rate": 1.152167427535141e-07, "loss": 0.0108, "step": 16495 }, { "epoch": 6.5, "learning_rate": 1.151936180804777e-07, "loss": 0.0117, "step": 16496 }, { "epoch": 6.5, "learning_rate": 1.1517049478965657e-07, "loss": 0.0108, "step": 16497 }, { "epoch": 6.5, "learning_rate": 1.151473728814274e-07, "loss": 0.0142, "step": 16498 }, { "epoch": 6.5, "learning_rate": 1.1512425235616716e-07, "loss": 0.0135, "step": 16499 }, { "epoch": 6.5, "learning_rate": 1.1510113321425246e-07, "loss": 0.0115, "step": 16500 }, { "epoch": 6.5, "learning_rate": 1.150780154560604e-07, "loss": 0.011, "step": 16501 }, { "epoch": 6.5, "learning_rate": 1.1505489908196752e-07, "loss": 0.0117, "step": 16502 }, { "epoch": 6.5, "learning_rate": 1.1503178409235082e-07, "loss": 0.013, "step": 16503 }, { "epoch": 6.5, "learning_rate": 1.1500867048758678e-07, "loss": 0.0089, "step": 16504 }, { "epoch": 6.5, "learning_rate": 1.1498555826805225e-07, "loss": 0.0105, "step": 16505 }, { "epoch": 6.51, "learning_rate": 1.1496244743412393e-07, "loss": 0.0139, "step": 16506 }, { "epoch": 6.51, "learning_rate": 1.1493933798617854e-07, "loss": 0.0103, "step": 16507 }, { "epoch": 6.51, "learning_rate": 1.1491622992459258e-07, "loss": 0.0103, "step": 16508 }, { "epoch": 6.51, "learning_rate": 1.1489312324974276e-07, "loss": 0.0136, "step": 16509 }, { "epoch": 6.51, "learning_rate": 1.1487001796200578e-07, "loss": 0.0109, "step": 16510 }, { "epoch": 6.51, "learning_rate": 1.1484691406175803e-07, "loss": 0.0133, "step": 16511 }, { "epoch": 6.51, "learning_rate": 1.1482381154937617e-07, "loss": 0.0113, "step": 16512 }, { "epoch": 6.51, "learning_rate": 1.1480071042523683e-07, "loss": 0.0101, "step": 16513 }, { "epoch": 6.51, "learning_rate": 1.147776106897163e-07, "loss": 0.0138, "step": 16514 }, { "epoch": 6.51, "learning_rate": 1.1475451234319123e-07, "loss": 0.0127, "step": 16515 }, { "epoch": 6.51, "learning_rate": 1.147314153860381e-07, "loss": 0.0129, "step": 16516 }, { "epoch": 6.51, "learning_rate": 1.1470831981863322e-07, "loss": 0.0106, "step": 16517 }, { "epoch": 6.51, "learning_rate": 1.1468522564135312e-07, "loss": 0.01, "step": 16518 }, { "epoch": 6.51, "learning_rate": 1.1466213285457417e-07, "loss": 0.0119, "step": 16519 }, { "epoch": 6.51, "learning_rate": 1.1463904145867284e-07, "loss": 0.0116, "step": 16520 }, { "epoch": 6.51, "learning_rate": 1.1461595145402534e-07, "loss": 0.0127, "step": 16521 }, { "epoch": 6.51, "learning_rate": 1.1459286284100811e-07, "loss": 0.0125, "step": 16522 }, { "epoch": 6.51, "learning_rate": 1.1456977561999726e-07, "loss": 0.0114, "step": 16523 }, { "epoch": 6.51, "learning_rate": 1.1454668979136939e-07, "loss": 0.013, "step": 16524 }, { "epoch": 6.51, "learning_rate": 1.1452360535550055e-07, "loss": 0.0116, "step": 16525 }, { "epoch": 6.51, "learning_rate": 1.1450052231276708e-07, "loss": 0.0179, "step": 16526 }, { "epoch": 6.51, "learning_rate": 1.1447744066354501e-07, "loss": 0.0113, "step": 16527 }, { "epoch": 6.51, "learning_rate": 1.1445436040821081e-07, "loss": 0.0133, "step": 16528 }, { "epoch": 6.51, "learning_rate": 1.1443128154714048e-07, "loss": 0.0107, "step": 16529 }, { "epoch": 6.51, "learning_rate": 1.1440820408071026e-07, "loss": 0.0116, "step": 16530 }, { "epoch": 6.52, "learning_rate": 1.1438512800929616e-07, "loss": 0.0103, "step": 16531 }, { "epoch": 6.52, "learning_rate": 1.1436205333327436e-07, "loss": 0.0102, "step": 16532 }, { "epoch": 6.52, "learning_rate": 1.14338980053021e-07, "loss": 0.0118, "step": 16533 }, { "epoch": 6.52, "learning_rate": 1.1431590816891201e-07, "loss": 0.0124, "step": 16534 }, { "epoch": 6.52, "learning_rate": 1.1429283768132348e-07, "loss": 0.0126, "step": 16535 }, { "epoch": 6.52, "learning_rate": 1.1426976859063144e-07, "loss": 0.0127, "step": 16536 }, { "epoch": 6.52, "learning_rate": 1.1424670089721197e-07, "loss": 0.0107, "step": 16537 }, { "epoch": 6.52, "learning_rate": 1.1422363460144087e-07, "loss": 0.0107, "step": 16538 }, { "epoch": 6.52, "learning_rate": 1.1420056970369421e-07, "loss": 0.011, "step": 16539 }, { "epoch": 6.52, "learning_rate": 1.1417750620434772e-07, "loss": 0.0116, "step": 16540 }, { "epoch": 6.52, "learning_rate": 1.1415444410377762e-07, "loss": 0.0132, "step": 16541 }, { "epoch": 6.52, "learning_rate": 1.141313834023595e-07, "loss": 0.0133, "step": 16542 }, { "epoch": 6.52, "learning_rate": 1.1410832410046943e-07, "loss": 0.0119, "step": 16543 }, { "epoch": 6.52, "learning_rate": 1.1408526619848302e-07, "loss": 0.014, "step": 16544 }, { "epoch": 6.52, "learning_rate": 1.1406220969677622e-07, "loss": 0.0107, "step": 16545 }, { "epoch": 6.52, "learning_rate": 1.1403915459572482e-07, "loss": 0.0134, "step": 16546 }, { "epoch": 6.52, "learning_rate": 1.1401610089570461e-07, "loss": 0.0122, "step": 16547 }, { "epoch": 6.52, "learning_rate": 1.139930485970912e-07, "loss": 0.0129, "step": 16548 }, { "epoch": 6.52, "learning_rate": 1.1396999770026037e-07, "loss": 0.0116, "step": 16549 }, { "epoch": 6.52, "learning_rate": 1.1394694820558783e-07, "loss": 0.0134, "step": 16550 }, { "epoch": 6.52, "learning_rate": 1.1392390011344932e-07, "loss": 0.0116, "step": 16551 }, { "epoch": 6.52, "learning_rate": 1.1390085342422035e-07, "loss": 0.0105, "step": 16552 }, { "epoch": 6.52, "learning_rate": 1.138778081382767e-07, "loss": 0.0108, "step": 16553 }, { "epoch": 6.52, "learning_rate": 1.138547642559938e-07, "loss": 0.0128, "step": 16554 }, { "epoch": 6.52, "learning_rate": 1.1383172177774731e-07, "loss": 0.01, "step": 16555 }, { "epoch": 6.52, "learning_rate": 1.1380868070391287e-07, "loss": 0.0123, "step": 16556 }, { "epoch": 6.53, "learning_rate": 1.1378564103486588e-07, "loss": 0.0138, "step": 16557 }, { "epoch": 6.53, "learning_rate": 1.137626027709819e-07, "loss": 0.01, "step": 16558 }, { "epoch": 6.53, "learning_rate": 1.1373956591263642e-07, "loss": 0.0128, "step": 16559 }, { "epoch": 6.53, "learning_rate": 1.1371653046020502e-07, "loss": 0.0115, "step": 16560 }, { "epoch": 6.53, "learning_rate": 1.1369349641406293e-07, "loss": 0.0132, "step": 16561 }, { "epoch": 6.53, "learning_rate": 1.1367046377458578e-07, "loss": 0.0106, "step": 16562 }, { "epoch": 6.53, "learning_rate": 1.136474325421487e-07, "loss": 0.015, "step": 16563 }, { "epoch": 6.53, "learning_rate": 1.1362440271712737e-07, "loss": 0.0088, "step": 16564 }, { "epoch": 6.53, "learning_rate": 1.1360137429989696e-07, "loss": 0.0127, "step": 16565 }, { "epoch": 6.53, "learning_rate": 1.1357834729083289e-07, "loss": 0.0109, "step": 16566 }, { "epoch": 6.53, "learning_rate": 1.1355532169031024e-07, "loss": 0.0149, "step": 16567 }, { "epoch": 6.53, "learning_rate": 1.1353229749870466e-07, "loss": 0.015, "step": 16568 }, { "epoch": 6.53, "learning_rate": 1.1350927471639109e-07, "loss": 0.0103, "step": 16569 }, { "epoch": 6.53, "learning_rate": 1.13486253343745e-07, "loss": 0.0089, "step": 16570 }, { "epoch": 6.53, "learning_rate": 1.1346323338114141e-07, "loss": 0.0122, "step": 16571 }, { "epoch": 6.53, "learning_rate": 1.1344021482895561e-07, "loss": 0.0144, "step": 16572 }, { "epoch": 6.53, "learning_rate": 1.1341719768756273e-07, "loss": 0.0126, "step": 16573 }, { "epoch": 6.53, "learning_rate": 1.1339418195733804e-07, "loss": 0.0121, "step": 16574 }, { "epoch": 6.53, "learning_rate": 1.1337116763865644e-07, "loss": 0.012, "step": 16575 }, { "epoch": 6.53, "learning_rate": 1.1334815473189318e-07, "loss": 0.013, "step": 16576 }, { "epoch": 6.53, "learning_rate": 1.1332514323742337e-07, "loss": 0.0121, "step": 16577 }, { "epoch": 6.53, "learning_rate": 1.1330213315562192e-07, "loss": 0.0114, "step": 16578 }, { "epoch": 6.53, "learning_rate": 1.1327912448686401e-07, "loss": 0.0142, "step": 16579 }, { "epoch": 6.53, "learning_rate": 1.1325611723152442e-07, "loss": 0.0135, "step": 16580 }, { "epoch": 6.53, "learning_rate": 1.1323311138997844e-07, "loss": 0.0143, "step": 16581 }, { "epoch": 6.54, "learning_rate": 1.1321010696260078e-07, "loss": 0.0132, "step": 16582 }, { "epoch": 6.54, "learning_rate": 1.1318710394976657e-07, "loss": 0.0134, "step": 16583 }, { "epoch": 6.54, "learning_rate": 1.1316410235185054e-07, "loss": 0.0117, "step": 16584 }, { "epoch": 6.54, "learning_rate": 1.1314110216922765e-07, "loss": 0.0123, "step": 16585 }, { "epoch": 6.54, "learning_rate": 1.1311810340227279e-07, "loss": 0.0096, "step": 16586 }, { "epoch": 6.54, "learning_rate": 1.130951060513609e-07, "loss": 0.0132, "step": 16587 }, { "epoch": 6.54, "learning_rate": 1.1307211011686662e-07, "loss": 0.0133, "step": 16588 }, { "epoch": 6.54, "learning_rate": 1.1304911559916481e-07, "loss": 0.0099, "step": 16589 }, { "epoch": 6.54, "learning_rate": 1.1302612249863026e-07, "loss": 0.0113, "step": 16590 }, { "epoch": 6.54, "learning_rate": 1.1300313081563783e-07, "loss": 0.0129, "step": 16591 }, { "epoch": 6.54, "learning_rate": 1.1298014055056206e-07, "loss": 0.0108, "step": 16592 }, { "epoch": 6.54, "learning_rate": 1.1295715170377781e-07, "loss": 0.014, "step": 16593 }, { "epoch": 6.54, "learning_rate": 1.1293416427565964e-07, "loss": 0.0124, "step": 16594 }, { "epoch": 6.54, "learning_rate": 1.1291117826658225e-07, "loss": 0.0109, "step": 16595 }, { "epoch": 6.54, "learning_rate": 1.1288819367692031e-07, "loss": 0.0142, "step": 16596 }, { "epoch": 6.54, "learning_rate": 1.1286521050704849e-07, "loss": 0.0102, "step": 16597 }, { "epoch": 6.54, "learning_rate": 1.1284222875734122e-07, "loss": 0.0171, "step": 16598 }, { "epoch": 6.54, "learning_rate": 1.1281924842817318e-07, "loss": 0.0098, "step": 16599 }, { "epoch": 6.54, "learning_rate": 1.1279626951991894e-07, "loss": 0.0102, "step": 16600 }, { "epoch": 6.54, "learning_rate": 1.1277329203295294e-07, "loss": 0.0105, "step": 16601 }, { "epoch": 6.54, "learning_rate": 1.1275031596764974e-07, "loss": 0.0136, "step": 16602 }, { "epoch": 6.54, "learning_rate": 1.1272734132438366e-07, "loss": 0.0126, "step": 16603 }, { "epoch": 6.54, "learning_rate": 1.1270436810352941e-07, "loss": 0.0116, "step": 16604 }, { "epoch": 6.54, "learning_rate": 1.1268139630546123e-07, "loss": 0.0141, "step": 16605 }, { "epoch": 6.54, "learning_rate": 1.1265842593055363e-07, "loss": 0.0122, "step": 16606 }, { "epoch": 6.54, "learning_rate": 1.1263545697918082e-07, "loss": 0.0116, "step": 16607 }, { "epoch": 6.55, "learning_rate": 1.1261248945171742e-07, "loss": 0.0144, "step": 16608 }, { "epoch": 6.55, "learning_rate": 1.1258952334853755e-07, "loss": 0.0107, "step": 16609 }, { "epoch": 6.55, "learning_rate": 1.125665586700157e-07, "loss": 0.0115, "step": 16610 }, { "epoch": 6.55, "learning_rate": 1.1254359541652588e-07, "loss": 0.0112, "step": 16611 }, { "epoch": 6.55, "learning_rate": 1.125206335884427e-07, "loss": 0.0132, "step": 16612 }, { "epoch": 6.55, "learning_rate": 1.1249767318614016e-07, "loss": 0.0133, "step": 16613 }, { "epoch": 6.55, "learning_rate": 1.1247471420999264e-07, "loss": 0.0126, "step": 16614 }, { "epoch": 6.55, "learning_rate": 1.1245175666037416e-07, "loss": 0.0111, "step": 16615 }, { "epoch": 6.55, "learning_rate": 1.1242880053765902e-07, "loss": 0.012, "step": 16616 }, { "epoch": 6.55, "learning_rate": 1.1240584584222132e-07, "loss": 0.0122, "step": 16617 }, { "epoch": 6.55, "learning_rate": 1.1238289257443527e-07, "loss": 0.0134, "step": 16618 }, { "epoch": 6.55, "learning_rate": 1.1235994073467488e-07, "loss": 0.0105, "step": 16619 }, { "epoch": 6.55, "learning_rate": 1.1233699032331421e-07, "loss": 0.0126, "step": 16620 }, { "epoch": 6.55, "learning_rate": 1.1231404134072743e-07, "loss": 0.018, "step": 16621 }, { "epoch": 6.55, "learning_rate": 1.1229109378728847e-07, "loss": 0.013, "step": 16622 }, { "epoch": 6.55, "learning_rate": 1.1226814766337147e-07, "loss": 0.0136, "step": 16623 }, { "epoch": 6.55, "learning_rate": 1.1224520296935022e-07, "loss": 0.0099, "step": 16624 }, { "epoch": 6.55, "learning_rate": 1.122222597055988e-07, "loss": 0.0101, "step": 16625 }, { "epoch": 6.55, "learning_rate": 1.121993178724911e-07, "loss": 0.0112, "step": 16626 }, { "epoch": 6.55, "learning_rate": 1.121763774704012e-07, "loss": 0.0085, "step": 16627 }, { "epoch": 6.55, "learning_rate": 1.121534384997028e-07, "loss": 0.0104, "step": 16628 }, { "epoch": 6.55, "learning_rate": 1.1213050096076982e-07, "loss": 0.0096, "step": 16629 }, { "epoch": 6.55, "learning_rate": 1.1210756485397612e-07, "loss": 0.0098, "step": 16630 }, { "epoch": 6.55, "learning_rate": 1.120846301796956e-07, "loss": 0.0108, "step": 16631 }, { "epoch": 6.55, "learning_rate": 1.1206169693830196e-07, "loss": 0.0135, "step": 16632 }, { "epoch": 6.56, "learning_rate": 1.1203876513016905e-07, "loss": 0.0124, "step": 16633 }, { "epoch": 6.56, "learning_rate": 1.1201583475567045e-07, "loss": 0.0113, "step": 16634 }, { "epoch": 6.56, "learning_rate": 1.1199290581518016e-07, "loss": 0.0128, "step": 16635 }, { "epoch": 6.56, "learning_rate": 1.1196997830907168e-07, "loss": 0.013, "step": 16636 }, { "epoch": 6.56, "learning_rate": 1.1194705223771884e-07, "loss": 0.0105, "step": 16637 }, { "epoch": 6.56, "learning_rate": 1.1192412760149514e-07, "loss": 0.0135, "step": 16638 }, { "epoch": 6.56, "learning_rate": 1.119012044007743e-07, "loss": 0.0124, "step": 16639 }, { "epoch": 6.56, "learning_rate": 1.1187828263592995e-07, "loss": 0.0116, "step": 16640 }, { "epoch": 6.56, "learning_rate": 1.1185536230733572e-07, "loss": 0.0108, "step": 16641 }, { "epoch": 6.56, "learning_rate": 1.1183244341536506e-07, "loss": 0.0125, "step": 16642 }, { "epoch": 6.56, "learning_rate": 1.1180952596039157e-07, "loss": 0.0145, "step": 16643 }, { "epoch": 6.56, "learning_rate": 1.1178660994278884e-07, "loss": 0.0104, "step": 16644 }, { "epoch": 6.56, "learning_rate": 1.1176369536293022e-07, "loss": 0.0095, "step": 16645 }, { "epoch": 6.56, "learning_rate": 1.1174078222118936e-07, "loss": 0.0092, "step": 16646 }, { "epoch": 6.56, "learning_rate": 1.1171787051793944e-07, "loss": 0.0127, "step": 16647 }, { "epoch": 6.56, "learning_rate": 1.1169496025355423e-07, "loss": 0.0132, "step": 16648 }, { "epoch": 6.56, "learning_rate": 1.1167205142840685e-07, "loss": 0.009, "step": 16649 }, { "epoch": 6.56, "learning_rate": 1.116491440428709e-07, "loss": 0.0121, "step": 16650 }, { "epoch": 6.56, "learning_rate": 1.1162623809731944e-07, "loss": 0.0101, "step": 16651 }, { "epoch": 6.56, "learning_rate": 1.1160333359212616e-07, "loss": 0.0142, "step": 16652 }, { "epoch": 6.56, "learning_rate": 1.115804305276641e-07, "loss": 0.0097, "step": 16653 }, { "epoch": 6.56, "learning_rate": 1.1155752890430674e-07, "loss": 0.0128, "step": 16654 }, { "epoch": 6.56, "learning_rate": 1.1153462872242716e-07, "loss": 0.012, "step": 16655 }, { "epoch": 6.56, "learning_rate": 1.1151172998239865e-07, "loss": 0.0121, "step": 16656 }, { "epoch": 6.56, "learning_rate": 1.114888326845945e-07, "loss": 0.0137, "step": 16657 }, { "epoch": 6.57, "learning_rate": 1.114659368293879e-07, "loss": 0.0105, "step": 16658 }, { "epoch": 6.57, "learning_rate": 1.1144304241715193e-07, "loss": 0.0115, "step": 16659 }, { "epoch": 6.57, "learning_rate": 1.1142014944825975e-07, "loss": 0.011, "step": 16660 }, { "epoch": 6.57, "learning_rate": 1.1139725792308459e-07, "loss": 0.0112, "step": 16661 }, { "epoch": 6.57, "learning_rate": 1.1137436784199938e-07, "loss": 0.0132, "step": 16662 }, { "epoch": 6.57, "learning_rate": 1.1135147920537728e-07, "loss": 0.0108, "step": 16663 }, { "epoch": 6.57, "learning_rate": 1.1132859201359144e-07, "loss": 0.0126, "step": 16664 }, { "epoch": 6.57, "learning_rate": 1.1130570626701468e-07, "loss": 0.015, "step": 16665 }, { "epoch": 6.57, "learning_rate": 1.112828219660201e-07, "loss": 0.0105, "step": 16666 }, { "epoch": 6.57, "learning_rate": 1.1125993911098079e-07, "loss": 0.0122, "step": 16667 }, { "epoch": 6.57, "learning_rate": 1.1123705770226952e-07, "loss": 0.0104, "step": 16668 }, { "epoch": 6.57, "learning_rate": 1.1121417774025932e-07, "loss": 0.0136, "step": 16669 }, { "epoch": 6.57, "learning_rate": 1.1119129922532306e-07, "loss": 0.0108, "step": 16670 }, { "epoch": 6.57, "learning_rate": 1.1116842215783373e-07, "loss": 0.009, "step": 16671 }, { "epoch": 6.57, "learning_rate": 1.1114554653816405e-07, "loss": 0.0126, "step": 16672 }, { "epoch": 6.57, "learning_rate": 1.1112267236668698e-07, "loss": 0.0136, "step": 16673 }, { "epoch": 6.57, "learning_rate": 1.1109979964377512e-07, "loss": 0.0108, "step": 16674 }, { "epoch": 6.57, "learning_rate": 1.1107692836980156e-07, "loss": 0.0124, "step": 16675 }, { "epoch": 6.57, "learning_rate": 1.1105405854513885e-07, "loss": 0.0105, "step": 16676 }, { "epoch": 6.57, "learning_rate": 1.1103119017015989e-07, "loss": 0.0104, "step": 16677 }, { "epoch": 6.57, "learning_rate": 1.1100832324523714e-07, "loss": 0.0114, "step": 16678 }, { "epoch": 6.57, "learning_rate": 1.1098545777074364e-07, "loss": 0.0131, "step": 16679 }, { "epoch": 6.57, "learning_rate": 1.1096259374705183e-07, "loss": 0.0116, "step": 16680 }, { "epoch": 6.57, "learning_rate": 1.1093973117453448e-07, "loss": 0.0117, "step": 16681 }, { "epoch": 6.57, "learning_rate": 1.1091687005356407e-07, "loss": 0.0124, "step": 16682 }, { "epoch": 6.57, "learning_rate": 1.1089401038451329e-07, "loss": 0.0119, "step": 16683 }, { "epoch": 6.58, "learning_rate": 1.108711521677548e-07, "loss": 0.0094, "step": 16684 }, { "epoch": 6.58, "learning_rate": 1.1084829540366099e-07, "loss": 0.0112, "step": 16685 }, { "epoch": 6.58, "learning_rate": 1.1082544009260448e-07, "loss": 0.0117, "step": 16686 }, { "epoch": 6.58, "learning_rate": 1.1080258623495775e-07, "loss": 0.0126, "step": 16687 }, { "epoch": 6.58, "learning_rate": 1.1077973383109338e-07, "loss": 0.0111, "step": 16688 }, { "epoch": 6.58, "learning_rate": 1.1075688288138368e-07, "loss": 0.0141, "step": 16689 }, { "epoch": 6.58, "learning_rate": 1.1073403338620122e-07, "loss": 0.0107, "step": 16690 }, { "epoch": 6.58, "learning_rate": 1.1071118534591818e-07, "loss": 0.0159, "step": 16691 }, { "epoch": 6.58, "learning_rate": 1.1068833876090731e-07, "loss": 0.0095, "step": 16692 }, { "epoch": 6.58, "learning_rate": 1.1066549363154068e-07, "loss": 0.0107, "step": 16693 }, { "epoch": 6.58, "learning_rate": 1.1064264995819082e-07, "loss": 0.016, "step": 16694 }, { "epoch": 6.58, "learning_rate": 1.1061980774122987e-07, "loss": 0.0107, "step": 16695 }, { "epoch": 6.58, "learning_rate": 1.105969669810302e-07, "loss": 0.0121, "step": 16696 }, { "epoch": 6.58, "learning_rate": 1.105741276779641e-07, "loss": 0.0121, "step": 16697 }, { "epoch": 6.58, "learning_rate": 1.1055128983240389e-07, "loss": 0.0117, "step": 16698 }, { "epoch": 6.58, "learning_rate": 1.1052845344472164e-07, "loss": 0.0097, "step": 16699 }, { "epoch": 6.58, "learning_rate": 1.1050561851528958e-07, "loss": 0.0112, "step": 16700 }, { "epoch": 6.58, "learning_rate": 1.1048278504447996e-07, "loss": 0.0113, "step": 16701 }, { "epoch": 6.58, "learning_rate": 1.1045995303266498e-07, "loss": 0.0099, "step": 16702 }, { "epoch": 6.58, "learning_rate": 1.1043712248021657e-07, "loss": 0.0142, "step": 16703 }, { "epoch": 6.58, "learning_rate": 1.1041429338750705e-07, "loss": 0.0134, "step": 16704 }, { "epoch": 6.58, "learning_rate": 1.103914657549083e-07, "loss": 0.0102, "step": 16705 }, { "epoch": 6.58, "learning_rate": 1.1036863958279248e-07, "loss": 0.0119, "step": 16706 }, { "epoch": 6.58, "learning_rate": 1.1034581487153167e-07, "loss": 0.0111, "step": 16707 }, { "epoch": 6.58, "learning_rate": 1.1032299162149774e-07, "loss": 0.0107, "step": 16708 }, { "epoch": 6.59, "learning_rate": 1.1030016983306279e-07, "loss": 0.0102, "step": 16709 }, { "epoch": 6.59, "learning_rate": 1.102773495065987e-07, "loss": 0.0108, "step": 16710 }, { "epoch": 6.59, "learning_rate": 1.1025453064247754e-07, "loss": 0.0135, "step": 16711 }, { "epoch": 6.59, "learning_rate": 1.1023171324107105e-07, "loss": 0.0122, "step": 16712 }, { "epoch": 6.59, "learning_rate": 1.1020889730275131e-07, "loss": 0.0123, "step": 16713 }, { "epoch": 6.59, "learning_rate": 1.1018608282788991e-07, "loss": 0.0135, "step": 16714 }, { "epoch": 6.59, "learning_rate": 1.1016326981685899e-07, "loss": 0.0131, "step": 16715 }, { "epoch": 6.59, "learning_rate": 1.1014045827003017e-07, "loss": 0.0146, "step": 16716 }, { "epoch": 6.59, "learning_rate": 1.101176481877754e-07, "loss": 0.0121, "step": 16717 }, { "epoch": 6.59, "learning_rate": 1.1009483957046619e-07, "loss": 0.0142, "step": 16718 }, { "epoch": 6.59, "learning_rate": 1.1007203241847462e-07, "loss": 0.01, "step": 16719 }, { "epoch": 6.59, "learning_rate": 1.1004922673217216e-07, "loss": 0.0102, "step": 16720 }, { "epoch": 6.59, "learning_rate": 1.1002642251193068e-07, "loss": 0.0142, "step": 16721 }, { "epoch": 6.59, "learning_rate": 1.1000361975812169e-07, "loss": 0.0146, "step": 16722 }, { "epoch": 6.59, "learning_rate": 1.0998081847111693e-07, "loss": 0.0106, "step": 16723 }, { "epoch": 6.59, "learning_rate": 1.09958018651288e-07, "loss": 0.0091, "step": 16724 }, { "epoch": 6.59, "learning_rate": 1.0993522029900659e-07, "loss": 0.0111, "step": 16725 }, { "epoch": 6.59, "learning_rate": 1.0991242341464415e-07, "loss": 0.0142, "step": 16726 }, { "epoch": 6.59, "learning_rate": 1.0988962799857226e-07, "loss": 0.0141, "step": 16727 }, { "epoch": 6.59, "learning_rate": 1.098668340511626e-07, "loss": 0.012, "step": 16728 }, { "epoch": 6.59, "learning_rate": 1.0984404157278646e-07, "loss": 0.0105, "step": 16729 }, { "epoch": 6.59, "learning_rate": 1.0982125056381549e-07, "loss": 0.0103, "step": 16730 }, { "epoch": 6.59, "learning_rate": 1.0979846102462094e-07, "loss": 0.0136, "step": 16731 }, { "epoch": 6.59, "learning_rate": 1.097756729555745e-07, "loss": 0.0122, "step": 16732 }, { "epoch": 6.59, "learning_rate": 1.0975288635704743e-07, "loss": 0.0122, "step": 16733 }, { "epoch": 6.6, "learning_rate": 1.0973010122941122e-07, "loss": 0.013, "step": 16734 }, { "epoch": 6.6, "learning_rate": 1.097073175730371e-07, "loss": 0.0131, "step": 16735 }, { "epoch": 6.6, "learning_rate": 1.0968453538829647e-07, "loss": 0.0104, "step": 16736 }, { "epoch": 6.6, "learning_rate": 1.0966175467556063e-07, "loss": 0.0118, "step": 16737 }, { "epoch": 6.6, "learning_rate": 1.0963897543520097e-07, "loss": 0.0109, "step": 16738 }, { "epoch": 6.6, "learning_rate": 1.0961619766758863e-07, "loss": 0.012, "step": 16739 }, { "epoch": 6.6, "learning_rate": 1.0959342137309486e-07, "loss": 0.012, "step": 16740 }, { "epoch": 6.6, "learning_rate": 1.0957064655209096e-07, "loss": 0.0128, "step": 16741 }, { "epoch": 6.6, "learning_rate": 1.0954787320494814e-07, "loss": 0.0141, "step": 16742 }, { "epoch": 6.6, "learning_rate": 1.0952510133203742e-07, "loss": 0.0121, "step": 16743 }, { "epoch": 6.6, "learning_rate": 1.0950233093373013e-07, "loss": 0.0091, "step": 16744 }, { "epoch": 6.6, "learning_rate": 1.0947956201039717e-07, "loss": 0.0126, "step": 16745 }, { "epoch": 6.6, "learning_rate": 1.0945679456240992e-07, "loss": 0.0124, "step": 16746 }, { "epoch": 6.6, "learning_rate": 1.0943402859013924e-07, "loss": 0.0122, "step": 16747 }, { "epoch": 6.6, "learning_rate": 1.0941126409395631e-07, "loss": 0.0118, "step": 16748 }, { "epoch": 6.6, "learning_rate": 1.0938850107423203e-07, "loss": 0.0116, "step": 16749 }, { "epoch": 6.6, "learning_rate": 1.0936573953133748e-07, "loss": 0.0127, "step": 16750 }, { "epoch": 6.6, "learning_rate": 1.0934297946564368e-07, "loss": 0.0127, "step": 16751 }, { "epoch": 6.6, "learning_rate": 1.0932022087752147e-07, "loss": 0.0135, "step": 16752 }, { "epoch": 6.6, "learning_rate": 1.092974637673419e-07, "loss": 0.013, "step": 16753 }, { "epoch": 6.6, "learning_rate": 1.0927470813547568e-07, "loss": 0.0117, "step": 16754 }, { "epoch": 6.6, "learning_rate": 1.0925195398229399e-07, "loss": 0.0123, "step": 16755 }, { "epoch": 6.6, "learning_rate": 1.0922920130816744e-07, "loss": 0.0116, "step": 16756 }, { "epoch": 6.6, "learning_rate": 1.0920645011346703e-07, "loss": 0.0114, "step": 16757 }, { "epoch": 6.6, "learning_rate": 1.0918370039856335e-07, "loss": 0.0105, "step": 16758 }, { "epoch": 6.6, "learning_rate": 1.0916095216382748e-07, "loss": 0.0101, "step": 16759 }, { "epoch": 6.61, "learning_rate": 1.0913820540962995e-07, "loss": 0.01, "step": 16760 }, { "epoch": 6.61, "learning_rate": 1.0911546013634165e-07, "loss": 0.0131, "step": 16761 }, { "epoch": 6.61, "learning_rate": 1.0909271634433315e-07, "loss": 0.0118, "step": 16762 }, { "epoch": 6.61, "learning_rate": 1.090699740339752e-07, "loss": 0.0091, "step": 16763 }, { "epoch": 6.61, "learning_rate": 1.0904723320563849e-07, "loss": 0.0115, "step": 16764 }, { "epoch": 6.61, "learning_rate": 1.0902449385969372e-07, "loss": 0.0133, "step": 16765 }, { "epoch": 6.61, "learning_rate": 1.0900175599651136e-07, "loss": 0.0131, "step": 16766 }, { "epoch": 6.61, "learning_rate": 1.0897901961646207e-07, "loss": 0.0109, "step": 16767 }, { "epoch": 6.61, "learning_rate": 1.0895628471991641e-07, "loss": 0.0128, "step": 16768 }, { "epoch": 6.61, "learning_rate": 1.0893355130724504e-07, "loss": 0.0113, "step": 16769 }, { "epoch": 6.61, "learning_rate": 1.0891081937881827e-07, "loss": 0.0155, "step": 16770 }, { "epoch": 6.61, "learning_rate": 1.0888808893500673e-07, "loss": 0.0132, "step": 16771 }, { "epoch": 6.61, "learning_rate": 1.0886535997618096e-07, "loss": 0.0121, "step": 16772 }, { "epoch": 6.61, "learning_rate": 1.088426325027112e-07, "loss": 0.012, "step": 16773 }, { "epoch": 6.61, "learning_rate": 1.0881990651496811e-07, "loss": 0.0101, "step": 16774 }, { "epoch": 6.61, "learning_rate": 1.0879718201332185e-07, "loss": 0.0115, "step": 16775 }, { "epoch": 6.61, "learning_rate": 1.0877445899814292e-07, "loss": 0.0128, "step": 16776 }, { "epoch": 6.61, "learning_rate": 1.0875173746980167e-07, "loss": 0.0086, "step": 16777 }, { "epoch": 6.61, "learning_rate": 1.0872901742866847e-07, "loss": 0.0129, "step": 16778 }, { "epoch": 6.61, "learning_rate": 1.0870629887511351e-07, "loss": 0.013, "step": 16779 }, { "epoch": 6.61, "learning_rate": 1.0868358180950714e-07, "loss": 0.0109, "step": 16780 }, { "epoch": 6.61, "learning_rate": 1.0866086623221957e-07, "loss": 0.0124, "step": 16781 }, { "epoch": 6.61, "learning_rate": 1.0863815214362118e-07, "loss": 0.0103, "step": 16782 }, { "epoch": 6.61, "learning_rate": 1.0861543954408193e-07, "loss": 0.0122, "step": 16783 }, { "epoch": 6.61, "learning_rate": 1.0859272843397223e-07, "loss": 0.0108, "step": 16784 }, { "epoch": 6.62, "learning_rate": 1.0857001881366199e-07, "loss": 0.0137, "step": 16785 }, { "epoch": 6.62, "learning_rate": 1.0854731068352161e-07, "loss": 0.0123, "step": 16786 }, { "epoch": 6.62, "learning_rate": 1.0852460404392103e-07, "loss": 0.0135, "step": 16787 }, { "epoch": 6.62, "learning_rate": 1.0850189889523042e-07, "loss": 0.0145, "step": 16788 }, { "epoch": 6.62, "learning_rate": 1.0847919523781977e-07, "loss": 0.0116, "step": 16789 }, { "epoch": 6.62, "learning_rate": 1.0845649307205911e-07, "loss": 0.0098, "step": 16790 }, { "epoch": 6.62, "learning_rate": 1.0843379239831846e-07, "loss": 0.0124, "step": 16791 }, { "epoch": 6.62, "learning_rate": 1.0841109321696795e-07, "loss": 0.0099, "step": 16792 }, { "epoch": 6.62, "learning_rate": 1.0838839552837733e-07, "loss": 0.014, "step": 16793 }, { "epoch": 6.62, "learning_rate": 1.0836569933291661e-07, "loss": 0.0093, "step": 16794 }, { "epoch": 6.62, "learning_rate": 1.0834300463095583e-07, "loss": 0.0107, "step": 16795 }, { "epoch": 6.62, "learning_rate": 1.0832031142286466e-07, "loss": 0.0107, "step": 16796 }, { "epoch": 6.62, "learning_rate": 1.0829761970901319e-07, "loss": 0.0098, "step": 16797 }, { "epoch": 6.62, "learning_rate": 1.0827492948977097e-07, "loss": 0.0123, "step": 16798 }, { "epoch": 6.62, "learning_rate": 1.0825224076550816e-07, "loss": 0.0128, "step": 16799 }, { "epoch": 6.62, "learning_rate": 1.0822955353659429e-07, "loss": 0.0113, "step": 16800 }, { "epoch": 6.62, "learning_rate": 1.0820686780339933e-07, "loss": 0.0142, "step": 16801 }, { "epoch": 6.62, "learning_rate": 1.081841835662928e-07, "loss": 0.0088, "step": 16802 }, { "epoch": 6.62, "learning_rate": 1.0816150082564457e-07, "loss": 0.0102, "step": 16803 }, { "epoch": 6.62, "learning_rate": 1.0813881958182429e-07, "loss": 0.0155, "step": 16804 }, { "epoch": 6.62, "learning_rate": 1.081161398352017e-07, "loss": 0.0081, "step": 16805 }, { "epoch": 6.62, "learning_rate": 1.0809346158614631e-07, "loss": 0.0107, "step": 16806 }, { "epoch": 6.62, "learning_rate": 1.0807078483502783e-07, "loss": 0.0115, "step": 16807 }, { "epoch": 6.62, "learning_rate": 1.080481095822158e-07, "loss": 0.014, "step": 16808 }, { "epoch": 6.62, "learning_rate": 1.0802543582807994e-07, "loss": 0.0119, "step": 16809 }, { "epoch": 6.62, "learning_rate": 1.080027635729896e-07, "loss": 0.0154, "step": 16810 }, { "epoch": 6.63, "learning_rate": 1.0798009281731442e-07, "loss": 0.0109, "step": 16811 }, { "epoch": 6.63, "learning_rate": 1.0795742356142395e-07, "loss": 0.0115, "step": 16812 }, { "epoch": 6.63, "learning_rate": 1.079347558056875e-07, "loss": 0.0101, "step": 16813 }, { "epoch": 6.63, "learning_rate": 1.0791208955047459e-07, "loss": 0.0119, "step": 16814 }, { "epoch": 6.63, "learning_rate": 1.0788942479615477e-07, "loss": 0.0118, "step": 16815 }, { "epoch": 6.63, "learning_rate": 1.0786676154309721e-07, "loss": 0.0117, "step": 16816 }, { "epoch": 6.63, "learning_rate": 1.0784409979167147e-07, "loss": 0.0099, "step": 16817 }, { "epoch": 6.63, "learning_rate": 1.078214395422469e-07, "loss": 0.0106, "step": 16818 }, { "epoch": 6.63, "learning_rate": 1.0779878079519267e-07, "loss": 0.0111, "step": 16819 }, { "epoch": 6.63, "learning_rate": 1.077761235508782e-07, "loss": 0.0106, "step": 16820 }, { "epoch": 6.63, "learning_rate": 1.0775346780967273e-07, "loss": 0.0137, "step": 16821 }, { "epoch": 6.63, "learning_rate": 1.0773081357194564e-07, "loss": 0.0135, "step": 16822 }, { "epoch": 6.63, "learning_rate": 1.0770816083806598e-07, "loss": 0.0104, "step": 16823 }, { "epoch": 6.63, "learning_rate": 1.076855096084031e-07, "loss": 0.0093, "step": 16824 }, { "epoch": 6.63, "learning_rate": 1.0766285988332593e-07, "loss": 0.0121, "step": 16825 }, { "epoch": 6.63, "learning_rate": 1.0764021166320401e-07, "loss": 0.0113, "step": 16826 }, { "epoch": 6.63, "learning_rate": 1.0761756494840617e-07, "loss": 0.0107, "step": 16827 }, { "epoch": 6.63, "learning_rate": 1.075949197393017e-07, "loss": 0.0092, "step": 16828 }, { "epoch": 6.63, "learning_rate": 1.0757227603625943e-07, "loss": 0.0108, "step": 16829 }, { "epoch": 6.63, "learning_rate": 1.0754963383964875e-07, "loss": 0.0114, "step": 16830 }, { "epoch": 6.63, "learning_rate": 1.0752699314983844e-07, "loss": 0.0126, "step": 16831 }, { "epoch": 6.63, "learning_rate": 1.0750435396719768e-07, "loss": 0.0104, "step": 16832 }, { "epoch": 6.63, "learning_rate": 1.074817162920953e-07, "loss": 0.0123, "step": 16833 }, { "epoch": 6.63, "learning_rate": 1.0745908012490035e-07, "loss": 0.011, "step": 16834 }, { "epoch": 6.63, "learning_rate": 1.0743644546598183e-07, "loss": 0.0115, "step": 16835 }, { "epoch": 6.64, "learning_rate": 1.0741381231570845e-07, "loss": 0.0123, "step": 16836 }, { "epoch": 6.64, "learning_rate": 1.0739118067444923e-07, "loss": 0.0118, "step": 16837 }, { "epoch": 6.64, "learning_rate": 1.0736855054257304e-07, "loss": 0.0115, "step": 16838 }, { "epoch": 6.64, "learning_rate": 1.0734592192044875e-07, "loss": 0.0107, "step": 16839 }, { "epoch": 6.64, "learning_rate": 1.0732329480844505e-07, "loss": 0.0137, "step": 16840 }, { "epoch": 6.64, "learning_rate": 1.0730066920693084e-07, "loss": 0.0125, "step": 16841 }, { "epoch": 6.64, "learning_rate": 1.0727804511627479e-07, "loss": 0.0118, "step": 16842 }, { "epoch": 6.64, "learning_rate": 1.0725542253684567e-07, "loss": 0.0114, "step": 16843 }, { "epoch": 6.64, "learning_rate": 1.0723280146901221e-07, "loss": 0.0126, "step": 16844 }, { "epoch": 6.64, "learning_rate": 1.0721018191314317e-07, "loss": 0.0117, "step": 16845 }, { "epoch": 6.64, "learning_rate": 1.0718756386960706e-07, "loss": 0.0128, "step": 16846 }, { "epoch": 6.64, "learning_rate": 1.0716494733877261e-07, "loss": 0.0112, "step": 16847 }, { "epoch": 6.64, "learning_rate": 1.0714233232100842e-07, "loss": 0.0117, "step": 16848 }, { "epoch": 6.64, "learning_rate": 1.0711971881668315e-07, "loss": 0.0118, "step": 16849 }, { "epoch": 6.64, "learning_rate": 1.0709710682616524e-07, "loss": 0.0114, "step": 16850 }, { "epoch": 6.64, "learning_rate": 1.070744963498233e-07, "loss": 0.0121, "step": 16851 }, { "epoch": 6.64, "learning_rate": 1.0705188738802582e-07, "loss": 0.0126, "step": 16852 }, { "epoch": 6.64, "learning_rate": 1.0702927994114141e-07, "loss": 0.0124, "step": 16853 }, { "epoch": 6.64, "learning_rate": 1.0700667400953835e-07, "loss": 0.0133, "step": 16854 }, { "epoch": 6.64, "learning_rate": 1.0698406959358526e-07, "loss": 0.0104, "step": 16855 }, { "epoch": 6.64, "learning_rate": 1.0696146669365037e-07, "loss": 0.0102, "step": 16856 }, { "epoch": 6.64, "learning_rate": 1.0693886531010217e-07, "loss": 0.0122, "step": 16857 }, { "epoch": 6.64, "learning_rate": 1.0691626544330912e-07, "loss": 0.0135, "step": 16858 }, { "epoch": 6.64, "learning_rate": 1.0689366709363939e-07, "loss": 0.0109, "step": 16859 }, { "epoch": 6.64, "learning_rate": 1.0687107026146138e-07, "loss": 0.0107, "step": 16860 }, { "epoch": 6.65, "learning_rate": 1.0684847494714338e-07, "loss": 0.0113, "step": 16861 }, { "epoch": 6.65, "learning_rate": 1.0682588115105375e-07, "loss": 0.013, "step": 16862 }, { "epoch": 6.65, "learning_rate": 1.0680328887356056e-07, "loss": 0.0105, "step": 16863 }, { "epoch": 6.65, "learning_rate": 1.0678069811503218e-07, "loss": 0.0137, "step": 16864 }, { "epoch": 6.65, "learning_rate": 1.0675810887583662e-07, "loss": 0.0116, "step": 16865 }, { "epoch": 6.65, "learning_rate": 1.0673552115634232e-07, "loss": 0.0112, "step": 16866 }, { "epoch": 6.65, "learning_rate": 1.0671293495691722e-07, "loss": 0.013, "step": 16867 }, { "epoch": 6.65, "learning_rate": 1.0669035027792955e-07, "loss": 0.011, "step": 16868 }, { "epoch": 6.65, "learning_rate": 1.066677671197472e-07, "loss": 0.014, "step": 16869 }, { "epoch": 6.65, "learning_rate": 1.0664518548273859e-07, "loss": 0.015, "step": 16870 }, { "epoch": 6.65, "learning_rate": 1.0662260536727146e-07, "loss": 0.0127, "step": 16871 }, { "epoch": 6.65, "learning_rate": 1.0660002677371403e-07, "loss": 0.0106, "step": 16872 }, { "epoch": 6.65, "learning_rate": 1.0657744970243414e-07, "loss": 0.012, "step": 16873 }, { "epoch": 6.65, "learning_rate": 1.0655487415379982e-07, "loss": 0.0095, "step": 16874 }, { "epoch": 6.65, "learning_rate": 1.0653230012817904e-07, "loss": 0.0122, "step": 16875 }, { "epoch": 6.65, "learning_rate": 1.0650972762593982e-07, "loss": 0.0122, "step": 16876 }, { "epoch": 6.65, "learning_rate": 1.0648715664744985e-07, "loss": 0.0094, "step": 16877 }, { "epoch": 6.65, "learning_rate": 1.0646458719307708e-07, "loss": 0.0125, "step": 16878 }, { "epoch": 6.65, "learning_rate": 1.0644201926318949e-07, "loss": 0.0105, "step": 16879 }, { "epoch": 6.65, "learning_rate": 1.064194528581547e-07, "loss": 0.0139, "step": 16880 }, { "epoch": 6.65, "learning_rate": 1.0639688797834069e-07, "loss": 0.0113, "step": 16881 }, { "epoch": 6.65, "learning_rate": 1.0637432462411506e-07, "loss": 0.0105, "step": 16882 }, { "epoch": 6.65, "learning_rate": 1.0635176279584563e-07, "loss": 0.0125, "step": 16883 }, { "epoch": 6.65, "learning_rate": 1.0632920249390016e-07, "loss": 0.012, "step": 16884 }, { "epoch": 6.65, "learning_rate": 1.0630664371864642e-07, "loss": 0.0128, "step": 16885 }, { "epoch": 6.65, "learning_rate": 1.0628408647045189e-07, "loss": 0.0098, "step": 16886 }, { "epoch": 6.66, "learning_rate": 1.0626153074968432e-07, "loss": 0.0109, "step": 16887 }, { "epoch": 6.66, "learning_rate": 1.0623897655671133e-07, "loss": 0.0102, "step": 16888 }, { "epoch": 6.66, "learning_rate": 1.062164238919006e-07, "loss": 0.0116, "step": 16889 }, { "epoch": 6.66, "learning_rate": 1.0619387275561955e-07, "loss": 0.0108, "step": 16890 }, { "epoch": 6.66, "learning_rate": 1.0617132314823581e-07, "loss": 0.0147, "step": 16891 }, { "epoch": 6.66, "learning_rate": 1.0614877507011688e-07, "loss": 0.0142, "step": 16892 }, { "epoch": 6.66, "learning_rate": 1.0612622852163038e-07, "loss": 0.0134, "step": 16893 }, { "epoch": 6.66, "learning_rate": 1.061036835031436e-07, "loss": 0.0113, "step": 16894 }, { "epoch": 6.66, "learning_rate": 1.0608114001502415e-07, "loss": 0.0123, "step": 16895 }, { "epoch": 6.66, "learning_rate": 1.0605859805763921e-07, "loss": 0.013, "step": 16896 }, { "epoch": 6.66, "learning_rate": 1.0603605763135653e-07, "loss": 0.0141, "step": 16897 }, { "epoch": 6.66, "learning_rate": 1.0601351873654319e-07, "loss": 0.0104, "step": 16898 }, { "epoch": 6.66, "learning_rate": 1.0599098137356675e-07, "loss": 0.0136, "step": 16899 }, { "epoch": 6.66, "learning_rate": 1.0596844554279436e-07, "loss": 0.0144, "step": 16900 }, { "epoch": 6.66, "learning_rate": 1.0594591124459337e-07, "loss": 0.0093, "step": 16901 }, { "epoch": 6.66, "learning_rate": 1.0592337847933119e-07, "loss": 0.0112, "step": 16902 }, { "epoch": 6.66, "learning_rate": 1.0590084724737489e-07, "loss": 0.0117, "step": 16903 }, { "epoch": 6.66, "learning_rate": 1.0587831754909173e-07, "loss": 0.013, "step": 16904 }, { "epoch": 6.66, "learning_rate": 1.0585578938484898e-07, "loss": 0.0117, "step": 16905 }, { "epoch": 6.66, "learning_rate": 1.0583326275501386e-07, "loss": 0.0107, "step": 16906 }, { "epoch": 6.66, "learning_rate": 1.0581073765995333e-07, "loss": 0.0118, "step": 16907 }, { "epoch": 6.66, "learning_rate": 1.0578821410003471e-07, "loss": 0.0108, "step": 16908 }, { "epoch": 6.66, "learning_rate": 1.0576569207562487e-07, "loss": 0.0086, "step": 16909 }, { "epoch": 6.66, "learning_rate": 1.0574317158709116e-07, "loss": 0.0104, "step": 16910 }, { "epoch": 6.66, "learning_rate": 1.0572065263480046e-07, "loss": 0.0094, "step": 16911 }, { "epoch": 6.67, "learning_rate": 1.056981352191199e-07, "loss": 0.0123, "step": 16912 }, { "epoch": 6.67, "learning_rate": 1.0567561934041633e-07, "loss": 0.0128, "step": 16913 }, { "epoch": 6.67, "learning_rate": 1.0565310499905681e-07, "loss": 0.0097, "step": 16914 }, { "epoch": 6.67, "learning_rate": 1.0563059219540828e-07, "loss": 0.0116, "step": 16915 }, { "epoch": 6.67, "learning_rate": 1.0560808092983778e-07, "loss": 0.0133, "step": 16916 }, { "epoch": 6.67, "learning_rate": 1.0558557120271196e-07, "loss": 0.0121, "step": 16917 }, { "epoch": 6.67, "learning_rate": 1.055630630143979e-07, "loss": 0.0133, "step": 16918 }, { "epoch": 6.67, "learning_rate": 1.0554055636526236e-07, "loss": 0.0129, "step": 16919 }, { "epoch": 6.67, "learning_rate": 1.0551805125567228e-07, "loss": 0.0123, "step": 16920 }, { "epoch": 6.67, "learning_rate": 1.0549554768599427e-07, "loss": 0.0135, "step": 16921 }, { "epoch": 6.67, "learning_rate": 1.0547304565659527e-07, "loss": 0.0122, "step": 16922 }, { "epoch": 6.67, "learning_rate": 1.0545054516784193e-07, "loss": 0.0152, "step": 16923 }, { "epoch": 6.67, "learning_rate": 1.0542804622010097e-07, "loss": 0.0115, "step": 16924 }, { "epoch": 6.67, "learning_rate": 1.0540554881373922e-07, "loss": 0.0152, "step": 16925 }, { "epoch": 6.67, "learning_rate": 1.0538305294912316e-07, "loss": 0.0113, "step": 16926 }, { "epoch": 6.67, "learning_rate": 1.0536055862661954e-07, "loss": 0.0119, "step": 16927 }, { "epoch": 6.67, "learning_rate": 1.0533806584659499e-07, "loss": 0.01, "step": 16928 }, { "epoch": 6.67, "learning_rate": 1.0531557460941618e-07, "loss": 0.0117, "step": 16929 }, { "epoch": 6.67, "learning_rate": 1.0529308491544951e-07, "loss": 0.0123, "step": 16930 }, { "epoch": 6.67, "learning_rate": 1.052705967650617e-07, "loss": 0.0111, "step": 16931 }, { "epoch": 6.67, "learning_rate": 1.0524811015861904e-07, "loss": 0.0135, "step": 16932 }, { "epoch": 6.67, "learning_rate": 1.052256250964883e-07, "loss": 0.0141, "step": 16933 }, { "epoch": 6.67, "learning_rate": 1.0520314157903579e-07, "loss": 0.0131, "step": 16934 }, { "epoch": 6.67, "learning_rate": 1.0518065960662807e-07, "loss": 0.0114, "step": 16935 }, { "epoch": 6.67, "learning_rate": 1.051581791796313e-07, "loss": 0.0123, "step": 16936 }, { "epoch": 6.68, "learning_rate": 1.0513570029841225e-07, "loss": 0.0089, "step": 16937 }, { "epoch": 6.68, "learning_rate": 1.0511322296333705e-07, "loss": 0.0132, "step": 16938 }, { "epoch": 6.68, "learning_rate": 1.0509074717477218e-07, "loss": 0.0106, "step": 16939 }, { "epoch": 6.68, "learning_rate": 1.0506827293308376e-07, "loss": 0.0115, "step": 16940 }, { "epoch": 6.68, "learning_rate": 1.0504580023863825e-07, "loss": 0.0104, "step": 16941 }, { "epoch": 6.68, "learning_rate": 1.050233290918019e-07, "loss": 0.0101, "step": 16942 }, { "epoch": 6.68, "learning_rate": 1.0500085949294097e-07, "loss": 0.0151, "step": 16943 }, { "epoch": 6.68, "learning_rate": 1.049783914424216e-07, "loss": 0.0126, "step": 16944 }, { "epoch": 6.68, "learning_rate": 1.0495592494061e-07, "loss": 0.0144, "step": 16945 }, { "epoch": 6.68, "learning_rate": 1.0493345998787249e-07, "loss": 0.0099, "step": 16946 }, { "epoch": 6.68, "learning_rate": 1.04910996584575e-07, "loss": 0.0136, "step": 16947 }, { "epoch": 6.68, "learning_rate": 1.0488853473108384e-07, "loss": 0.0116, "step": 16948 }, { "epoch": 6.68, "learning_rate": 1.0486607442776483e-07, "loss": 0.0104, "step": 16949 }, { "epoch": 6.68, "learning_rate": 1.0484361567498438e-07, "loss": 0.0121, "step": 16950 }, { "epoch": 6.68, "learning_rate": 1.0482115847310833e-07, "loss": 0.0114, "step": 16951 }, { "epoch": 6.68, "learning_rate": 1.0479870282250281e-07, "loss": 0.0136, "step": 16952 }, { "epoch": 6.68, "learning_rate": 1.0477624872353364e-07, "loss": 0.0109, "step": 16953 }, { "epoch": 6.68, "learning_rate": 1.047537961765669e-07, "loss": 0.0117, "step": 16954 }, { "epoch": 6.68, "learning_rate": 1.0473134518196853e-07, "loss": 0.0152, "step": 16955 }, { "epoch": 6.68, "learning_rate": 1.047088957401045e-07, "loss": 0.0119, "step": 16956 }, { "epoch": 6.68, "learning_rate": 1.046864478513406e-07, "loss": 0.0082, "step": 16957 }, { "epoch": 6.68, "learning_rate": 1.0466400151604273e-07, "loss": 0.0121, "step": 16958 }, { "epoch": 6.68, "learning_rate": 1.0464155673457673e-07, "loss": 0.0105, "step": 16959 }, { "epoch": 6.68, "learning_rate": 1.0461911350730852e-07, "loss": 0.0153, "step": 16960 }, { "epoch": 6.68, "learning_rate": 1.0459667183460372e-07, "loss": 0.0111, "step": 16961 }, { "epoch": 6.68, "learning_rate": 1.0457423171682825e-07, "loss": 0.0131, "step": 16962 }, { "epoch": 6.69, "learning_rate": 1.0455179315434768e-07, "loss": 0.0093, "step": 16963 }, { "epoch": 6.69, "learning_rate": 1.0452935614752782e-07, "loss": 0.0121, "step": 16964 }, { "epoch": 6.69, "learning_rate": 1.0450692069673435e-07, "loss": 0.0121, "step": 16965 }, { "epoch": 6.69, "learning_rate": 1.0448448680233303e-07, "loss": 0.012, "step": 16966 }, { "epoch": 6.69, "learning_rate": 1.0446205446468934e-07, "loss": 0.0127, "step": 16967 }, { "epoch": 6.69, "learning_rate": 1.0443962368416894e-07, "loss": 0.0097, "step": 16968 }, { "epoch": 6.69, "learning_rate": 1.0441719446113752e-07, "loss": 0.0115, "step": 16969 }, { "epoch": 6.69, "learning_rate": 1.0439476679596051e-07, "loss": 0.0114, "step": 16970 }, { "epoch": 6.69, "learning_rate": 1.0437234068900354e-07, "loss": 0.013, "step": 16971 }, { "epoch": 6.69, "learning_rate": 1.0434991614063196e-07, "loss": 0.0114, "step": 16972 }, { "epoch": 6.69, "learning_rate": 1.0432749315121148e-07, "loss": 0.0127, "step": 16973 }, { "epoch": 6.69, "learning_rate": 1.043050717211074e-07, "loss": 0.0125, "step": 16974 }, { "epoch": 6.69, "learning_rate": 1.0428265185068533e-07, "loss": 0.0134, "step": 16975 }, { "epoch": 6.69, "learning_rate": 1.0426023354031035e-07, "loss": 0.0102, "step": 16976 }, { "epoch": 6.69, "learning_rate": 1.0423781679034825e-07, "loss": 0.0115, "step": 16977 }, { "epoch": 6.69, "learning_rate": 1.042154016011641e-07, "loss": 0.012, "step": 16978 }, { "epoch": 6.69, "learning_rate": 1.0419298797312341e-07, "loss": 0.0119, "step": 16979 }, { "epoch": 6.69, "learning_rate": 1.0417057590659124e-07, "loss": 0.0135, "step": 16980 }, { "epoch": 6.69, "learning_rate": 1.041481654019332e-07, "loss": 0.0118, "step": 16981 }, { "epoch": 6.69, "learning_rate": 1.0412575645951432e-07, "loss": 0.0111, "step": 16982 }, { "epoch": 6.69, "learning_rate": 1.0410334907969995e-07, "loss": 0.0106, "step": 16983 }, { "epoch": 6.69, "learning_rate": 1.0408094326285519e-07, "loss": 0.013, "step": 16984 }, { "epoch": 6.69, "learning_rate": 1.0405853900934528e-07, "loss": 0.0146, "step": 16985 }, { "epoch": 6.69, "learning_rate": 1.0403613631953543e-07, "loss": 0.0124, "step": 16986 }, { "epoch": 6.69, "learning_rate": 1.0401373519379064e-07, "loss": 0.0108, "step": 16987 }, { "epoch": 6.7, "learning_rate": 1.0399133563247607e-07, "loss": 0.0124, "step": 16988 }, { "epoch": 6.7, "learning_rate": 1.0396893763595682e-07, "loss": 0.0108, "step": 16989 }, { "epoch": 6.7, "learning_rate": 1.0394654120459803e-07, "loss": 0.0115, "step": 16990 }, { "epoch": 6.7, "learning_rate": 1.0392414633876454e-07, "loss": 0.0125, "step": 16991 }, { "epoch": 6.7, "learning_rate": 1.0390175303882154e-07, "loss": 0.0137, "step": 16992 }, { "epoch": 6.7, "learning_rate": 1.0387936130513381e-07, "loss": 0.0114, "step": 16993 }, { "epoch": 6.7, "learning_rate": 1.038569711380664e-07, "loss": 0.0105, "step": 16994 }, { "epoch": 6.7, "learning_rate": 1.0383458253798428e-07, "loss": 0.012, "step": 16995 }, { "epoch": 6.7, "learning_rate": 1.0381219550525236e-07, "loss": 0.0094, "step": 16996 }, { "epoch": 6.7, "learning_rate": 1.0378981004023537e-07, "loss": 0.0129, "step": 16997 }, { "epoch": 6.7, "learning_rate": 1.0376742614329827e-07, "loss": 0.009, "step": 16998 }, { "epoch": 6.7, "learning_rate": 1.0374504381480589e-07, "loss": 0.013, "step": 16999 }, { "epoch": 6.7, "learning_rate": 1.0372266305512307e-07, "loss": 0.0126, "step": 17000 }, { "epoch": 6.7, "learning_rate": 1.0370028386461443e-07, "loss": 0.0124, "step": 17001 }, { "epoch": 6.7, "learning_rate": 1.036779062436449e-07, "loss": 0.0115, "step": 17002 }, { "epoch": 6.7, "learning_rate": 1.0365553019257891e-07, "loss": 0.0101, "step": 17003 }, { "epoch": 6.7, "learning_rate": 1.0363315571178155e-07, "loss": 0.0126, "step": 17004 }, { "epoch": 6.7, "learning_rate": 1.0361078280161719e-07, "loss": 0.0101, "step": 17005 }, { "epoch": 6.7, "learning_rate": 1.0358841146245066e-07, "loss": 0.0117, "step": 17006 }, { "epoch": 6.7, "learning_rate": 1.0356604169464642e-07, "loss": 0.0125, "step": 17007 }, { "epoch": 6.7, "learning_rate": 1.0354367349856913e-07, "loss": 0.0113, "step": 17008 }, { "epoch": 6.7, "learning_rate": 1.0352130687458345e-07, "loss": 0.0104, "step": 17009 }, { "epoch": 6.7, "learning_rate": 1.0349894182305377e-07, "loss": 0.0138, "step": 17010 }, { "epoch": 6.7, "learning_rate": 1.0347657834434467e-07, "loss": 0.012, "step": 17011 }, { "epoch": 6.7, "learning_rate": 1.0345421643882062e-07, "loss": 0.0126, "step": 17012 }, { "epoch": 6.7, "learning_rate": 1.0343185610684622e-07, "loss": 0.0141, "step": 17013 }, { "epoch": 6.71, "learning_rate": 1.0340949734878568e-07, "loss": 0.0125, "step": 17014 }, { "epoch": 6.71, "learning_rate": 1.0338714016500362e-07, "loss": 0.0141, "step": 17015 }, { "epoch": 6.71, "learning_rate": 1.0336478455586422e-07, "loss": 0.0126, "step": 17016 }, { "epoch": 6.71, "learning_rate": 1.0334243052173209e-07, "loss": 0.0109, "step": 17017 }, { "epoch": 6.71, "learning_rate": 1.0332007806297136e-07, "loss": 0.0134, "step": 17018 }, { "epoch": 6.71, "learning_rate": 1.0329772717994652e-07, "loss": 0.0104, "step": 17019 }, { "epoch": 6.71, "learning_rate": 1.0327537787302158e-07, "loss": 0.013, "step": 17020 }, { "epoch": 6.71, "learning_rate": 1.0325303014256115e-07, "loss": 0.0146, "step": 17021 }, { "epoch": 6.71, "learning_rate": 1.032306839889292e-07, "loss": 0.0118, "step": 17022 }, { "epoch": 6.71, "learning_rate": 1.0320833941249012e-07, "loss": 0.0129, "step": 17023 }, { "epoch": 6.71, "learning_rate": 1.031859964136079e-07, "loss": 0.0112, "step": 17024 }, { "epoch": 6.71, "learning_rate": 1.031636549926468e-07, "loss": 0.0096, "step": 17025 }, { "epoch": 6.71, "learning_rate": 1.0314131514997097e-07, "loss": 0.0118, "step": 17026 }, { "epoch": 6.71, "learning_rate": 1.0311897688594457e-07, "loss": 0.0119, "step": 17027 }, { "epoch": 6.71, "learning_rate": 1.0309664020093152e-07, "loss": 0.0126, "step": 17028 }, { "epoch": 6.71, "learning_rate": 1.0307430509529595e-07, "loss": 0.0123, "step": 17029 }, { "epoch": 6.71, "learning_rate": 1.0305197156940199e-07, "loss": 0.0104, "step": 17030 }, { "epoch": 6.71, "learning_rate": 1.0302963962361347e-07, "loss": 0.0111, "step": 17031 }, { "epoch": 6.71, "learning_rate": 1.0300730925829444e-07, "loss": 0.0135, "step": 17032 }, { "epoch": 6.71, "learning_rate": 1.0298498047380895e-07, "loss": 0.0117, "step": 17033 }, { "epoch": 6.71, "learning_rate": 1.0296265327052074e-07, "loss": 0.0087, "step": 17034 }, { "epoch": 6.71, "learning_rate": 1.029403276487938e-07, "loss": 0.0125, "step": 17035 }, { "epoch": 6.71, "learning_rate": 1.0291800360899208e-07, "loss": 0.0103, "step": 17036 }, { "epoch": 6.71, "learning_rate": 1.0289568115147926e-07, "loss": 0.0129, "step": 17037 }, { "epoch": 6.71, "learning_rate": 1.0287336027661927e-07, "loss": 0.0138, "step": 17038 }, { "epoch": 6.72, "learning_rate": 1.0285104098477592e-07, "loss": 0.0135, "step": 17039 }, { "epoch": 6.72, "learning_rate": 1.0282872327631303e-07, "loss": 0.0114, "step": 17040 }, { "epoch": 6.72, "learning_rate": 1.0280640715159417e-07, "loss": 0.0119, "step": 17041 }, { "epoch": 6.72, "learning_rate": 1.0278409261098321e-07, "loss": 0.0102, "step": 17042 }, { "epoch": 6.72, "learning_rate": 1.0276177965484367e-07, "loss": 0.0144, "step": 17043 }, { "epoch": 6.72, "learning_rate": 1.0273946828353946e-07, "loss": 0.0114, "step": 17044 }, { "epoch": 6.72, "learning_rate": 1.0271715849743406e-07, "loss": 0.0097, "step": 17045 }, { "epoch": 6.72, "learning_rate": 1.0269485029689118e-07, "loss": 0.009, "step": 17046 }, { "epoch": 6.72, "learning_rate": 1.0267254368227418e-07, "loss": 0.0118, "step": 17047 }, { "epoch": 6.72, "learning_rate": 1.0265023865394697e-07, "loss": 0.0114, "step": 17048 }, { "epoch": 6.72, "learning_rate": 1.0262793521227285e-07, "loss": 0.0093, "step": 17049 }, { "epoch": 6.72, "learning_rate": 1.0260563335761548e-07, "loss": 0.0123, "step": 17050 }, { "epoch": 6.72, "learning_rate": 1.0258333309033816e-07, "loss": 0.0121, "step": 17051 }, { "epoch": 6.72, "learning_rate": 1.0256103441080446e-07, "loss": 0.0112, "step": 17052 }, { "epoch": 6.72, "learning_rate": 1.0253873731937788e-07, "loss": 0.0106, "step": 17053 }, { "epoch": 6.72, "learning_rate": 1.025164418164217e-07, "loss": 0.0141, "step": 17054 }, { "epoch": 6.72, "learning_rate": 1.0249414790229935e-07, "loss": 0.0118, "step": 17055 }, { "epoch": 6.72, "learning_rate": 1.0247185557737421e-07, "loss": 0.0125, "step": 17056 }, { "epoch": 6.72, "learning_rate": 1.0244956484200966e-07, "loss": 0.0111, "step": 17057 }, { "epoch": 6.72, "learning_rate": 1.0242727569656886e-07, "loss": 0.0087, "step": 17058 }, { "epoch": 6.72, "learning_rate": 1.0240498814141527e-07, "loss": 0.0111, "step": 17059 }, { "epoch": 6.72, "learning_rate": 1.0238270217691187e-07, "loss": 0.0103, "step": 17060 }, { "epoch": 6.72, "learning_rate": 1.0236041780342224e-07, "loss": 0.0154, "step": 17061 }, { "epoch": 6.72, "learning_rate": 1.0233813502130932e-07, "loss": 0.0099, "step": 17062 }, { "epoch": 6.72, "learning_rate": 1.0231585383093646e-07, "loss": 0.0113, "step": 17063 }, { "epoch": 6.73, "learning_rate": 1.0229357423266663e-07, "loss": 0.0121, "step": 17064 }, { "epoch": 6.73, "learning_rate": 1.0227129622686304e-07, "loss": 0.0113, "step": 17065 }, { "epoch": 6.73, "learning_rate": 1.0224901981388878e-07, "loss": 0.0131, "step": 17066 }, { "epoch": 6.73, "learning_rate": 1.0222674499410701e-07, "loss": 0.0109, "step": 17067 }, { "epoch": 6.73, "learning_rate": 1.0220447176788062e-07, "loss": 0.0125, "step": 17068 }, { "epoch": 6.73, "learning_rate": 1.0218220013557272e-07, "loss": 0.0104, "step": 17069 }, { "epoch": 6.73, "learning_rate": 1.021599300975463e-07, "loss": 0.0098, "step": 17070 }, { "epoch": 6.73, "learning_rate": 1.0213766165416439e-07, "loss": 0.0113, "step": 17071 }, { "epoch": 6.73, "learning_rate": 1.0211539480578979e-07, "loss": 0.0115, "step": 17072 }, { "epoch": 6.73, "learning_rate": 1.0209312955278558e-07, "loss": 0.0102, "step": 17073 }, { "epoch": 6.73, "learning_rate": 1.0207086589551445e-07, "loss": 0.012, "step": 17074 }, { "epoch": 6.73, "learning_rate": 1.0204860383433937e-07, "loss": 0.0113, "step": 17075 }, { "epoch": 6.73, "learning_rate": 1.0202634336962325e-07, "loss": 0.0102, "step": 17076 }, { "epoch": 6.73, "learning_rate": 1.0200408450172879e-07, "loss": 0.0122, "step": 17077 }, { "epoch": 6.73, "learning_rate": 1.019818272310188e-07, "loss": 0.013, "step": 17078 }, { "epoch": 6.73, "learning_rate": 1.0195957155785604e-07, "loss": 0.0116, "step": 17079 }, { "epoch": 6.73, "learning_rate": 1.0193731748260334e-07, "loss": 0.0128, "step": 17080 }, { "epoch": 6.73, "learning_rate": 1.0191506500562326e-07, "loss": 0.0129, "step": 17081 }, { "epoch": 6.73, "learning_rate": 1.0189281412727861e-07, "loss": 0.0113, "step": 17082 }, { "epoch": 6.73, "learning_rate": 1.0187056484793187e-07, "loss": 0.0098, "step": 17083 }, { "epoch": 6.73, "learning_rate": 1.0184831716794588e-07, "loss": 0.0126, "step": 17084 }, { "epoch": 6.73, "learning_rate": 1.0182607108768312e-07, "loss": 0.0096, "step": 17085 }, { "epoch": 6.73, "learning_rate": 1.0180382660750626e-07, "loss": 0.011, "step": 17086 }, { "epoch": 6.73, "learning_rate": 1.017815837277776e-07, "loss": 0.0096, "step": 17087 }, { "epoch": 6.73, "learning_rate": 1.0175934244886004e-07, "loss": 0.0151, "step": 17088 }, { "epoch": 6.73, "learning_rate": 1.017371027711158e-07, "loss": 0.0133, "step": 17089 }, { "epoch": 6.74, "learning_rate": 1.0171486469490751e-07, "loss": 0.0125, "step": 17090 }, { "epoch": 6.74, "learning_rate": 1.0169262822059747e-07, "loss": 0.0114, "step": 17091 }, { "epoch": 6.74, "learning_rate": 1.0167039334854819e-07, "loss": 0.0109, "step": 17092 }, { "epoch": 6.74, "learning_rate": 1.0164816007912205e-07, "loss": 0.0095, "step": 17093 }, { "epoch": 6.74, "learning_rate": 1.0162592841268152e-07, "loss": 0.0127, "step": 17094 }, { "epoch": 6.74, "learning_rate": 1.0160369834958875e-07, "loss": 0.0127, "step": 17095 }, { "epoch": 6.74, "learning_rate": 1.0158146989020615e-07, "loss": 0.01, "step": 17096 }, { "epoch": 6.74, "learning_rate": 1.0155924303489611e-07, "loss": 0.0122, "step": 17097 }, { "epoch": 6.74, "learning_rate": 1.0153701778402068e-07, "loss": 0.0114, "step": 17098 }, { "epoch": 6.74, "learning_rate": 1.0151479413794233e-07, "loss": 0.013, "step": 17099 }, { "epoch": 6.74, "learning_rate": 1.0149257209702298e-07, "loss": 0.0111, "step": 17100 }, { "epoch": 6.74, "learning_rate": 1.0147035166162515e-07, "loss": 0.0116, "step": 17101 }, { "epoch": 6.74, "learning_rate": 1.0144813283211075e-07, "loss": 0.0136, "step": 17102 }, { "epoch": 6.74, "learning_rate": 1.014259156088421e-07, "loss": 0.015, "step": 17103 }, { "epoch": 6.74, "learning_rate": 1.0140369999218113e-07, "loss": 0.0103, "step": 17104 }, { "epoch": 6.74, "learning_rate": 1.0138148598248999e-07, "loss": 0.0108, "step": 17105 }, { "epoch": 6.74, "learning_rate": 1.0135927358013075e-07, "loss": 0.012, "step": 17106 }, { "epoch": 6.74, "learning_rate": 1.0133706278546553e-07, "loss": 0.0143, "step": 17107 }, { "epoch": 6.74, "learning_rate": 1.0131485359885612e-07, "loss": 0.0105, "step": 17108 }, { "epoch": 6.74, "learning_rate": 1.0129264602066466e-07, "loss": 0.0156, "step": 17109 }, { "epoch": 6.74, "learning_rate": 1.01270440051253e-07, "loss": 0.0102, "step": 17110 }, { "epoch": 6.74, "learning_rate": 1.0124823569098322e-07, "loss": 0.0112, "step": 17111 }, { "epoch": 6.74, "learning_rate": 1.0122603294021703e-07, "loss": 0.0097, "step": 17112 }, { "epoch": 6.74, "learning_rate": 1.0120383179931649e-07, "loss": 0.0146, "step": 17113 }, { "epoch": 6.74, "learning_rate": 1.0118163226864322e-07, "loss": 0.0114, "step": 17114 }, { "epoch": 6.75, "learning_rate": 1.011594343485592e-07, "loss": 0.0137, "step": 17115 }, { "epoch": 6.75, "learning_rate": 1.0113723803942616e-07, "loss": 0.0117, "step": 17116 }, { "epoch": 6.75, "learning_rate": 1.0111504334160594e-07, "loss": 0.0114, "step": 17117 }, { "epoch": 6.75, "learning_rate": 1.0109285025546018e-07, "loss": 0.0114, "step": 17118 }, { "epoch": 6.75, "learning_rate": 1.0107065878135062e-07, "loss": 0.0115, "step": 17119 }, { "epoch": 6.75, "learning_rate": 1.0104846891963905e-07, "loss": 0.0104, "step": 17120 }, { "epoch": 6.75, "learning_rate": 1.0102628067068697e-07, "loss": 0.0119, "step": 17121 }, { "epoch": 6.75, "learning_rate": 1.0100409403485617e-07, "loss": 0.0157, "step": 17122 }, { "epoch": 6.75, "learning_rate": 1.00981909012508e-07, "loss": 0.0115, "step": 17123 }, { "epoch": 6.75, "learning_rate": 1.009597256040044e-07, "loss": 0.0131, "step": 17124 }, { "epoch": 6.75, "learning_rate": 1.0093754380970669e-07, "loss": 0.0119, "step": 17125 }, { "epoch": 6.75, "learning_rate": 1.0091536362997653e-07, "loss": 0.0098, "step": 17126 }, { "epoch": 6.75, "learning_rate": 1.0089318506517519e-07, "loss": 0.0114, "step": 17127 }, { "epoch": 6.75, "learning_rate": 1.0087100811566447e-07, "loss": 0.0121, "step": 17128 }, { "epoch": 6.75, "learning_rate": 1.0084883278180557e-07, "loss": 0.0128, "step": 17129 }, { "epoch": 6.75, "learning_rate": 1.0082665906396007e-07, "loss": 0.0118, "step": 17130 }, { "epoch": 6.75, "learning_rate": 1.0080448696248924e-07, "loss": 0.0115, "step": 17131 }, { "epoch": 6.75, "learning_rate": 1.007823164777545e-07, "loss": 0.0113, "step": 17132 }, { "epoch": 6.75, "learning_rate": 1.0076014761011724e-07, "loss": 0.011, "step": 17133 }, { "epoch": 6.75, "learning_rate": 1.007379803599388e-07, "loss": 0.0117, "step": 17134 }, { "epoch": 6.75, "learning_rate": 1.0071581472758032e-07, "loss": 0.0119, "step": 17135 }, { "epoch": 6.75, "learning_rate": 1.006936507134032e-07, "loss": 0.0121, "step": 17136 }, { "epoch": 6.75, "learning_rate": 1.0067148831776871e-07, "loss": 0.0127, "step": 17137 }, { "epoch": 6.75, "learning_rate": 1.0064932754103793e-07, "loss": 0.013, "step": 17138 }, { "epoch": 6.75, "learning_rate": 1.006271683835721e-07, "loss": 0.0098, "step": 17139 }, { "epoch": 6.76, "learning_rate": 1.0060501084573241e-07, "loss": 0.0101, "step": 17140 }, { "epoch": 6.76, "learning_rate": 1.0058285492788004e-07, "loss": 0.009, "step": 17141 }, { "epoch": 6.76, "learning_rate": 1.0056070063037595e-07, "loss": 0.0134, "step": 17142 }, { "epoch": 6.76, "learning_rate": 1.0053854795358143e-07, "loss": 0.0105, "step": 17143 }, { "epoch": 6.76, "learning_rate": 1.0051639689785734e-07, "loss": 0.0098, "step": 17144 }, { "epoch": 6.76, "learning_rate": 1.0049424746356474e-07, "loss": 0.0129, "step": 17145 }, { "epoch": 6.76, "learning_rate": 1.0047209965106469e-07, "loss": 0.0115, "step": 17146 }, { "epoch": 6.76, "learning_rate": 1.0044995346071824e-07, "loss": 0.0109, "step": 17147 }, { "epoch": 6.76, "learning_rate": 1.0042780889288616e-07, "loss": 0.0128, "step": 17148 }, { "epoch": 6.76, "learning_rate": 1.0040566594792946e-07, "loss": 0.0122, "step": 17149 }, { "epoch": 6.76, "learning_rate": 1.0038352462620905e-07, "loss": 0.0126, "step": 17150 }, { "epoch": 6.76, "learning_rate": 1.0036138492808586e-07, "loss": 0.0096, "step": 17151 }, { "epoch": 6.76, "learning_rate": 1.0033924685392058e-07, "loss": 0.0121, "step": 17152 }, { "epoch": 6.76, "learning_rate": 1.0031711040407422e-07, "loss": 0.0137, "step": 17153 }, { "epoch": 6.76, "learning_rate": 1.0029497557890728e-07, "loss": 0.0134, "step": 17154 }, { "epoch": 6.76, "learning_rate": 1.0027284237878084e-07, "loss": 0.0099, "step": 17155 }, { "epoch": 6.76, "learning_rate": 1.0025071080405544e-07, "loss": 0.0116, "step": 17156 }, { "epoch": 6.76, "learning_rate": 1.0022858085509194e-07, "loss": 0.013, "step": 17157 }, { "epoch": 6.76, "learning_rate": 1.0020645253225084e-07, "loss": 0.0112, "step": 17158 }, { "epoch": 6.76, "learning_rate": 1.0018432583589291e-07, "loss": 0.0126, "step": 17159 }, { "epoch": 6.76, "learning_rate": 1.0016220076637875e-07, "loss": 0.0117, "step": 17160 }, { "epoch": 6.76, "learning_rate": 1.0014007732406908e-07, "loss": 0.0116, "step": 17161 }, { "epoch": 6.76, "learning_rate": 1.0011795550932429e-07, "loss": 0.0115, "step": 17162 }, { "epoch": 6.76, "learning_rate": 1.0009583532250502e-07, "loss": 0.0147, "step": 17163 }, { "epoch": 6.76, "learning_rate": 1.0007371676397189e-07, "loss": 0.0113, "step": 17164 }, { "epoch": 6.76, "learning_rate": 1.0005159983408522e-07, "loss": 0.0109, "step": 17165 }, { "epoch": 6.77, "learning_rate": 1.0002948453320565e-07, "loss": 0.0122, "step": 17166 }, { "epoch": 6.77, "learning_rate": 1.000073708616934e-07, "loss": 0.0121, "step": 17167 }, { "epoch": 6.77, "learning_rate": 9.998525881990918e-08, "loss": 0.0094, "step": 17168 }, { "epoch": 6.77, "learning_rate": 9.996314840821314e-08, "loss": 0.0099, "step": 17169 }, { "epoch": 6.77, "learning_rate": 9.994103962696585e-08, "loss": 0.0101, "step": 17170 }, { "epoch": 6.77, "learning_rate": 9.991893247652742e-08, "loss": 0.011, "step": 17171 }, { "epoch": 6.77, "learning_rate": 9.989682695725834e-08, "loss": 0.0127, "step": 17172 }, { "epoch": 6.77, "learning_rate": 9.987472306951881e-08, "loss": 0.0108, "step": 17173 }, { "epoch": 6.77, "learning_rate": 9.98526208136692e-08, "loss": 0.0103, "step": 17174 }, { "epoch": 6.77, "learning_rate": 9.98305201900696e-08, "loss": 0.0148, "step": 17175 }, { "epoch": 6.77, "learning_rate": 9.980842119908027e-08, "loss": 0.0112, "step": 17176 }, { "epoch": 6.77, "learning_rate": 9.978632384106138e-08, "loss": 0.012, "step": 17177 }, { "epoch": 6.77, "learning_rate": 9.976422811637322e-08, "loss": 0.0139, "step": 17178 }, { "epoch": 6.77, "learning_rate": 9.974213402537571e-08, "loss": 0.0127, "step": 17179 }, { "epoch": 6.77, "learning_rate": 9.972004156842904e-08, "loss": 0.0117, "step": 17180 }, { "epoch": 6.77, "learning_rate": 9.969795074589335e-08, "loss": 0.0097, "step": 17181 }, { "epoch": 6.77, "learning_rate": 9.967586155812855e-08, "loss": 0.0142, "step": 17182 }, { "epoch": 6.77, "learning_rate": 9.965377400549473e-08, "loss": 0.0079, "step": 17183 }, { "epoch": 6.77, "learning_rate": 9.963168808835196e-08, "loss": 0.0113, "step": 17184 }, { "epoch": 6.77, "learning_rate": 9.960960380706008e-08, "loss": 0.0119, "step": 17185 }, { "epoch": 6.77, "learning_rate": 9.958752116197906e-08, "loss": 0.0109, "step": 17186 }, { "epoch": 6.77, "learning_rate": 9.956544015346891e-08, "loss": 0.014, "step": 17187 }, { "epoch": 6.77, "learning_rate": 9.954336078188937e-08, "loss": 0.0117, "step": 17188 }, { "epoch": 6.77, "learning_rate": 9.952128304760037e-08, "loss": 0.0127, "step": 17189 }, { "epoch": 6.77, "learning_rate": 9.949920695096176e-08, "loss": 0.0124, "step": 17190 }, { "epoch": 6.78, "learning_rate": 9.94771324923334e-08, "loss": 0.0115, "step": 17191 }, { "epoch": 6.78, "learning_rate": 9.945505967207493e-08, "loss": 0.0099, "step": 17192 }, { "epoch": 6.78, "learning_rate": 9.943298849054624e-08, "loss": 0.0114, "step": 17193 }, { "epoch": 6.78, "learning_rate": 9.941091894810684e-08, "loss": 0.0122, "step": 17194 }, { "epoch": 6.78, "learning_rate": 9.938885104511677e-08, "loss": 0.0117, "step": 17195 }, { "epoch": 6.78, "learning_rate": 9.936678478193541e-08, "loss": 0.0092, "step": 17196 }, { "epoch": 6.78, "learning_rate": 9.93447201589226e-08, "loss": 0.0101, "step": 17197 }, { "epoch": 6.78, "learning_rate": 9.932265717643771e-08, "loss": 0.0109, "step": 17198 }, { "epoch": 6.78, "learning_rate": 9.930059583484067e-08, "loss": 0.0099, "step": 17199 }, { "epoch": 6.78, "learning_rate": 9.927853613449078e-08, "loss": 0.0109, "step": 17200 }, { "epoch": 6.78, "learning_rate": 9.925647807574775e-08, "loss": 0.0116, "step": 17201 }, { "epoch": 6.78, "learning_rate": 9.923442165897094e-08, "loss": 0.0113, "step": 17202 }, { "epoch": 6.78, "learning_rate": 9.92123668845199e-08, "loss": 0.0155, "step": 17203 }, { "epoch": 6.78, "learning_rate": 9.91903137527542e-08, "loss": 0.0113, "step": 17204 }, { "epoch": 6.78, "learning_rate": 9.916826226403308e-08, "loss": 0.0138, "step": 17205 }, { "epoch": 6.78, "learning_rate": 9.914621241871604e-08, "loss": 0.0109, "step": 17206 }, { "epoch": 6.78, "learning_rate": 9.912416421716246e-08, "loss": 0.0124, "step": 17207 }, { "epoch": 6.78, "learning_rate": 9.910211765973175e-08, "loss": 0.011, "step": 17208 }, { "epoch": 6.78, "learning_rate": 9.908007274678311e-08, "loss": 0.0117, "step": 17209 }, { "epoch": 6.78, "learning_rate": 9.905802947867599e-08, "loss": 0.0127, "step": 17210 }, { "epoch": 6.78, "learning_rate": 9.90359878557695e-08, "loss": 0.0132, "step": 17211 }, { "epoch": 6.78, "learning_rate": 9.901394787842294e-08, "loss": 0.0104, "step": 17212 }, { "epoch": 6.78, "learning_rate": 9.899190954699556e-08, "loss": 0.0118, "step": 17213 }, { "epoch": 6.78, "learning_rate": 9.896987286184663e-08, "loss": 0.0145, "step": 17214 }, { "epoch": 6.78, "learning_rate": 9.894783782333513e-08, "loss": 0.0147, "step": 17215 }, { "epoch": 6.78, "learning_rate": 9.892580443182031e-08, "loss": 0.0108, "step": 17216 }, { "epoch": 6.79, "learning_rate": 9.890377268766124e-08, "loss": 0.0133, "step": 17217 }, { "epoch": 6.79, "learning_rate": 9.888174259121711e-08, "loss": 0.0131, "step": 17218 }, { "epoch": 6.79, "learning_rate": 9.885971414284682e-08, "loss": 0.0127, "step": 17219 }, { "epoch": 6.79, "learning_rate": 9.883768734290947e-08, "loss": 0.0115, "step": 17220 }, { "epoch": 6.79, "learning_rate": 9.88156621917641e-08, "loss": 0.0122, "step": 17221 }, { "epoch": 6.79, "learning_rate": 9.879363868976969e-08, "loss": 0.0133, "step": 17222 }, { "epoch": 6.79, "learning_rate": 9.877161683728508e-08, "loss": 0.01, "step": 17223 }, { "epoch": 6.79, "learning_rate": 9.874959663466936e-08, "loss": 0.0149, "step": 17224 }, { "epoch": 6.79, "learning_rate": 9.872757808228124e-08, "loss": 0.0107, "step": 17225 }, { "epoch": 6.79, "learning_rate": 9.870556118047969e-08, "loss": 0.0118, "step": 17226 }, { "epoch": 6.79, "learning_rate": 9.868354592962363e-08, "loss": 0.0106, "step": 17227 }, { "epoch": 6.79, "learning_rate": 9.866153233007169e-08, "loss": 0.0132, "step": 17228 }, { "epoch": 6.79, "learning_rate": 9.863952038218276e-08, "loss": 0.0117, "step": 17229 }, { "epoch": 6.79, "learning_rate": 9.861751008631561e-08, "loss": 0.0119, "step": 17230 }, { "epoch": 6.79, "learning_rate": 9.859550144282902e-08, "loss": 0.011, "step": 17231 }, { "epoch": 6.79, "learning_rate": 9.857349445208159e-08, "loss": 0.0118, "step": 17232 }, { "epoch": 6.79, "learning_rate": 9.85514891144321e-08, "loss": 0.0095, "step": 17233 }, { "epoch": 6.79, "learning_rate": 9.8529485430239e-08, "loss": 0.0142, "step": 17234 }, { "epoch": 6.79, "learning_rate": 9.850748339986124e-08, "loss": 0.0104, "step": 17235 }, { "epoch": 6.79, "learning_rate": 9.848548302365718e-08, "loss": 0.0142, "step": 17236 }, { "epoch": 6.79, "learning_rate": 9.846348430198552e-08, "loss": 0.0115, "step": 17237 }, { "epoch": 6.79, "learning_rate": 9.84414872352046e-08, "loss": 0.0103, "step": 17238 }, { "epoch": 6.79, "learning_rate": 9.841949182367324e-08, "loss": 0.0124, "step": 17239 }, { "epoch": 6.79, "learning_rate": 9.83974980677497e-08, "loss": 0.013, "step": 17240 }, { "epoch": 6.79, "learning_rate": 9.83755059677926e-08, "loss": 0.0089, "step": 17241 }, { "epoch": 6.8, "learning_rate": 9.835351552416025e-08, "loss": 0.0124, "step": 17242 }, { "epoch": 6.8, "learning_rate": 9.833152673721107e-08, "loss": 0.0118, "step": 17243 }, { "epoch": 6.8, "learning_rate": 9.83095396073035e-08, "loss": 0.0102, "step": 17244 }, { "epoch": 6.8, "learning_rate": 9.828755413479597e-08, "loss": 0.0127, "step": 17245 }, { "epoch": 6.8, "learning_rate": 9.826557032004665e-08, "loss": 0.0122, "step": 17246 }, { "epoch": 6.8, "learning_rate": 9.824358816341391e-08, "loss": 0.0134, "step": 17247 }, { "epoch": 6.8, "learning_rate": 9.822160766525608e-08, "loss": 0.0121, "step": 17248 }, { "epoch": 6.8, "learning_rate": 9.819962882593134e-08, "loss": 0.0117, "step": 17249 }, { "epoch": 6.8, "learning_rate": 9.817765164579797e-08, "loss": 0.0146, "step": 17250 }, { "epoch": 6.8, "learning_rate": 9.815567612521405e-08, "loss": 0.0097, "step": 17251 }, { "epoch": 6.8, "learning_rate": 9.813370226453783e-08, "loss": 0.0144, "step": 17252 }, { "epoch": 6.8, "learning_rate": 9.811173006412743e-08, "loss": 0.0115, "step": 17253 }, { "epoch": 6.8, "learning_rate": 9.808975952434108e-08, "loss": 0.0103, "step": 17254 }, { "epoch": 6.8, "learning_rate": 9.806779064553667e-08, "loss": 0.0105, "step": 17255 }, { "epoch": 6.8, "learning_rate": 9.804582342807236e-08, "loss": 0.0119, "step": 17256 }, { "epoch": 6.8, "learning_rate": 9.802385787230618e-08, "loss": 0.0122, "step": 17257 }, { "epoch": 6.8, "learning_rate": 9.80018939785962e-08, "loss": 0.0116, "step": 17258 }, { "epoch": 6.8, "learning_rate": 9.797993174730024e-08, "loss": 0.0097, "step": 17259 }, { "epoch": 6.8, "learning_rate": 9.795797117877636e-08, "loss": 0.0159, "step": 17260 }, { "epoch": 6.8, "learning_rate": 9.793601227338247e-08, "loss": 0.0129, "step": 17261 }, { "epoch": 6.8, "learning_rate": 9.791405503147651e-08, "loss": 0.0094, "step": 17262 }, { "epoch": 6.8, "learning_rate": 9.789209945341624e-08, "loss": 0.012, "step": 17263 }, { "epoch": 6.8, "learning_rate": 9.787014553955963e-08, "loss": 0.0101, "step": 17264 }, { "epoch": 6.8, "learning_rate": 9.784819329026428e-08, "loss": 0.0127, "step": 17265 }, { "epoch": 6.8, "learning_rate": 9.782624270588828e-08, "loss": 0.0128, "step": 17266 }, { "epoch": 6.81, "learning_rate": 9.780429378678916e-08, "loss": 0.0143, "step": 17267 }, { "epoch": 6.81, "learning_rate": 9.778234653332483e-08, "loss": 0.0101, "step": 17268 }, { "epoch": 6.81, "learning_rate": 9.776040094585278e-08, "loss": 0.012, "step": 17269 }, { "epoch": 6.81, "learning_rate": 9.773845702473085e-08, "loss": 0.0118, "step": 17270 }, { "epoch": 6.81, "learning_rate": 9.771651477031671e-08, "loss": 0.0121, "step": 17271 }, { "epoch": 6.81, "learning_rate": 9.769457418296784e-08, "loss": 0.0127, "step": 17272 }, { "epoch": 6.81, "learning_rate": 9.767263526304201e-08, "loss": 0.0106, "step": 17273 }, { "epoch": 6.81, "learning_rate": 9.765069801089656e-08, "loss": 0.0129, "step": 17274 }, { "epoch": 6.81, "learning_rate": 9.762876242688931e-08, "loss": 0.0113, "step": 17275 }, { "epoch": 6.81, "learning_rate": 9.760682851137758e-08, "loss": 0.0139, "step": 17276 }, { "epoch": 6.81, "learning_rate": 9.758489626471901e-08, "loss": 0.0101, "step": 17277 }, { "epoch": 6.81, "learning_rate": 9.756296568727082e-08, "loss": 0.0125, "step": 17278 }, { "epoch": 6.81, "learning_rate": 9.754103677939076e-08, "loss": 0.0135, "step": 17279 }, { "epoch": 6.81, "learning_rate": 9.7519109541436e-08, "loss": 0.0121, "step": 17280 }, { "epoch": 6.81, "learning_rate": 9.749718397376408e-08, "loss": 0.0106, "step": 17281 }, { "epoch": 6.81, "learning_rate": 9.747526007673222e-08, "loss": 0.0121, "step": 17282 }, { "epoch": 6.81, "learning_rate": 9.745333785069777e-08, "loss": 0.0106, "step": 17283 }, { "epoch": 6.81, "learning_rate": 9.743141729601807e-08, "loss": 0.0127, "step": 17284 }, { "epoch": 6.81, "learning_rate": 9.740949841305044e-08, "loss": 0.0125, "step": 17285 }, { "epoch": 6.81, "learning_rate": 9.7387581202152e-08, "loss": 0.0095, "step": 17286 }, { "epoch": 6.81, "learning_rate": 9.736566566368006e-08, "loss": 0.011, "step": 17287 }, { "epoch": 6.81, "learning_rate": 9.734375179799175e-08, "loss": 0.0109, "step": 17288 }, { "epoch": 6.81, "learning_rate": 9.732183960544436e-08, "loss": 0.0092, "step": 17289 }, { "epoch": 6.81, "learning_rate": 9.729992908639486e-08, "loss": 0.0113, "step": 17290 }, { "epoch": 6.81, "learning_rate": 9.727802024120049e-08, "loss": 0.0123, "step": 17291 }, { "epoch": 6.81, "learning_rate": 9.725611307021822e-08, "loss": 0.0118, "step": 17292 }, { "epoch": 6.82, "learning_rate": 9.723420757380511e-08, "loss": 0.0112, "step": 17293 }, { "epoch": 6.82, "learning_rate": 9.721230375231834e-08, "loss": 0.0124, "step": 17294 }, { "epoch": 6.82, "learning_rate": 9.719040160611471e-08, "loss": 0.0126, "step": 17295 }, { "epoch": 6.82, "learning_rate": 9.716850113555131e-08, "loss": 0.0109, "step": 17296 }, { "epoch": 6.82, "learning_rate": 9.714660234098501e-08, "loss": 0.0154, "step": 17297 }, { "epoch": 6.82, "learning_rate": 9.712470522277287e-08, "loss": 0.0119, "step": 17298 }, { "epoch": 6.82, "learning_rate": 9.710280978127161e-08, "loss": 0.0107, "step": 17299 }, { "epoch": 6.82, "learning_rate": 9.708091601683816e-08, "loss": 0.0137, "step": 17300 }, { "epoch": 6.82, "learning_rate": 9.705902392982939e-08, "loss": 0.0092, "step": 17301 }, { "epoch": 6.82, "learning_rate": 9.703713352060212e-08, "loss": 0.0167, "step": 17302 }, { "epoch": 6.82, "learning_rate": 9.701524478951301e-08, "loss": 0.0118, "step": 17303 }, { "epoch": 6.82, "learning_rate": 9.699335773691897e-08, "loss": 0.0107, "step": 17304 }, { "epoch": 6.82, "learning_rate": 9.697147236317651e-08, "loss": 0.0129, "step": 17305 }, { "epoch": 6.82, "learning_rate": 9.69495886686426e-08, "loss": 0.0087, "step": 17306 }, { "epoch": 6.82, "learning_rate": 9.692770665367371e-08, "loss": 0.0123, "step": 17307 }, { "epoch": 6.82, "learning_rate": 9.690582631862662e-08, "loss": 0.0096, "step": 17308 }, { "epoch": 6.82, "learning_rate": 9.688394766385781e-08, "loss": 0.0117, "step": 17309 }, { "epoch": 6.82, "learning_rate": 9.686207068972395e-08, "loss": 0.0114, "step": 17310 }, { "epoch": 6.82, "learning_rate": 9.684019539658155e-08, "loss": 0.0128, "step": 17311 }, { "epoch": 6.82, "learning_rate": 9.681832178478726e-08, "loss": 0.0105, "step": 17312 }, { "epoch": 6.82, "learning_rate": 9.679644985469746e-08, "loss": 0.008, "step": 17313 }, { "epoch": 6.82, "learning_rate": 9.677457960666863e-08, "loss": 0.0124, "step": 17314 }, { "epoch": 6.82, "learning_rate": 9.67527110410574e-08, "loss": 0.0135, "step": 17315 }, { "epoch": 6.82, "learning_rate": 9.673084415821994e-08, "loss": 0.0116, "step": 17316 }, { "epoch": 6.82, "learning_rate": 9.670897895851287e-08, "loss": 0.0117, "step": 17317 }, { "epoch": 6.83, "learning_rate": 9.668711544229231e-08, "loss": 0.0138, "step": 17318 }, { "epoch": 6.83, "learning_rate": 9.666525360991489e-08, "loss": 0.0111, "step": 17319 }, { "epoch": 6.83, "learning_rate": 9.664339346173671e-08, "loss": 0.0114, "step": 17320 }, { "epoch": 6.83, "learning_rate": 9.662153499811421e-08, "loss": 0.0152, "step": 17321 }, { "epoch": 6.83, "learning_rate": 9.659967821940349e-08, "loss": 0.0122, "step": 17322 }, { "epoch": 6.83, "learning_rate": 9.657782312596088e-08, "loss": 0.0113, "step": 17323 }, { "epoch": 6.83, "learning_rate": 9.655596971814255e-08, "loss": 0.0132, "step": 17324 }, { "epoch": 6.83, "learning_rate": 9.653411799630479e-08, "loss": 0.0092, "step": 17325 }, { "epoch": 6.83, "learning_rate": 9.651226796080358e-08, "loss": 0.0129, "step": 17326 }, { "epoch": 6.83, "learning_rate": 9.64904196119951e-08, "loss": 0.0118, "step": 17327 }, { "epoch": 6.83, "learning_rate": 9.646857295023549e-08, "loss": 0.0098, "step": 17328 }, { "epoch": 6.83, "learning_rate": 9.644672797588085e-08, "loss": 0.0126, "step": 17329 }, { "epoch": 6.83, "learning_rate": 9.642488468928708e-08, "loss": 0.0114, "step": 17330 }, { "epoch": 6.83, "learning_rate": 9.640304309081038e-08, "loss": 0.0112, "step": 17331 }, { "epoch": 6.83, "learning_rate": 9.638120318080652e-08, "loss": 0.0121, "step": 17332 }, { "epoch": 6.83, "learning_rate": 9.635936495963157e-08, "loss": 0.0101, "step": 17333 }, { "epoch": 6.83, "learning_rate": 9.633752842764147e-08, "loss": 0.0095, "step": 17334 }, { "epoch": 6.83, "learning_rate": 9.631569358519218e-08, "loss": 0.0116, "step": 17335 }, { "epoch": 6.83, "learning_rate": 9.629386043263943e-08, "loss": 0.0097, "step": 17336 }, { "epoch": 6.83, "learning_rate": 9.627202897033912e-08, "loss": 0.0094, "step": 17337 }, { "epoch": 6.83, "learning_rate": 9.625019919864718e-08, "loss": 0.0112, "step": 17338 }, { "epoch": 6.83, "learning_rate": 9.622837111791924e-08, "loss": 0.0134, "step": 17339 }, { "epoch": 6.83, "learning_rate": 9.620654472851113e-08, "loss": 0.0128, "step": 17340 }, { "epoch": 6.83, "learning_rate": 9.61847200307786e-08, "loss": 0.011, "step": 17341 }, { "epoch": 6.83, "learning_rate": 9.61628970250774e-08, "loss": 0.0118, "step": 17342 }, { "epoch": 6.84, "learning_rate": 9.614107571176313e-08, "loss": 0.0145, "step": 17343 }, { "epoch": 6.84, "learning_rate": 9.611925609119153e-08, "loss": 0.0116, "step": 17344 }, { "epoch": 6.84, "learning_rate": 9.609743816371803e-08, "loss": 0.0141, "step": 17345 }, { "epoch": 6.84, "learning_rate": 9.607562192969851e-08, "loss": 0.0128, "step": 17346 }, { "epoch": 6.84, "learning_rate": 9.605380738948836e-08, "loss": 0.0104, "step": 17347 }, { "epoch": 6.84, "learning_rate": 9.603199454344322e-08, "loss": 0.0145, "step": 17348 }, { "epoch": 6.84, "learning_rate": 9.601018339191843e-08, "loss": 0.011, "step": 17349 }, { "epoch": 6.84, "learning_rate": 9.598837393526973e-08, "loss": 0.0136, "step": 17350 }, { "epoch": 6.84, "learning_rate": 9.596656617385239e-08, "loss": 0.0119, "step": 17351 }, { "epoch": 6.84, "learning_rate": 9.594476010802197e-08, "loss": 0.0124, "step": 17352 }, { "epoch": 6.84, "learning_rate": 9.592295573813376e-08, "loss": 0.0133, "step": 17353 }, { "epoch": 6.84, "learning_rate": 9.590115306454319e-08, "loss": 0.0129, "step": 17354 }, { "epoch": 6.84, "learning_rate": 9.587935208760567e-08, "loss": 0.0119, "step": 17355 }, { "epoch": 6.84, "learning_rate": 9.58575528076764e-08, "loss": 0.0102, "step": 17356 }, { "epoch": 6.84, "learning_rate": 9.583575522511076e-08, "loss": 0.0111, "step": 17357 }, { "epoch": 6.84, "learning_rate": 9.581395934026397e-08, "loss": 0.0114, "step": 17358 }, { "epoch": 6.84, "learning_rate": 9.57921651534914e-08, "loss": 0.0094, "step": 17359 }, { "epoch": 6.84, "learning_rate": 9.577037266514807e-08, "loss": 0.009, "step": 17360 }, { "epoch": 6.84, "learning_rate": 9.574858187558933e-08, "loss": 0.0133, "step": 17361 }, { "epoch": 6.84, "learning_rate": 9.57267927851702e-08, "loss": 0.0109, "step": 17362 }, { "epoch": 6.84, "learning_rate": 9.570500539424585e-08, "loss": 0.0128, "step": 17363 }, { "epoch": 6.84, "learning_rate": 9.568321970317141e-08, "loss": 0.0132, "step": 17364 }, { "epoch": 6.84, "learning_rate": 9.566143571230201e-08, "loss": 0.0085, "step": 17365 }, { "epoch": 6.84, "learning_rate": 9.563965342199258e-08, "loss": 0.0115, "step": 17366 }, { "epoch": 6.84, "learning_rate": 9.561787283259817e-08, "loss": 0.0115, "step": 17367 }, { "epoch": 6.84, "learning_rate": 9.559609394447382e-08, "loss": 0.0091, "step": 17368 }, { "epoch": 6.85, "learning_rate": 9.557431675797449e-08, "loss": 0.0118, "step": 17369 }, { "epoch": 6.85, "learning_rate": 9.555254127345504e-08, "loss": 0.0105, "step": 17370 }, { "epoch": 6.85, "learning_rate": 9.553076749127052e-08, "loss": 0.0122, "step": 17371 }, { "epoch": 6.85, "learning_rate": 9.550899541177553e-08, "loss": 0.0102, "step": 17372 }, { "epoch": 6.85, "learning_rate": 9.548722503532527e-08, "loss": 0.0127, "step": 17373 }, { "epoch": 6.85, "learning_rate": 9.546545636227433e-08, "loss": 0.0127, "step": 17374 }, { "epoch": 6.85, "learning_rate": 9.544368939297764e-08, "loss": 0.0137, "step": 17375 }, { "epoch": 6.85, "learning_rate": 9.542192412778984e-08, "loss": 0.0113, "step": 17376 }, { "epoch": 6.85, "learning_rate": 9.540016056706575e-08, "loss": 0.0147, "step": 17377 }, { "epoch": 6.85, "learning_rate": 9.537839871116011e-08, "loss": 0.012, "step": 17378 }, { "epoch": 6.85, "learning_rate": 9.535663856042753e-08, "loss": 0.0134, "step": 17379 }, { "epoch": 6.85, "learning_rate": 9.533488011522269e-08, "loss": 0.0113, "step": 17380 }, { "epoch": 6.85, "learning_rate": 9.531312337590021e-08, "loss": 0.0116, "step": 17381 }, { "epoch": 6.85, "learning_rate": 9.529136834281479e-08, "loss": 0.0123, "step": 17382 }, { "epoch": 6.85, "learning_rate": 9.526961501632087e-08, "loss": 0.0096, "step": 17383 }, { "epoch": 6.85, "learning_rate": 9.524786339677313e-08, "loss": 0.0115, "step": 17384 }, { "epoch": 6.85, "learning_rate": 9.522611348452586e-08, "loss": 0.0146, "step": 17385 }, { "epoch": 6.85, "learning_rate": 9.520436527993382e-08, "loss": 0.011, "step": 17386 }, { "epoch": 6.85, "learning_rate": 9.518261878335132e-08, "loss": 0.0113, "step": 17387 }, { "epoch": 6.85, "learning_rate": 9.516087399513289e-08, "loss": 0.0112, "step": 17388 }, { "epoch": 6.85, "learning_rate": 9.513913091563273e-08, "loss": 0.0107, "step": 17389 }, { "epoch": 6.85, "learning_rate": 9.511738954520551e-08, "loss": 0.0096, "step": 17390 }, { "epoch": 6.85, "learning_rate": 9.509564988420538e-08, "loss": 0.0119, "step": 17391 }, { "epoch": 6.85, "learning_rate": 9.507391193298677e-08, "loss": 0.0133, "step": 17392 }, { "epoch": 6.85, "learning_rate": 9.505217569190387e-08, "loss": 0.011, "step": 17393 }, { "epoch": 6.86, "learning_rate": 9.503044116131098e-08, "loss": 0.0117, "step": 17394 }, { "epoch": 6.86, "learning_rate": 9.500870834156238e-08, "loss": 0.0125, "step": 17395 }, { "epoch": 6.86, "learning_rate": 9.498697723301234e-08, "loss": 0.0128, "step": 17396 }, { "epoch": 6.86, "learning_rate": 9.496524783601489e-08, "loss": 0.0129, "step": 17397 }, { "epoch": 6.86, "learning_rate": 9.494352015092429e-08, "loss": 0.0124, "step": 17398 }, { "epoch": 6.86, "learning_rate": 9.492179417809469e-08, "loss": 0.0128, "step": 17399 }, { "epoch": 6.86, "learning_rate": 9.490006991788009e-08, "loss": 0.0099, "step": 17400 }, { "epoch": 6.86, "learning_rate": 9.487834737063467e-08, "loss": 0.013, "step": 17401 }, { "epoch": 6.86, "learning_rate": 9.485662653671236e-08, "loss": 0.0138, "step": 17402 }, { "epoch": 6.86, "learning_rate": 9.483490741646725e-08, "loss": 0.0114, "step": 17403 }, { "epoch": 6.86, "learning_rate": 9.48131900102533e-08, "loss": 0.0098, "step": 17404 }, { "epoch": 6.86, "learning_rate": 9.479147431842456e-08, "loss": 0.012, "step": 17405 }, { "epoch": 6.86, "learning_rate": 9.476976034133482e-08, "loss": 0.0152, "step": 17406 }, { "epoch": 6.86, "learning_rate": 9.474804807933805e-08, "loss": 0.0107, "step": 17407 }, { "epoch": 6.86, "learning_rate": 9.472633753278813e-08, "loss": 0.0113, "step": 17408 }, { "epoch": 6.86, "learning_rate": 9.470462870203899e-08, "loss": 0.0141, "step": 17409 }, { "epoch": 6.86, "learning_rate": 9.468292158744431e-08, "loss": 0.0141, "step": 17410 }, { "epoch": 6.86, "learning_rate": 9.4661216189358e-08, "loss": 0.0134, "step": 17411 }, { "epoch": 6.86, "learning_rate": 9.463951250813364e-08, "loss": 0.0146, "step": 17412 }, { "epoch": 6.86, "learning_rate": 9.461781054412524e-08, "loss": 0.0104, "step": 17413 }, { "epoch": 6.86, "learning_rate": 9.459611029768629e-08, "loss": 0.0123, "step": 17414 }, { "epoch": 6.86, "learning_rate": 9.45744117691706e-08, "loss": 0.0114, "step": 17415 }, { "epoch": 6.86, "learning_rate": 9.455271495893165e-08, "loss": 0.0112, "step": 17416 }, { "epoch": 6.86, "learning_rate": 9.453101986732331e-08, "loss": 0.0131, "step": 17417 }, { "epoch": 6.86, "learning_rate": 9.450932649469899e-08, "loss": 0.0122, "step": 17418 }, { "epoch": 6.86, "learning_rate": 9.44876348414124e-08, "loss": 0.0109, "step": 17419 }, { "epoch": 6.87, "learning_rate": 9.446594490781692e-08, "loss": 0.0121, "step": 17420 }, { "epoch": 6.87, "learning_rate": 9.444425669426615e-08, "loss": 0.0113, "step": 17421 }, { "epoch": 6.87, "learning_rate": 9.442257020111364e-08, "loss": 0.0095, "step": 17422 }, { "epoch": 6.87, "learning_rate": 9.44008854287127e-08, "loss": 0.0104, "step": 17423 }, { "epoch": 6.87, "learning_rate": 9.43792023774169e-08, "loss": 0.0142, "step": 17424 }, { "epoch": 6.87, "learning_rate": 9.435752104757944e-08, "loss": 0.0128, "step": 17425 }, { "epoch": 6.87, "learning_rate": 9.433584143955398e-08, "loss": 0.0122, "step": 17426 }, { "epoch": 6.87, "learning_rate": 9.431416355369363e-08, "loss": 0.0121, "step": 17427 }, { "epoch": 6.87, "learning_rate": 9.429248739035187e-08, "loss": 0.0121, "step": 17428 }, { "epoch": 6.87, "learning_rate": 9.427081294988177e-08, "loss": 0.0145, "step": 17429 }, { "epoch": 6.87, "learning_rate": 9.424914023263686e-08, "loss": 0.012, "step": 17430 }, { "epoch": 6.87, "learning_rate": 9.422746923897018e-08, "loss": 0.0121, "step": 17431 }, { "epoch": 6.87, "learning_rate": 9.420579996923505e-08, "loss": 0.0124, "step": 17432 }, { "epoch": 6.87, "learning_rate": 9.418413242378455e-08, "loss": 0.0102, "step": 17433 }, { "epoch": 6.87, "learning_rate": 9.416246660297184e-08, "loss": 0.0168, "step": 17434 }, { "epoch": 6.87, "learning_rate": 9.414080250715009e-08, "loss": 0.0111, "step": 17435 }, { "epoch": 6.87, "learning_rate": 9.411914013667243e-08, "loss": 0.0121, "step": 17436 }, { "epoch": 6.87, "learning_rate": 9.409747949189182e-08, "loss": 0.0114, "step": 17437 }, { "epoch": 6.87, "learning_rate": 9.40758205731613e-08, "loss": 0.0151, "step": 17438 }, { "epoch": 6.87, "learning_rate": 9.405416338083394e-08, "loss": 0.0108, "step": 17439 }, { "epoch": 6.87, "learning_rate": 9.403250791526278e-08, "loss": 0.0114, "step": 17440 }, { "epoch": 6.87, "learning_rate": 9.401085417680061e-08, "loss": 0.0125, "step": 17441 }, { "epoch": 6.87, "learning_rate": 9.39892021658005e-08, "loss": 0.0149, "step": 17442 }, { "epoch": 6.87, "learning_rate": 9.39675518826152e-08, "loss": 0.0142, "step": 17443 }, { "epoch": 6.87, "learning_rate": 9.394590332759766e-08, "loss": 0.0116, "step": 17444 }, { "epoch": 6.88, "learning_rate": 9.392425650110079e-08, "loss": 0.0098, "step": 17445 }, { "epoch": 6.88, "learning_rate": 9.390261140347724e-08, "loss": 0.0109, "step": 17446 }, { "epoch": 6.88, "learning_rate": 9.388096803507988e-08, "loss": 0.0137, "step": 17447 }, { "epoch": 6.88, "learning_rate": 9.385932639626144e-08, "loss": 0.0109, "step": 17448 }, { "epoch": 6.88, "learning_rate": 9.383768648737476e-08, "loss": 0.0112, "step": 17449 }, { "epoch": 6.88, "learning_rate": 9.381604830877236e-08, "loss": 0.0145, "step": 17450 }, { "epoch": 6.88, "learning_rate": 9.379441186080707e-08, "loss": 0.0117, "step": 17451 }, { "epoch": 6.88, "learning_rate": 9.377277714383132e-08, "loss": 0.0113, "step": 17452 }, { "epoch": 6.88, "learning_rate": 9.375114415819798e-08, "loss": 0.011, "step": 17453 }, { "epoch": 6.88, "learning_rate": 9.372951290425943e-08, "loss": 0.0092, "step": 17454 }, { "epoch": 6.88, "learning_rate": 9.370788338236841e-08, "loss": 0.0123, "step": 17455 }, { "epoch": 6.88, "learning_rate": 9.368625559287717e-08, "loss": 0.0119, "step": 17456 }, { "epoch": 6.88, "learning_rate": 9.366462953613856e-08, "loss": 0.0148, "step": 17457 }, { "epoch": 6.88, "learning_rate": 9.364300521250478e-08, "loss": 0.0124, "step": 17458 }, { "epoch": 6.88, "learning_rate": 9.362138262232846e-08, "loss": 0.0103, "step": 17459 }, { "epoch": 6.88, "learning_rate": 9.359976176596185e-08, "loss": 0.011, "step": 17460 }, { "epoch": 6.88, "learning_rate": 9.35781426437574e-08, "loss": 0.0128, "step": 17461 }, { "epoch": 6.88, "learning_rate": 9.35565252560675e-08, "loss": 0.0119, "step": 17462 }, { "epoch": 6.88, "learning_rate": 9.353490960324453e-08, "loss": 0.0114, "step": 17463 }, { "epoch": 6.88, "learning_rate": 9.351329568564064e-08, "loss": 0.0133, "step": 17464 }, { "epoch": 6.88, "learning_rate": 9.34916835036082e-08, "loss": 0.0106, "step": 17465 }, { "epoch": 6.88, "learning_rate": 9.347007305749952e-08, "loss": 0.01, "step": 17466 }, { "epoch": 6.88, "learning_rate": 9.344846434766666e-08, "loss": 0.0113, "step": 17467 }, { "epoch": 6.88, "learning_rate": 9.342685737446198e-08, "loss": 0.0108, "step": 17468 }, { "epoch": 6.88, "learning_rate": 9.34052521382374e-08, "loss": 0.0097, "step": 17469 }, { "epoch": 6.89, "learning_rate": 9.338364863934533e-08, "loss": 0.0115, "step": 17470 }, { "epoch": 6.89, "learning_rate": 9.336204687813771e-08, "loss": 0.0117, "step": 17471 }, { "epoch": 6.89, "learning_rate": 9.334044685496671e-08, "loss": 0.0148, "step": 17472 }, { "epoch": 6.89, "learning_rate": 9.331884857018426e-08, "loss": 0.0134, "step": 17473 }, { "epoch": 6.89, "learning_rate": 9.329725202414243e-08, "loss": 0.0106, "step": 17474 }, { "epoch": 6.89, "learning_rate": 9.327565721719322e-08, "loss": 0.0144, "step": 17475 }, { "epoch": 6.89, "learning_rate": 9.325406414968868e-08, "loss": 0.0089, "step": 17476 }, { "epoch": 6.89, "learning_rate": 9.323247282198059e-08, "loss": 0.0108, "step": 17477 }, { "epoch": 6.89, "learning_rate": 9.321088323442091e-08, "loss": 0.0123, "step": 17478 }, { "epoch": 6.89, "learning_rate": 9.318929538736153e-08, "loss": 0.0101, "step": 17479 }, { "epoch": 6.89, "learning_rate": 9.316770928115438e-08, "loss": 0.011, "step": 17480 }, { "epoch": 6.89, "learning_rate": 9.314612491615112e-08, "loss": 0.0108, "step": 17481 }, { "epoch": 6.89, "learning_rate": 9.312454229270372e-08, "loss": 0.013, "step": 17482 }, { "epoch": 6.89, "learning_rate": 9.310296141116378e-08, "loss": 0.0101, "step": 17483 }, { "epoch": 6.89, "learning_rate": 9.308138227188309e-08, "loss": 0.0146, "step": 17484 }, { "epoch": 6.89, "learning_rate": 9.305980487521337e-08, "loss": 0.012, "step": 17485 }, { "epoch": 6.89, "learning_rate": 9.303822922150637e-08, "loss": 0.012, "step": 17486 }, { "epoch": 6.89, "learning_rate": 9.30166553111136e-08, "loss": 0.0078, "step": 17487 }, { "epoch": 6.89, "learning_rate": 9.299508314438675e-08, "loss": 0.0121, "step": 17488 }, { "epoch": 6.89, "learning_rate": 9.297351272167751e-08, "loss": 0.0149, "step": 17489 }, { "epoch": 6.89, "learning_rate": 9.295194404333724e-08, "loss": 0.0128, "step": 17490 }, { "epoch": 6.89, "learning_rate": 9.293037710971769e-08, "loss": 0.0097, "step": 17491 }, { "epoch": 6.89, "learning_rate": 9.29088119211701e-08, "loss": 0.0142, "step": 17492 }, { "epoch": 6.89, "learning_rate": 9.288724847804627e-08, "loss": 0.0142, "step": 17493 }, { "epoch": 6.89, "learning_rate": 9.286568678069743e-08, "loss": 0.0118, "step": 17494 }, { "epoch": 6.89, "learning_rate": 9.284412682947516e-08, "loss": 0.0106, "step": 17495 }, { "epoch": 6.9, "learning_rate": 9.282256862473058e-08, "loss": 0.0138, "step": 17496 }, { "epoch": 6.9, "learning_rate": 9.280101216681541e-08, "loss": 0.0116, "step": 17497 }, { "epoch": 6.9, "learning_rate": 9.277945745608075e-08, "loss": 0.0121, "step": 17498 }, { "epoch": 6.9, "learning_rate": 9.275790449287804e-08, "loss": 0.014, "step": 17499 }, { "epoch": 6.9, "learning_rate": 9.273635327755835e-08, "loss": 0.0122, "step": 17500 }, { "epoch": 6.9, "learning_rate": 9.271480381047323e-08, "loss": 0.0125, "step": 17501 }, { "epoch": 6.9, "learning_rate": 9.269325609197367e-08, "loss": 0.0125, "step": 17502 }, { "epoch": 6.9, "learning_rate": 9.267171012241104e-08, "loss": 0.01, "step": 17503 }, { "epoch": 6.9, "learning_rate": 9.265016590213632e-08, "loss": 0.0144, "step": 17504 }, { "epoch": 6.9, "learning_rate": 9.262862343150078e-08, "loss": 0.0105, "step": 17505 }, { "epoch": 6.9, "learning_rate": 9.260708271085554e-08, "loss": 0.0119, "step": 17506 }, { "epoch": 6.9, "learning_rate": 9.258554374055154e-08, "loss": 0.0118, "step": 17507 }, { "epoch": 6.9, "learning_rate": 9.256400652093994e-08, "loss": 0.0113, "step": 17508 }, { "epoch": 6.9, "learning_rate": 9.254247105237177e-08, "loss": 0.0101, "step": 17509 }, { "epoch": 6.9, "learning_rate": 9.252093733519806e-08, "loss": 0.0116, "step": 17510 }, { "epoch": 6.9, "learning_rate": 9.249940536976965e-08, "loss": 0.0094, "step": 17511 }, { "epoch": 6.9, "learning_rate": 9.247787515643764e-08, "loss": 0.01, "step": 17512 }, { "epoch": 6.9, "learning_rate": 9.245634669555275e-08, "loss": 0.013, "step": 17513 }, { "epoch": 6.9, "learning_rate": 9.243481998746598e-08, "loss": 0.0098, "step": 17514 }, { "epoch": 6.9, "learning_rate": 9.241329503252815e-08, "loss": 0.0115, "step": 17515 }, { "epoch": 6.9, "learning_rate": 9.239177183109019e-08, "loss": 0.0112, "step": 17516 }, { "epoch": 6.9, "learning_rate": 9.237025038350272e-08, "loss": 0.0124, "step": 17517 }, { "epoch": 6.9, "learning_rate": 9.234873069011659e-08, "loss": 0.0097, "step": 17518 }, { "epoch": 6.9, "learning_rate": 9.232721275128255e-08, "loss": 0.012, "step": 17519 }, { "epoch": 6.9, "learning_rate": 9.230569656735134e-08, "loss": 0.0112, "step": 17520 }, { "epoch": 6.91, "learning_rate": 9.228418213867355e-08, "loss": 0.0115, "step": 17521 }, { "epoch": 6.91, "learning_rate": 9.226266946559996e-08, "loss": 0.0128, "step": 17522 }, { "epoch": 6.91, "learning_rate": 9.224115854848095e-08, "loss": 0.0113, "step": 17523 }, { "epoch": 6.91, "learning_rate": 9.221964938766744e-08, "loss": 0.0115, "step": 17524 }, { "epoch": 6.91, "learning_rate": 9.219814198350977e-08, "loss": 0.0123, "step": 17525 }, { "epoch": 6.91, "learning_rate": 9.217663633635862e-08, "loss": 0.0127, "step": 17526 }, { "epoch": 6.91, "learning_rate": 9.215513244656434e-08, "loss": 0.0139, "step": 17527 }, { "epoch": 6.91, "learning_rate": 9.213363031447749e-08, "loss": 0.0123, "step": 17528 }, { "epoch": 6.91, "learning_rate": 9.211212994044864e-08, "loss": 0.0134, "step": 17529 }, { "epoch": 6.91, "learning_rate": 9.209063132482799e-08, "loss": 0.0126, "step": 17530 }, { "epoch": 6.91, "learning_rate": 9.206913446796605e-08, "loss": 0.013, "step": 17531 }, { "epoch": 6.91, "learning_rate": 9.204763937021319e-08, "loss": 0.0104, "step": 17532 }, { "epoch": 6.91, "learning_rate": 9.202614603191981e-08, "loss": 0.0098, "step": 17533 }, { "epoch": 6.91, "learning_rate": 9.200465445343608e-08, "loss": 0.0153, "step": 17534 }, { "epoch": 6.91, "learning_rate": 9.198316463511242e-08, "loss": 0.0111, "step": 17535 }, { "epoch": 6.91, "learning_rate": 9.196167657729885e-08, "loss": 0.0114, "step": 17536 }, { "epoch": 6.91, "learning_rate": 9.194019028034592e-08, "loss": 0.0114, "step": 17537 }, { "epoch": 6.91, "learning_rate": 9.191870574460356e-08, "loss": 0.0146, "step": 17538 }, { "epoch": 6.91, "learning_rate": 9.189722297042213e-08, "loss": 0.0105, "step": 17539 }, { "epoch": 6.91, "learning_rate": 9.187574195815152e-08, "loss": 0.011, "step": 17540 }, { "epoch": 6.91, "learning_rate": 9.185426270814209e-08, "loss": 0.0092, "step": 17541 }, { "epoch": 6.91, "learning_rate": 9.183278522074377e-08, "loss": 0.0134, "step": 17542 }, { "epoch": 6.91, "learning_rate": 9.181130949630675e-08, "loss": 0.0131, "step": 17543 }, { "epoch": 6.91, "learning_rate": 9.178983553518087e-08, "loss": 0.0101, "step": 17544 }, { "epoch": 6.91, "learning_rate": 9.17683633377162e-08, "loss": 0.0097, "step": 17545 }, { "epoch": 6.92, "learning_rate": 9.174689290426271e-08, "loss": 0.0114, "step": 17546 }, { "epoch": 6.92, "learning_rate": 9.172542423517045e-08, "loss": 0.0136, "step": 17547 }, { "epoch": 6.92, "learning_rate": 9.17039573307891e-08, "loss": 0.0101, "step": 17548 }, { "epoch": 6.92, "learning_rate": 9.168249219146869e-08, "loss": 0.0115, "step": 17549 }, { "epoch": 6.92, "learning_rate": 9.166102881755908e-08, "loss": 0.0113, "step": 17550 }, { "epoch": 6.92, "learning_rate": 9.163956720940999e-08, "loss": 0.0128, "step": 17551 }, { "epoch": 6.92, "learning_rate": 9.161810736737132e-08, "loss": 0.0124, "step": 17552 }, { "epoch": 6.92, "learning_rate": 9.159664929179272e-08, "loss": 0.0129, "step": 17553 }, { "epoch": 6.92, "learning_rate": 9.157519298302399e-08, "loss": 0.0105, "step": 17554 }, { "epoch": 6.92, "learning_rate": 9.155373844141481e-08, "loss": 0.0121, "step": 17555 }, { "epoch": 6.92, "learning_rate": 9.153228566731494e-08, "loss": 0.0108, "step": 17556 }, { "epoch": 6.92, "learning_rate": 9.15108346610739e-08, "loss": 0.0117, "step": 17557 }, { "epoch": 6.92, "learning_rate": 9.14893854230414e-08, "loss": 0.0115, "step": 17558 }, { "epoch": 6.92, "learning_rate": 9.146793795356696e-08, "loss": 0.0111, "step": 17559 }, { "epoch": 6.92, "learning_rate": 9.144649225300025e-08, "loss": 0.0123, "step": 17560 }, { "epoch": 6.92, "learning_rate": 9.142504832169069e-08, "loss": 0.0138, "step": 17561 }, { "epoch": 6.92, "learning_rate": 9.140360615998788e-08, "loss": 0.0116, "step": 17562 }, { "epoch": 6.92, "learning_rate": 9.138216576824109e-08, "loss": 0.0095, "step": 17563 }, { "epoch": 6.92, "learning_rate": 9.13607271468001e-08, "loss": 0.0123, "step": 17564 }, { "epoch": 6.92, "learning_rate": 9.133929029601403e-08, "loss": 0.0105, "step": 17565 }, { "epoch": 6.92, "learning_rate": 9.13178552162325e-08, "loss": 0.0113, "step": 17566 }, { "epoch": 6.92, "learning_rate": 9.129642190780457e-08, "loss": 0.0124, "step": 17567 }, { "epoch": 6.92, "learning_rate": 9.127499037107991e-08, "loss": 0.0102, "step": 17568 }, { "epoch": 6.92, "learning_rate": 9.125356060640759e-08, "loss": 0.0109, "step": 17569 }, { "epoch": 6.92, "learning_rate": 9.123213261413702e-08, "loss": 0.01, "step": 17570 }, { "epoch": 6.92, "learning_rate": 9.121070639461732e-08, "loss": 0.0159, "step": 17571 }, { "epoch": 6.93, "learning_rate": 9.118928194819777e-08, "loss": 0.0127, "step": 17572 }, { "epoch": 6.93, "learning_rate": 9.116785927522761e-08, "loss": 0.0123, "step": 17573 }, { "epoch": 6.93, "learning_rate": 9.11464383760559e-08, "loss": 0.0106, "step": 17574 }, { "epoch": 6.93, "learning_rate": 9.112501925103176e-08, "loss": 0.0103, "step": 17575 }, { "epoch": 6.93, "learning_rate": 9.110360190050435e-08, "loss": 0.0105, "step": 17576 }, { "epoch": 6.93, "learning_rate": 9.108218632482281e-08, "loss": 0.0092, "step": 17577 }, { "epoch": 6.93, "learning_rate": 9.106077252433602e-08, "loss": 0.0142, "step": 17578 }, { "epoch": 6.93, "learning_rate": 9.103936049939314e-08, "loss": 0.0113, "step": 17579 }, { "epoch": 6.93, "learning_rate": 9.101795025034291e-08, "loss": 0.0134, "step": 17580 }, { "epoch": 6.93, "learning_rate": 9.099654177753464e-08, "loss": 0.0134, "step": 17581 }, { "epoch": 6.93, "learning_rate": 9.0975135081317e-08, "loss": 0.0144, "step": 17582 }, { "epoch": 6.93, "learning_rate": 9.095373016203903e-08, "loss": 0.0133, "step": 17583 }, { "epoch": 6.93, "learning_rate": 9.093232702004942e-08, "loss": 0.0102, "step": 17584 }, { "epoch": 6.93, "learning_rate": 9.091092565569713e-08, "loss": 0.0104, "step": 17585 }, { "epoch": 6.93, "learning_rate": 9.088952606933097e-08, "loss": 0.0102, "step": 17586 }, { "epoch": 6.93, "learning_rate": 9.08681282612998e-08, "loss": 0.014, "step": 17587 }, { "epoch": 6.93, "learning_rate": 9.084673223195216e-08, "loss": 0.0091, "step": 17588 }, { "epoch": 6.93, "learning_rate": 9.082533798163692e-08, "loss": 0.0137, "step": 17589 }, { "epoch": 6.93, "learning_rate": 9.080394551070274e-08, "loss": 0.0102, "step": 17590 }, { "epoch": 6.93, "learning_rate": 9.078255481949835e-08, "loss": 0.0104, "step": 17591 }, { "epoch": 6.93, "learning_rate": 9.076116590837226e-08, "loss": 0.0113, "step": 17592 }, { "epoch": 6.93, "learning_rate": 9.073977877767322e-08, "loss": 0.0121, "step": 17593 }, { "epoch": 6.93, "learning_rate": 9.071839342774965e-08, "loss": 0.0093, "step": 17594 }, { "epoch": 6.93, "learning_rate": 9.069700985895017e-08, "loss": 0.0133, "step": 17595 }, { "epoch": 6.93, "learning_rate": 9.067562807162338e-08, "loss": 0.0126, "step": 17596 }, { "epoch": 6.94, "learning_rate": 9.065424806611766e-08, "loss": 0.011, "step": 17597 }, { "epoch": 6.94, "learning_rate": 9.063286984278149e-08, "loss": 0.0147, "step": 17598 }, { "epoch": 6.94, "learning_rate": 9.06114934019633e-08, "loss": 0.0123, "step": 17599 }, { "epoch": 6.94, "learning_rate": 9.059011874401161e-08, "loss": 0.0126, "step": 17600 }, { "epoch": 6.94, "learning_rate": 9.056874586927463e-08, "loss": 0.0098, "step": 17601 }, { "epoch": 6.94, "learning_rate": 9.054737477810085e-08, "loss": 0.0111, "step": 17602 }, { "epoch": 6.94, "learning_rate": 9.052600547083836e-08, "loss": 0.011, "step": 17603 }, { "epoch": 6.94, "learning_rate": 9.050463794783574e-08, "loss": 0.013, "step": 17604 }, { "epoch": 6.94, "learning_rate": 9.048327220944106e-08, "loss": 0.0102, "step": 17605 }, { "epoch": 6.94, "learning_rate": 9.046190825600264e-08, "loss": 0.0122, "step": 17606 }, { "epoch": 6.94, "learning_rate": 9.04405460878685e-08, "loss": 0.0135, "step": 17607 }, { "epoch": 6.94, "learning_rate": 9.041918570538709e-08, "loss": 0.0154, "step": 17608 }, { "epoch": 6.94, "learning_rate": 9.039782710890636e-08, "loss": 0.0118, "step": 17609 }, { "epoch": 6.94, "learning_rate": 9.037647029877451e-08, "loss": 0.0116, "step": 17610 }, { "epoch": 6.94, "learning_rate": 9.035511527533952e-08, "loss": 0.0128, "step": 17611 }, { "epoch": 6.94, "learning_rate": 9.033376203894953e-08, "loss": 0.0136, "step": 17612 }, { "epoch": 6.94, "learning_rate": 9.031241058995253e-08, "loss": 0.0116, "step": 17613 }, { "epoch": 6.94, "learning_rate": 9.02910609286966e-08, "loss": 0.0161, "step": 17614 }, { "epoch": 6.94, "learning_rate": 9.026971305552959e-08, "loss": 0.0109, "step": 17615 }, { "epoch": 6.94, "learning_rate": 9.024836697079948e-08, "loss": 0.0112, "step": 17616 }, { "epoch": 6.94, "learning_rate": 9.022702267485425e-08, "loss": 0.0119, "step": 17617 }, { "epoch": 6.94, "learning_rate": 9.020568016804162e-08, "loss": 0.0111, "step": 17618 }, { "epoch": 6.94, "learning_rate": 9.018433945070964e-08, "loss": 0.0128, "step": 17619 }, { "epoch": 6.94, "learning_rate": 9.016300052320587e-08, "loss": 0.0132, "step": 17620 }, { "epoch": 6.94, "learning_rate": 9.01416633858784e-08, "loss": 0.0136, "step": 17621 }, { "epoch": 6.94, "learning_rate": 9.012032803907478e-08, "loss": 0.0116, "step": 17622 }, { "epoch": 6.95, "learning_rate": 9.009899448314289e-08, "loss": 0.0125, "step": 17623 }, { "epoch": 6.95, "learning_rate": 9.007766271843029e-08, "loss": 0.0116, "step": 17624 }, { "epoch": 6.95, "learning_rate": 9.005633274528472e-08, "loss": 0.0138, "step": 17625 }, { "epoch": 6.95, "learning_rate": 9.003500456405384e-08, "loss": 0.0092, "step": 17626 }, { "epoch": 6.95, "learning_rate": 9.001367817508532e-08, "loss": 0.0121, "step": 17627 }, { "epoch": 6.95, "learning_rate": 8.99923535787266e-08, "loss": 0.0105, "step": 17628 }, { "epoch": 6.95, "learning_rate": 8.997103077532536e-08, "loss": 0.0131, "step": 17629 }, { "epoch": 6.95, "learning_rate": 8.994970976522907e-08, "loss": 0.0117, "step": 17630 }, { "epoch": 6.95, "learning_rate": 8.992839054878534e-08, "loss": 0.0145, "step": 17631 }, { "epoch": 6.95, "learning_rate": 8.990707312634147e-08, "loss": 0.0108, "step": 17632 }, { "epoch": 6.95, "learning_rate": 8.988575749824505e-08, "loss": 0.012, "step": 17633 }, { "epoch": 6.95, "learning_rate": 8.986444366484338e-08, "loss": 0.011, "step": 17634 }, { "epoch": 6.95, "learning_rate": 8.984313162648388e-08, "loss": 0.0131, "step": 17635 }, { "epoch": 6.95, "learning_rate": 8.982182138351391e-08, "loss": 0.0089, "step": 17636 }, { "epoch": 6.95, "learning_rate": 8.980051293628088e-08, "loss": 0.0122, "step": 17637 }, { "epoch": 6.95, "learning_rate": 8.977920628513192e-08, "loss": 0.0128, "step": 17638 }, { "epoch": 6.95, "learning_rate": 8.975790143041439e-08, "loss": 0.0109, "step": 17639 }, { "epoch": 6.95, "learning_rate": 8.973659837247558e-08, "loss": 0.0121, "step": 17640 }, { "epoch": 6.95, "learning_rate": 8.971529711166259e-08, "loss": 0.0107, "step": 17641 }, { "epoch": 6.95, "learning_rate": 8.969399764832271e-08, "loss": 0.0137, "step": 17642 }, { "epoch": 6.95, "learning_rate": 8.967269998280284e-08, "loss": 0.0114, "step": 17643 }, { "epoch": 6.95, "learning_rate": 8.965140411545047e-08, "loss": 0.0118, "step": 17644 }, { "epoch": 6.95, "learning_rate": 8.96301100466124e-08, "loss": 0.0099, "step": 17645 }, { "epoch": 6.95, "learning_rate": 8.96088177766359e-08, "loss": 0.0119, "step": 17646 }, { "epoch": 6.95, "learning_rate": 8.958752730586771e-08, "loss": 0.0095, "step": 17647 }, { "epoch": 6.96, "learning_rate": 8.95662386346552e-08, "loss": 0.0111, "step": 17648 }, { "epoch": 6.96, "learning_rate": 8.954495176334507e-08, "loss": 0.0138, "step": 17649 }, { "epoch": 6.96, "learning_rate": 8.952366669228442e-08, "loss": 0.0106, "step": 17650 }, { "epoch": 6.96, "learning_rate": 8.950238342182004e-08, "loss": 0.0123, "step": 17651 }, { "epoch": 6.96, "learning_rate": 8.948110195229885e-08, "loss": 0.0103, "step": 17652 }, { "epoch": 6.96, "learning_rate": 8.945982228406777e-08, "loss": 0.0132, "step": 17653 }, { "epoch": 6.96, "learning_rate": 8.943854441747363e-08, "loss": 0.0101, "step": 17654 }, { "epoch": 6.96, "learning_rate": 8.941726835286309e-08, "loss": 0.0099, "step": 17655 }, { "epoch": 6.96, "learning_rate": 8.939599409058305e-08, "loss": 0.0104, "step": 17656 }, { "epoch": 6.96, "learning_rate": 8.937472163098025e-08, "loss": 0.0131, "step": 17657 }, { "epoch": 6.96, "learning_rate": 8.93534509744013e-08, "loss": 0.0121, "step": 17658 }, { "epoch": 6.96, "learning_rate": 8.933218212119294e-08, "loss": 0.0154, "step": 17659 }, { "epoch": 6.96, "learning_rate": 8.931091507170178e-08, "loss": 0.0114, "step": 17660 }, { "epoch": 6.96, "learning_rate": 8.928964982627458e-08, "loss": 0.0113, "step": 17661 }, { "epoch": 6.96, "learning_rate": 8.926838638525776e-08, "loss": 0.0113, "step": 17662 }, { "epoch": 6.96, "learning_rate": 8.924712474899801e-08, "loss": 0.013, "step": 17663 }, { "epoch": 6.96, "learning_rate": 8.922586491784174e-08, "loss": 0.015, "step": 17664 }, { "epoch": 6.96, "learning_rate": 8.920460689213549e-08, "loss": 0.0108, "step": 17665 }, { "epoch": 6.96, "learning_rate": 8.918335067222578e-08, "loss": 0.0121, "step": 17666 }, { "epoch": 6.96, "learning_rate": 8.91620962584591e-08, "loss": 0.0122, "step": 17667 }, { "epoch": 6.96, "learning_rate": 8.914084365118173e-08, "loss": 0.0094, "step": 17668 }, { "epoch": 6.96, "learning_rate": 8.911959285074011e-08, "loss": 0.0139, "step": 17669 }, { "epoch": 6.96, "learning_rate": 8.90983438574806e-08, "loss": 0.0123, "step": 17670 }, { "epoch": 6.96, "learning_rate": 8.907709667174963e-08, "loss": 0.0145, "step": 17671 }, { "epoch": 6.96, "learning_rate": 8.905585129389333e-08, "loss": 0.0119, "step": 17672 }, { "epoch": 6.97, "learning_rate": 8.903460772425808e-08, "loss": 0.011, "step": 17673 }, { "epoch": 6.97, "learning_rate": 8.901336596318996e-08, "loss": 0.0108, "step": 17674 }, { "epoch": 6.97, "learning_rate": 8.899212601103542e-08, "loss": 0.0115, "step": 17675 }, { "epoch": 6.97, "learning_rate": 8.897088786814046e-08, "loss": 0.0106, "step": 17676 }, { "epoch": 6.97, "learning_rate": 8.894965153485135e-08, "loss": 0.0113, "step": 17677 }, { "epoch": 6.97, "learning_rate": 8.892841701151405e-08, "loss": 0.011, "step": 17678 }, { "epoch": 6.97, "learning_rate": 8.890718429847477e-08, "loss": 0.0123, "step": 17679 }, { "epoch": 6.97, "learning_rate": 8.88859533960796e-08, "loss": 0.0125, "step": 17680 }, { "epoch": 6.97, "learning_rate": 8.886472430467444e-08, "loss": 0.0117, "step": 17681 }, { "epoch": 6.97, "learning_rate": 8.884349702460538e-08, "loss": 0.0128, "step": 17682 }, { "epoch": 6.97, "learning_rate": 8.882227155621838e-08, "loss": 0.0136, "step": 17683 }, { "epoch": 6.97, "learning_rate": 8.880104789985945e-08, "loss": 0.0109, "step": 17684 }, { "epoch": 6.97, "learning_rate": 8.877982605587438e-08, "loss": 0.0132, "step": 17685 }, { "epoch": 6.97, "learning_rate": 8.875860602460916e-08, "loss": 0.0102, "step": 17686 }, { "epoch": 6.97, "learning_rate": 8.873738780640947e-08, "loss": 0.0111, "step": 17687 }, { "epoch": 6.97, "learning_rate": 8.871617140162137e-08, "loss": 0.0119, "step": 17688 }, { "epoch": 6.97, "learning_rate": 8.869495681059052e-08, "loss": 0.0122, "step": 17689 }, { "epoch": 6.97, "learning_rate": 8.867374403366275e-08, "loss": 0.0109, "step": 17690 }, { "epoch": 6.97, "learning_rate": 8.865253307118368e-08, "loss": 0.0104, "step": 17691 }, { "epoch": 6.97, "learning_rate": 8.86313239234991e-08, "loss": 0.0103, "step": 17692 }, { "epoch": 6.97, "learning_rate": 8.861011659095467e-08, "loss": 0.013, "step": 17693 }, { "epoch": 6.97, "learning_rate": 8.858891107389614e-08, "loss": 0.0128, "step": 17694 }, { "epoch": 6.97, "learning_rate": 8.856770737266897e-08, "loss": 0.0116, "step": 17695 }, { "epoch": 6.97, "learning_rate": 8.854650548761877e-08, "loss": 0.0118, "step": 17696 }, { "epoch": 6.97, "learning_rate": 8.852530541909118e-08, "loss": 0.0104, "step": 17697 }, { "epoch": 6.97, "learning_rate": 8.850410716743175e-08, "loss": 0.013, "step": 17698 }, { "epoch": 6.98, "learning_rate": 8.848291073298584e-08, "loss": 0.0109, "step": 17699 }, { "epoch": 6.98, "learning_rate": 8.846171611609908e-08, "loss": 0.0105, "step": 17700 }, { "epoch": 6.98, "learning_rate": 8.844052331711673e-08, "loss": 0.0131, "step": 17701 }, { "epoch": 6.98, "learning_rate": 8.841933233638432e-08, "loss": 0.012, "step": 17702 }, { "epoch": 6.98, "learning_rate": 8.839814317424721e-08, "loss": 0.0108, "step": 17703 }, { "epoch": 6.98, "learning_rate": 8.837695583105083e-08, "loss": 0.0119, "step": 17704 }, { "epoch": 6.98, "learning_rate": 8.835577030714034e-08, "loss": 0.0131, "step": 17705 }, { "epoch": 6.98, "learning_rate": 8.833458660286113e-08, "loss": 0.0104, "step": 17706 }, { "epoch": 6.98, "learning_rate": 8.83134047185585e-08, "loss": 0.0107, "step": 17707 }, { "epoch": 6.98, "learning_rate": 8.829222465457756e-08, "loss": 0.0105, "step": 17708 }, { "epoch": 6.98, "learning_rate": 8.827104641126359e-08, "loss": 0.0107, "step": 17709 }, { "epoch": 6.98, "learning_rate": 8.824986998896174e-08, "loss": 0.0134, "step": 17710 }, { "epoch": 6.98, "learning_rate": 8.822869538801725e-08, "loss": 0.0089, "step": 17711 }, { "epoch": 6.98, "learning_rate": 8.820752260877509e-08, "loss": 0.0103, "step": 17712 }, { "epoch": 6.98, "learning_rate": 8.818635165158046e-08, "loss": 0.0116, "step": 17713 }, { "epoch": 6.98, "learning_rate": 8.816518251677822e-08, "loss": 0.0153, "step": 17714 }, { "epoch": 6.98, "learning_rate": 8.81440152047137e-08, "loss": 0.0128, "step": 17715 }, { "epoch": 6.98, "learning_rate": 8.812284971573165e-08, "loss": 0.0137, "step": 17716 }, { "epoch": 6.98, "learning_rate": 8.810168605017718e-08, "loss": 0.0108, "step": 17717 }, { "epoch": 6.98, "learning_rate": 8.8080524208395e-08, "loss": 0.0106, "step": 17718 }, { "epoch": 6.98, "learning_rate": 8.805936419073034e-08, "loss": 0.0156, "step": 17719 }, { "epoch": 6.98, "learning_rate": 8.803820599752785e-08, "loss": 0.0129, "step": 17720 }, { "epoch": 6.98, "learning_rate": 8.801704962913252e-08, "loss": 0.0114, "step": 17721 }, { "epoch": 6.98, "learning_rate": 8.799589508588899e-08, "loss": 0.0116, "step": 17722 }, { "epoch": 6.98, "learning_rate": 8.797474236814215e-08, "loss": 0.0128, "step": 17723 }, { "epoch": 6.99, "learning_rate": 8.795359147623682e-08, "loss": 0.0124, "step": 17724 }, { "epoch": 6.99, "learning_rate": 8.79324424105176e-08, "loss": 0.0155, "step": 17725 }, { "epoch": 6.99, "learning_rate": 8.791129517132925e-08, "loss": 0.0131, "step": 17726 }, { "epoch": 6.99, "learning_rate": 8.78901497590164e-08, "loss": 0.0105, "step": 17727 }, { "epoch": 6.99, "learning_rate": 8.786900617392383e-08, "loss": 0.0106, "step": 17728 }, { "epoch": 6.99, "learning_rate": 8.784786441639593e-08, "loss": 0.0133, "step": 17729 }, { "epoch": 6.99, "learning_rate": 8.78267244867775e-08, "loss": 0.011, "step": 17730 }, { "epoch": 6.99, "learning_rate": 8.780558638541289e-08, "loss": 0.0112, "step": 17731 }, { "epoch": 6.99, "learning_rate": 8.77844501126467e-08, "loss": 0.0113, "step": 17732 }, { "epoch": 6.99, "learning_rate": 8.776331566882343e-08, "loss": 0.0111, "step": 17733 }, { "epoch": 6.99, "learning_rate": 8.77421830542876e-08, "loss": 0.0123, "step": 17734 }, { "epoch": 6.99, "learning_rate": 8.772105226938351e-08, "loss": 0.0115, "step": 17735 }, { "epoch": 6.99, "learning_rate": 8.769992331445562e-08, "loss": 0.0102, "step": 17736 }, { "epoch": 6.99, "learning_rate": 8.76787961898483e-08, "loss": 0.0109, "step": 17737 }, { "epoch": 6.99, "learning_rate": 8.765767089590596e-08, "loss": 0.0122, "step": 17738 }, { "epoch": 6.99, "learning_rate": 8.763654743297279e-08, "loss": 0.0102, "step": 17739 }, { "epoch": 6.99, "learning_rate": 8.761542580139321e-08, "loss": 0.0138, "step": 17740 }, { "epoch": 6.99, "learning_rate": 8.759430600151121e-08, "loss": 0.011, "step": 17741 }, { "epoch": 6.99, "learning_rate": 8.757318803367133e-08, "loss": 0.0104, "step": 17742 }, { "epoch": 6.99, "learning_rate": 8.755207189821755e-08, "loss": 0.0105, "step": 17743 }, { "epoch": 6.99, "learning_rate": 8.753095759549419e-08, "loss": 0.0142, "step": 17744 }, { "epoch": 6.99, "learning_rate": 8.750984512584518e-08, "loss": 0.0116, "step": 17745 }, { "epoch": 6.99, "learning_rate": 8.748873448961474e-08, "loss": 0.0139, "step": 17746 }, { "epoch": 6.99, "learning_rate": 8.746762568714701e-08, "loss": 0.0129, "step": 17747 }, { "epoch": 6.99, "learning_rate": 8.744651871878587e-08, "loss": 0.0151, "step": 17748 }, { "epoch": 7.0, "learning_rate": 8.742541358487544e-08, "loss": 0.0146, "step": 17749 }, { "epoch": 7.0, "learning_rate": 8.740431028575965e-08, "loss": 0.0132, "step": 17750 }, { "epoch": 7.0, "learning_rate": 8.738320882178256e-08, "loss": 0.0115, "step": 17751 }, { "epoch": 7.0, "learning_rate": 8.736210919328793e-08, "loss": 0.0128, "step": 17752 }, { "epoch": 7.0, "learning_rate": 8.73410114006198e-08, "loss": 0.0141, "step": 17753 }, { "epoch": 7.0, "learning_rate": 8.731991544412183e-08, "loss": 0.0105, "step": 17754 }, { "epoch": 7.0, "learning_rate": 8.72988213241381e-08, "loss": 0.0121, "step": 17755 }, { "epoch": 7.0, "learning_rate": 8.727772904101226e-08, "loss": 0.0108, "step": 17756 }, { "epoch": 7.0, "learning_rate": 8.725663859508817e-08, "loss": 0.0093, "step": 17757 }, { "epoch": 7.0, "learning_rate": 8.723554998670939e-08, "loss": 0.0091, "step": 17758 }, { "epoch": 7.0, "learning_rate": 8.721446321621988e-08, "loss": 0.0116, "step": 17759 }, { "epoch": 7.0, "learning_rate": 8.719337828396316e-08, "loss": 0.0091, "step": 17760 }, { "epoch": 7.0, "learning_rate": 8.717229519028299e-08, "loss": 0.0103, "step": 17761 }, { "epoch": 7.0, "learning_rate": 8.715121393552287e-08, "loss": 0.0126, "step": 17762 }, { "epoch": 7.0, "learning_rate": 8.713013452002642e-08, "loss": 0.01, "step": 17763 }, { "epoch": 7.0, "learning_rate": 8.710905694413726e-08, "loss": 0.011, "step": 17764 }, { "epoch": 7.0, "learning_rate": 8.708798120819898e-08, "loss": 0.0124, "step": 17765 }, { "epoch": 7.0, "learning_rate": 8.706690731255491e-08, "loss": 0.0133, "step": 17766 }, { "epoch": 7.0, "learning_rate": 8.704583525754861e-08, "loss": 0.0125, "step": 17767 }, { "epoch": 7.0, "learning_rate": 8.70247650435236e-08, "loss": 0.0101, "step": 17768 }, { "epoch": 7.0, "learning_rate": 8.700369667082317e-08, "loss": 0.0095, "step": 17769 }, { "epoch": 7.0, "learning_rate": 8.69826301397908e-08, "loss": 0.0106, "step": 17770 }, { "epoch": 7.0, "learning_rate": 8.696156545076971e-08, "loss": 0.0113, "step": 17771 }, { "epoch": 7.0, "learning_rate": 8.69405026041033e-08, "loss": 0.013, "step": 17772 }, { "epoch": 7.0, "learning_rate": 8.691944160013488e-08, "loss": 0.0112, "step": 17773 }, { "epoch": 7.0, "learning_rate": 8.68983824392078e-08, "loss": 0.0111, "step": 17774 }, { "epoch": 7.01, "learning_rate": 8.687732512166508e-08, "loss": 0.0105, "step": 17775 }, { "epoch": 7.01, "learning_rate": 8.685626964785004e-08, "loss": 0.0091, "step": 17776 }, { "epoch": 7.01, "learning_rate": 8.683521601810586e-08, "loss": 0.0084, "step": 17777 }, { "epoch": 7.01, "learning_rate": 8.681416423277572e-08, "loss": 0.0118, "step": 17778 }, { "epoch": 7.01, "learning_rate": 8.679311429220264e-08, "loss": 0.0106, "step": 17779 }, { "epoch": 7.01, "learning_rate": 8.677206619672979e-08, "loss": 0.0103, "step": 17780 }, { "epoch": 7.01, "learning_rate": 8.675101994670002e-08, "loss": 0.0102, "step": 17781 }, { "epoch": 7.01, "learning_rate": 8.672997554245667e-08, "loss": 0.0105, "step": 17782 }, { "epoch": 7.01, "learning_rate": 8.67089329843425e-08, "loss": 0.0119, "step": 17783 }, { "epoch": 7.01, "learning_rate": 8.668789227270061e-08, "loss": 0.0121, "step": 17784 }, { "epoch": 7.01, "learning_rate": 8.666685340787379e-08, "loss": 0.01, "step": 17785 }, { "epoch": 7.01, "learning_rate": 8.664581639020499e-08, "loss": 0.0105, "step": 17786 }, { "epoch": 7.01, "learning_rate": 8.66247812200371e-08, "loss": 0.0078, "step": 17787 }, { "epoch": 7.01, "learning_rate": 8.660374789771308e-08, "loss": 0.0107, "step": 17788 }, { "epoch": 7.01, "learning_rate": 8.658271642357553e-08, "loss": 0.0131, "step": 17789 }, { "epoch": 7.01, "learning_rate": 8.656168679796734e-08, "loss": 0.0108, "step": 17790 }, { "epoch": 7.01, "learning_rate": 8.65406590212313e-08, "loss": 0.0123, "step": 17791 }, { "epoch": 7.01, "learning_rate": 8.651963309371002e-08, "loss": 0.0097, "step": 17792 }, { "epoch": 7.01, "learning_rate": 8.649860901574633e-08, "loss": 0.0124, "step": 17793 }, { "epoch": 7.01, "learning_rate": 8.647758678768264e-08, "loss": 0.0094, "step": 17794 }, { "epoch": 7.01, "learning_rate": 8.645656640986192e-08, "loss": 0.0094, "step": 17795 }, { "epoch": 7.01, "learning_rate": 8.643554788262652e-08, "loss": 0.0098, "step": 17796 }, { "epoch": 7.01, "learning_rate": 8.641453120631919e-08, "loss": 0.0112, "step": 17797 }, { "epoch": 7.01, "learning_rate": 8.639351638128218e-08, "loss": 0.0098, "step": 17798 }, { "epoch": 7.01, "learning_rate": 8.637250340785838e-08, "loss": 0.0109, "step": 17799 }, { "epoch": 7.02, "learning_rate": 8.635149228638998e-08, "loss": 0.0122, "step": 17800 }, { "epoch": 7.02, "learning_rate": 8.633048301721961e-08, "loss": 0.0105, "step": 17801 }, { "epoch": 7.02, "learning_rate": 8.630947560068953e-08, "loss": 0.0107, "step": 17802 }, { "epoch": 7.02, "learning_rate": 8.628847003714222e-08, "loss": 0.0119, "step": 17803 }, { "epoch": 7.02, "learning_rate": 8.626746632692001e-08, "loss": 0.0103, "step": 17804 }, { "epoch": 7.02, "learning_rate": 8.624646447036534e-08, "loss": 0.0113, "step": 17805 }, { "epoch": 7.02, "learning_rate": 8.622546446782033e-08, "loss": 0.0099, "step": 17806 }, { "epoch": 7.02, "learning_rate": 8.620446631962732e-08, "loss": 0.0106, "step": 17807 }, { "epoch": 7.02, "learning_rate": 8.618347002612858e-08, "loss": 0.0152, "step": 17808 }, { "epoch": 7.02, "learning_rate": 8.616247558766637e-08, "loss": 0.0114, "step": 17809 }, { "epoch": 7.02, "learning_rate": 8.614148300458271e-08, "loss": 0.0142, "step": 17810 }, { "epoch": 7.02, "learning_rate": 8.612049227721989e-08, "loss": 0.0089, "step": 17811 }, { "epoch": 7.02, "learning_rate": 8.609950340591991e-08, "loss": 0.0107, "step": 17812 }, { "epoch": 7.02, "learning_rate": 8.607851639102494e-08, "loss": 0.0125, "step": 17813 }, { "epoch": 7.02, "learning_rate": 8.605753123287707e-08, "loss": 0.0137, "step": 17814 }, { "epoch": 7.02, "learning_rate": 8.603654793181818e-08, "loss": 0.0099, "step": 17815 }, { "epoch": 7.02, "learning_rate": 8.601556648819037e-08, "loss": 0.0134, "step": 17816 }, { "epoch": 7.02, "learning_rate": 8.599458690233559e-08, "loss": 0.0083, "step": 17817 }, { "epoch": 7.02, "learning_rate": 8.597360917459584e-08, "loss": 0.0113, "step": 17818 }, { "epoch": 7.02, "learning_rate": 8.595263330531289e-08, "loss": 0.0137, "step": 17819 }, { "epoch": 7.02, "learning_rate": 8.593165929482876e-08, "loss": 0.0126, "step": 17820 }, { "epoch": 7.02, "learning_rate": 8.591068714348506e-08, "loss": 0.0149, "step": 17821 }, { "epoch": 7.02, "learning_rate": 8.588971685162392e-08, "loss": 0.0073, "step": 17822 }, { "epoch": 7.02, "learning_rate": 8.586874841958691e-08, "loss": 0.0101, "step": 17823 }, { "epoch": 7.02, "learning_rate": 8.584778184771588e-08, "loss": 0.0109, "step": 17824 }, { "epoch": 7.02, "learning_rate": 8.58268171363524e-08, "loss": 0.0107, "step": 17825 }, { "epoch": 7.03, "learning_rate": 8.580585428583842e-08, "loss": 0.0127, "step": 17826 }, { "epoch": 7.03, "learning_rate": 8.578489329651538e-08, "loss": 0.0115, "step": 17827 }, { "epoch": 7.03, "learning_rate": 8.576393416872504e-08, "loss": 0.0106, "step": 17828 }, { "epoch": 7.03, "learning_rate": 8.574297690280892e-08, "loss": 0.0122, "step": 17829 }, { "epoch": 7.03, "learning_rate": 8.572202149910861e-08, "loss": 0.0113, "step": 17830 }, { "epoch": 7.03, "learning_rate": 8.570106795796568e-08, "loss": 0.0123, "step": 17831 }, { "epoch": 7.03, "learning_rate": 8.568011627972169e-08, "loss": 0.0108, "step": 17832 }, { "epoch": 7.03, "learning_rate": 8.565916646471798e-08, "loss": 0.0123, "step": 17833 }, { "epoch": 7.03, "learning_rate": 8.563821851329609e-08, "loss": 0.0108, "step": 17834 }, { "epoch": 7.03, "learning_rate": 8.561727242579748e-08, "loss": 0.0102, "step": 17835 }, { "epoch": 7.03, "learning_rate": 8.559632820256344e-08, "loss": 0.0109, "step": 17836 }, { "epoch": 7.03, "learning_rate": 8.557538584393544e-08, "loss": 0.0098, "step": 17837 }, { "epoch": 7.03, "learning_rate": 8.55544453502546e-08, "loss": 0.0117, "step": 17838 }, { "epoch": 7.03, "learning_rate": 8.55335067218625e-08, "loss": 0.0105, "step": 17839 }, { "epoch": 7.03, "learning_rate": 8.551256995910021e-08, "loss": 0.0146, "step": 17840 }, { "epoch": 7.03, "learning_rate": 8.54916350623091e-08, "loss": 0.0119, "step": 17841 }, { "epoch": 7.03, "learning_rate": 8.547070203183024e-08, "loss": 0.0114, "step": 17842 }, { "epoch": 7.03, "learning_rate": 8.544977086800486e-08, "loss": 0.0094, "step": 17843 }, { "epoch": 7.03, "learning_rate": 8.542884157117411e-08, "loss": 0.011, "step": 17844 }, { "epoch": 7.03, "learning_rate": 8.54079141416792e-08, "loss": 0.0121, "step": 17845 }, { "epoch": 7.03, "learning_rate": 8.538698857986107e-08, "loss": 0.0082, "step": 17846 }, { "epoch": 7.03, "learning_rate": 8.536606488606082e-08, "loss": 0.0118, "step": 17847 }, { "epoch": 7.03, "learning_rate": 8.534514306061948e-08, "loss": 0.0109, "step": 17848 }, { "epoch": 7.03, "learning_rate": 8.532422310387814e-08, "loss": 0.0139, "step": 17849 }, { "epoch": 7.03, "learning_rate": 8.53033050161776e-08, "loss": 0.0095, "step": 17850 }, { "epoch": 7.04, "learning_rate": 8.528238879785894e-08, "loss": 0.0103, "step": 17851 }, { "epoch": 7.04, "learning_rate": 8.526147444926292e-08, "loss": 0.0118, "step": 17852 }, { "epoch": 7.04, "learning_rate": 8.524056197073049e-08, "loss": 0.0103, "step": 17853 }, { "epoch": 7.04, "learning_rate": 8.521965136260248e-08, "loss": 0.0128, "step": 17854 }, { "epoch": 7.04, "learning_rate": 8.519874262521978e-08, "loss": 0.0148, "step": 17855 }, { "epoch": 7.04, "learning_rate": 8.517783575892304e-08, "loss": 0.0105, "step": 17856 }, { "epoch": 7.04, "learning_rate": 8.515693076405303e-08, "loss": 0.0107, "step": 17857 }, { "epoch": 7.04, "learning_rate": 8.51360276409506e-08, "loss": 0.0095, "step": 17858 }, { "epoch": 7.04, "learning_rate": 8.511512638995629e-08, "loss": 0.0119, "step": 17859 }, { "epoch": 7.04, "learning_rate": 8.509422701141087e-08, "loss": 0.011, "step": 17860 }, { "epoch": 7.04, "learning_rate": 8.507332950565475e-08, "loss": 0.0106, "step": 17861 }, { "epoch": 7.04, "learning_rate": 8.505243387302887e-08, "loss": 0.0112, "step": 17862 }, { "epoch": 7.04, "learning_rate": 8.503154011387352e-08, "loss": 0.009, "step": 17863 }, { "epoch": 7.04, "learning_rate": 8.501064822852944e-08, "loss": 0.0097, "step": 17864 }, { "epoch": 7.04, "learning_rate": 8.498975821733685e-08, "loss": 0.0105, "step": 17865 }, { "epoch": 7.04, "learning_rate": 8.496887008063657e-08, "loss": 0.011, "step": 17866 }, { "epoch": 7.04, "learning_rate": 8.49479838187688e-08, "loss": 0.01, "step": 17867 }, { "epoch": 7.04, "learning_rate": 8.492709943207412e-08, "loss": 0.0104, "step": 17868 }, { "epoch": 7.04, "learning_rate": 8.490621692089265e-08, "loss": 0.0092, "step": 17869 }, { "epoch": 7.04, "learning_rate": 8.488533628556511e-08, "loss": 0.0103, "step": 17870 }, { "epoch": 7.04, "learning_rate": 8.486445752643155e-08, "loss": 0.0125, "step": 17871 }, { "epoch": 7.04, "learning_rate": 8.484358064383239e-08, "loss": 0.0111, "step": 17872 }, { "epoch": 7.04, "learning_rate": 8.48227056381078e-08, "loss": 0.0112, "step": 17873 }, { "epoch": 7.04, "learning_rate": 8.480183250959807e-08, "loss": 0.0127, "step": 17874 }, { "epoch": 7.04, "learning_rate": 8.478096125864344e-08, "loss": 0.0098, "step": 17875 }, { "epoch": 7.05, "learning_rate": 8.476009188558396e-08, "loss": 0.0101, "step": 17876 }, { "epoch": 7.05, "learning_rate": 8.473922439075985e-08, "loss": 0.0105, "step": 17877 }, { "epoch": 7.05, "learning_rate": 8.471835877451118e-08, "loss": 0.0083, "step": 17878 }, { "epoch": 7.05, "learning_rate": 8.469749503717816e-08, "loss": 0.0106, "step": 17879 }, { "epoch": 7.05, "learning_rate": 8.467663317910063e-08, "loss": 0.0097, "step": 17880 }, { "epoch": 7.05, "learning_rate": 8.465577320061879e-08, "loss": 0.0118, "step": 17881 }, { "epoch": 7.05, "learning_rate": 8.463491510207249e-08, "loss": 0.0133, "step": 17882 }, { "epoch": 7.05, "learning_rate": 8.461405888380171e-08, "loss": 0.0102, "step": 17883 }, { "epoch": 7.05, "learning_rate": 8.459320454614643e-08, "loss": 0.0094, "step": 17884 }, { "epoch": 7.05, "learning_rate": 8.457235208944658e-08, "loss": 0.012, "step": 17885 }, { "epoch": 7.05, "learning_rate": 8.455150151404187e-08, "loss": 0.0121, "step": 17886 }, { "epoch": 7.05, "learning_rate": 8.453065282027222e-08, "loss": 0.0104, "step": 17887 }, { "epoch": 7.05, "learning_rate": 8.450980600847748e-08, "loss": 0.0131, "step": 17888 }, { "epoch": 7.05, "learning_rate": 8.448896107899742e-08, "loss": 0.0123, "step": 17889 }, { "epoch": 7.05, "learning_rate": 8.446811803217166e-08, "loss": 0.0096, "step": 17890 }, { "epoch": 7.05, "learning_rate": 8.444727686834008e-08, "loss": 0.0115, "step": 17891 }, { "epoch": 7.05, "learning_rate": 8.442643758784211e-08, "loss": 0.0111, "step": 17892 }, { "epoch": 7.05, "learning_rate": 8.44056001910177e-08, "loss": 0.0101, "step": 17893 }, { "epoch": 7.05, "learning_rate": 8.438476467820626e-08, "loss": 0.0098, "step": 17894 }, { "epoch": 7.05, "learning_rate": 8.43639310497475e-08, "loss": 0.0114, "step": 17895 }, { "epoch": 7.05, "learning_rate": 8.434309930598086e-08, "loss": 0.0112, "step": 17896 }, { "epoch": 7.05, "learning_rate": 8.432226944724592e-08, "loss": 0.0106, "step": 17897 }, { "epoch": 7.05, "learning_rate": 8.430144147388225e-08, "loss": 0.011, "step": 17898 }, { "epoch": 7.05, "learning_rate": 8.428061538622918e-08, "loss": 0.0127, "step": 17899 }, { "epoch": 7.05, "learning_rate": 8.425979118462618e-08, "loss": 0.0118, "step": 17900 }, { "epoch": 7.05, "learning_rate": 8.423896886941269e-08, "loss": 0.0105, "step": 17901 }, { "epoch": 7.06, "learning_rate": 8.421814844092814e-08, "loss": 0.01, "step": 17902 }, { "epoch": 7.06, "learning_rate": 8.419732989951173e-08, "loss": 0.0129, "step": 17903 }, { "epoch": 7.06, "learning_rate": 8.417651324550292e-08, "loss": 0.0097, "step": 17904 }, { "epoch": 7.06, "learning_rate": 8.415569847924078e-08, "loss": 0.0144, "step": 17905 }, { "epoch": 7.06, "learning_rate": 8.413488560106481e-08, "loss": 0.0113, "step": 17906 }, { "epoch": 7.06, "learning_rate": 8.411407461131405e-08, "loss": 0.0107, "step": 17907 }, { "epoch": 7.06, "learning_rate": 8.409326551032783e-08, "loss": 0.0103, "step": 17908 }, { "epoch": 7.06, "learning_rate": 8.407245829844506e-08, "loss": 0.012, "step": 17909 }, { "epoch": 7.06, "learning_rate": 8.40516529760052e-08, "loss": 0.0082, "step": 17910 }, { "epoch": 7.06, "learning_rate": 8.40308495433471e-08, "loss": 0.011, "step": 17911 }, { "epoch": 7.06, "learning_rate": 8.401004800080995e-08, "loss": 0.0114, "step": 17912 }, { "epoch": 7.06, "learning_rate": 8.398924834873266e-08, "loss": 0.0117, "step": 17913 }, { "epoch": 7.06, "learning_rate": 8.396845058745432e-08, "loss": 0.0097, "step": 17914 }, { "epoch": 7.06, "learning_rate": 8.394765471731388e-08, "loss": 0.0099, "step": 17915 }, { "epoch": 7.06, "learning_rate": 8.392686073865039e-08, "loss": 0.0116, "step": 17916 }, { "epoch": 7.06, "learning_rate": 8.390606865180257e-08, "loss": 0.01, "step": 17917 }, { "epoch": 7.06, "learning_rate": 8.388527845710938e-08, "loss": 0.0133, "step": 17918 }, { "epoch": 7.06, "learning_rate": 8.386449015490975e-08, "loss": 0.0114, "step": 17919 }, { "epoch": 7.06, "learning_rate": 8.384370374554235e-08, "loss": 0.0094, "step": 17920 }, { "epoch": 7.06, "learning_rate": 8.382291922934613e-08, "loss": 0.0099, "step": 17921 }, { "epoch": 7.06, "learning_rate": 8.380213660665966e-08, "loss": 0.0092, "step": 17922 }, { "epoch": 7.06, "learning_rate": 8.378135587782178e-08, "loss": 0.0093, "step": 17923 }, { "epoch": 7.06, "learning_rate": 8.376057704317117e-08, "loss": 0.0094, "step": 17924 }, { "epoch": 7.06, "learning_rate": 8.373980010304656e-08, "loss": 0.0107, "step": 17925 }, { "epoch": 7.06, "learning_rate": 8.371902505778644e-08, "loss": 0.0109, "step": 17926 }, { "epoch": 7.07, "learning_rate": 8.369825190772946e-08, "loss": 0.0107, "step": 17927 }, { "epoch": 7.07, "learning_rate": 8.367748065321426e-08, "loss": 0.011, "step": 17928 }, { "epoch": 7.07, "learning_rate": 8.365671129457936e-08, "loss": 0.0105, "step": 17929 }, { "epoch": 7.07, "learning_rate": 8.363594383216322e-08, "loss": 0.0117, "step": 17930 }, { "epoch": 7.07, "learning_rate": 8.361517826630437e-08, "loss": 0.0126, "step": 17931 }, { "epoch": 7.07, "learning_rate": 8.359441459734111e-08, "loss": 0.0105, "step": 17932 }, { "epoch": 7.07, "learning_rate": 8.357365282561213e-08, "loss": 0.0111, "step": 17933 }, { "epoch": 7.07, "learning_rate": 8.355289295145558e-08, "loss": 0.0101, "step": 17934 }, { "epoch": 7.07, "learning_rate": 8.353213497520996e-08, "loss": 0.012, "step": 17935 }, { "epoch": 7.07, "learning_rate": 8.351137889721338e-08, "loss": 0.011, "step": 17936 }, { "epoch": 7.07, "learning_rate": 8.349062471780444e-08, "loss": 0.0091, "step": 17937 }, { "epoch": 7.07, "learning_rate": 8.346987243732117e-08, "loss": 0.0132, "step": 17938 }, { "epoch": 7.07, "learning_rate": 8.344912205610194e-08, "loss": 0.0114, "step": 17939 }, { "epoch": 7.07, "learning_rate": 8.342837357448482e-08, "loss": 0.0106, "step": 17940 }, { "epoch": 7.07, "learning_rate": 8.340762699280803e-08, "loss": 0.011, "step": 17941 }, { "epoch": 7.07, "learning_rate": 8.33868823114098e-08, "loss": 0.0087, "step": 17942 }, { "epoch": 7.07, "learning_rate": 8.33661395306281e-08, "loss": 0.0089, "step": 17943 }, { "epoch": 7.07, "learning_rate": 8.33453986508011e-08, "loss": 0.0097, "step": 17944 }, { "epoch": 7.07, "learning_rate": 8.33246596722667e-08, "loss": 0.0086, "step": 17945 }, { "epoch": 7.07, "learning_rate": 8.330392259536314e-08, "loss": 0.012, "step": 17946 }, { "epoch": 7.07, "learning_rate": 8.328318742042821e-08, "loss": 0.0108, "step": 17947 }, { "epoch": 7.07, "learning_rate": 8.326245414780002e-08, "loss": 0.0121, "step": 17948 }, { "epoch": 7.07, "learning_rate": 8.324172277781627e-08, "loss": 0.0127, "step": 17949 }, { "epoch": 7.07, "learning_rate": 8.322099331081512e-08, "loss": 0.0104, "step": 17950 }, { "epoch": 7.07, "learning_rate": 8.320026574713421e-08, "loss": 0.0097, "step": 17951 }, { "epoch": 7.08, "learning_rate": 8.317954008711155e-08, "loss": 0.0117, "step": 17952 }, { "epoch": 7.08, "learning_rate": 8.315881633108475e-08, "loss": 0.0109, "step": 17953 }, { "epoch": 7.08, "learning_rate": 8.313809447939169e-08, "loss": 0.0098, "step": 17954 }, { "epoch": 7.08, "learning_rate": 8.311737453237005e-08, "loss": 0.01, "step": 17955 }, { "epoch": 7.08, "learning_rate": 8.309665649035765e-08, "loss": 0.0126, "step": 17956 }, { "epoch": 7.08, "learning_rate": 8.3075940353692e-08, "loss": 0.0101, "step": 17957 }, { "epoch": 7.08, "learning_rate": 8.305522612271083e-08, "loss": 0.0134, "step": 17958 }, { "epoch": 7.08, "learning_rate": 8.303451379775173e-08, "loss": 0.0106, "step": 17959 }, { "epoch": 7.08, "learning_rate": 8.301380337915236e-08, "loss": 0.0125, "step": 17960 }, { "epoch": 7.08, "learning_rate": 8.299309486725012e-08, "loss": 0.0103, "step": 17961 }, { "epoch": 7.08, "learning_rate": 8.297238826238268e-08, "loss": 0.0096, "step": 17962 }, { "epoch": 7.08, "learning_rate": 8.295168356488737e-08, "loss": 0.0117, "step": 17963 }, { "epoch": 7.08, "learning_rate": 8.293098077510174e-08, "loss": 0.0117, "step": 17964 }, { "epoch": 7.08, "learning_rate": 8.291027989336328e-08, "loss": 0.0095, "step": 17965 }, { "epoch": 7.08, "learning_rate": 8.28895809200092e-08, "loss": 0.0121, "step": 17966 }, { "epoch": 7.08, "learning_rate": 8.286888385537697e-08, "loss": 0.0097, "step": 17967 }, { "epoch": 7.08, "learning_rate": 8.284818869980394e-08, "loss": 0.0101, "step": 17968 }, { "epoch": 7.08, "learning_rate": 8.282749545362744e-08, "loss": 0.0092, "step": 17969 }, { "epoch": 7.08, "learning_rate": 8.28068041171846e-08, "loss": 0.0123, "step": 17970 }, { "epoch": 7.08, "learning_rate": 8.278611469081282e-08, "loss": 0.0107, "step": 17971 }, { "epoch": 7.08, "learning_rate": 8.27654271748491e-08, "loss": 0.011, "step": 17972 }, { "epoch": 7.08, "learning_rate": 8.274474156963085e-08, "loss": 0.0095, "step": 17973 }, { "epoch": 7.08, "learning_rate": 8.272405787549506e-08, "loss": 0.0156, "step": 17974 }, { "epoch": 7.08, "learning_rate": 8.270337609277893e-08, "loss": 0.0095, "step": 17975 }, { "epoch": 7.08, "learning_rate": 8.268269622181938e-08, "loss": 0.0107, "step": 17976 }, { "epoch": 7.08, "learning_rate": 8.266201826295373e-08, "loss": 0.0116, "step": 17977 }, { "epoch": 7.09, "learning_rate": 8.264134221651878e-08, "loss": 0.0096, "step": 17978 }, { "epoch": 7.09, "learning_rate": 8.262066808285162e-08, "loss": 0.009, "step": 17979 }, { "epoch": 7.09, "learning_rate": 8.259999586228914e-08, "loss": 0.0113, "step": 17980 }, { "epoch": 7.09, "learning_rate": 8.257932555516829e-08, "loss": 0.0114, "step": 17981 }, { "epoch": 7.09, "learning_rate": 8.255865716182597e-08, "loss": 0.0112, "step": 17982 }, { "epoch": 7.09, "learning_rate": 8.253799068259913e-08, "loss": 0.0092, "step": 17983 }, { "epoch": 7.09, "learning_rate": 8.251732611782443e-08, "loss": 0.0136, "step": 17984 }, { "epoch": 7.09, "learning_rate": 8.249666346783877e-08, "loss": 0.0096, "step": 17985 }, { "epoch": 7.09, "learning_rate": 8.247600273297899e-08, "loss": 0.0145, "step": 17986 }, { "epoch": 7.09, "learning_rate": 8.245534391358166e-08, "loss": 0.0096, "step": 17987 }, { "epoch": 7.09, "learning_rate": 8.243468700998366e-08, "loss": 0.0108, "step": 17988 }, { "epoch": 7.09, "learning_rate": 8.241403202252146e-08, "loss": 0.0119, "step": 17989 }, { "epoch": 7.09, "learning_rate": 8.239337895153193e-08, "loss": 0.0079, "step": 17990 }, { "epoch": 7.09, "learning_rate": 8.237272779735153e-08, "loss": 0.0107, "step": 17991 }, { "epoch": 7.09, "learning_rate": 8.235207856031698e-08, "loss": 0.0103, "step": 17992 }, { "epoch": 7.09, "learning_rate": 8.233143124076467e-08, "loss": 0.0092, "step": 17993 }, { "epoch": 7.09, "learning_rate": 8.231078583903118e-08, "loss": 0.0117, "step": 17994 }, { "epoch": 7.09, "learning_rate": 8.229014235545301e-08, "loss": 0.0117, "step": 17995 }, { "epoch": 7.09, "learning_rate": 8.226950079036672e-08, "loss": 0.008, "step": 17996 }, { "epoch": 7.09, "learning_rate": 8.224886114410854e-08, "loss": 0.0127, "step": 17997 }, { "epoch": 7.09, "learning_rate": 8.222822341701498e-08, "loss": 0.0106, "step": 17998 }, { "epoch": 7.09, "learning_rate": 8.220758760942237e-08, "loss": 0.0108, "step": 17999 }, { "epoch": 7.09, "learning_rate": 8.218695372166713e-08, "loss": 0.0123, "step": 18000 }, { "epoch": 7.09, "learning_rate": 8.216632175408545e-08, "loss": 0.0123, "step": 18001 }, { "epoch": 7.09, "learning_rate": 8.214569170701367e-08, "loss": 0.0113, "step": 18002 }, { "epoch": 7.1, "learning_rate": 8.212506358078795e-08, "loss": 0.0102, "step": 18003 }, { "epoch": 7.1, "learning_rate": 8.210443737574453e-08, "loss": 0.0105, "step": 18004 }, { "epoch": 7.1, "learning_rate": 8.208381309221959e-08, "loss": 0.0099, "step": 18005 }, { "epoch": 7.1, "learning_rate": 8.206319073054937e-08, "loss": 0.0118, "step": 18006 }, { "epoch": 7.1, "learning_rate": 8.204257029106984e-08, "loss": 0.009, "step": 18007 }, { "epoch": 7.1, "learning_rate": 8.202195177411712e-08, "loss": 0.0117, "step": 18008 }, { "epoch": 7.1, "learning_rate": 8.200133518002737e-08, "loss": 0.0097, "step": 18009 }, { "epoch": 7.1, "learning_rate": 8.198072050913643e-08, "loss": 0.0126, "step": 18010 }, { "epoch": 7.1, "learning_rate": 8.196010776178046e-08, "loss": 0.0119, "step": 18011 }, { "epoch": 7.1, "learning_rate": 8.193949693829518e-08, "loss": 0.0118, "step": 18012 }, { "epoch": 7.1, "learning_rate": 8.191888803901681e-08, "loss": 0.0105, "step": 18013 }, { "epoch": 7.1, "learning_rate": 8.189828106428105e-08, "loss": 0.0118, "step": 18014 }, { "epoch": 7.1, "learning_rate": 8.187767601442388e-08, "loss": 0.014, "step": 18015 }, { "epoch": 7.1, "learning_rate": 8.18570728897809e-08, "loss": 0.0095, "step": 18016 }, { "epoch": 7.1, "learning_rate": 8.183647169068823e-08, "loss": 0.0129, "step": 18017 }, { "epoch": 7.1, "learning_rate": 8.181587241748143e-08, "loss": 0.0068, "step": 18018 }, { "epoch": 7.1, "learning_rate": 8.179527507049635e-08, "loss": 0.0128, "step": 18019 }, { "epoch": 7.1, "learning_rate": 8.177467965006857e-08, "loss": 0.0111, "step": 18020 }, { "epoch": 7.1, "learning_rate": 8.17540861565338e-08, "loss": 0.009, "step": 18021 }, { "epoch": 7.1, "learning_rate": 8.173349459022776e-08, "loss": 0.012, "step": 18022 }, { "epoch": 7.1, "learning_rate": 8.171290495148607e-08, "loss": 0.0109, "step": 18023 }, { "epoch": 7.1, "learning_rate": 8.169231724064416e-08, "loss": 0.0093, "step": 18024 }, { "epoch": 7.1, "learning_rate": 8.167173145803767e-08, "loss": 0.0126, "step": 18025 }, { "epoch": 7.1, "learning_rate": 8.165114760400223e-08, "loss": 0.0115, "step": 18026 }, { "epoch": 7.1, "learning_rate": 8.163056567887309e-08, "loss": 0.0139, "step": 18027 }, { "epoch": 7.1, "learning_rate": 8.160998568298587e-08, "loss": 0.0124, "step": 18028 }, { "epoch": 7.11, "learning_rate": 8.158940761667594e-08, "loss": 0.0134, "step": 18029 }, { "epoch": 7.11, "learning_rate": 8.156883148027878e-08, "loss": 0.01, "step": 18030 }, { "epoch": 7.11, "learning_rate": 8.154825727412959e-08, "loss": 0.0123, "step": 18031 }, { "epoch": 7.11, "learning_rate": 8.152768499856386e-08, "loss": 0.0083, "step": 18032 }, { "epoch": 7.11, "learning_rate": 8.150711465391673e-08, "loss": 0.0133, "step": 18033 }, { "epoch": 7.11, "learning_rate": 8.148654624052353e-08, "loss": 0.0127, "step": 18034 }, { "epoch": 7.11, "learning_rate": 8.146597975871954e-08, "loss": 0.0081, "step": 18035 }, { "epoch": 7.11, "learning_rate": 8.144541520883998e-08, "loss": 0.0123, "step": 18036 }, { "epoch": 7.11, "learning_rate": 8.14248525912199e-08, "loss": 0.0135, "step": 18037 }, { "epoch": 7.11, "learning_rate": 8.140429190619451e-08, "loss": 0.0096, "step": 18038 }, { "epoch": 7.11, "learning_rate": 8.13837331540989e-08, "loss": 0.0094, "step": 18039 }, { "epoch": 7.11, "learning_rate": 8.136317633526826e-08, "loss": 0.0087, "step": 18040 }, { "epoch": 7.11, "learning_rate": 8.134262145003745e-08, "loss": 0.0101, "step": 18041 }, { "epoch": 7.11, "learning_rate": 8.132206849874166e-08, "loss": 0.0098, "step": 18042 }, { "epoch": 7.11, "learning_rate": 8.130151748171563e-08, "loss": 0.011, "step": 18043 }, { "epoch": 7.11, "learning_rate": 8.128096839929462e-08, "loss": 0.0108, "step": 18044 }, { "epoch": 7.11, "learning_rate": 8.126042125181332e-08, "loss": 0.012, "step": 18045 }, { "epoch": 7.11, "learning_rate": 8.123987603960674e-08, "loss": 0.0102, "step": 18046 }, { "epoch": 7.11, "learning_rate": 8.121933276300963e-08, "loss": 0.0084, "step": 18047 }, { "epoch": 7.11, "learning_rate": 8.119879142235687e-08, "loss": 0.0076, "step": 18048 }, { "epoch": 7.11, "learning_rate": 8.117825201798332e-08, "loss": 0.0116, "step": 18049 }, { "epoch": 7.11, "learning_rate": 8.11577145502236e-08, "loss": 0.0083, "step": 18050 }, { "epoch": 7.11, "learning_rate": 8.113717901941252e-08, "loss": 0.0118, "step": 18051 }, { "epoch": 7.11, "learning_rate": 8.111664542588477e-08, "loss": 0.0099, "step": 18052 }, { "epoch": 7.11, "learning_rate": 8.109611376997509e-08, "loss": 0.0104, "step": 18053 }, { "epoch": 7.12, "learning_rate": 8.107558405201795e-08, "loss": 0.011, "step": 18054 }, { "epoch": 7.12, "learning_rate": 8.105505627234814e-08, "loss": 0.0133, "step": 18055 }, { "epoch": 7.12, "learning_rate": 8.103453043129999e-08, "loss": 0.0113, "step": 18056 }, { "epoch": 7.12, "learning_rate": 8.101400652920834e-08, "loss": 0.0092, "step": 18057 }, { "epoch": 7.12, "learning_rate": 8.099348456640746e-08, "loss": 0.0107, "step": 18058 }, { "epoch": 7.12, "learning_rate": 8.097296454323199e-08, "loss": 0.0119, "step": 18059 }, { "epoch": 7.12, "learning_rate": 8.095244646001622e-08, "loss": 0.0119, "step": 18060 }, { "epoch": 7.12, "learning_rate": 8.093193031709467e-08, "loss": 0.0112, "step": 18061 }, { "epoch": 7.12, "learning_rate": 8.091141611480166e-08, "loss": 0.0117, "step": 18062 }, { "epoch": 7.12, "learning_rate": 8.089090385347165e-08, "loss": 0.0109, "step": 18063 }, { "epoch": 7.12, "learning_rate": 8.08703935334388e-08, "loss": 0.01, "step": 18064 }, { "epoch": 7.12, "learning_rate": 8.084988515503753e-08, "loss": 0.011, "step": 18065 }, { "epoch": 7.12, "learning_rate": 8.082937871860198e-08, "loss": 0.0119, "step": 18066 }, { "epoch": 7.12, "learning_rate": 8.080887422446656e-08, "loss": 0.0108, "step": 18067 }, { "epoch": 7.12, "learning_rate": 8.078837167296523e-08, "loss": 0.0113, "step": 18068 }, { "epoch": 7.12, "learning_rate": 8.076787106443228e-08, "loss": 0.0101, "step": 18069 }, { "epoch": 7.12, "learning_rate": 8.07473723992019e-08, "loss": 0.013, "step": 18070 }, { "epoch": 7.12, "learning_rate": 8.072687567760803e-08, "loss": 0.0149, "step": 18071 }, { "epoch": 7.12, "learning_rate": 8.070638089998488e-08, "loss": 0.0111, "step": 18072 }, { "epoch": 7.12, "learning_rate": 8.068588806666634e-08, "loss": 0.0086, "step": 18073 }, { "epoch": 7.12, "learning_rate": 8.066539717798648e-08, "loss": 0.0101, "step": 18074 }, { "epoch": 7.12, "learning_rate": 8.06449082342793e-08, "loss": 0.0098, "step": 18075 }, { "epoch": 7.12, "learning_rate": 8.062442123587878e-08, "loss": 0.0108, "step": 18076 }, { "epoch": 7.12, "learning_rate": 8.06039361831187e-08, "loss": 0.012, "step": 18077 }, { "epoch": 7.12, "learning_rate": 8.058345307633298e-08, "loss": 0.0112, "step": 18078 }, { "epoch": 7.13, "learning_rate": 8.05629719158555e-08, "loss": 0.0117, "step": 18079 }, { "epoch": 7.13, "learning_rate": 8.054249270202012e-08, "loss": 0.0131, "step": 18080 }, { "epoch": 7.13, "learning_rate": 8.052201543516049e-08, "loss": 0.0135, "step": 18081 }, { "epoch": 7.13, "learning_rate": 8.050154011561048e-08, "loss": 0.0114, "step": 18082 }, { "epoch": 7.13, "learning_rate": 8.04810667437036e-08, "loss": 0.0115, "step": 18083 }, { "epoch": 7.13, "learning_rate": 8.046059531977385e-08, "loss": 0.0116, "step": 18084 }, { "epoch": 7.13, "learning_rate": 8.044012584415462e-08, "loss": 0.0122, "step": 18085 }, { "epoch": 7.13, "learning_rate": 8.041965831717971e-08, "loss": 0.0101, "step": 18086 }, { "epoch": 7.13, "learning_rate": 8.039919273918249e-08, "loss": 0.0131, "step": 18087 }, { "epoch": 7.13, "learning_rate": 8.037872911049682e-08, "loss": 0.0122, "step": 18088 }, { "epoch": 7.13, "learning_rate": 8.035826743145598e-08, "loss": 0.011, "step": 18089 }, { "epoch": 7.13, "learning_rate": 8.03378077023936e-08, "loss": 0.0108, "step": 18090 }, { "epoch": 7.13, "learning_rate": 8.031734992364301e-08, "loss": 0.0101, "step": 18091 }, { "epoch": 7.13, "learning_rate": 8.029689409553775e-08, "loss": 0.0129, "step": 18092 }, { "epoch": 7.13, "learning_rate": 8.027644021841125e-08, "loss": 0.0098, "step": 18093 }, { "epoch": 7.13, "learning_rate": 8.025598829259672e-08, "loss": 0.0117, "step": 18094 }, { "epoch": 7.13, "learning_rate": 8.023553831842769e-08, "loss": 0.0112, "step": 18095 }, { "epoch": 7.13, "learning_rate": 8.02150902962372e-08, "loss": 0.0107, "step": 18096 }, { "epoch": 7.13, "learning_rate": 8.019464422635886e-08, "loss": 0.0131, "step": 18097 }, { "epoch": 7.13, "learning_rate": 8.017420010912563e-08, "loss": 0.0086, "step": 18098 }, { "epoch": 7.13, "learning_rate": 8.015375794487091e-08, "loss": 0.0094, "step": 18099 }, { "epoch": 7.13, "learning_rate": 8.013331773392773e-08, "loss": 0.0112, "step": 18100 }, { "epoch": 7.13, "learning_rate": 8.011287947662929e-08, "loss": 0.0111, "step": 18101 }, { "epoch": 7.13, "learning_rate": 8.00924431733087e-08, "loss": 0.0133, "step": 18102 }, { "epoch": 7.13, "learning_rate": 8.007200882429912e-08, "loss": 0.0097, "step": 18103 }, { "epoch": 7.13, "learning_rate": 8.005157642993345e-08, "loss": 0.0109, "step": 18104 }, { "epoch": 7.14, "learning_rate": 8.003114599054481e-08, "loss": 0.009, "step": 18105 }, { "epoch": 7.14, "learning_rate": 8.001071750646613e-08, "loss": 0.0111, "step": 18106 }, { "epoch": 7.14, "learning_rate": 7.999029097803047e-08, "loss": 0.0096, "step": 18107 }, { "epoch": 7.14, "learning_rate": 7.996986640557062e-08, "loss": 0.0136, "step": 18108 }, { "epoch": 7.14, "learning_rate": 7.99494437894195e-08, "loss": 0.0111, "step": 18109 }, { "epoch": 7.14, "learning_rate": 7.992902312990999e-08, "loss": 0.0092, "step": 18110 }, { "epoch": 7.14, "learning_rate": 7.9908604427375e-08, "loss": 0.0104, "step": 18111 }, { "epoch": 7.14, "learning_rate": 7.988818768214716e-08, "loss": 0.0107, "step": 18112 }, { "epoch": 7.14, "learning_rate": 7.986777289455939e-08, "loss": 0.009, "step": 18113 }, { "epoch": 7.14, "learning_rate": 7.984736006494426e-08, "loss": 0.0112, "step": 18114 }, { "epoch": 7.14, "learning_rate": 7.982694919363457e-08, "loss": 0.0094, "step": 18115 }, { "epoch": 7.14, "learning_rate": 7.980654028096303e-08, "loss": 0.0121, "step": 18116 }, { "epoch": 7.14, "learning_rate": 7.978613332726211e-08, "loss": 0.0129, "step": 18117 }, { "epoch": 7.14, "learning_rate": 7.976572833286454e-08, "loss": 0.0105, "step": 18118 }, { "epoch": 7.14, "learning_rate": 7.974532529810287e-08, "loss": 0.0128, "step": 18119 }, { "epoch": 7.14, "learning_rate": 7.97249242233097e-08, "loss": 0.0103, "step": 18120 }, { "epoch": 7.14, "learning_rate": 7.970452510881738e-08, "loss": 0.0105, "step": 18121 }, { "epoch": 7.14, "learning_rate": 7.968412795495856e-08, "loss": 0.0105, "step": 18122 }, { "epoch": 7.14, "learning_rate": 7.966373276206544e-08, "loss": 0.0111, "step": 18123 }, { "epoch": 7.14, "learning_rate": 7.964333953047073e-08, "loss": 0.009, "step": 18124 }, { "epoch": 7.14, "learning_rate": 7.962294826050662e-08, "loss": 0.0112, "step": 18125 }, { "epoch": 7.14, "learning_rate": 7.960255895250554e-08, "loss": 0.0079, "step": 18126 }, { "epoch": 7.14, "learning_rate": 7.958217160679964e-08, "loss": 0.0109, "step": 18127 }, { "epoch": 7.14, "learning_rate": 7.956178622372145e-08, "loss": 0.0128, "step": 18128 }, { "epoch": 7.14, "learning_rate": 7.954140280360304e-08, "loss": 0.0101, "step": 18129 }, { "epoch": 7.15, "learning_rate": 7.952102134677678e-08, "loss": 0.0145, "step": 18130 }, { "epoch": 7.15, "learning_rate": 7.950064185357468e-08, "loss": 0.0096, "step": 18131 }, { "epoch": 7.15, "learning_rate": 7.948026432432897e-08, "loss": 0.0118, "step": 18132 }, { "epoch": 7.15, "learning_rate": 7.94598887593718e-08, "loss": 0.0091, "step": 18133 }, { "epoch": 7.15, "learning_rate": 7.943951515903528e-08, "loss": 0.0112, "step": 18134 }, { "epoch": 7.15, "learning_rate": 7.94191435236514e-08, "loss": 0.0103, "step": 18135 }, { "epoch": 7.15, "learning_rate": 7.939877385355221e-08, "loss": 0.0133, "step": 18136 }, { "epoch": 7.15, "learning_rate": 7.937840614906977e-08, "loss": 0.0109, "step": 18137 }, { "epoch": 7.15, "learning_rate": 7.935804041053593e-08, "loss": 0.0089, "step": 18138 }, { "epoch": 7.15, "learning_rate": 7.933767663828274e-08, "loss": 0.0087, "step": 18139 }, { "epoch": 7.15, "learning_rate": 7.931731483264197e-08, "loss": 0.0134, "step": 18140 }, { "epoch": 7.15, "learning_rate": 7.929695499394554e-08, "loss": 0.0081, "step": 18141 }, { "epoch": 7.15, "learning_rate": 7.92765971225253e-08, "loss": 0.0111, "step": 18142 }, { "epoch": 7.15, "learning_rate": 7.925624121871312e-08, "loss": 0.0121, "step": 18143 }, { "epoch": 7.15, "learning_rate": 7.923588728284063e-08, "loss": 0.0112, "step": 18144 }, { "epoch": 7.15, "learning_rate": 7.921553531523966e-08, "loss": 0.0115, "step": 18145 }, { "epoch": 7.15, "learning_rate": 7.919518531624185e-08, "loss": 0.0093, "step": 18146 }, { "epoch": 7.15, "learning_rate": 7.917483728617902e-08, "loss": 0.012, "step": 18147 }, { "epoch": 7.15, "learning_rate": 7.915449122538262e-08, "loss": 0.0093, "step": 18148 }, { "epoch": 7.15, "learning_rate": 7.913414713418436e-08, "loss": 0.011, "step": 18149 }, { "epoch": 7.15, "learning_rate": 7.911380501291581e-08, "loss": 0.0121, "step": 18150 }, { "epoch": 7.15, "learning_rate": 7.909346486190857e-08, "loss": 0.009, "step": 18151 }, { "epoch": 7.15, "learning_rate": 7.907312668149405e-08, "loss": 0.013, "step": 18152 }, { "epoch": 7.15, "learning_rate": 7.905279047200382e-08, "loss": 0.0111, "step": 18153 }, { "epoch": 7.15, "learning_rate": 7.903245623376922e-08, "loss": 0.0128, "step": 18154 }, { "epoch": 7.16, "learning_rate": 7.901212396712173e-08, "loss": 0.0102, "step": 18155 }, { "epoch": 7.16, "learning_rate": 7.899179367239275e-08, "loss": 0.0095, "step": 18156 }, { "epoch": 7.16, "learning_rate": 7.89714653499137e-08, "loss": 0.0079, "step": 18157 }, { "epoch": 7.16, "learning_rate": 7.895113900001573e-08, "loss": 0.011, "step": 18158 }, { "epoch": 7.16, "learning_rate": 7.893081462303023e-08, "loss": 0.0128, "step": 18159 }, { "epoch": 7.16, "learning_rate": 7.891049221928854e-08, "loss": 0.0119, "step": 18160 }, { "epoch": 7.16, "learning_rate": 7.88901717891217e-08, "loss": 0.0124, "step": 18161 }, { "epoch": 7.16, "learning_rate": 7.886985333286107e-08, "loss": 0.013, "step": 18162 }, { "epoch": 7.16, "learning_rate": 7.88495368508376e-08, "loss": 0.0122, "step": 18163 }, { "epoch": 7.16, "learning_rate": 7.882922234338268e-08, "loss": 0.0111, "step": 18164 }, { "epoch": 7.16, "learning_rate": 7.88089098108272e-08, "loss": 0.0105, "step": 18165 }, { "epoch": 7.16, "learning_rate": 7.878859925350238e-08, "loss": 0.009, "step": 18166 }, { "epoch": 7.16, "learning_rate": 7.876829067173904e-08, "loss": 0.011, "step": 18167 }, { "epoch": 7.16, "learning_rate": 7.874798406586844e-08, "loss": 0.0102, "step": 18168 }, { "epoch": 7.16, "learning_rate": 7.872767943622136e-08, "loss": 0.0118, "step": 18169 }, { "epoch": 7.16, "learning_rate": 7.870737678312887e-08, "loss": 0.0098, "step": 18170 }, { "epoch": 7.16, "learning_rate": 7.868707610692169e-08, "loss": 0.0115, "step": 18171 }, { "epoch": 7.16, "learning_rate": 7.866677740793081e-08, "loss": 0.0107, "step": 18172 }, { "epoch": 7.16, "learning_rate": 7.864648068648707e-08, "loss": 0.0127, "step": 18173 }, { "epoch": 7.16, "learning_rate": 7.862618594292131e-08, "loss": 0.0098, "step": 18174 }, { "epoch": 7.16, "learning_rate": 7.860589317756418e-08, "loss": 0.0113, "step": 18175 }, { "epoch": 7.16, "learning_rate": 7.858560239074651e-08, "loss": 0.0098, "step": 18176 }, { "epoch": 7.16, "learning_rate": 7.856531358279907e-08, "loss": 0.0103, "step": 18177 }, { "epoch": 7.16, "learning_rate": 7.854502675405236e-08, "loss": 0.0087, "step": 18178 }, { "epoch": 7.16, "learning_rate": 7.852474190483716e-08, "loss": 0.0101, "step": 18179 }, { "epoch": 7.16, "learning_rate": 7.850445903548409e-08, "loss": 0.0098, "step": 18180 }, { "epoch": 7.17, "learning_rate": 7.848417814632362e-08, "loss": 0.0085, "step": 18181 }, { "epoch": 7.17, "learning_rate": 7.846389923768635e-08, "loss": 0.0092, "step": 18182 }, { "epoch": 7.17, "learning_rate": 7.844362230990289e-08, "loss": 0.0134, "step": 18183 }, { "epoch": 7.17, "learning_rate": 7.842334736330358e-08, "loss": 0.0109, "step": 18184 }, { "epoch": 7.17, "learning_rate": 7.840307439821894e-08, "loss": 0.0132, "step": 18185 }, { "epoch": 7.17, "learning_rate": 7.838280341497936e-08, "loss": 0.0109, "step": 18186 }, { "epoch": 7.17, "learning_rate": 7.836253441391532e-08, "loss": 0.0119, "step": 18187 }, { "epoch": 7.17, "learning_rate": 7.834226739535703e-08, "loss": 0.0127, "step": 18188 }, { "epoch": 7.17, "learning_rate": 7.832200235963492e-08, "loss": 0.0107, "step": 18189 }, { "epoch": 7.17, "learning_rate": 7.83017393070792e-08, "loss": 0.0124, "step": 18190 }, { "epoch": 7.17, "learning_rate": 7.828147823802025e-08, "loss": 0.0119, "step": 18191 }, { "epoch": 7.17, "learning_rate": 7.826121915278814e-08, "loss": 0.0139, "step": 18192 }, { "epoch": 7.17, "learning_rate": 7.824096205171323e-08, "loss": 0.0113, "step": 18193 }, { "epoch": 7.17, "learning_rate": 7.822070693512541e-08, "loss": 0.0081, "step": 18194 }, { "epoch": 7.17, "learning_rate": 7.820045380335514e-08, "loss": 0.0121, "step": 18195 }, { "epoch": 7.17, "learning_rate": 7.818020265673229e-08, "loss": 0.0122, "step": 18196 }, { "epoch": 7.17, "learning_rate": 7.815995349558706e-08, "loss": 0.0121, "step": 18197 }, { "epoch": 7.17, "learning_rate": 7.813970632024936e-08, "loss": 0.0091, "step": 18198 }, { "epoch": 7.17, "learning_rate": 7.811946113104921e-08, "loss": 0.0125, "step": 18199 }, { "epoch": 7.17, "learning_rate": 7.809921792831671e-08, "loss": 0.0117, "step": 18200 }, { "epoch": 7.17, "learning_rate": 7.807897671238159e-08, "loss": 0.0093, "step": 18201 }, { "epoch": 7.17, "learning_rate": 7.805873748357386e-08, "loss": 0.0115, "step": 18202 }, { "epoch": 7.17, "learning_rate": 7.803850024222338e-08, "loss": 0.0125, "step": 18203 }, { "epoch": 7.17, "learning_rate": 7.801826498866007e-08, "loss": 0.0118, "step": 18204 }, { "epoch": 7.17, "learning_rate": 7.799803172321355e-08, "loss": 0.0108, "step": 18205 }, { "epoch": 7.18, "learning_rate": 7.797780044621378e-08, "loss": 0.0113, "step": 18206 }, { "epoch": 7.18, "learning_rate": 7.795757115799029e-08, "loss": 0.0112, "step": 18207 }, { "epoch": 7.18, "learning_rate": 7.793734385887303e-08, "loss": 0.0111, "step": 18208 }, { "epoch": 7.18, "learning_rate": 7.791711854919148e-08, "loss": 0.0105, "step": 18209 }, { "epoch": 7.18, "learning_rate": 7.789689522927541e-08, "loss": 0.0137, "step": 18210 }, { "epoch": 7.18, "learning_rate": 7.787667389945432e-08, "loss": 0.01, "step": 18211 }, { "epoch": 7.18, "learning_rate": 7.785645456005783e-08, "loss": 0.012, "step": 18212 }, { "epoch": 7.18, "learning_rate": 7.78362372114155e-08, "loss": 0.0121, "step": 18213 }, { "epoch": 7.18, "learning_rate": 7.781602185385689e-08, "loss": 0.0112, "step": 18214 }, { "epoch": 7.18, "learning_rate": 7.779580848771136e-08, "loss": 0.009, "step": 18215 }, { "epoch": 7.18, "learning_rate": 7.777559711330846e-08, "loss": 0.0105, "step": 18216 }, { "epoch": 7.18, "learning_rate": 7.775538773097752e-08, "loss": 0.0136, "step": 18217 }, { "epoch": 7.18, "learning_rate": 7.773518034104807e-08, "loss": 0.0115, "step": 18218 }, { "epoch": 7.18, "learning_rate": 7.771497494384927e-08, "loss": 0.0112, "step": 18219 }, { "epoch": 7.18, "learning_rate": 7.76947715397106e-08, "loss": 0.0112, "step": 18220 }, { "epoch": 7.18, "learning_rate": 7.767457012896122e-08, "loss": 0.011, "step": 18221 }, { "epoch": 7.18, "learning_rate": 7.765437071193042e-08, "loss": 0.0106, "step": 18222 }, { "epoch": 7.18, "learning_rate": 7.763417328894741e-08, "loss": 0.0112, "step": 18223 }, { "epoch": 7.18, "learning_rate": 7.761397786034152e-08, "loss": 0.0087, "step": 18224 }, { "epoch": 7.18, "learning_rate": 7.759378442644167e-08, "loss": 0.0134, "step": 18225 }, { "epoch": 7.18, "learning_rate": 7.757359298757713e-08, "loss": 0.0115, "step": 18226 }, { "epoch": 7.18, "learning_rate": 7.7553403544077e-08, "loss": 0.0105, "step": 18227 }, { "epoch": 7.18, "learning_rate": 7.753321609627024e-08, "loss": 0.0112, "step": 18228 }, { "epoch": 7.18, "learning_rate": 7.751303064448593e-08, "loss": 0.0121, "step": 18229 }, { "epoch": 7.18, "learning_rate": 7.749284718905307e-08, "loss": 0.0108, "step": 18230 }, { "epoch": 7.18, "learning_rate": 7.747266573030067e-08, "loss": 0.0102, "step": 18231 }, { "epoch": 7.19, "learning_rate": 7.745248626855754e-08, "loss": 0.0104, "step": 18232 }, { "epoch": 7.19, "learning_rate": 7.743230880415268e-08, "loss": 0.011, "step": 18233 }, { "epoch": 7.19, "learning_rate": 7.741213333741479e-08, "loss": 0.011, "step": 18234 }, { "epoch": 7.19, "learning_rate": 7.739195986867292e-08, "loss": 0.0123, "step": 18235 }, { "epoch": 7.19, "learning_rate": 7.737178839825574e-08, "loss": 0.0102, "step": 18236 }, { "epoch": 7.19, "learning_rate": 7.735161892649207e-08, "loss": 0.0104, "step": 18237 }, { "epoch": 7.19, "learning_rate": 7.733145145371047e-08, "loss": 0.0139, "step": 18238 }, { "epoch": 7.19, "learning_rate": 7.731128598023993e-08, "loss": 0.0126, "step": 18239 }, { "epoch": 7.19, "learning_rate": 7.729112250640888e-08, "loss": 0.0097, "step": 18240 }, { "epoch": 7.19, "learning_rate": 7.727096103254614e-08, "loss": 0.0118, "step": 18241 }, { "epoch": 7.19, "learning_rate": 7.72508015589801e-08, "loss": 0.0107, "step": 18242 }, { "epoch": 7.19, "learning_rate": 7.723064408603945e-08, "loss": 0.0125, "step": 18243 }, { "epoch": 7.19, "learning_rate": 7.721048861405277e-08, "loss": 0.0092, "step": 18244 }, { "epoch": 7.19, "learning_rate": 7.719033514334846e-08, "loss": 0.0087, "step": 18245 }, { "epoch": 7.19, "learning_rate": 7.717018367425503e-08, "loss": 0.0096, "step": 18246 }, { "epoch": 7.19, "learning_rate": 7.715003420710094e-08, "loss": 0.0087, "step": 18247 }, { "epoch": 7.19, "learning_rate": 7.712988674221463e-08, "loss": 0.0103, "step": 18248 }, { "epoch": 7.19, "learning_rate": 7.710974127992436e-08, "loss": 0.0106, "step": 18249 }, { "epoch": 7.19, "learning_rate": 7.708959782055859e-08, "loss": 0.0107, "step": 18250 }, { "epoch": 7.19, "learning_rate": 7.706945636444553e-08, "loss": 0.0099, "step": 18251 }, { "epoch": 7.19, "learning_rate": 7.70493169119135e-08, "loss": 0.0118, "step": 18252 }, { "epoch": 7.19, "learning_rate": 7.702917946329073e-08, "loss": 0.0097, "step": 18253 }, { "epoch": 7.19, "learning_rate": 7.700904401890551e-08, "loss": 0.0127, "step": 18254 }, { "epoch": 7.19, "learning_rate": 7.698891057908587e-08, "loss": 0.0094, "step": 18255 }, { "epoch": 7.19, "learning_rate": 7.696877914416004e-08, "loss": 0.01, "step": 18256 }, { "epoch": 7.2, "learning_rate": 7.694864971445614e-08, "loss": 0.0104, "step": 18257 }, { "epoch": 7.2, "learning_rate": 7.692852229030227e-08, "loss": 0.01, "step": 18258 }, { "epoch": 7.2, "learning_rate": 7.69083968720264e-08, "loss": 0.0106, "step": 18259 }, { "epoch": 7.2, "learning_rate": 7.688827345995663e-08, "loss": 0.0121, "step": 18260 }, { "epoch": 7.2, "learning_rate": 7.686815205442076e-08, "loss": 0.0117, "step": 18261 }, { "epoch": 7.2, "learning_rate": 7.684803265574702e-08, "loss": 0.0158, "step": 18262 }, { "epoch": 7.2, "learning_rate": 7.682791526426311e-08, "loss": 0.0089, "step": 18263 }, { "epoch": 7.2, "learning_rate": 7.680779988029704e-08, "loss": 0.0115, "step": 18264 }, { "epoch": 7.2, "learning_rate": 7.678768650417655e-08, "loss": 0.0114, "step": 18265 }, { "epoch": 7.2, "learning_rate": 7.676757513622948e-08, "loss": 0.0122, "step": 18266 }, { "epoch": 7.2, "learning_rate": 7.674746577678375e-08, "loss": 0.0103, "step": 18267 }, { "epoch": 7.2, "learning_rate": 7.672735842616692e-08, "loss": 0.0084, "step": 18268 }, { "epoch": 7.2, "learning_rate": 7.670725308470678e-08, "loss": 0.0121, "step": 18269 }, { "epoch": 7.2, "learning_rate": 7.668714975273105e-08, "loss": 0.0086, "step": 18270 }, { "epoch": 7.2, "learning_rate": 7.666704843056744e-08, "loss": 0.0101, "step": 18271 }, { "epoch": 7.2, "learning_rate": 7.664694911854341e-08, "loss": 0.0124, "step": 18272 }, { "epoch": 7.2, "learning_rate": 7.66268518169867e-08, "loss": 0.0099, "step": 18273 }, { "epoch": 7.2, "learning_rate": 7.660675652622468e-08, "loss": 0.0099, "step": 18274 }, { "epoch": 7.2, "learning_rate": 7.658666324658511e-08, "loss": 0.0108, "step": 18275 }, { "epoch": 7.2, "learning_rate": 7.656657197839531e-08, "loss": 0.0135, "step": 18276 }, { "epoch": 7.2, "learning_rate": 7.654648272198286e-08, "loss": 0.0115, "step": 18277 }, { "epoch": 7.2, "learning_rate": 7.652639547767495e-08, "loss": 0.0122, "step": 18278 }, { "epoch": 7.2, "learning_rate": 7.65063102457993e-08, "loss": 0.0116, "step": 18279 }, { "epoch": 7.2, "learning_rate": 7.648622702668301e-08, "loss": 0.0104, "step": 18280 }, { "epoch": 7.2, "learning_rate": 7.64661458206536e-08, "loss": 0.0096, "step": 18281 }, { "epoch": 7.21, "learning_rate": 7.644606662803819e-08, "loss": 0.0112, "step": 18282 }, { "epoch": 7.21, "learning_rate": 7.64259894491641e-08, "loss": 0.0105, "step": 18283 }, { "epoch": 7.21, "learning_rate": 7.640591428435856e-08, "loss": 0.0119, "step": 18284 }, { "epoch": 7.21, "learning_rate": 7.638584113394886e-08, "loss": 0.0108, "step": 18285 }, { "epoch": 7.21, "learning_rate": 7.6365769998262e-08, "loss": 0.0107, "step": 18286 }, { "epoch": 7.21, "learning_rate": 7.634570087762522e-08, "loss": 0.0101, "step": 18287 }, { "epoch": 7.21, "learning_rate": 7.632563377236563e-08, "loss": 0.0108, "step": 18288 }, { "epoch": 7.21, "learning_rate": 7.630556868281015e-08, "loss": 0.0106, "step": 18289 }, { "epoch": 7.21, "learning_rate": 7.628550560928602e-08, "loss": 0.0115, "step": 18290 }, { "epoch": 7.21, "learning_rate": 7.626544455212003e-08, "loss": 0.011, "step": 18291 }, { "epoch": 7.21, "learning_rate": 7.624538551163924e-08, "loss": 0.0093, "step": 18292 }, { "epoch": 7.21, "learning_rate": 7.622532848817059e-08, "loss": 0.0111, "step": 18293 }, { "epoch": 7.21, "learning_rate": 7.620527348204102e-08, "loss": 0.0124, "step": 18294 }, { "epoch": 7.21, "learning_rate": 7.618522049357729e-08, "loss": 0.0122, "step": 18295 }, { "epoch": 7.21, "learning_rate": 7.61651695231063e-08, "loss": 0.0099, "step": 18296 }, { "epoch": 7.21, "learning_rate": 7.614512057095484e-08, "loss": 0.0093, "step": 18297 }, { "epoch": 7.21, "learning_rate": 7.612507363744971e-08, "loss": 0.0089, "step": 18298 }, { "epoch": 7.21, "learning_rate": 7.610502872291758e-08, "loss": 0.0109, "step": 18299 }, { "epoch": 7.21, "learning_rate": 7.608498582768523e-08, "loss": 0.0124, "step": 18300 }, { "epoch": 7.21, "learning_rate": 7.606494495207916e-08, "loss": 0.0116, "step": 18301 }, { "epoch": 7.21, "learning_rate": 7.604490609642626e-08, "loss": 0.0109, "step": 18302 }, { "epoch": 7.21, "learning_rate": 7.602486926105294e-08, "loss": 0.0109, "step": 18303 }, { "epoch": 7.21, "learning_rate": 7.600483444628588e-08, "loss": 0.0109, "step": 18304 }, { "epoch": 7.21, "learning_rate": 7.598480165245151e-08, "loss": 0.0094, "step": 18305 }, { "epoch": 7.21, "learning_rate": 7.596477087987638e-08, "loss": 0.0118, "step": 18306 }, { "epoch": 7.21, "learning_rate": 7.5944742128887e-08, "loss": 0.0099, "step": 18307 }, { "epoch": 7.22, "learning_rate": 7.592471539980982e-08, "loss": 0.011, "step": 18308 }, { "epoch": 7.22, "learning_rate": 7.590469069297116e-08, "loss": 0.012, "step": 18309 }, { "epoch": 7.22, "learning_rate": 7.588466800869742e-08, "loss": 0.0109, "step": 18310 }, { "epoch": 7.22, "learning_rate": 7.586464734731504e-08, "loss": 0.0109, "step": 18311 }, { "epoch": 7.22, "learning_rate": 7.584462870915019e-08, "loss": 0.0103, "step": 18312 }, { "epoch": 7.22, "learning_rate": 7.582461209452924e-08, "loss": 0.0123, "step": 18313 }, { "epoch": 7.22, "learning_rate": 7.580459750377826e-08, "loss": 0.0092, "step": 18314 }, { "epoch": 7.22, "learning_rate": 7.57845849372237e-08, "loss": 0.0111, "step": 18315 }, { "epoch": 7.22, "learning_rate": 7.576457439519159e-08, "loss": 0.0117, "step": 18316 }, { "epoch": 7.22, "learning_rate": 7.574456587800815e-08, "loss": 0.0102, "step": 18317 }, { "epoch": 7.22, "learning_rate": 7.57245593859993e-08, "loss": 0.0107, "step": 18318 }, { "epoch": 7.22, "learning_rate": 7.570455491949139e-08, "loss": 0.011, "step": 18319 }, { "epoch": 7.22, "learning_rate": 7.568455247881028e-08, "loss": 0.01, "step": 18320 }, { "epoch": 7.22, "learning_rate": 7.566455206428208e-08, "loss": 0.0088, "step": 18321 }, { "epoch": 7.22, "learning_rate": 7.564455367623264e-08, "loss": 0.0084, "step": 18322 }, { "epoch": 7.22, "learning_rate": 7.562455731498798e-08, "loss": 0.0116, "step": 18323 }, { "epoch": 7.22, "learning_rate": 7.560456298087401e-08, "loss": 0.0105, "step": 18324 }, { "epoch": 7.22, "learning_rate": 7.558457067421668e-08, "loss": 0.0108, "step": 18325 }, { "epoch": 7.22, "learning_rate": 7.556458039534168e-08, "loss": 0.0128, "step": 18326 }, { "epoch": 7.22, "learning_rate": 7.554459214457489e-08, "loss": 0.0117, "step": 18327 }, { "epoch": 7.22, "learning_rate": 7.552460592224219e-08, "loss": 0.0104, "step": 18328 }, { "epoch": 7.22, "learning_rate": 7.550462172866916e-08, "loss": 0.0127, "step": 18329 }, { "epoch": 7.22, "learning_rate": 7.548463956418156e-08, "loss": 0.0097, "step": 18330 }, { "epoch": 7.22, "learning_rate": 7.546465942910519e-08, "loss": 0.0098, "step": 18331 }, { "epoch": 7.22, "learning_rate": 7.544468132376551e-08, "loss": 0.0122, "step": 18332 }, { "epoch": 7.23, "learning_rate": 7.542470524848822e-08, "loss": 0.012, "step": 18333 }, { "epoch": 7.23, "learning_rate": 7.5404731203599e-08, "loss": 0.0096, "step": 18334 }, { "epoch": 7.23, "learning_rate": 7.538475918942322e-08, "loss": 0.0139, "step": 18335 }, { "epoch": 7.23, "learning_rate": 7.536478920628646e-08, "loss": 0.0106, "step": 18336 }, { "epoch": 7.23, "learning_rate": 7.53448212545142e-08, "loss": 0.0103, "step": 18337 }, { "epoch": 7.23, "learning_rate": 7.5324855334432e-08, "loss": 0.0114, "step": 18338 }, { "epoch": 7.23, "learning_rate": 7.53048914463651e-08, "loss": 0.0117, "step": 18339 }, { "epoch": 7.23, "learning_rate": 7.5284929590639e-08, "loss": 0.0097, "step": 18340 }, { "epoch": 7.23, "learning_rate": 7.526496976757887e-08, "loss": 0.0094, "step": 18341 }, { "epoch": 7.23, "learning_rate": 7.52450119775103e-08, "loss": 0.0109, "step": 18342 }, { "epoch": 7.23, "learning_rate": 7.522505622075836e-08, "loss": 0.0126, "step": 18343 }, { "epoch": 7.23, "learning_rate": 7.520510249764844e-08, "loss": 0.0137, "step": 18344 }, { "epoch": 7.23, "learning_rate": 7.518515080850556e-08, "loss": 0.0092, "step": 18345 }, { "epoch": 7.23, "learning_rate": 7.516520115365514e-08, "loss": 0.0104, "step": 18346 }, { "epoch": 7.23, "learning_rate": 7.514525353342216e-08, "loss": 0.0098, "step": 18347 }, { "epoch": 7.23, "learning_rate": 7.512530794813186e-08, "loss": 0.0122, "step": 18348 }, { "epoch": 7.23, "learning_rate": 7.510536439810914e-08, "loss": 0.0091, "step": 18349 }, { "epoch": 7.23, "learning_rate": 7.508542288367923e-08, "loss": 0.0113, "step": 18350 }, { "epoch": 7.23, "learning_rate": 7.506548340516702e-08, "loss": 0.01, "step": 18351 }, { "epoch": 7.23, "learning_rate": 7.504554596289767e-08, "loss": 0.0105, "step": 18352 }, { "epoch": 7.23, "learning_rate": 7.502561055719592e-08, "loss": 0.011, "step": 18353 }, { "epoch": 7.23, "learning_rate": 7.500567718838677e-08, "loss": 0.0123, "step": 18354 }, { "epoch": 7.23, "learning_rate": 7.498574585679518e-08, "loss": 0.0128, "step": 18355 }, { "epoch": 7.23, "learning_rate": 7.496581656274588e-08, "loss": 0.0103, "step": 18356 }, { "epoch": 7.23, "learning_rate": 7.49458893065638e-08, "loss": 0.0123, "step": 18357 }, { "epoch": 7.24, "learning_rate": 7.492596408857354e-08, "loss": 0.0104, "step": 18358 }, { "epoch": 7.24, "learning_rate": 7.490604090910009e-08, "loss": 0.0097, "step": 18359 }, { "epoch": 7.24, "learning_rate": 7.488611976846799e-08, "loss": 0.0125, "step": 18360 }, { "epoch": 7.24, "learning_rate": 7.486620066700206e-08, "loss": 0.0111, "step": 18361 }, { "epoch": 7.24, "learning_rate": 7.484628360502682e-08, "loss": 0.0108, "step": 18362 }, { "epoch": 7.24, "learning_rate": 7.482636858286693e-08, "loss": 0.0114, "step": 18363 }, { "epoch": 7.24, "learning_rate": 7.4806455600847e-08, "loss": 0.0127, "step": 18364 }, { "epoch": 7.24, "learning_rate": 7.478654465929163e-08, "loss": 0.011, "step": 18365 }, { "epoch": 7.24, "learning_rate": 7.476663575852523e-08, "loss": 0.0103, "step": 18366 }, { "epoch": 7.24, "learning_rate": 7.474672889887232e-08, "loss": 0.0087, "step": 18367 }, { "epoch": 7.24, "learning_rate": 7.472682408065736e-08, "loss": 0.0106, "step": 18368 }, { "epoch": 7.24, "learning_rate": 7.470692130420484e-08, "loss": 0.0087, "step": 18369 }, { "epoch": 7.24, "learning_rate": 7.468702056983903e-08, "loss": 0.0111, "step": 18370 }, { "epoch": 7.24, "learning_rate": 7.46671218778844e-08, "loss": 0.0105, "step": 18371 }, { "epoch": 7.24, "learning_rate": 7.464722522866511e-08, "loss": 0.0111, "step": 18372 }, { "epoch": 7.24, "learning_rate": 7.462733062250555e-08, "loss": 0.0121, "step": 18373 }, { "epoch": 7.24, "learning_rate": 7.460743805972993e-08, "loss": 0.0109, "step": 18374 }, { "epoch": 7.24, "learning_rate": 7.458754754066259e-08, "loss": 0.0092, "step": 18375 }, { "epoch": 7.24, "learning_rate": 7.456765906562756e-08, "loss": 0.0142, "step": 18376 }, { "epoch": 7.24, "learning_rate": 7.454777263494902e-08, "loss": 0.0089, "step": 18377 }, { "epoch": 7.24, "learning_rate": 7.452788824895122e-08, "loss": 0.0101, "step": 18378 }, { "epoch": 7.24, "learning_rate": 7.450800590795807e-08, "loss": 0.0077, "step": 18379 }, { "epoch": 7.24, "learning_rate": 7.448812561229376e-08, "loss": 0.0097, "step": 18380 }, { "epoch": 7.24, "learning_rate": 7.44682473622821e-08, "loss": 0.0115, "step": 18381 }, { "epoch": 7.24, "learning_rate": 7.444837115824738e-08, "loss": 0.0115, "step": 18382 }, { "epoch": 7.24, "learning_rate": 7.44284970005133e-08, "loss": 0.0145, "step": 18383 }, { "epoch": 7.25, "learning_rate": 7.440862488940396e-08, "loss": 0.0095, "step": 18384 }, { "epoch": 7.25, "learning_rate": 7.438875482524301e-08, "loss": 0.01, "step": 18385 }, { "epoch": 7.25, "learning_rate": 7.43688868083546e-08, "loss": 0.0126, "step": 18386 }, { "epoch": 7.25, "learning_rate": 7.434902083906232e-08, "loss": 0.0119, "step": 18387 }, { "epoch": 7.25, "learning_rate": 7.43291569176901e-08, "loss": 0.0113, "step": 18388 }, { "epoch": 7.25, "learning_rate": 7.43092950445615e-08, "loss": 0.0119, "step": 18389 }, { "epoch": 7.25, "learning_rate": 7.428943522000049e-08, "loss": 0.0127, "step": 18390 }, { "epoch": 7.25, "learning_rate": 7.426957744433055e-08, "loss": 0.0166, "step": 18391 }, { "epoch": 7.25, "learning_rate": 7.424972171787548e-08, "loss": 0.0093, "step": 18392 }, { "epoch": 7.25, "learning_rate": 7.422986804095876e-08, "loss": 0.0129, "step": 18393 }, { "epoch": 7.25, "learning_rate": 7.421001641390406e-08, "loss": 0.0108, "step": 18394 }, { "epoch": 7.25, "learning_rate": 7.419016683703497e-08, "loss": 0.011, "step": 18395 }, { "epoch": 7.25, "learning_rate": 7.417031931067488e-08, "loss": 0.0133, "step": 18396 }, { "epoch": 7.25, "learning_rate": 7.415047383514735e-08, "loss": 0.0107, "step": 18397 }, { "epoch": 7.25, "learning_rate": 7.413063041077583e-08, "loss": 0.0127, "step": 18398 }, { "epoch": 7.25, "learning_rate": 7.41107890378838e-08, "loss": 0.0114, "step": 18399 }, { "epoch": 7.25, "learning_rate": 7.409094971679451e-08, "loss": 0.0105, "step": 18400 }, { "epoch": 7.25, "learning_rate": 7.407111244783144e-08, "loss": 0.0114, "step": 18401 }, { "epoch": 7.25, "learning_rate": 7.40512772313178e-08, "loss": 0.0123, "step": 18402 }, { "epoch": 7.25, "learning_rate": 7.40314440675769e-08, "loss": 0.0116, "step": 18403 }, { "epoch": 7.25, "learning_rate": 7.401161295693204e-08, "loss": 0.011, "step": 18404 }, { "epoch": 7.25, "learning_rate": 7.399178389970644e-08, "loss": 0.0099, "step": 18405 }, { "epoch": 7.25, "learning_rate": 7.39719568962232e-08, "loss": 0.009, "step": 18406 }, { "epoch": 7.25, "learning_rate": 7.395213194680554e-08, "loss": 0.0148, "step": 18407 }, { "epoch": 7.25, "learning_rate": 7.393230905177654e-08, "loss": 0.0093, "step": 18408 }, { "epoch": 7.26, "learning_rate": 7.391248821145937e-08, "loss": 0.0126, "step": 18409 }, { "epoch": 7.26, "learning_rate": 7.389266942617694e-08, "loss": 0.0115, "step": 18410 }, { "epoch": 7.26, "learning_rate": 7.387285269625241e-08, "loss": 0.0091, "step": 18411 }, { "epoch": 7.26, "learning_rate": 7.385303802200855e-08, "loss": 0.0108, "step": 18412 }, { "epoch": 7.26, "learning_rate": 7.38332254037686e-08, "loss": 0.0109, "step": 18413 }, { "epoch": 7.26, "learning_rate": 7.381341484185524e-08, "loss": 0.0109, "step": 18414 }, { "epoch": 7.26, "learning_rate": 7.379360633659151e-08, "loss": 0.0114, "step": 18415 }, { "epoch": 7.26, "learning_rate": 7.377379988830012e-08, "loss": 0.012, "step": 18416 }, { "epoch": 7.26, "learning_rate": 7.375399549730394e-08, "loss": 0.0104, "step": 18417 }, { "epoch": 7.26, "learning_rate": 7.373419316392584e-08, "loss": 0.0085, "step": 18418 }, { "epoch": 7.26, "learning_rate": 7.371439288848842e-08, "loss": 0.0094, "step": 18419 }, { "epoch": 7.26, "learning_rate": 7.369459467131448e-08, "loss": 0.0094, "step": 18420 }, { "epoch": 7.26, "learning_rate": 7.367479851272664e-08, "loss": 0.0131, "step": 18421 }, { "epoch": 7.26, "learning_rate": 7.365500441304771e-08, "loss": 0.0141, "step": 18422 }, { "epoch": 7.26, "learning_rate": 7.363521237260009e-08, "loss": 0.0118, "step": 18423 }, { "epoch": 7.26, "learning_rate": 7.361542239170653e-08, "loss": 0.0099, "step": 18424 }, { "epoch": 7.26, "learning_rate": 7.359563447068935e-08, "loss": 0.0117, "step": 18425 }, { "epoch": 7.26, "learning_rate": 7.357584860987139e-08, "loss": 0.0112, "step": 18426 }, { "epoch": 7.26, "learning_rate": 7.355606480957485e-08, "loss": 0.0105, "step": 18427 }, { "epoch": 7.26, "learning_rate": 7.353628307012236e-08, "loss": 0.0121, "step": 18428 }, { "epoch": 7.26, "learning_rate": 7.351650339183612e-08, "loss": 0.0116, "step": 18429 }, { "epoch": 7.26, "learning_rate": 7.349672577503878e-08, "loss": 0.0119, "step": 18430 }, { "epoch": 7.26, "learning_rate": 7.347695022005247e-08, "loss": 0.011, "step": 18431 }, { "epoch": 7.26, "learning_rate": 7.345717672719966e-08, "loss": 0.0113, "step": 18432 }, { "epoch": 7.26, "learning_rate": 7.343740529680247e-08, "loss": 0.0093, "step": 18433 }, { "epoch": 7.26, "learning_rate": 7.341763592918322e-08, "loss": 0.0112, "step": 18434 }, { "epoch": 7.27, "learning_rate": 7.339786862466413e-08, "loss": 0.0114, "step": 18435 }, { "epoch": 7.27, "learning_rate": 7.337810338356742e-08, "loss": 0.0131, "step": 18436 }, { "epoch": 7.27, "learning_rate": 7.335834020621514e-08, "loss": 0.0101, "step": 18437 }, { "epoch": 7.27, "learning_rate": 7.333857909292941e-08, "loss": 0.0093, "step": 18438 }, { "epoch": 7.27, "learning_rate": 7.331882004403244e-08, "loss": 0.0123, "step": 18439 }, { "epoch": 7.27, "learning_rate": 7.329906305984608e-08, "loss": 0.009, "step": 18440 }, { "epoch": 7.27, "learning_rate": 7.32793081406925e-08, "loss": 0.0084, "step": 18441 }, { "epoch": 7.27, "learning_rate": 7.325955528689354e-08, "loss": 0.0095, "step": 18442 }, { "epoch": 7.27, "learning_rate": 7.323980449877121e-08, "loss": 0.0107, "step": 18443 }, { "epoch": 7.27, "learning_rate": 7.32200557766474e-08, "loss": 0.0102, "step": 18444 }, { "epoch": 7.27, "learning_rate": 7.320030912084408e-08, "loss": 0.0118, "step": 18445 }, { "epoch": 7.27, "learning_rate": 7.31805645316829e-08, "loss": 0.0124, "step": 18446 }, { "epoch": 7.27, "learning_rate": 7.316082200948583e-08, "loss": 0.0087, "step": 18447 }, { "epoch": 7.27, "learning_rate": 7.314108155457457e-08, "loss": 0.0135, "step": 18448 }, { "epoch": 7.27, "learning_rate": 7.312134316727093e-08, "loss": 0.0118, "step": 18449 }, { "epoch": 7.27, "learning_rate": 7.31016068478965e-08, "loss": 0.0111, "step": 18450 }, { "epoch": 7.27, "learning_rate": 7.30818725967731e-08, "loss": 0.0089, "step": 18451 }, { "epoch": 7.27, "learning_rate": 7.306214041422211e-08, "loss": 0.0122, "step": 18452 }, { "epoch": 7.27, "learning_rate": 7.304241030056548e-08, "loss": 0.0122, "step": 18453 }, { "epoch": 7.27, "learning_rate": 7.302268225612453e-08, "loss": 0.0103, "step": 18454 }, { "epoch": 7.27, "learning_rate": 7.300295628122095e-08, "loss": 0.0105, "step": 18455 }, { "epoch": 7.27, "learning_rate": 7.29832323761761e-08, "loss": 0.0092, "step": 18456 }, { "epoch": 7.27, "learning_rate": 7.29635105413115e-08, "loss": 0.0087, "step": 18457 }, { "epoch": 7.27, "learning_rate": 7.294379077694863e-08, "loss": 0.0111, "step": 18458 }, { "epoch": 7.27, "learning_rate": 7.292407308340891e-08, "loss": 0.0102, "step": 18459 }, { "epoch": 7.28, "learning_rate": 7.290435746101362e-08, "loss": 0.0113, "step": 18460 }, { "epoch": 7.28, "learning_rate": 7.288464391008413e-08, "loss": 0.0111, "step": 18461 }, { "epoch": 7.28, "learning_rate": 7.286493243094183e-08, "loss": 0.0103, "step": 18462 }, { "epoch": 7.28, "learning_rate": 7.284522302390783e-08, "loss": 0.0104, "step": 18463 }, { "epoch": 7.28, "learning_rate": 7.282551568930351e-08, "loss": 0.0096, "step": 18464 }, { "epoch": 7.28, "learning_rate": 7.280581042744987e-08, "loss": 0.0113, "step": 18465 }, { "epoch": 7.28, "learning_rate": 7.278610723866833e-08, "loss": 0.0129, "step": 18466 }, { "epoch": 7.28, "learning_rate": 7.276640612327984e-08, "loss": 0.0097, "step": 18467 }, { "epoch": 7.28, "learning_rate": 7.274670708160562e-08, "loss": 0.0135, "step": 18468 }, { "epoch": 7.28, "learning_rate": 7.272701011396663e-08, "loss": 0.0116, "step": 18469 }, { "epoch": 7.28, "learning_rate": 7.27073152206839e-08, "loss": 0.0114, "step": 18470 }, { "epoch": 7.28, "learning_rate": 7.268762240207851e-08, "loss": 0.0109, "step": 18471 }, { "epoch": 7.28, "learning_rate": 7.266793165847142e-08, "loss": 0.0111, "step": 18472 }, { "epoch": 7.28, "learning_rate": 7.264824299018346e-08, "loss": 0.0128, "step": 18473 }, { "epoch": 7.28, "learning_rate": 7.26285563975356e-08, "loss": 0.013, "step": 18474 }, { "epoch": 7.28, "learning_rate": 7.260887188084868e-08, "loss": 0.0111, "step": 18475 }, { "epoch": 7.28, "learning_rate": 7.25891894404436e-08, "loss": 0.0105, "step": 18476 }, { "epoch": 7.28, "learning_rate": 7.256950907664104e-08, "loss": 0.0098, "step": 18477 }, { "epoch": 7.28, "learning_rate": 7.254983078976178e-08, "loss": 0.0097, "step": 18478 }, { "epoch": 7.28, "learning_rate": 7.25301545801266e-08, "loss": 0.0105, "step": 18479 }, { "epoch": 7.28, "learning_rate": 7.251048044805624e-08, "loss": 0.0126, "step": 18480 }, { "epoch": 7.28, "learning_rate": 7.249080839387121e-08, "loss": 0.0091, "step": 18481 }, { "epoch": 7.28, "learning_rate": 7.247113841789228e-08, "loss": 0.01, "step": 18482 }, { "epoch": 7.28, "learning_rate": 7.245147052043989e-08, "loss": 0.0091, "step": 18483 }, { "epoch": 7.28, "learning_rate": 7.243180470183469e-08, "loss": 0.0109, "step": 18484 }, { "epoch": 7.29, "learning_rate": 7.241214096239728e-08, "loss": 0.012, "step": 18485 }, { "epoch": 7.29, "learning_rate": 7.239247930244796e-08, "loss": 0.0118, "step": 18486 }, { "epoch": 7.29, "learning_rate": 7.23728197223073e-08, "loss": 0.0111, "step": 18487 }, { "epoch": 7.29, "learning_rate": 7.23531622222957e-08, "loss": 0.0113, "step": 18488 }, { "epoch": 7.29, "learning_rate": 7.233350680273364e-08, "loss": 0.0083, "step": 18489 }, { "epoch": 7.29, "learning_rate": 7.231385346394132e-08, "loss": 0.0144, "step": 18490 }, { "epoch": 7.29, "learning_rate": 7.229420220623919e-08, "loss": 0.0122, "step": 18491 }, { "epoch": 7.29, "learning_rate": 7.227455302994731e-08, "loss": 0.0114, "step": 18492 }, { "epoch": 7.29, "learning_rate": 7.225490593538627e-08, "loss": 0.0136, "step": 18493 }, { "epoch": 7.29, "learning_rate": 7.223526092287604e-08, "loss": 0.0102, "step": 18494 }, { "epoch": 7.29, "learning_rate": 7.221561799273692e-08, "loss": 0.013, "step": 18495 }, { "epoch": 7.29, "learning_rate": 7.219597714528892e-08, "loss": 0.0091, "step": 18496 }, { "epoch": 7.29, "learning_rate": 7.217633838085236e-08, "loss": 0.0121, "step": 18497 }, { "epoch": 7.29, "learning_rate": 7.215670169974714e-08, "loss": 0.0093, "step": 18498 }, { "epoch": 7.29, "learning_rate": 7.213706710229348e-08, "loss": 0.0103, "step": 18499 }, { "epoch": 7.29, "learning_rate": 7.21174345888112e-08, "loss": 0.0108, "step": 18500 }, { "epoch": 7.29, "learning_rate": 7.209780415962039e-08, "loss": 0.0125, "step": 18501 }, { "epoch": 7.29, "learning_rate": 7.207817581504098e-08, "loss": 0.0081, "step": 18502 }, { "epoch": 7.29, "learning_rate": 7.205854955539297e-08, "loss": 0.011, "step": 18503 }, { "epoch": 7.29, "learning_rate": 7.203892538099607e-08, "loss": 0.0104, "step": 18504 }, { "epoch": 7.29, "learning_rate": 7.20193032921702e-08, "loss": 0.0099, "step": 18505 }, { "epoch": 7.29, "learning_rate": 7.199968328923527e-08, "loss": 0.0106, "step": 18506 }, { "epoch": 7.29, "learning_rate": 7.198006537251087e-08, "loss": 0.0115, "step": 18507 }, { "epoch": 7.29, "learning_rate": 7.196044954231693e-08, "loss": 0.0081, "step": 18508 }, { "epoch": 7.29, "learning_rate": 7.194083579897299e-08, "loss": 0.0086, "step": 18509 }, { "epoch": 7.29, "learning_rate": 7.192122414279877e-08, "loss": 0.0094, "step": 18510 }, { "epoch": 7.3, "learning_rate": 7.190161457411396e-08, "loss": 0.0087, "step": 18511 }, { "epoch": 7.3, "learning_rate": 7.188200709323822e-08, "loss": 0.0112, "step": 18512 }, { "epoch": 7.3, "learning_rate": 7.186240170049098e-08, "loss": 0.0102, "step": 18513 }, { "epoch": 7.3, "learning_rate": 7.184279839619185e-08, "loss": 0.0116, "step": 18514 }, { "epoch": 7.3, "learning_rate": 7.182319718066032e-08, "loss": 0.0099, "step": 18515 }, { "epoch": 7.3, "learning_rate": 7.180359805421594e-08, "loss": 0.0104, "step": 18516 }, { "epoch": 7.3, "learning_rate": 7.178400101717798e-08, "loss": 0.0122, "step": 18517 }, { "epoch": 7.3, "learning_rate": 7.176440606986597e-08, "loss": 0.0101, "step": 18518 }, { "epoch": 7.3, "learning_rate": 7.174481321259924e-08, "loss": 0.0106, "step": 18519 }, { "epoch": 7.3, "learning_rate": 7.172522244569721e-08, "loss": 0.0105, "step": 18520 }, { "epoch": 7.3, "learning_rate": 7.170563376947901e-08, "loss": 0.0103, "step": 18521 }, { "epoch": 7.3, "learning_rate": 7.168604718426407e-08, "loss": 0.0129, "step": 18522 }, { "epoch": 7.3, "learning_rate": 7.166646269037149e-08, "loss": 0.0113, "step": 18523 }, { "epoch": 7.3, "learning_rate": 7.164688028812049e-08, "loss": 0.0118, "step": 18524 }, { "epoch": 7.3, "learning_rate": 7.16272999778303e-08, "loss": 0.0131, "step": 18525 }, { "epoch": 7.3, "learning_rate": 7.160772175982006e-08, "loss": 0.0108, "step": 18526 }, { "epoch": 7.3, "learning_rate": 7.158814563440878e-08, "loss": 0.0099, "step": 18527 }, { "epoch": 7.3, "learning_rate": 7.156857160191554e-08, "loss": 0.0116, "step": 18528 }, { "epoch": 7.3, "learning_rate": 7.154899966265946e-08, "loss": 0.0097, "step": 18529 }, { "epoch": 7.3, "learning_rate": 7.152942981695943e-08, "loss": 0.0111, "step": 18530 }, { "epoch": 7.3, "learning_rate": 7.150986206513445e-08, "loss": 0.0107, "step": 18531 }, { "epoch": 7.3, "learning_rate": 7.149029640750334e-08, "loss": 0.011, "step": 18532 }, { "epoch": 7.3, "learning_rate": 7.14707328443852e-08, "loss": 0.0093, "step": 18533 }, { "epoch": 7.3, "learning_rate": 7.145117137609873e-08, "loss": 0.0126, "step": 18534 }, { "epoch": 7.3, "learning_rate": 7.143161200296284e-08, "loss": 0.0113, "step": 18535 }, { "epoch": 7.31, "learning_rate": 7.141205472529614e-08, "loss": 0.0142, "step": 18536 }, { "epoch": 7.31, "learning_rate": 7.139249954341769e-08, "loss": 0.0127, "step": 18537 }, { "epoch": 7.31, "learning_rate": 7.137294645764591e-08, "loss": 0.0131, "step": 18538 }, { "epoch": 7.31, "learning_rate": 7.135339546829972e-08, "loss": 0.0115, "step": 18539 }, { "epoch": 7.31, "learning_rate": 7.133384657569758e-08, "loss": 0.0118, "step": 18540 }, { "epoch": 7.31, "learning_rate": 7.13142997801582e-08, "loss": 0.0129, "step": 18541 }, { "epoch": 7.31, "learning_rate": 7.129475508200014e-08, "loss": 0.0083, "step": 18542 }, { "epoch": 7.31, "learning_rate": 7.127521248154205e-08, "loss": 0.0109, "step": 18543 }, { "epoch": 7.31, "learning_rate": 7.125567197910227e-08, "loss": 0.0088, "step": 18544 }, { "epoch": 7.31, "learning_rate": 7.123613357499938e-08, "loss": 0.01, "step": 18545 }, { "epoch": 7.31, "learning_rate": 7.121659726955187e-08, "loss": 0.0129, "step": 18546 }, { "epoch": 7.31, "learning_rate": 7.119706306307804e-08, "loss": 0.0123, "step": 18547 }, { "epoch": 7.31, "learning_rate": 7.117753095589631e-08, "loss": 0.0108, "step": 18548 }, { "epoch": 7.31, "learning_rate": 7.115800094832514e-08, "loss": 0.0101, "step": 18549 }, { "epoch": 7.31, "learning_rate": 7.113847304068263e-08, "loss": 0.0112, "step": 18550 }, { "epoch": 7.31, "learning_rate": 7.111894723328719e-08, "loss": 0.0108, "step": 18551 }, { "epoch": 7.31, "learning_rate": 7.109942352645708e-08, "loss": 0.0127, "step": 18552 }, { "epoch": 7.31, "learning_rate": 7.10799019205104e-08, "loss": 0.0106, "step": 18553 }, { "epoch": 7.31, "learning_rate": 7.106038241576539e-08, "loss": 0.0104, "step": 18554 }, { "epoch": 7.31, "learning_rate": 7.104086501254018e-08, "loss": 0.0108, "step": 18555 }, { "epoch": 7.31, "learning_rate": 7.102134971115295e-08, "loss": 0.0125, "step": 18556 }, { "epoch": 7.31, "learning_rate": 7.10018365119216e-08, "loss": 0.0098, "step": 18557 }, { "epoch": 7.31, "learning_rate": 7.098232541516429e-08, "loss": 0.0133, "step": 18558 }, { "epoch": 7.31, "learning_rate": 7.096281642119897e-08, "loss": 0.0113, "step": 18559 }, { "epoch": 7.31, "learning_rate": 7.094330953034371e-08, "loss": 0.0102, "step": 18560 }, { "epoch": 7.32, "learning_rate": 7.092380474291628e-08, "loss": 0.0132, "step": 18561 }, { "epoch": 7.32, "learning_rate": 7.090430205923472e-08, "loss": 0.0092, "step": 18562 }, { "epoch": 7.32, "learning_rate": 7.088480147961673e-08, "loss": 0.0093, "step": 18563 }, { "epoch": 7.32, "learning_rate": 7.086530300438036e-08, "loss": 0.0117, "step": 18564 }, { "epoch": 7.32, "learning_rate": 7.084580663384321e-08, "loss": 0.0096, "step": 18565 }, { "epoch": 7.32, "learning_rate": 7.082631236832322e-08, "loss": 0.0132, "step": 18566 }, { "epoch": 7.32, "learning_rate": 7.08068202081379e-08, "loss": 0.0098, "step": 18567 }, { "epoch": 7.32, "learning_rate": 7.07873301536051e-08, "loss": 0.0149, "step": 18568 }, { "epoch": 7.32, "learning_rate": 7.076784220504251e-08, "loss": 0.0086, "step": 18569 }, { "epoch": 7.32, "learning_rate": 7.074835636276759e-08, "loss": 0.0113, "step": 18570 }, { "epoch": 7.32, "learning_rate": 7.072887262709803e-08, "loss": 0.0101, "step": 18571 }, { "epoch": 7.32, "learning_rate": 7.070939099835139e-08, "loss": 0.0098, "step": 18572 }, { "epoch": 7.32, "learning_rate": 7.068991147684522e-08, "loss": 0.0119, "step": 18573 }, { "epoch": 7.32, "learning_rate": 7.06704340628969e-08, "loss": 0.0122, "step": 18574 }, { "epoch": 7.32, "learning_rate": 7.065095875682403e-08, "loss": 0.0101, "step": 18575 }, { "epoch": 7.32, "learning_rate": 7.06314855589438e-08, "loss": 0.012, "step": 18576 }, { "epoch": 7.32, "learning_rate": 7.061201446957387e-08, "loss": 0.0117, "step": 18577 }, { "epoch": 7.32, "learning_rate": 7.05925454890314e-08, "loss": 0.0081, "step": 18578 }, { "epoch": 7.32, "learning_rate": 7.057307861763382e-08, "loss": 0.0127, "step": 18579 }, { "epoch": 7.32, "learning_rate": 7.055361385569829e-08, "loss": 0.0108, "step": 18580 }, { "epoch": 7.32, "learning_rate": 7.053415120354212e-08, "loss": 0.0119, "step": 18581 }, { "epoch": 7.32, "learning_rate": 7.051469066148253e-08, "loss": 0.0099, "step": 18582 }, { "epoch": 7.32, "learning_rate": 7.049523222983673e-08, "loss": 0.0096, "step": 18583 }, { "epoch": 7.32, "learning_rate": 7.047577590892177e-08, "loss": 0.0111, "step": 18584 }, { "epoch": 7.32, "learning_rate": 7.04563216990548e-08, "loss": 0.0088, "step": 18585 }, { "epoch": 7.32, "learning_rate": 7.043686960055291e-08, "loss": 0.0101, "step": 18586 }, { "epoch": 7.33, "learning_rate": 7.041741961373319e-08, "loss": 0.0127, "step": 18587 }, { "epoch": 7.33, "learning_rate": 7.039797173891253e-08, "loss": 0.0125, "step": 18588 }, { "epoch": 7.33, "learning_rate": 7.037852597640801e-08, "loss": 0.0094, "step": 18589 }, { "epoch": 7.33, "learning_rate": 7.035908232653646e-08, "loss": 0.0104, "step": 18590 }, { "epoch": 7.33, "learning_rate": 7.033964078961483e-08, "loss": 0.0131, "step": 18591 }, { "epoch": 7.33, "learning_rate": 7.032020136596007e-08, "loss": 0.0097, "step": 18592 }, { "epoch": 7.33, "learning_rate": 7.030076405588887e-08, "loss": 0.0113, "step": 18593 }, { "epoch": 7.33, "learning_rate": 7.028132885971809e-08, "loss": 0.0107, "step": 18594 }, { "epoch": 7.33, "learning_rate": 7.026189577776451e-08, "loss": 0.0114, "step": 18595 }, { "epoch": 7.33, "learning_rate": 7.024246481034491e-08, "loss": 0.0113, "step": 18596 }, { "epoch": 7.33, "learning_rate": 7.022303595777586e-08, "loss": 0.012, "step": 18597 }, { "epoch": 7.33, "learning_rate": 7.020360922037407e-08, "loss": 0.0124, "step": 18598 }, { "epoch": 7.33, "learning_rate": 7.018418459845618e-08, "loss": 0.0119, "step": 18599 }, { "epoch": 7.33, "learning_rate": 7.016476209233888e-08, "loss": 0.0145, "step": 18600 }, { "epoch": 7.33, "learning_rate": 7.014534170233855e-08, "loss": 0.0097, "step": 18601 }, { "epoch": 7.33, "learning_rate": 7.012592342877186e-08, "loss": 0.012, "step": 18602 }, { "epoch": 7.33, "learning_rate": 7.010650727195509e-08, "loss": 0.011, "step": 18603 }, { "epoch": 7.33, "learning_rate": 7.008709323220498e-08, "loss": 0.0132, "step": 18604 }, { "epoch": 7.33, "learning_rate": 7.006768130983774e-08, "loss": 0.0103, "step": 18605 }, { "epoch": 7.33, "learning_rate": 7.004827150516988e-08, "loss": 0.014, "step": 18606 }, { "epoch": 7.33, "learning_rate": 7.002886381851756e-08, "loss": 0.0101, "step": 18607 }, { "epoch": 7.33, "learning_rate": 7.000945825019737e-08, "loss": 0.0092, "step": 18608 }, { "epoch": 7.33, "learning_rate": 6.999005480052535e-08, "loss": 0.0089, "step": 18609 }, { "epoch": 7.33, "learning_rate": 6.997065346981794e-08, "loss": 0.0107, "step": 18610 }, { "epoch": 7.33, "learning_rate": 6.99512542583912e-08, "loss": 0.0098, "step": 18611 }, { "epoch": 7.34, "learning_rate": 6.993185716656133e-08, "loss": 0.0118, "step": 18612 }, { "epoch": 7.34, "learning_rate": 6.991246219464459e-08, "loss": 0.0103, "step": 18613 }, { "epoch": 7.34, "learning_rate": 6.989306934295696e-08, "loss": 0.0106, "step": 18614 }, { "epoch": 7.34, "learning_rate": 6.987367861181459e-08, "loss": 0.0104, "step": 18615 }, { "epoch": 7.34, "learning_rate": 6.985429000153337e-08, "loss": 0.0095, "step": 18616 }, { "epoch": 7.34, "learning_rate": 6.983490351242958e-08, "loss": 0.0097, "step": 18617 }, { "epoch": 7.34, "learning_rate": 6.981551914481894e-08, "loss": 0.0096, "step": 18618 }, { "epoch": 7.34, "learning_rate": 6.979613689901756e-08, "loss": 0.0113, "step": 18619 }, { "epoch": 7.34, "learning_rate": 6.977675677534118e-08, "loss": 0.011, "step": 18620 }, { "epoch": 7.34, "learning_rate": 6.975737877410577e-08, "loss": 0.0112, "step": 18621 }, { "epoch": 7.34, "learning_rate": 6.973800289562713e-08, "loss": 0.0099, "step": 18622 }, { "epoch": 7.34, "learning_rate": 6.971862914022112e-08, "loss": 0.0112, "step": 18623 }, { "epoch": 7.34, "learning_rate": 6.96992575082034e-08, "loss": 0.0111, "step": 18624 }, { "epoch": 7.34, "learning_rate": 6.967988799988973e-08, "loss": 0.0117, "step": 18625 }, { "epoch": 7.34, "learning_rate": 6.966052061559584e-08, "loss": 0.0126, "step": 18626 }, { "epoch": 7.34, "learning_rate": 6.964115535563744e-08, "loss": 0.0131, "step": 18627 }, { "epoch": 7.34, "learning_rate": 6.962179222033e-08, "loss": 0.0104, "step": 18628 }, { "epoch": 7.34, "learning_rate": 6.960243120998928e-08, "loss": 0.0116, "step": 18629 }, { "epoch": 7.34, "learning_rate": 6.958307232493062e-08, "loss": 0.0107, "step": 18630 }, { "epoch": 7.34, "learning_rate": 6.956371556546978e-08, "loss": 0.0078, "step": 18631 }, { "epoch": 7.34, "learning_rate": 6.954436093192209e-08, "loss": 0.0095, "step": 18632 }, { "epoch": 7.34, "learning_rate": 6.952500842460311e-08, "loss": 0.0116, "step": 18633 }, { "epoch": 7.34, "learning_rate": 6.950565804382811e-08, "loss": 0.0094, "step": 18634 }, { "epoch": 7.34, "learning_rate": 6.948630978991257e-08, "loss": 0.0107, "step": 18635 }, { "epoch": 7.34, "learning_rate": 6.946696366317188e-08, "loss": 0.0131, "step": 18636 }, { "epoch": 7.34, "learning_rate": 6.944761966392125e-08, "loss": 0.009, "step": 18637 }, { "epoch": 7.35, "learning_rate": 6.9428277792476e-08, "loss": 0.0129, "step": 18638 }, { "epoch": 7.35, "learning_rate": 6.940893804915136e-08, "loss": 0.0093, "step": 18639 }, { "epoch": 7.35, "learning_rate": 6.938960043426263e-08, "loss": 0.0124, "step": 18640 }, { "epoch": 7.35, "learning_rate": 6.937026494812483e-08, "loss": 0.0117, "step": 18641 }, { "epoch": 7.35, "learning_rate": 6.935093159105325e-08, "loss": 0.0105, "step": 18642 }, { "epoch": 7.35, "learning_rate": 6.933160036336278e-08, "loss": 0.0116, "step": 18643 }, { "epoch": 7.35, "learning_rate": 6.931227126536879e-08, "loss": 0.0121, "step": 18644 }, { "epoch": 7.35, "learning_rate": 6.929294429738606e-08, "loss": 0.0107, "step": 18645 }, { "epoch": 7.35, "learning_rate": 6.927361945972976e-08, "loss": 0.0121, "step": 18646 }, { "epoch": 7.35, "learning_rate": 6.925429675271466e-08, "loss": 0.013, "step": 18647 }, { "epoch": 7.35, "learning_rate": 6.923497617665595e-08, "loss": 0.0104, "step": 18648 }, { "epoch": 7.35, "learning_rate": 6.92156577318683e-08, "loss": 0.0098, "step": 18649 }, { "epoch": 7.35, "learning_rate": 6.919634141866672e-08, "loss": 0.0131, "step": 18650 }, { "epoch": 7.35, "learning_rate": 6.917702723736594e-08, "loss": 0.0098, "step": 18651 }, { "epoch": 7.35, "learning_rate": 6.915771518828077e-08, "loss": 0.0106, "step": 18652 }, { "epoch": 7.35, "learning_rate": 6.913840527172598e-08, "loss": 0.0114, "step": 18653 }, { "epoch": 7.35, "learning_rate": 6.911909748801638e-08, "loss": 0.0096, "step": 18654 }, { "epoch": 7.35, "learning_rate": 6.90997918374665e-08, "loss": 0.0114, "step": 18655 }, { "epoch": 7.35, "learning_rate": 6.908048832039109e-08, "loss": 0.0108, "step": 18656 }, { "epoch": 7.35, "learning_rate": 6.906118693710479e-08, "loss": 0.0132, "step": 18657 }, { "epoch": 7.35, "learning_rate": 6.904188768792207e-08, "loss": 0.0121, "step": 18658 }, { "epoch": 7.35, "learning_rate": 6.90225905731576e-08, "loss": 0.0132, "step": 18659 }, { "epoch": 7.35, "learning_rate": 6.90032955931258e-08, "loss": 0.0101, "step": 18660 }, { "epoch": 7.35, "learning_rate": 6.898400274814117e-08, "loss": 0.0093, "step": 18661 }, { "epoch": 7.35, "learning_rate": 6.89647120385182e-08, "loss": 0.0141, "step": 18662 }, { "epoch": 7.36, "learning_rate": 6.894542346457133e-08, "loss": 0.0103, "step": 18663 }, { "epoch": 7.36, "learning_rate": 6.89261370266148e-08, "loss": 0.0127, "step": 18664 }, { "epoch": 7.36, "learning_rate": 6.890685272496303e-08, "loss": 0.0117, "step": 18665 }, { "epoch": 7.36, "learning_rate": 6.888757055993033e-08, "loss": 0.0115, "step": 18666 }, { "epoch": 7.36, "learning_rate": 6.886829053183103e-08, "loss": 0.0107, "step": 18667 }, { "epoch": 7.36, "learning_rate": 6.884901264097923e-08, "loss": 0.0116, "step": 18668 }, { "epoch": 7.36, "learning_rate": 6.882973688768926e-08, "loss": 0.0107, "step": 18669 }, { "epoch": 7.36, "learning_rate": 6.881046327227507e-08, "loss": 0.0106, "step": 18670 }, { "epoch": 7.36, "learning_rate": 6.87911917950511e-08, "loss": 0.0094, "step": 18671 }, { "epoch": 7.36, "learning_rate": 6.877192245633122e-08, "loss": 0.0125, "step": 18672 }, { "epoch": 7.36, "learning_rate": 6.875265525642962e-08, "loss": 0.0092, "step": 18673 }, { "epoch": 7.36, "learning_rate": 6.87333901956602e-08, "loss": 0.0122, "step": 18674 }, { "epoch": 7.36, "learning_rate": 6.871412727433703e-08, "loss": 0.0116, "step": 18675 }, { "epoch": 7.36, "learning_rate": 6.869486649277405e-08, "loss": 0.0134, "step": 18676 }, { "epoch": 7.36, "learning_rate": 6.867560785128523e-08, "loss": 0.014, "step": 18677 }, { "epoch": 7.36, "learning_rate": 6.865635135018438e-08, "loss": 0.0148, "step": 18678 }, { "epoch": 7.36, "learning_rate": 6.863709698978537e-08, "loss": 0.0121, "step": 18679 }, { "epoch": 7.36, "learning_rate": 6.861784477040209e-08, "loss": 0.0122, "step": 18680 }, { "epoch": 7.36, "learning_rate": 6.859859469234819e-08, "loss": 0.0098, "step": 18681 }, { "epoch": 7.36, "learning_rate": 6.857934675593758e-08, "loss": 0.0109, "step": 18682 }, { "epoch": 7.36, "learning_rate": 6.856010096148373e-08, "loss": 0.0109, "step": 18683 }, { "epoch": 7.36, "learning_rate": 6.85408573093006e-08, "loss": 0.0113, "step": 18684 }, { "epoch": 7.36, "learning_rate": 6.852161579970164e-08, "loss": 0.008, "step": 18685 }, { "epoch": 7.36, "learning_rate": 6.850237643300058e-08, "loss": 0.0097, "step": 18686 }, { "epoch": 7.36, "learning_rate": 6.848313920951081e-08, "loss": 0.009, "step": 18687 }, { "epoch": 7.37, "learning_rate": 6.846390412954612e-08, "loss": 0.0135, "step": 18688 }, { "epoch": 7.37, "learning_rate": 6.844467119341981e-08, "loss": 0.0116, "step": 18689 }, { "epoch": 7.37, "learning_rate": 6.842544040144547e-08, "loss": 0.0103, "step": 18690 }, { "epoch": 7.37, "learning_rate": 6.840621175393644e-08, "loss": 0.0113, "step": 18691 }, { "epoch": 7.37, "learning_rate": 6.838698525120614e-08, "loss": 0.0085, "step": 18692 }, { "epoch": 7.37, "learning_rate": 6.836776089356795e-08, "loss": 0.0112, "step": 18693 }, { "epoch": 7.37, "learning_rate": 6.834853868133528e-08, "loss": 0.0109, "step": 18694 }, { "epoch": 7.37, "learning_rate": 6.832931861482128e-08, "loss": 0.0092, "step": 18695 }, { "epoch": 7.37, "learning_rate": 6.831010069433924e-08, "loss": 0.0123, "step": 18696 }, { "epoch": 7.37, "learning_rate": 6.829088492020251e-08, "loss": 0.0101, "step": 18697 }, { "epoch": 7.37, "learning_rate": 6.82716712927241e-08, "loss": 0.011, "step": 18698 }, { "epoch": 7.37, "learning_rate": 6.825245981221725e-08, "loss": 0.0087, "step": 18699 }, { "epoch": 7.37, "learning_rate": 6.823325047899515e-08, "loss": 0.0105, "step": 18700 }, { "epoch": 7.37, "learning_rate": 6.821404329337072e-08, "loss": 0.0121, "step": 18701 }, { "epoch": 7.37, "learning_rate": 6.81948382556571e-08, "loss": 0.0097, "step": 18702 }, { "epoch": 7.37, "learning_rate": 6.817563536616735e-08, "loss": 0.0114, "step": 18703 }, { "epoch": 7.37, "learning_rate": 6.815643462521435e-08, "loss": 0.0111, "step": 18704 }, { "epoch": 7.37, "learning_rate": 6.813723603311108e-08, "loss": 0.0098, "step": 18705 }, { "epoch": 7.37, "learning_rate": 6.811803959017044e-08, "loss": 0.0113, "step": 18706 }, { "epoch": 7.37, "learning_rate": 6.80988452967054e-08, "loss": 0.0105, "step": 18707 }, { "epoch": 7.37, "learning_rate": 6.807965315302864e-08, "loss": 0.0116, "step": 18708 }, { "epoch": 7.37, "learning_rate": 6.806046315945308e-08, "loss": 0.0099, "step": 18709 }, { "epoch": 7.37, "learning_rate": 6.804127531629135e-08, "loss": 0.0093, "step": 18710 }, { "epoch": 7.37, "learning_rate": 6.80220896238564e-08, "loss": 0.0124, "step": 18711 }, { "epoch": 7.37, "learning_rate": 6.800290608246073e-08, "loss": 0.0145, "step": 18712 }, { "epoch": 7.37, "learning_rate": 6.798372469241715e-08, "loss": 0.0096, "step": 18713 }, { "epoch": 7.38, "learning_rate": 6.796454545403809e-08, "loss": 0.0126, "step": 18714 }, { "epoch": 7.38, "learning_rate": 6.79453683676364e-08, "loss": 0.0104, "step": 18715 }, { "epoch": 7.38, "learning_rate": 6.792619343352444e-08, "loss": 0.0123, "step": 18716 }, { "epoch": 7.38, "learning_rate": 6.790702065201485e-08, "loss": 0.01, "step": 18717 }, { "epoch": 7.38, "learning_rate": 6.788785002342001e-08, "loss": 0.0111, "step": 18718 }, { "epoch": 7.38, "learning_rate": 6.786868154805244e-08, "loss": 0.012, "step": 18719 }, { "epoch": 7.38, "learning_rate": 6.78495152262246e-08, "loss": 0.0102, "step": 18720 }, { "epoch": 7.38, "learning_rate": 6.783035105824874e-08, "loss": 0.0097, "step": 18721 }, { "epoch": 7.38, "learning_rate": 6.78111890444373e-08, "loss": 0.009, "step": 18722 }, { "epoch": 7.38, "learning_rate": 6.779202918510256e-08, "loss": 0.0103, "step": 18723 }, { "epoch": 7.38, "learning_rate": 6.777287148055688e-08, "loss": 0.0108, "step": 18724 }, { "epoch": 7.38, "learning_rate": 6.775371593111236e-08, "loss": 0.0121, "step": 18725 }, { "epoch": 7.38, "learning_rate": 6.773456253708136e-08, "loss": 0.0119, "step": 18726 }, { "epoch": 7.38, "learning_rate": 6.771541129877583e-08, "loss": 0.011, "step": 18727 }, { "epoch": 7.38, "learning_rate": 6.76962622165082e-08, "loss": 0.0128, "step": 18728 }, { "epoch": 7.38, "learning_rate": 6.767711529059032e-08, "loss": 0.0093, "step": 18729 }, { "epoch": 7.38, "learning_rate": 6.765797052133441e-08, "loss": 0.009, "step": 18730 }, { "epoch": 7.38, "learning_rate": 6.763882790905242e-08, "loss": 0.0109, "step": 18731 }, { "epoch": 7.38, "learning_rate": 6.761968745405635e-08, "loss": 0.0115, "step": 18732 }, { "epoch": 7.38, "learning_rate": 6.760054915665818e-08, "loss": 0.0112, "step": 18733 }, { "epoch": 7.38, "learning_rate": 6.758141301716989e-08, "loss": 0.0133, "step": 18734 }, { "epoch": 7.38, "learning_rate": 6.756227903590326e-08, "loss": 0.0108, "step": 18735 }, { "epoch": 7.38, "learning_rate": 6.75431472131702e-08, "loss": 0.0112, "step": 18736 }, { "epoch": 7.38, "learning_rate": 6.752401754928252e-08, "loss": 0.0098, "step": 18737 }, { "epoch": 7.38, "learning_rate": 6.750489004455209e-08, "loss": 0.0091, "step": 18738 }, { "epoch": 7.39, "learning_rate": 6.74857646992905e-08, "loss": 0.0109, "step": 18739 }, { "epoch": 7.39, "learning_rate": 6.746664151380963e-08, "loss": 0.0105, "step": 18740 }, { "epoch": 7.39, "learning_rate": 6.744752048842101e-08, "loss": 0.0109, "step": 18741 }, { "epoch": 7.39, "learning_rate": 6.742840162343635e-08, "loss": 0.0133, "step": 18742 }, { "epoch": 7.39, "learning_rate": 6.740928491916733e-08, "loss": 0.0104, "step": 18743 }, { "epoch": 7.39, "learning_rate": 6.739017037592536e-08, "loss": 0.0097, "step": 18744 }, { "epoch": 7.39, "learning_rate": 6.737105799402209e-08, "loss": 0.0119, "step": 18745 }, { "epoch": 7.39, "learning_rate": 6.7351947773769e-08, "loss": 0.0119, "step": 18746 }, { "epoch": 7.39, "learning_rate": 6.733283971547761e-08, "loss": 0.0093, "step": 18747 }, { "epoch": 7.39, "learning_rate": 6.731373381945926e-08, "loss": 0.0128, "step": 18748 }, { "epoch": 7.39, "learning_rate": 6.729463008602542e-08, "loss": 0.0088, "step": 18749 }, { "epoch": 7.39, "learning_rate": 6.727552851548732e-08, "loss": 0.0132, "step": 18750 }, { "epoch": 7.39, "learning_rate": 6.725642910815652e-08, "loss": 0.0087, "step": 18751 }, { "epoch": 7.39, "learning_rate": 6.723733186434411e-08, "loss": 0.01, "step": 18752 }, { "epoch": 7.39, "learning_rate": 6.721823678436145e-08, "loss": 0.0111, "step": 18753 }, { "epoch": 7.39, "learning_rate": 6.719914386851961e-08, "loss": 0.0081, "step": 18754 }, { "epoch": 7.39, "learning_rate": 6.718005311713004e-08, "loss": 0.0113, "step": 18755 }, { "epoch": 7.39, "learning_rate": 6.716096453050366e-08, "loss": 0.01, "step": 18756 }, { "epoch": 7.39, "learning_rate": 6.714187810895173e-08, "loss": 0.0099, "step": 18757 }, { "epoch": 7.39, "learning_rate": 6.712279385278514e-08, "loss": 0.0103, "step": 18758 }, { "epoch": 7.39, "learning_rate": 6.71037117623152e-08, "loss": 0.0089, "step": 18759 }, { "epoch": 7.39, "learning_rate": 6.708463183785272e-08, "loss": 0.011, "step": 18760 }, { "epoch": 7.39, "learning_rate": 6.706555407970877e-08, "loss": 0.0128, "step": 18761 }, { "epoch": 7.39, "learning_rate": 6.70464784881942e-08, "loss": 0.0114, "step": 18762 }, { "epoch": 7.39, "learning_rate": 6.702740506361997e-08, "loss": 0.0103, "step": 18763 }, { "epoch": 7.4, "learning_rate": 6.7008333806297e-08, "loss": 0.0117, "step": 18764 }, { "epoch": 7.4, "learning_rate": 6.6989264716536e-08, "loss": 0.0102, "step": 18765 }, { "epoch": 7.4, "learning_rate": 6.697019779464783e-08, "loss": 0.0119, "step": 18766 }, { "epoch": 7.4, "learning_rate": 6.695113304094326e-08, "loss": 0.0113, "step": 18767 }, { "epoch": 7.4, "learning_rate": 6.693207045573308e-08, "loss": 0.0107, "step": 18768 }, { "epoch": 7.4, "learning_rate": 6.691301003932783e-08, "loss": 0.0101, "step": 18769 }, { "epoch": 7.4, "learning_rate": 6.689395179203834e-08, "loss": 0.0109, "step": 18770 }, { "epoch": 7.4, "learning_rate": 6.687489571417503e-08, "loss": 0.0113, "step": 18771 }, { "epoch": 7.4, "learning_rate": 6.685584180604864e-08, "loss": 0.0151, "step": 18772 }, { "epoch": 7.4, "learning_rate": 6.683679006796965e-08, "loss": 0.0095, "step": 18773 }, { "epoch": 7.4, "learning_rate": 6.681774050024867e-08, "loss": 0.0116, "step": 18774 }, { "epoch": 7.4, "learning_rate": 6.679869310319601e-08, "loss": 0.0116, "step": 18775 }, { "epoch": 7.4, "learning_rate": 6.677964787712224e-08, "loss": 0.0159, "step": 18776 }, { "epoch": 7.4, "learning_rate": 6.676060482233775e-08, "loss": 0.0123, "step": 18777 }, { "epoch": 7.4, "learning_rate": 6.674156393915293e-08, "loss": 0.0123, "step": 18778 }, { "epoch": 7.4, "learning_rate": 6.672252522787805e-08, "loss": 0.0106, "step": 18779 }, { "epoch": 7.4, "learning_rate": 6.670348868882349e-08, "loss": 0.013, "step": 18780 }, { "epoch": 7.4, "learning_rate": 6.668445432229937e-08, "loss": 0.0138, "step": 18781 }, { "epoch": 7.4, "learning_rate": 6.666542212861616e-08, "loss": 0.0073, "step": 18782 }, { "epoch": 7.4, "learning_rate": 6.664639210808385e-08, "loss": 0.0106, "step": 18783 }, { "epoch": 7.4, "learning_rate": 6.662736426101274e-08, "loss": 0.0101, "step": 18784 }, { "epoch": 7.4, "learning_rate": 6.660833858771282e-08, "loss": 0.0105, "step": 18785 }, { "epoch": 7.4, "learning_rate": 6.658931508849426e-08, "loss": 0.0106, "step": 18786 }, { "epoch": 7.4, "learning_rate": 6.657029376366717e-08, "loss": 0.0107, "step": 18787 }, { "epoch": 7.4, "learning_rate": 6.655127461354142e-08, "loss": 0.0122, "step": 18788 }, { "epoch": 7.4, "learning_rate": 6.65322576384271e-08, "loss": 0.0098, "step": 18789 }, { "epoch": 7.41, "learning_rate": 6.651324283863411e-08, "loss": 0.0109, "step": 18790 }, { "epoch": 7.41, "learning_rate": 6.649423021447245e-08, "loss": 0.013, "step": 18791 }, { "epoch": 7.41, "learning_rate": 6.647521976625188e-08, "loss": 0.0132, "step": 18792 }, { "epoch": 7.41, "learning_rate": 6.645621149428235e-08, "loss": 0.0086, "step": 18793 }, { "epoch": 7.41, "learning_rate": 6.643720539887347e-08, "loss": 0.0119, "step": 18794 }, { "epoch": 7.41, "learning_rate": 6.64182014803353e-08, "loss": 0.0107, "step": 18795 }, { "epoch": 7.41, "learning_rate": 6.639919973897736e-08, "loss": 0.0102, "step": 18796 }, { "epoch": 7.41, "learning_rate": 6.638020017510946e-08, "loss": 0.0099, "step": 18797 }, { "epoch": 7.41, "learning_rate": 6.636120278904111e-08, "loss": 0.0125, "step": 18798 }, { "epoch": 7.41, "learning_rate": 6.634220758108216e-08, "loss": 0.0108, "step": 18799 }, { "epoch": 7.41, "learning_rate": 6.632321455154204e-08, "loss": 0.0102, "step": 18800 }, { "epoch": 7.41, "learning_rate": 6.630422370073041e-08, "loss": 0.0114, "step": 18801 }, { "epoch": 7.41, "learning_rate": 6.628523502895669e-08, "loss": 0.0113, "step": 18802 }, { "epoch": 7.41, "learning_rate": 6.62662485365304e-08, "loss": 0.0146, "step": 18803 }, { "epoch": 7.41, "learning_rate": 6.624726422376101e-08, "loss": 0.0098, "step": 18804 }, { "epoch": 7.41, "learning_rate": 6.622828209095799e-08, "loss": 0.0103, "step": 18805 }, { "epoch": 7.41, "learning_rate": 6.620930213843061e-08, "loss": 0.0099, "step": 18806 }, { "epoch": 7.41, "learning_rate": 6.619032436648826e-08, "loss": 0.0111, "step": 18807 }, { "epoch": 7.41, "learning_rate": 6.617134877544033e-08, "loss": 0.0143, "step": 18808 }, { "epoch": 7.41, "learning_rate": 6.615237536559594e-08, "loss": 0.0108, "step": 18809 }, { "epoch": 7.41, "learning_rate": 6.613340413726449e-08, "loss": 0.0116, "step": 18810 }, { "epoch": 7.41, "learning_rate": 6.611443509075502e-08, "loss": 0.0094, "step": 18811 }, { "epoch": 7.41, "learning_rate": 6.60954682263768e-08, "loss": 0.0102, "step": 18812 }, { "epoch": 7.41, "learning_rate": 6.60765035444389e-08, "loss": 0.0122, "step": 18813 }, { "epoch": 7.41, "learning_rate": 6.605754104525052e-08, "loss": 0.0087, "step": 18814 }, { "epoch": 7.42, "learning_rate": 6.603858072912061e-08, "loss": 0.0073, "step": 18815 }, { "epoch": 7.42, "learning_rate": 6.601962259635823e-08, "loss": 0.0102, "step": 18816 }, { "epoch": 7.42, "learning_rate": 6.600066664727238e-08, "loss": 0.0109, "step": 18817 }, { "epoch": 7.42, "learning_rate": 6.598171288217205e-08, "loss": 0.0105, "step": 18818 }, { "epoch": 7.42, "learning_rate": 6.596276130136606e-08, "loss": 0.0102, "step": 18819 }, { "epoch": 7.42, "learning_rate": 6.594381190516343e-08, "loss": 0.0096, "step": 18820 }, { "epoch": 7.42, "learning_rate": 6.592486469387277e-08, "loss": 0.0128, "step": 18821 }, { "epoch": 7.42, "learning_rate": 6.590591966780319e-08, "loss": 0.0137, "step": 18822 }, { "epoch": 7.42, "learning_rate": 6.588697682726326e-08, "loss": 0.013, "step": 18823 }, { "epoch": 7.42, "learning_rate": 6.586803617256183e-08, "loss": 0.0091, "step": 18824 }, { "epoch": 7.42, "learning_rate": 6.584909770400748e-08, "loss": 0.0123, "step": 18825 }, { "epoch": 7.42, "learning_rate": 6.583016142190898e-08, "loss": 0.0114, "step": 18826 }, { "epoch": 7.42, "learning_rate": 6.581122732657491e-08, "loss": 0.0094, "step": 18827 }, { "epoch": 7.42, "learning_rate": 6.579229541831397e-08, "loss": 0.0119, "step": 18828 }, { "epoch": 7.42, "learning_rate": 6.577336569743457e-08, "loss": 0.0108, "step": 18829 }, { "epoch": 7.42, "learning_rate": 6.575443816424531e-08, "loss": 0.0136, "step": 18830 }, { "epoch": 7.42, "learning_rate": 6.573551281905474e-08, "loss": 0.0096, "step": 18831 }, { "epoch": 7.42, "learning_rate": 6.571658966217119e-08, "loss": 0.011, "step": 18832 }, { "epoch": 7.42, "learning_rate": 6.56976686939032e-08, "loss": 0.01, "step": 18833 }, { "epoch": 7.42, "learning_rate": 6.567874991455897e-08, "loss": 0.0114, "step": 18834 }, { "epoch": 7.42, "learning_rate": 6.565983332444711e-08, "loss": 0.0101, "step": 18835 }, { "epoch": 7.42, "learning_rate": 6.564091892387573e-08, "loss": 0.0115, "step": 18836 }, { "epoch": 7.42, "learning_rate": 6.562200671315323e-08, "loss": 0.0118, "step": 18837 }, { "epoch": 7.42, "learning_rate": 6.560309669258766e-08, "loss": 0.0115, "step": 18838 }, { "epoch": 7.42, "learning_rate": 6.558418886248747e-08, "loss": 0.0113, "step": 18839 }, { "epoch": 7.42, "learning_rate": 6.556528322316066e-08, "loss": 0.0115, "step": 18840 }, { "epoch": 7.43, "learning_rate": 6.554637977491549e-08, "loss": 0.0115, "step": 18841 }, { "epoch": 7.43, "learning_rate": 6.552747851805988e-08, "loss": 0.0091, "step": 18842 }, { "epoch": 7.43, "learning_rate": 6.550857945290205e-08, "loss": 0.0096, "step": 18843 }, { "epoch": 7.43, "learning_rate": 6.548968257974994e-08, "loss": 0.012, "step": 18844 }, { "epoch": 7.43, "learning_rate": 6.547078789891162e-08, "loss": 0.0092, "step": 18845 }, { "epoch": 7.43, "learning_rate": 6.545189541069496e-08, "loss": 0.0108, "step": 18846 }, { "epoch": 7.43, "learning_rate": 6.543300511540793e-08, "loss": 0.01, "step": 18847 }, { "epoch": 7.43, "learning_rate": 6.541411701335844e-08, "loss": 0.0109, "step": 18848 }, { "epoch": 7.43, "learning_rate": 6.539523110485421e-08, "loss": 0.0096, "step": 18849 }, { "epoch": 7.43, "learning_rate": 6.537634739020317e-08, "loss": 0.0114, "step": 18850 }, { "epoch": 7.43, "learning_rate": 6.535746586971311e-08, "loss": 0.0096, "step": 18851 }, { "epoch": 7.43, "learning_rate": 6.533858654369166e-08, "loss": 0.0093, "step": 18852 }, { "epoch": 7.43, "learning_rate": 6.53197094124466e-08, "loss": 0.0089, "step": 18853 }, { "epoch": 7.43, "learning_rate": 6.530083447628565e-08, "loss": 0.0126, "step": 18854 }, { "epoch": 7.43, "learning_rate": 6.52819617355163e-08, "loss": 0.0099, "step": 18855 }, { "epoch": 7.43, "learning_rate": 6.526309119044625e-08, "loss": 0.0131, "step": 18856 }, { "epoch": 7.43, "learning_rate": 6.524422284138302e-08, "loss": 0.0098, "step": 18857 }, { "epoch": 7.43, "learning_rate": 6.522535668863422e-08, "loss": 0.0094, "step": 18858 }, { "epoch": 7.43, "learning_rate": 6.52064927325072e-08, "loss": 0.0114, "step": 18859 }, { "epoch": 7.43, "learning_rate": 6.518763097330957e-08, "loss": 0.0106, "step": 18860 }, { "epoch": 7.43, "learning_rate": 6.516877141134851e-08, "loss": 0.0089, "step": 18861 }, { "epoch": 7.43, "learning_rate": 6.51499140469317e-08, "loss": 0.0127, "step": 18862 }, { "epoch": 7.43, "learning_rate": 6.513105888036628e-08, "loss": 0.0106, "step": 18863 }, { "epoch": 7.43, "learning_rate": 6.511220591195967e-08, "loss": 0.0089, "step": 18864 }, { "epoch": 7.43, "learning_rate": 6.509335514201899e-08, "loss": 0.0114, "step": 18865 }, { "epoch": 7.44, "learning_rate": 6.50745065708517e-08, "loss": 0.0113, "step": 18866 }, { "epoch": 7.44, "learning_rate": 6.505566019876483e-08, "loss": 0.0089, "step": 18867 }, { "epoch": 7.44, "learning_rate": 6.503681602606567e-08, "loss": 0.0102, "step": 18868 }, { "epoch": 7.44, "learning_rate": 6.501797405306125e-08, "loss": 0.0123, "step": 18869 }, { "epoch": 7.44, "learning_rate": 6.499913428005868e-08, "loss": 0.0091, "step": 18870 }, { "epoch": 7.44, "learning_rate": 6.49802967073651e-08, "loss": 0.0142, "step": 18871 }, { "epoch": 7.44, "learning_rate": 6.496146133528742e-08, "loss": 0.0109, "step": 18872 }, { "epoch": 7.44, "learning_rate": 6.49426281641327e-08, "loss": 0.0104, "step": 18873 }, { "epoch": 7.44, "learning_rate": 6.492379719420786e-08, "loss": 0.0102, "step": 18874 }, { "epoch": 7.44, "learning_rate": 6.49049684258199e-08, "loss": 0.0115, "step": 18875 }, { "epoch": 7.44, "learning_rate": 6.488614185927556e-08, "loss": 0.0098, "step": 18876 }, { "epoch": 7.44, "learning_rate": 6.486731749488185e-08, "loss": 0.0124, "step": 18877 }, { "epoch": 7.44, "learning_rate": 6.484849533294534e-08, "loss": 0.0103, "step": 18878 }, { "epoch": 7.44, "learning_rate": 6.482967537377308e-08, "loss": 0.0098, "step": 18879 }, { "epoch": 7.44, "learning_rate": 6.481085761767163e-08, "loss": 0.0086, "step": 18880 }, { "epoch": 7.44, "learning_rate": 6.479204206494777e-08, "loss": 0.0108, "step": 18881 }, { "epoch": 7.44, "learning_rate": 6.477322871590809e-08, "loss": 0.0127, "step": 18882 }, { "epoch": 7.44, "learning_rate": 6.475441757085925e-08, "loss": 0.0107, "step": 18883 }, { "epoch": 7.44, "learning_rate": 6.473560863010785e-08, "loss": 0.0127, "step": 18884 }, { "epoch": 7.44, "learning_rate": 6.471680189396053e-08, "loss": 0.0097, "step": 18885 }, { "epoch": 7.44, "learning_rate": 6.469799736272365e-08, "loss": 0.0115, "step": 18886 }, { "epoch": 7.44, "learning_rate": 6.46791950367038e-08, "loss": 0.0113, "step": 18887 }, { "epoch": 7.44, "learning_rate": 6.466039491620739e-08, "loss": 0.0096, "step": 18888 }, { "epoch": 7.44, "learning_rate": 6.464159700154092e-08, "loss": 0.01, "step": 18889 }, { "epoch": 7.44, "learning_rate": 6.462280129301063e-08, "loss": 0.0128, "step": 18890 }, { "epoch": 7.45, "learning_rate": 6.460400779092301e-08, "loss": 0.0094, "step": 18891 }, { "epoch": 7.45, "learning_rate": 6.458521649558419e-08, "loss": 0.0115, "step": 18892 }, { "epoch": 7.45, "learning_rate": 6.456642740730055e-08, "loss": 0.013, "step": 18893 }, { "epoch": 7.45, "learning_rate": 6.454764052637831e-08, "loss": 0.012, "step": 18894 }, { "epoch": 7.45, "learning_rate": 6.452885585312373e-08, "loss": 0.008, "step": 18895 }, { "epoch": 7.45, "learning_rate": 6.451007338784285e-08, "loss": 0.0134, "step": 18896 }, { "epoch": 7.45, "learning_rate": 6.449129313084185e-08, "loss": 0.0134, "step": 18897 }, { "epoch": 7.45, "learning_rate": 6.447251508242688e-08, "loss": 0.0093, "step": 18898 }, { "epoch": 7.45, "learning_rate": 6.445373924290387e-08, "loss": 0.0113, "step": 18899 }, { "epoch": 7.45, "learning_rate": 6.443496561257898e-08, "loss": 0.0102, "step": 18900 }, { "epoch": 7.45, "learning_rate": 6.441619419175799e-08, "loss": 0.0104, "step": 18901 }, { "epoch": 7.45, "learning_rate": 6.439742498074707e-08, "loss": 0.0118, "step": 18902 }, { "epoch": 7.45, "learning_rate": 6.4378657979852e-08, "loss": 0.0096, "step": 18903 }, { "epoch": 7.45, "learning_rate": 6.435989318937874e-08, "loss": 0.0085, "step": 18904 }, { "epoch": 7.45, "learning_rate": 6.434113060963293e-08, "loss": 0.01, "step": 18905 }, { "epoch": 7.45, "learning_rate": 6.432237024092065e-08, "loss": 0.0116, "step": 18906 }, { "epoch": 7.45, "learning_rate": 6.430361208354747e-08, "loss": 0.0114, "step": 18907 }, { "epoch": 7.45, "learning_rate": 6.428485613781923e-08, "loss": 0.0104, "step": 18908 }, { "epoch": 7.45, "learning_rate": 6.42661024040415e-08, "loss": 0.0122, "step": 18909 }, { "epoch": 7.45, "learning_rate": 6.424735088252002e-08, "loss": 0.0136, "step": 18910 }, { "epoch": 7.45, "learning_rate": 6.42286015735604e-08, "loss": 0.0097, "step": 18911 }, { "epoch": 7.45, "learning_rate": 6.42098544774683e-08, "loss": 0.0092, "step": 18912 }, { "epoch": 7.45, "learning_rate": 6.419110959454912e-08, "loss": 0.0111, "step": 18913 }, { "epoch": 7.45, "learning_rate": 6.417236692510845e-08, "loss": 0.0099, "step": 18914 }, { "epoch": 7.45, "learning_rate": 6.415362646945184e-08, "loss": 0.0113, "step": 18915 }, { "epoch": 7.45, "learning_rate": 6.413488822788458e-08, "loss": 0.0093, "step": 18916 }, { "epoch": 7.46, "learning_rate": 6.411615220071216e-08, "loss": 0.0125, "step": 18917 }, { "epoch": 7.46, "learning_rate": 6.409741838823992e-08, "loss": 0.0103, "step": 18918 }, { "epoch": 7.46, "learning_rate": 6.407868679077328e-08, "loss": 0.0088, "step": 18919 }, { "epoch": 7.46, "learning_rate": 6.405995740861742e-08, "loss": 0.0117, "step": 18920 }, { "epoch": 7.46, "learning_rate": 6.40412302420777e-08, "loss": 0.0129, "step": 18921 }, { "epoch": 7.46, "learning_rate": 6.402250529145925e-08, "loss": 0.0102, "step": 18922 }, { "epoch": 7.46, "learning_rate": 6.400378255706728e-08, "loss": 0.0125, "step": 18923 }, { "epoch": 7.46, "learning_rate": 6.398506203920697e-08, "loss": 0.0115, "step": 18924 }, { "epoch": 7.46, "learning_rate": 6.39663437381835e-08, "loss": 0.0111, "step": 18925 }, { "epoch": 7.46, "learning_rate": 6.394762765430182e-08, "loss": 0.0104, "step": 18926 }, { "epoch": 7.46, "learning_rate": 6.392891378786703e-08, "loss": 0.0125, "step": 18927 }, { "epoch": 7.46, "learning_rate": 6.391020213918412e-08, "loss": 0.0135, "step": 18928 }, { "epoch": 7.46, "learning_rate": 6.389149270855816e-08, "loss": 0.0114, "step": 18929 }, { "epoch": 7.46, "learning_rate": 6.387278549629396e-08, "loss": 0.0105, "step": 18930 }, { "epoch": 7.46, "learning_rate": 6.38540805026965e-08, "loss": 0.009, "step": 18931 }, { "epoch": 7.46, "learning_rate": 6.383537772807051e-08, "loss": 0.0111, "step": 18932 }, { "epoch": 7.46, "learning_rate": 6.381667717272103e-08, "loss": 0.0098, "step": 18933 }, { "epoch": 7.46, "learning_rate": 6.379797883695269e-08, "loss": 0.0119, "step": 18934 }, { "epoch": 7.46, "learning_rate": 6.377928272107032e-08, "loss": 0.0098, "step": 18935 }, { "epoch": 7.46, "learning_rate": 6.376058882537858e-08, "loss": 0.0127, "step": 18936 }, { "epoch": 7.46, "learning_rate": 6.374189715018217e-08, "loss": 0.0112, "step": 18937 }, { "epoch": 7.46, "learning_rate": 6.372320769578581e-08, "loss": 0.0111, "step": 18938 }, { "epoch": 7.46, "learning_rate": 6.370452046249397e-08, "loss": 0.0118, "step": 18939 }, { "epoch": 7.46, "learning_rate": 6.368583545061133e-08, "loss": 0.0104, "step": 18940 }, { "epoch": 7.46, "learning_rate": 6.366715266044238e-08, "loss": 0.0092, "step": 18941 }, { "epoch": 7.47, "learning_rate": 6.364847209229171e-08, "loss": 0.0118, "step": 18942 }, { "epoch": 7.47, "learning_rate": 6.362979374646364e-08, "loss": 0.0097, "step": 18943 }, { "epoch": 7.47, "learning_rate": 6.361111762326276e-08, "loss": 0.012, "step": 18944 }, { "epoch": 7.47, "learning_rate": 6.359244372299324e-08, "loss": 0.0104, "step": 18945 }, { "epoch": 7.47, "learning_rate": 6.357377204595968e-08, "loss": 0.0098, "step": 18946 }, { "epoch": 7.47, "learning_rate": 6.355510259246624e-08, "loss": 0.0099, "step": 18947 }, { "epoch": 7.47, "learning_rate": 6.353643536281732e-08, "loss": 0.0104, "step": 18948 }, { "epoch": 7.47, "learning_rate": 6.351777035731705e-08, "loss": 0.0082, "step": 18949 }, { "epoch": 7.47, "learning_rate": 6.349910757626969e-08, "loss": 0.0099, "step": 18950 }, { "epoch": 7.47, "learning_rate": 6.34804470199794e-08, "loss": 0.0084, "step": 18951 }, { "epoch": 7.47, "learning_rate": 6.346178868875043e-08, "loss": 0.0108, "step": 18952 }, { "epoch": 7.47, "learning_rate": 6.34431325828867e-08, "loss": 0.0105, "step": 18953 }, { "epoch": 7.47, "learning_rate": 6.342447870269239e-08, "loss": 0.0102, "step": 18954 }, { "epoch": 7.47, "learning_rate": 6.340582704847149e-08, "loss": 0.0099, "step": 18955 }, { "epoch": 7.47, "learning_rate": 6.338717762052808e-08, "loss": 0.0114, "step": 18956 }, { "epoch": 7.47, "learning_rate": 6.336853041916597e-08, "loss": 0.0124, "step": 18957 }, { "epoch": 7.47, "learning_rate": 6.334988544468918e-08, "loss": 0.0093, "step": 18958 }, { "epoch": 7.47, "learning_rate": 6.333124269740162e-08, "loss": 0.0113, "step": 18959 }, { "epoch": 7.47, "learning_rate": 6.331260217760704e-08, "loss": 0.0078, "step": 18960 }, { "epoch": 7.47, "learning_rate": 6.329396388560935e-08, "loss": 0.0134, "step": 18961 }, { "epoch": 7.47, "learning_rate": 6.327532782171223e-08, "loss": 0.0123, "step": 18962 }, { "epoch": 7.47, "learning_rate": 6.325669398621947e-08, "loss": 0.0107, "step": 18963 }, { "epoch": 7.47, "learning_rate": 6.323806237943476e-08, "loss": 0.008, "step": 18964 }, { "epoch": 7.47, "learning_rate": 6.321943300166185e-08, "loss": 0.01, "step": 18965 }, { "epoch": 7.47, "learning_rate": 6.320080585320424e-08, "loss": 0.0109, "step": 18966 }, { "epoch": 7.48, "learning_rate": 6.318218093436559e-08, "loss": 0.0107, "step": 18967 }, { "epoch": 7.48, "learning_rate": 6.316355824544943e-08, "loss": 0.0114, "step": 18968 }, { "epoch": 7.48, "learning_rate": 6.314493778675938e-08, "loss": 0.0085, "step": 18969 }, { "epoch": 7.48, "learning_rate": 6.312631955859877e-08, "loss": 0.0094, "step": 18970 }, { "epoch": 7.48, "learning_rate": 6.310770356127122e-08, "loss": 0.0127, "step": 18971 }, { "epoch": 7.48, "learning_rate": 6.30890897950799e-08, "loss": 0.0123, "step": 18972 }, { "epoch": 7.48, "learning_rate": 6.307047826032848e-08, "loss": 0.0098, "step": 18973 }, { "epoch": 7.48, "learning_rate": 6.305186895732006e-08, "loss": 0.0094, "step": 18974 }, { "epoch": 7.48, "learning_rate": 6.303326188635813e-08, "loss": 0.0114, "step": 18975 }, { "epoch": 7.48, "learning_rate": 6.30146570477458e-08, "loss": 0.0106, "step": 18976 }, { "epoch": 7.48, "learning_rate": 6.299605444178635e-08, "loss": 0.0113, "step": 18977 }, { "epoch": 7.48, "learning_rate": 6.297745406878301e-08, "loss": 0.0084, "step": 18978 }, { "epoch": 7.48, "learning_rate": 6.295885592903898e-08, "loss": 0.0104, "step": 18979 }, { "epoch": 7.48, "learning_rate": 6.294026002285726e-08, "loss": 0.0126, "step": 18980 }, { "epoch": 7.48, "learning_rate": 6.292166635054096e-08, "loss": 0.0097, "step": 18981 }, { "epoch": 7.48, "learning_rate": 6.290307491239326e-08, "loss": 0.011, "step": 18982 }, { "epoch": 7.48, "learning_rate": 6.2884485708717e-08, "loss": 0.0122, "step": 18983 }, { "epoch": 7.48, "learning_rate": 6.286589873981532e-08, "loss": 0.0112, "step": 18984 }, { "epoch": 7.48, "learning_rate": 6.284731400599093e-08, "loss": 0.0114, "step": 18985 }, { "epoch": 7.48, "learning_rate": 6.282873150754698e-08, "loss": 0.0112, "step": 18986 }, { "epoch": 7.48, "learning_rate": 6.281015124478619e-08, "loss": 0.0108, "step": 18987 }, { "epoch": 7.48, "learning_rate": 6.279157321801147e-08, "loss": 0.0094, "step": 18988 }, { "epoch": 7.48, "learning_rate": 6.277299742752552e-08, "loss": 0.0118, "step": 18989 }, { "epoch": 7.48, "learning_rate": 6.275442387363116e-08, "loss": 0.0108, "step": 18990 }, { "epoch": 7.48, "learning_rate": 6.27358525566311e-08, "loss": 0.0092, "step": 18991 }, { "epoch": 7.48, "learning_rate": 6.271728347682809e-08, "loss": 0.0116, "step": 18992 }, { "epoch": 7.49, "learning_rate": 6.269871663452464e-08, "loss": 0.0128, "step": 18993 }, { "epoch": 7.49, "learning_rate": 6.268015203002346e-08, "loss": 0.0127, "step": 18994 }, { "epoch": 7.49, "learning_rate": 6.266158966362707e-08, "loss": 0.0129, "step": 18995 }, { "epoch": 7.49, "learning_rate": 6.264302953563812e-08, "loss": 0.0096, "step": 18996 }, { "epoch": 7.49, "learning_rate": 6.262447164635895e-08, "loss": 0.0116, "step": 18997 }, { "epoch": 7.49, "learning_rate": 6.260591599609213e-08, "loss": 0.0108, "step": 18998 }, { "epoch": 7.49, "learning_rate": 6.258736258514011e-08, "loss": 0.0104, "step": 18999 }, { "epoch": 7.49, "learning_rate": 6.256881141380517e-08, "loss": 0.0122, "step": 19000 }, { "epoch": 7.49, "learning_rate": 6.255026248238972e-08, "loss": 0.012, "step": 19001 }, { "epoch": 7.49, "learning_rate": 6.253171579119618e-08, "loss": 0.0083, "step": 19002 }, { "epoch": 7.49, "learning_rate": 6.25131713405267e-08, "loss": 0.0146, "step": 19003 }, { "epoch": 7.49, "learning_rate": 6.249462913068353e-08, "loss": 0.0121, "step": 19004 }, { "epoch": 7.49, "learning_rate": 6.2476089161969e-08, "loss": 0.0126, "step": 19005 }, { "epoch": 7.49, "learning_rate": 6.245755143468513e-08, "loss": 0.0117, "step": 19006 }, { "epoch": 7.49, "learning_rate": 6.243901594913416e-08, "loss": 0.0123, "step": 19007 }, { "epoch": 7.49, "learning_rate": 6.242048270561815e-08, "loss": 0.0119, "step": 19008 }, { "epoch": 7.49, "learning_rate": 6.240195170443922e-08, "loss": 0.0098, "step": 19009 }, { "epoch": 7.49, "learning_rate": 6.238342294589931e-08, "loss": 0.0094, "step": 19010 }, { "epoch": 7.49, "learning_rate": 6.236489643030052e-08, "loss": 0.0117, "step": 19011 }, { "epoch": 7.49, "learning_rate": 6.234637215794459e-08, "loss": 0.0118, "step": 19012 }, { "epoch": 7.49, "learning_rate": 6.232785012913373e-08, "loss": 0.0087, "step": 19013 }, { "epoch": 7.49, "learning_rate": 6.230933034416959e-08, "loss": 0.0115, "step": 19014 }, { "epoch": 7.49, "learning_rate": 6.229081280335418e-08, "loss": 0.0099, "step": 19015 }, { "epoch": 7.49, "learning_rate": 6.22722975069891e-08, "loss": 0.011, "step": 19016 }, { "epoch": 7.49, "learning_rate": 6.225378445537637e-08, "loss": 0.0104, "step": 19017 }, { "epoch": 7.5, "learning_rate": 6.223527364881754e-08, "loss": 0.0139, "step": 19018 }, { "epoch": 7.5, "learning_rate": 6.221676508761444e-08, "loss": 0.0109, "step": 19019 }, { "epoch": 7.5, "learning_rate": 6.21982587720686e-08, "loss": 0.0099, "step": 19020 }, { "epoch": 7.5, "learning_rate": 6.217975470248169e-08, "loss": 0.0122, "step": 19021 }, { "epoch": 7.5, "learning_rate": 6.216125287915532e-08, "loss": 0.0104, "step": 19022 }, { "epoch": 7.5, "learning_rate": 6.214275330239111e-08, "loss": 0.0109, "step": 19023 }, { "epoch": 7.5, "learning_rate": 6.212425597249044e-08, "loss": 0.0105, "step": 19024 }, { "epoch": 7.5, "learning_rate": 6.210576088975486e-08, "loss": 0.01, "step": 19025 }, { "epoch": 7.5, "learning_rate": 6.208726805448586e-08, "loss": 0.0121, "step": 19026 }, { "epoch": 7.5, "learning_rate": 6.20687774669847e-08, "loss": 0.0145, "step": 19027 }, { "epoch": 7.5, "learning_rate": 6.205028912755292e-08, "loss": 0.0107, "step": 19028 }, { "epoch": 7.5, "learning_rate": 6.203180303649171e-08, "loss": 0.0126, "step": 19029 }, { "epoch": 7.5, "learning_rate": 6.201331919410242e-08, "loss": 0.0093, "step": 19030 }, { "epoch": 7.5, "learning_rate": 6.199483760068629e-08, "loss": 0.0124, "step": 19031 }, { "epoch": 7.5, "learning_rate": 6.197635825654465e-08, "loss": 0.0124, "step": 19032 }, { "epoch": 7.5, "learning_rate": 6.195788116197853e-08, "loss": 0.0112, "step": 19033 }, { "epoch": 7.5, "learning_rate": 6.193940631728916e-08, "loss": 0.0094, "step": 19034 }, { "epoch": 7.5, "learning_rate": 6.19209337227776e-08, "loss": 0.011, "step": 19035 }, { "epoch": 7.5, "learning_rate": 6.190246337874507e-08, "loss": 0.0097, "step": 19036 }, { "epoch": 7.5, "learning_rate": 6.188399528549242e-08, "loss": 0.0088, "step": 19037 }, { "epoch": 7.5, "learning_rate": 6.186552944332073e-08, "loss": 0.0126, "step": 19038 }, { "epoch": 7.5, "learning_rate": 6.1847065852531e-08, "loss": 0.01, "step": 19039 }, { "epoch": 7.5, "learning_rate": 6.182860451342416e-08, "loss": 0.0124, "step": 19040 }, { "epoch": 7.5, "learning_rate": 6.181014542630101e-08, "loss": 0.0103, "step": 19041 }, { "epoch": 7.5, "learning_rate": 6.179168859146253e-08, "loss": 0.0117, "step": 19042 }, { "epoch": 7.51, "learning_rate": 6.177323400920941e-08, "loss": 0.0132, "step": 19043 }, { "epoch": 7.51, "learning_rate": 6.17547816798425e-08, "loss": 0.0113, "step": 19044 }, { "epoch": 7.51, "learning_rate": 6.173633160366254e-08, "loss": 0.0114, "step": 19045 }, { "epoch": 7.51, "learning_rate": 6.171788378097032e-08, "loss": 0.0129, "step": 19046 }, { "epoch": 7.51, "learning_rate": 6.169943821206633e-08, "loss": 0.0092, "step": 19047 }, { "epoch": 7.51, "learning_rate": 6.168099489725132e-08, "loss": 0.0117, "step": 19048 }, { "epoch": 7.51, "learning_rate": 6.166255383682594e-08, "loss": 0.0142, "step": 19049 }, { "epoch": 7.51, "learning_rate": 6.164411503109062e-08, "loss": 0.0101, "step": 19050 }, { "epoch": 7.51, "learning_rate": 6.162567848034601e-08, "loss": 0.0098, "step": 19051 }, { "epoch": 7.51, "learning_rate": 6.16072441848924e-08, "loss": 0.0118, "step": 19052 }, { "epoch": 7.51, "learning_rate": 6.158881214503053e-08, "loss": 0.0114, "step": 19053 }, { "epoch": 7.51, "learning_rate": 6.157038236106058e-08, "loss": 0.0096, "step": 19054 }, { "epoch": 7.51, "learning_rate": 6.155195483328306e-08, "loss": 0.0103, "step": 19055 }, { "epoch": 7.51, "learning_rate": 6.153352956199815e-08, "loss": 0.0121, "step": 19056 }, { "epoch": 7.51, "learning_rate": 6.151510654750638e-08, "loss": 0.0106, "step": 19057 }, { "epoch": 7.51, "learning_rate": 6.149668579010785e-08, "loss": 0.0116, "step": 19058 }, { "epoch": 7.51, "learning_rate": 6.14782672901029e-08, "loss": 0.0099, "step": 19059 }, { "epoch": 7.51, "learning_rate": 6.14598510477916e-08, "loss": 0.0102, "step": 19060 }, { "epoch": 7.51, "learning_rate": 6.144143706347416e-08, "loss": 0.0105, "step": 19061 }, { "epoch": 7.51, "learning_rate": 6.142302533745072e-08, "loss": 0.0141, "step": 19062 }, { "epoch": 7.51, "learning_rate": 6.140461587002141e-08, "loss": 0.0118, "step": 19063 }, { "epoch": 7.51, "learning_rate": 6.138620866148619e-08, "loss": 0.0129, "step": 19064 }, { "epoch": 7.51, "learning_rate": 6.136780371214507e-08, "loss": 0.0108, "step": 19065 }, { "epoch": 7.51, "learning_rate": 6.13494010222981e-08, "loss": 0.011, "step": 19066 }, { "epoch": 7.51, "learning_rate": 6.133100059224514e-08, "loss": 0.0114, "step": 19067 }, { "epoch": 7.51, "learning_rate": 6.13126024222861e-08, "loss": 0.0101, "step": 19068 }, { "epoch": 7.52, "learning_rate": 6.129420651272095e-08, "loss": 0.011, "step": 19069 }, { "epoch": 7.52, "learning_rate": 6.127581286384934e-08, "loss": 0.0128, "step": 19070 }, { "epoch": 7.52, "learning_rate": 6.125742147597113e-08, "loss": 0.0107, "step": 19071 }, { "epoch": 7.52, "learning_rate": 6.123903234938617e-08, "loss": 0.0133, "step": 19072 }, { "epoch": 7.52, "learning_rate": 6.1220645484394e-08, "loss": 0.0123, "step": 19073 }, { "epoch": 7.52, "learning_rate": 6.120226088129444e-08, "loss": 0.0114, "step": 19074 }, { "epoch": 7.52, "learning_rate": 6.118387854038703e-08, "loss": 0.0103, "step": 19075 }, { "epoch": 7.52, "learning_rate": 6.11654984619715e-08, "loss": 0.0111, "step": 19076 }, { "epoch": 7.52, "learning_rate": 6.114712064634728e-08, "loss": 0.012, "step": 19077 }, { "epoch": 7.52, "learning_rate": 6.112874509381396e-08, "loss": 0.0139, "step": 19078 }, { "epoch": 7.52, "learning_rate": 6.111037180467105e-08, "loss": 0.0131, "step": 19079 }, { "epoch": 7.52, "learning_rate": 6.109200077921804e-08, "loss": 0.0119, "step": 19080 }, { "epoch": 7.52, "learning_rate": 6.107363201775424e-08, "loss": 0.0118, "step": 19081 }, { "epoch": 7.52, "learning_rate": 6.105526552057915e-08, "loss": 0.0109, "step": 19082 }, { "epoch": 7.52, "learning_rate": 6.103690128799195e-08, "loss": 0.0111, "step": 19083 }, { "epoch": 7.52, "learning_rate": 6.101853932029218e-08, "loss": 0.01, "step": 19084 }, { "epoch": 7.52, "learning_rate": 6.100017961777892e-08, "loss": 0.0116, "step": 19085 }, { "epoch": 7.52, "learning_rate": 6.098182218075156e-08, "loss": 0.0107, "step": 19086 }, { "epoch": 7.52, "learning_rate": 6.096346700950914e-08, "loss": 0.0105, "step": 19087 }, { "epoch": 7.52, "learning_rate": 6.094511410435091e-08, "loss": 0.0089, "step": 19088 }, { "epoch": 7.52, "learning_rate": 6.092676346557601e-08, "loss": 0.0115, "step": 19089 }, { "epoch": 7.52, "learning_rate": 6.090841509348345e-08, "loss": 0.0105, "step": 19090 }, { "epoch": 7.52, "learning_rate": 6.089006898837234e-08, "loss": 0.0107, "step": 19091 }, { "epoch": 7.52, "learning_rate": 6.087172515054168e-08, "loss": 0.0098, "step": 19092 }, { "epoch": 7.52, "learning_rate": 6.08533835802905e-08, "loss": 0.0111, "step": 19093 }, { "epoch": 7.53, "learning_rate": 6.083504427791764e-08, "loss": 0.0098, "step": 19094 }, { "epoch": 7.53, "learning_rate": 6.081670724372211e-08, "loss": 0.0109, "step": 19095 }, { "epoch": 7.53, "learning_rate": 6.07983724780026e-08, "loss": 0.0101, "step": 19096 }, { "epoch": 7.53, "learning_rate": 6.078003998105818e-08, "loss": 0.0142, "step": 19097 }, { "epoch": 7.53, "learning_rate": 6.076170975318744e-08, "loss": 0.0114, "step": 19098 }, { "epoch": 7.53, "learning_rate": 6.07433817946893e-08, "loss": 0.0109, "step": 19099 }, { "epoch": 7.53, "learning_rate": 6.072505610586231e-08, "loss": 0.0107, "step": 19100 }, { "epoch": 7.53, "learning_rate": 6.070673268700523e-08, "loss": 0.0094, "step": 19101 }, { "epoch": 7.53, "learning_rate": 6.068841153841671e-08, "loss": 0.0106, "step": 19102 }, { "epoch": 7.53, "learning_rate": 6.067009266039542e-08, "loss": 0.0114, "step": 19103 }, { "epoch": 7.53, "learning_rate": 6.065177605323981e-08, "loss": 0.0125, "step": 19104 }, { "epoch": 7.53, "learning_rate": 6.063346171724843e-08, "loss": 0.0114, "step": 19105 }, { "epoch": 7.53, "learning_rate": 6.061514965271986e-08, "loss": 0.0119, "step": 19106 }, { "epoch": 7.53, "learning_rate": 6.059683985995254e-08, "loss": 0.0147, "step": 19107 }, { "epoch": 7.53, "learning_rate": 6.057853233924482e-08, "loss": 0.0088, "step": 19108 }, { "epoch": 7.53, "learning_rate": 6.056022709089517e-08, "loss": 0.0138, "step": 19109 }, { "epoch": 7.53, "learning_rate": 6.054192411520183e-08, "loss": 0.0119, "step": 19110 }, { "epoch": 7.53, "learning_rate": 6.052362341246317e-08, "loss": 0.0107, "step": 19111 }, { "epoch": 7.53, "learning_rate": 6.050532498297756e-08, "loss": 0.0095, "step": 19112 }, { "epoch": 7.53, "learning_rate": 6.048702882704306e-08, "loss": 0.011, "step": 19113 }, { "epoch": 7.53, "learning_rate": 6.046873494495796e-08, "loss": 0.0107, "step": 19114 }, { "epoch": 7.53, "learning_rate": 6.045044333702043e-08, "loss": 0.0102, "step": 19115 }, { "epoch": 7.53, "learning_rate": 6.043215400352863e-08, "loss": 0.0105, "step": 19116 }, { "epoch": 7.53, "learning_rate": 6.041386694478054e-08, "loss": 0.0141, "step": 19117 }, { "epoch": 7.53, "learning_rate": 6.039558216107427e-08, "loss": 0.009, "step": 19118 }, { "epoch": 7.53, "learning_rate": 6.037729965270784e-08, "loss": 0.0121, "step": 19119 }, { "epoch": 7.54, "learning_rate": 6.03590194199793e-08, "loss": 0.0099, "step": 19120 }, { "epoch": 7.54, "learning_rate": 6.034074146318644e-08, "loss": 0.0112, "step": 19121 }, { "epoch": 7.54, "learning_rate": 6.03224657826273e-08, "loss": 0.0106, "step": 19122 }, { "epoch": 7.54, "learning_rate": 6.030419237859954e-08, "loss": 0.0094, "step": 19123 }, { "epoch": 7.54, "learning_rate": 6.028592125140128e-08, "loss": 0.0176, "step": 19124 }, { "epoch": 7.54, "learning_rate": 6.026765240133009e-08, "loss": 0.0105, "step": 19125 }, { "epoch": 7.54, "learning_rate": 6.024938582868386e-08, "loss": 0.0121, "step": 19126 }, { "epoch": 7.54, "learning_rate": 6.023112153376017e-08, "loss": 0.0103, "step": 19127 }, { "epoch": 7.54, "learning_rate": 6.02128595168568e-08, "loss": 0.0082, "step": 19128 }, { "epoch": 7.54, "learning_rate": 6.019459977827135e-08, "loss": 0.0097, "step": 19129 }, { "epoch": 7.54, "learning_rate": 6.017634231830151e-08, "loss": 0.0097, "step": 19130 }, { "epoch": 7.54, "learning_rate": 6.015808713724473e-08, "loss": 0.0108, "step": 19131 }, { "epoch": 7.54, "learning_rate": 6.013983423539858e-08, "loss": 0.008, "step": 19132 }, { "epoch": 7.54, "learning_rate": 6.012158361306066e-08, "loss": 0.0107, "step": 19133 }, { "epoch": 7.54, "learning_rate": 6.010333527052824e-08, "loss": 0.0102, "step": 19134 }, { "epoch": 7.54, "learning_rate": 6.008508920809891e-08, "loss": 0.0101, "step": 19135 }, { "epoch": 7.54, "learning_rate": 6.006684542606985e-08, "loss": 0.0109, "step": 19136 }, { "epoch": 7.54, "learning_rate": 6.004860392473868e-08, "loss": 0.01, "step": 19137 }, { "epoch": 7.54, "learning_rate": 6.003036470440248e-08, "loss": 0.0099, "step": 19138 }, { "epoch": 7.54, "learning_rate": 6.001212776535869e-08, "loss": 0.0111, "step": 19139 }, { "epoch": 7.54, "learning_rate": 5.999389310790437e-08, "loss": 0.013, "step": 19140 }, { "epoch": 7.54, "learning_rate": 5.997566073233682e-08, "loss": 0.0105, "step": 19141 }, { "epoch": 7.54, "learning_rate": 5.99574306389532e-08, "loss": 0.0087, "step": 19142 }, { "epoch": 7.54, "learning_rate": 5.993920282805067e-08, "loss": 0.0124, "step": 19143 }, { "epoch": 7.54, "learning_rate": 5.99209772999262e-08, "loss": 0.0095, "step": 19144 }, { "epoch": 7.55, "learning_rate": 5.990275405487692e-08, "loss": 0.0143, "step": 19145 }, { "epoch": 7.55, "learning_rate": 5.988453309319981e-08, "loss": 0.0141, "step": 19146 }, { "epoch": 7.55, "learning_rate": 5.986631441519194e-08, "loss": 0.0106, "step": 19147 }, { "epoch": 7.55, "learning_rate": 5.98480980211501e-08, "loss": 0.0116, "step": 19148 }, { "epoch": 7.55, "learning_rate": 5.982988391137129e-08, "loss": 0.0099, "step": 19149 }, { "epoch": 7.55, "learning_rate": 5.981167208615223e-08, "loss": 0.0101, "step": 19150 }, { "epoch": 7.55, "learning_rate": 5.979346254578999e-08, "loss": 0.0123, "step": 19151 }, { "epoch": 7.55, "learning_rate": 5.977525529058117e-08, "loss": 0.0095, "step": 19152 }, { "epoch": 7.55, "learning_rate": 5.975705032082262e-08, "loss": 0.0093, "step": 19153 }, { "epoch": 7.55, "learning_rate": 5.973884763681091e-08, "loss": 0.0111, "step": 19154 }, { "epoch": 7.55, "learning_rate": 5.972064723884284e-08, "loss": 0.0101, "step": 19155 }, { "epoch": 7.55, "learning_rate": 5.97024491272151e-08, "loss": 0.0136, "step": 19156 }, { "epoch": 7.55, "learning_rate": 5.968425330222413e-08, "loss": 0.0124, "step": 19157 }, { "epoch": 7.55, "learning_rate": 5.966605976416656e-08, "loss": 0.01, "step": 19158 }, { "epoch": 7.55, "learning_rate": 5.964786851333896e-08, "loss": 0.0115, "step": 19159 }, { "epoch": 7.55, "learning_rate": 5.962967955003784e-08, "loss": 0.0124, "step": 19160 }, { "epoch": 7.55, "learning_rate": 5.961149287455956e-08, "loss": 0.0094, "step": 19161 }, { "epoch": 7.55, "learning_rate": 5.959330848720061e-08, "loss": 0.013, "step": 19162 }, { "epoch": 7.55, "learning_rate": 5.9575126388257235e-08, "loss": 0.014, "step": 19163 }, { "epoch": 7.55, "learning_rate": 5.9556946578026015e-08, "loss": 0.0121, "step": 19164 }, { "epoch": 7.55, "learning_rate": 5.953876905680304e-08, "loss": 0.0117, "step": 19165 }, { "epoch": 7.55, "learning_rate": 5.9520593824884703e-08, "loss": 0.013, "step": 19166 }, { "epoch": 7.55, "learning_rate": 5.950242088256707e-08, "loss": 0.0111, "step": 19167 }, { "epoch": 7.55, "learning_rate": 5.948425023014658e-08, "loss": 0.0115, "step": 19168 }, { "epoch": 7.55, "learning_rate": 5.946608186791917e-08, "loss": 0.0099, "step": 19169 }, { "epoch": 7.56, "learning_rate": 5.944791579618111e-08, "loss": 0.0126, "step": 19170 }, { "epoch": 7.56, "learning_rate": 5.942975201522835e-08, "loss": 0.0108, "step": 19171 }, { "epoch": 7.56, "learning_rate": 5.941159052535698e-08, "loss": 0.0111, "step": 19172 }, { "epoch": 7.56, "learning_rate": 5.939343132686303e-08, "loss": 0.0109, "step": 19173 }, { "epoch": 7.56, "learning_rate": 5.937527442004251e-08, "loss": 0.0097, "step": 19174 }, { "epoch": 7.56, "learning_rate": 5.935711980519122e-08, "loss": 0.0134, "step": 19175 }, { "epoch": 7.56, "learning_rate": 5.933896748260514e-08, "loss": 0.0124, "step": 19176 }, { "epoch": 7.56, "learning_rate": 5.932081745258015e-08, "loss": 0.0136, "step": 19177 }, { "epoch": 7.56, "learning_rate": 5.930266971541198e-08, "loss": 0.0109, "step": 19178 }, { "epoch": 7.56, "learning_rate": 5.928452427139652e-08, "loss": 0.0116, "step": 19179 }, { "epoch": 7.56, "learning_rate": 5.9266381120829404e-08, "loss": 0.0136, "step": 19180 }, { "epoch": 7.56, "learning_rate": 5.9248240264006345e-08, "loss": 0.0092, "step": 19181 }, { "epoch": 7.56, "learning_rate": 5.923010170122309e-08, "loss": 0.0113, "step": 19182 }, { "epoch": 7.56, "learning_rate": 5.921196543277527e-08, "loss": 0.0102, "step": 19183 }, { "epoch": 7.56, "learning_rate": 5.91938314589584e-08, "loss": 0.0142, "step": 19184 }, { "epoch": 7.56, "learning_rate": 5.9175699780068064e-08, "loss": 0.0093, "step": 19185 }, { "epoch": 7.56, "learning_rate": 5.9157570396399794e-08, "loss": 0.0129, "step": 19186 }, { "epoch": 7.56, "learning_rate": 5.9139443308249117e-08, "loss": 0.0098, "step": 19187 }, { "epoch": 7.56, "learning_rate": 5.91213185159114e-08, "loss": 0.0086, "step": 19188 }, { "epoch": 7.56, "learning_rate": 5.910319601968212e-08, "loss": 0.0091, "step": 19189 }, { "epoch": 7.56, "learning_rate": 5.908507581985649e-08, "loss": 0.0101, "step": 19190 }, { "epoch": 7.56, "learning_rate": 5.9066957916730085e-08, "loss": 0.0091, "step": 19191 }, { "epoch": 7.56, "learning_rate": 5.904884231059799e-08, "loss": 0.0106, "step": 19192 }, { "epoch": 7.56, "learning_rate": 5.903072900175563e-08, "loss": 0.0097, "step": 19193 }, { "epoch": 7.56, "learning_rate": 5.901261799049806e-08, "loss": 0.0091, "step": 19194 }, { "epoch": 7.56, "learning_rate": 5.899450927712055e-08, "loss": 0.0109, "step": 19195 }, { "epoch": 7.57, "learning_rate": 5.897640286191825e-08, "loss": 0.0106, "step": 19196 }, { "epoch": 7.57, "learning_rate": 5.8958298745186295e-08, "loss": 0.009, "step": 19197 }, { "epoch": 7.57, "learning_rate": 5.894019692721966e-08, "loss": 0.0134, "step": 19198 }, { "epoch": 7.57, "learning_rate": 5.892209740831342e-08, "loss": 0.0117, "step": 19199 }, { "epoch": 7.57, "learning_rate": 5.8904000188762625e-08, "loss": 0.0116, "step": 19200 }, { "epoch": 7.57, "learning_rate": 5.888590526886215e-08, "loss": 0.0103, "step": 19201 }, { "epoch": 7.57, "learning_rate": 5.886781264890701e-08, "loss": 0.0132, "step": 19202 }, { "epoch": 7.57, "learning_rate": 5.8849722329191895e-08, "loss": 0.0109, "step": 19203 }, { "epoch": 7.57, "learning_rate": 5.8831634310011904e-08, "loss": 0.0113, "step": 19204 }, { "epoch": 7.57, "learning_rate": 5.881354859166168e-08, "loss": 0.011, "step": 19205 }, { "epoch": 7.57, "learning_rate": 5.8795465174436077e-08, "loss": 0.0085, "step": 19206 }, { "epoch": 7.57, "learning_rate": 5.877738405862966e-08, "loss": 0.0093, "step": 19207 }, { "epoch": 7.57, "learning_rate": 5.8759305244537384e-08, "loss": 0.0098, "step": 19208 }, { "epoch": 7.57, "learning_rate": 5.8741228732453705e-08, "loss": 0.0105, "step": 19209 }, { "epoch": 7.57, "learning_rate": 5.872315452267338e-08, "loss": 0.0101, "step": 19210 }, { "epoch": 7.57, "learning_rate": 5.8705082615490855e-08, "loss": 0.01, "step": 19211 }, { "epoch": 7.57, "learning_rate": 5.8687013011200716e-08, "loss": 0.0089, "step": 19212 }, { "epoch": 7.57, "learning_rate": 5.866894571009753e-08, "loss": 0.0093, "step": 19213 }, { "epoch": 7.57, "learning_rate": 5.865088071247577e-08, "loss": 0.012, "step": 19214 }, { "epoch": 7.57, "learning_rate": 5.8632818018629756e-08, "loss": 0.0109, "step": 19215 }, { "epoch": 7.57, "learning_rate": 5.8614757628853993e-08, "loss": 0.0138, "step": 19216 }, { "epoch": 7.57, "learning_rate": 5.859669954344284e-08, "loss": 0.0098, "step": 19217 }, { "epoch": 7.57, "learning_rate": 5.85786437626905e-08, "loss": 0.0085, "step": 19218 }, { "epoch": 7.57, "learning_rate": 5.8560590286891355e-08, "loss": 0.0114, "step": 19219 }, { "epoch": 7.57, "learning_rate": 5.854253911633966e-08, "loss": 0.0098, "step": 19220 }, { "epoch": 7.58, "learning_rate": 5.852449025132955e-08, "loss": 0.0102, "step": 19221 }, { "epoch": 7.58, "learning_rate": 5.850644369215523e-08, "loss": 0.0095, "step": 19222 }, { "epoch": 7.58, "learning_rate": 5.848839943911088e-08, "loss": 0.0094, "step": 19223 }, { "epoch": 7.58, "learning_rate": 5.84703574924905e-08, "loss": 0.0113, "step": 19224 }, { "epoch": 7.58, "learning_rate": 5.845231785258817e-08, "loss": 0.011, "step": 19225 }, { "epoch": 7.58, "learning_rate": 5.843428051969797e-08, "loss": 0.0113, "step": 19226 }, { "epoch": 7.58, "learning_rate": 5.841624549411386e-08, "loss": 0.012, "step": 19227 }, { "epoch": 7.58, "learning_rate": 5.839821277612973e-08, "loss": 0.0108, "step": 19228 }, { "epoch": 7.58, "learning_rate": 5.8380182366039586e-08, "loss": 0.0127, "step": 19229 }, { "epoch": 7.58, "learning_rate": 5.8362154264137086e-08, "loss": 0.0142, "step": 19230 }, { "epoch": 7.58, "learning_rate": 5.834412847071635e-08, "loss": 0.0103, "step": 19231 }, { "epoch": 7.58, "learning_rate": 5.832610498607096e-08, "loss": 0.011, "step": 19232 }, { "epoch": 7.58, "learning_rate": 5.830808381049481e-08, "loss": 0.0106, "step": 19233 }, { "epoch": 7.58, "learning_rate": 5.829006494428142e-08, "loss": 0.0119, "step": 19234 }, { "epoch": 7.58, "learning_rate": 5.82720483877247e-08, "loss": 0.0113, "step": 19235 }, { "epoch": 7.58, "learning_rate": 5.8254034141118155e-08, "loss": 0.011, "step": 19236 }, { "epoch": 7.58, "learning_rate": 5.823602220475548e-08, "loss": 0.0123, "step": 19237 }, { "epoch": 7.58, "learning_rate": 5.821801257893011e-08, "loss": 0.0104, "step": 19238 }, { "epoch": 7.58, "learning_rate": 5.820000526393565e-08, "loss": 0.0099, "step": 19239 }, { "epoch": 7.58, "learning_rate": 5.8182000260065675e-08, "loss": 0.0148, "step": 19240 }, { "epoch": 7.58, "learning_rate": 5.8163997567613473e-08, "loss": 0.0091, "step": 19241 }, { "epoch": 7.58, "learning_rate": 5.814599718687257e-08, "loss": 0.0089, "step": 19242 }, { "epoch": 7.58, "learning_rate": 5.8127999118136306e-08, "loss": 0.0104, "step": 19243 }, { "epoch": 7.58, "learning_rate": 5.811000336169809e-08, "loss": 0.0111, "step": 19244 }, { "epoch": 7.58, "learning_rate": 5.809200991785112e-08, "loss": 0.013, "step": 19245 }, { "epoch": 7.59, "learning_rate": 5.807401878688878e-08, "loss": 0.0153, "step": 19246 }, { "epoch": 7.59, "learning_rate": 5.8056029969104106e-08, "loss": 0.0112, "step": 19247 }, { "epoch": 7.59, "learning_rate": 5.8038043464790533e-08, "loss": 0.0097, "step": 19248 }, { "epoch": 7.59, "learning_rate": 5.802005927424103e-08, "loss": 0.0104, "step": 19249 }, { "epoch": 7.59, "learning_rate": 5.8002077397748825e-08, "loss": 0.0127, "step": 19250 }, { "epoch": 7.59, "learning_rate": 5.7984097835606916e-08, "loss": 0.0132, "step": 19251 }, { "epoch": 7.59, "learning_rate": 5.7966120588108345e-08, "loss": 0.0107, "step": 19252 }, { "epoch": 7.59, "learning_rate": 5.7948145655546155e-08, "loss": 0.01, "step": 19253 }, { "epoch": 7.59, "learning_rate": 5.793017303821335e-08, "loss": 0.0099, "step": 19254 }, { "epoch": 7.59, "learning_rate": 5.791220273640276e-08, "loss": 0.0098, "step": 19255 }, { "epoch": 7.59, "learning_rate": 5.789423475040731e-08, "loss": 0.0105, "step": 19256 }, { "epoch": 7.59, "learning_rate": 5.7876269080519837e-08, "loss": 0.0093, "step": 19257 }, { "epoch": 7.59, "learning_rate": 5.785830572703325e-08, "loss": 0.0108, "step": 19258 }, { "epoch": 7.59, "learning_rate": 5.78403446902402e-08, "loss": 0.0127, "step": 19259 }, { "epoch": 7.59, "learning_rate": 5.782238597043352e-08, "loss": 0.0102, "step": 19260 }, { "epoch": 7.59, "learning_rate": 5.780442956790581e-08, "loss": 0.0124, "step": 19261 }, { "epoch": 7.59, "learning_rate": 5.778647548294979e-08, "loss": 0.0119, "step": 19262 }, { "epoch": 7.59, "learning_rate": 5.776852371585812e-08, "loss": 0.0115, "step": 19263 }, { "epoch": 7.59, "learning_rate": 5.775057426692331e-08, "loss": 0.0123, "step": 19264 }, { "epoch": 7.59, "learning_rate": 5.773262713643794e-08, "loss": 0.0142, "step": 19265 }, { "epoch": 7.59, "learning_rate": 5.771468232469452e-08, "loss": 0.0113, "step": 19266 }, { "epoch": 7.59, "learning_rate": 5.7696739831985576e-08, "loss": 0.0103, "step": 19267 }, { "epoch": 7.59, "learning_rate": 5.767879965860345e-08, "loss": 0.0115, "step": 19268 }, { "epoch": 7.59, "learning_rate": 5.766086180484064e-08, "loss": 0.0104, "step": 19269 }, { "epoch": 7.59, "learning_rate": 5.764292627098934e-08, "loss": 0.0124, "step": 19270 }, { "epoch": 7.59, "learning_rate": 5.762499305734212e-08, "loss": 0.0083, "step": 19271 }, { "epoch": 7.6, "learning_rate": 5.7607062164191044e-08, "loss": 0.0116, "step": 19272 }, { "epoch": 7.6, "learning_rate": 5.758913359182851e-08, "loss": 0.0112, "step": 19273 }, { "epoch": 7.6, "learning_rate": 5.757120734054655e-08, "loss": 0.0093, "step": 19274 }, { "epoch": 7.6, "learning_rate": 5.755328341063754e-08, "loss": 0.0094, "step": 19275 }, { "epoch": 7.6, "learning_rate": 5.753536180239347e-08, "loss": 0.0108, "step": 19276 }, { "epoch": 7.6, "learning_rate": 5.751744251610655e-08, "loss": 0.0124, "step": 19277 }, { "epoch": 7.6, "learning_rate": 5.7499525552068695e-08, "loss": 0.0091, "step": 19278 }, { "epoch": 7.6, "learning_rate": 5.7481610910572e-08, "loss": 0.0108, "step": 19279 }, { "epoch": 7.6, "learning_rate": 5.746369859190845e-08, "loss": 0.0103, "step": 19280 }, { "epoch": 7.6, "learning_rate": 5.744578859637004e-08, "loss": 0.0092, "step": 19281 }, { "epoch": 7.6, "learning_rate": 5.742788092424855e-08, "loss": 0.0096, "step": 19282 }, { "epoch": 7.6, "learning_rate": 5.740997557583591e-08, "loss": 0.0095, "step": 19283 }, { "epoch": 7.6, "learning_rate": 5.7392072551424045e-08, "loss": 0.0119, "step": 19284 }, { "epoch": 7.6, "learning_rate": 5.7374171851304574e-08, "loss": 0.0098, "step": 19285 }, { "epoch": 7.6, "learning_rate": 5.735627347576941e-08, "loss": 0.0092, "step": 19286 }, { "epoch": 7.6, "learning_rate": 5.733837742511005e-08, "loss": 0.012, "step": 19287 }, { "epoch": 7.6, "learning_rate": 5.732048369961846e-08, "loss": 0.0099, "step": 19288 }, { "epoch": 7.6, "learning_rate": 5.730259229958607e-08, "loss": 0.01, "step": 19289 }, { "epoch": 7.6, "learning_rate": 5.728470322530463e-08, "loss": 0.0113, "step": 19290 }, { "epoch": 7.6, "learning_rate": 5.726681647706555e-08, "loss": 0.009, "step": 19291 }, { "epoch": 7.6, "learning_rate": 5.724893205516042e-08, "loss": 0.0144, "step": 19292 }, { "epoch": 7.6, "learning_rate": 5.723104995988075e-08, "loss": 0.0135, "step": 19293 }, { "epoch": 7.6, "learning_rate": 5.721317019151806e-08, "loss": 0.0099, "step": 19294 }, { "epoch": 7.6, "learning_rate": 5.719529275036361e-08, "loss": 0.011, "step": 19295 }, { "epoch": 7.6, "learning_rate": 5.717741763670885e-08, "loss": 0.0108, "step": 19296 }, { "epoch": 7.61, "learning_rate": 5.715954485084511e-08, "loss": 0.0122, "step": 19297 }, { "epoch": 7.61, "learning_rate": 5.7141674393063765e-08, "loss": 0.0095, "step": 19298 }, { "epoch": 7.61, "learning_rate": 5.712380626365594e-08, "loss": 0.012, "step": 19299 }, { "epoch": 7.61, "learning_rate": 5.710594046291299e-08, "loss": 0.0098, "step": 19300 }, { "epoch": 7.61, "learning_rate": 5.708807699112592e-08, "loss": 0.011, "step": 19301 }, { "epoch": 7.61, "learning_rate": 5.7070215848586114e-08, "loss": 0.0113, "step": 19302 }, { "epoch": 7.61, "learning_rate": 5.705235703558449e-08, "loss": 0.0106, "step": 19303 }, { "epoch": 7.61, "learning_rate": 5.703450055241226e-08, "loss": 0.0116, "step": 19304 }, { "epoch": 7.61, "learning_rate": 5.7016646399360325e-08, "loss": 0.009, "step": 19305 }, { "epoch": 7.61, "learning_rate": 5.6998794576719766e-08, "loss": 0.0108, "step": 19306 }, { "epoch": 7.61, "learning_rate": 5.698094508478155e-08, "loss": 0.0116, "step": 19307 }, { "epoch": 7.61, "learning_rate": 5.6963097923836515e-08, "loss": 0.0086, "step": 19308 }, { "epoch": 7.61, "learning_rate": 5.6945253094175594e-08, "loss": 0.0116, "step": 19309 }, { "epoch": 7.61, "learning_rate": 5.692741059608963e-08, "loss": 0.0119, "step": 19310 }, { "epoch": 7.61, "learning_rate": 5.690957042986948e-08, "loss": 0.0113, "step": 19311 }, { "epoch": 7.61, "learning_rate": 5.689173259580582e-08, "loss": 0.0149, "step": 19312 }, { "epoch": 7.61, "learning_rate": 5.687389709418946e-08, "loss": 0.0116, "step": 19313 }, { "epoch": 7.61, "learning_rate": 5.685606392531093e-08, "loss": 0.0115, "step": 19314 }, { "epoch": 7.61, "learning_rate": 5.683823308946114e-08, "loss": 0.0105, "step": 19315 }, { "epoch": 7.61, "learning_rate": 5.6820404586930495e-08, "loss": 0.0104, "step": 19316 }, { "epoch": 7.61, "learning_rate": 5.680257841800971e-08, "loss": 0.013, "step": 19317 }, { "epoch": 7.61, "learning_rate": 5.6784754582989216e-08, "loss": 0.0099, "step": 19318 }, { "epoch": 7.61, "learning_rate": 5.676693308215954e-08, "loss": 0.0111, "step": 19319 }, { "epoch": 7.61, "learning_rate": 5.674911391581116e-08, "loss": 0.0106, "step": 19320 }, { "epoch": 7.61, "learning_rate": 5.6731297084234587e-08, "loss": 0.0097, "step": 19321 }, { "epoch": 7.61, "learning_rate": 5.6713482587720065e-08, "loss": 0.0117, "step": 19322 }, { "epoch": 7.62, "learning_rate": 5.6695670426558004e-08, "loss": 0.0113, "step": 19323 }, { "epoch": 7.62, "learning_rate": 5.667786060103872e-08, "loss": 0.0097, "step": 19324 }, { "epoch": 7.62, "learning_rate": 5.666005311145254e-08, "loss": 0.0116, "step": 19325 }, { "epoch": 7.62, "learning_rate": 5.664224795808958e-08, "loss": 0.0112, "step": 19326 }, { "epoch": 7.62, "learning_rate": 5.662444514124012e-08, "loss": 0.0097, "step": 19327 }, { "epoch": 7.62, "learning_rate": 5.660664466119433e-08, "loss": 0.0128, "step": 19328 }, { "epoch": 7.62, "learning_rate": 5.658884651824225e-08, "loss": 0.0107, "step": 19329 }, { "epoch": 7.62, "learning_rate": 5.6571050712674073e-08, "loss": 0.0102, "step": 19330 }, { "epoch": 7.62, "learning_rate": 5.6553257244779726e-08, "loss": 0.0105, "step": 19331 }, { "epoch": 7.62, "learning_rate": 5.6535466114849276e-08, "loss": 0.0088, "step": 19332 }, { "epoch": 7.62, "learning_rate": 5.651767732317267e-08, "loss": 0.0115, "step": 19333 }, { "epoch": 7.62, "learning_rate": 5.6499890870039925e-08, "loss": 0.0112, "step": 19334 }, { "epoch": 7.62, "learning_rate": 5.648210675574081e-08, "loss": 0.0131, "step": 19335 }, { "epoch": 7.62, "learning_rate": 5.6464324980565236e-08, "loss": 0.0133, "step": 19336 }, { "epoch": 7.62, "learning_rate": 5.6446545544802995e-08, "loss": 0.0111, "step": 19337 }, { "epoch": 7.62, "learning_rate": 5.6428768448743954e-08, "loss": 0.0114, "step": 19338 }, { "epoch": 7.62, "learning_rate": 5.641099369267772e-08, "loss": 0.0112, "step": 19339 }, { "epoch": 7.62, "learning_rate": 5.6393221276894146e-08, "loss": 0.0097, "step": 19340 }, { "epoch": 7.62, "learning_rate": 5.6375451201682656e-08, "loss": 0.0108, "step": 19341 }, { "epoch": 7.62, "learning_rate": 5.635768346733316e-08, "loss": 0.0114, "step": 19342 }, { "epoch": 7.62, "learning_rate": 5.633991807413505e-08, "loss": 0.0105, "step": 19343 }, { "epoch": 7.62, "learning_rate": 5.6322155022378e-08, "loss": 0.0094, "step": 19344 }, { "epoch": 7.62, "learning_rate": 5.630439431235141e-08, "loss": 0.0096, "step": 19345 }, { "epoch": 7.62, "learning_rate": 5.6286635944344777e-08, "loss": 0.0109, "step": 19346 }, { "epoch": 7.62, "learning_rate": 5.6268879918647575e-08, "loss": 0.0111, "step": 19347 }, { "epoch": 7.63, "learning_rate": 5.625112623554924e-08, "loss": 0.0124, "step": 19348 }, { "epoch": 7.63, "learning_rate": 5.6233374895339016e-08, "loss": 0.0107, "step": 19349 }, { "epoch": 7.63, "learning_rate": 5.621562589830626e-08, "loss": 0.0099, "step": 19350 }, { "epoch": 7.63, "learning_rate": 5.619787924474035e-08, "loss": 0.0104, "step": 19351 }, { "epoch": 7.63, "learning_rate": 5.618013493493041e-08, "loss": 0.0098, "step": 19352 }, { "epoch": 7.63, "learning_rate": 5.616239296916574e-08, "loss": 0.0114, "step": 19353 }, { "epoch": 7.63, "learning_rate": 5.6144653347735324e-08, "loss": 0.0113, "step": 19354 }, { "epoch": 7.63, "learning_rate": 5.6126916070928564e-08, "loss": 0.0107, "step": 19355 }, { "epoch": 7.63, "learning_rate": 5.6109181139034335e-08, "loss": 0.013, "step": 19356 }, { "epoch": 7.63, "learning_rate": 5.609144855234183e-08, "loss": 0.0105, "step": 19357 }, { "epoch": 7.63, "learning_rate": 5.607371831113994e-08, "loss": 0.012, "step": 19358 }, { "epoch": 7.63, "learning_rate": 5.605599041571769e-08, "loss": 0.0102, "step": 19359 }, { "epoch": 7.63, "learning_rate": 5.6038264866364026e-08, "loss": 0.0115, "step": 19360 }, { "epoch": 7.63, "learning_rate": 5.602054166336791e-08, "loss": 0.0099, "step": 19361 }, { "epoch": 7.63, "learning_rate": 5.60028208070181e-08, "loss": 0.0109, "step": 19362 }, { "epoch": 7.63, "learning_rate": 5.5985102297603445e-08, "loss": 0.0116, "step": 19363 }, { "epoch": 7.63, "learning_rate": 5.596738613541274e-08, "loss": 0.0098, "step": 19364 }, { "epoch": 7.63, "learning_rate": 5.59496723207348e-08, "loss": 0.0121, "step": 19365 }, { "epoch": 7.63, "learning_rate": 5.593196085385821e-08, "loss": 0.0117, "step": 19366 }, { "epoch": 7.63, "learning_rate": 5.591425173507171e-08, "loss": 0.0115, "step": 19367 }, { "epoch": 7.63, "learning_rate": 5.589654496466396e-08, "loss": 0.0128, "step": 19368 }, { "epoch": 7.63, "learning_rate": 5.5878840542923465e-08, "loss": 0.0104, "step": 19369 }, { "epoch": 7.63, "learning_rate": 5.5861138470138845e-08, "loss": 0.0096, "step": 19370 }, { "epoch": 7.63, "learning_rate": 5.584343874659865e-08, "loss": 0.0124, "step": 19371 }, { "epoch": 7.63, "learning_rate": 5.5825741372591265e-08, "loss": 0.012, "step": 19372 }, { "epoch": 7.64, "learning_rate": 5.5808046348405146e-08, "loss": 0.0118, "step": 19373 }, { "epoch": 7.64, "learning_rate": 5.579035367432881e-08, "loss": 0.0109, "step": 19374 }, { "epoch": 7.64, "learning_rate": 5.577266335065047e-08, "loss": 0.0122, "step": 19375 }, { "epoch": 7.64, "learning_rate": 5.575497537765852e-08, "loss": 0.0093, "step": 19376 }, { "epoch": 7.64, "learning_rate": 5.573728975564122e-08, "loss": 0.0089, "step": 19377 }, { "epoch": 7.64, "learning_rate": 5.571960648488694e-08, "loss": 0.014, "step": 19378 }, { "epoch": 7.64, "learning_rate": 5.570192556568369e-08, "loss": 0.0122, "step": 19379 }, { "epoch": 7.64, "learning_rate": 5.568424699831984e-08, "loss": 0.0089, "step": 19380 }, { "epoch": 7.64, "learning_rate": 5.56665707830833e-08, "loss": 0.0106, "step": 19381 }, { "epoch": 7.64, "learning_rate": 5.5648896920262424e-08, "loss": 0.0119, "step": 19382 }, { "epoch": 7.64, "learning_rate": 5.5631225410145065e-08, "loss": 0.0102, "step": 19383 }, { "epoch": 7.64, "learning_rate": 5.561355625301938e-08, "loss": 0.0095, "step": 19384 }, { "epoch": 7.64, "learning_rate": 5.559588944917317e-08, "loss": 0.0108, "step": 19385 }, { "epoch": 7.64, "learning_rate": 5.557822499889462e-08, "loss": 0.0088, "step": 19386 }, { "epoch": 7.64, "learning_rate": 5.556056290247142e-08, "loss": 0.0116, "step": 19387 }, { "epoch": 7.64, "learning_rate": 5.5542903160191614e-08, "loss": 0.0108, "step": 19388 }, { "epoch": 7.64, "learning_rate": 5.552524577234286e-08, "loss": 0.0152, "step": 19389 }, { "epoch": 7.64, "learning_rate": 5.550759073921303e-08, "loss": 0.0113, "step": 19390 }, { "epoch": 7.64, "learning_rate": 5.5489938061089925e-08, "loss": 0.0116, "step": 19391 }, { "epoch": 7.64, "learning_rate": 5.5472287738261136e-08, "loss": 0.0102, "step": 19392 }, { "epoch": 7.64, "learning_rate": 5.545463977101439e-08, "loss": 0.0095, "step": 19393 }, { "epoch": 7.64, "learning_rate": 5.543699415963734e-08, "loss": 0.0095, "step": 19394 }, { "epoch": 7.64, "learning_rate": 5.541935090441765e-08, "loss": 0.0124, "step": 19395 }, { "epoch": 7.64, "learning_rate": 5.540171000564273e-08, "loss": 0.0104, "step": 19396 }, { "epoch": 7.64, "learning_rate": 5.5384071463600244e-08, "loss": 0.0103, "step": 19397 }, { "epoch": 7.64, "learning_rate": 5.536643527857752e-08, "loss": 0.0121, "step": 19398 }, { "epoch": 7.65, "learning_rate": 5.534880145086209e-08, "loss": 0.0095, "step": 19399 }, { "epoch": 7.65, "learning_rate": 5.5331169980741365e-08, "loss": 0.0107, "step": 19400 }, { "epoch": 7.65, "learning_rate": 5.531354086850275e-08, "loss": 0.0095, "step": 19401 }, { "epoch": 7.65, "learning_rate": 5.529591411443346e-08, "loss": 0.0118, "step": 19402 }, { "epoch": 7.65, "learning_rate": 5.527828971882085e-08, "loss": 0.0091, "step": 19403 }, { "epoch": 7.65, "learning_rate": 5.5260667681952166e-08, "loss": 0.0116, "step": 19404 }, { "epoch": 7.65, "learning_rate": 5.524304800411468e-08, "loss": 0.0117, "step": 19405 }, { "epoch": 7.65, "learning_rate": 5.522543068559544e-08, "loss": 0.0116, "step": 19406 }, { "epoch": 7.65, "learning_rate": 5.520781572668165e-08, "loss": 0.0093, "step": 19407 }, { "epoch": 7.65, "learning_rate": 5.519020312766041e-08, "loss": 0.0109, "step": 19408 }, { "epoch": 7.65, "learning_rate": 5.517259288881884e-08, "loss": 0.0101, "step": 19409 }, { "epoch": 7.65, "learning_rate": 5.515498501044382e-08, "loss": 0.0129, "step": 19410 }, { "epoch": 7.65, "learning_rate": 5.513737949282247e-08, "loss": 0.0124, "step": 19411 }, { "epoch": 7.65, "learning_rate": 5.511977633624161e-08, "loss": 0.0114, "step": 19412 }, { "epoch": 7.65, "learning_rate": 5.5102175540988196e-08, "loss": 0.015, "step": 19413 }, { "epoch": 7.65, "learning_rate": 5.508457710734917e-08, "loss": 0.0102, "step": 19414 }, { "epoch": 7.65, "learning_rate": 5.506698103561123e-08, "loss": 0.0115, "step": 19415 }, { "epoch": 7.65, "learning_rate": 5.504938732606122e-08, "loss": 0.0122, "step": 19416 }, { "epoch": 7.65, "learning_rate": 5.503179597898591e-08, "loss": 0.0115, "step": 19417 }, { "epoch": 7.65, "learning_rate": 5.501420699467203e-08, "loss": 0.0103, "step": 19418 }, { "epoch": 7.65, "learning_rate": 5.49966203734062e-08, "loss": 0.0108, "step": 19419 }, { "epoch": 7.65, "learning_rate": 5.497903611547512e-08, "loss": 0.0129, "step": 19420 }, { "epoch": 7.65, "learning_rate": 5.4961454221165223e-08, "loss": 0.0105, "step": 19421 }, { "epoch": 7.65, "learning_rate": 5.4943874690763314e-08, "loss": 0.0092, "step": 19422 }, { "epoch": 7.65, "learning_rate": 5.492629752455571e-08, "loss": 0.0109, "step": 19423 }, { "epoch": 7.66, "learning_rate": 5.490872272282905e-08, "loss": 0.0129, "step": 19424 }, { "epoch": 7.66, "learning_rate": 5.489115028586957e-08, "loss": 0.0095, "step": 19425 }, { "epoch": 7.66, "learning_rate": 5.4873580213963934e-08, "loss": 0.01, "step": 19426 }, { "epoch": 7.66, "learning_rate": 5.485601250739829e-08, "loss": 0.01, "step": 19427 }, { "epoch": 7.66, "learning_rate": 5.483844716645913e-08, "loss": 0.0147, "step": 19428 }, { "epoch": 7.66, "learning_rate": 5.4820884191432584e-08, "loss": 0.0124, "step": 19429 }, { "epoch": 7.66, "learning_rate": 5.4803323582604996e-08, "loss": 0.0092, "step": 19430 }, { "epoch": 7.66, "learning_rate": 5.478576534026254e-08, "loss": 0.0111, "step": 19431 }, { "epoch": 7.66, "learning_rate": 5.4768209464691475e-08, "loss": 0.0092, "step": 19432 }, { "epoch": 7.66, "learning_rate": 5.4750655956177825e-08, "loss": 0.0114, "step": 19433 }, { "epoch": 7.66, "learning_rate": 5.47331048150077e-08, "loss": 0.0083, "step": 19434 }, { "epoch": 7.66, "learning_rate": 5.471555604146727e-08, "loss": 0.0116, "step": 19435 }, { "epoch": 7.66, "learning_rate": 5.469800963584241e-08, "loss": 0.0116, "step": 19436 }, { "epoch": 7.66, "learning_rate": 5.4680465598419145e-08, "loss": 0.0115, "step": 19437 }, { "epoch": 7.66, "learning_rate": 5.466292392948351e-08, "loss": 0.0119, "step": 19438 }, { "epoch": 7.66, "learning_rate": 5.464538462932125e-08, "loss": 0.0096, "step": 19439 }, { "epoch": 7.66, "learning_rate": 5.462784769821831e-08, "loss": 0.0113, "step": 19440 }, { "epoch": 7.66, "learning_rate": 5.461031313646058e-08, "loss": 0.0099, "step": 19441 }, { "epoch": 7.66, "learning_rate": 5.459278094433371e-08, "loss": 0.0101, "step": 19442 }, { "epoch": 7.66, "learning_rate": 5.457525112212353e-08, "loss": 0.0105, "step": 19443 }, { "epoch": 7.66, "learning_rate": 5.455772367011573e-08, "loss": 0.0109, "step": 19444 }, { "epoch": 7.66, "learning_rate": 5.454019858859604e-08, "loss": 0.0131, "step": 19445 }, { "epoch": 7.66, "learning_rate": 5.4522675877849975e-08, "loss": 0.008, "step": 19446 }, { "epoch": 7.66, "learning_rate": 5.4505155538163214e-08, "loss": 0.011, "step": 19447 }, { "epoch": 7.66, "learning_rate": 5.448763756982126e-08, "loss": 0.0089, "step": 19448 }, { "epoch": 7.67, "learning_rate": 5.4470121973109764e-08, "loss": 0.0111, "step": 19449 }, { "epoch": 7.67, "learning_rate": 5.445260874831401e-08, "loss": 0.0107, "step": 19450 }, { "epoch": 7.67, "learning_rate": 5.4435097895719586e-08, "loss": 0.0121, "step": 19451 }, { "epoch": 7.67, "learning_rate": 5.441758941561172e-08, "loss": 0.0151, "step": 19452 }, { "epoch": 7.67, "learning_rate": 5.4400083308276014e-08, "loss": 0.0131, "step": 19453 }, { "epoch": 7.67, "learning_rate": 5.438257957399759e-08, "loss": 0.0129, "step": 19454 }, { "epoch": 7.67, "learning_rate": 5.436507821306189e-08, "loss": 0.013, "step": 19455 }, { "epoch": 7.67, "learning_rate": 5.434757922575399e-08, "loss": 0.0095, "step": 19456 }, { "epoch": 7.67, "learning_rate": 5.4330082612359184e-08, "loss": 0.0107, "step": 19457 }, { "epoch": 7.67, "learning_rate": 5.4312588373162707e-08, "loss": 0.0156, "step": 19458 }, { "epoch": 7.67, "learning_rate": 5.429509650844957e-08, "loss": 0.0123, "step": 19459 }, { "epoch": 7.67, "learning_rate": 5.42776070185049e-08, "loss": 0.0099, "step": 19460 }, { "epoch": 7.67, "learning_rate": 5.426011990361377e-08, "loss": 0.0113, "step": 19461 }, { "epoch": 7.67, "learning_rate": 5.424263516406122e-08, "loss": 0.0117, "step": 19462 }, { "epoch": 7.67, "learning_rate": 5.422515280013216e-08, "loss": 0.0114, "step": 19463 }, { "epoch": 7.67, "learning_rate": 5.4207672812111605e-08, "loss": 0.0102, "step": 19464 }, { "epoch": 7.67, "learning_rate": 5.419019520028429e-08, "loss": 0.011, "step": 19465 }, { "epoch": 7.67, "learning_rate": 5.417271996493531e-08, "loss": 0.0129, "step": 19466 }, { "epoch": 7.67, "learning_rate": 5.415524710634931e-08, "loss": 0.0143, "step": 19467 }, { "epoch": 7.67, "learning_rate": 5.413777662481116e-08, "loss": 0.0118, "step": 19468 }, { "epoch": 7.67, "learning_rate": 5.412030852060554e-08, "loss": 0.0093, "step": 19469 }, { "epoch": 7.67, "learning_rate": 5.4102842794017156e-08, "loss": 0.0119, "step": 19470 }, { "epoch": 7.67, "learning_rate": 5.408537944533069e-08, "loss": 0.012, "step": 19471 }, { "epoch": 7.67, "learning_rate": 5.406791847483085e-08, "loss": 0.0119, "step": 19472 }, { "epoch": 7.67, "learning_rate": 5.405045988280208e-08, "loss": 0.011, "step": 19473 }, { "epoch": 7.67, "learning_rate": 5.403300366952899e-08, "loss": 0.0089, "step": 19474 }, { "epoch": 7.68, "learning_rate": 5.4015549835296105e-08, "loss": 0.0099, "step": 19475 }, { "epoch": 7.68, "learning_rate": 5.3998098380387954e-08, "loss": 0.01, "step": 19476 }, { "epoch": 7.68, "learning_rate": 5.398064930508883e-08, "loss": 0.0076, "step": 19477 }, { "epoch": 7.68, "learning_rate": 5.3963202609683254e-08, "loss": 0.01, "step": 19478 }, { "epoch": 7.68, "learning_rate": 5.394575829445549e-08, "loss": 0.0119, "step": 19479 }, { "epoch": 7.68, "learning_rate": 5.392831635968986e-08, "loss": 0.0094, "step": 19480 }, { "epoch": 7.68, "learning_rate": 5.3910876805670746e-08, "loss": 0.009, "step": 19481 }, { "epoch": 7.68, "learning_rate": 5.389343963268225e-08, "loss": 0.0089, "step": 19482 }, { "epoch": 7.68, "learning_rate": 5.3876004841008624e-08, "loss": 0.01, "step": 19483 }, { "epoch": 7.68, "learning_rate": 5.3858572430934036e-08, "loss": 0.0101, "step": 19484 }, { "epoch": 7.68, "learning_rate": 5.384114240274269e-08, "loss": 0.0131, "step": 19485 }, { "epoch": 7.68, "learning_rate": 5.38237147567185e-08, "loss": 0.0154, "step": 19486 }, { "epoch": 7.68, "learning_rate": 5.380628949314561e-08, "loss": 0.0107, "step": 19487 }, { "epoch": 7.68, "learning_rate": 5.378886661230802e-08, "loss": 0.0117, "step": 19488 }, { "epoch": 7.68, "learning_rate": 5.3771446114489713e-08, "loss": 0.0122, "step": 19489 }, { "epoch": 7.68, "learning_rate": 5.375402799997457e-08, "loss": 0.009, "step": 19490 }, { "epoch": 7.68, "learning_rate": 5.373661226904656e-08, "loss": 0.0118, "step": 19491 }, { "epoch": 7.68, "learning_rate": 5.371919892198933e-08, "loss": 0.0098, "step": 19492 }, { "epoch": 7.68, "learning_rate": 5.370178795908698e-08, "loss": 0.0086, "step": 19493 }, { "epoch": 7.68, "learning_rate": 5.368437938062307e-08, "loss": 0.0147, "step": 19494 }, { "epoch": 7.68, "learning_rate": 5.366697318688145e-08, "loss": 0.0127, "step": 19495 }, { "epoch": 7.68, "learning_rate": 5.364956937814573e-08, "loss": 0.0108, "step": 19496 }, { "epoch": 7.68, "learning_rate": 5.363216795469958e-08, "loss": 0.0128, "step": 19497 }, { "epoch": 7.68, "learning_rate": 5.3614768916826655e-08, "loss": 0.0127, "step": 19498 }, { "epoch": 7.68, "learning_rate": 5.359737226481056e-08, "loss": 0.0099, "step": 19499 }, { "epoch": 7.69, "learning_rate": 5.3579977998934744e-08, "loss": 0.0113, "step": 19500 }, { "epoch": 7.69, "learning_rate": 5.356258611948277e-08, "loss": 0.0123, "step": 19501 }, { "epoch": 7.69, "learning_rate": 5.354519662673811e-08, "loss": 0.014, "step": 19502 }, { "epoch": 7.69, "learning_rate": 5.352780952098412e-08, "loss": 0.0093, "step": 19503 }, { "epoch": 7.69, "learning_rate": 5.351042480250427e-08, "loss": 0.0094, "step": 19504 }, { "epoch": 7.69, "learning_rate": 5.349304247158175e-08, "loss": 0.0123, "step": 19505 }, { "epoch": 7.69, "learning_rate": 5.3475662528500086e-08, "loss": 0.0102, "step": 19506 }, { "epoch": 7.69, "learning_rate": 5.3458284973542366e-08, "loss": 0.0118, "step": 19507 }, { "epoch": 7.69, "learning_rate": 5.344090980699196e-08, "loss": 0.0112, "step": 19508 }, { "epoch": 7.69, "learning_rate": 5.3423537029131895e-08, "loss": 0.0106, "step": 19509 }, { "epoch": 7.69, "learning_rate": 5.340616664024542e-08, "loss": 0.0084, "step": 19510 }, { "epoch": 7.69, "learning_rate": 5.338879864061563e-08, "loss": 0.0107, "step": 19511 }, { "epoch": 7.69, "learning_rate": 5.337143303052567e-08, "loss": 0.0092, "step": 19512 }, { "epoch": 7.69, "learning_rate": 5.335406981025843e-08, "loss": 0.0136, "step": 19513 }, { "epoch": 7.69, "learning_rate": 5.333670898009699e-08, "loss": 0.0101, "step": 19514 }, { "epoch": 7.69, "learning_rate": 5.3319350540324284e-08, "loss": 0.0116, "step": 19515 }, { "epoch": 7.69, "learning_rate": 5.330199449122328e-08, "loss": 0.0112, "step": 19516 }, { "epoch": 7.69, "learning_rate": 5.328464083307676e-08, "loss": 0.0117, "step": 19517 }, { "epoch": 7.69, "learning_rate": 5.326728956616768e-08, "loss": 0.0107, "step": 19518 }, { "epoch": 7.69, "learning_rate": 5.3249940690778706e-08, "loss": 0.0138, "step": 19519 }, { "epoch": 7.69, "learning_rate": 5.3232594207192664e-08, "loss": 0.0114, "step": 19520 }, { "epoch": 7.69, "learning_rate": 5.321525011569228e-08, "loss": 0.0135, "step": 19521 }, { "epoch": 7.69, "learning_rate": 5.319790841656029e-08, "loss": 0.0107, "step": 19522 }, { "epoch": 7.69, "learning_rate": 5.31805691100792e-08, "loss": 0.0112, "step": 19523 }, { "epoch": 7.69, "learning_rate": 5.316323219653171e-08, "loss": 0.0092, "step": 19524 }, { "epoch": 7.69, "learning_rate": 5.314589767620043e-08, "loss": 0.0093, "step": 19525 }, { "epoch": 7.7, "learning_rate": 5.312856554936778e-08, "loss": 0.0101, "step": 19526 }, { "epoch": 7.7, "learning_rate": 5.311123581631627e-08, "loss": 0.0103, "step": 19527 }, { "epoch": 7.7, "learning_rate": 5.3093908477328373e-08, "loss": 0.0088, "step": 19528 }, { "epoch": 7.7, "learning_rate": 5.307658353268658e-08, "loss": 0.0136, "step": 19529 }, { "epoch": 7.7, "learning_rate": 5.3059260982673107e-08, "loss": 0.0104, "step": 19530 }, { "epoch": 7.7, "learning_rate": 5.3041940827570407e-08, "loss": 0.0121, "step": 19531 }, { "epoch": 7.7, "learning_rate": 5.3024623067660604e-08, "loss": 0.0105, "step": 19532 }, { "epoch": 7.7, "learning_rate": 5.300730770322617e-08, "loss": 0.0108, "step": 19533 }, { "epoch": 7.7, "learning_rate": 5.2989994734549174e-08, "loss": 0.011, "step": 19534 }, { "epoch": 7.7, "learning_rate": 5.2972684161911895e-08, "loss": 0.0097, "step": 19535 }, { "epoch": 7.7, "learning_rate": 5.29553759855963e-08, "loss": 0.0129, "step": 19536 }, { "epoch": 7.7, "learning_rate": 5.29380702058847e-08, "loss": 0.0105, "step": 19537 }, { "epoch": 7.7, "learning_rate": 5.2920766823058996e-08, "loss": 0.0099, "step": 19538 }, { "epoch": 7.7, "learning_rate": 5.290346583740131e-08, "loss": 0.0114, "step": 19539 }, { "epoch": 7.7, "learning_rate": 5.2886167249193505e-08, "loss": 0.0134, "step": 19540 }, { "epoch": 7.7, "learning_rate": 5.28688710587176e-08, "loss": 0.009, "step": 19541 }, { "epoch": 7.7, "learning_rate": 5.285157726625553e-08, "loss": 0.0126, "step": 19542 }, { "epoch": 7.7, "learning_rate": 5.2834285872089065e-08, "loss": 0.011, "step": 19543 }, { "epoch": 7.7, "learning_rate": 5.281699687650005e-08, "loss": 0.0109, "step": 19544 }, { "epoch": 7.7, "learning_rate": 5.279971027977033e-08, "loss": 0.0097, "step": 19545 }, { "epoch": 7.7, "learning_rate": 5.2782426082181644e-08, "loss": 0.0124, "step": 19546 }, { "epoch": 7.7, "learning_rate": 5.2765144284015616e-08, "loss": 0.0102, "step": 19547 }, { "epoch": 7.7, "learning_rate": 5.274786488555403e-08, "loss": 0.0092, "step": 19548 }, { "epoch": 7.7, "learning_rate": 5.2730587887078383e-08, "loss": 0.0105, "step": 19549 }, { "epoch": 7.7, "learning_rate": 5.2713313288870344e-08, "loss": 0.0089, "step": 19550 }, { "epoch": 7.71, "learning_rate": 5.269604109121142e-08, "loss": 0.0132, "step": 19551 }, { "epoch": 7.71, "learning_rate": 5.267877129438323e-08, "loss": 0.0119, "step": 19552 }, { "epoch": 7.71, "learning_rate": 5.2661503898667105e-08, "loss": 0.0113, "step": 19553 }, { "epoch": 7.71, "learning_rate": 5.264423890434453e-08, "loss": 0.0124, "step": 19554 }, { "epoch": 7.71, "learning_rate": 5.2626976311696926e-08, "loss": 0.0106, "step": 19555 }, { "epoch": 7.71, "learning_rate": 5.2609716121005666e-08, "loss": 0.0139, "step": 19556 }, { "epoch": 7.71, "learning_rate": 5.259245833255197e-08, "loss": 0.0122, "step": 19557 }, { "epoch": 7.71, "learning_rate": 5.257520294661724e-08, "loss": 0.0114, "step": 19558 }, { "epoch": 7.71, "learning_rate": 5.255794996348251e-08, "loss": 0.011, "step": 19559 }, { "epoch": 7.71, "learning_rate": 5.2540699383429224e-08, "loss": 0.0106, "step": 19560 }, { "epoch": 7.71, "learning_rate": 5.252345120673838e-08, "loss": 0.0106, "step": 19561 }, { "epoch": 7.71, "learning_rate": 5.2506205433691195e-08, "loss": 0.0136, "step": 19562 }, { "epoch": 7.71, "learning_rate": 5.248896206456863e-08, "loss": 0.0105, "step": 19563 }, { "epoch": 7.71, "learning_rate": 5.2471721099651814e-08, "loss": 0.0093, "step": 19564 }, { "epoch": 7.71, "learning_rate": 5.245448253922171e-08, "loss": 0.0105, "step": 19565 }, { "epoch": 7.71, "learning_rate": 5.2437246383559376e-08, "loss": 0.0092, "step": 19566 }, { "epoch": 7.71, "learning_rate": 5.2420012632945574e-08, "loss": 0.0099, "step": 19567 }, { "epoch": 7.71, "learning_rate": 5.2402781287661265e-08, "loss": 0.0116, "step": 19568 }, { "epoch": 7.71, "learning_rate": 5.2385552347987386e-08, "loss": 0.0122, "step": 19569 }, { "epoch": 7.71, "learning_rate": 5.236832581420458e-08, "loss": 0.0091, "step": 19570 }, { "epoch": 7.71, "learning_rate": 5.2351101686593756e-08, "loss": 0.0105, "step": 19571 }, { "epoch": 7.71, "learning_rate": 5.2333879965435436e-08, "loss": 0.0103, "step": 19572 }, { "epoch": 7.71, "learning_rate": 5.231666065101059e-08, "loss": 0.0092, "step": 19573 }, { "epoch": 7.71, "learning_rate": 5.2299443743599646e-08, "loss": 0.0089, "step": 19574 }, { "epoch": 7.71, "learning_rate": 5.2282229243483354e-08, "loss": 0.0078, "step": 19575 }, { "epoch": 7.72, "learning_rate": 5.226501715094211e-08, "loss": 0.0095, "step": 19576 }, { "epoch": 7.72, "learning_rate": 5.224780746625666e-08, "loss": 0.0089, "step": 19577 }, { "epoch": 7.72, "learning_rate": 5.2230600189707333e-08, "loss": 0.0113, "step": 19578 }, { "epoch": 7.72, "learning_rate": 5.221339532157472e-08, "loss": 0.0083, "step": 19579 }, { "epoch": 7.72, "learning_rate": 5.219619286213908e-08, "loss": 0.0137, "step": 19580 }, { "epoch": 7.72, "learning_rate": 5.2178992811680854e-08, "loss": 0.0138, "step": 19581 }, { "epoch": 7.72, "learning_rate": 5.216179517048038e-08, "loss": 0.0095, "step": 19582 }, { "epoch": 7.72, "learning_rate": 5.2144599938818036e-08, "loss": 0.0113, "step": 19583 }, { "epoch": 7.72, "learning_rate": 5.212740711697392e-08, "loss": 0.0097, "step": 19584 }, { "epoch": 7.72, "learning_rate": 5.211021670522833e-08, "loss": 0.0103, "step": 19585 }, { "epoch": 7.72, "learning_rate": 5.2093028703861494e-08, "loss": 0.0126, "step": 19586 }, { "epoch": 7.72, "learning_rate": 5.207584311315345e-08, "loss": 0.0115, "step": 19587 }, { "epoch": 7.72, "learning_rate": 5.205865993338434e-08, "loss": 0.0113, "step": 19588 }, { "epoch": 7.72, "learning_rate": 5.20414791648343e-08, "loss": 0.0098, "step": 19589 }, { "epoch": 7.72, "learning_rate": 5.202430080778322e-08, "loss": 0.014, "step": 19590 }, { "epoch": 7.72, "learning_rate": 5.2007124862511135e-08, "loss": 0.0102, "step": 19591 }, { "epoch": 7.72, "learning_rate": 5.1989951329298046e-08, "loss": 0.0107, "step": 19592 }, { "epoch": 7.72, "learning_rate": 5.197278020842373e-08, "loss": 0.0125, "step": 19593 }, { "epoch": 7.72, "learning_rate": 5.195561150016814e-08, "loss": 0.0091, "step": 19594 }, { "epoch": 7.72, "learning_rate": 5.1938445204811056e-08, "loss": 0.0112, "step": 19595 }, { "epoch": 7.72, "learning_rate": 5.192128132263236e-08, "loss": 0.013, "step": 19596 }, { "epoch": 7.72, "learning_rate": 5.1904119853911676e-08, "loss": 0.013, "step": 19597 }, { "epoch": 7.72, "learning_rate": 5.188696079892882e-08, "loss": 0.0142, "step": 19598 }, { "epoch": 7.72, "learning_rate": 5.186980415796327e-08, "loss": 0.0124, "step": 19599 }, { "epoch": 7.72, "learning_rate": 5.18526499312949e-08, "loss": 0.0118, "step": 19600 }, { "epoch": 7.72, "learning_rate": 5.183549811920311e-08, "loss": 0.0118, "step": 19601 }, { "epoch": 7.73, "learning_rate": 5.181834872196757e-08, "loss": 0.0118, "step": 19602 }, { "epoch": 7.73, "learning_rate": 5.180120173986764e-08, "loss": 0.0139, "step": 19603 }, { "epoch": 7.73, "learning_rate": 5.178405717318299e-08, "loss": 0.0087, "step": 19604 }, { "epoch": 7.73, "learning_rate": 5.1766915022192904e-08, "loss": 0.0147, "step": 19605 }, { "epoch": 7.73, "learning_rate": 5.1749775287176854e-08, "loss": 0.0128, "step": 19606 }, { "epoch": 7.73, "learning_rate": 5.1732637968414096e-08, "loss": 0.0111, "step": 19607 }, { "epoch": 7.73, "learning_rate": 5.171550306618402e-08, "loss": 0.0108, "step": 19608 }, { "epoch": 7.73, "learning_rate": 5.169837058076591e-08, "loss": 0.0144, "step": 19609 }, { "epoch": 7.73, "learning_rate": 5.1681240512438914e-08, "loss": 0.0115, "step": 19610 }, { "epoch": 7.73, "learning_rate": 5.166411286148229e-08, "loss": 0.0109, "step": 19611 }, { "epoch": 7.73, "learning_rate": 5.164698762817516e-08, "loss": 0.0106, "step": 19612 }, { "epoch": 7.73, "learning_rate": 5.162986481279674e-08, "loss": 0.0105, "step": 19613 }, { "epoch": 7.73, "learning_rate": 5.1612744415625976e-08, "loss": 0.0101, "step": 19614 }, { "epoch": 7.73, "learning_rate": 5.1595626436942e-08, "loss": 0.0119, "step": 19615 }, { "epoch": 7.73, "learning_rate": 5.157851087702365e-08, "loss": 0.0104, "step": 19616 }, { "epoch": 7.73, "learning_rate": 5.156139773615011e-08, "loss": 0.0113, "step": 19617 }, { "epoch": 7.73, "learning_rate": 5.154428701460012e-08, "loss": 0.0116, "step": 19618 }, { "epoch": 7.73, "learning_rate": 5.1527178712652705e-08, "loss": 0.0116, "step": 19619 }, { "epoch": 7.73, "learning_rate": 5.151007283058655e-08, "loss": 0.0112, "step": 19620 }, { "epoch": 7.73, "learning_rate": 5.149296936868053e-08, "loss": 0.0085, "step": 19621 }, { "epoch": 7.73, "learning_rate": 5.147586832721342e-08, "loss": 0.0096, "step": 19622 }, { "epoch": 7.73, "learning_rate": 5.1458769706463965e-08, "loss": 0.0096, "step": 19623 }, { "epoch": 7.73, "learning_rate": 5.1441673506710756e-08, "loss": 0.0108, "step": 19624 }, { "epoch": 7.73, "learning_rate": 5.142457972823249e-08, "loss": 0.0089, "step": 19625 }, { "epoch": 7.73, "learning_rate": 5.1407488371307753e-08, "loss": 0.0088, "step": 19626 }, { "epoch": 7.74, "learning_rate": 5.1390399436215195e-08, "loss": 0.0093, "step": 19627 }, { "epoch": 7.74, "learning_rate": 5.137331292323319e-08, "loss": 0.0122, "step": 19628 }, { "epoch": 7.74, "learning_rate": 5.135622883264039e-08, "loss": 0.0092, "step": 19629 }, { "epoch": 7.74, "learning_rate": 5.133914716471506e-08, "loss": 0.0096, "step": 19630 }, { "epoch": 7.74, "learning_rate": 5.1322067919735704e-08, "loss": 0.0132, "step": 19631 }, { "epoch": 7.74, "learning_rate": 5.1304991097980724e-08, "loss": 0.0107, "step": 19632 }, { "epoch": 7.74, "learning_rate": 5.128791669972834e-08, "loss": 0.0124, "step": 19633 }, { "epoch": 7.74, "learning_rate": 5.127084472525691e-08, "loss": 0.0099, "step": 19634 }, { "epoch": 7.74, "learning_rate": 5.125377517484466e-08, "loss": 0.0106, "step": 19635 }, { "epoch": 7.74, "learning_rate": 5.123670804876987e-08, "loss": 0.0119, "step": 19636 }, { "epoch": 7.74, "learning_rate": 5.1219643347310594e-08, "loss": 0.0108, "step": 19637 }, { "epoch": 7.74, "learning_rate": 5.1202581070745046e-08, "loss": 0.0093, "step": 19638 }, { "epoch": 7.74, "learning_rate": 5.118552121935118e-08, "loss": 0.0111, "step": 19639 }, { "epoch": 7.74, "learning_rate": 5.116846379340727e-08, "loss": 0.0097, "step": 19640 }, { "epoch": 7.74, "learning_rate": 5.115140879319113e-08, "loss": 0.0097, "step": 19641 }, { "epoch": 7.74, "learning_rate": 5.1134356218980854e-08, "loss": 0.0106, "step": 19642 }, { "epoch": 7.74, "learning_rate": 5.1117306071054224e-08, "loss": 0.0092, "step": 19643 }, { "epoch": 7.74, "learning_rate": 5.110025834968934e-08, "loss": 0.0108, "step": 19644 }, { "epoch": 7.74, "learning_rate": 5.1083213055163874e-08, "loss": 0.0119, "step": 19645 }, { "epoch": 7.74, "learning_rate": 5.106617018775579e-08, "loss": 0.0104, "step": 19646 }, { "epoch": 7.74, "learning_rate": 5.1049129747742694e-08, "loss": 0.0116, "step": 19647 }, { "epoch": 7.74, "learning_rate": 5.103209173540242e-08, "loss": 0.0108, "step": 19648 }, { "epoch": 7.74, "learning_rate": 5.101505615101265e-08, "loss": 0.0083, "step": 19649 }, { "epoch": 7.74, "learning_rate": 5.0998022994851074e-08, "loss": 0.0113, "step": 19650 }, { "epoch": 7.74, "learning_rate": 5.098099226719521e-08, "loss": 0.0114, "step": 19651 }, { "epoch": 7.75, "learning_rate": 5.0963963968322696e-08, "loss": 0.0094, "step": 19652 }, { "epoch": 7.75, "learning_rate": 5.0946938098511136e-08, "loss": 0.0103, "step": 19653 }, { "epoch": 7.75, "learning_rate": 5.0929914658037866e-08, "loss": 0.0088, "step": 19654 }, { "epoch": 7.75, "learning_rate": 5.0912893647180496e-08, "loss": 0.0105, "step": 19655 }, { "epoch": 7.75, "learning_rate": 5.089587506621626e-08, "loss": 0.0119, "step": 19656 }, { "epoch": 7.75, "learning_rate": 5.087885891542276e-08, "loss": 0.0095, "step": 19657 }, { "epoch": 7.75, "learning_rate": 5.086184519507717e-08, "loss": 0.011, "step": 19658 }, { "epoch": 7.75, "learning_rate": 5.0844833905456885e-08, "loss": 0.0128, "step": 19659 }, { "epoch": 7.75, "learning_rate": 5.0827825046839067e-08, "loss": 0.0108, "step": 19660 }, { "epoch": 7.75, "learning_rate": 5.0810818619500984e-08, "loss": 0.0121, "step": 19661 }, { "epoch": 7.75, "learning_rate": 5.079381462371981e-08, "loss": 0.0112, "step": 19662 }, { "epoch": 7.75, "learning_rate": 5.077681305977275e-08, "loss": 0.0104, "step": 19663 }, { "epoch": 7.75, "learning_rate": 5.07598139279368e-08, "loss": 0.0134, "step": 19664 }, { "epoch": 7.75, "learning_rate": 5.0742817228489054e-08, "loss": 0.0124, "step": 19665 }, { "epoch": 7.75, "learning_rate": 5.0725822961706534e-08, "loss": 0.0124, "step": 19666 }, { "epoch": 7.75, "learning_rate": 5.0708831127866304e-08, "loss": 0.0117, "step": 19667 }, { "epoch": 7.75, "learning_rate": 5.0691841727245165e-08, "loss": 0.0113, "step": 19668 }, { "epoch": 7.75, "learning_rate": 5.067485476012014e-08, "loss": 0.0115, "step": 19669 }, { "epoch": 7.75, "learning_rate": 5.065787022676797e-08, "loss": 0.0096, "step": 19670 }, { "epoch": 7.75, "learning_rate": 5.064088812746556e-08, "loss": 0.008, "step": 19671 }, { "epoch": 7.75, "learning_rate": 5.062390846248965e-08, "loss": 0.0099, "step": 19672 }, { "epoch": 7.75, "learning_rate": 5.0606931232117076e-08, "loss": 0.0126, "step": 19673 }, { "epoch": 7.75, "learning_rate": 5.0589956436624425e-08, "loss": 0.0134, "step": 19674 }, { "epoch": 7.75, "learning_rate": 5.0572984076288407e-08, "loss": 0.0145, "step": 19675 }, { "epoch": 7.75, "learning_rate": 5.055601415138568e-08, "loss": 0.0079, "step": 19676 }, { "epoch": 7.75, "learning_rate": 5.053904666219275e-08, "loss": 0.0115, "step": 19677 }, { "epoch": 7.76, "learning_rate": 5.052208160898626e-08, "loss": 0.0117, "step": 19678 }, { "epoch": 7.76, "learning_rate": 5.050511899204255e-08, "loss": 0.0117, "step": 19679 }, { "epoch": 7.76, "learning_rate": 5.0488158811638305e-08, "loss": 0.0092, "step": 19680 }, { "epoch": 7.76, "learning_rate": 5.0471201068049784e-08, "loss": 0.0118, "step": 19681 }, { "epoch": 7.76, "learning_rate": 5.0454245761553484e-08, "loss": 0.0135, "step": 19682 }, { "epoch": 7.76, "learning_rate": 5.043729289242557e-08, "loss": 0.0115, "step": 19683 }, { "epoch": 7.76, "learning_rate": 5.04203424609426e-08, "loss": 0.0088, "step": 19684 }, { "epoch": 7.76, "learning_rate": 5.040339446738067e-08, "loss": 0.0109, "step": 19685 }, { "epoch": 7.76, "learning_rate": 5.038644891201607e-08, "loss": 0.0123, "step": 19686 }, { "epoch": 7.76, "learning_rate": 5.036950579512487e-08, "loss": 0.0106, "step": 19687 }, { "epoch": 7.76, "learning_rate": 5.035256511698345e-08, "loss": 0.0142, "step": 19688 }, { "epoch": 7.76, "learning_rate": 5.03356268778677e-08, "loss": 0.0096, "step": 19689 }, { "epoch": 7.76, "learning_rate": 5.031869107805382e-08, "loss": 0.0148, "step": 19690 }, { "epoch": 7.76, "learning_rate": 5.030175771781773e-08, "loss": 0.0099, "step": 19691 }, { "epoch": 7.76, "learning_rate": 5.028482679743549e-08, "loss": 0.0109, "step": 19692 }, { "epoch": 7.76, "learning_rate": 5.026789831718301e-08, "loss": 0.0103, "step": 19693 }, { "epoch": 7.76, "learning_rate": 5.0250972277336304e-08, "loss": 0.0098, "step": 19694 }, { "epoch": 7.76, "learning_rate": 5.023404867817109e-08, "loss": 0.0099, "step": 19695 }, { "epoch": 7.76, "learning_rate": 5.0217127519963255e-08, "loss": 0.0104, "step": 19696 }, { "epoch": 7.76, "learning_rate": 5.0200208802988676e-08, "loss": 0.0095, "step": 19697 }, { "epoch": 7.76, "learning_rate": 5.0183292527522935e-08, "loss": 0.0102, "step": 19698 }, { "epoch": 7.76, "learning_rate": 5.0166378693841926e-08, "loss": 0.0092, "step": 19699 }, { "epoch": 7.76, "learning_rate": 5.014946730222114e-08, "loss": 0.0086, "step": 19700 }, { "epoch": 7.76, "learning_rate": 5.013255835293628e-08, "loss": 0.0113, "step": 19701 }, { "epoch": 7.76, "learning_rate": 5.011565184626298e-08, "loss": 0.0121, "step": 19702 }, { "epoch": 7.77, "learning_rate": 5.00987477824768e-08, "loss": 0.011, "step": 19703 }, { "epoch": 7.77, "learning_rate": 5.0081846161853135e-08, "loss": 0.0084, "step": 19704 }, { "epoch": 7.77, "learning_rate": 5.0064946984667544e-08, "loss": 0.0117, "step": 19705 }, { "epoch": 7.77, "learning_rate": 5.004805025119545e-08, "loss": 0.0109, "step": 19706 }, { "epoch": 7.77, "learning_rate": 5.003115596171228e-08, "loss": 0.0132, "step": 19707 }, { "epoch": 7.77, "learning_rate": 5.0014264116493274e-08, "loss": 0.0128, "step": 19708 }, { "epoch": 7.77, "learning_rate": 4.9997374715813886e-08, "loss": 0.013, "step": 19709 }, { "epoch": 7.77, "learning_rate": 4.998048775994919e-08, "loss": 0.0109, "step": 19710 }, { "epoch": 7.77, "learning_rate": 4.996360324917466e-08, "loss": 0.012, "step": 19711 }, { "epoch": 7.77, "learning_rate": 4.9946721183765304e-08, "loss": 0.0105, "step": 19712 }, { "epoch": 7.77, "learning_rate": 4.99298415639964e-08, "loss": 0.0103, "step": 19713 }, { "epoch": 7.77, "learning_rate": 4.991296439014294e-08, "loss": 0.0126, "step": 19714 }, { "epoch": 7.77, "learning_rate": 4.989608966248007e-08, "loss": 0.013, "step": 19715 }, { "epoch": 7.77, "learning_rate": 4.987921738128278e-08, "loss": 0.0124, "step": 19716 }, { "epoch": 7.77, "learning_rate": 4.986234754682617e-08, "loss": 0.0089, "step": 19717 }, { "epoch": 7.77, "learning_rate": 4.984548015938503e-08, "loss": 0.009, "step": 19718 }, { "epoch": 7.77, "learning_rate": 4.982861521923438e-08, "loss": 0.0095, "step": 19719 }, { "epoch": 7.77, "learning_rate": 4.98117527266491e-08, "loss": 0.0116, "step": 19720 }, { "epoch": 7.77, "learning_rate": 4.9794892681903956e-08, "loss": 0.0101, "step": 19721 }, { "epoch": 7.77, "learning_rate": 4.977803508527383e-08, "loss": 0.0116, "step": 19722 }, { "epoch": 7.77, "learning_rate": 4.976117993703328e-08, "loss": 0.0095, "step": 19723 }, { "epoch": 7.77, "learning_rate": 4.9744327237457316e-08, "loss": 0.0132, "step": 19724 }, { "epoch": 7.77, "learning_rate": 4.972747698682036e-08, "loss": 0.0112, "step": 19725 }, { "epoch": 7.77, "learning_rate": 4.971062918539723e-08, "loss": 0.011, "step": 19726 }, { "epoch": 7.77, "learning_rate": 4.96937838334623e-08, "loss": 0.0091, "step": 19727 }, { "epoch": 7.77, "learning_rate": 4.9676940931290355e-08, "loss": 0.0117, "step": 19728 }, { "epoch": 7.78, "learning_rate": 4.9660100479155786e-08, "loss": 0.0103, "step": 19729 }, { "epoch": 7.78, "learning_rate": 4.964326247733312e-08, "loss": 0.013, "step": 19730 }, { "epoch": 7.78, "learning_rate": 4.962642692609671e-08, "loss": 0.0104, "step": 19731 }, { "epoch": 7.78, "learning_rate": 4.960959382572101e-08, "loss": 0.0127, "step": 19732 }, { "epoch": 7.78, "learning_rate": 4.9592763176480334e-08, "loss": 0.0105, "step": 19733 }, { "epoch": 7.78, "learning_rate": 4.9575934978649114e-08, "loss": 0.0103, "step": 19734 }, { "epoch": 7.78, "learning_rate": 4.9559109232501464e-08, "loss": 0.0099, "step": 19735 }, { "epoch": 7.78, "learning_rate": 4.954228593831169e-08, "loss": 0.0126, "step": 19736 }, { "epoch": 7.78, "learning_rate": 4.952546509635405e-08, "loss": 0.0095, "step": 19737 }, { "epoch": 7.78, "learning_rate": 4.950864670690256e-08, "loss": 0.0121, "step": 19738 }, { "epoch": 7.78, "learning_rate": 4.949183077023143e-08, "loss": 0.0105, "step": 19739 }, { "epoch": 7.78, "learning_rate": 4.947501728661476e-08, "loss": 0.0097, "step": 19740 }, { "epoch": 7.78, "learning_rate": 4.945820625632646e-08, "loss": 0.0092, "step": 19741 }, { "epoch": 7.78, "learning_rate": 4.944139767964064e-08, "loss": 0.0099, "step": 19742 }, { "epoch": 7.78, "learning_rate": 4.9424591556831253e-08, "loss": 0.0109, "step": 19743 }, { "epoch": 7.78, "learning_rate": 4.94077878881721e-08, "loss": 0.0101, "step": 19744 }, { "epoch": 7.78, "learning_rate": 4.939098667393715e-08, "loss": 0.0105, "step": 19745 }, { "epoch": 7.78, "learning_rate": 4.937418791440022e-08, "loss": 0.0093, "step": 19746 }, { "epoch": 7.78, "learning_rate": 4.9357391609835167e-08, "loss": 0.0127, "step": 19747 }, { "epoch": 7.78, "learning_rate": 4.93405977605156e-08, "loss": 0.009, "step": 19748 }, { "epoch": 7.78, "learning_rate": 4.932380636671538e-08, "loss": 0.0099, "step": 19749 }, { "epoch": 7.78, "learning_rate": 4.9307017428708e-08, "loss": 0.0096, "step": 19750 }, { "epoch": 7.78, "learning_rate": 4.929023094676732e-08, "loss": 0.011, "step": 19751 }, { "epoch": 7.78, "learning_rate": 4.927344692116677e-08, "loss": 0.011, "step": 19752 }, { "epoch": 7.78, "learning_rate": 4.9256665352180025e-08, "loss": 0.0087, "step": 19753 }, { "epoch": 7.79, "learning_rate": 4.92398862400804e-08, "loss": 0.0107, "step": 19754 }, { "epoch": 7.79, "learning_rate": 4.922310958514162e-08, "loss": 0.013, "step": 19755 }, { "epoch": 7.79, "learning_rate": 4.920633538763694e-08, "loss": 0.0109, "step": 19756 }, { "epoch": 7.79, "learning_rate": 4.9189563647839884e-08, "loss": 0.0122, "step": 19757 }, { "epoch": 7.79, "learning_rate": 4.917279436602366e-08, "loss": 0.0115, "step": 19758 }, { "epoch": 7.79, "learning_rate": 4.915602754246169e-08, "loss": 0.0106, "step": 19759 }, { "epoch": 7.79, "learning_rate": 4.9139263177427226e-08, "loss": 0.0086, "step": 19760 }, { "epoch": 7.79, "learning_rate": 4.912250127119346e-08, "loss": 0.0123, "step": 19761 }, { "epoch": 7.79, "learning_rate": 4.910574182403362e-08, "loss": 0.0123, "step": 19762 }, { "epoch": 7.79, "learning_rate": 4.9088984836220835e-08, "loss": 0.0089, "step": 19763 }, { "epoch": 7.79, "learning_rate": 4.907223030802832e-08, "loss": 0.0118, "step": 19764 }, { "epoch": 7.79, "learning_rate": 4.9055478239729e-08, "loss": 0.0096, "step": 19765 }, { "epoch": 7.79, "learning_rate": 4.903872863159601e-08, "loss": 0.0105, "step": 19766 }, { "epoch": 7.79, "learning_rate": 4.902198148390222e-08, "loss": 0.0096, "step": 19767 }, { "epoch": 7.79, "learning_rate": 4.900523679692079e-08, "loss": 0.0109, "step": 19768 }, { "epoch": 7.79, "learning_rate": 4.8988494570924446e-08, "loss": 0.0113, "step": 19769 }, { "epoch": 7.79, "learning_rate": 4.8971754806186184e-08, "loss": 0.0121, "step": 19770 }, { "epoch": 7.79, "learning_rate": 4.8955017502978725e-08, "loss": 0.0097, "step": 19771 }, { "epoch": 7.79, "learning_rate": 4.893828266157491e-08, "loss": 0.0109, "step": 19772 }, { "epoch": 7.79, "learning_rate": 4.892155028224754e-08, "loss": 0.0094, "step": 19773 }, { "epoch": 7.79, "learning_rate": 4.890482036526931e-08, "loss": 0.0096, "step": 19774 }, { "epoch": 7.79, "learning_rate": 4.888809291091283e-08, "loss": 0.0109, "step": 19775 }, { "epoch": 7.79, "learning_rate": 4.8871367919450746e-08, "loss": 0.0103, "step": 19776 }, { "epoch": 7.79, "learning_rate": 4.885464539115569e-08, "loss": 0.0092, "step": 19777 }, { "epoch": 7.79, "learning_rate": 4.883792532630027e-08, "loss": 0.012, "step": 19778 }, { "epoch": 7.8, "learning_rate": 4.882120772515685e-08, "loss": 0.0132, "step": 19779 }, { "epoch": 7.8, "learning_rate": 4.880449258799806e-08, "loss": 0.0113, "step": 19780 }, { "epoch": 7.8, "learning_rate": 4.878777991509617e-08, "loss": 0.0131, "step": 19781 }, { "epoch": 7.8, "learning_rate": 4.8771069706723666e-08, "loss": 0.0101, "step": 19782 }, { "epoch": 7.8, "learning_rate": 4.8754361963152926e-08, "loss": 0.0117, "step": 19783 }, { "epoch": 7.8, "learning_rate": 4.873765668465617e-08, "loss": 0.009, "step": 19784 }, { "epoch": 7.8, "learning_rate": 4.872095387150572e-08, "loss": 0.0115, "step": 19785 }, { "epoch": 7.8, "learning_rate": 4.87042535239738e-08, "loss": 0.0103, "step": 19786 }, { "epoch": 7.8, "learning_rate": 4.8687555642332665e-08, "loss": 0.0086, "step": 19787 }, { "epoch": 7.8, "learning_rate": 4.867086022685434e-08, "loss": 0.0088, "step": 19788 }, { "epoch": 7.8, "learning_rate": 4.865416727781107e-08, "loss": 0.0095, "step": 19789 }, { "epoch": 7.8, "learning_rate": 4.863747679547474e-08, "loss": 0.0103, "step": 19790 }, { "epoch": 7.8, "learning_rate": 4.86207887801176e-08, "loss": 0.0098, "step": 19791 }, { "epoch": 7.8, "learning_rate": 4.8604103232011476e-08, "loss": 0.0115, "step": 19792 }, { "epoch": 7.8, "learning_rate": 4.8587420151428426e-08, "loss": 0.0119, "step": 19793 }, { "epoch": 7.8, "learning_rate": 4.8570739538640194e-08, "loss": 0.0122, "step": 19794 }, { "epoch": 7.8, "learning_rate": 4.855406139391889e-08, "loss": 0.0112, "step": 19795 }, { "epoch": 7.8, "learning_rate": 4.853738571753614e-08, "loss": 0.0133, "step": 19796 }, { "epoch": 7.8, "learning_rate": 4.8520712509763885e-08, "loss": 0.0139, "step": 19797 }, { "epoch": 7.8, "learning_rate": 4.8504041770873725e-08, "loss": 0.0112, "step": 19798 }, { "epoch": 7.8, "learning_rate": 4.848737350113743e-08, "loss": 0.011, "step": 19799 }, { "epoch": 7.8, "learning_rate": 4.847070770082668e-08, "loss": 0.0111, "step": 19800 }, { "epoch": 7.8, "learning_rate": 4.8454044370213146e-08, "loss": 0.0099, "step": 19801 }, { "epoch": 7.8, "learning_rate": 4.8437383509568316e-08, "loss": 0.0099, "step": 19802 }, { "epoch": 7.8, "learning_rate": 4.8420725119163794e-08, "loss": 0.0129, "step": 19803 }, { "epoch": 7.8, "learning_rate": 4.840406919927111e-08, "loss": 0.0106, "step": 19804 }, { "epoch": 7.81, "learning_rate": 4.838741575016168e-08, "loss": 0.0101, "step": 19805 }, { "epoch": 7.81, "learning_rate": 4.8370764772106975e-08, "loss": 0.0112, "step": 19806 }, { "epoch": 7.81, "learning_rate": 4.835411626537824e-08, "loss": 0.0092, "step": 19807 }, { "epoch": 7.81, "learning_rate": 4.833747023024708e-08, "loss": 0.0118, "step": 19808 }, { "epoch": 7.81, "learning_rate": 4.832082666698456e-08, "loss": 0.0107, "step": 19809 }, { "epoch": 7.81, "learning_rate": 4.830418557586211e-08, "loss": 0.0097, "step": 19810 }, { "epoch": 7.81, "learning_rate": 4.8287546957150846e-08, "loss": 0.0089, "step": 19811 }, { "epoch": 7.81, "learning_rate": 4.8270910811121953e-08, "loss": 0.0103, "step": 19812 }, { "epoch": 7.81, "learning_rate": 4.825427713804664e-08, "loss": 0.0155, "step": 19813 }, { "epoch": 7.81, "learning_rate": 4.8237645938196034e-08, "loss": 0.0103, "step": 19814 }, { "epoch": 7.81, "learning_rate": 4.822101721184109e-08, "loss": 0.0089, "step": 19815 }, { "epoch": 7.81, "learning_rate": 4.8204390959252886e-08, "loss": 0.0121, "step": 19816 }, { "epoch": 7.81, "learning_rate": 4.818776718070241e-08, "loss": 0.0146, "step": 19817 }, { "epoch": 7.81, "learning_rate": 4.817114587646065e-08, "loss": 0.0108, "step": 19818 }, { "epoch": 7.81, "learning_rate": 4.815452704679841e-08, "loss": 0.0111, "step": 19819 }, { "epoch": 7.81, "learning_rate": 4.813791069198667e-08, "loss": 0.0109, "step": 19820 }, { "epoch": 7.81, "learning_rate": 4.812129681229604e-08, "loss": 0.0121, "step": 19821 }, { "epoch": 7.81, "learning_rate": 4.810468540799757e-08, "loss": 0.0105, "step": 19822 }, { "epoch": 7.81, "learning_rate": 4.808807647936182e-08, "loss": 0.0134, "step": 19823 }, { "epoch": 7.81, "learning_rate": 4.8071470026659586e-08, "loss": 0.0117, "step": 19824 }, { "epoch": 7.81, "learning_rate": 4.805486605016145e-08, "loss": 0.0118, "step": 19825 }, { "epoch": 7.81, "learning_rate": 4.803826455013806e-08, "loss": 0.014, "step": 19826 }, { "epoch": 7.81, "learning_rate": 4.8021665526860066e-08, "loss": 0.0104, "step": 19827 }, { "epoch": 7.81, "learning_rate": 4.800506898059786e-08, "loss": 0.0103, "step": 19828 }, { "epoch": 7.81, "learning_rate": 4.79884749116221e-08, "loss": 0.0108, "step": 19829 }, { "epoch": 7.82, "learning_rate": 4.797188332020306e-08, "loss": 0.0129, "step": 19830 }, { "epoch": 7.82, "learning_rate": 4.7955294206611376e-08, "loss": 0.0091, "step": 19831 }, { "epoch": 7.82, "learning_rate": 4.7938707571117245e-08, "loss": 0.0133, "step": 19832 }, { "epoch": 7.82, "learning_rate": 4.792212341399116e-08, "loss": 0.0145, "step": 19833 }, { "epoch": 7.82, "learning_rate": 4.790554173550319e-08, "loss": 0.01, "step": 19834 }, { "epoch": 7.82, "learning_rate": 4.788896253592385e-08, "loss": 0.0116, "step": 19835 }, { "epoch": 7.82, "learning_rate": 4.7872385815523196e-08, "loss": 0.0078, "step": 19836 }, { "epoch": 7.82, "learning_rate": 4.785581157457148e-08, "loss": 0.0123, "step": 19837 }, { "epoch": 7.82, "learning_rate": 4.783923981333877e-08, "loss": 0.0087, "step": 19838 }, { "epoch": 7.82, "learning_rate": 4.7822670532095164e-08, "loss": 0.0114, "step": 19839 }, { "epoch": 7.82, "learning_rate": 4.780610373111078e-08, "loss": 0.01, "step": 19840 }, { "epoch": 7.82, "learning_rate": 4.778953941065564e-08, "loss": 0.0148, "step": 19841 }, { "epoch": 7.82, "learning_rate": 4.7772977570999603e-08, "loss": 0.0135, "step": 19842 }, { "epoch": 7.82, "learning_rate": 4.77564182124127e-08, "loss": 0.011, "step": 19843 }, { "epoch": 7.82, "learning_rate": 4.7739861335164765e-08, "loss": 0.0096, "step": 19844 }, { "epoch": 7.82, "learning_rate": 4.772330693952578e-08, "loss": 0.0114, "step": 19845 }, { "epoch": 7.82, "learning_rate": 4.7706755025765356e-08, "loss": 0.0105, "step": 19846 }, { "epoch": 7.82, "learning_rate": 4.76902055941534e-08, "loss": 0.0143, "step": 19847 }, { "epoch": 7.82, "learning_rate": 4.767365864495965e-08, "loss": 0.0107, "step": 19848 }, { "epoch": 7.82, "learning_rate": 4.7657114178453706e-08, "loss": 0.0153, "step": 19849 }, { "epoch": 7.82, "learning_rate": 4.764057219490532e-08, "loss": 0.0138, "step": 19850 }, { "epoch": 7.82, "learning_rate": 4.762403269458397e-08, "loss": 0.0086, "step": 19851 }, { "epoch": 7.82, "learning_rate": 4.760749567775932e-08, "loss": 0.0096, "step": 19852 }, { "epoch": 7.82, "learning_rate": 4.759096114470089e-08, "loss": 0.0091, "step": 19853 }, { "epoch": 7.82, "learning_rate": 4.7574429095678196e-08, "loss": 0.011, "step": 19854 }, { "epoch": 7.83, "learning_rate": 4.755789953096061e-08, "loss": 0.011, "step": 19855 }, { "epoch": 7.83, "learning_rate": 4.754137245081757e-08, "loss": 0.0096, "step": 19856 }, { "epoch": 7.83, "learning_rate": 4.752484785551845e-08, "loss": 0.0094, "step": 19857 }, { "epoch": 7.83, "learning_rate": 4.750832574533261e-08, "loss": 0.0118, "step": 19858 }, { "epoch": 7.83, "learning_rate": 4.749180612052928e-08, "loss": 0.0103, "step": 19859 }, { "epoch": 7.83, "learning_rate": 4.747528898137774e-08, "loss": 0.0108, "step": 19860 }, { "epoch": 7.83, "learning_rate": 4.745877432814709e-08, "loss": 0.0097, "step": 19861 }, { "epoch": 7.83, "learning_rate": 4.744226216110669e-08, "loss": 0.0121, "step": 19862 }, { "epoch": 7.83, "learning_rate": 4.7425752480525514e-08, "loss": 0.0137, "step": 19863 }, { "epoch": 7.83, "learning_rate": 4.7409245286672716e-08, "loss": 0.0135, "step": 19864 }, { "epoch": 7.83, "learning_rate": 4.739274057981728e-08, "loss": 0.0092, "step": 19865 }, { "epoch": 7.83, "learning_rate": 4.737623836022821e-08, "loss": 0.009, "step": 19866 }, { "epoch": 7.83, "learning_rate": 4.735973862817451e-08, "loss": 0.013, "step": 19867 }, { "epoch": 7.83, "learning_rate": 4.734324138392514e-08, "loss": 0.012, "step": 19868 }, { "epoch": 7.83, "learning_rate": 4.732674662774887e-08, "loss": 0.0113, "step": 19869 }, { "epoch": 7.83, "learning_rate": 4.731025435991458e-08, "loss": 0.0132, "step": 19870 }, { "epoch": 7.83, "learning_rate": 4.7293764580691144e-08, "loss": 0.0114, "step": 19871 }, { "epoch": 7.83, "learning_rate": 4.7277277290347204e-08, "loss": 0.0102, "step": 19872 }, { "epoch": 7.83, "learning_rate": 4.72607924891516e-08, "loss": 0.0112, "step": 19873 }, { "epoch": 7.83, "learning_rate": 4.7244310177372804e-08, "loss": 0.0094, "step": 19874 }, { "epoch": 7.83, "learning_rate": 4.722783035527973e-08, "loss": 0.0083, "step": 19875 }, { "epoch": 7.83, "learning_rate": 4.721135302314076e-08, "loss": 0.0131, "step": 19876 }, { "epoch": 7.83, "learning_rate": 4.7194878181224586e-08, "loss": 0.0102, "step": 19877 }, { "epoch": 7.83, "learning_rate": 4.717840582979959e-08, "loss": 0.0137, "step": 19878 }, { "epoch": 7.83, "learning_rate": 4.716193596913434e-08, "loss": 0.0106, "step": 19879 }, { "epoch": 7.83, "learning_rate": 4.714546859949724e-08, "loss": 0.012, "step": 19880 }, { "epoch": 7.84, "learning_rate": 4.712900372115673e-08, "loss": 0.0105, "step": 19881 }, { "epoch": 7.84, "learning_rate": 4.7112541334381095e-08, "loss": 0.0121, "step": 19882 }, { "epoch": 7.84, "learning_rate": 4.709608143943864e-08, "loss": 0.0128, "step": 19883 }, { "epoch": 7.84, "learning_rate": 4.707962403659769e-08, "loss": 0.0119, "step": 19884 }, { "epoch": 7.84, "learning_rate": 4.7063169126126515e-08, "loss": 0.0118, "step": 19885 }, { "epoch": 7.84, "learning_rate": 4.7046716708293185e-08, "loss": 0.0101, "step": 19886 }, { "epoch": 7.84, "learning_rate": 4.7030266783365905e-08, "loss": 0.0096, "step": 19887 }, { "epoch": 7.84, "learning_rate": 4.701381935161286e-08, "loss": 0.0112, "step": 19888 }, { "epoch": 7.84, "learning_rate": 4.6997374413301985e-08, "loss": 0.0146, "step": 19889 }, { "epoch": 7.84, "learning_rate": 4.6980931968701365e-08, "loss": 0.0112, "step": 19890 }, { "epoch": 7.84, "learning_rate": 4.696449201807904e-08, "loss": 0.0117, "step": 19891 }, { "epoch": 7.84, "learning_rate": 4.694805456170288e-08, "loss": 0.0107, "step": 19892 }, { "epoch": 7.84, "learning_rate": 4.6931619599840775e-08, "loss": 0.0108, "step": 19893 }, { "epoch": 7.84, "learning_rate": 4.6915187132760725e-08, "loss": 0.0108, "step": 19894 }, { "epoch": 7.84, "learning_rate": 4.68987571607304e-08, "loss": 0.0121, "step": 19895 }, { "epoch": 7.84, "learning_rate": 4.688232968401764e-08, "loss": 0.0115, "step": 19896 }, { "epoch": 7.84, "learning_rate": 4.68659047028902e-08, "loss": 0.0111, "step": 19897 }, { "epoch": 7.84, "learning_rate": 4.684948221761582e-08, "loss": 0.0113, "step": 19898 }, { "epoch": 7.84, "learning_rate": 4.6833062228462084e-08, "loss": 0.014, "step": 19899 }, { "epoch": 7.84, "learning_rate": 4.681664473569669e-08, "loss": 0.0127, "step": 19900 }, { "epoch": 7.84, "learning_rate": 4.680022973958706e-08, "loss": 0.0115, "step": 19901 }, { "epoch": 7.84, "learning_rate": 4.678381724040095e-08, "loss": 0.0081, "step": 19902 }, { "epoch": 7.84, "learning_rate": 4.6767407238405707e-08, "loss": 0.0119, "step": 19903 }, { "epoch": 7.84, "learning_rate": 4.675099973386891e-08, "loss": 0.012, "step": 19904 }, { "epoch": 7.84, "learning_rate": 4.673459472705779e-08, "loss": 0.0109, "step": 19905 }, { "epoch": 7.85, "learning_rate": 4.6718192218239935e-08, "loss": 0.0112, "step": 19906 }, { "epoch": 7.85, "learning_rate": 4.6701792207682556e-08, "loss": 0.0104, "step": 19907 }, { "epoch": 7.85, "learning_rate": 4.668539469565302e-08, "loss": 0.0088, "step": 19908 }, { "epoch": 7.85, "learning_rate": 4.666899968241849e-08, "loss": 0.0113, "step": 19909 }, { "epoch": 7.85, "learning_rate": 4.665260716824622e-08, "loss": 0.0118, "step": 19910 }, { "epoch": 7.85, "learning_rate": 4.6636217153403445e-08, "loss": 0.0088, "step": 19911 }, { "epoch": 7.85, "learning_rate": 4.661982963815718e-08, "loss": 0.0124, "step": 19912 }, { "epoch": 7.85, "learning_rate": 4.6603444622774615e-08, "loss": 0.0117, "step": 19913 }, { "epoch": 7.85, "learning_rate": 4.6587062107522725e-08, "loss": 0.0091, "step": 19914 }, { "epoch": 7.85, "learning_rate": 4.657068209266866e-08, "loss": 0.0099, "step": 19915 }, { "epoch": 7.85, "learning_rate": 4.65543045784792e-08, "loss": 0.0093, "step": 19916 }, { "epoch": 7.85, "learning_rate": 4.653792956522142e-08, "loss": 0.0134, "step": 19917 }, { "epoch": 7.85, "learning_rate": 4.6521557053162116e-08, "loss": 0.0091, "step": 19918 }, { "epoch": 7.85, "learning_rate": 4.6505187042568185e-08, "loss": 0.0097, "step": 19919 }, { "epoch": 7.85, "learning_rate": 4.648881953370638e-08, "loss": 0.0103, "step": 19920 }, { "epoch": 7.85, "learning_rate": 4.647245452684359e-08, "loss": 0.0113, "step": 19921 }, { "epoch": 7.85, "learning_rate": 4.6456092022246406e-08, "loss": 0.0118, "step": 19922 }, { "epoch": 7.85, "learning_rate": 4.6439732020181544e-08, "loss": 0.0111, "step": 19923 }, { "epoch": 7.85, "learning_rate": 4.642337452091567e-08, "loss": 0.0089, "step": 19924 }, { "epoch": 7.85, "learning_rate": 4.640701952471544e-08, "loss": 0.0116, "step": 19925 }, { "epoch": 7.85, "learning_rate": 4.6390667031847305e-08, "loss": 0.0129, "step": 19926 }, { "epoch": 7.85, "learning_rate": 4.637431704257784e-08, "loss": 0.0109, "step": 19927 }, { "epoch": 7.85, "learning_rate": 4.6357969557173505e-08, "loss": 0.0122, "step": 19928 }, { "epoch": 7.85, "learning_rate": 4.634162457590083e-08, "loss": 0.0119, "step": 19929 }, { "epoch": 7.85, "learning_rate": 4.632528209902606e-08, "loss": 0.0122, "step": 19930 }, { "epoch": 7.85, "learning_rate": 4.630894212681573e-08, "loss": 0.0101, "step": 19931 }, { "epoch": 7.86, "learning_rate": 4.629260465953597e-08, "loss": 0.0106, "step": 19932 }, { "epoch": 7.86, "learning_rate": 4.6276269697453176e-08, "loss": 0.0116, "step": 19933 }, { "epoch": 7.86, "learning_rate": 4.625993724083359e-08, "loss": 0.0085, "step": 19934 }, { "epoch": 7.86, "learning_rate": 4.62436072899433e-08, "loss": 0.009, "step": 19935 }, { "epoch": 7.86, "learning_rate": 4.622727984504855e-08, "loss": 0.0109, "step": 19936 }, { "epoch": 7.86, "learning_rate": 4.6210954906415424e-08, "loss": 0.0105, "step": 19937 }, { "epoch": 7.86, "learning_rate": 4.619463247431006e-08, "loss": 0.0104, "step": 19938 }, { "epoch": 7.86, "learning_rate": 4.6178312548998355e-08, "loss": 0.0094, "step": 19939 }, { "epoch": 7.86, "learning_rate": 4.616199513074646e-08, "loss": 0.012, "step": 19940 }, { "epoch": 7.86, "learning_rate": 4.6145680219820103e-08, "loss": 0.0077, "step": 19941 }, { "epoch": 7.86, "learning_rate": 4.612936781648544e-08, "loss": 0.0113, "step": 19942 }, { "epoch": 7.86, "learning_rate": 4.611305792100817e-08, "loss": 0.0124, "step": 19943 }, { "epoch": 7.86, "learning_rate": 4.609675053365425e-08, "loss": 0.0117, "step": 19944 }, { "epoch": 7.86, "learning_rate": 4.608044565468925e-08, "loss": 0.0124, "step": 19945 }, { "epoch": 7.86, "learning_rate": 4.606414328437917e-08, "loss": 0.0089, "step": 19946 }, { "epoch": 7.86, "learning_rate": 4.6047843422989576e-08, "loss": 0.0129, "step": 19947 }, { "epoch": 7.86, "learning_rate": 4.6031546070786185e-08, "loss": 0.01, "step": 19948 }, { "epoch": 7.86, "learning_rate": 4.6015251228034533e-08, "loss": 0.0096, "step": 19949 }, { "epoch": 7.86, "learning_rate": 4.599895889500025e-08, "loss": 0.011, "step": 19950 }, { "epoch": 7.86, "learning_rate": 4.59826690719489e-08, "loss": 0.013, "step": 19951 }, { "epoch": 7.86, "learning_rate": 4.5966381759146e-08, "loss": 0.0107, "step": 19952 }, { "epoch": 7.86, "learning_rate": 4.5950096956856945e-08, "loss": 0.0147, "step": 19953 }, { "epoch": 7.86, "learning_rate": 4.593381466534716e-08, "loss": 0.0105, "step": 19954 }, { "epoch": 7.86, "learning_rate": 4.59175348848821e-08, "loss": 0.0094, "step": 19955 }, { "epoch": 7.86, "learning_rate": 4.5901257615726984e-08, "loss": 0.0097, "step": 19956 }, { "epoch": 7.87, "learning_rate": 4.5884982858147215e-08, "loss": 0.0115, "step": 19957 }, { "epoch": 7.87, "learning_rate": 4.5868710612407956e-08, "loss": 0.0125, "step": 19958 }, { "epoch": 7.87, "learning_rate": 4.585244087877447e-08, "loss": 0.0101, "step": 19959 }, { "epoch": 7.87, "learning_rate": 4.583617365751191e-08, "loss": 0.0139, "step": 19960 }, { "epoch": 7.87, "learning_rate": 4.5819908948885476e-08, "loss": 0.0097, "step": 19961 }, { "epoch": 7.87, "learning_rate": 4.580364675316013e-08, "loss": 0.0097, "step": 19962 }, { "epoch": 7.87, "learning_rate": 4.578738707060099e-08, "loss": 0.0114, "step": 19963 }, { "epoch": 7.87, "learning_rate": 4.5771129901473096e-08, "loss": 0.0113, "step": 19964 }, { "epoch": 7.87, "learning_rate": 4.5754875246041404e-08, "loss": 0.0116, "step": 19965 }, { "epoch": 7.87, "learning_rate": 4.5738623104570776e-08, "loss": 0.0094, "step": 19966 }, { "epoch": 7.87, "learning_rate": 4.572237347732615e-08, "loss": 0.0091, "step": 19967 }, { "epoch": 7.87, "learning_rate": 4.5706126364572364e-08, "loss": 0.011, "step": 19968 }, { "epoch": 7.87, "learning_rate": 4.5689881766574266e-08, "loss": 0.0106, "step": 19969 }, { "epoch": 7.87, "learning_rate": 4.56736396835965e-08, "loss": 0.0116, "step": 19970 }, { "epoch": 7.87, "learning_rate": 4.565740011590393e-08, "loss": 0.0104, "step": 19971 }, { "epoch": 7.87, "learning_rate": 4.564116306376105e-08, "loss": 0.0108, "step": 19972 }, { "epoch": 7.87, "learning_rate": 4.5624928527432736e-08, "loss": 0.0117, "step": 19973 }, { "epoch": 7.87, "learning_rate": 4.5608696507183394e-08, "loss": 0.0123, "step": 19974 }, { "epoch": 7.87, "learning_rate": 4.55924670032777e-08, "loss": 0.011, "step": 19975 }, { "epoch": 7.87, "learning_rate": 4.5576240015980085e-08, "loss": 0.0128, "step": 19976 }, { "epoch": 7.87, "learning_rate": 4.556001554555506e-08, "loss": 0.0114, "step": 19977 }, { "epoch": 7.87, "learning_rate": 4.55437935922671e-08, "loss": 0.0117, "step": 19978 }, { "epoch": 7.87, "learning_rate": 4.552757415638051e-08, "loss": 0.0109, "step": 19979 }, { "epoch": 7.87, "learning_rate": 4.5511357238159686e-08, "loss": 0.0152, "step": 19980 }, { "epoch": 7.87, "learning_rate": 4.549514283786895e-08, "loss": 0.0149, "step": 19981 }, { "epoch": 7.88, "learning_rate": 4.547893095577262e-08, "loss": 0.0147, "step": 19982 }, { "epoch": 7.88, "learning_rate": 4.546272159213482e-08, "loss": 0.0109, "step": 19983 }, { "epoch": 7.88, "learning_rate": 4.5446514747219836e-08, "loss": 0.0119, "step": 19984 }, { "epoch": 7.88, "learning_rate": 4.543031042129166e-08, "loss": 0.0108, "step": 19985 }, { "epoch": 7.88, "learning_rate": 4.541410861461461e-08, "loss": 0.0142, "step": 19986 }, { "epoch": 7.88, "learning_rate": 4.53979093274526e-08, "loss": 0.0103, "step": 19987 }, { "epoch": 7.88, "learning_rate": 4.538171256006975e-08, "loss": 0.0082, "step": 19988 }, { "epoch": 7.88, "learning_rate": 4.5365518312729947e-08, "loss": 0.0126, "step": 19989 }, { "epoch": 7.88, "learning_rate": 4.5349326585697167e-08, "loss": 0.0086, "step": 19990 }, { "epoch": 7.88, "learning_rate": 4.533313737923532e-08, "loss": 0.0096, "step": 19991 }, { "epoch": 7.88, "learning_rate": 4.5316950693608326e-08, "loss": 0.011, "step": 19992 }, { "epoch": 7.88, "learning_rate": 4.530076652907988e-08, "loss": 0.0125, "step": 19993 }, { "epoch": 7.88, "learning_rate": 4.528458488591385e-08, "loss": 0.0095, "step": 19994 }, { "epoch": 7.88, "learning_rate": 4.526840576437392e-08, "loss": 0.0104, "step": 19995 }, { "epoch": 7.88, "learning_rate": 4.5252229164723884e-08, "loss": 0.0104, "step": 19996 }, { "epoch": 7.88, "learning_rate": 4.523605508722725e-08, "loss": 0.0089, "step": 19997 }, { "epoch": 7.88, "learning_rate": 4.5219883532147784e-08, "loss": 0.0108, "step": 19998 }, { "epoch": 7.88, "learning_rate": 4.520371449974889e-08, "loss": 0.0096, "step": 19999 }, { "epoch": 7.88, "learning_rate": 4.518754799029421e-08, "loss": 0.0102, "step": 20000 }, { "epoch": 7.88, "learning_rate": 4.5171384004047254e-08, "loss": 0.0109, "step": 20001 }, { "epoch": 7.88, "learning_rate": 4.515522254127138e-08, "loss": 0.0126, "step": 20002 }, { "epoch": 7.88, "learning_rate": 4.513906360223003e-08, "loss": 0.0093, "step": 20003 }, { "epoch": 7.88, "learning_rate": 4.512290718718661e-08, "loss": 0.011, "step": 20004 }, { "epoch": 7.88, "learning_rate": 4.5106753296404455e-08, "loss": 0.0109, "step": 20005 }, { "epoch": 7.88, "learning_rate": 4.509060193014673e-08, "loss": 0.0107, "step": 20006 }, { "epoch": 7.88, "learning_rate": 4.507445308867686e-08, "loss": 0.0121, "step": 20007 }, { "epoch": 7.89, "learning_rate": 4.5058306772257796e-08, "loss": 0.0102, "step": 20008 }, { "epoch": 7.89, "learning_rate": 4.504216298115298e-08, "loss": 0.0099, "step": 20009 }, { "epoch": 7.89, "learning_rate": 4.502602171562533e-08, "loss": 0.0113, "step": 20010 }, { "epoch": 7.89, "learning_rate": 4.500988297593804e-08, "loss": 0.0106, "step": 20011 }, { "epoch": 7.89, "learning_rate": 4.4993746762354016e-08, "loss": 0.0098, "step": 20012 }, { "epoch": 7.89, "learning_rate": 4.4977613075136435e-08, "loss": 0.0107, "step": 20013 }, { "epoch": 7.89, "learning_rate": 4.4961481914548096e-08, "loss": 0.0102, "step": 20014 }, { "epoch": 7.89, "learning_rate": 4.4945353280852024e-08, "loss": 0.012, "step": 20015 }, { "epoch": 7.89, "learning_rate": 4.492922717431098e-08, "loss": 0.0133, "step": 20016 }, { "epoch": 7.89, "learning_rate": 4.4913103595187873e-08, "loss": 0.0087, "step": 20017 }, { "epoch": 7.89, "learning_rate": 4.489698254374545e-08, "loss": 0.0126, "step": 20018 }, { "epoch": 7.89, "learning_rate": 4.4880864020246555e-08, "loss": 0.01, "step": 20019 }, { "epoch": 7.89, "learning_rate": 4.486474802495375e-08, "loss": 0.0104, "step": 20020 }, { "epoch": 7.89, "learning_rate": 4.484863455812977e-08, "loss": 0.0108, "step": 20021 }, { "epoch": 7.89, "learning_rate": 4.4832523620037334e-08, "loss": 0.0098, "step": 20022 }, { "epoch": 7.89, "learning_rate": 4.481641521093884e-08, "loss": 0.0131, "step": 20023 }, { "epoch": 7.89, "learning_rate": 4.4800309331097e-08, "loss": 0.0092, "step": 20024 }, { "epoch": 7.89, "learning_rate": 4.4784205980774124e-08, "loss": 0.0102, "step": 20025 }, { "epoch": 7.89, "learning_rate": 4.476810516023293e-08, "loss": 0.0082, "step": 20026 }, { "epoch": 7.89, "learning_rate": 4.475200686973561e-08, "loss": 0.0122, "step": 20027 }, { "epoch": 7.89, "learning_rate": 4.4735911109544686e-08, "loss": 0.0111, "step": 20028 }, { "epoch": 7.89, "learning_rate": 4.471981787992238e-08, "loss": 0.0102, "step": 20029 }, { "epoch": 7.89, "learning_rate": 4.4703727181131044e-08, "loss": 0.0107, "step": 20030 }, { "epoch": 7.89, "learning_rate": 4.4687639013432954e-08, "loss": 0.0128, "step": 20031 }, { "epoch": 7.89, "learning_rate": 4.4671553377090344e-08, "loss": 0.0094, "step": 20032 }, { "epoch": 7.9, "learning_rate": 4.465547027236527e-08, "loss": 0.01, "step": 20033 }, { "epoch": 7.9, "learning_rate": 4.4639389699519944e-08, "loss": 0.0111, "step": 20034 }, { "epoch": 7.9, "learning_rate": 4.4623311658816454e-08, "loss": 0.0101, "step": 20035 }, { "epoch": 7.9, "learning_rate": 4.4607236150516894e-08, "loss": 0.0104, "step": 20036 }, { "epoch": 7.9, "learning_rate": 4.459116317488314e-08, "loss": 0.0103, "step": 20037 }, { "epoch": 7.9, "learning_rate": 4.45750927321773e-08, "loss": 0.0088, "step": 20038 }, { "epoch": 7.9, "learning_rate": 4.4559024822661185e-08, "loss": 0.0096, "step": 20039 }, { "epoch": 7.9, "learning_rate": 4.454295944659672e-08, "loss": 0.0082, "step": 20040 }, { "epoch": 7.9, "learning_rate": 4.4526896604245735e-08, "loss": 0.0126, "step": 20041 }, { "epoch": 7.9, "learning_rate": 4.451083629587009e-08, "loss": 0.0097, "step": 20042 }, { "epoch": 7.9, "learning_rate": 4.449477852173147e-08, "loss": 0.0098, "step": 20043 }, { "epoch": 7.9, "learning_rate": 4.447872328209159e-08, "loss": 0.0103, "step": 20044 }, { "epoch": 7.9, "learning_rate": 4.446267057721222e-08, "loss": 0.0121, "step": 20045 }, { "epoch": 7.9, "learning_rate": 4.44466204073549e-08, "loss": 0.011, "step": 20046 }, { "epoch": 7.9, "learning_rate": 4.443057277278128e-08, "loss": 0.0128, "step": 20047 }, { "epoch": 7.9, "learning_rate": 4.44145276737528e-08, "loss": 0.0134, "step": 20048 }, { "epoch": 7.9, "learning_rate": 4.4398485110531146e-08, "loss": 0.0098, "step": 20049 }, { "epoch": 7.9, "learning_rate": 4.438244508337765e-08, "loss": 0.0122, "step": 20050 }, { "epoch": 7.9, "learning_rate": 4.436640759255383e-08, "loss": 0.0101, "step": 20051 }, { "epoch": 7.9, "learning_rate": 4.4350372638320935e-08, "loss": 0.0119, "step": 20052 }, { "epoch": 7.9, "learning_rate": 4.433434022094051e-08, "loss": 0.0083, "step": 20053 }, { "epoch": 7.9, "learning_rate": 4.4318310340673706e-08, "loss": 0.0128, "step": 20054 }, { "epoch": 7.9, "learning_rate": 4.4302282997781895e-08, "loss": 0.0082, "step": 20055 }, { "epoch": 7.9, "learning_rate": 4.4286258192526115e-08, "loss": 0.0127, "step": 20056 }, { "epoch": 7.9, "learning_rate": 4.4270235925167764e-08, "loss": 0.0097, "step": 20057 }, { "epoch": 7.91, "learning_rate": 4.425421619596785e-08, "loss": 0.0112, "step": 20058 }, { "epoch": 7.91, "learning_rate": 4.423819900518757e-08, "loss": 0.0101, "step": 20059 }, { "epoch": 7.91, "learning_rate": 4.4222184353087844e-08, "loss": 0.0107, "step": 20060 }, { "epoch": 7.91, "learning_rate": 4.420617223992977e-08, "loss": 0.0103, "step": 20061 }, { "epoch": 7.91, "learning_rate": 4.4190162665974354e-08, "loss": 0.0108, "step": 20062 }, { "epoch": 7.91, "learning_rate": 4.4174155631482434e-08, "loss": 0.0105, "step": 20063 }, { "epoch": 7.91, "learning_rate": 4.415815113671495e-08, "loss": 0.0088, "step": 20064 }, { "epoch": 7.91, "learning_rate": 4.4142149181932755e-08, "loss": 0.0088, "step": 20065 }, { "epoch": 7.91, "learning_rate": 4.412614976739668e-08, "loss": 0.0117, "step": 20066 }, { "epoch": 7.91, "learning_rate": 4.4110152893367434e-08, "loss": 0.0126, "step": 20067 }, { "epoch": 7.91, "learning_rate": 4.409415856010581e-08, "loss": 0.0111, "step": 20068 }, { "epoch": 7.91, "learning_rate": 4.407816676787242e-08, "loss": 0.0125, "step": 20069 }, { "epoch": 7.91, "learning_rate": 4.406217751692791e-08, "loss": 0.0115, "step": 20070 }, { "epoch": 7.91, "learning_rate": 4.404619080753293e-08, "loss": 0.0098, "step": 20071 }, { "epoch": 7.91, "learning_rate": 4.403020663994805e-08, "loss": 0.0088, "step": 20072 }, { "epoch": 7.91, "learning_rate": 4.401422501443373e-08, "loss": 0.0107, "step": 20073 }, { "epoch": 7.91, "learning_rate": 4.399824593125045e-08, "loss": 0.0132, "step": 20074 }, { "epoch": 7.91, "learning_rate": 4.3982269390658675e-08, "loss": 0.009, "step": 20075 }, { "epoch": 7.91, "learning_rate": 4.396629539291883e-08, "loss": 0.0106, "step": 20076 }, { "epoch": 7.91, "learning_rate": 4.395032393829119e-08, "loss": 0.0106, "step": 20077 }, { "epoch": 7.91, "learning_rate": 4.393435502703615e-08, "loss": 0.0111, "step": 20078 }, { "epoch": 7.91, "learning_rate": 4.391838865941382e-08, "loss": 0.0116, "step": 20079 }, { "epoch": 7.91, "learning_rate": 4.3902424835684646e-08, "loss": 0.0107, "step": 20080 }, { "epoch": 7.91, "learning_rate": 4.3886463556108654e-08, "loss": 0.0143, "step": 20081 }, { "epoch": 7.91, "learning_rate": 4.3870504820946096e-08, "loss": 0.0118, "step": 20082 }, { "epoch": 7.91, "learning_rate": 4.3854548630456945e-08, "loss": 0.0103, "step": 20083 }, { "epoch": 7.92, "learning_rate": 4.3838594984901366e-08, "loss": 0.0106, "step": 20084 }, { "epoch": 7.92, "learning_rate": 4.3822643884539404e-08, "loss": 0.0122, "step": 20085 }, { "epoch": 7.92, "learning_rate": 4.380669532963091e-08, "loss": 0.012, "step": 20086 }, { "epoch": 7.92, "learning_rate": 4.379074932043591e-08, "loss": 0.0123, "step": 20087 }, { "epoch": 7.92, "learning_rate": 4.377480585721427e-08, "loss": 0.0097, "step": 20088 }, { "epoch": 7.92, "learning_rate": 4.375886494022594e-08, "loss": 0.0152, "step": 20089 }, { "epoch": 7.92, "learning_rate": 4.374292656973058e-08, "loss": 0.0117, "step": 20090 }, { "epoch": 7.92, "learning_rate": 4.372699074598809e-08, "loss": 0.01, "step": 20091 }, { "epoch": 7.92, "learning_rate": 4.371105746925803e-08, "loss": 0.0116, "step": 20092 }, { "epoch": 7.92, "learning_rate": 4.369512673980031e-08, "loss": 0.013, "step": 20093 }, { "epoch": 7.92, "learning_rate": 4.367919855787443e-08, "loss": 0.0116, "step": 20094 }, { "epoch": 7.92, "learning_rate": 4.366327292374008e-08, "loss": 0.0093, "step": 20095 }, { "epoch": 7.92, "learning_rate": 4.364734983765665e-08, "loss": 0.0102, "step": 20096 }, { "epoch": 7.92, "learning_rate": 4.3631429299883925e-08, "loss": 0.0142, "step": 20097 }, { "epoch": 7.92, "learning_rate": 4.3615511310681194e-08, "loss": 0.0136, "step": 20098 }, { "epoch": 7.92, "learning_rate": 4.359959587030799e-08, "loss": 0.0107, "step": 20099 }, { "epoch": 7.92, "learning_rate": 4.3583682979023616e-08, "loss": 0.0107, "step": 20100 }, { "epoch": 7.92, "learning_rate": 4.3567772637087484e-08, "loss": 0.0104, "step": 20101 }, { "epoch": 7.92, "learning_rate": 4.3551864844758923e-08, "loss": 0.0071, "step": 20102 }, { "epoch": 7.92, "learning_rate": 4.3535959602297233e-08, "loss": 0.0114, "step": 20103 }, { "epoch": 7.92, "learning_rate": 4.3520056909961566e-08, "loss": 0.013, "step": 20104 }, { "epoch": 7.92, "learning_rate": 4.350415676801113e-08, "loss": 0.0119, "step": 20105 }, { "epoch": 7.92, "learning_rate": 4.348825917670515e-08, "loss": 0.0086, "step": 20106 }, { "epoch": 7.92, "learning_rate": 4.347236413630264e-08, "loss": 0.0114, "step": 20107 }, { "epoch": 7.92, "learning_rate": 4.345647164706267e-08, "loss": 0.0106, "step": 20108 }, { "epoch": 7.93, "learning_rate": 4.3440581709244384e-08, "loss": 0.0096, "step": 20109 }, { "epoch": 7.93, "learning_rate": 4.342469432310658e-08, "loss": 0.01, "step": 20110 }, { "epoch": 7.93, "learning_rate": 4.340880948890833e-08, "loss": 0.0129, "step": 20111 }, { "epoch": 7.93, "learning_rate": 4.3392927206908527e-08, "loss": 0.011, "step": 20112 }, { "epoch": 7.93, "learning_rate": 4.337704747736592e-08, "loss": 0.0119, "step": 20113 }, { "epoch": 7.93, "learning_rate": 4.336117030053943e-08, "loss": 0.0093, "step": 20114 }, { "epoch": 7.93, "learning_rate": 4.3345295676687786e-08, "loss": 0.0102, "step": 20115 }, { "epoch": 7.93, "learning_rate": 4.332942360606982e-08, "loss": 0.0122, "step": 20116 }, { "epoch": 7.93, "learning_rate": 4.3313554088944035e-08, "loss": 0.0098, "step": 20117 }, { "epoch": 7.93, "learning_rate": 4.329768712556927e-08, "loss": 0.0106, "step": 20118 }, { "epoch": 7.93, "learning_rate": 4.328182271620393e-08, "loss": 0.0102, "step": 20119 }, { "epoch": 7.93, "learning_rate": 4.326596086110679e-08, "loss": 0.0121, "step": 20120 }, { "epoch": 7.93, "learning_rate": 4.325010156053623e-08, "loss": 0.0109, "step": 20121 }, { "epoch": 7.93, "learning_rate": 4.323424481475084e-08, "loss": 0.0099, "step": 20122 }, { "epoch": 7.93, "learning_rate": 4.3218390624008916e-08, "loss": 0.011, "step": 20123 }, { "epoch": 7.93, "learning_rate": 4.320253898856903e-08, "loss": 0.0121, "step": 20124 }, { "epoch": 7.93, "learning_rate": 4.318668990868939e-08, "loss": 0.0115, "step": 20125 }, { "epoch": 7.93, "learning_rate": 4.317084338462844e-08, "loss": 0.0103, "step": 20126 }, { "epoch": 7.93, "learning_rate": 4.315499941664436e-08, "loss": 0.01, "step": 20127 }, { "epoch": 7.93, "learning_rate": 4.313915800499537e-08, "loss": 0.0107, "step": 20128 }, { "epoch": 7.93, "learning_rate": 4.312331914993979e-08, "loss": 0.0145, "step": 20129 }, { "epoch": 7.93, "learning_rate": 4.310748285173562e-08, "loss": 0.0117, "step": 20130 }, { "epoch": 7.93, "learning_rate": 4.309164911064103e-08, "loss": 0.012, "step": 20131 }, { "epoch": 7.93, "learning_rate": 4.307581792691408e-08, "loss": 0.0107, "step": 20132 }, { "epoch": 7.93, "learning_rate": 4.305998930081285e-08, "loss": 0.0116, "step": 20133 }, { "epoch": 7.93, "learning_rate": 4.3044163232595235e-08, "loss": 0.0099, "step": 20134 }, { "epoch": 7.94, "learning_rate": 4.302833972251927e-08, "loss": 0.0115, "step": 20135 }, { "epoch": 7.94, "learning_rate": 4.3012518770842666e-08, "loss": 0.0107, "step": 20136 }, { "epoch": 7.94, "learning_rate": 4.299670037782355e-08, "loss": 0.01, "step": 20137 }, { "epoch": 7.94, "learning_rate": 4.298088454371951e-08, "loss": 0.0102, "step": 20138 }, { "epoch": 7.94, "learning_rate": 4.29650712687885e-08, "loss": 0.0122, "step": 20139 }, { "epoch": 7.94, "learning_rate": 4.294926055328809e-08, "loss": 0.0103, "step": 20140 }, { "epoch": 7.94, "learning_rate": 4.293345239747606e-08, "loss": 0.0111, "step": 20141 }, { "epoch": 7.94, "learning_rate": 4.2917646801610005e-08, "loss": 0.0114, "step": 20142 }, { "epoch": 7.94, "learning_rate": 4.2901843765947657e-08, "loss": 0.008, "step": 20143 }, { "epoch": 7.94, "learning_rate": 4.288604329074643e-08, "loss": 0.0106, "step": 20144 }, { "epoch": 7.94, "learning_rate": 4.2870245376263915e-08, "loss": 0.0135, "step": 20145 }, { "epoch": 7.94, "learning_rate": 4.2854450022757584e-08, "loss": 0.0106, "step": 20146 }, { "epoch": 7.94, "learning_rate": 4.283865723048494e-08, "loss": 0.0103, "step": 20147 }, { "epoch": 7.94, "learning_rate": 4.282286699970328e-08, "loss": 0.0113, "step": 20148 }, { "epoch": 7.94, "learning_rate": 4.280707933067007e-08, "loss": 0.0132, "step": 20149 }, { "epoch": 7.94, "learning_rate": 4.2791294223642496e-08, "loss": 0.0114, "step": 20150 }, { "epoch": 7.94, "learning_rate": 4.2775511678877874e-08, "loss": 0.0143, "step": 20151 }, { "epoch": 7.94, "learning_rate": 4.2759731696633536e-08, "loss": 0.0121, "step": 20152 }, { "epoch": 7.94, "learning_rate": 4.2743954277166525e-08, "loss": 0.0095, "step": 20153 }, { "epoch": 7.94, "learning_rate": 4.272817942073408e-08, "loss": 0.01, "step": 20154 }, { "epoch": 7.94, "learning_rate": 4.271240712759328e-08, "loss": 0.0095, "step": 20155 }, { "epoch": 7.94, "learning_rate": 4.269663739800122e-08, "loss": 0.0113, "step": 20156 }, { "epoch": 7.94, "learning_rate": 4.268087023221487e-08, "loss": 0.0098, "step": 20157 }, { "epoch": 7.94, "learning_rate": 4.266510563049128e-08, "loss": 0.0105, "step": 20158 }, { "epoch": 7.94, "learning_rate": 4.264934359308725e-08, "loss": 0.0099, "step": 20159 }, { "epoch": 7.95, "learning_rate": 4.2633584120259855e-08, "loss": 0.0108, "step": 20160 }, { "epoch": 7.95, "learning_rate": 4.2617827212265854e-08, "loss": 0.0115, "step": 20161 }, { "epoch": 7.95, "learning_rate": 4.26020728693621e-08, "loss": 0.01, "step": 20162 }, { "epoch": 7.95, "learning_rate": 4.2586321091805244e-08, "loss": 0.011, "step": 20163 }, { "epoch": 7.95, "learning_rate": 4.257057187985222e-08, "loss": 0.012, "step": 20164 }, { "epoch": 7.95, "learning_rate": 4.2554825233759574e-08, "loss": 0.0116, "step": 20165 }, { "epoch": 7.95, "learning_rate": 4.253908115378402e-08, "loss": 0.0107, "step": 20166 }, { "epoch": 7.95, "learning_rate": 4.2523339640182064e-08, "loss": 0.0098, "step": 20167 }, { "epoch": 7.95, "learning_rate": 4.250760069321035e-08, "loss": 0.0106, "step": 20168 }, { "epoch": 7.95, "learning_rate": 4.249186431312539e-08, "loss": 0.0107, "step": 20169 }, { "epoch": 7.95, "learning_rate": 4.2476130500183706e-08, "loss": 0.0082, "step": 20170 }, { "epoch": 7.95, "learning_rate": 4.2460399254641645e-08, "loss": 0.012, "step": 20171 }, { "epoch": 7.95, "learning_rate": 4.244467057675563e-08, "loss": 0.0109, "step": 20172 }, { "epoch": 7.95, "learning_rate": 4.242894446678207e-08, "loss": 0.0106, "step": 20173 }, { "epoch": 7.95, "learning_rate": 4.241322092497719e-08, "loss": 0.0109, "step": 20174 }, { "epoch": 7.95, "learning_rate": 4.239749995159738e-08, "loss": 0.0115, "step": 20175 }, { "epoch": 7.95, "learning_rate": 4.238178154689866e-08, "loss": 0.014, "step": 20176 }, { "epoch": 7.95, "learning_rate": 4.236606571113746e-08, "loss": 0.01, "step": 20177 }, { "epoch": 7.95, "learning_rate": 4.235035244456977e-08, "loss": 0.011, "step": 20178 }, { "epoch": 7.95, "learning_rate": 4.233464174745178e-08, "loss": 0.0104, "step": 20179 }, { "epoch": 7.95, "learning_rate": 4.2318933620039445e-08, "loss": 0.011, "step": 20180 }, { "epoch": 7.95, "learning_rate": 4.230322806258886e-08, "loss": 0.0128, "step": 20181 }, { "epoch": 7.95, "learning_rate": 4.228752507535598e-08, "loss": 0.0142, "step": 20182 }, { "epoch": 7.95, "learning_rate": 4.22718246585968e-08, "loss": 0.008, "step": 20183 }, { "epoch": 7.95, "learning_rate": 4.2256126812567096e-08, "loss": 0.013, "step": 20184 }, { "epoch": 7.96, "learning_rate": 4.224043153752279e-08, "loss": 0.0106, "step": 20185 }, { "epoch": 7.96, "learning_rate": 4.2224738833719686e-08, "loss": 0.0109, "step": 20186 }, { "epoch": 7.96, "learning_rate": 4.22090487014136e-08, "loss": 0.0103, "step": 20187 }, { "epoch": 7.96, "learning_rate": 4.219336114086014e-08, "loss": 0.0104, "step": 20188 }, { "epoch": 7.96, "learning_rate": 4.217767615231513e-08, "loss": 0.0097, "step": 20189 }, { "epoch": 7.96, "learning_rate": 4.216199373603406e-08, "loss": 0.0113, "step": 20190 }, { "epoch": 7.96, "learning_rate": 4.2146313892272636e-08, "loss": 0.0116, "step": 20191 }, { "epoch": 7.96, "learning_rate": 4.213063662128635e-08, "loss": 0.012, "step": 20192 }, { "epoch": 7.96, "learning_rate": 4.211496192333084e-08, "loss": 0.0086, "step": 20193 }, { "epoch": 7.96, "learning_rate": 4.209928979866142e-08, "loss": 0.0096, "step": 20194 }, { "epoch": 7.96, "learning_rate": 4.20836202475336e-08, "loss": 0.0093, "step": 20195 }, { "epoch": 7.96, "learning_rate": 4.206795327020281e-08, "loss": 0.0101, "step": 20196 }, { "epoch": 7.96, "learning_rate": 4.2052288866924316e-08, "loss": 0.0117, "step": 20197 }, { "epoch": 7.96, "learning_rate": 4.20366270379535e-08, "loss": 0.0098, "step": 20198 }, { "epoch": 7.96, "learning_rate": 4.202096778354547e-08, "loss": 0.0102, "step": 20199 }, { "epoch": 7.96, "learning_rate": 4.2005311103955686e-08, "loss": 0.0103, "step": 20200 }, { "epoch": 7.96, "learning_rate": 4.198965699943915e-08, "loss": 0.0105, "step": 20201 }, { "epoch": 7.96, "learning_rate": 4.197400547025112e-08, "loss": 0.0104, "step": 20202 }, { "epoch": 7.96, "learning_rate": 4.195835651664652e-08, "loss": 0.011, "step": 20203 }, { "epoch": 7.96, "learning_rate": 4.194271013888062e-08, "loss": 0.0099, "step": 20204 }, { "epoch": 7.96, "learning_rate": 4.192706633720826e-08, "loss": 0.0119, "step": 20205 }, { "epoch": 7.96, "learning_rate": 4.1911425111884546e-08, "loss": 0.0107, "step": 20206 }, { "epoch": 7.96, "learning_rate": 4.189578646316427e-08, "loss": 0.0117, "step": 20207 }, { "epoch": 7.96, "learning_rate": 4.188015039130237e-08, "loss": 0.0098, "step": 20208 }, { "epoch": 7.96, "learning_rate": 4.1864516896553725e-08, "loss": 0.0095, "step": 20209 }, { "epoch": 7.96, "learning_rate": 4.184888597917318e-08, "loss": 0.01, "step": 20210 }, { "epoch": 7.97, "learning_rate": 4.1833257639415346e-08, "loss": 0.0111, "step": 20211 }, { "epoch": 7.97, "learning_rate": 4.181763187753504e-08, "loss": 0.0105, "step": 20212 }, { "epoch": 7.97, "learning_rate": 4.180200869378698e-08, "loss": 0.0108, "step": 20213 }, { "epoch": 7.97, "learning_rate": 4.178638808842572e-08, "loss": 0.0116, "step": 20214 }, { "epoch": 7.97, "learning_rate": 4.177077006170584e-08, "loss": 0.0105, "step": 20215 }, { "epoch": 7.97, "learning_rate": 4.175515461388193e-08, "loss": 0.0093, "step": 20216 }, { "epoch": 7.97, "learning_rate": 4.1739541745208573e-08, "loss": 0.0111, "step": 20217 }, { "epoch": 7.97, "learning_rate": 4.1723931455940086e-08, "loss": 0.011, "step": 20218 }, { "epoch": 7.97, "learning_rate": 4.170832374633102e-08, "loss": 0.0105, "step": 20219 }, { "epoch": 7.97, "learning_rate": 4.1692718616635636e-08, "loss": 0.0097, "step": 20220 }, { "epoch": 7.97, "learning_rate": 4.1677116067108356e-08, "loss": 0.008, "step": 20221 }, { "epoch": 7.97, "learning_rate": 4.1661516098003456e-08, "loss": 0.0123, "step": 20222 }, { "epoch": 7.97, "learning_rate": 4.164591870957526e-08, "loss": 0.0125, "step": 20223 }, { "epoch": 7.97, "learning_rate": 4.1630323902077837e-08, "loss": 0.0095, "step": 20224 }, { "epoch": 7.97, "learning_rate": 4.1614731675765456e-08, "loss": 0.012, "step": 20225 }, { "epoch": 7.97, "learning_rate": 4.159914203089225e-08, "loss": 0.0117, "step": 20226 }, { "epoch": 7.97, "learning_rate": 4.1583554967712306e-08, "loss": 0.0087, "step": 20227 }, { "epoch": 7.97, "learning_rate": 4.156797048647962e-08, "loss": 0.0118, "step": 20228 }, { "epoch": 7.97, "learning_rate": 4.155238858744827e-08, "loss": 0.0131, "step": 20229 }, { "epoch": 7.97, "learning_rate": 4.153680927087209e-08, "loss": 0.0107, "step": 20230 }, { "epoch": 7.97, "learning_rate": 4.152123253700517e-08, "loss": 0.0108, "step": 20231 }, { "epoch": 7.97, "learning_rate": 4.150565838610125e-08, "loss": 0.0108, "step": 20232 }, { "epoch": 7.97, "learning_rate": 4.149008681841426e-08, "loss": 0.0106, "step": 20233 }, { "epoch": 7.97, "learning_rate": 4.147451783419791e-08, "loss": 0.012, "step": 20234 }, { "epoch": 7.97, "learning_rate": 4.145895143370597e-08, "loss": 0.01, "step": 20235 }, { "epoch": 7.98, "learning_rate": 4.144338761719217e-08, "loss": 0.0105, "step": 20236 }, { "epoch": 7.98, "learning_rate": 4.1427826384910224e-08, "loss": 0.0116, "step": 20237 }, { "epoch": 7.98, "learning_rate": 4.1412267737113657e-08, "loss": 0.01, "step": 20238 }, { "epoch": 7.98, "learning_rate": 4.1396711674056114e-08, "loss": 0.0093, "step": 20239 }, { "epoch": 7.98, "learning_rate": 4.138115819599117e-08, "loss": 0.0109, "step": 20240 }, { "epoch": 7.98, "learning_rate": 4.136560730317222e-08, "loss": 0.0106, "step": 20241 }, { "epoch": 7.98, "learning_rate": 4.135005899585282e-08, "loss": 0.0096, "step": 20242 }, { "epoch": 7.98, "learning_rate": 4.133451327428623e-08, "loss": 0.0106, "step": 20243 }, { "epoch": 7.98, "learning_rate": 4.131897013872605e-08, "loss": 0.0101, "step": 20244 }, { "epoch": 7.98, "learning_rate": 4.130342958942543e-08, "loss": 0.0092, "step": 20245 }, { "epoch": 7.98, "learning_rate": 4.128789162663775e-08, "loss": 0.0123, "step": 20246 }, { "epoch": 7.98, "learning_rate": 4.1272356250616206e-08, "loss": 0.0083, "step": 20247 }, { "epoch": 7.98, "learning_rate": 4.125682346161399e-08, "loss": 0.012, "step": 20248 }, { "epoch": 7.98, "learning_rate": 4.12412932598843e-08, "loss": 0.0111, "step": 20249 }, { "epoch": 7.98, "learning_rate": 4.12257656456803e-08, "loss": 0.0111, "step": 20250 }, { "epoch": 7.98, "learning_rate": 4.1210240619254935e-08, "loss": 0.0094, "step": 20251 }, { "epoch": 7.98, "learning_rate": 4.119471818086133e-08, "loss": 0.0102, "step": 20252 }, { "epoch": 7.98, "learning_rate": 4.1179198330752474e-08, "loss": 0.0084, "step": 20253 }, { "epoch": 7.98, "learning_rate": 4.1163681069181355e-08, "loss": 0.0099, "step": 20254 }, { "epoch": 7.98, "learning_rate": 4.114816639640078e-08, "loss": 0.0094, "step": 20255 }, { "epoch": 7.98, "learning_rate": 4.113265431266364e-08, "loss": 0.0109, "step": 20256 }, { "epoch": 7.98, "learning_rate": 4.111714481822286e-08, "loss": 0.0108, "step": 20257 }, { "epoch": 7.98, "learning_rate": 4.11016379133311e-08, "loss": 0.0098, "step": 20258 }, { "epoch": 7.98, "learning_rate": 4.1086133598241114e-08, "loss": 0.0089, "step": 20259 }, { "epoch": 7.98, "learning_rate": 4.107063187320572e-08, "loss": 0.0094, "step": 20260 }, { "epoch": 7.99, "learning_rate": 4.1055132738477405e-08, "loss": 0.0106, "step": 20261 }, { "epoch": 7.99, "learning_rate": 4.103963619430886e-08, "loss": 0.0119, "step": 20262 }, { "epoch": 7.99, "learning_rate": 4.10241422409527e-08, "loss": 0.0109, "step": 20263 }, { "epoch": 7.99, "learning_rate": 4.1008650878661364e-08, "loss": 0.0098, "step": 20264 }, { "epoch": 7.99, "learning_rate": 4.0993162107687375e-08, "loss": 0.0084, "step": 20265 }, { "epoch": 7.99, "learning_rate": 4.0977675928283183e-08, "loss": 0.0128, "step": 20266 }, { "epoch": 7.99, "learning_rate": 4.096219234070124e-08, "loss": 0.0106, "step": 20267 }, { "epoch": 7.99, "learning_rate": 4.0946711345193805e-08, "loss": 0.0108, "step": 20268 }, { "epoch": 7.99, "learning_rate": 4.0931232942013285e-08, "loss": 0.0117, "step": 20269 }, { "epoch": 7.99, "learning_rate": 4.091575713141182e-08, "loss": 0.0101, "step": 20270 }, { "epoch": 7.99, "learning_rate": 4.090028391364182e-08, "loss": 0.0109, "step": 20271 }, { "epoch": 7.99, "learning_rate": 4.0884813288955346e-08, "loss": 0.0092, "step": 20272 }, { "epoch": 7.99, "learning_rate": 4.086934525760464e-08, "loss": 0.0113, "step": 20273 }, { "epoch": 7.99, "learning_rate": 4.0853879819841655e-08, "loss": 0.0104, "step": 20274 }, { "epoch": 7.99, "learning_rate": 4.083841697591866e-08, "loss": 0.0123, "step": 20275 }, { "epoch": 7.99, "learning_rate": 4.082295672608751e-08, "loss": 0.0122, "step": 20276 }, { "epoch": 7.99, "learning_rate": 4.08074990706003e-08, "loss": 0.0128, "step": 20277 }, { "epoch": 7.99, "learning_rate": 4.079204400970886e-08, "loss": 0.0113, "step": 20278 }, { "epoch": 7.99, "learning_rate": 4.0776591543665126e-08, "loss": 0.0135, "step": 20279 }, { "epoch": 7.99, "learning_rate": 4.0761141672721023e-08, "loss": 0.013, "step": 20280 }, { "epoch": 7.99, "learning_rate": 4.074569439712823e-08, "loss": 0.0091, "step": 20281 }, { "epoch": 7.99, "learning_rate": 4.073024971713859e-08, "loss": 0.0107, "step": 20282 }, { "epoch": 7.99, "learning_rate": 4.0714807633003814e-08, "loss": 0.0116, "step": 20283 }, { "epoch": 7.99, "learning_rate": 4.069936814497563e-08, "loss": 0.0092, "step": 20284 }, { "epoch": 7.99, "learning_rate": 4.06839312533056e-08, "loss": 0.0098, "step": 20285 }, { "epoch": 7.99, "learning_rate": 4.066849695824541e-08, "loss": 0.01, "step": 20286 }, { "epoch": 8.0, "learning_rate": 4.0653065260046504e-08, "loss": 0.0109, "step": 20287 }, { "epoch": 8.0, "learning_rate": 4.0637636158960456e-08, "loss": 0.0101, "step": 20288 }, { "epoch": 8.0, "learning_rate": 4.062220965523872e-08, "loss": 0.0112, "step": 20289 }, { "epoch": 8.0, "learning_rate": 4.0606785749132835e-08, "loss": 0.0129, "step": 20290 }, { "epoch": 8.0, "learning_rate": 4.0591364440893996e-08, "loss": 0.0142, "step": 20291 }, { "epoch": 8.0, "learning_rate": 4.0575945730773676e-08, "loss": 0.0128, "step": 20292 }, { "epoch": 8.0, "learning_rate": 4.056052961902312e-08, "loss": 0.0085, "step": 20293 }, { "epoch": 8.0, "learning_rate": 4.0545116105893684e-08, "loss": 0.0121, "step": 20294 }, { "epoch": 8.0, "learning_rate": 4.0529705191636435e-08, "loss": 0.0099, "step": 20295 }, { "epoch": 8.0, "learning_rate": 4.051429687650265e-08, "loss": 0.0114, "step": 20296 }, { "epoch": 8.0, "learning_rate": 4.049889116074341e-08, "loss": 0.0102, "step": 20297 }, { "epoch": 8.0, "learning_rate": 4.0483488044609906e-08, "loss": 0.0115, "step": 20298 }, { "epoch": 8.0, "learning_rate": 4.046808752835305e-08, "loss": 0.0115, "step": 20299 }, { "epoch": 8.0, "learning_rate": 4.045268961222395e-08, "loss": 0.0095, "step": 20300 }, { "epoch": 8.0, "learning_rate": 4.043729429647349e-08, "loss": 0.0089, "step": 20301 }, { "epoch": 8.0, "learning_rate": 4.0421901581352614e-08, "loss": 0.0088, "step": 20302 }, { "epoch": 8.0, "learning_rate": 4.0406511467112246e-08, "loss": 0.0101, "step": 20303 }, { "epoch": 8.0, "learning_rate": 4.039112395400317e-08, "loss": 0.0102, "step": 20304 }, { "epoch": 8.0, "learning_rate": 4.037573904227618e-08, "loss": 0.0086, "step": 20305 }, { "epoch": 8.0, "learning_rate": 4.0360356732182036e-08, "loss": 0.0109, "step": 20306 }, { "epoch": 8.0, "learning_rate": 4.0344977023971506e-08, "loss": 0.011, "step": 20307 }, { "epoch": 8.0, "learning_rate": 4.0329599917895173e-08, "loss": 0.0104, "step": 20308 }, { "epoch": 8.0, "learning_rate": 4.0314225414203725e-08, "loss": 0.0091, "step": 20309 }, { "epoch": 8.0, "learning_rate": 4.029885351314759e-08, "loss": 0.0103, "step": 20310 }, { "epoch": 8.0, "learning_rate": 4.028348421497756e-08, "loss": 0.0082, "step": 20311 }, { "epoch": 8.01, "learning_rate": 4.026811751994392e-08, "loss": 0.0122, "step": 20312 }, { "epoch": 8.01, "learning_rate": 4.025275342829726e-08, "loss": 0.0105, "step": 20313 }, { "epoch": 8.01, "learning_rate": 4.0237391940287834e-08, "loss": 0.0078, "step": 20314 }, { "epoch": 8.01, "learning_rate": 4.022203305616621e-08, "loss": 0.0118, "step": 20315 }, { "epoch": 8.01, "learning_rate": 4.0206676776182545e-08, "loss": 0.0105, "step": 20316 }, { "epoch": 8.01, "learning_rate": 4.019132310058726e-08, "loss": 0.0085, "step": 20317 }, { "epoch": 8.01, "learning_rate": 4.0175972029630456e-08, "loss": 0.0106, "step": 20318 }, { "epoch": 8.01, "learning_rate": 4.0160623563562427e-08, "loss": 0.0081, "step": 20319 }, { "epoch": 8.01, "learning_rate": 4.01452777026333e-08, "loss": 0.0119, "step": 20320 }, { "epoch": 8.01, "learning_rate": 4.012993444709323e-08, "loss": 0.0123, "step": 20321 }, { "epoch": 8.01, "learning_rate": 4.011459379719222e-08, "loss": 0.0087, "step": 20322 }, { "epoch": 8.01, "learning_rate": 4.009925575318032e-08, "loss": 0.0089, "step": 20323 }, { "epoch": 8.01, "learning_rate": 4.008392031530761e-08, "loss": 0.0106, "step": 20324 }, { "epoch": 8.01, "learning_rate": 4.0068587483823867e-08, "loss": 0.0105, "step": 20325 }, { "epoch": 8.01, "learning_rate": 4.0053257258979146e-08, "loss": 0.0105, "step": 20326 }, { "epoch": 8.01, "learning_rate": 4.00379296410232e-08, "loss": 0.0109, "step": 20327 }, { "epoch": 8.01, "learning_rate": 4.002260463020588e-08, "loss": 0.0105, "step": 20328 }, { "epoch": 8.01, "learning_rate": 4.0007282226776964e-08, "loss": 0.0093, "step": 20329 }, { "epoch": 8.01, "learning_rate": 3.999196243098624e-08, "loss": 0.0127, "step": 20330 }, { "epoch": 8.01, "learning_rate": 3.9976645243083286e-08, "loss": 0.0113, "step": 20331 }, { "epoch": 8.01, "learning_rate": 3.9961330663317816e-08, "loss": 0.0084, "step": 20332 }, { "epoch": 8.01, "learning_rate": 3.9946018691939406e-08, "loss": 0.0112, "step": 20333 }, { "epoch": 8.01, "learning_rate": 3.993070932919771e-08, "loss": 0.0111, "step": 20334 }, { "epoch": 8.01, "learning_rate": 3.991540257534212e-08, "loss": 0.0104, "step": 20335 }, { "epoch": 8.01, "learning_rate": 3.9900098430622165e-08, "loss": 0.0138, "step": 20336 }, { "epoch": 8.01, "learning_rate": 3.9884796895287273e-08, "loss": 0.0126, "step": 20337 }, { "epoch": 8.02, "learning_rate": 3.9869497969586916e-08, "loss": 0.0087, "step": 20338 }, { "epoch": 8.02, "learning_rate": 3.985420165377031e-08, "loss": 0.0081, "step": 20339 }, { "epoch": 8.02, "learning_rate": 3.983890794808687e-08, "loss": 0.0078, "step": 20340 }, { "epoch": 8.02, "learning_rate": 3.982361685278573e-08, "loss": 0.0091, "step": 20341 }, { "epoch": 8.02, "learning_rate": 3.980832836811627e-08, "loss": 0.0105, "step": 20342 }, { "epoch": 8.02, "learning_rate": 3.979304249432755e-08, "loss": 0.0093, "step": 20343 }, { "epoch": 8.02, "learning_rate": 3.977775923166882e-08, "loss": 0.0113, "step": 20344 }, { "epoch": 8.02, "learning_rate": 3.9762478580389036e-08, "loss": 0.0131, "step": 20345 }, { "epoch": 8.02, "learning_rate": 3.9747200540737326e-08, "loss": 0.0104, "step": 20346 }, { "epoch": 8.02, "learning_rate": 3.973192511296275e-08, "loss": 0.0105, "step": 20347 }, { "epoch": 8.02, "learning_rate": 3.9716652297314136e-08, "loss": 0.0093, "step": 20348 }, { "epoch": 8.02, "learning_rate": 3.970138209404057e-08, "loss": 0.0078, "step": 20349 }, { "epoch": 8.02, "learning_rate": 3.968611450339072e-08, "loss": 0.008, "step": 20350 }, { "epoch": 8.02, "learning_rate": 3.967084952561368e-08, "loss": 0.0107, "step": 20351 }, { "epoch": 8.02, "learning_rate": 3.965558716095805e-08, "loss": 0.0108, "step": 20352 }, { "epoch": 8.02, "learning_rate": 3.964032740967271e-08, "loss": 0.0104, "step": 20353 }, { "epoch": 8.02, "learning_rate": 3.9625070272006236e-08, "loss": 0.0099, "step": 20354 }, { "epoch": 8.02, "learning_rate": 3.9609815748207455e-08, "loss": 0.0151, "step": 20355 }, { "epoch": 8.02, "learning_rate": 3.9594563838524866e-08, "loss": 0.009, "step": 20356 }, { "epoch": 8.02, "learning_rate": 3.957931454320715e-08, "loss": 0.0105, "step": 20357 }, { "epoch": 8.02, "learning_rate": 3.956406786250273e-08, "loss": 0.0106, "step": 20358 }, { "epoch": 8.02, "learning_rate": 3.954882379666018e-08, "loss": 0.0092, "step": 20359 }, { "epoch": 8.02, "learning_rate": 3.9533582345927965e-08, "loss": 0.0114, "step": 20360 }, { "epoch": 8.02, "learning_rate": 3.95183435105545e-08, "loss": 0.011, "step": 20361 }, { "epoch": 8.02, "learning_rate": 3.9503107290788096e-08, "loss": 0.0106, "step": 20362 }, { "epoch": 8.03, "learning_rate": 3.9487873686877094e-08, "loss": 0.0103, "step": 20363 }, { "epoch": 8.03, "learning_rate": 3.947264269906983e-08, "loss": 0.0105, "step": 20364 }, { "epoch": 8.03, "learning_rate": 3.9457414327614556e-08, "loss": 0.0081, "step": 20365 }, { "epoch": 8.03, "learning_rate": 3.944218857275936e-08, "loss": 0.0096, "step": 20366 }, { "epoch": 8.03, "learning_rate": 3.942696543475254e-08, "loss": 0.0155, "step": 20367 }, { "epoch": 8.03, "learning_rate": 3.9411744913842096e-08, "loss": 0.0108, "step": 20368 }, { "epoch": 8.03, "learning_rate": 3.9396527010276134e-08, "loss": 0.0096, "step": 20369 }, { "epoch": 8.03, "learning_rate": 3.938131172430277e-08, "loss": 0.0097, "step": 20370 }, { "epoch": 8.03, "learning_rate": 3.936609905616983e-08, "loss": 0.0094, "step": 20371 }, { "epoch": 8.03, "learning_rate": 3.935088900612535e-08, "loss": 0.0112, "step": 20372 }, { "epoch": 8.03, "learning_rate": 3.933568157441722e-08, "loss": 0.0105, "step": 20373 }, { "epoch": 8.03, "learning_rate": 3.932047676129335e-08, "loss": 0.0101, "step": 20374 }, { "epoch": 8.03, "learning_rate": 3.9305274567001454e-08, "loss": 0.0117, "step": 20375 }, { "epoch": 8.03, "learning_rate": 3.9290074991789336e-08, "loss": 0.009, "step": 20376 }, { "epoch": 8.03, "learning_rate": 3.927487803590477e-08, "loss": 0.0118, "step": 20377 }, { "epoch": 8.03, "learning_rate": 3.925968369959547e-08, "loss": 0.0122, "step": 20378 }, { "epoch": 8.03, "learning_rate": 3.9244491983108954e-08, "loss": 0.0106, "step": 20379 }, { "epoch": 8.03, "learning_rate": 3.922930288669297e-08, "loss": 0.0088, "step": 20380 }, { "epoch": 8.03, "learning_rate": 3.921411641059489e-08, "loss": 0.0106, "step": 20381 }, { "epoch": 8.03, "learning_rate": 3.919893255506246e-08, "loss": 0.01, "step": 20382 }, { "epoch": 8.03, "learning_rate": 3.9183751320342996e-08, "loss": 0.0101, "step": 20383 }, { "epoch": 8.03, "learning_rate": 3.916857270668401e-08, "loss": 0.0109, "step": 20384 }, { "epoch": 8.03, "learning_rate": 3.9153396714332796e-08, "loss": 0.011, "step": 20385 }, { "epoch": 8.03, "learning_rate": 3.9138223343536755e-08, "loss": 0.0092, "step": 20386 }, { "epoch": 8.03, "learning_rate": 3.91230525945432e-08, "loss": 0.0108, "step": 20387 }, { "epoch": 8.04, "learning_rate": 3.910788446759945e-08, "loss": 0.0089, "step": 20388 }, { "epoch": 8.04, "learning_rate": 3.909271896295259e-08, "loss": 0.0097, "step": 20389 }, { "epoch": 8.04, "learning_rate": 3.907755608084986e-08, "loss": 0.0095, "step": 20390 }, { "epoch": 8.04, "learning_rate": 3.906239582153845e-08, "loss": 0.01, "step": 20391 }, { "epoch": 8.04, "learning_rate": 3.904723818526534e-08, "loss": 0.0088, "step": 20392 }, { "epoch": 8.04, "learning_rate": 3.9032083172277666e-08, "loss": 0.0086, "step": 20393 }, { "epoch": 8.04, "learning_rate": 3.9016930782822287e-08, "loss": 0.0114, "step": 20394 }, { "epoch": 8.04, "learning_rate": 3.9001781017146396e-08, "loss": 0.0107, "step": 20395 }, { "epoch": 8.04, "learning_rate": 3.8986633875496724e-08, "loss": 0.0111, "step": 20396 }, { "epoch": 8.04, "learning_rate": 3.897148935812025e-08, "loss": 0.0094, "step": 20397 }, { "epoch": 8.04, "learning_rate": 3.8956347465263726e-08, "loss": 0.0084, "step": 20398 }, { "epoch": 8.04, "learning_rate": 3.894120819717397e-08, "loss": 0.0101, "step": 20399 }, { "epoch": 8.04, "learning_rate": 3.8926071554097726e-08, "loss": 0.01, "step": 20400 }, { "epoch": 8.04, "learning_rate": 3.8910937536281785e-08, "loss": 0.0084, "step": 20401 }, { "epoch": 8.04, "learning_rate": 3.8895806143972676e-08, "loss": 0.0135, "step": 20402 }, { "epoch": 8.04, "learning_rate": 3.888067737741709e-08, "loss": 0.0111, "step": 20403 }, { "epoch": 8.04, "learning_rate": 3.886555123686157e-08, "loss": 0.0147, "step": 20404 }, { "epoch": 8.04, "learning_rate": 3.8850427722552735e-08, "loss": 0.0116, "step": 20405 }, { "epoch": 8.04, "learning_rate": 3.883530683473695e-08, "loss": 0.0092, "step": 20406 }, { "epoch": 8.04, "learning_rate": 3.882018857366076e-08, "loss": 0.0077, "step": 20407 }, { "epoch": 8.04, "learning_rate": 3.8805072939570515e-08, "loss": 0.012, "step": 20408 }, { "epoch": 8.04, "learning_rate": 3.878995993271257e-08, "loss": 0.0127, "step": 20409 }, { "epoch": 8.04, "learning_rate": 3.8774849553333256e-08, "loss": 0.0105, "step": 20410 }, { "epoch": 8.04, "learning_rate": 3.8759741801678937e-08, "loss": 0.0102, "step": 20411 }, { "epoch": 8.04, "learning_rate": 3.87446366779957e-08, "loss": 0.0115, "step": 20412 }, { "epoch": 8.04, "learning_rate": 3.87295341825298e-08, "loss": 0.0096, "step": 20413 }, { "epoch": 8.05, "learning_rate": 3.8714434315527454e-08, "loss": 0.0086, "step": 20414 }, { "epoch": 8.05, "learning_rate": 3.869933707723465e-08, "loss": 0.0111, "step": 20415 }, { "epoch": 8.05, "learning_rate": 3.86842424678975e-08, "loss": 0.0117, "step": 20416 }, { "epoch": 8.05, "learning_rate": 3.8669150487762025e-08, "loss": 0.007, "step": 20417 }, { "epoch": 8.05, "learning_rate": 3.8654061137074276e-08, "loss": 0.0098, "step": 20418 }, { "epoch": 8.05, "learning_rate": 3.863897441608004e-08, "loss": 0.0086, "step": 20419 }, { "epoch": 8.05, "learning_rate": 3.862389032502533e-08, "loss": 0.0089, "step": 20420 }, { "epoch": 8.05, "learning_rate": 3.860880886415585e-08, "loss": 0.0093, "step": 20421 }, { "epoch": 8.05, "learning_rate": 3.85937300337176e-08, "loss": 0.0118, "step": 20422 }, { "epoch": 8.05, "learning_rate": 3.857865383395618e-08, "loss": 0.0088, "step": 20423 }, { "epoch": 8.05, "learning_rate": 3.8563580265117434e-08, "loss": 0.0106, "step": 20424 }, { "epoch": 8.05, "learning_rate": 3.8548509327446866e-08, "loss": 0.0099, "step": 20425 }, { "epoch": 8.05, "learning_rate": 3.853344102119032e-08, "loss": 0.0125, "step": 20426 }, { "epoch": 8.05, "learning_rate": 3.8518375346593234e-08, "loss": 0.0099, "step": 20427 }, { "epoch": 8.05, "learning_rate": 3.850331230390127e-08, "loss": 0.0109, "step": 20428 }, { "epoch": 8.05, "learning_rate": 3.8488251893359804e-08, "loss": 0.0092, "step": 20429 }, { "epoch": 8.05, "learning_rate": 3.8473194115214367e-08, "loss": 0.0092, "step": 20430 }, { "epoch": 8.05, "learning_rate": 3.845813896971042e-08, "loss": 0.012, "step": 20431 }, { "epoch": 8.05, "learning_rate": 3.8443086457093247e-08, "loss": 0.0119, "step": 20432 }, { "epoch": 8.05, "learning_rate": 3.842803657760823e-08, "loss": 0.0114, "step": 20433 }, { "epoch": 8.05, "learning_rate": 3.841298933150063e-08, "loss": 0.0096, "step": 20434 }, { "epoch": 8.05, "learning_rate": 3.839794471901581e-08, "loss": 0.0106, "step": 20435 }, { "epoch": 8.05, "learning_rate": 3.8382902740398794e-08, "loss": 0.0117, "step": 20436 }, { "epoch": 8.05, "learning_rate": 3.8367863395894904e-08, "loss": 0.009, "step": 20437 }, { "epoch": 8.05, "learning_rate": 3.835282668574915e-08, "loss": 0.0107, "step": 20438 }, { "epoch": 8.06, "learning_rate": 3.833779261020662e-08, "loss": 0.0087, "step": 20439 }, { "epoch": 8.06, "learning_rate": 3.832276116951239e-08, "loss": 0.0103, "step": 20440 }, { "epoch": 8.06, "learning_rate": 3.830773236391147e-08, "loss": 0.0092, "step": 20441 }, { "epoch": 8.06, "learning_rate": 3.829270619364873e-08, "loss": 0.0087, "step": 20442 }, { "epoch": 8.06, "learning_rate": 3.827768265896911e-08, "loss": 0.0115, "step": 20443 }, { "epoch": 8.06, "learning_rate": 3.826266176011748e-08, "loss": 0.0101, "step": 20444 }, { "epoch": 8.06, "learning_rate": 3.8247643497338685e-08, "loss": 0.0104, "step": 20445 }, { "epoch": 8.06, "learning_rate": 3.823262787087742e-08, "loss": 0.0103, "step": 20446 }, { "epoch": 8.06, "learning_rate": 3.821761488097852e-08, "loss": 0.0107, "step": 20447 }, { "epoch": 8.06, "learning_rate": 3.8202604527886504e-08, "loss": 0.012, "step": 20448 }, { "epoch": 8.06, "learning_rate": 3.818759681184625e-08, "loss": 0.0092, "step": 20449 }, { "epoch": 8.06, "learning_rate": 3.8172591733102164e-08, "loss": 0.0131, "step": 20450 }, { "epoch": 8.06, "learning_rate": 3.8157589291898963e-08, "loss": 0.0117, "step": 20451 }, { "epoch": 8.06, "learning_rate": 3.814258948848101e-08, "loss": 0.0086, "step": 20452 }, { "epoch": 8.06, "learning_rate": 3.812759232309286e-08, "loss": 0.0094, "step": 20453 }, { "epoch": 8.06, "learning_rate": 3.8112597795978995e-08, "loss": 0.0099, "step": 20454 }, { "epoch": 8.06, "learning_rate": 3.809760590738369e-08, "loss": 0.0103, "step": 20455 }, { "epoch": 8.06, "learning_rate": 3.808261665755135e-08, "loss": 0.0092, "step": 20456 }, { "epoch": 8.06, "learning_rate": 3.806763004672626e-08, "loss": 0.0106, "step": 20457 }, { "epoch": 8.06, "learning_rate": 3.805264607515277e-08, "loss": 0.0116, "step": 20458 }, { "epoch": 8.06, "learning_rate": 3.803766474307495e-08, "loss": 0.0105, "step": 20459 }, { "epoch": 8.06, "learning_rate": 3.802268605073708e-08, "loss": 0.0128, "step": 20460 }, { "epoch": 8.06, "learning_rate": 3.8007709998383165e-08, "loss": 0.0105, "step": 20461 }, { "epoch": 8.06, "learning_rate": 3.799273658625746e-08, "loss": 0.0094, "step": 20462 }, { "epoch": 8.06, "learning_rate": 3.79777658146039e-08, "loss": 0.0103, "step": 20463 }, { "epoch": 8.07, "learning_rate": 3.796279768366655e-08, "loss": 0.0105, "step": 20464 }, { "epoch": 8.07, "learning_rate": 3.794783219368922e-08, "loss": 0.0113, "step": 20465 }, { "epoch": 8.07, "learning_rate": 3.793286934491604e-08, "loss": 0.0113, "step": 20466 }, { "epoch": 8.07, "learning_rate": 3.791790913759072e-08, "loss": 0.0118, "step": 20467 }, { "epoch": 8.07, "learning_rate": 3.790295157195722e-08, "loss": 0.0091, "step": 20468 }, { "epoch": 8.07, "learning_rate": 3.788799664825917e-08, "loss": 0.0103, "step": 20469 }, { "epoch": 8.07, "learning_rate": 3.7873044366740416e-08, "loss": 0.0091, "step": 20470 }, { "epoch": 8.07, "learning_rate": 3.785809472764463e-08, "loss": 0.0091, "step": 20471 }, { "epoch": 8.07, "learning_rate": 3.7843147731215506e-08, "loss": 0.0132, "step": 20472 }, { "epoch": 8.07, "learning_rate": 3.7828203377696585e-08, "loss": 0.0108, "step": 20473 }, { "epoch": 8.07, "learning_rate": 3.7813261667331476e-08, "loss": 0.0106, "step": 20474 }, { "epoch": 8.07, "learning_rate": 3.7798322600363753e-08, "loss": 0.0096, "step": 20475 }, { "epoch": 8.07, "learning_rate": 3.7783386177036827e-08, "loss": 0.0102, "step": 20476 }, { "epoch": 8.07, "learning_rate": 3.7768452397594185e-08, "loss": 0.0123, "step": 20477 }, { "epoch": 8.07, "learning_rate": 3.775352126227918e-08, "loss": 0.0102, "step": 20478 }, { "epoch": 8.07, "learning_rate": 3.7738592771335175e-08, "loss": 0.009, "step": 20479 }, { "epoch": 8.07, "learning_rate": 3.772366692500548e-08, "loss": 0.0097, "step": 20480 }, { "epoch": 8.07, "learning_rate": 3.770874372353348e-08, "loss": 0.0119, "step": 20481 }, { "epoch": 8.07, "learning_rate": 3.769382316716221e-08, "loss": 0.0101, "step": 20482 }, { "epoch": 8.07, "learning_rate": 3.767890525613498e-08, "loss": 0.0106, "step": 20483 }, { "epoch": 8.07, "learning_rate": 3.766398999069487e-08, "loss": 0.012, "step": 20484 }, { "epoch": 8.07, "learning_rate": 3.7649077371085045e-08, "loss": 0.0082, "step": 20485 }, { "epoch": 8.07, "learning_rate": 3.7634167397548477e-08, "loss": 0.0125, "step": 20486 }, { "epoch": 8.07, "learning_rate": 3.7619260070328276e-08, "loss": 0.0095, "step": 20487 }, { "epoch": 8.07, "learning_rate": 3.760435538966722e-08, "loss": 0.01, "step": 20488 }, { "epoch": 8.07, "learning_rate": 3.7589453355808474e-08, "loss": 0.011, "step": 20489 }, { "epoch": 8.08, "learning_rate": 3.757455396899476e-08, "loss": 0.0121, "step": 20490 }, { "epoch": 8.08, "learning_rate": 3.755965722946899e-08, "loss": 0.0102, "step": 20491 }, { "epoch": 8.08, "learning_rate": 3.7544763137473835e-08, "loss": 0.0099, "step": 20492 }, { "epoch": 8.08, "learning_rate": 3.752987169325224e-08, "loss": 0.0102, "step": 20493 }, { "epoch": 8.08, "learning_rate": 3.751498289704675e-08, "loss": 0.01, "step": 20494 }, { "epoch": 8.08, "learning_rate": 3.750009674910015e-08, "loss": 0.011, "step": 20495 }, { "epoch": 8.08, "learning_rate": 3.748521324965495e-08, "loss": 0.0112, "step": 20496 }, { "epoch": 8.08, "learning_rate": 3.7470332398953786e-08, "loss": 0.0115, "step": 20497 }, { "epoch": 8.08, "learning_rate": 3.745545419723923e-08, "loss": 0.0081, "step": 20498 }, { "epoch": 8.08, "learning_rate": 3.744057864475367e-08, "loss": 0.0096, "step": 20499 }, { "epoch": 8.08, "learning_rate": 3.742570574173969e-08, "loss": 0.0087, "step": 20500 }, { "epoch": 8.08, "learning_rate": 3.7410835488439505e-08, "loss": 0.0106, "step": 20501 }, { "epoch": 8.08, "learning_rate": 3.739596788509569e-08, "loss": 0.0126, "step": 20502 }, { "epoch": 8.08, "learning_rate": 3.7381102931950424e-08, "loss": 0.011, "step": 20503 }, { "epoch": 8.08, "learning_rate": 3.7366240629246095e-08, "loss": 0.0106, "step": 20504 }, { "epoch": 8.08, "learning_rate": 3.735138097722475e-08, "loss": 0.0107, "step": 20505 }, { "epoch": 8.08, "learning_rate": 3.733652397612881e-08, "loss": 0.0111, "step": 20506 }, { "epoch": 8.08, "learning_rate": 3.7321669626200247e-08, "loss": 0.0095, "step": 20507 }, { "epoch": 8.08, "learning_rate": 3.7306817927681286e-08, "loss": 0.0071, "step": 20508 }, { "epoch": 8.08, "learning_rate": 3.7291968880813876e-08, "loss": 0.0094, "step": 20509 }, { "epoch": 8.08, "learning_rate": 3.7277122485840075e-08, "loss": 0.0072, "step": 20510 }, { "epoch": 8.08, "learning_rate": 3.726227874300187e-08, "loss": 0.0105, "step": 20511 }, { "epoch": 8.08, "learning_rate": 3.724743765254126e-08, "loss": 0.0079, "step": 20512 }, { "epoch": 8.08, "learning_rate": 3.723259921469999e-08, "loss": 0.0091, "step": 20513 }, { "epoch": 8.08, "learning_rate": 3.721776342971998e-08, "loss": 0.009, "step": 20514 }, { "epoch": 8.09, "learning_rate": 3.7202930297843026e-08, "loss": 0.0104, "step": 20515 }, { "epoch": 8.09, "learning_rate": 3.7188099819310926e-08, "loss": 0.009, "step": 20516 }, { "epoch": 8.09, "learning_rate": 3.7173271994365304e-08, "loss": 0.0091, "step": 20517 }, { "epoch": 8.09, "learning_rate": 3.7158446823247955e-08, "loss": 0.0129, "step": 20518 }, { "epoch": 8.09, "learning_rate": 3.714362430620037e-08, "loss": 0.011, "step": 20519 }, { "epoch": 8.09, "learning_rate": 3.712880444346418e-08, "loss": 0.0097, "step": 20520 }, { "epoch": 8.09, "learning_rate": 3.7113987235281007e-08, "loss": 0.0125, "step": 20521 }, { "epoch": 8.09, "learning_rate": 3.709917268189222e-08, "loss": 0.0102, "step": 20522 }, { "epoch": 8.09, "learning_rate": 3.7084360783539356e-08, "loss": 0.0082, "step": 20523 }, { "epoch": 8.09, "learning_rate": 3.7069551540463804e-08, "loss": 0.0092, "step": 20524 }, { "epoch": 8.09, "learning_rate": 3.705474495290697e-08, "loss": 0.0116, "step": 20525 }, { "epoch": 8.09, "learning_rate": 3.703994102111012e-08, "loss": 0.0143, "step": 20526 }, { "epoch": 8.09, "learning_rate": 3.7025139745314584e-08, "loss": 0.0106, "step": 20527 }, { "epoch": 8.09, "learning_rate": 3.701034112576149e-08, "loss": 0.0084, "step": 20528 }, { "epoch": 8.09, "learning_rate": 3.699554516269223e-08, "loss": 0.0121, "step": 20529 }, { "epoch": 8.09, "learning_rate": 3.698075185634777e-08, "loss": 0.0119, "step": 20530 }, { "epoch": 8.09, "learning_rate": 3.6965961206969376e-08, "loss": 0.0132, "step": 20531 }, { "epoch": 8.09, "learning_rate": 3.6951173214797904e-08, "loss": 0.0111, "step": 20532 }, { "epoch": 8.09, "learning_rate": 3.6936387880074625e-08, "loss": 0.0082, "step": 20533 }, { "epoch": 8.09, "learning_rate": 3.692160520304033e-08, "loss": 0.0087, "step": 20534 }, { "epoch": 8.09, "learning_rate": 3.6906825183936065e-08, "loss": 0.013, "step": 20535 }, { "epoch": 8.09, "learning_rate": 3.689204782300264e-08, "loss": 0.0095, "step": 20536 }, { "epoch": 8.09, "learning_rate": 3.6877273120480945e-08, "loss": 0.0113, "step": 20537 }, { "epoch": 8.09, "learning_rate": 3.6862501076611755e-08, "loss": 0.0087, "step": 20538 }, { "epoch": 8.09, "learning_rate": 3.684773169163593e-08, "loss": 0.0147, "step": 20539 }, { "epoch": 8.09, "learning_rate": 3.6832964965794045e-08, "loss": 0.0112, "step": 20540 }, { "epoch": 8.1, "learning_rate": 3.681820089932686e-08, "loss": 0.011, "step": 20541 }, { "epoch": 8.1, "learning_rate": 3.680343949247504e-08, "loss": 0.0102, "step": 20542 }, { "epoch": 8.1, "learning_rate": 3.6788680745479093e-08, "loss": 0.0107, "step": 20543 }, { "epoch": 8.1, "learning_rate": 3.677392465857962e-08, "loss": 0.0108, "step": 20544 }, { "epoch": 8.1, "learning_rate": 3.675917123201702e-08, "loss": 0.0103, "step": 20545 }, { "epoch": 8.1, "learning_rate": 3.6744420466031925e-08, "loss": 0.01, "step": 20546 }, { "epoch": 8.1, "learning_rate": 3.6729672360864596e-08, "loss": 0.012, "step": 20547 }, { "epoch": 8.1, "learning_rate": 3.671492691675553e-08, "loss": 0.0105, "step": 20548 }, { "epoch": 8.1, "learning_rate": 3.6700184133944937e-08, "loss": 0.0107, "step": 20549 }, { "epoch": 8.1, "learning_rate": 3.668544401267315e-08, "loss": 0.0112, "step": 20550 }, { "epoch": 8.1, "learning_rate": 3.6670706553180406e-08, "loss": 0.0092, "step": 20551 }, { "epoch": 8.1, "learning_rate": 3.665597175570698e-08, "loss": 0.0099, "step": 20552 }, { "epoch": 8.1, "learning_rate": 3.6641239620492904e-08, "loss": 0.0114, "step": 20553 }, { "epoch": 8.1, "learning_rate": 3.6626510147778355e-08, "loss": 0.0125, "step": 20554 }, { "epoch": 8.1, "learning_rate": 3.661178333780337e-08, "loss": 0.0084, "step": 20555 }, { "epoch": 8.1, "learning_rate": 3.6597059190808064e-08, "loss": 0.0084, "step": 20556 }, { "epoch": 8.1, "learning_rate": 3.65823377070323e-08, "loss": 0.0125, "step": 20557 }, { "epoch": 8.1, "learning_rate": 3.6567618886716114e-08, "loss": 0.0134, "step": 20558 }, { "epoch": 8.1, "learning_rate": 3.65529027300993e-08, "loss": 0.0119, "step": 20559 }, { "epoch": 8.1, "learning_rate": 3.653818923742176e-08, "loss": 0.0078, "step": 20560 }, { "epoch": 8.1, "learning_rate": 3.652347840892329e-08, "loss": 0.011, "step": 20561 }, { "epoch": 8.1, "learning_rate": 3.650877024484373e-08, "loss": 0.0097, "step": 20562 }, { "epoch": 8.1, "learning_rate": 3.649406474542267e-08, "loss": 0.01, "step": 20563 }, { "epoch": 8.1, "learning_rate": 3.6479361910899884e-08, "loss": 0.0094, "step": 20564 }, { "epoch": 8.1, "learning_rate": 3.646466174151501e-08, "loss": 0.0096, "step": 20565 }, { "epoch": 8.11, "learning_rate": 3.644996423750755e-08, "loss": 0.0124, "step": 20566 }, { "epoch": 8.11, "learning_rate": 3.6435269399117165e-08, "loss": 0.0091, "step": 20567 }, { "epoch": 8.11, "learning_rate": 3.642057722658318e-08, "loss": 0.0114, "step": 20568 }, { "epoch": 8.11, "learning_rate": 3.64058877201453e-08, "loss": 0.011, "step": 20569 }, { "epoch": 8.11, "learning_rate": 3.639120088004275e-08, "loss": 0.0083, "step": 20570 }, { "epoch": 8.11, "learning_rate": 3.6376516706515025e-08, "loss": 0.0094, "step": 20571 }, { "epoch": 8.11, "learning_rate": 3.63618351998013e-08, "loss": 0.0091, "step": 20572 }, { "epoch": 8.11, "learning_rate": 3.634715636014105e-08, "loss": 0.0098, "step": 20573 }, { "epoch": 8.11, "learning_rate": 3.633248018777337e-08, "loss": 0.0099, "step": 20574 }, { "epoch": 8.11, "learning_rate": 3.631780668293758e-08, "loss": 0.0111, "step": 20575 }, { "epoch": 8.11, "learning_rate": 3.630313584587268e-08, "loss": 0.009, "step": 20576 }, { "epoch": 8.11, "learning_rate": 3.628846767681797e-08, "loss": 0.0096, "step": 20577 }, { "epoch": 8.11, "learning_rate": 3.627380217601237e-08, "loss": 0.0085, "step": 20578 }, { "epoch": 8.11, "learning_rate": 3.625913934369502e-08, "loss": 0.0099, "step": 20579 }, { "epoch": 8.11, "learning_rate": 3.624447918010478e-08, "loss": 0.0106, "step": 20580 }, { "epoch": 8.11, "learning_rate": 3.6229821685480656e-08, "loss": 0.0128, "step": 20581 }, { "epoch": 8.11, "learning_rate": 3.621516686006159e-08, "loss": 0.0096, "step": 20582 }, { "epoch": 8.11, "learning_rate": 3.6200514704086315e-08, "loss": 0.0088, "step": 20583 }, { "epoch": 8.11, "learning_rate": 3.61858652177937e-08, "loss": 0.011, "step": 20584 }, { "epoch": 8.11, "learning_rate": 3.617121840142252e-08, "loss": 0.015, "step": 20585 }, { "epoch": 8.11, "learning_rate": 3.615657425521155e-08, "loss": 0.0089, "step": 20586 }, { "epoch": 8.11, "learning_rate": 3.6141932779399363e-08, "loss": 0.0117, "step": 20587 }, { "epoch": 8.11, "learning_rate": 3.612729397422467e-08, "loss": 0.0132, "step": 20588 }, { "epoch": 8.11, "learning_rate": 3.6112657839925984e-08, "loss": 0.011, "step": 20589 }, { "epoch": 8.11, "learning_rate": 3.6098024376741876e-08, "loss": 0.0097, "step": 20590 }, { "epoch": 8.12, "learning_rate": 3.60833935849109e-08, "loss": 0.0084, "step": 20591 }, { "epoch": 8.12, "learning_rate": 3.60687654646715e-08, "loss": 0.0135, "step": 20592 }, { "epoch": 8.12, "learning_rate": 3.605414001626206e-08, "loss": 0.0139, "step": 20593 }, { "epoch": 8.12, "learning_rate": 3.6039517239920956e-08, "loss": 0.0106, "step": 20594 }, { "epoch": 8.12, "learning_rate": 3.602489713588652e-08, "loss": 0.0094, "step": 20595 }, { "epoch": 8.12, "learning_rate": 3.6010279704397095e-08, "loss": 0.0101, "step": 20596 }, { "epoch": 8.12, "learning_rate": 3.599566494569084e-08, "loss": 0.011, "step": 20597 }, { "epoch": 8.12, "learning_rate": 3.598105286000603e-08, "loss": 0.0101, "step": 20598 }, { "epoch": 8.12, "learning_rate": 3.5966443447580684e-08, "loss": 0.0109, "step": 20599 }, { "epoch": 8.12, "learning_rate": 3.595183670865312e-08, "loss": 0.0083, "step": 20600 }, { "epoch": 8.12, "learning_rate": 3.5937232643461225e-08, "loss": 0.0116, "step": 20601 }, { "epoch": 8.12, "learning_rate": 3.592263125224313e-08, "loss": 0.0097, "step": 20602 }, { "epoch": 8.12, "learning_rate": 3.5908032535236754e-08, "loss": 0.0119, "step": 20603 }, { "epoch": 8.12, "learning_rate": 3.589343649268004e-08, "loss": 0.011, "step": 20604 }, { "epoch": 8.12, "learning_rate": 3.587884312481095e-08, "loss": 0.012, "step": 20605 }, { "epoch": 8.12, "learning_rate": 3.5864252431867236e-08, "loss": 0.0094, "step": 20606 }, { "epoch": 8.12, "learning_rate": 3.5849664414086746e-08, "loss": 0.0096, "step": 20607 }, { "epoch": 8.12, "learning_rate": 3.583507907170724e-08, "loss": 0.0097, "step": 20608 }, { "epoch": 8.12, "learning_rate": 3.5820496404966517e-08, "loss": 0.0099, "step": 20609 }, { "epoch": 8.12, "learning_rate": 3.580591641410211e-08, "loss": 0.0089, "step": 20610 }, { "epoch": 8.12, "learning_rate": 3.579133909935177e-08, "loss": 0.0112, "step": 20611 }, { "epoch": 8.12, "learning_rate": 3.577676446095292e-08, "loss": 0.0092, "step": 20612 }, { "epoch": 8.12, "learning_rate": 3.576219249914334e-08, "loss": 0.0109, "step": 20613 }, { "epoch": 8.12, "learning_rate": 3.574762321416036e-08, "loss": 0.0097, "step": 20614 }, { "epoch": 8.12, "learning_rate": 3.573305660624153e-08, "loss": 0.012, "step": 20615 }, { "epoch": 8.12, "learning_rate": 3.5718492675624125e-08, "loss": 0.0117, "step": 20616 }, { "epoch": 8.13, "learning_rate": 3.570393142254571e-08, "loss": 0.0103, "step": 20617 }, { "epoch": 8.13, "learning_rate": 3.568937284724345e-08, "loss": 0.0119, "step": 20618 }, { "epoch": 8.13, "learning_rate": 3.567481694995473e-08, "loss": 0.0122, "step": 20619 }, { "epoch": 8.13, "learning_rate": 3.566026373091673e-08, "loss": 0.01, "step": 20620 }, { "epoch": 8.13, "learning_rate": 3.564571319036664e-08, "loss": 0.0107, "step": 20621 }, { "epoch": 8.13, "learning_rate": 3.5631165328541647e-08, "loss": 0.0099, "step": 20622 }, { "epoch": 8.13, "learning_rate": 3.56166201456789e-08, "loss": 0.0108, "step": 20623 }, { "epoch": 8.13, "learning_rate": 3.560207764201535e-08, "loss": 0.0147, "step": 20624 }, { "epoch": 8.13, "learning_rate": 3.558753781778809e-08, "loss": 0.0086, "step": 20625 }, { "epoch": 8.13, "learning_rate": 3.557300067323412e-08, "loss": 0.0105, "step": 20626 }, { "epoch": 8.13, "learning_rate": 3.555846620859031e-08, "loss": 0.0106, "step": 20627 }, { "epoch": 8.13, "learning_rate": 3.5543934424093603e-08, "loss": 0.0099, "step": 20628 }, { "epoch": 8.13, "learning_rate": 3.5529405319980784e-08, "loss": 0.0108, "step": 20629 }, { "epoch": 8.13, "learning_rate": 3.55148788964887e-08, "loss": 0.0095, "step": 20630 }, { "epoch": 8.13, "learning_rate": 3.55003551538541e-08, "loss": 0.0111, "step": 20631 }, { "epoch": 8.13, "learning_rate": 3.548583409231374e-08, "loss": 0.0097, "step": 20632 }, { "epoch": 8.13, "learning_rate": 3.54713157121042e-08, "loss": 0.0107, "step": 20633 }, { "epoch": 8.13, "learning_rate": 3.545680001346216e-08, "loss": 0.0119, "step": 20634 }, { "epoch": 8.13, "learning_rate": 3.544228699662421e-08, "loss": 0.0114, "step": 20635 }, { "epoch": 8.13, "learning_rate": 3.542777666182692e-08, "loss": 0.0112, "step": 20636 }, { "epoch": 8.13, "learning_rate": 3.5413269009306725e-08, "loss": 0.0109, "step": 20637 }, { "epoch": 8.13, "learning_rate": 3.5398764039300134e-08, "loss": 0.0104, "step": 20638 }, { "epoch": 8.13, "learning_rate": 3.538426175204341e-08, "loss": 0.0104, "step": 20639 }, { "epoch": 8.13, "learning_rate": 3.5369762147773165e-08, "loss": 0.0088, "step": 20640 }, { "epoch": 8.13, "learning_rate": 3.535526522672552e-08, "loss": 0.0103, "step": 20641 }, { "epoch": 8.14, "learning_rate": 3.534077098913686e-08, "loss": 0.0086, "step": 20642 }, { "epoch": 8.14, "learning_rate": 3.532627943524329e-08, "loss": 0.0101, "step": 20643 }, { "epoch": 8.14, "learning_rate": 3.53117905652812e-08, "loss": 0.0091, "step": 20644 }, { "epoch": 8.14, "learning_rate": 3.529730437948657e-08, "loss": 0.0123, "step": 20645 }, { "epoch": 8.14, "learning_rate": 3.52828208780956e-08, "loss": 0.0105, "step": 20646 }, { "epoch": 8.14, "learning_rate": 3.526834006134427e-08, "loss": 0.0093, "step": 20647 }, { "epoch": 8.14, "learning_rate": 3.525386192946864e-08, "loss": 0.0074, "step": 20648 }, { "epoch": 8.14, "learning_rate": 3.52393864827047e-08, "loss": 0.0117, "step": 20649 }, { "epoch": 8.14, "learning_rate": 3.522491372128833e-08, "loss": 0.0096, "step": 20650 }, { "epoch": 8.14, "learning_rate": 3.5210443645455426e-08, "loss": 0.0081, "step": 20651 }, { "epoch": 8.14, "learning_rate": 3.519597625544184e-08, "loss": 0.0117, "step": 20652 }, { "epoch": 8.14, "learning_rate": 3.518151155148341e-08, "loss": 0.0087, "step": 20653 }, { "epoch": 8.14, "learning_rate": 3.516704953381579e-08, "loss": 0.0115, "step": 20654 }, { "epoch": 8.14, "learning_rate": 3.5152590202674826e-08, "loss": 0.0102, "step": 20655 }, { "epoch": 8.14, "learning_rate": 3.513813355829598e-08, "loss": 0.0133, "step": 20656 }, { "epoch": 8.14, "learning_rate": 3.5123679600915114e-08, "loss": 0.0111, "step": 20657 }, { "epoch": 8.14, "learning_rate": 3.510922833076762e-08, "loss": 0.0101, "step": 20658 }, { "epoch": 8.14, "learning_rate": 3.5094779748089146e-08, "loss": 0.0107, "step": 20659 }, { "epoch": 8.14, "learning_rate": 3.5080333853115085e-08, "loss": 0.0101, "step": 20660 }, { "epoch": 8.14, "learning_rate": 3.506589064608094e-08, "loss": 0.0126, "step": 20661 }, { "epoch": 8.14, "learning_rate": 3.5051450127222124e-08, "loss": 0.01, "step": 20662 }, { "epoch": 8.14, "learning_rate": 3.5037012296774e-08, "loss": 0.0118, "step": 20663 }, { "epoch": 8.14, "learning_rate": 3.5022577154971834e-08, "loss": 0.0108, "step": 20664 }, { "epoch": 8.14, "learning_rate": 3.5008144702050915e-08, "loss": 0.0092, "step": 20665 }, { "epoch": 8.14, "learning_rate": 3.4993714938246455e-08, "loss": 0.0104, "step": 20666 }, { "epoch": 8.15, "learning_rate": 3.497928786379374e-08, "loss": 0.0092, "step": 20667 }, { "epoch": 8.15, "learning_rate": 3.4964863478927776e-08, "loss": 0.008, "step": 20668 }, { "epoch": 8.15, "learning_rate": 3.495044178388375e-08, "loss": 0.0118, "step": 20669 }, { "epoch": 8.15, "learning_rate": 3.493602277889662e-08, "loss": 0.0083, "step": 20670 }, { "epoch": 8.15, "learning_rate": 3.492160646420148e-08, "loss": 0.0099, "step": 20671 }, { "epoch": 8.15, "learning_rate": 3.4907192840033296e-08, "loss": 0.012, "step": 20672 }, { "epoch": 8.15, "learning_rate": 3.489278190662692e-08, "loss": 0.0094, "step": 20673 }, { "epoch": 8.15, "learning_rate": 3.4878373664217243e-08, "loss": 0.0098, "step": 20674 }, { "epoch": 8.15, "learning_rate": 3.486396811303913e-08, "loss": 0.0108, "step": 20675 }, { "epoch": 8.15, "learning_rate": 3.4849565253327406e-08, "loss": 0.0147, "step": 20676 }, { "epoch": 8.15, "learning_rate": 3.483516508531672e-08, "loss": 0.0093, "step": 20677 }, { "epoch": 8.15, "learning_rate": 3.482076760924189e-08, "loss": 0.0121, "step": 20678 }, { "epoch": 8.15, "learning_rate": 3.480637282533738e-08, "loss": 0.0109, "step": 20679 }, { "epoch": 8.15, "learning_rate": 3.4791980733838043e-08, "loss": 0.0082, "step": 20680 }, { "epoch": 8.15, "learning_rate": 3.4777591334978264e-08, "loss": 0.0098, "step": 20681 }, { "epoch": 8.15, "learning_rate": 3.4763204628992714e-08, "loss": 0.0077, "step": 20682 }, { "epoch": 8.15, "learning_rate": 3.4748820616115683e-08, "loss": 0.0094, "step": 20683 }, { "epoch": 8.15, "learning_rate": 3.473443929658182e-08, "loss": 0.012, "step": 20684 }, { "epoch": 8.15, "learning_rate": 3.472006067062536e-08, "loss": 0.0095, "step": 20685 }, { "epoch": 8.15, "learning_rate": 3.470568473848079e-08, "loss": 0.0128, "step": 20686 }, { "epoch": 8.15, "learning_rate": 3.469131150038227e-08, "loss": 0.0103, "step": 20687 }, { "epoch": 8.15, "learning_rate": 3.467694095656415e-08, "loss": 0.0081, "step": 20688 }, { "epoch": 8.15, "learning_rate": 3.466257310726062e-08, "loss": 0.0108, "step": 20689 }, { "epoch": 8.15, "learning_rate": 3.464820795270591e-08, "loss": 0.009, "step": 20690 }, { "epoch": 8.15, "learning_rate": 3.463384549313406e-08, "loss": 0.0123, "step": 20691 }, { "epoch": 8.15, "learning_rate": 3.461948572877919e-08, "loss": 0.0131, "step": 20692 }, { "epoch": 8.16, "learning_rate": 3.460512865987544e-08, "loss": 0.0113, "step": 20693 }, { "epoch": 8.16, "learning_rate": 3.4590774286656645e-08, "loss": 0.0101, "step": 20694 }, { "epoch": 8.16, "learning_rate": 3.457642260935687e-08, "loss": 0.0089, "step": 20695 }, { "epoch": 8.16, "learning_rate": 3.456207362820991e-08, "loss": 0.0089, "step": 20696 }, { "epoch": 8.16, "learning_rate": 3.454772734344982e-08, "loss": 0.0114, "step": 20697 }, { "epoch": 8.16, "learning_rate": 3.453338375531025e-08, "loss": 0.0107, "step": 20698 }, { "epoch": 8.16, "learning_rate": 3.4519042864025116e-08, "loss": 0.0088, "step": 20699 }, { "epoch": 8.16, "learning_rate": 3.4504704669828025e-08, "loss": 0.0089, "step": 20700 }, { "epoch": 8.16, "learning_rate": 3.4490369172952695e-08, "loss": 0.0093, "step": 20701 }, { "epoch": 8.16, "learning_rate": 3.447603637363283e-08, "loss": 0.0111, "step": 20702 }, { "epoch": 8.16, "learning_rate": 3.4461706272102055e-08, "loss": 0.0124, "step": 20703 }, { "epoch": 8.16, "learning_rate": 3.444737886859381e-08, "loss": 0.0093, "step": 20704 }, { "epoch": 8.16, "learning_rate": 3.44330541633417e-08, "loss": 0.0103, "step": 20705 }, { "epoch": 8.16, "learning_rate": 3.4418732156579155e-08, "loss": 0.009, "step": 20706 }, { "epoch": 8.16, "learning_rate": 3.44044128485397e-08, "loss": 0.01, "step": 20707 }, { "epoch": 8.16, "learning_rate": 3.439009623945657e-08, "loss": 0.0093, "step": 20708 }, { "epoch": 8.16, "learning_rate": 3.437578232956322e-08, "loss": 0.0123, "step": 20709 }, { "epoch": 8.16, "learning_rate": 3.4361471119092846e-08, "loss": 0.0087, "step": 20710 }, { "epoch": 8.16, "learning_rate": 3.4347162608278746e-08, "loss": 0.0072, "step": 20711 }, { "epoch": 8.16, "learning_rate": 3.433285679735416e-08, "loss": 0.0103, "step": 20712 }, { "epoch": 8.16, "learning_rate": 3.4318553686552256e-08, "loss": 0.0102, "step": 20713 }, { "epoch": 8.16, "learning_rate": 3.4304253276106064e-08, "loss": 0.0101, "step": 20714 }, { "epoch": 8.16, "learning_rate": 3.428995556624872e-08, "loss": 0.0113, "step": 20715 }, { "epoch": 8.16, "learning_rate": 3.427566055721329e-08, "loss": 0.0096, "step": 20716 }, { "epoch": 8.16, "learning_rate": 3.426136824923269e-08, "loss": 0.0101, "step": 20717 }, { "epoch": 8.17, "learning_rate": 3.424707864253995e-08, "loss": 0.009, "step": 20718 }, { "epoch": 8.17, "learning_rate": 3.423279173736779e-08, "loss": 0.0116, "step": 20719 }, { "epoch": 8.17, "learning_rate": 3.42185075339493e-08, "loss": 0.01, "step": 20720 }, { "epoch": 8.17, "learning_rate": 3.420422603251711e-08, "loss": 0.0113, "step": 20721 }, { "epoch": 8.17, "learning_rate": 3.418994723330413e-08, "loss": 0.0115, "step": 20722 }, { "epoch": 8.17, "learning_rate": 3.4175671136542894e-08, "loss": 0.0101, "step": 20723 }, { "epoch": 8.17, "learning_rate": 3.4161397742466295e-08, "loss": 0.0086, "step": 20724 }, { "epoch": 8.17, "learning_rate": 3.414712705130682e-08, "loss": 0.01, "step": 20725 }, { "epoch": 8.17, "learning_rate": 3.413285906329717e-08, "loss": 0.0121, "step": 20726 }, { "epoch": 8.17, "learning_rate": 3.411859377866977e-08, "loss": 0.0115, "step": 20727 }, { "epoch": 8.17, "learning_rate": 3.4104331197657165e-08, "loss": 0.0084, "step": 20728 }, { "epoch": 8.17, "learning_rate": 3.4090071320491845e-08, "loss": 0.0136, "step": 20729 }, { "epoch": 8.17, "learning_rate": 3.407581414740626e-08, "loss": 0.0091, "step": 20730 }, { "epoch": 8.17, "learning_rate": 3.4061559678632665e-08, "loss": 0.0101, "step": 20731 }, { "epoch": 8.17, "learning_rate": 3.404730791440347e-08, "loss": 0.0131, "step": 20732 }, { "epoch": 8.17, "learning_rate": 3.4033058854950955e-08, "loss": 0.0076, "step": 20733 }, { "epoch": 8.17, "learning_rate": 3.401881250050731e-08, "loss": 0.0101, "step": 20734 }, { "epoch": 8.17, "learning_rate": 3.400456885130476e-08, "loss": 0.0123, "step": 20735 }, { "epoch": 8.17, "learning_rate": 3.399032790757546e-08, "loss": 0.0109, "step": 20736 }, { "epoch": 8.17, "learning_rate": 3.3976089669551544e-08, "loss": 0.0128, "step": 20737 }, { "epoch": 8.17, "learning_rate": 3.3961854137464993e-08, "loss": 0.0102, "step": 20738 }, { "epoch": 8.17, "learning_rate": 3.394762131154794e-08, "loss": 0.01, "step": 20739 }, { "epoch": 8.17, "learning_rate": 3.393339119203222e-08, "loss": 0.0089, "step": 20740 }, { "epoch": 8.17, "learning_rate": 3.391916377914983e-08, "loss": 0.0116, "step": 20741 }, { "epoch": 8.17, "learning_rate": 3.3904939073132655e-08, "loss": 0.0114, "step": 20742 }, { "epoch": 8.17, "learning_rate": 3.389071707421261e-08, "loss": 0.0074, "step": 20743 }, { "epoch": 8.18, "learning_rate": 3.387649778262134e-08, "loss": 0.0078, "step": 20744 }, { "epoch": 8.18, "learning_rate": 3.386228119859069e-08, "loss": 0.0108, "step": 20745 }, { "epoch": 8.18, "learning_rate": 3.384806732235237e-08, "loss": 0.0079, "step": 20746 }, { "epoch": 8.18, "learning_rate": 3.383385615413805e-08, "loss": 0.0115, "step": 20747 }, { "epoch": 8.18, "learning_rate": 3.381964769417931e-08, "loss": 0.0098, "step": 20748 }, { "epoch": 8.18, "learning_rate": 3.380544194270778e-08, "loss": 0.0111, "step": 20749 }, { "epoch": 8.18, "learning_rate": 3.3791238899954875e-08, "loss": 0.0092, "step": 20750 }, { "epoch": 8.18, "learning_rate": 3.377703856615226e-08, "loss": 0.0089, "step": 20751 }, { "epoch": 8.18, "learning_rate": 3.3762840941531236e-08, "loss": 0.0112, "step": 20752 }, { "epoch": 8.18, "learning_rate": 3.37486460263233e-08, "loss": 0.0092, "step": 20753 }, { "epoch": 8.18, "learning_rate": 3.3734453820759724e-08, "loss": 0.0111, "step": 20754 }, { "epoch": 8.18, "learning_rate": 3.372026432507185e-08, "loss": 0.0105, "step": 20755 }, { "epoch": 8.18, "learning_rate": 3.3706077539490927e-08, "loss": 0.0091, "step": 20756 }, { "epoch": 8.18, "learning_rate": 3.369189346424828e-08, "loss": 0.0094, "step": 20757 }, { "epoch": 8.18, "learning_rate": 3.367771209957495e-08, "loss": 0.0091, "step": 20758 }, { "epoch": 8.18, "learning_rate": 3.366353344570214e-08, "loss": 0.0124, "step": 20759 }, { "epoch": 8.18, "learning_rate": 3.364935750286098e-08, "loss": 0.0084, "step": 20760 }, { "epoch": 8.18, "learning_rate": 3.363518427128243e-08, "loss": 0.0123, "step": 20761 }, { "epoch": 8.18, "learning_rate": 3.362101375119755e-08, "loss": 0.0079, "step": 20762 }, { "epoch": 8.18, "learning_rate": 3.36068459428372e-08, "loss": 0.009, "step": 20763 }, { "epoch": 8.18, "learning_rate": 3.359268084643247e-08, "loss": 0.0096, "step": 20764 }, { "epoch": 8.18, "learning_rate": 3.357851846221409e-08, "loss": 0.012, "step": 20765 }, { "epoch": 8.18, "learning_rate": 3.3564358790412974e-08, "loss": 0.0102, "step": 20766 }, { "epoch": 8.18, "learning_rate": 3.3550201831259785e-08, "loss": 0.0103, "step": 20767 }, { "epoch": 8.18, "learning_rate": 3.353604758498536e-08, "loss": 0.0105, "step": 20768 }, { "epoch": 8.19, "learning_rate": 3.3521896051820345e-08, "loss": 0.009, "step": 20769 }, { "epoch": 8.19, "learning_rate": 3.3507747231995475e-08, "loss": 0.0113, "step": 20770 }, { "epoch": 8.19, "learning_rate": 3.3493601125741224e-08, "loss": 0.0132, "step": 20771 }, { "epoch": 8.19, "learning_rate": 3.3479457733288216e-08, "loss": 0.013, "step": 20772 }, { "epoch": 8.19, "learning_rate": 3.346531705486695e-08, "loss": 0.0096, "step": 20773 }, { "epoch": 8.19, "learning_rate": 3.3451179090707984e-08, "loss": 0.0105, "step": 20774 }, { "epoch": 8.19, "learning_rate": 3.3437043841041626e-08, "loss": 0.0083, "step": 20775 }, { "epoch": 8.19, "learning_rate": 3.3422911306098334e-08, "loss": 0.012, "step": 20776 }, { "epoch": 8.19, "learning_rate": 3.3408781486108395e-08, "loss": 0.0098, "step": 20777 }, { "epoch": 8.19, "learning_rate": 3.339465438130211e-08, "loss": 0.0102, "step": 20778 }, { "epoch": 8.19, "learning_rate": 3.3380529991909745e-08, "loss": 0.0087, "step": 20779 }, { "epoch": 8.19, "learning_rate": 3.3366408318161576e-08, "loss": 0.0118, "step": 20780 }, { "epoch": 8.19, "learning_rate": 3.3352289360287644e-08, "loss": 0.0104, "step": 20781 }, { "epoch": 8.19, "learning_rate": 3.33381731185181e-08, "loss": 0.0093, "step": 20782 }, { "epoch": 8.19, "learning_rate": 3.332405959308311e-08, "loss": 0.0091, "step": 20783 }, { "epoch": 8.19, "learning_rate": 3.330994878421256e-08, "loss": 0.0125, "step": 20784 }, { "epoch": 8.19, "learning_rate": 3.329584069213653e-08, "loss": 0.0086, "step": 20785 }, { "epoch": 8.19, "learning_rate": 3.3281735317084916e-08, "loss": 0.0108, "step": 20786 }, { "epoch": 8.19, "learning_rate": 3.326763265928769e-08, "loss": 0.0094, "step": 20787 }, { "epoch": 8.19, "learning_rate": 3.32535327189746e-08, "loss": 0.0097, "step": 20788 }, { "epoch": 8.19, "learning_rate": 3.3239435496375555e-08, "loss": 0.0096, "step": 20789 }, { "epoch": 8.19, "learning_rate": 3.322534099172016e-08, "loss": 0.0078, "step": 20790 }, { "epoch": 8.19, "learning_rate": 3.321124920523835e-08, "loss": 0.014, "step": 20791 }, { "epoch": 8.19, "learning_rate": 3.319716013715961e-08, "loss": 0.0083, "step": 20792 }, { "epoch": 8.19, "learning_rate": 3.318307378771372e-08, "loss": 0.0093, "step": 20793 }, { "epoch": 8.2, "learning_rate": 3.3168990157130085e-08, "loss": 0.0121, "step": 20794 }, { "epoch": 8.2, "learning_rate": 3.3154909245638465e-08, "loss": 0.0109, "step": 20795 }, { "epoch": 8.2, "learning_rate": 3.314083105346821e-08, "loss": 0.0127, "step": 20796 }, { "epoch": 8.2, "learning_rate": 3.312675558084883e-08, "loss": 0.0098, "step": 20797 }, { "epoch": 8.2, "learning_rate": 3.31126828280097e-08, "loss": 0.0081, "step": 20798 }, { "epoch": 8.2, "learning_rate": 3.3098612795180184e-08, "loss": 0.0112, "step": 20799 }, { "epoch": 8.2, "learning_rate": 3.3084545482589655e-08, "loss": 0.008, "step": 20800 }, { "epoch": 8.2, "learning_rate": 3.307048089046731e-08, "loss": 0.012, "step": 20801 }, { "epoch": 8.2, "learning_rate": 3.305641901904244e-08, "loss": 0.0108, "step": 20802 }, { "epoch": 8.2, "learning_rate": 3.30423598685442e-08, "loss": 0.0119, "step": 20803 }, { "epoch": 8.2, "learning_rate": 3.30283034392018e-08, "loss": 0.0087, "step": 20804 }, { "epoch": 8.2, "learning_rate": 3.301424973124425e-08, "loss": 0.0086, "step": 20805 }, { "epoch": 8.2, "learning_rate": 3.3000198744900674e-08, "loss": 0.0108, "step": 20806 }, { "epoch": 8.2, "learning_rate": 3.298615048039999e-08, "loss": 0.0103, "step": 20807 }, { "epoch": 8.2, "learning_rate": 3.2972104937971244e-08, "loss": 0.0099, "step": 20808 }, { "epoch": 8.2, "learning_rate": 3.295806211784331e-08, "loss": 0.0092, "step": 20809 }, { "epoch": 8.2, "learning_rate": 3.2944022020245154e-08, "loss": 0.0074, "step": 20810 }, { "epoch": 8.2, "learning_rate": 3.292998464540551e-08, "loss": 0.0108, "step": 20811 }, { "epoch": 8.2, "learning_rate": 3.2915949993553185e-08, "loss": 0.0108, "step": 20812 }, { "epoch": 8.2, "learning_rate": 3.290191806491691e-08, "loss": 0.0092, "step": 20813 }, { "epoch": 8.2, "learning_rate": 3.2887888859725485e-08, "loss": 0.0073, "step": 20814 }, { "epoch": 8.2, "learning_rate": 3.287386237820744e-08, "loss": 0.012, "step": 20815 }, { "epoch": 8.2, "learning_rate": 3.285983862059149e-08, "loss": 0.0122, "step": 20816 }, { "epoch": 8.2, "learning_rate": 3.284581758710605e-08, "loss": 0.0109, "step": 20817 }, { "epoch": 8.2, "learning_rate": 3.283179927797983e-08, "loss": 0.01, "step": 20818 }, { "epoch": 8.2, "learning_rate": 3.281778369344115e-08, "loss": 0.0094, "step": 20819 }, { "epoch": 8.21, "learning_rate": 3.280377083371859e-08, "loss": 0.0139, "step": 20820 }, { "epoch": 8.21, "learning_rate": 3.2789760699040384e-08, "loss": 0.0123, "step": 20821 }, { "epoch": 8.21, "learning_rate": 3.2775753289634955e-08, "loss": 0.0111, "step": 20822 }, { "epoch": 8.21, "learning_rate": 3.276174860573064e-08, "loss": 0.0106, "step": 20823 }, { "epoch": 8.21, "learning_rate": 3.27477466475556e-08, "loss": 0.0116, "step": 20824 }, { "epoch": 8.21, "learning_rate": 3.2733747415338095e-08, "loss": 0.0078, "step": 20825 }, { "epoch": 8.21, "learning_rate": 3.27197509093063e-08, "loss": 0.0096, "step": 20826 }, { "epoch": 8.21, "learning_rate": 3.2705757129688375e-08, "loss": 0.0093, "step": 20827 }, { "epoch": 8.21, "learning_rate": 3.269176607671231e-08, "loss": 0.0101, "step": 20828 }, { "epoch": 8.21, "learning_rate": 3.267777775060621e-08, "loss": 0.011, "step": 20829 }, { "epoch": 8.21, "learning_rate": 3.266379215159796e-08, "loss": 0.0108, "step": 20830 }, { "epoch": 8.21, "learning_rate": 3.264980927991565e-08, "loss": 0.009, "step": 20831 }, { "epoch": 8.21, "learning_rate": 3.263582913578706e-08, "loss": 0.0106, "step": 20832 }, { "epoch": 8.21, "learning_rate": 3.262185171944014e-08, "loss": 0.0139, "step": 20833 }, { "epoch": 8.21, "learning_rate": 3.2607877031102574e-08, "loss": 0.0105, "step": 20834 }, { "epoch": 8.21, "learning_rate": 3.259390507100228e-08, "loss": 0.0104, "step": 20835 }, { "epoch": 8.21, "learning_rate": 3.257993583936687e-08, "loss": 0.0089, "step": 20836 }, { "epoch": 8.21, "learning_rate": 3.2565969336424085e-08, "loss": 0.0104, "step": 20837 }, { "epoch": 8.21, "learning_rate": 3.25520055624015e-08, "loss": 0.0082, "step": 20838 }, { "epoch": 8.21, "learning_rate": 3.253804451752673e-08, "loss": 0.0096, "step": 20839 }, { "epoch": 8.21, "learning_rate": 3.252408620202731e-08, "loss": 0.0094, "step": 20840 }, { "epoch": 8.21, "learning_rate": 3.251013061613081e-08, "loss": 0.0133, "step": 20841 }, { "epoch": 8.21, "learning_rate": 3.249617776006459e-08, "loss": 0.0083, "step": 20842 }, { "epoch": 8.21, "learning_rate": 3.2482227634056084e-08, "loss": 0.0105, "step": 20843 }, { "epoch": 8.21, "learning_rate": 3.2468280238332704e-08, "loss": 0.0123, "step": 20844 }, { "epoch": 8.22, "learning_rate": 3.2454335573121694e-08, "loss": 0.0127, "step": 20845 }, { "epoch": 8.22, "learning_rate": 3.244039363865043e-08, "loss": 0.0098, "step": 20846 }, { "epoch": 8.22, "learning_rate": 3.2426454435146045e-08, "loss": 0.0102, "step": 20847 }, { "epoch": 8.22, "learning_rate": 3.2412517962835773e-08, "loss": 0.0104, "step": 20848 }, { "epoch": 8.22, "learning_rate": 3.2398584221946744e-08, "loss": 0.0091, "step": 20849 }, { "epoch": 8.22, "learning_rate": 3.2384653212706113e-08, "loss": 0.0107, "step": 20850 }, { "epoch": 8.22, "learning_rate": 3.2370724935340854e-08, "loss": 0.0104, "step": 20851 }, { "epoch": 8.22, "learning_rate": 3.2356799390078004e-08, "loss": 0.0118, "step": 20852 }, { "epoch": 8.22, "learning_rate": 3.234287657714452e-08, "loss": 0.0108, "step": 20853 }, { "epoch": 8.22, "learning_rate": 3.232895649676741e-08, "loss": 0.0117, "step": 20854 }, { "epoch": 8.22, "learning_rate": 3.231503914917342e-08, "loss": 0.01, "step": 20855 }, { "epoch": 8.22, "learning_rate": 3.2301124534589485e-08, "loss": 0.0088, "step": 20856 }, { "epoch": 8.22, "learning_rate": 3.2287212653242256e-08, "loss": 0.0124, "step": 20857 }, { "epoch": 8.22, "learning_rate": 3.2273303505358664e-08, "loss": 0.0106, "step": 20858 }, { "epoch": 8.22, "learning_rate": 3.225939709116525e-08, "loss": 0.0116, "step": 20859 }, { "epoch": 8.22, "learning_rate": 3.2245493410888804e-08, "loss": 0.0126, "step": 20860 }, { "epoch": 8.22, "learning_rate": 3.223159246475578e-08, "loss": 0.0086, "step": 20861 }, { "epoch": 8.22, "learning_rate": 3.221769425299283e-08, "loss": 0.013, "step": 20862 }, { "epoch": 8.22, "learning_rate": 3.2203798775826464e-08, "loss": 0.009, "step": 20863 }, { "epoch": 8.22, "learning_rate": 3.2189906033483215e-08, "loss": 0.009, "step": 20864 }, { "epoch": 8.22, "learning_rate": 3.21760160261894e-08, "loss": 0.0124, "step": 20865 }, { "epoch": 8.22, "learning_rate": 3.2162128754171456e-08, "loss": 0.0114, "step": 20866 }, { "epoch": 8.22, "learning_rate": 3.2148244217655786e-08, "loss": 0.0095, "step": 20867 }, { "epoch": 8.22, "learning_rate": 3.213436241686858e-08, "loss": 0.0083, "step": 20868 }, { "epoch": 8.22, "learning_rate": 3.212048335203617e-08, "loss": 0.0082, "step": 20869 }, { "epoch": 8.23, "learning_rate": 3.2106607023384655e-08, "loss": 0.0109, "step": 20870 }, { "epoch": 8.23, "learning_rate": 3.2092733431140365e-08, "loss": 0.0103, "step": 20871 }, { "epoch": 8.23, "learning_rate": 3.2078862575529296e-08, "loss": 0.0097, "step": 20872 }, { "epoch": 8.23, "learning_rate": 3.206499445677759e-08, "loss": 0.0146, "step": 20873 }, { "epoch": 8.23, "learning_rate": 3.205112907511116e-08, "loss": 0.0082, "step": 20874 }, { "epoch": 8.23, "learning_rate": 3.203726643075617e-08, "loss": 0.0101, "step": 20875 }, { "epoch": 8.23, "learning_rate": 3.20234065239384e-08, "loss": 0.011, "step": 20876 }, { "epoch": 8.23, "learning_rate": 3.200954935488387e-08, "loss": 0.0115, "step": 20877 }, { "epoch": 8.23, "learning_rate": 3.19956949238183e-08, "loss": 0.0102, "step": 20878 }, { "epoch": 8.23, "learning_rate": 3.1981843230967595e-08, "loss": 0.0092, "step": 20879 }, { "epoch": 8.23, "learning_rate": 3.196799427655747e-08, "loss": 0.01, "step": 20880 }, { "epoch": 8.23, "learning_rate": 3.1954148060813713e-08, "loss": 0.0102, "step": 20881 }, { "epoch": 8.23, "learning_rate": 3.194030458396188e-08, "loss": 0.0114, "step": 20882 }, { "epoch": 8.23, "learning_rate": 3.192646384622768e-08, "loss": 0.0114, "step": 20883 }, { "epoch": 8.23, "learning_rate": 3.1912625847836675e-08, "loss": 0.0097, "step": 20884 }, { "epoch": 8.23, "learning_rate": 3.1898790589014435e-08, "loss": 0.009, "step": 20885 }, { "epoch": 8.23, "learning_rate": 3.188495806998639e-08, "loss": 0.0105, "step": 20886 }, { "epoch": 8.23, "learning_rate": 3.187112829097807e-08, "loss": 0.0097, "step": 20887 }, { "epoch": 8.23, "learning_rate": 3.18573012522148e-08, "loss": 0.0123, "step": 20888 }, { "epoch": 8.23, "learning_rate": 3.1843476953921956e-08, "loss": 0.0098, "step": 20889 }, { "epoch": 8.23, "learning_rate": 3.1829655396324916e-08, "loss": 0.0074, "step": 20890 }, { "epoch": 8.23, "learning_rate": 3.181583657964886e-08, "loss": 0.0117, "step": 20891 }, { "epoch": 8.23, "learning_rate": 3.1802020504119046e-08, "loss": 0.0112, "step": 20892 }, { "epoch": 8.23, "learning_rate": 3.178820716996067e-08, "loss": 0.0105, "step": 20893 }, { "epoch": 8.23, "learning_rate": 3.177439657739893e-08, "loss": 0.0091, "step": 20894 }, { "epoch": 8.23, "learning_rate": 3.176058872665879e-08, "loss": 0.0124, "step": 20895 }, { "epoch": 8.24, "learning_rate": 3.1746783617965413e-08, "loss": 0.0092, "step": 20896 }, { "epoch": 8.24, "learning_rate": 3.1732981251543647e-08, "loss": 0.0094, "step": 20897 }, { "epoch": 8.24, "learning_rate": 3.171918162761864e-08, "loss": 0.0098, "step": 20898 }, { "epoch": 8.24, "learning_rate": 3.1705384746415176e-08, "loss": 0.0104, "step": 20899 }, { "epoch": 8.24, "learning_rate": 3.16915906081582e-08, "loss": 0.0106, "step": 20900 }, { "epoch": 8.24, "learning_rate": 3.167779921307241e-08, "loss": 0.0117, "step": 20901 }, { "epoch": 8.24, "learning_rate": 3.166401056138277e-08, "loss": 0.0131, "step": 20902 }, { "epoch": 8.24, "learning_rate": 3.1650224653313886e-08, "loss": 0.0079, "step": 20903 }, { "epoch": 8.24, "learning_rate": 3.1636441489090506e-08, "loss": 0.0121, "step": 20904 }, { "epoch": 8.24, "learning_rate": 3.1622661068937185e-08, "loss": 0.0082, "step": 20905 }, { "epoch": 8.24, "learning_rate": 3.160888339307861e-08, "loss": 0.0107, "step": 20906 }, { "epoch": 8.24, "learning_rate": 3.159510846173929e-08, "loss": 0.011, "step": 20907 }, { "epoch": 8.24, "learning_rate": 3.158133627514381e-08, "loss": 0.0104, "step": 20908 }, { "epoch": 8.24, "learning_rate": 3.1567566833516554e-08, "loss": 0.0126, "step": 20909 }, { "epoch": 8.24, "learning_rate": 3.1553800137081954e-08, "loss": 0.0116, "step": 20910 }, { "epoch": 8.24, "learning_rate": 3.1540036186064447e-08, "loss": 0.0093, "step": 20911 }, { "epoch": 8.24, "learning_rate": 3.152627498068827e-08, "loss": 0.0092, "step": 20912 }, { "epoch": 8.24, "learning_rate": 3.1512516521177814e-08, "loss": 0.0094, "step": 20913 }, { "epoch": 8.24, "learning_rate": 3.1498760807757174e-08, "loss": 0.0132, "step": 20914 }, { "epoch": 8.24, "learning_rate": 3.148500784065073e-08, "loss": 0.011, "step": 20915 }, { "epoch": 8.24, "learning_rate": 3.1471257620082516e-08, "loss": 0.01, "step": 20916 }, { "epoch": 8.24, "learning_rate": 3.145751014627671e-08, "loss": 0.0104, "step": 20917 }, { "epoch": 8.24, "learning_rate": 3.144376541945728e-08, "loss": 0.0127, "step": 20918 }, { "epoch": 8.24, "learning_rate": 3.143002343984828e-08, "loss": 0.0118, "step": 20919 }, { "epoch": 8.24, "learning_rate": 3.1416284207673724e-08, "loss": 0.0083, "step": 20920 }, { "epoch": 8.25, "learning_rate": 3.140254772315756e-08, "loss": 0.0112, "step": 20921 }, { "epoch": 8.25, "learning_rate": 3.138881398652358e-08, "loss": 0.0127, "step": 20922 }, { "epoch": 8.25, "learning_rate": 3.137508299799567e-08, "loss": 0.0099, "step": 20923 }, { "epoch": 8.25, "learning_rate": 3.136135475779764e-08, "loss": 0.0113, "step": 20924 }, { "epoch": 8.25, "learning_rate": 3.1347629266153265e-08, "loss": 0.0103, "step": 20925 }, { "epoch": 8.25, "learning_rate": 3.133390652328616e-08, "loss": 0.0078, "step": 20926 }, { "epoch": 8.25, "learning_rate": 3.132018652942008e-08, "loss": 0.0087, "step": 20927 }, { "epoch": 8.25, "learning_rate": 3.130646928477856e-08, "loss": 0.0108, "step": 20928 }, { "epoch": 8.25, "learning_rate": 3.1292754789585196e-08, "loss": 0.0106, "step": 20929 }, { "epoch": 8.25, "learning_rate": 3.127904304406352e-08, "loss": 0.0096, "step": 20930 }, { "epoch": 8.25, "learning_rate": 3.126533404843707e-08, "loss": 0.0085, "step": 20931 }, { "epoch": 8.25, "learning_rate": 3.1251627802929184e-08, "loss": 0.0113, "step": 20932 }, { "epoch": 8.25, "learning_rate": 3.123792430776331e-08, "loss": 0.0083, "step": 20933 }, { "epoch": 8.25, "learning_rate": 3.122422356316281e-08, "loss": 0.0117, "step": 20934 }, { "epoch": 8.25, "learning_rate": 3.121052556935091e-08, "loss": 0.0112, "step": 20935 }, { "epoch": 8.25, "learning_rate": 3.119683032655096e-08, "loss": 0.012, "step": 20936 }, { "epoch": 8.25, "learning_rate": 3.118313783498603e-08, "loss": 0.0116, "step": 20937 }, { "epoch": 8.25, "learning_rate": 3.116944809487947e-08, "loss": 0.0105, "step": 20938 }, { "epoch": 8.25, "learning_rate": 3.1155761106454285e-08, "loss": 0.0132, "step": 20939 }, { "epoch": 8.25, "learning_rate": 3.11420768699336e-08, "loss": 0.011, "step": 20940 }, { "epoch": 8.25, "learning_rate": 3.112839538554035e-08, "loss": 0.0102, "step": 20941 }, { "epoch": 8.25, "learning_rate": 3.11147166534977e-08, "loss": 0.012, "step": 20942 }, { "epoch": 8.25, "learning_rate": 3.1101040674028434e-08, "loss": 0.011, "step": 20943 }, { "epoch": 8.25, "learning_rate": 3.108736744735556e-08, "loss": 0.0118, "step": 20944 }, { "epoch": 8.25, "learning_rate": 3.107369697370179e-08, "loss": 0.0101, "step": 20945 }, { "epoch": 8.25, "learning_rate": 3.1060029253290096e-08, "loss": 0.0099, "step": 20946 }, { "epoch": 8.26, "learning_rate": 3.104636428634315e-08, "loss": 0.0101, "step": 20947 }, { "epoch": 8.26, "learning_rate": 3.1032702073083705e-08, "loss": 0.0092, "step": 20948 }, { "epoch": 8.26, "learning_rate": 3.101904261373438e-08, "loss": 0.0111, "step": 20949 }, { "epoch": 8.26, "learning_rate": 3.1005385908517844e-08, "loss": 0.0103, "step": 20950 }, { "epoch": 8.26, "learning_rate": 3.099173195765672e-08, "loss": 0.0106, "step": 20951 }, { "epoch": 8.26, "learning_rate": 3.097808076137345e-08, "loss": 0.0111, "step": 20952 }, { "epoch": 8.26, "learning_rate": 3.0964432319890586e-08, "loss": 0.0097, "step": 20953 }, { "epoch": 8.26, "learning_rate": 3.0950786633430556e-08, "loss": 0.0111, "step": 20954 }, { "epoch": 8.26, "learning_rate": 3.093714370221583e-08, "loss": 0.0092, "step": 20955 }, { "epoch": 8.26, "learning_rate": 3.0923503526468686e-08, "loss": 0.0099, "step": 20956 }, { "epoch": 8.26, "learning_rate": 3.0909866106411506e-08, "loss": 0.0108, "step": 20957 }, { "epoch": 8.26, "learning_rate": 3.089623144226647e-08, "loss": 0.0104, "step": 20958 }, { "epoch": 8.26, "learning_rate": 3.088259953425585e-08, "loss": 0.0084, "step": 20959 }, { "epoch": 8.26, "learning_rate": 3.086897038260183e-08, "loss": 0.012, "step": 20960 }, { "epoch": 8.26, "learning_rate": 3.0855343987526604e-08, "loss": 0.0143, "step": 20961 }, { "epoch": 8.26, "learning_rate": 3.084172034925214e-08, "loss": 0.0094, "step": 20962 }, { "epoch": 8.26, "learning_rate": 3.082809946800056e-08, "loss": 0.0134, "step": 20963 }, { "epoch": 8.26, "learning_rate": 3.081448134399382e-08, "loss": 0.0116, "step": 20964 }, { "epoch": 8.26, "learning_rate": 3.080086597745397e-08, "loss": 0.0111, "step": 20965 }, { "epoch": 8.26, "learning_rate": 3.078725336860279e-08, "loss": 0.0105, "step": 20966 }, { "epoch": 8.26, "learning_rate": 3.077364351766227e-08, "loss": 0.0092, "step": 20967 }, { "epoch": 8.26, "learning_rate": 3.0760036424854055e-08, "loss": 0.0103, "step": 20968 }, { "epoch": 8.26, "learning_rate": 3.074643209040013e-08, "loss": 0.0096, "step": 20969 }, { "epoch": 8.26, "learning_rate": 3.073283051452209e-08, "loss": 0.0099, "step": 20970 }, { "epoch": 8.26, "learning_rate": 3.0719231697441684e-08, "loss": 0.0099, "step": 20971 }, { "epoch": 8.27, "learning_rate": 3.070563563938049e-08, "loss": 0.0137, "step": 20972 }, { "epoch": 8.27, "learning_rate": 3.069204234056013e-08, "loss": 0.0108, "step": 20973 }, { "epoch": 8.27, "learning_rate": 3.067845180120223e-08, "loss": 0.0137, "step": 20974 }, { "epoch": 8.27, "learning_rate": 3.066486402152817e-08, "loss": 0.0091, "step": 20975 }, { "epoch": 8.27, "learning_rate": 3.0651279001759453e-08, "loss": 0.0139, "step": 20976 }, { "epoch": 8.27, "learning_rate": 3.0637696742117516e-08, "loss": 0.0105, "step": 20977 }, { "epoch": 8.27, "learning_rate": 3.0624117242823763e-08, "loss": 0.0088, "step": 20978 }, { "epoch": 8.27, "learning_rate": 3.061054050409944e-08, "loss": 0.012, "step": 20979 }, { "epoch": 8.27, "learning_rate": 3.0596966526165903e-08, "loss": 0.0096, "step": 20980 }, { "epoch": 8.27, "learning_rate": 3.058339530924425e-08, "loss": 0.0086, "step": 20981 }, { "epoch": 8.27, "learning_rate": 3.0569826853555894e-08, "loss": 0.0111, "step": 20982 }, { "epoch": 8.27, "learning_rate": 3.0556261159321775e-08, "loss": 0.0089, "step": 20983 }, { "epoch": 8.27, "learning_rate": 3.054269822676314e-08, "loss": 0.0094, "step": 20984 }, { "epoch": 8.27, "learning_rate": 3.0529138056100864e-08, "loss": 0.0103, "step": 20985 }, { "epoch": 8.27, "learning_rate": 3.05155806475562e-08, "loss": 0.0121, "step": 20986 }, { "epoch": 8.27, "learning_rate": 3.05020260013499e-08, "loss": 0.0089, "step": 20987 }, { "epoch": 8.27, "learning_rate": 3.048847411770303e-08, "loss": 0.0085, "step": 20988 }, { "epoch": 8.27, "learning_rate": 3.047492499683637e-08, "loss": 0.0095, "step": 20989 }, { "epoch": 8.27, "learning_rate": 3.0461378638970784e-08, "loss": 0.0104, "step": 20990 }, { "epoch": 8.27, "learning_rate": 3.044783504432704e-08, "loss": 0.0092, "step": 20991 }, { "epoch": 8.27, "learning_rate": 3.043429421312593e-08, "loss": 0.0108, "step": 20992 }, { "epoch": 8.27, "learning_rate": 3.04207561455881e-08, "loss": 0.0104, "step": 20993 }, { "epoch": 8.27, "learning_rate": 3.040722084193421e-08, "loss": 0.0118, "step": 20994 }, { "epoch": 8.27, "learning_rate": 3.039368830238489e-08, "loss": 0.0108, "step": 20995 }, { "epoch": 8.27, "learning_rate": 3.038015852716065e-08, "loss": 0.0103, "step": 20996 }, { "epoch": 8.28, "learning_rate": 3.0366631516482086e-08, "loss": 0.0103, "step": 20997 }, { "epoch": 8.28, "learning_rate": 3.0353107270569566e-08, "loss": 0.0126, "step": 20998 }, { "epoch": 8.28, "learning_rate": 3.033958578964355e-08, "loss": 0.01, "step": 20999 }, { "epoch": 8.28, "learning_rate": 3.0326067073924444e-08, "loss": 0.0112, "step": 21000 }, { "epoch": 8.28, "learning_rate": 3.0312551123632625e-08, "loss": 0.0114, "step": 21001 }, { "epoch": 8.28, "learning_rate": 3.029903793898827e-08, "loss": 0.0088, "step": 21002 }, { "epoch": 8.28, "learning_rate": 3.028552752021167e-08, "loss": 0.0104, "step": 21003 }, { "epoch": 8.28, "learning_rate": 3.0272019867523056e-08, "loss": 0.0114, "step": 21004 }, { "epoch": 8.28, "learning_rate": 3.025851498114258e-08, "loss": 0.0102, "step": 21005 }, { "epoch": 8.28, "learning_rate": 3.0245012861290306e-08, "loss": 0.0123, "step": 21006 }, { "epoch": 8.28, "learning_rate": 3.023151350818638e-08, "loss": 0.0102, "step": 21007 }, { "epoch": 8.28, "learning_rate": 3.0218016922050657e-08, "loss": 0.0084, "step": 21008 }, { "epoch": 8.28, "learning_rate": 3.020452310310331e-08, "loss": 0.0095, "step": 21009 }, { "epoch": 8.28, "learning_rate": 3.019103205156415e-08, "loss": 0.0084, "step": 21010 }, { "epoch": 8.28, "learning_rate": 3.017754376765311e-08, "loss": 0.0084, "step": 21011 }, { "epoch": 8.28, "learning_rate": 3.016405825158992e-08, "loss": 0.0119, "step": 21012 }, { "epoch": 8.28, "learning_rate": 3.015057550359457e-08, "loss": 0.0099, "step": 21013 }, { "epoch": 8.28, "learning_rate": 3.0137095523886636e-08, "loss": 0.0097, "step": 21014 }, { "epoch": 8.28, "learning_rate": 3.012361831268593e-08, "loss": 0.0102, "step": 21015 }, { "epoch": 8.28, "learning_rate": 3.011014387021202e-08, "loss": 0.0104, "step": 21016 }, { "epoch": 8.28, "learning_rate": 3.009667219668455e-08, "loss": 0.0136, "step": 21017 }, { "epoch": 8.28, "learning_rate": 3.008320329232317e-08, "loss": 0.0124, "step": 21018 }, { "epoch": 8.28, "learning_rate": 3.0069737157347265e-08, "loss": 0.0099, "step": 21019 }, { "epoch": 8.28, "learning_rate": 3.005627379197644e-08, "loss": 0.0093, "step": 21020 }, { "epoch": 8.28, "learning_rate": 3.004281319642996e-08, "loss": 0.0128, "step": 21021 }, { "epoch": 8.28, "learning_rate": 3.0029355370927435e-08, "loss": 0.0118, "step": 21022 }, { "epoch": 8.29, "learning_rate": 3.001590031568801e-08, "loss": 0.0109, "step": 21023 }, { "epoch": 8.29, "learning_rate": 3.000244803093115e-08, "loss": 0.0089, "step": 21024 }, { "epoch": 8.29, "learning_rate": 2.9988998516875907e-08, "loss": 0.0089, "step": 21025 }, { "epoch": 8.29, "learning_rate": 2.997555177374169e-08, "loss": 0.0097, "step": 21026 }, { "epoch": 8.29, "learning_rate": 2.996210780174753e-08, "loss": 0.0097, "step": 21027 }, { "epoch": 8.29, "learning_rate": 2.994866660111264e-08, "loss": 0.0097, "step": 21028 }, { "epoch": 8.29, "learning_rate": 2.993522817205598e-08, "loss": 0.0091, "step": 21029 }, { "epoch": 8.29, "learning_rate": 2.992179251479665e-08, "loss": 0.0109, "step": 21030 }, { "epoch": 8.29, "learning_rate": 2.990835962955358e-08, "loss": 0.0114, "step": 21031 }, { "epoch": 8.29, "learning_rate": 2.989492951654582e-08, "loss": 0.0087, "step": 21032 }, { "epoch": 8.29, "learning_rate": 2.98815021759921e-08, "loss": 0.0083, "step": 21033 }, { "epoch": 8.29, "learning_rate": 2.9868077608111364e-08, "loss": 0.0119, "step": 21034 }, { "epoch": 8.29, "learning_rate": 2.9854655813122384e-08, "loss": 0.0105, "step": 21035 }, { "epoch": 8.29, "learning_rate": 2.9841236791243974e-08, "loss": 0.0065, "step": 21036 }, { "epoch": 8.29, "learning_rate": 2.982782054269475e-08, "loss": 0.0138, "step": 21037 }, { "epoch": 8.29, "learning_rate": 2.981440706769347e-08, "loss": 0.0112, "step": 21038 }, { "epoch": 8.29, "learning_rate": 2.9800996366458652e-08, "loss": 0.0103, "step": 21039 }, { "epoch": 8.29, "learning_rate": 2.9787588439208923e-08, "loss": 0.0107, "step": 21040 }, { "epoch": 8.29, "learning_rate": 2.9774183286162836e-08, "loss": 0.0104, "step": 21041 }, { "epoch": 8.29, "learning_rate": 2.9760780907538817e-08, "loss": 0.0124, "step": 21042 }, { "epoch": 8.29, "learning_rate": 2.974738130355534e-08, "loss": 0.0099, "step": 21043 }, { "epoch": 8.29, "learning_rate": 2.9733984474430785e-08, "loss": 0.0091, "step": 21044 }, { "epoch": 8.29, "learning_rate": 2.972059042038355e-08, "loss": 0.0081, "step": 21045 }, { "epoch": 8.29, "learning_rate": 2.9707199141631866e-08, "loss": 0.0085, "step": 21046 }, { "epoch": 8.29, "learning_rate": 2.9693810638394045e-08, "loss": 0.0076, "step": 21047 }, { "epoch": 8.3, "learning_rate": 2.96804249108882e-08, "loss": 0.0093, "step": 21048 }, { "epoch": 8.3, "learning_rate": 2.966704195933267e-08, "loss": 0.011, "step": 21049 }, { "epoch": 8.3, "learning_rate": 2.9653661783945417e-08, "loss": 0.01, "step": 21050 }, { "epoch": 8.3, "learning_rate": 2.9640284384944635e-08, "loss": 0.0124, "step": 21051 }, { "epoch": 8.3, "learning_rate": 2.9626909762548202e-08, "loss": 0.0078, "step": 21052 }, { "epoch": 8.3, "learning_rate": 2.9613537916974318e-08, "loss": 0.0112, "step": 21053 }, { "epoch": 8.3, "learning_rate": 2.9600168848440743e-08, "loss": 0.0089, "step": 21054 }, { "epoch": 8.3, "learning_rate": 2.958680255716548e-08, "loss": 0.0105, "step": 21055 }, { "epoch": 8.3, "learning_rate": 2.9573439043366287e-08, "loss": 0.0108, "step": 21056 }, { "epoch": 8.3, "learning_rate": 2.956007830726104e-08, "loss": 0.0075, "step": 21057 }, { "epoch": 8.3, "learning_rate": 2.9546720349067466e-08, "loss": 0.0108, "step": 21058 }, { "epoch": 8.3, "learning_rate": 2.9533365169003333e-08, "loss": 0.0106, "step": 21059 }, { "epoch": 8.3, "learning_rate": 2.9520012767286218e-08, "loss": 0.01, "step": 21060 }, { "epoch": 8.3, "learning_rate": 2.9506663144133816e-08, "loss": 0.0098, "step": 21061 }, { "epoch": 8.3, "learning_rate": 2.9493316299763725e-08, "loss": 0.0139, "step": 21062 }, { "epoch": 8.3, "learning_rate": 2.947997223439338e-08, "loss": 0.0094, "step": 21063 }, { "epoch": 8.3, "learning_rate": 2.9466630948240377e-08, "loss": 0.0101, "step": 21064 }, { "epoch": 8.3, "learning_rate": 2.9453292441522015e-08, "loss": 0.0115, "step": 21065 }, { "epoch": 8.3, "learning_rate": 2.943995671445587e-08, "loss": 0.0111, "step": 21066 }, { "epoch": 8.3, "learning_rate": 2.9426623767259194e-08, "loss": 0.009, "step": 21067 }, { "epoch": 8.3, "learning_rate": 2.9413293600149324e-08, "loss": 0.0098, "step": 21068 }, { "epoch": 8.3, "learning_rate": 2.9399966213343463e-08, "loss": 0.0126, "step": 21069 }, { "epoch": 8.3, "learning_rate": 2.9386641607058883e-08, "loss": 0.0107, "step": 21070 }, { "epoch": 8.3, "learning_rate": 2.937331978151274e-08, "loss": 0.0124, "step": 21071 }, { "epoch": 8.3, "learning_rate": 2.936000073692222e-08, "loss": 0.0103, "step": 21072 }, { "epoch": 8.31, "learning_rate": 2.934668447350428e-08, "loss": 0.0095, "step": 21073 }, { "epoch": 8.31, "learning_rate": 2.9333370991476014e-08, "loss": 0.0104, "step": 21074 }, { "epoch": 8.31, "learning_rate": 2.9320060291054427e-08, "loss": 0.0125, "step": 21075 }, { "epoch": 8.31, "learning_rate": 2.93067523724565e-08, "loss": 0.0103, "step": 21076 }, { "epoch": 8.31, "learning_rate": 2.929344723589906e-08, "loss": 0.0085, "step": 21077 }, { "epoch": 8.31, "learning_rate": 2.9280144881599e-08, "loss": 0.0082, "step": 21078 }, { "epoch": 8.31, "learning_rate": 2.926684530977308e-08, "loss": 0.0084, "step": 21079 }, { "epoch": 8.31, "learning_rate": 2.9253548520638105e-08, "loss": 0.01, "step": 21080 }, { "epoch": 8.31, "learning_rate": 2.9240254514410766e-08, "loss": 0.0114, "step": 21081 }, { "epoch": 8.31, "learning_rate": 2.9226963291307828e-08, "loss": 0.0108, "step": 21082 }, { "epoch": 8.31, "learning_rate": 2.9213674851545778e-08, "loss": 0.0087, "step": 21083 }, { "epoch": 8.31, "learning_rate": 2.9200389195341248e-08, "loss": 0.0099, "step": 21084 }, { "epoch": 8.31, "learning_rate": 2.918710632291086e-08, "loss": 0.0115, "step": 21085 }, { "epoch": 8.31, "learning_rate": 2.9173826234470976e-08, "loss": 0.0108, "step": 21086 }, { "epoch": 8.31, "learning_rate": 2.916054893023814e-08, "loss": 0.0123, "step": 21087 }, { "epoch": 8.31, "learning_rate": 2.9147274410428634e-08, "loss": 0.0118, "step": 21088 }, { "epoch": 8.31, "learning_rate": 2.9134002675258962e-08, "loss": 0.0121, "step": 21089 }, { "epoch": 8.31, "learning_rate": 2.9120733724945325e-08, "loss": 0.009, "step": 21090 }, { "epoch": 8.31, "learning_rate": 2.910746755970408e-08, "loss": 0.0101, "step": 21091 }, { "epoch": 8.31, "learning_rate": 2.909420417975128e-08, "loss": 0.0124, "step": 21092 }, { "epoch": 8.31, "learning_rate": 2.9080943585303286e-08, "loss": 0.0086, "step": 21093 }, { "epoch": 8.31, "learning_rate": 2.9067685776576123e-08, "loss": 0.0116, "step": 21094 }, { "epoch": 8.31, "learning_rate": 2.905443075378593e-08, "loss": 0.009, "step": 21095 }, { "epoch": 8.31, "learning_rate": 2.904117851714867e-08, "loss": 0.0106, "step": 21096 }, { "epoch": 8.31, "learning_rate": 2.9027929066880343e-08, "loss": 0.0098, "step": 21097 }, { "epoch": 8.31, "learning_rate": 2.9014682403196954e-08, "loss": 0.0118, "step": 21098 }, { "epoch": 8.32, "learning_rate": 2.9001438526314404e-08, "loss": 0.0102, "step": 21099 }, { "epoch": 8.32, "learning_rate": 2.898819743644849e-08, "loss": 0.0089, "step": 21100 }, { "epoch": 8.32, "learning_rate": 2.8974959133815046e-08, "loss": 0.012, "step": 21101 }, { "epoch": 8.32, "learning_rate": 2.8961723618629874e-08, "loss": 0.0122, "step": 21102 }, { "epoch": 8.32, "learning_rate": 2.8948490891108624e-08, "loss": 0.0102, "step": 21103 }, { "epoch": 8.32, "learning_rate": 2.893526095146699e-08, "loss": 0.0109, "step": 21104 }, { "epoch": 8.32, "learning_rate": 2.892203379992062e-08, "loss": 0.008, "step": 21105 }, { "epoch": 8.32, "learning_rate": 2.8908809436685146e-08, "loss": 0.0114, "step": 21106 }, { "epoch": 8.32, "learning_rate": 2.889558786197599e-08, "loss": 0.0117, "step": 21107 }, { "epoch": 8.32, "learning_rate": 2.888236907600876e-08, "loss": 0.0119, "step": 21108 }, { "epoch": 8.32, "learning_rate": 2.8869153078998797e-08, "loss": 0.0102, "step": 21109 }, { "epoch": 8.32, "learning_rate": 2.885593987116155e-08, "loss": 0.0108, "step": 21110 }, { "epoch": 8.32, "learning_rate": 2.8842729452712378e-08, "loss": 0.0107, "step": 21111 }, { "epoch": 8.32, "learning_rate": 2.882952182386662e-08, "loss": 0.011, "step": 21112 }, { "epoch": 8.32, "learning_rate": 2.881631698483946e-08, "loss": 0.0095, "step": 21113 }, { "epoch": 8.32, "learning_rate": 2.8803114935846175e-08, "loss": 0.01, "step": 21114 }, { "epoch": 8.32, "learning_rate": 2.8789915677101916e-08, "loss": 0.0112, "step": 21115 }, { "epoch": 8.32, "learning_rate": 2.8776719208821875e-08, "loss": 0.0104, "step": 21116 }, { "epoch": 8.32, "learning_rate": 2.8763525531221057e-08, "loss": 0.0115, "step": 21117 }, { "epoch": 8.32, "learning_rate": 2.8750334644514553e-08, "loss": 0.0087, "step": 21118 }, { "epoch": 8.32, "learning_rate": 2.873714654891721e-08, "loss": 0.0116, "step": 21119 }, { "epoch": 8.32, "learning_rate": 2.8723961244644202e-08, "loss": 0.0123, "step": 21120 }, { "epoch": 8.32, "learning_rate": 2.8710778731910278e-08, "loss": 0.0112, "step": 21121 }, { "epoch": 8.32, "learning_rate": 2.8697599010930362e-08, "loss": 0.0097, "step": 21122 }, { "epoch": 8.32, "learning_rate": 2.868442208191919e-08, "loss": 0.0118, "step": 21123 }, { "epoch": 8.33, "learning_rate": 2.867124794509157e-08, "loss": 0.0105, "step": 21124 }, { "epoch": 8.33, "learning_rate": 2.8658076600662263e-08, "loss": 0.0102, "step": 21125 }, { "epoch": 8.33, "learning_rate": 2.8644908048845852e-08, "loss": 0.0126, "step": 21126 }, { "epoch": 8.33, "learning_rate": 2.863174228985701e-08, "loss": 0.0114, "step": 21127 }, { "epoch": 8.33, "learning_rate": 2.861857932391032e-08, "loss": 0.0092, "step": 21128 }, { "epoch": 8.33, "learning_rate": 2.8605419151220367e-08, "loss": 0.0122, "step": 21129 }, { "epoch": 8.33, "learning_rate": 2.8592261772001537e-08, "loss": 0.01, "step": 21130 }, { "epoch": 8.33, "learning_rate": 2.857910718646839e-08, "loss": 0.0129, "step": 21131 }, { "epoch": 8.33, "learning_rate": 2.8565955394835172e-08, "loss": 0.0105, "step": 21132 }, { "epoch": 8.33, "learning_rate": 2.855280639731641e-08, "loss": 0.0114, "step": 21133 }, { "epoch": 8.33, "learning_rate": 2.8539660194126303e-08, "loss": 0.0106, "step": 21134 }, { "epoch": 8.33, "learning_rate": 2.852651678547917e-08, "loss": 0.01, "step": 21135 }, { "epoch": 8.33, "learning_rate": 2.8513376171589178e-08, "loss": 0.009, "step": 21136 }, { "epoch": 8.33, "learning_rate": 2.8500238352670526e-08, "loss": 0.0121, "step": 21137 }, { "epoch": 8.33, "learning_rate": 2.8487103328937333e-08, "loss": 0.0106, "step": 21138 }, { "epoch": 8.33, "learning_rate": 2.8473971100603723e-08, "loss": 0.0096, "step": 21139 }, { "epoch": 8.33, "learning_rate": 2.846084166788365e-08, "loss": 0.0114, "step": 21140 }, { "epoch": 8.33, "learning_rate": 2.844771503099117e-08, "loss": 0.0109, "step": 21141 }, { "epoch": 8.33, "learning_rate": 2.8434591190140177e-08, "loss": 0.0081, "step": 21142 }, { "epoch": 8.33, "learning_rate": 2.8421470145544657e-08, "loss": 0.01, "step": 21143 }, { "epoch": 8.33, "learning_rate": 2.840835189741837e-08, "loss": 0.0104, "step": 21144 }, { "epoch": 8.33, "learning_rate": 2.8395236445975145e-08, "loss": 0.0105, "step": 21145 }, { "epoch": 8.33, "learning_rate": 2.8382123791428793e-08, "loss": 0.0091, "step": 21146 }, { "epoch": 8.33, "learning_rate": 2.8369013933992958e-08, "loss": 0.0115, "step": 21147 }, { "epoch": 8.33, "learning_rate": 2.835590687388141e-08, "loss": 0.0098, "step": 21148 }, { "epoch": 8.33, "learning_rate": 2.8342802611307636e-08, "loss": 0.0124, "step": 21149 }, { "epoch": 8.34, "learning_rate": 2.8329701146485296e-08, "loss": 0.0092, "step": 21150 }, { "epoch": 8.34, "learning_rate": 2.8316602479627927e-08, "loss": 0.0105, "step": 21151 }, { "epoch": 8.34, "learning_rate": 2.8303506610949067e-08, "loss": 0.0101, "step": 21152 }, { "epoch": 8.34, "learning_rate": 2.829041354066204e-08, "loss": 0.0115, "step": 21153 }, { "epoch": 8.34, "learning_rate": 2.8277323268980315e-08, "loss": 0.0082, "step": 21154 }, { "epoch": 8.34, "learning_rate": 2.8264235796117207e-08, "loss": 0.0115, "step": 21155 }, { "epoch": 8.34, "learning_rate": 2.825115112228611e-08, "loss": 0.0109, "step": 21156 }, { "epoch": 8.34, "learning_rate": 2.823806924770018e-08, "loss": 0.0113, "step": 21157 }, { "epoch": 8.34, "learning_rate": 2.8224990172572716e-08, "loss": 0.0136, "step": 21158 }, { "epoch": 8.34, "learning_rate": 2.8211913897116745e-08, "loss": 0.011, "step": 21159 }, { "epoch": 8.34, "learning_rate": 2.819884042154559e-08, "loss": 0.0085, "step": 21160 }, { "epoch": 8.34, "learning_rate": 2.8185769746072186e-08, "loss": 0.0121, "step": 21161 }, { "epoch": 8.34, "learning_rate": 2.8172701870909632e-08, "loss": 0.0112, "step": 21162 }, { "epoch": 8.34, "learning_rate": 2.8159636796270824e-08, "loss": 0.0101, "step": 21163 }, { "epoch": 8.34, "learning_rate": 2.814657452236884e-08, "loss": 0.0099, "step": 21164 }, { "epoch": 8.34, "learning_rate": 2.8133515049416478e-08, "loss": 0.0134, "step": 21165 }, { "epoch": 8.34, "learning_rate": 2.8120458377626644e-08, "loss": 0.0095, "step": 21166 }, { "epoch": 8.34, "learning_rate": 2.8107404507212075e-08, "loss": 0.0122, "step": 21167 }, { "epoch": 8.34, "learning_rate": 2.8094353438385555e-08, "loss": 0.0093, "step": 21168 }, { "epoch": 8.34, "learning_rate": 2.808130517135987e-08, "loss": 0.0106, "step": 21169 }, { "epoch": 8.34, "learning_rate": 2.8068259706347564e-08, "loss": 0.0097, "step": 21170 }, { "epoch": 8.34, "learning_rate": 2.8055217043561374e-08, "loss": 0.0088, "step": 21171 }, { "epoch": 8.34, "learning_rate": 2.804217718321371e-08, "loss": 0.0122, "step": 21172 }, { "epoch": 8.34, "learning_rate": 2.8029140125517336e-08, "loss": 0.0129, "step": 21173 }, { "epoch": 8.34, "learning_rate": 2.8016105870684547e-08, "loss": 0.0095, "step": 21174 }, { "epoch": 8.35, "learning_rate": 2.8003074418927884e-08, "loss": 0.0124, "step": 21175 }, { "epoch": 8.35, "learning_rate": 2.799004577045967e-08, "loss": 0.01, "step": 21176 }, { "epoch": 8.35, "learning_rate": 2.7977019925492285e-08, "loss": 0.0106, "step": 21177 }, { "epoch": 8.35, "learning_rate": 2.796399688423803e-08, "loss": 0.0127, "step": 21178 }, { "epoch": 8.35, "learning_rate": 2.7950976646909197e-08, "loss": 0.0111, "step": 21179 }, { "epoch": 8.35, "learning_rate": 2.7937959213717932e-08, "loss": 0.0103, "step": 21180 }, { "epoch": 8.35, "learning_rate": 2.792494458487642e-08, "loss": 0.0092, "step": 21181 }, { "epoch": 8.35, "learning_rate": 2.791193276059678e-08, "loss": 0.008, "step": 21182 }, { "epoch": 8.35, "learning_rate": 2.789892374109115e-08, "loss": 0.0096, "step": 21183 }, { "epoch": 8.35, "learning_rate": 2.7885917526571434e-08, "loss": 0.011, "step": 21184 }, { "epoch": 8.35, "learning_rate": 2.7872914117249702e-08, "loss": 0.0094, "step": 21185 }, { "epoch": 8.35, "learning_rate": 2.785991351333785e-08, "loss": 0.0096, "step": 21186 }, { "epoch": 8.35, "learning_rate": 2.7846915715047847e-08, "loss": 0.0104, "step": 21187 }, { "epoch": 8.35, "learning_rate": 2.783392072259141e-08, "loss": 0.011, "step": 21188 }, { "epoch": 8.35, "learning_rate": 2.7820928536180453e-08, "loss": 0.0082, "step": 21189 }, { "epoch": 8.35, "learning_rate": 2.780793915602664e-08, "loss": 0.0104, "step": 21190 }, { "epoch": 8.35, "learning_rate": 2.7794952582341725e-08, "loss": 0.0102, "step": 21191 }, { "epoch": 8.35, "learning_rate": 2.7781968815337388e-08, "loss": 0.012, "step": 21192 }, { "epoch": 8.35, "learning_rate": 2.776898785522519e-08, "loss": 0.0134, "step": 21193 }, { "epoch": 8.35, "learning_rate": 2.775600970221672e-08, "loss": 0.0104, "step": 21194 }, { "epoch": 8.35, "learning_rate": 2.774303435652352e-08, "loss": 0.011, "step": 21195 }, { "epoch": 8.35, "learning_rate": 2.7730061818357087e-08, "loss": 0.0091, "step": 21196 }, { "epoch": 8.35, "learning_rate": 2.7717092087928785e-08, "loss": 0.0084, "step": 21197 }, { "epoch": 8.35, "learning_rate": 2.770412516545011e-08, "loss": 0.0091, "step": 21198 }, { "epoch": 8.35, "learning_rate": 2.7691161051132227e-08, "loss": 0.0104, "step": 21199 }, { "epoch": 8.36, "learning_rate": 2.7678199745186637e-08, "loss": 0.0094, "step": 21200 }, { "epoch": 8.36, "learning_rate": 2.7665241247824434e-08, "loss": 0.0096, "step": 21201 }, { "epoch": 8.36, "learning_rate": 2.7652285559256938e-08, "loss": 0.0111, "step": 21202 }, { "epoch": 8.36, "learning_rate": 2.7639332679695137e-08, "loss": 0.0108, "step": 21203 }, { "epoch": 8.36, "learning_rate": 2.7626382609350375e-08, "loss": 0.0089, "step": 21204 }, { "epoch": 8.36, "learning_rate": 2.7613435348433544e-08, "loss": 0.011, "step": 21205 }, { "epoch": 8.36, "learning_rate": 2.7600490897155748e-08, "loss": 0.0105, "step": 21206 }, { "epoch": 8.36, "learning_rate": 2.7587549255727903e-08, "loss": 0.0091, "step": 21207 }, { "epoch": 8.36, "learning_rate": 2.7574610424360977e-08, "loss": 0.0104, "step": 21208 }, { "epoch": 8.36, "learning_rate": 2.7561674403265823e-08, "loss": 0.013, "step": 21209 }, { "epoch": 8.36, "learning_rate": 2.7548741192653357e-08, "loss": 0.0125, "step": 21210 }, { "epoch": 8.36, "learning_rate": 2.7535810792734282e-08, "loss": 0.0078, "step": 21211 }, { "epoch": 8.36, "learning_rate": 2.752288320371936e-08, "loss": 0.0102, "step": 21212 }, { "epoch": 8.36, "learning_rate": 2.7509958425819357e-08, "loss": 0.0099, "step": 21213 }, { "epoch": 8.36, "learning_rate": 2.7497036459244816e-08, "loss": 0.0104, "step": 21214 }, { "epoch": 8.36, "learning_rate": 2.7484117304206477e-08, "loss": 0.0108, "step": 21215 }, { "epoch": 8.36, "learning_rate": 2.7471200960914776e-08, "loss": 0.0112, "step": 21216 }, { "epoch": 8.36, "learning_rate": 2.7458287429580272e-08, "loss": 0.0095, "step": 21217 }, { "epoch": 8.36, "learning_rate": 2.744537671041347e-08, "loss": 0.0099, "step": 21218 }, { "epoch": 8.36, "learning_rate": 2.74324688036248e-08, "loss": 0.0124, "step": 21219 }, { "epoch": 8.36, "learning_rate": 2.7419563709424596e-08, "loss": 0.0106, "step": 21220 }, { "epoch": 8.36, "learning_rate": 2.740666142802319e-08, "loss": 0.0117, "step": 21221 }, { "epoch": 8.36, "learning_rate": 2.73937619596309e-08, "loss": 0.0085, "step": 21222 }, { "epoch": 8.36, "learning_rate": 2.7380865304458e-08, "loss": 0.0094, "step": 21223 }, { "epoch": 8.36, "learning_rate": 2.7367971462714612e-08, "loss": 0.0115, "step": 21224 }, { "epoch": 8.36, "learning_rate": 2.73550804346109e-08, "loss": 0.0108, "step": 21225 }, { "epoch": 8.37, "learning_rate": 2.7342192220357007e-08, "loss": 0.0112, "step": 21226 }, { "epoch": 8.37, "learning_rate": 2.732930682016299e-08, "loss": 0.0108, "step": 21227 }, { "epoch": 8.37, "learning_rate": 2.7316424234238833e-08, "loss": 0.0091, "step": 21228 }, { "epoch": 8.37, "learning_rate": 2.7303544462794524e-08, "loss": 0.0124, "step": 21229 }, { "epoch": 8.37, "learning_rate": 2.7290667506039945e-08, "loss": 0.0109, "step": 21230 }, { "epoch": 8.37, "learning_rate": 2.7277793364184987e-08, "loss": 0.0114, "step": 21231 }, { "epoch": 8.37, "learning_rate": 2.726492203743951e-08, "loss": 0.0115, "step": 21232 }, { "epoch": 8.37, "learning_rate": 2.7252053526013295e-08, "loss": 0.0092, "step": 21233 }, { "epoch": 8.37, "learning_rate": 2.723918783011603e-08, "loss": 0.0092, "step": 21234 }, { "epoch": 8.37, "learning_rate": 2.7226324949957445e-08, "loss": 0.0092, "step": 21235 }, { "epoch": 8.37, "learning_rate": 2.7213464885747206e-08, "loss": 0.0079, "step": 21236 }, { "epoch": 8.37, "learning_rate": 2.720060763769485e-08, "loss": 0.0094, "step": 21237 }, { "epoch": 8.37, "learning_rate": 2.7187753206010012e-08, "loss": 0.0122, "step": 21238 }, { "epoch": 8.37, "learning_rate": 2.7174901590902054e-08, "loss": 0.0098, "step": 21239 }, { "epoch": 8.37, "learning_rate": 2.7162052792580616e-08, "loss": 0.0107, "step": 21240 }, { "epoch": 8.37, "learning_rate": 2.714920681125501e-08, "loss": 0.0111, "step": 21241 }, { "epoch": 8.37, "learning_rate": 2.7136363647134653e-08, "loss": 0.0136, "step": 21242 }, { "epoch": 8.37, "learning_rate": 2.7123523300428774e-08, "loss": 0.0097, "step": 21243 }, { "epoch": 8.37, "learning_rate": 2.7110685771346807e-08, "loss": 0.0071, "step": 21244 }, { "epoch": 8.37, "learning_rate": 2.7097851060097852e-08, "loss": 0.0084, "step": 21245 }, { "epoch": 8.37, "learning_rate": 2.708501916689121e-08, "loss": 0.0105, "step": 21246 }, { "epoch": 8.37, "learning_rate": 2.7072190091935888e-08, "loss": 0.0112, "step": 21247 }, { "epoch": 8.37, "learning_rate": 2.705936383544105e-08, "loss": 0.0107, "step": 21248 }, { "epoch": 8.37, "learning_rate": 2.7046540397615757e-08, "loss": 0.0088, "step": 21249 }, { "epoch": 8.37, "learning_rate": 2.7033719778669016e-08, "loss": 0.0102, "step": 21250 }, { "epoch": 8.38, "learning_rate": 2.7020901978809753e-08, "loss": 0.0118, "step": 21251 }, { "epoch": 8.38, "learning_rate": 2.700808699824686e-08, "loss": 0.0104, "step": 21252 }, { "epoch": 8.38, "learning_rate": 2.6995274837189285e-08, "loss": 0.011, "step": 21253 }, { "epoch": 8.38, "learning_rate": 2.698246549584575e-08, "loss": 0.0102, "step": 21254 }, { "epoch": 8.38, "learning_rate": 2.6969658974425067e-08, "loss": 0.0098, "step": 21255 }, { "epoch": 8.38, "learning_rate": 2.6956855273136025e-08, "loss": 0.0091, "step": 21256 }, { "epoch": 8.38, "learning_rate": 2.6944054392187187e-08, "loss": 0.0147, "step": 21257 }, { "epoch": 8.38, "learning_rate": 2.6931256331787256e-08, "loss": 0.0074, "step": 21258 }, { "epoch": 8.38, "learning_rate": 2.691846109214486e-08, "loss": 0.0149, "step": 21259 }, { "epoch": 8.38, "learning_rate": 2.690566867346844e-08, "loss": 0.0091, "step": 21260 }, { "epoch": 8.38, "learning_rate": 2.6892879075966556e-08, "loss": 0.0113, "step": 21261 }, { "epoch": 8.38, "learning_rate": 2.6880092299847644e-08, "loss": 0.0112, "step": 21262 }, { "epoch": 8.38, "learning_rate": 2.6867308345320163e-08, "loss": 0.0107, "step": 21263 }, { "epoch": 8.38, "learning_rate": 2.685452721259238e-08, "loss": 0.0084, "step": 21264 }, { "epoch": 8.38, "learning_rate": 2.684174890187265e-08, "loss": 0.0108, "step": 21265 }, { "epoch": 8.38, "learning_rate": 2.682897341336925e-08, "loss": 0.0105, "step": 21266 }, { "epoch": 8.38, "learning_rate": 2.6816200747290428e-08, "loss": 0.0095, "step": 21267 }, { "epoch": 8.38, "learning_rate": 2.6803430903844294e-08, "loss": 0.0103, "step": 21268 }, { "epoch": 8.38, "learning_rate": 2.679066388323905e-08, "loss": 0.0135, "step": 21269 }, { "epoch": 8.38, "learning_rate": 2.6777899685682625e-08, "loss": 0.0096, "step": 21270 }, { "epoch": 8.38, "learning_rate": 2.6765138311383274e-08, "loss": 0.01, "step": 21271 }, { "epoch": 8.38, "learning_rate": 2.675237976054885e-08, "loss": 0.0134, "step": 21272 }, { "epoch": 8.38, "learning_rate": 2.6739624033387364e-08, "loss": 0.0118, "step": 21273 }, { "epoch": 8.38, "learning_rate": 2.6726871130106633e-08, "loss": 0.0091, "step": 21274 }, { "epoch": 8.38, "learning_rate": 2.671412105091455e-08, "loss": 0.0095, "step": 21275 }, { "epoch": 8.39, "learning_rate": 2.6701373796018977e-08, "loss": 0.0105, "step": 21276 }, { "epoch": 8.39, "learning_rate": 2.668862936562759e-08, "loss": 0.0115, "step": 21277 }, { "epoch": 8.39, "learning_rate": 2.6675887759948157e-08, "loss": 0.0095, "step": 21278 }, { "epoch": 8.39, "learning_rate": 2.666314897918831e-08, "loss": 0.0082, "step": 21279 }, { "epoch": 8.39, "learning_rate": 2.6650413023555728e-08, "loss": 0.0101, "step": 21280 }, { "epoch": 8.39, "learning_rate": 2.6637679893257938e-08, "loss": 0.0113, "step": 21281 }, { "epoch": 8.39, "learning_rate": 2.66249495885025e-08, "loss": 0.0117, "step": 21282 }, { "epoch": 8.39, "learning_rate": 2.6612222109496808e-08, "loss": 0.0078, "step": 21283 }, { "epoch": 8.39, "learning_rate": 2.659949745644847e-08, "loss": 0.0107, "step": 21284 }, { "epoch": 8.39, "learning_rate": 2.6586775629564752e-08, "loss": 0.0115, "step": 21285 }, { "epoch": 8.39, "learning_rate": 2.6574056629053053e-08, "loss": 0.0089, "step": 21286 }, { "epoch": 8.39, "learning_rate": 2.6561340455120617e-08, "loss": 0.0088, "step": 21287 }, { "epoch": 8.39, "learning_rate": 2.6548627107974743e-08, "loss": 0.0099, "step": 21288 }, { "epoch": 8.39, "learning_rate": 2.6535916587822614e-08, "loss": 0.009, "step": 21289 }, { "epoch": 8.39, "learning_rate": 2.6523208894871474e-08, "loss": 0.0091, "step": 21290 }, { "epoch": 8.39, "learning_rate": 2.651050402932833e-08, "loss": 0.0116, "step": 21291 }, { "epoch": 8.39, "learning_rate": 2.6497801991400283e-08, "loss": 0.0093, "step": 21292 }, { "epoch": 8.39, "learning_rate": 2.6485102781294387e-08, "loss": 0.0107, "step": 21293 }, { "epoch": 8.39, "learning_rate": 2.647240639921764e-08, "loss": 0.009, "step": 21294 }, { "epoch": 8.39, "learning_rate": 2.6459712845376893e-08, "loss": 0.0087, "step": 21295 }, { "epoch": 8.39, "learning_rate": 2.6447022119979112e-08, "loss": 0.0094, "step": 21296 }, { "epoch": 8.39, "learning_rate": 2.6434334223231048e-08, "loss": 0.0102, "step": 21297 }, { "epoch": 8.39, "learning_rate": 2.6421649155339576e-08, "loss": 0.0121, "step": 21298 }, { "epoch": 8.39, "learning_rate": 2.6408966916511398e-08, "loss": 0.0116, "step": 21299 }, { "epoch": 8.39, "learning_rate": 2.6396287506953265e-08, "loss": 0.0107, "step": 21300 }, { "epoch": 8.39, "learning_rate": 2.6383610926871756e-08, "loss": 0.0092, "step": 21301 }, { "epoch": 8.4, "learning_rate": 2.637093717647354e-08, "loss": 0.0106, "step": 21302 }, { "epoch": 8.4, "learning_rate": 2.63582662559652e-08, "loss": 0.0092, "step": 21303 }, { "epoch": 8.4, "learning_rate": 2.6345598165553173e-08, "loss": 0.0086, "step": 21304 }, { "epoch": 8.4, "learning_rate": 2.6332932905443962e-08, "loss": 0.0099, "step": 21305 }, { "epoch": 8.4, "learning_rate": 2.632027047584402e-08, "loss": 0.0116, "step": 21306 }, { "epoch": 8.4, "learning_rate": 2.6307610876959718e-08, "loss": 0.0103, "step": 21307 }, { "epoch": 8.4, "learning_rate": 2.6294954108997358e-08, "loss": 0.0112, "step": 21308 }, { "epoch": 8.4, "learning_rate": 2.6282300172163284e-08, "loss": 0.0107, "step": 21309 }, { "epoch": 8.4, "learning_rate": 2.62696490666636e-08, "loss": 0.0084, "step": 21310 }, { "epoch": 8.4, "learning_rate": 2.62570007927047e-08, "loss": 0.0109, "step": 21311 }, { "epoch": 8.4, "learning_rate": 2.6244355350492586e-08, "loss": 0.0104, "step": 21312 }, { "epoch": 8.4, "learning_rate": 2.6231712740233435e-08, "loss": 0.0103, "step": 21313 }, { "epoch": 8.4, "learning_rate": 2.6219072962133172e-08, "loss": 0.0123, "step": 21314 }, { "epoch": 8.4, "learning_rate": 2.6206436016398005e-08, "loss": 0.0094, "step": 21315 }, { "epoch": 8.4, "learning_rate": 2.6193801903233748e-08, "loss": 0.0086, "step": 21316 }, { "epoch": 8.4, "learning_rate": 2.6181170622846394e-08, "loss": 0.01, "step": 21317 }, { "epoch": 8.4, "learning_rate": 2.6168542175441756e-08, "loss": 0.009, "step": 21318 }, { "epoch": 8.4, "learning_rate": 2.615591656122569e-08, "loss": 0.0099, "step": 21319 }, { "epoch": 8.4, "learning_rate": 2.6143293780404008e-08, "loss": 0.0133, "step": 21320 }, { "epoch": 8.4, "learning_rate": 2.613067383318237e-08, "loss": 0.0097, "step": 21321 }, { "epoch": 8.4, "learning_rate": 2.6118056719766502e-08, "loss": 0.0086, "step": 21322 }, { "epoch": 8.4, "learning_rate": 2.6105442440362013e-08, "loss": 0.0087, "step": 21323 }, { "epoch": 8.4, "learning_rate": 2.6092830995174585e-08, "loss": 0.0124, "step": 21324 }, { "epoch": 8.4, "learning_rate": 2.6080222384409655e-08, "loss": 0.0096, "step": 21325 }, { "epoch": 8.4, "learning_rate": 2.6067616608272834e-08, "loss": 0.0099, "step": 21326 }, { "epoch": 8.41, "learning_rate": 2.605501366696945e-08, "loss": 0.0106, "step": 21327 }, { "epoch": 8.41, "learning_rate": 2.604241356070498e-08, "loss": 0.0106, "step": 21328 }, { "epoch": 8.41, "learning_rate": 2.602981628968479e-08, "loss": 0.0092, "step": 21329 }, { "epoch": 8.41, "learning_rate": 2.6017221854114234e-08, "loss": 0.0117, "step": 21330 }, { "epoch": 8.41, "learning_rate": 2.60046302541985e-08, "loss": 0.0082, "step": 21331 }, { "epoch": 8.41, "learning_rate": 2.5992041490142846e-08, "loss": 0.0103, "step": 21332 }, { "epoch": 8.41, "learning_rate": 2.597945556215244e-08, "loss": 0.0104, "step": 21333 }, { "epoch": 8.41, "learning_rate": 2.5966872470432476e-08, "loss": 0.0121, "step": 21334 }, { "epoch": 8.41, "learning_rate": 2.5954292215187968e-08, "loss": 0.0113, "step": 21335 }, { "epoch": 8.41, "learning_rate": 2.5941714796624015e-08, "loss": 0.0102, "step": 21336 }, { "epoch": 8.41, "learning_rate": 2.5929140214945478e-08, "loss": 0.0083, "step": 21337 }, { "epoch": 8.41, "learning_rate": 2.591656847035748e-08, "loss": 0.0117, "step": 21338 }, { "epoch": 8.41, "learning_rate": 2.5903999563064792e-08, "loss": 0.0099, "step": 21339 }, { "epoch": 8.41, "learning_rate": 2.5891433493272362e-08, "loss": 0.0102, "step": 21340 }, { "epoch": 8.41, "learning_rate": 2.5878870261184917e-08, "loss": 0.0097, "step": 21341 }, { "epoch": 8.41, "learning_rate": 2.5866309867007262e-08, "loss": 0.009, "step": 21342 }, { "epoch": 8.41, "learning_rate": 2.5853752310944132e-08, "loss": 0.0106, "step": 21343 }, { "epoch": 8.41, "learning_rate": 2.5841197593200138e-08, "loss": 0.0094, "step": 21344 }, { "epoch": 8.41, "learning_rate": 2.5828645713979912e-08, "loss": 0.0069, "step": 21345 }, { "epoch": 8.41, "learning_rate": 2.581609667348805e-08, "loss": 0.0116, "step": 21346 }, { "epoch": 8.41, "learning_rate": 2.5803550471929147e-08, "loss": 0.0113, "step": 21347 }, { "epoch": 8.41, "learning_rate": 2.5791007109507567e-08, "loss": 0.0083, "step": 21348 }, { "epoch": 8.41, "learning_rate": 2.577846658642786e-08, "loss": 0.01, "step": 21349 }, { "epoch": 8.41, "learning_rate": 2.576592890289426e-08, "loss": 0.0099, "step": 21350 }, { "epoch": 8.41, "learning_rate": 2.5753394059111298e-08, "loss": 0.0105, "step": 21351 }, { "epoch": 8.41, "learning_rate": 2.5740862055283165e-08, "loss": 0.0122, "step": 21352 }, { "epoch": 8.42, "learning_rate": 2.5728332891614158e-08, "loss": 0.0106, "step": 21353 }, { "epoch": 8.42, "learning_rate": 2.5715806568308406e-08, "loss": 0.0105, "step": 21354 }, { "epoch": 8.42, "learning_rate": 2.5703283085570193e-08, "loss": 0.0077, "step": 21355 }, { "epoch": 8.42, "learning_rate": 2.5690762443603508e-08, "loss": 0.0105, "step": 21356 }, { "epoch": 8.42, "learning_rate": 2.5678244642612545e-08, "loss": 0.0109, "step": 21357 }, { "epoch": 8.42, "learning_rate": 2.5665729682801186e-08, "loss": 0.0115, "step": 21358 }, { "epoch": 8.42, "learning_rate": 2.5653217564373485e-08, "loss": 0.0089, "step": 21359 }, { "epoch": 8.42, "learning_rate": 2.5640708287533375e-08, "loss": 0.0106, "step": 21360 }, { "epoch": 8.42, "learning_rate": 2.5628201852484753e-08, "loss": 0.0127, "step": 21361 }, { "epoch": 8.42, "learning_rate": 2.5615698259431394e-08, "loss": 0.0128, "step": 21362 }, { "epoch": 8.42, "learning_rate": 2.560319750857711e-08, "loss": 0.011, "step": 21363 }, { "epoch": 8.42, "learning_rate": 2.55906996001257e-08, "loss": 0.0129, "step": 21364 }, { "epoch": 8.42, "learning_rate": 2.5578204534280767e-08, "loss": 0.0109, "step": 21365 }, { "epoch": 8.42, "learning_rate": 2.5565712311246068e-08, "loss": 0.0094, "step": 21366 }, { "epoch": 8.42, "learning_rate": 2.5553222931225106e-08, "loss": 0.0128, "step": 21367 }, { "epoch": 8.42, "learning_rate": 2.554073639442147e-08, "loss": 0.0105, "step": 21368 }, { "epoch": 8.42, "learning_rate": 2.5528252701038666e-08, "loss": 0.0081, "step": 21369 }, { "epoch": 8.42, "learning_rate": 2.5515771851280243e-08, "loss": 0.0091, "step": 21370 }, { "epoch": 8.42, "learning_rate": 2.5503293845349507e-08, "loss": 0.0109, "step": 21371 }, { "epoch": 8.42, "learning_rate": 2.549081868344989e-08, "loss": 0.0099, "step": 21372 }, { "epoch": 8.42, "learning_rate": 2.5478346365784676e-08, "loss": 0.0112, "step": 21373 }, { "epoch": 8.42, "learning_rate": 2.5465876892557237e-08, "loss": 0.0077, "step": 21374 }, { "epoch": 8.42, "learning_rate": 2.54534102639707e-08, "loss": 0.0109, "step": 21375 }, { "epoch": 8.42, "learning_rate": 2.544094648022832e-08, "loss": 0.0112, "step": 21376 }, { "epoch": 8.42, "learning_rate": 2.542848554153314e-08, "loss": 0.0086, "step": 21377 }, { "epoch": 8.43, "learning_rate": 2.5416027448088416e-08, "loss": 0.0129, "step": 21378 }, { "epoch": 8.43, "learning_rate": 2.5403572200097078e-08, "loss": 0.0089, "step": 21379 }, { "epoch": 8.43, "learning_rate": 2.5391119797762183e-08, "loss": 0.0102, "step": 21380 }, { "epoch": 8.43, "learning_rate": 2.5378670241286615e-08, "loss": 0.0101, "step": 21381 }, { "epoch": 8.43, "learning_rate": 2.5366223530873322e-08, "loss": 0.011, "step": 21382 }, { "epoch": 8.43, "learning_rate": 2.5353779666725185e-08, "loss": 0.0097, "step": 21383 }, { "epoch": 8.43, "learning_rate": 2.5341338649045064e-08, "loss": 0.0111, "step": 21384 }, { "epoch": 8.43, "learning_rate": 2.5328900478035596e-08, "loss": 0.0101, "step": 21385 }, { "epoch": 8.43, "learning_rate": 2.5316465153899602e-08, "loss": 0.009, "step": 21386 }, { "epoch": 8.43, "learning_rate": 2.530403267683978e-08, "loss": 0.011, "step": 21387 }, { "epoch": 8.43, "learning_rate": 2.5291603047058684e-08, "loss": 0.0112, "step": 21388 }, { "epoch": 8.43, "learning_rate": 2.527917626475895e-08, "loss": 0.0118, "step": 21389 }, { "epoch": 8.43, "learning_rate": 2.5266752330143037e-08, "loss": 0.0099, "step": 21390 }, { "epoch": 8.43, "learning_rate": 2.5254331243413564e-08, "loss": 0.0087, "step": 21391 }, { "epoch": 8.43, "learning_rate": 2.5241913004772876e-08, "loss": 0.01, "step": 21392 }, { "epoch": 8.43, "learning_rate": 2.5229497614423456e-08, "loss": 0.0098, "step": 21393 }, { "epoch": 8.43, "learning_rate": 2.52170850725675e-08, "loss": 0.0092, "step": 21394 }, { "epoch": 8.43, "learning_rate": 2.5204675379407513e-08, "loss": 0.0078, "step": 21395 }, { "epoch": 8.43, "learning_rate": 2.5192268535145623e-08, "loss": 0.0103, "step": 21396 }, { "epoch": 8.43, "learning_rate": 2.5179864539984107e-08, "loss": 0.0112, "step": 21397 }, { "epoch": 8.43, "learning_rate": 2.5167463394125054e-08, "loss": 0.0101, "step": 21398 }, { "epoch": 8.43, "learning_rate": 2.5155065097770657e-08, "loss": 0.0093, "step": 21399 }, { "epoch": 8.43, "learning_rate": 2.514266965112295e-08, "loss": 0.0079, "step": 21400 }, { "epoch": 8.43, "learning_rate": 2.5130277054384018e-08, "loss": 0.0085, "step": 21401 }, { "epoch": 8.43, "learning_rate": 2.511788730775577e-08, "loss": 0.0098, "step": 21402 }, { "epoch": 8.44, "learning_rate": 2.510550041144015e-08, "loss": 0.0111, "step": 21403 }, { "epoch": 8.44, "learning_rate": 2.5093116365639112e-08, "loss": 0.0094, "step": 21404 }, { "epoch": 8.44, "learning_rate": 2.5080735170554425e-08, "loss": 0.0103, "step": 21405 }, { "epoch": 8.44, "learning_rate": 2.5068356826387883e-08, "loss": 0.0094, "step": 21406 }, { "epoch": 8.44, "learning_rate": 2.5055981333341303e-08, "loss": 0.0106, "step": 21407 }, { "epoch": 8.44, "learning_rate": 2.5043608691616324e-08, "loss": 0.0116, "step": 21408 }, { "epoch": 8.44, "learning_rate": 2.5031238901414587e-08, "loss": 0.0128, "step": 21409 }, { "epoch": 8.44, "learning_rate": 2.5018871962937815e-08, "loss": 0.0095, "step": 21410 }, { "epoch": 8.44, "learning_rate": 2.5006507876387405e-08, "loss": 0.0096, "step": 21411 }, { "epoch": 8.44, "learning_rate": 2.4994146641964996e-08, "loss": 0.01, "step": 21412 }, { "epoch": 8.44, "learning_rate": 2.4981788259872005e-08, "loss": 0.0097, "step": 21413 }, { "epoch": 8.44, "learning_rate": 2.496943273030989e-08, "loss": 0.0123, "step": 21414 }, { "epoch": 8.44, "learning_rate": 2.4957080053479985e-08, "loss": 0.0092, "step": 21415 }, { "epoch": 8.44, "learning_rate": 2.4944730229583677e-08, "loss": 0.0108, "step": 21416 }, { "epoch": 8.44, "learning_rate": 2.4932383258822144e-08, "loss": 0.0104, "step": 21417 }, { "epoch": 8.44, "learning_rate": 2.4920039141396755e-08, "loss": 0.0109, "step": 21418 }, { "epoch": 8.44, "learning_rate": 2.4907697877508594e-08, "loss": 0.009, "step": 21419 }, { "epoch": 8.44, "learning_rate": 2.48953594673589e-08, "loss": 0.0104, "step": 21420 }, { "epoch": 8.44, "learning_rate": 2.4883023911148626e-08, "loss": 0.0121, "step": 21421 }, { "epoch": 8.44, "learning_rate": 2.4870691209079008e-08, "loss": 0.009, "step": 21422 }, { "epoch": 8.44, "learning_rate": 2.4858361361350912e-08, "loss": 0.0107, "step": 21423 }, { "epoch": 8.44, "learning_rate": 2.4846034368165392e-08, "loss": 0.0118, "step": 21424 }, { "epoch": 8.44, "learning_rate": 2.483371022972325e-08, "loss": 0.0089, "step": 21425 }, { "epoch": 8.44, "learning_rate": 2.482138894622543e-08, "loss": 0.0096, "step": 21426 }, { "epoch": 8.44, "learning_rate": 2.480907051787271e-08, "loss": 0.0094, "step": 21427 }, { "epoch": 8.44, "learning_rate": 2.479675494486593e-08, "loss": 0.0113, "step": 21428 }, { "epoch": 8.45, "learning_rate": 2.4784442227405722e-08, "loss": 0.0115, "step": 21429 }, { "epoch": 8.45, "learning_rate": 2.477213236569282e-08, "loss": 0.0088, "step": 21430 }, { "epoch": 8.45, "learning_rate": 2.4759825359927866e-08, "loss": 0.0102, "step": 21431 }, { "epoch": 8.45, "learning_rate": 2.4747521210311406e-08, "loss": 0.0103, "step": 21432 }, { "epoch": 8.45, "learning_rate": 2.4735219917044015e-08, "loss": 0.0109, "step": 21433 }, { "epoch": 8.45, "learning_rate": 2.4722921480326087e-08, "loss": 0.0104, "step": 21434 }, { "epoch": 8.45, "learning_rate": 2.4710625900358262e-08, "loss": 0.0129, "step": 21435 }, { "epoch": 8.45, "learning_rate": 2.4698333177340758e-08, "loss": 0.0097, "step": 21436 }, { "epoch": 8.45, "learning_rate": 2.4686043311474036e-08, "loss": 0.0091, "step": 21437 }, { "epoch": 8.45, "learning_rate": 2.4673756302958315e-08, "loss": 0.013, "step": 21438 }, { "epoch": 8.45, "learning_rate": 2.466147215199392e-08, "loss": 0.0119, "step": 21439 }, { "epoch": 8.45, "learning_rate": 2.4649190858781032e-08, "loss": 0.0106, "step": 21440 }, { "epoch": 8.45, "learning_rate": 2.4636912423519884e-08, "loss": 0.0105, "step": 21441 }, { "epoch": 8.45, "learning_rate": 2.4624636846410497e-08, "loss": 0.0076, "step": 21442 }, { "epoch": 8.45, "learning_rate": 2.4612364127652973e-08, "loss": 0.0084, "step": 21443 }, { "epoch": 8.45, "learning_rate": 2.460009426744738e-08, "loss": 0.0091, "step": 21444 }, { "epoch": 8.45, "learning_rate": 2.458782726599371e-08, "loss": 0.0095, "step": 21445 }, { "epoch": 8.45, "learning_rate": 2.4575563123491806e-08, "loss": 0.0072, "step": 21446 }, { "epoch": 8.45, "learning_rate": 2.456330184014166e-08, "loss": 0.009, "step": 21447 }, { "epoch": 8.45, "learning_rate": 2.4551043416143002e-08, "loss": 0.0109, "step": 21448 }, { "epoch": 8.45, "learning_rate": 2.4538787851695695e-08, "loss": 0.0133, "step": 21449 }, { "epoch": 8.45, "learning_rate": 2.4526535146999493e-08, "loss": 0.0099, "step": 21450 }, { "epoch": 8.45, "learning_rate": 2.45142853022541e-08, "loss": 0.0126, "step": 21451 }, { "epoch": 8.45, "learning_rate": 2.4502038317659112e-08, "loss": 0.011, "step": 21452 }, { "epoch": 8.45, "learning_rate": 2.448979419341417e-08, "loss": 0.0097, "step": 21453 }, { "epoch": 8.46, "learning_rate": 2.4477552929718893e-08, "loss": 0.0103, "step": 21454 }, { "epoch": 8.46, "learning_rate": 2.4465314526772697e-08, "loss": 0.0111, "step": 21455 }, { "epoch": 8.46, "learning_rate": 2.445307898477511e-08, "loss": 0.0127, "step": 21456 }, { "epoch": 8.46, "learning_rate": 2.4440846303925468e-08, "loss": 0.0099, "step": 21457 }, { "epoch": 8.46, "learning_rate": 2.442861648442329e-08, "loss": 0.0084, "step": 21458 }, { "epoch": 8.46, "learning_rate": 2.4416389526467806e-08, "loss": 0.0118, "step": 21459 }, { "epoch": 8.46, "learning_rate": 2.4404165430258337e-08, "loss": 0.0093, "step": 21460 }, { "epoch": 8.46, "learning_rate": 2.439194419599402e-08, "loss": 0.0096, "step": 21461 }, { "epoch": 8.46, "learning_rate": 2.43797258238742e-08, "loss": 0.0103, "step": 21462 }, { "epoch": 8.46, "learning_rate": 2.4367510314097895e-08, "loss": 0.0107, "step": 21463 }, { "epoch": 8.46, "learning_rate": 2.4355297666864282e-08, "loss": 0.0092, "step": 21464 }, { "epoch": 8.46, "learning_rate": 2.4343087882372293e-08, "loss": 0.0121, "step": 21465 }, { "epoch": 8.46, "learning_rate": 2.4330880960821076e-08, "loss": 0.008, "step": 21466 }, { "epoch": 8.46, "learning_rate": 2.4318676902409473e-08, "loss": 0.0094, "step": 21467 }, { "epoch": 8.46, "learning_rate": 2.4306475707336484e-08, "loss": 0.0125, "step": 21468 }, { "epoch": 8.46, "learning_rate": 2.4294277375800852e-08, "loss": 0.011, "step": 21469 }, { "epoch": 8.46, "learning_rate": 2.4282081908001474e-08, "loss": 0.0074, "step": 21470 }, { "epoch": 8.46, "learning_rate": 2.4269889304137138e-08, "loss": 0.0132, "step": 21471 }, { "epoch": 8.46, "learning_rate": 2.4257699564406486e-08, "loss": 0.0095, "step": 21472 }, { "epoch": 8.46, "learning_rate": 2.424551268900825e-08, "loss": 0.0118, "step": 21473 }, { "epoch": 8.46, "learning_rate": 2.423332867814103e-08, "loss": 0.012, "step": 21474 }, { "epoch": 8.46, "learning_rate": 2.4221147532003437e-08, "loss": 0.0097, "step": 21475 }, { "epoch": 8.46, "learning_rate": 2.4208969250793964e-08, "loss": 0.0102, "step": 21476 }, { "epoch": 8.46, "learning_rate": 2.419679383471118e-08, "loss": 0.0122, "step": 21477 }, { "epoch": 8.46, "learning_rate": 2.41846212839534e-08, "loss": 0.0102, "step": 21478 }, { "epoch": 8.47, "learning_rate": 2.417245159871908e-08, "loss": 0.0082, "step": 21479 }, { "epoch": 8.47, "learning_rate": 2.4160284779206596e-08, "loss": 0.0091, "step": 21480 }, { "epoch": 8.47, "learning_rate": 2.414812082561426e-08, "loss": 0.0121, "step": 21481 }, { "epoch": 8.47, "learning_rate": 2.4135959738140243e-08, "loss": 0.0097, "step": 21482 }, { "epoch": 8.47, "learning_rate": 2.4123801516982788e-08, "loss": 0.0104, "step": 21483 }, { "epoch": 8.47, "learning_rate": 2.4111646162340072e-08, "loss": 0.0129, "step": 21484 }, { "epoch": 8.47, "learning_rate": 2.4099493674410264e-08, "loss": 0.01, "step": 21485 }, { "epoch": 8.47, "learning_rate": 2.4087344053391322e-08, "loss": 0.0102, "step": 21486 }, { "epoch": 8.47, "learning_rate": 2.4075197299481355e-08, "loss": 0.0136, "step": 21487 }, { "epoch": 8.47, "learning_rate": 2.40630534128782e-08, "loss": 0.0091, "step": 21488 }, { "epoch": 8.47, "learning_rate": 2.405091239377999e-08, "loss": 0.0099, "step": 21489 }, { "epoch": 8.47, "learning_rate": 2.403877424238443e-08, "loss": 0.01, "step": 21490 }, { "epoch": 8.47, "learning_rate": 2.402663895888948e-08, "loss": 0.0103, "step": 21491 }, { "epoch": 8.47, "learning_rate": 2.4014506543492796e-08, "loss": 0.0106, "step": 21492 }, { "epoch": 8.47, "learning_rate": 2.4002376996392225e-08, "loss": 0.0132, "step": 21493 }, { "epoch": 8.47, "learning_rate": 2.3990250317785432e-08, "loss": 0.0115, "step": 21494 }, { "epoch": 8.47, "learning_rate": 2.3978126507870035e-08, "loss": 0.0096, "step": 21495 }, { "epoch": 8.47, "learning_rate": 2.396600556684365e-08, "loss": 0.0093, "step": 21496 }, { "epoch": 8.47, "learning_rate": 2.3953887494903836e-08, "loss": 0.0084, "step": 21497 }, { "epoch": 8.47, "learning_rate": 2.3941772292248142e-08, "loss": 0.009, "step": 21498 }, { "epoch": 8.47, "learning_rate": 2.3929659959073923e-08, "loss": 0.011, "step": 21499 }, { "epoch": 8.47, "learning_rate": 2.391755049557871e-08, "loss": 0.0109, "step": 21500 }, { "epoch": 8.47, "learning_rate": 2.39054439019597e-08, "loss": 0.0083, "step": 21501 }, { "epoch": 8.47, "learning_rate": 2.3893340178414424e-08, "loss": 0.0089, "step": 21502 }, { "epoch": 8.47, "learning_rate": 2.3881239325140013e-08, "loss": 0.0125, "step": 21503 }, { "epoch": 8.47, "learning_rate": 2.3869141342333778e-08, "loss": 0.009, "step": 21504 }, { "epoch": 8.48, "learning_rate": 2.3857046230192755e-08, "loss": 0.0114, "step": 21505 }, { "epoch": 8.48, "learning_rate": 2.3844953988914262e-08, "loss": 0.0086, "step": 21506 }, { "epoch": 8.48, "learning_rate": 2.3832864618695247e-08, "loss": 0.013, "step": 21507 }, { "epoch": 8.48, "learning_rate": 2.3820778119732843e-08, "loss": 0.0115, "step": 21508 }, { "epoch": 8.48, "learning_rate": 2.3808694492223935e-08, "loss": 0.0114, "step": 21509 }, { "epoch": 8.48, "learning_rate": 2.3796613736365545e-08, "loss": 0.0105, "step": 21510 }, { "epoch": 8.48, "learning_rate": 2.378453585235456e-08, "loss": 0.0097, "step": 21511 }, { "epoch": 8.48, "learning_rate": 2.3772460840387842e-08, "loss": 0.011, "step": 21512 }, { "epoch": 8.48, "learning_rate": 2.3760388700662147e-08, "loss": 0.0098, "step": 21513 }, { "epoch": 8.48, "learning_rate": 2.3748319433374274e-08, "loss": 0.0092, "step": 21514 }, { "epoch": 8.48, "learning_rate": 2.3736253038720956e-08, "loss": 0.0118, "step": 21515 }, { "epoch": 8.48, "learning_rate": 2.3724189516898764e-08, "loss": 0.0091, "step": 21516 }, { "epoch": 8.48, "learning_rate": 2.3712128868104432e-08, "loss": 0.0085, "step": 21517 }, { "epoch": 8.48, "learning_rate": 2.3700071092534445e-08, "loss": 0.0097, "step": 21518 }, { "epoch": 8.48, "learning_rate": 2.368801619038534e-08, "loss": 0.0114, "step": 21519 }, { "epoch": 8.48, "learning_rate": 2.36759641618536e-08, "loss": 0.0123, "step": 21520 }, { "epoch": 8.48, "learning_rate": 2.3663915007135715e-08, "loss": 0.0082, "step": 21521 }, { "epoch": 8.48, "learning_rate": 2.365186872642797e-08, "loss": 0.0114, "step": 21522 }, { "epoch": 8.48, "learning_rate": 2.3639825319926742e-08, "loss": 0.012, "step": 21523 }, { "epoch": 8.48, "learning_rate": 2.362778478782832e-08, "loss": 0.0112, "step": 21524 }, { "epoch": 8.48, "learning_rate": 2.3615747130329012e-08, "loss": 0.0089, "step": 21525 }, { "epoch": 8.48, "learning_rate": 2.3603712347624904e-08, "loss": 0.0104, "step": 21526 }, { "epoch": 8.48, "learning_rate": 2.359168043991222e-08, "loss": 0.0105, "step": 21527 }, { "epoch": 8.48, "learning_rate": 2.357965140738698e-08, "loss": 0.0121, "step": 21528 }, { "epoch": 8.48, "learning_rate": 2.356762525024536e-08, "loss": 0.0103, "step": 21529 }, { "epoch": 8.49, "learning_rate": 2.355560196868327e-08, "loss": 0.0114, "step": 21530 }, { "epoch": 8.49, "learning_rate": 2.3543581562896732e-08, "loss": 0.0103, "step": 21531 }, { "epoch": 8.49, "learning_rate": 2.3531564033081585e-08, "loss": 0.0105, "step": 21532 }, { "epoch": 8.49, "learning_rate": 2.351954937943377e-08, "loss": 0.0109, "step": 21533 }, { "epoch": 8.49, "learning_rate": 2.350753760214905e-08, "loss": 0.0107, "step": 21534 }, { "epoch": 8.49, "learning_rate": 2.3495528701423284e-08, "loss": 0.013, "step": 21535 }, { "epoch": 8.49, "learning_rate": 2.3483522677452104e-08, "loss": 0.0112, "step": 21536 }, { "epoch": 8.49, "learning_rate": 2.3471519530431228e-08, "loss": 0.0113, "step": 21537 }, { "epoch": 8.49, "learning_rate": 2.3459519260556336e-08, "loss": 0.0122, "step": 21538 }, { "epoch": 8.49, "learning_rate": 2.3447521868022922e-08, "loss": 0.0108, "step": 21539 }, { "epoch": 8.49, "learning_rate": 2.343552735302663e-08, "loss": 0.0071, "step": 21540 }, { "epoch": 8.49, "learning_rate": 2.3423535715762788e-08, "loss": 0.0127, "step": 21541 }, { "epoch": 8.49, "learning_rate": 2.3411546956427043e-08, "loss": 0.0083, "step": 21542 }, { "epoch": 8.49, "learning_rate": 2.339956107521468e-08, "loss": 0.0102, "step": 21543 }, { "epoch": 8.49, "learning_rate": 2.3387578072321078e-08, "loss": 0.0101, "step": 21544 }, { "epoch": 8.49, "learning_rate": 2.3375597947941505e-08, "loss": 0.0118, "step": 21545 }, { "epoch": 8.49, "learning_rate": 2.3363620702271246e-08, "loss": 0.0103, "step": 21546 }, { "epoch": 8.49, "learning_rate": 2.3351646335505505e-08, "loss": 0.0096, "step": 21547 }, { "epoch": 8.49, "learning_rate": 2.333967484783952e-08, "loss": 0.0094, "step": 21548 }, { "epoch": 8.49, "learning_rate": 2.332770623946827e-08, "loss": 0.0092, "step": 21549 }, { "epoch": 8.49, "learning_rate": 2.3315740510586913e-08, "loss": 0.0106, "step": 21550 }, { "epoch": 8.49, "learning_rate": 2.3303777661390466e-08, "loss": 0.0109, "step": 21551 }, { "epoch": 8.49, "learning_rate": 2.3291817692073933e-08, "loss": 0.0111, "step": 21552 }, { "epoch": 8.49, "learning_rate": 2.3279860602832156e-08, "loss": 0.0105, "step": 21553 }, { "epoch": 8.49, "learning_rate": 2.326790639386007e-08, "loss": 0.0112, "step": 21554 }, { "epoch": 8.49, "learning_rate": 2.3255955065352495e-08, "loss": 0.0093, "step": 21555 }, { "epoch": 8.5, "learning_rate": 2.3244006617504298e-08, "loss": 0.0111, "step": 21556 }, { "epoch": 8.5, "learning_rate": 2.32320610505101e-08, "loss": 0.0099, "step": 21557 }, { "epoch": 8.5, "learning_rate": 2.32201183645647e-08, "loss": 0.0122, "step": 21558 }, { "epoch": 8.5, "learning_rate": 2.3208178559862658e-08, "loss": 0.0108, "step": 21559 }, { "epoch": 8.5, "learning_rate": 2.319624163659859e-08, "loss": 0.0125, "step": 21560 }, { "epoch": 8.5, "learning_rate": 2.3184307594967122e-08, "loss": 0.0118, "step": 21561 }, { "epoch": 8.5, "learning_rate": 2.317237643516268e-08, "loss": 0.0099, "step": 21562 }, { "epoch": 8.5, "learning_rate": 2.3160448157379765e-08, "loss": 0.0084, "step": 21563 }, { "epoch": 8.5, "learning_rate": 2.314852276181276e-08, "loss": 0.0092, "step": 21564 }, { "epoch": 8.5, "learning_rate": 2.313660024865609e-08, "loss": 0.0103, "step": 21565 }, { "epoch": 8.5, "learning_rate": 2.3124680618103997e-08, "loss": 0.0116, "step": 21566 }, { "epoch": 8.5, "learning_rate": 2.3112763870350838e-08, "loss": 0.012, "step": 21567 }, { "epoch": 8.5, "learning_rate": 2.3100850005590723e-08, "loss": 0.0109, "step": 21568 }, { "epoch": 8.5, "learning_rate": 2.3088939024017963e-08, "loss": 0.0092, "step": 21569 }, { "epoch": 8.5, "learning_rate": 2.307703092582658e-08, "loss": 0.0101, "step": 21570 }, { "epoch": 8.5, "learning_rate": 2.306512571121073e-08, "loss": 0.012, "step": 21571 }, { "epoch": 8.5, "learning_rate": 2.305322338036435e-08, "loss": 0.0109, "step": 21572 }, { "epoch": 8.5, "learning_rate": 2.3041323933481594e-08, "loss": 0.0099, "step": 21573 }, { "epoch": 8.5, "learning_rate": 2.302942737075626e-08, "loss": 0.0123, "step": 21574 }, { "epoch": 8.5, "learning_rate": 2.301753369238235e-08, "loss": 0.0109, "step": 21575 }, { "epoch": 8.5, "learning_rate": 2.30056428985536e-08, "loss": 0.0091, "step": 21576 }, { "epoch": 8.5, "learning_rate": 2.2993754989463875e-08, "loss": 0.0105, "step": 21577 }, { "epoch": 8.5, "learning_rate": 2.2981869965306932e-08, "loss": 0.011, "step": 21578 }, { "epoch": 8.5, "learning_rate": 2.2969987826276503e-08, "loss": 0.0078, "step": 21579 }, { "epoch": 8.5, "learning_rate": 2.295810857256617e-08, "loss": 0.0132, "step": 21580 }, { "epoch": 8.51, "learning_rate": 2.294623220436962e-08, "loss": 0.011, "step": 21581 }, { "epoch": 8.51, "learning_rate": 2.293435872188041e-08, "loss": 0.0102, "step": 21582 }, { "epoch": 8.51, "learning_rate": 2.2922488125292027e-08, "loss": 0.0091, "step": 21583 }, { "epoch": 8.51, "learning_rate": 2.2910620414797987e-08, "loss": 0.0105, "step": 21584 }, { "epoch": 8.51, "learning_rate": 2.2898755590591645e-08, "loss": 0.012, "step": 21585 }, { "epoch": 8.51, "learning_rate": 2.2886893652866424e-08, "loss": 0.0106, "step": 21586 }, { "epoch": 8.51, "learning_rate": 2.287503460181568e-08, "loss": 0.0107, "step": 21587 }, { "epoch": 8.51, "learning_rate": 2.286317843763268e-08, "loss": 0.0093, "step": 21588 }, { "epoch": 8.51, "learning_rate": 2.2851325160510628e-08, "loss": 0.0089, "step": 21589 }, { "epoch": 8.51, "learning_rate": 2.283947477064274e-08, "loss": 0.0119, "step": 21590 }, { "epoch": 8.51, "learning_rate": 2.282762726822216e-08, "loss": 0.0081, "step": 21591 }, { "epoch": 8.51, "learning_rate": 2.281578265344204e-08, "loss": 0.0095, "step": 21592 }, { "epoch": 8.51, "learning_rate": 2.2803940926495313e-08, "loss": 0.0129, "step": 21593 }, { "epoch": 8.51, "learning_rate": 2.279210208757505e-08, "loss": 0.0086, "step": 21594 }, { "epoch": 8.51, "learning_rate": 2.2780266136874203e-08, "loss": 0.0104, "step": 21595 }, { "epoch": 8.51, "learning_rate": 2.2768433074585713e-08, "loss": 0.0092, "step": 21596 }, { "epoch": 8.51, "learning_rate": 2.2756602900902355e-08, "loss": 0.0144, "step": 21597 }, { "epoch": 8.51, "learning_rate": 2.2744775616017043e-08, "loss": 0.0102, "step": 21598 }, { "epoch": 8.51, "learning_rate": 2.2732951220122442e-08, "loss": 0.0097, "step": 21599 }, { "epoch": 8.51, "learning_rate": 2.2721129713411313e-08, "loss": 0.012, "step": 21600 }, { "epoch": 8.51, "learning_rate": 2.2709311096076345e-08, "loss": 0.0154, "step": 21601 }, { "epoch": 8.51, "learning_rate": 2.2697495368310205e-08, "loss": 0.0097, "step": 21602 }, { "epoch": 8.51, "learning_rate": 2.2685682530305362e-08, "loss": 0.0085, "step": 21603 }, { "epoch": 8.51, "learning_rate": 2.2673872582254416e-08, "loss": 0.0102, "step": 21604 }, { "epoch": 8.51, "learning_rate": 2.2662065524349905e-08, "loss": 0.0116, "step": 21605 }, { "epoch": 8.52, "learning_rate": 2.265026135678414e-08, "loss": 0.0126, "step": 21606 }, { "epoch": 8.52, "learning_rate": 2.263846007974961e-08, "loss": 0.0096, "step": 21607 }, { "epoch": 8.52, "learning_rate": 2.2626661693438543e-08, "loss": 0.0086, "step": 21608 }, { "epoch": 8.52, "learning_rate": 2.26148661980434e-08, "loss": 0.0115, "step": 21609 }, { "epoch": 8.52, "learning_rate": 2.2603073593756306e-08, "loss": 0.011, "step": 21610 }, { "epoch": 8.52, "learning_rate": 2.259128388076954e-08, "loss": 0.0125, "step": 21611 }, { "epoch": 8.52, "learning_rate": 2.2579497059275153e-08, "loss": 0.0127, "step": 21612 }, { "epoch": 8.52, "learning_rate": 2.256771312946537e-08, "loss": 0.01, "step": 21613 }, { "epoch": 8.52, "learning_rate": 2.2555932091532172e-08, "loss": 0.0106, "step": 21614 }, { "epoch": 8.52, "learning_rate": 2.254415394566762e-08, "loss": 0.0119, "step": 21615 }, { "epoch": 8.52, "learning_rate": 2.2532378692063635e-08, "loss": 0.0096, "step": 21616 }, { "epoch": 8.52, "learning_rate": 2.2520606330912152e-08, "loss": 0.0133, "step": 21617 }, { "epoch": 8.52, "learning_rate": 2.2508836862405034e-08, "loss": 0.0102, "step": 21618 }, { "epoch": 8.52, "learning_rate": 2.2497070286734177e-08, "loss": 0.0098, "step": 21619 }, { "epoch": 8.52, "learning_rate": 2.2485306604091246e-08, "loss": 0.0091, "step": 21620 }, { "epoch": 8.52, "learning_rate": 2.247354581466805e-08, "loss": 0.0107, "step": 21621 }, { "epoch": 8.52, "learning_rate": 2.246178791865627e-08, "loss": 0.0127, "step": 21622 }, { "epoch": 8.52, "learning_rate": 2.2450032916247473e-08, "loss": 0.0127, "step": 21623 }, { "epoch": 8.52, "learning_rate": 2.2438280807633324e-08, "loss": 0.0122, "step": 21624 }, { "epoch": 8.52, "learning_rate": 2.2426531593005337e-08, "loss": 0.0104, "step": 21625 }, { "epoch": 8.52, "learning_rate": 2.2414785272555003e-08, "loss": 0.0081, "step": 21626 }, { "epoch": 8.52, "learning_rate": 2.2403041846473745e-08, "loss": 0.0106, "step": 21627 }, { "epoch": 8.52, "learning_rate": 2.2391301314953037e-08, "loss": 0.0094, "step": 21628 }, { "epoch": 8.52, "learning_rate": 2.2379563678184144e-08, "loss": 0.0115, "step": 21629 }, { "epoch": 8.52, "learning_rate": 2.2367828936358402e-08, "loss": 0.0104, "step": 21630 }, { "epoch": 8.52, "learning_rate": 2.235609708966708e-08, "loss": 0.0113, "step": 21631 }, { "epoch": 8.53, "learning_rate": 2.234436813830145e-08, "loss": 0.0095, "step": 21632 }, { "epoch": 8.53, "learning_rate": 2.2332642082452557e-08, "loss": 0.0091, "step": 21633 }, { "epoch": 8.53, "learning_rate": 2.232091892231156e-08, "loss": 0.0095, "step": 21634 }, { "epoch": 8.53, "learning_rate": 2.2309198658069573e-08, "loss": 0.011, "step": 21635 }, { "epoch": 8.53, "learning_rate": 2.2297481289917595e-08, "loss": 0.0104, "step": 21636 }, { "epoch": 8.53, "learning_rate": 2.2285766818046593e-08, "loss": 0.0126, "step": 21637 }, { "epoch": 8.53, "learning_rate": 2.2274055242647516e-08, "loss": 0.0084, "step": 21638 }, { "epoch": 8.53, "learning_rate": 2.226234656391115e-08, "loss": 0.0104, "step": 21639 }, { "epoch": 8.53, "learning_rate": 2.2250640782028472e-08, "loss": 0.009, "step": 21640 }, { "epoch": 8.53, "learning_rate": 2.223893789719018e-08, "loss": 0.0075, "step": 21641 }, { "epoch": 8.53, "learning_rate": 2.2227237909587072e-08, "loss": 0.0085, "step": 21642 }, { "epoch": 8.53, "learning_rate": 2.221554081940975e-08, "loss": 0.0094, "step": 21643 }, { "epoch": 8.53, "learning_rate": 2.220384662684893e-08, "loss": 0.0109, "step": 21644 }, { "epoch": 8.53, "learning_rate": 2.219215533209522e-08, "loss": 0.0097, "step": 21645 }, { "epoch": 8.53, "learning_rate": 2.2180466935339104e-08, "loss": 0.0111, "step": 21646 }, { "epoch": 8.53, "learning_rate": 2.216878143677112e-08, "loss": 0.0106, "step": 21647 }, { "epoch": 8.53, "learning_rate": 2.215709883658172e-08, "loss": 0.0121, "step": 21648 }, { "epoch": 8.53, "learning_rate": 2.2145419134961375e-08, "loss": 0.0092, "step": 21649 }, { "epoch": 8.53, "learning_rate": 2.2133742332100326e-08, "loss": 0.0099, "step": 21650 }, { "epoch": 8.53, "learning_rate": 2.212206842818902e-08, "loss": 0.0099, "step": 21651 }, { "epoch": 8.53, "learning_rate": 2.211039742341756e-08, "loss": 0.0127, "step": 21652 }, { "epoch": 8.53, "learning_rate": 2.2098729317976338e-08, "loss": 0.0127, "step": 21653 }, { "epoch": 8.53, "learning_rate": 2.208706411205541e-08, "loss": 0.0103, "step": 21654 }, { "epoch": 8.53, "learning_rate": 2.2075401805844994e-08, "loss": 0.0102, "step": 21655 }, { "epoch": 8.53, "learning_rate": 2.2063742399535057e-08, "loss": 0.0118, "step": 21656 }, { "epoch": 8.54, "learning_rate": 2.205208589331571e-08, "loss": 0.0131, "step": 21657 }, { "epoch": 8.54, "learning_rate": 2.2040432287376888e-08, "loss": 0.0098, "step": 21658 }, { "epoch": 8.54, "learning_rate": 2.2028781581908595e-08, "loss": 0.009, "step": 21659 }, { "epoch": 8.54, "learning_rate": 2.201713377710066e-08, "loss": 0.0103, "step": 21660 }, { "epoch": 8.54, "learning_rate": 2.2005488873142907e-08, "loss": 0.0098, "step": 21661 }, { "epoch": 8.54, "learning_rate": 2.1993846870225185e-08, "loss": 0.0088, "step": 21662 }, { "epoch": 8.54, "learning_rate": 2.1982207768537277e-08, "loss": 0.0136, "step": 21663 }, { "epoch": 8.54, "learning_rate": 2.1970571568268757e-08, "loss": 0.0119, "step": 21664 }, { "epoch": 8.54, "learning_rate": 2.1958938269609394e-08, "loss": 0.0133, "step": 21665 }, { "epoch": 8.54, "learning_rate": 2.1947307872748722e-08, "loss": 0.0096, "step": 21666 }, { "epoch": 8.54, "learning_rate": 2.193568037787632e-08, "loss": 0.0104, "step": 21667 }, { "epoch": 8.54, "learning_rate": 2.1924055785181727e-08, "loss": 0.0113, "step": 21668 }, { "epoch": 8.54, "learning_rate": 2.191243409485435e-08, "loss": 0.0118, "step": 21669 }, { "epoch": 8.54, "learning_rate": 2.1900815307083632e-08, "loss": 0.0076, "step": 21670 }, { "epoch": 8.54, "learning_rate": 2.1889199422058934e-08, "loss": 0.0097, "step": 21671 }, { "epoch": 8.54, "learning_rate": 2.187758643996964e-08, "loss": 0.0091, "step": 21672 }, { "epoch": 8.54, "learning_rate": 2.1865976361004934e-08, "loss": 0.0111, "step": 21673 }, { "epoch": 8.54, "learning_rate": 2.1854369185354083e-08, "loss": 0.0102, "step": 21674 }, { "epoch": 8.54, "learning_rate": 2.1842764913206245e-08, "loss": 0.01, "step": 21675 }, { "epoch": 8.54, "learning_rate": 2.183116354475063e-08, "loss": 0.0102, "step": 21676 }, { "epoch": 8.54, "learning_rate": 2.1819565080176218e-08, "loss": 0.0133, "step": 21677 }, { "epoch": 8.54, "learning_rate": 2.1807969519672143e-08, "loss": 0.0117, "step": 21678 }, { "epoch": 8.54, "learning_rate": 2.1796376863427256e-08, "loss": 0.0105, "step": 21679 }, { "epoch": 8.54, "learning_rate": 2.1784787111630675e-08, "loss": 0.008, "step": 21680 }, { "epoch": 8.54, "learning_rate": 2.177320026447116e-08, "loss": 0.0091, "step": 21681 }, { "epoch": 8.55, "learning_rate": 2.1761616322137645e-08, "loss": 0.0083, "step": 21682 }, { "epoch": 8.55, "learning_rate": 2.1750035284818803e-08, "loss": 0.0074, "step": 21683 }, { "epoch": 8.55, "learning_rate": 2.1738457152703572e-08, "loss": 0.0138, "step": 21684 }, { "epoch": 8.55, "learning_rate": 2.1726881925980513e-08, "loss": 0.0125, "step": 21685 }, { "epoch": 8.55, "learning_rate": 2.1715309604838384e-08, "loss": 0.0086, "step": 21686 }, { "epoch": 8.55, "learning_rate": 2.1703740189465703e-08, "loss": 0.0127, "step": 21687 }, { "epoch": 8.55, "learning_rate": 2.1692173680051074e-08, "loss": 0.0109, "step": 21688 }, { "epoch": 8.55, "learning_rate": 2.168061007678308e-08, "loss": 0.0103, "step": 21689 }, { "epoch": 8.55, "learning_rate": 2.166904937985006e-08, "loss": 0.0112, "step": 21690 }, { "epoch": 8.55, "learning_rate": 2.1657491589440548e-08, "loss": 0.0099, "step": 21691 }, { "epoch": 8.55, "learning_rate": 2.164593670574282e-08, "loss": 0.0134, "step": 21692 }, { "epoch": 8.55, "learning_rate": 2.16343847289453e-08, "loss": 0.0105, "step": 21693 }, { "epoch": 8.55, "learning_rate": 2.1622835659236195e-08, "loss": 0.0111, "step": 21694 }, { "epoch": 8.55, "learning_rate": 2.161128949680382e-08, "loss": 0.0083, "step": 21695 }, { "epoch": 8.55, "learning_rate": 2.159974624183627e-08, "loss": 0.01, "step": 21696 }, { "epoch": 8.55, "learning_rate": 2.1588205894521704e-08, "loss": 0.0088, "step": 21697 }, { "epoch": 8.55, "learning_rate": 2.157666845504822e-08, "loss": 0.0096, "step": 21698 }, { "epoch": 8.55, "learning_rate": 2.156513392360393e-08, "loss": 0.0121, "step": 21699 }, { "epoch": 8.55, "learning_rate": 2.1553602300376704e-08, "loss": 0.0097, "step": 21700 }, { "epoch": 8.55, "learning_rate": 2.1542073585554577e-08, "loss": 0.0111, "step": 21701 }, { "epoch": 8.55, "learning_rate": 2.1530547779325415e-08, "loss": 0.009, "step": 21702 }, { "epoch": 8.55, "learning_rate": 2.1519024881877113e-08, "loss": 0.0104, "step": 21703 }, { "epoch": 8.55, "learning_rate": 2.150750489339741e-08, "loss": 0.0091, "step": 21704 }, { "epoch": 8.55, "learning_rate": 2.1495987814074156e-08, "loss": 0.0103, "step": 21705 }, { "epoch": 8.55, "learning_rate": 2.1484473644094915e-08, "loss": 0.0098, "step": 21706 }, { "epoch": 8.55, "learning_rate": 2.1472962383647507e-08, "loss": 0.0098, "step": 21707 }, { "epoch": 8.56, "learning_rate": 2.1461454032919458e-08, "loss": 0.0109, "step": 21708 }, { "epoch": 8.56, "learning_rate": 2.1449948592098387e-08, "loss": 0.0123, "step": 21709 }, { "epoch": 8.56, "learning_rate": 2.1438446061371772e-08, "loss": 0.0084, "step": 21710 }, { "epoch": 8.56, "learning_rate": 2.1426946440927084e-08, "loss": 0.0116, "step": 21711 }, { "epoch": 8.56, "learning_rate": 2.1415449730951818e-08, "loss": 0.0111, "step": 21712 }, { "epoch": 8.56, "learning_rate": 2.1403955931633268e-08, "loss": 0.0094, "step": 21713 }, { "epoch": 8.56, "learning_rate": 2.139246504315879e-08, "loss": 0.0104, "step": 21714 }, { "epoch": 8.56, "learning_rate": 2.138097706571569e-08, "loss": 0.0097, "step": 21715 }, { "epoch": 8.56, "learning_rate": 2.136949199949123e-08, "loss": 0.0106, "step": 21716 }, { "epoch": 8.56, "learning_rate": 2.1358009844672532e-08, "loss": 0.01, "step": 21717 }, { "epoch": 8.56, "learning_rate": 2.13465306014468e-08, "loss": 0.01, "step": 21718 }, { "epoch": 8.56, "learning_rate": 2.133505427000102e-08, "loss": 0.0087, "step": 21719 }, { "epoch": 8.56, "learning_rate": 2.1323580850522393e-08, "loss": 0.0136, "step": 21720 }, { "epoch": 8.56, "learning_rate": 2.1312110343197796e-08, "loss": 0.0094, "step": 21721 }, { "epoch": 8.56, "learning_rate": 2.130064274821428e-08, "loss": 0.0097, "step": 21722 }, { "epoch": 8.56, "learning_rate": 2.1289178065758606e-08, "loss": 0.011, "step": 21723 }, { "epoch": 8.56, "learning_rate": 2.12777162960178e-08, "loss": 0.0103, "step": 21724 }, { "epoch": 8.56, "learning_rate": 2.126625743917856e-08, "loss": 0.0096, "step": 21725 }, { "epoch": 8.56, "learning_rate": 2.125480149542771e-08, "loss": 0.0097, "step": 21726 }, { "epoch": 8.56, "learning_rate": 2.1243348464951905e-08, "loss": 0.0094, "step": 21727 }, { "epoch": 8.56, "learning_rate": 2.1231898347937836e-08, "loss": 0.0118, "step": 21728 }, { "epoch": 8.56, "learning_rate": 2.1220451144572115e-08, "loss": 0.0098, "step": 21729 }, { "epoch": 8.56, "learning_rate": 2.120900685504139e-08, "loss": 0.0097, "step": 21730 }, { "epoch": 8.56, "learning_rate": 2.1197565479532065e-08, "loss": 0.009, "step": 21731 }, { "epoch": 8.56, "learning_rate": 2.1186127018230683e-08, "loss": 0.0097, "step": 21732 }, { "epoch": 8.57, "learning_rate": 2.117469147132369e-08, "loss": 0.0095, "step": 21733 }, { "epoch": 8.57, "learning_rate": 2.1163258838997412e-08, "loss": 0.0101, "step": 21734 }, { "epoch": 8.57, "learning_rate": 2.1151829121438268e-08, "loss": 0.0092, "step": 21735 }, { "epoch": 8.57, "learning_rate": 2.1140402318832428e-08, "loss": 0.0101, "step": 21736 }, { "epoch": 8.57, "learning_rate": 2.1128978431366206e-08, "loss": 0.0093, "step": 21737 }, { "epoch": 8.57, "learning_rate": 2.1117557459225787e-08, "loss": 0.0099, "step": 21738 }, { "epoch": 8.57, "learning_rate": 2.1106139402597356e-08, "loss": 0.0092, "step": 21739 }, { "epoch": 8.57, "learning_rate": 2.109472426166692e-08, "loss": 0.0096, "step": 21740 }, { "epoch": 8.57, "learning_rate": 2.108331203662057e-08, "loss": 0.0084, "step": 21741 }, { "epoch": 8.57, "learning_rate": 2.107190272764432e-08, "loss": 0.0096, "step": 21742 }, { "epoch": 8.57, "learning_rate": 2.106049633492415e-08, "loss": 0.0114, "step": 21743 }, { "epoch": 8.57, "learning_rate": 2.104909285864589e-08, "loss": 0.0105, "step": 21744 }, { "epoch": 8.57, "learning_rate": 2.103769229899548e-08, "loss": 0.0113, "step": 21745 }, { "epoch": 8.57, "learning_rate": 2.1026294656158616e-08, "loss": 0.0125, "step": 21746 }, { "epoch": 8.57, "learning_rate": 2.1014899930321218e-08, "loss": 0.0104, "step": 21747 }, { "epoch": 8.57, "learning_rate": 2.100350812166891e-08, "loss": 0.0114, "step": 21748 }, { "epoch": 8.57, "learning_rate": 2.0992119230387395e-08, "loss": 0.0101, "step": 21749 }, { "epoch": 8.57, "learning_rate": 2.098073325666223e-08, "loss": 0.0115, "step": 21750 }, { "epoch": 8.57, "learning_rate": 2.096935020067905e-08, "loss": 0.009, "step": 21751 }, { "epoch": 8.57, "learning_rate": 2.095797006262334e-08, "loss": 0.0111, "step": 21752 }, { "epoch": 8.57, "learning_rate": 2.0946592842680678e-08, "loss": 0.0102, "step": 21753 }, { "epoch": 8.57, "learning_rate": 2.093521854103637e-08, "loss": 0.013, "step": 21754 }, { "epoch": 8.57, "learning_rate": 2.092384715787585e-08, "loss": 0.0085, "step": 21755 }, { "epoch": 8.57, "learning_rate": 2.0912478693384505e-08, "loss": 0.0105, "step": 21756 }, { "epoch": 8.57, "learning_rate": 2.0901113147747516e-08, "loss": 0.0096, "step": 21757 }, { "epoch": 8.58, "learning_rate": 2.0889750521150253e-08, "loss": 0.0082, "step": 21758 }, { "epoch": 8.58, "learning_rate": 2.0878390813777734e-08, "loss": 0.0115, "step": 21759 }, { "epoch": 8.58, "learning_rate": 2.086703402581531e-08, "loss": 0.0109, "step": 21760 }, { "epoch": 8.58, "learning_rate": 2.0855680157447942e-08, "loss": 0.0077, "step": 21761 }, { "epoch": 8.58, "learning_rate": 2.0844329208860746e-08, "loss": 0.011, "step": 21762 }, { "epoch": 8.58, "learning_rate": 2.0832981180238618e-08, "loss": 0.0109, "step": 21763 }, { "epoch": 8.58, "learning_rate": 2.082163607176668e-08, "loss": 0.0089, "step": 21764 }, { "epoch": 8.58, "learning_rate": 2.0810293883629714e-08, "loss": 0.0103, "step": 21765 }, { "epoch": 8.58, "learning_rate": 2.0798954616012663e-08, "loss": 0.0094, "step": 21766 }, { "epoch": 8.58, "learning_rate": 2.0787618269100248e-08, "loss": 0.0105, "step": 21767 }, { "epoch": 8.58, "learning_rate": 2.0776284843077295e-08, "loss": 0.0116, "step": 21768 }, { "epoch": 8.58, "learning_rate": 2.0764954338128505e-08, "loss": 0.0109, "step": 21769 }, { "epoch": 8.58, "learning_rate": 2.0753626754438593e-08, "loss": 0.013, "step": 21770 }, { "epoch": 8.58, "learning_rate": 2.07423020921921e-08, "loss": 0.0091, "step": 21771 }, { "epoch": 8.58, "learning_rate": 2.0730980351573657e-08, "loss": 0.0088, "step": 21772 }, { "epoch": 8.58, "learning_rate": 2.0719661532767806e-08, "loss": 0.0109, "step": 21773 }, { "epoch": 8.58, "learning_rate": 2.0708345635958958e-08, "loss": 0.0096, "step": 21774 }, { "epoch": 8.58, "learning_rate": 2.069703266133158e-08, "loss": 0.0082, "step": 21775 }, { "epoch": 8.58, "learning_rate": 2.068572260907009e-08, "loss": 0.011, "step": 21776 }, { "epoch": 8.58, "learning_rate": 2.0674415479358775e-08, "loss": 0.011, "step": 21777 }, { "epoch": 8.58, "learning_rate": 2.066311127238194e-08, "loss": 0.0095, "step": 21778 }, { "epoch": 8.58, "learning_rate": 2.0651809988323877e-08, "loss": 0.0093, "step": 21779 }, { "epoch": 8.58, "learning_rate": 2.0640511627368683e-08, "loss": 0.0082, "step": 21780 }, { "epoch": 8.58, "learning_rate": 2.0629216189700548e-08, "loss": 0.0114, "step": 21781 }, { "epoch": 8.58, "learning_rate": 2.061792367550359e-08, "loss": 0.0095, "step": 21782 }, { "epoch": 8.58, "learning_rate": 2.060663408496188e-08, "loss": 0.0105, "step": 21783 }, { "epoch": 8.59, "learning_rate": 2.0595347418259345e-08, "loss": 0.0103, "step": 21784 }, { "epoch": 8.59, "learning_rate": 2.0584063675580033e-08, "loss": 0.0093, "step": 21785 }, { "epoch": 8.59, "learning_rate": 2.0572782857107708e-08, "loss": 0.0098, "step": 21786 }, { "epoch": 8.59, "learning_rate": 2.0561504963026423e-08, "loss": 0.0109, "step": 21787 }, { "epoch": 8.59, "learning_rate": 2.0550229993519853e-08, "loss": 0.011, "step": 21788 }, { "epoch": 8.59, "learning_rate": 2.053895794877183e-08, "loss": 0.0096, "step": 21789 }, { "epoch": 8.59, "learning_rate": 2.052768882896596e-08, "loss": 0.0086, "step": 21790 }, { "epoch": 8.59, "learning_rate": 2.051642263428608e-08, "loss": 0.0082, "step": 21791 }, { "epoch": 8.59, "learning_rate": 2.0505159364915682e-08, "loss": 0.0097, "step": 21792 }, { "epoch": 8.59, "learning_rate": 2.049389902103844e-08, "loss": 0.0108, "step": 21793 }, { "epoch": 8.59, "learning_rate": 2.048264160283777e-08, "loss": 0.0116, "step": 21794 }, { "epoch": 8.59, "learning_rate": 2.0471387110497208e-08, "loss": 0.011, "step": 21795 }, { "epoch": 8.59, "learning_rate": 2.046013554420023e-08, "loss": 0.0101, "step": 21796 }, { "epoch": 8.59, "learning_rate": 2.0448886904130135e-08, "loss": 0.0091, "step": 21797 }, { "epoch": 8.59, "learning_rate": 2.0437641190470312e-08, "loss": 0.0114, "step": 21798 }, { "epoch": 8.59, "learning_rate": 2.0426398403404012e-08, "loss": 0.0097, "step": 21799 }, { "epoch": 8.59, "learning_rate": 2.0415158543114553e-08, "loss": 0.0094, "step": 21800 }, { "epoch": 8.59, "learning_rate": 2.0403921609785014e-08, "loss": 0.0123, "step": 21801 }, { "epoch": 8.59, "learning_rate": 2.0392687603598645e-08, "loss": 0.0115, "step": 21802 }, { "epoch": 8.59, "learning_rate": 2.0381456524738415e-08, "loss": 0.009, "step": 21803 }, { "epoch": 8.59, "learning_rate": 2.0370228373387555e-08, "loss": 0.0106, "step": 21804 }, { "epoch": 8.59, "learning_rate": 2.0359003149728914e-08, "loss": 0.0124, "step": 21805 }, { "epoch": 8.59, "learning_rate": 2.0347780853945528e-08, "loss": 0.0122, "step": 21806 }, { "epoch": 8.59, "learning_rate": 2.033656148622025e-08, "loss": 0.0078, "step": 21807 }, { "epoch": 8.59, "learning_rate": 2.0325345046735952e-08, "loss": 0.0102, "step": 21808 }, { "epoch": 8.6, "learning_rate": 2.031413153567545e-08, "loss": 0.0132, "step": 21809 }, { "epoch": 8.6, "learning_rate": 2.030292095322157e-08, "loss": 0.0086, "step": 21810 }, { "epoch": 8.6, "learning_rate": 2.0291713299556923e-08, "loss": 0.0085, "step": 21811 }, { "epoch": 8.6, "learning_rate": 2.0280508574864208e-08, "loss": 0.0102, "step": 21812 }, { "epoch": 8.6, "learning_rate": 2.026930677932608e-08, "loss": 0.0109, "step": 21813 }, { "epoch": 8.6, "learning_rate": 2.025810791312512e-08, "loss": 0.0097, "step": 21814 }, { "epoch": 8.6, "learning_rate": 2.0246911976443816e-08, "loss": 0.0105, "step": 21815 }, { "epoch": 8.6, "learning_rate": 2.0235718969464654e-08, "loss": 0.011, "step": 21816 }, { "epoch": 8.6, "learning_rate": 2.022452889237005e-08, "loss": 0.0093, "step": 21817 }, { "epoch": 8.6, "learning_rate": 2.021334174534237e-08, "loss": 0.0102, "step": 21818 }, { "epoch": 8.6, "learning_rate": 2.0202157528564045e-08, "loss": 0.0135, "step": 21819 }, { "epoch": 8.6, "learning_rate": 2.0190976242217238e-08, "loss": 0.0091, "step": 21820 }, { "epoch": 8.6, "learning_rate": 2.0179797886484252e-08, "loss": 0.012, "step": 21821 }, { "epoch": 8.6, "learning_rate": 2.0168622461547246e-08, "loss": 0.0101, "step": 21822 }, { "epoch": 8.6, "learning_rate": 2.0157449967588434e-08, "loss": 0.0099, "step": 21823 }, { "epoch": 8.6, "learning_rate": 2.0146280404789806e-08, "loss": 0.0097, "step": 21824 }, { "epoch": 8.6, "learning_rate": 2.01351137733335e-08, "loss": 0.0086, "step": 21825 }, { "epoch": 8.6, "learning_rate": 2.012395007340142e-08, "loss": 0.0107, "step": 21826 }, { "epoch": 8.6, "learning_rate": 2.0112789305175636e-08, "loss": 0.0092, "step": 21827 }, { "epoch": 8.6, "learning_rate": 2.0101631468837943e-08, "loss": 0.0102, "step": 21828 }, { "epoch": 8.6, "learning_rate": 2.0090476564570302e-08, "loss": 0.008, "step": 21829 }, { "epoch": 8.6, "learning_rate": 2.007932459255437e-08, "loss": 0.0128, "step": 21830 }, { "epoch": 8.6, "learning_rate": 2.006817555297209e-08, "loss": 0.0086, "step": 21831 }, { "epoch": 8.6, "learning_rate": 2.0057029446005025e-08, "loss": 0.0094, "step": 21832 }, { "epoch": 8.6, "learning_rate": 2.0045886271834945e-08, "loss": 0.0118, "step": 21833 }, { "epoch": 8.6, "learning_rate": 2.0034746030643345e-08, "loss": 0.0111, "step": 21834 }, { "epoch": 8.61, "learning_rate": 2.0023608722611952e-08, "loss": 0.0108, "step": 21835 }, { "epoch": 8.61, "learning_rate": 2.0012474347922148e-08, "loss": 0.0099, "step": 21836 }, { "epoch": 8.61, "learning_rate": 2.0001342906755525e-08, "loss": 0.0112, "step": 21837 }, { "epoch": 8.61, "learning_rate": 1.9990214399293404e-08, "loss": 0.0093, "step": 21838 }, { "epoch": 8.61, "learning_rate": 1.9979088825717196e-08, "loss": 0.0095, "step": 21839 }, { "epoch": 8.61, "learning_rate": 1.9967966186208262e-08, "loss": 0.0121, "step": 21840 }, { "epoch": 8.61, "learning_rate": 1.995684648094784e-08, "loss": 0.0099, "step": 21841 }, { "epoch": 8.61, "learning_rate": 1.9945729710117187e-08, "loss": 0.0104, "step": 21842 }, { "epoch": 8.61, "learning_rate": 1.993461587389751e-08, "loss": 0.0119, "step": 21843 }, { "epoch": 8.61, "learning_rate": 1.992350497246993e-08, "loss": 0.0097, "step": 21844 }, { "epoch": 8.61, "learning_rate": 1.9912397006015503e-08, "loss": 0.009, "step": 21845 }, { "epoch": 8.61, "learning_rate": 1.990129197471535e-08, "loss": 0.0117, "step": 21846 }, { "epoch": 8.61, "learning_rate": 1.9890189878750396e-08, "loss": 0.01, "step": 21847 }, { "epoch": 8.61, "learning_rate": 1.9879090718301582e-08, "loss": 0.0097, "step": 21848 }, { "epoch": 8.61, "learning_rate": 1.9867994493549857e-08, "loss": 0.0106, "step": 21849 }, { "epoch": 8.61, "learning_rate": 1.985690120467609e-08, "loss": 0.0113, "step": 21850 }, { "epoch": 8.61, "learning_rate": 1.984581085186099e-08, "loss": 0.0094, "step": 21851 }, { "epoch": 8.61, "learning_rate": 1.9834723435285384e-08, "loss": 0.0118, "step": 21852 }, { "epoch": 8.61, "learning_rate": 1.982363895512997e-08, "loss": 0.0102, "step": 21853 }, { "epoch": 8.61, "learning_rate": 1.9812557411575438e-08, "loss": 0.0104, "step": 21854 }, { "epoch": 8.61, "learning_rate": 1.980147880480232e-08, "loss": 0.0093, "step": 21855 }, { "epoch": 8.61, "learning_rate": 1.9790403134991273e-08, "loss": 0.0091, "step": 21856 }, { "epoch": 8.61, "learning_rate": 1.977933040232269e-08, "loss": 0.0089, "step": 21857 }, { "epoch": 8.61, "learning_rate": 1.9768260606977205e-08, "loss": 0.0118, "step": 21858 }, { "epoch": 8.61, "learning_rate": 1.9757193749135114e-08, "loss": 0.0085, "step": 21859 }, { "epoch": 8.62, "learning_rate": 1.9746129828976853e-08, "loss": 0.0117, "step": 21860 }, { "epoch": 8.62, "learning_rate": 1.973506884668268e-08, "loss": 0.0108, "step": 21861 }, { "epoch": 8.62, "learning_rate": 1.9724010802432932e-08, "loss": 0.0109, "step": 21862 }, { "epoch": 8.62, "learning_rate": 1.971295569640783e-08, "loss": 0.011, "step": 21863 }, { "epoch": 8.62, "learning_rate": 1.9701903528787534e-08, "loss": 0.011, "step": 21864 }, { "epoch": 8.62, "learning_rate": 1.969085429975217e-08, "loss": 0.0127, "step": 21865 }, { "epoch": 8.62, "learning_rate": 1.967980800948186e-08, "loss": 0.0105, "step": 21866 }, { "epoch": 8.62, "learning_rate": 1.9668764658156657e-08, "loss": 0.0137, "step": 21867 }, { "epoch": 8.62, "learning_rate": 1.9657724245956465e-08, "loss": 0.0115, "step": 21868 }, { "epoch": 8.62, "learning_rate": 1.9646686773061338e-08, "loss": 0.0101, "step": 21869 }, { "epoch": 8.62, "learning_rate": 1.9635652239651025e-08, "loss": 0.0118, "step": 21870 }, { "epoch": 8.62, "learning_rate": 1.9624620645905533e-08, "loss": 0.0124, "step": 21871 }, { "epoch": 8.62, "learning_rate": 1.9613591992004563e-08, "loss": 0.0089, "step": 21872 }, { "epoch": 8.62, "learning_rate": 1.9602566278127908e-08, "loss": 0.012, "step": 21873 }, { "epoch": 8.62, "learning_rate": 1.9591543504455178e-08, "loss": 0.0109, "step": 21874 }, { "epoch": 8.62, "learning_rate": 1.9580523671166182e-08, "loss": 0.011, "step": 21875 }, { "epoch": 8.62, "learning_rate": 1.9569506778440382e-08, "loss": 0.0123, "step": 21876 }, { "epoch": 8.62, "learning_rate": 1.9558492826457475e-08, "loss": 0.0095, "step": 21877 }, { "epoch": 8.62, "learning_rate": 1.9547481815396827e-08, "loss": 0.01, "step": 21878 }, { "epoch": 8.62, "learning_rate": 1.9536473745437988e-08, "loss": 0.0102, "step": 21879 }, { "epoch": 8.62, "learning_rate": 1.9525468616760342e-08, "loss": 0.0106, "step": 21880 }, { "epoch": 8.62, "learning_rate": 1.9514466429543306e-08, "loss": 0.0096, "step": 21881 }, { "epoch": 8.62, "learning_rate": 1.950346718396614e-08, "loss": 0.0095, "step": 21882 }, { "epoch": 8.62, "learning_rate": 1.9492470880208112e-08, "loss": 0.0105, "step": 21883 }, { "epoch": 8.62, "learning_rate": 1.948147751844851e-08, "loss": 0.0122, "step": 21884 }, { "epoch": 8.63, "learning_rate": 1.947048709886645e-08, "loss": 0.0092, "step": 21885 }, { "epoch": 8.63, "learning_rate": 1.9459499621641106e-08, "loss": 0.0114, "step": 21886 }, { "epoch": 8.63, "learning_rate": 1.9448515086951468e-08, "loss": 0.0105, "step": 21887 }, { "epoch": 8.63, "learning_rate": 1.9437533494976655e-08, "loss": 0.0146, "step": 21888 }, { "epoch": 8.63, "learning_rate": 1.9426554845895616e-08, "loss": 0.01, "step": 21889 }, { "epoch": 8.63, "learning_rate": 1.9415579139887317e-08, "loss": 0.0113, "step": 21890 }, { "epoch": 8.63, "learning_rate": 1.9404606377130594e-08, "loss": 0.0123, "step": 21891 }, { "epoch": 8.63, "learning_rate": 1.9393636557804327e-08, "loss": 0.0087, "step": 21892 }, { "epoch": 8.63, "learning_rate": 1.9382669682087283e-08, "loss": 0.0093, "step": 21893 }, { "epoch": 8.63, "learning_rate": 1.937170575015825e-08, "loss": 0.0096, "step": 21894 }, { "epoch": 8.63, "learning_rate": 1.9360744762195846e-08, "loss": 0.0099, "step": 21895 }, { "epoch": 8.63, "learning_rate": 1.9349786718378813e-08, "loss": 0.0095, "step": 21896 }, { "epoch": 8.63, "learning_rate": 1.9338831618885633e-08, "loss": 0.0127, "step": 21897 }, { "epoch": 8.63, "learning_rate": 1.932787946389498e-08, "loss": 0.0106, "step": 21898 }, { "epoch": 8.63, "learning_rate": 1.9316930253585274e-08, "loss": 0.0103, "step": 21899 }, { "epoch": 8.63, "learning_rate": 1.9305983988135033e-08, "loss": 0.0092, "step": 21900 }, { "epoch": 8.63, "learning_rate": 1.9295040667722585e-08, "loss": 0.0159, "step": 21901 }, { "epoch": 8.63, "learning_rate": 1.928410029252634e-08, "loss": 0.0105, "step": 21902 }, { "epoch": 8.63, "learning_rate": 1.9273162862724625e-08, "loss": 0.0099, "step": 21903 }, { "epoch": 8.63, "learning_rate": 1.9262228378495694e-08, "loss": 0.0128, "step": 21904 }, { "epoch": 8.63, "learning_rate": 1.9251296840017718e-08, "loss": 0.0109, "step": 21905 }, { "epoch": 8.63, "learning_rate": 1.9240368247468885e-08, "loss": 0.0118, "step": 21906 }, { "epoch": 8.63, "learning_rate": 1.922944260102739e-08, "loss": 0.0109, "step": 21907 }, { "epoch": 8.63, "learning_rate": 1.92185199008712e-08, "loss": 0.0096, "step": 21908 }, { "epoch": 8.63, "learning_rate": 1.9207600147178394e-08, "loss": 0.0122, "step": 21909 }, { "epoch": 8.63, "learning_rate": 1.9196683340126872e-08, "loss": 0.0113, "step": 21910 }, { "epoch": 8.64, "learning_rate": 1.91857694798947e-08, "loss": 0.0114, "step": 21911 }, { "epoch": 8.64, "learning_rate": 1.9174858566659634e-08, "loss": 0.0128, "step": 21912 }, { "epoch": 8.64, "learning_rate": 1.9163950600599586e-08, "loss": 0.0094, "step": 21913 }, { "epoch": 8.64, "learning_rate": 1.9153045581892256e-08, "loss": 0.0104, "step": 21914 }, { "epoch": 8.64, "learning_rate": 1.914214351071548e-08, "loss": 0.011, "step": 21915 }, { "epoch": 8.64, "learning_rate": 1.9131244387246846e-08, "loss": 0.01, "step": 21916 }, { "epoch": 8.64, "learning_rate": 1.9120348211664107e-08, "loss": 0.011, "step": 21917 }, { "epoch": 8.64, "learning_rate": 1.9109454984144713e-08, "loss": 0.009, "step": 21918 }, { "epoch": 8.64, "learning_rate": 1.9098564704866303e-08, "loss": 0.0104, "step": 21919 }, { "epoch": 8.64, "learning_rate": 1.908767737400634e-08, "loss": 0.0103, "step": 21920 }, { "epoch": 8.64, "learning_rate": 1.907679299174234e-08, "loss": 0.0124, "step": 21921 }, { "epoch": 8.64, "learning_rate": 1.906591155825159e-08, "loss": 0.0106, "step": 21922 }, { "epoch": 8.64, "learning_rate": 1.905503307371148e-08, "loss": 0.0107, "step": 21923 }, { "epoch": 8.64, "learning_rate": 1.9044157538299376e-08, "loss": 0.0102, "step": 21924 }, { "epoch": 8.64, "learning_rate": 1.9033284952192452e-08, "loss": 0.0085, "step": 21925 }, { "epoch": 8.64, "learning_rate": 1.902241531556794e-08, "loss": 0.0094, "step": 21926 }, { "epoch": 8.64, "learning_rate": 1.9011548628603036e-08, "loss": 0.011, "step": 21927 }, { "epoch": 8.64, "learning_rate": 1.9000684891474795e-08, "loss": 0.0135, "step": 21928 }, { "epoch": 8.64, "learning_rate": 1.8989824104360296e-08, "loss": 0.0091, "step": 21929 }, { "epoch": 8.64, "learning_rate": 1.897896626743658e-08, "loss": 0.0105, "step": 21930 }, { "epoch": 8.64, "learning_rate": 1.896811138088059e-08, "loss": 0.0098, "step": 21931 }, { "epoch": 8.64, "learning_rate": 1.895725944486921e-08, "loss": 0.0089, "step": 21932 }, { "epoch": 8.64, "learning_rate": 1.8946410459579386e-08, "loss": 0.0107, "step": 21933 }, { "epoch": 8.64, "learning_rate": 1.893556442518791e-08, "loss": 0.0129, "step": 21934 }, { "epoch": 8.64, "learning_rate": 1.892472134187153e-08, "loss": 0.0085, "step": 21935 }, { "epoch": 8.65, "learning_rate": 1.8913881209807015e-08, "loss": 0.0085, "step": 21936 }, { "epoch": 8.65, "learning_rate": 1.8903044029170933e-08, "loss": 0.0123, "step": 21937 }, { "epoch": 8.65, "learning_rate": 1.8892209800140103e-08, "loss": 0.0105, "step": 21938 }, { "epoch": 8.65, "learning_rate": 1.8881378522890935e-08, "loss": 0.0104, "step": 21939 }, { "epoch": 8.65, "learning_rate": 1.8870550197600087e-08, "loss": 0.0103, "step": 21940 }, { "epoch": 8.65, "learning_rate": 1.885972482444389e-08, "loss": 0.0104, "step": 21941 }, { "epoch": 8.65, "learning_rate": 1.8848902403598955e-08, "loss": 0.0096, "step": 21942 }, { "epoch": 8.65, "learning_rate": 1.883808293524156e-08, "loss": 0.0115, "step": 21943 }, { "epoch": 8.65, "learning_rate": 1.8827266419548127e-08, "loss": 0.0093, "step": 21944 }, { "epoch": 8.65, "learning_rate": 1.8816452856694863e-08, "loss": 0.0111, "step": 21945 }, { "epoch": 8.65, "learning_rate": 1.8805642246858033e-08, "loss": 0.0145, "step": 21946 }, { "epoch": 8.65, "learning_rate": 1.879483459021389e-08, "loss": 0.0078, "step": 21947 }, { "epoch": 8.65, "learning_rate": 1.8784029886938523e-08, "loss": 0.0104, "step": 21948 }, { "epoch": 8.65, "learning_rate": 1.877322813720803e-08, "loss": 0.0107, "step": 21949 }, { "epoch": 8.65, "learning_rate": 1.8762429341198517e-08, "loss": 0.0096, "step": 21950 }, { "epoch": 8.65, "learning_rate": 1.8751633499085974e-08, "loss": 0.0115, "step": 21951 }, { "epoch": 8.65, "learning_rate": 1.8740840611046283e-08, "loss": 0.0115, "step": 21952 }, { "epoch": 8.65, "learning_rate": 1.8730050677255482e-08, "loss": 0.0127, "step": 21953 }, { "epoch": 8.65, "learning_rate": 1.8719263697889254e-08, "loss": 0.0075, "step": 21954 }, { "epoch": 8.65, "learning_rate": 1.870847967312359e-08, "loss": 0.01, "step": 21955 }, { "epoch": 8.65, "learning_rate": 1.8697698603134147e-08, "loss": 0.01, "step": 21956 }, { "epoch": 8.65, "learning_rate": 1.8686920488096702e-08, "loss": 0.0077, "step": 21957 }, { "epoch": 8.65, "learning_rate": 1.8676145328186864e-08, "loss": 0.0105, "step": 21958 }, { "epoch": 8.65, "learning_rate": 1.8665373123580276e-08, "loss": 0.0112, "step": 21959 }, { "epoch": 8.65, "learning_rate": 1.8654603874452502e-08, "loss": 0.0087, "step": 21960 }, { "epoch": 8.66, "learning_rate": 1.8643837580979094e-08, "loss": 0.0108, "step": 21961 }, { "epoch": 8.66, "learning_rate": 1.863307424333549e-08, "loss": 0.0121, "step": 21962 }, { "epoch": 8.66, "learning_rate": 1.8622313861697103e-08, "loss": 0.0112, "step": 21963 }, { "epoch": 8.66, "learning_rate": 1.8611556436239374e-08, "loss": 0.0117, "step": 21964 }, { "epoch": 8.66, "learning_rate": 1.8600801967137602e-08, "loss": 0.0074, "step": 21965 }, { "epoch": 8.66, "learning_rate": 1.8590050454567053e-08, "loss": 0.014, "step": 21966 }, { "epoch": 8.66, "learning_rate": 1.857930189870298e-08, "loss": 0.0134, "step": 21967 }, { "epoch": 8.66, "learning_rate": 1.856855629972054e-08, "loss": 0.012, "step": 21968 }, { "epoch": 8.66, "learning_rate": 1.8557813657794874e-08, "loss": 0.0115, "step": 21969 }, { "epoch": 8.66, "learning_rate": 1.854707397310109e-08, "loss": 0.0098, "step": 21970 }, { "epoch": 8.66, "learning_rate": 1.853633724581427e-08, "loss": 0.0088, "step": 21971 }, { "epoch": 8.66, "learning_rate": 1.852560347610932e-08, "loss": 0.0109, "step": 21972 }, { "epoch": 8.66, "learning_rate": 1.8514872664161207e-08, "loss": 0.0086, "step": 21973 }, { "epoch": 8.66, "learning_rate": 1.8504144810144905e-08, "loss": 0.013, "step": 21974 }, { "epoch": 8.66, "learning_rate": 1.8493419914235142e-08, "loss": 0.0117, "step": 21975 }, { "epoch": 8.66, "learning_rate": 1.8482697976606843e-08, "loss": 0.0083, "step": 21976 }, { "epoch": 8.66, "learning_rate": 1.84719789974346e-08, "loss": 0.0108, "step": 21977 }, { "epoch": 8.66, "learning_rate": 1.8461262976893276e-08, "loss": 0.0101, "step": 21978 }, { "epoch": 8.66, "learning_rate": 1.845054991515742e-08, "loss": 0.0115, "step": 21979 }, { "epoch": 8.66, "learning_rate": 1.8439839812401737e-08, "loss": 0.0132, "step": 21980 }, { "epoch": 8.66, "learning_rate": 1.8429132668800617e-08, "loss": 0.0117, "step": 21981 }, { "epoch": 8.66, "learning_rate": 1.841842848452877e-08, "loss": 0.0112, "step": 21982 }, { "epoch": 8.66, "learning_rate": 1.840772725976054e-08, "loss": 0.0081, "step": 21983 }, { "epoch": 8.66, "learning_rate": 1.8397028994670394e-08, "loss": 0.0096, "step": 21984 }, { "epoch": 8.66, "learning_rate": 1.838633368943263e-08, "loss": 0.0089, "step": 21985 }, { "epoch": 8.66, "learning_rate": 1.8375641344221626e-08, "loss": 0.0108, "step": 21986 }, { "epoch": 8.67, "learning_rate": 1.8364951959211615e-08, "loss": 0.0125, "step": 21987 }, { "epoch": 8.67, "learning_rate": 1.835426553457686e-08, "loss": 0.0098, "step": 21988 }, { "epoch": 8.67, "learning_rate": 1.8343582070491492e-08, "loss": 0.0102, "step": 21989 }, { "epoch": 8.67, "learning_rate": 1.8332901567129634e-08, "loss": 0.011, "step": 21990 }, { "epoch": 8.67, "learning_rate": 1.8322224024665433e-08, "loss": 0.0098, "step": 21991 }, { "epoch": 8.67, "learning_rate": 1.83115494432728e-08, "loss": 0.0081, "step": 21992 }, { "epoch": 8.67, "learning_rate": 1.8300877823125793e-08, "loss": 0.0106, "step": 21993 }, { "epoch": 8.67, "learning_rate": 1.829020916439832e-08, "loss": 0.0093, "step": 21994 }, { "epoch": 8.67, "learning_rate": 1.8279543467264325e-08, "loss": 0.0097, "step": 21995 }, { "epoch": 8.67, "learning_rate": 1.8268880731897517e-08, "loss": 0.0109, "step": 21996 }, { "epoch": 8.67, "learning_rate": 1.8258220958471826e-08, "loss": 0.0129, "step": 21997 }, { "epoch": 8.67, "learning_rate": 1.8247564147160866e-08, "loss": 0.0097, "step": 21998 }, { "epoch": 8.67, "learning_rate": 1.8236910298138363e-08, "loss": 0.0098, "step": 21999 }, { "epoch": 8.67, "learning_rate": 1.8226259411578005e-08, "loss": 0.0108, "step": 22000 }, { "epoch": 8.67, "learning_rate": 1.821561148765336e-08, "loss": 0.0092, "step": 22001 }, { "epoch": 8.67, "learning_rate": 1.8204966526537957e-08, "loss": 0.0113, "step": 22002 }, { "epoch": 8.67, "learning_rate": 1.819432452840528e-08, "loss": 0.0109, "step": 22003 }, { "epoch": 8.67, "learning_rate": 1.8183685493428814e-08, "loss": 0.0088, "step": 22004 }, { "epoch": 8.67, "learning_rate": 1.8173049421781994e-08, "loss": 0.0093, "step": 22005 }, { "epoch": 8.67, "learning_rate": 1.8162416313638064e-08, "loss": 0.0107, "step": 22006 }, { "epoch": 8.67, "learning_rate": 1.8151786169170437e-08, "loss": 0.0117, "step": 22007 }, { "epoch": 8.67, "learning_rate": 1.8141158988552242e-08, "loss": 0.0109, "step": 22008 }, { "epoch": 8.67, "learning_rate": 1.813053477195683e-08, "loss": 0.0113, "step": 22009 }, { "epoch": 8.67, "learning_rate": 1.8119913519557263e-08, "loss": 0.0095, "step": 22010 }, { "epoch": 8.67, "learning_rate": 1.810929523152671e-08, "loss": 0.0091, "step": 22011 }, { "epoch": 8.68, "learning_rate": 1.809867990803817e-08, "loss": 0.0112, "step": 22012 }, { "epoch": 8.68, "learning_rate": 1.8088067549264686e-08, "loss": 0.0109, "step": 22013 }, { "epoch": 8.68, "learning_rate": 1.8077458155379265e-08, "loss": 0.0126, "step": 22014 }, { "epoch": 8.68, "learning_rate": 1.8066851726554734e-08, "loss": 0.0101, "step": 22015 }, { "epoch": 8.68, "learning_rate": 1.8056248262964035e-08, "loss": 0.0108, "step": 22016 }, { "epoch": 8.68, "learning_rate": 1.8045647764779948e-08, "loss": 0.0127, "step": 22017 }, { "epoch": 8.68, "learning_rate": 1.803505023217533e-08, "loss": 0.0091, "step": 22018 }, { "epoch": 8.68, "learning_rate": 1.802445566532278e-08, "loss": 0.0099, "step": 22019 }, { "epoch": 8.68, "learning_rate": 1.801386406439507e-08, "loss": 0.0095, "step": 22020 }, { "epoch": 8.68, "learning_rate": 1.8003275429564723e-08, "loss": 0.0131, "step": 22021 }, { "epoch": 8.68, "learning_rate": 1.799268976100443e-08, "loss": 0.0118, "step": 22022 }, { "epoch": 8.68, "learning_rate": 1.7982107058886676e-08, "loss": 0.0087, "step": 22023 }, { "epoch": 8.68, "learning_rate": 1.7971527323383963e-08, "loss": 0.0112, "step": 22024 }, { "epoch": 8.68, "learning_rate": 1.7960950554668665e-08, "loss": 0.0105, "step": 22025 }, { "epoch": 8.68, "learning_rate": 1.79503767529132e-08, "loss": 0.0112, "step": 22026 }, { "epoch": 8.68, "learning_rate": 1.793980591828992e-08, "loss": 0.01, "step": 22027 }, { "epoch": 8.68, "learning_rate": 1.792923805097113e-08, "loss": 0.0107, "step": 22028 }, { "epoch": 8.68, "learning_rate": 1.7918673151129003e-08, "loss": 0.0113, "step": 22029 }, { "epoch": 8.68, "learning_rate": 1.7908111218935785e-08, "loss": 0.0084, "step": 22030 }, { "epoch": 8.68, "learning_rate": 1.7897552254563618e-08, "loss": 0.0104, "step": 22031 }, { "epoch": 8.68, "learning_rate": 1.7886996258184594e-08, "loss": 0.0089, "step": 22032 }, { "epoch": 8.68, "learning_rate": 1.787644322997073e-08, "loss": 0.011, "step": 22033 }, { "epoch": 8.68, "learning_rate": 1.7865893170094038e-08, "loss": 0.0129, "step": 22034 }, { "epoch": 8.68, "learning_rate": 1.785534607872652e-08, "loss": 0.0097, "step": 22035 }, { "epoch": 8.68, "learning_rate": 1.784480195604001e-08, "loss": 0.0096, "step": 22036 }, { "epoch": 8.68, "learning_rate": 1.78342608022064e-08, "loss": 0.0095, "step": 22037 }, { "epoch": 8.69, "learning_rate": 1.782372261739744e-08, "loss": 0.0077, "step": 22038 }, { "epoch": 8.69, "learning_rate": 1.7813187401784946e-08, "loss": 0.0094, "step": 22039 }, { "epoch": 8.69, "learning_rate": 1.780265515554058e-08, "loss": 0.0104, "step": 22040 }, { "epoch": 8.69, "learning_rate": 1.779212587883607e-08, "loss": 0.0109, "step": 22041 }, { "epoch": 8.69, "learning_rate": 1.7781599571842952e-08, "loss": 0.0102, "step": 22042 }, { "epoch": 8.69, "learning_rate": 1.777107623473284e-08, "loss": 0.0088, "step": 22043 }, { "epoch": 8.69, "learning_rate": 1.7760555867677195e-08, "loss": 0.0088, "step": 22044 }, { "epoch": 8.69, "learning_rate": 1.7750038470847572e-08, "loss": 0.0109, "step": 22045 }, { "epoch": 8.69, "learning_rate": 1.77395240444153e-08, "loss": 0.0096, "step": 22046 }, { "epoch": 8.69, "learning_rate": 1.7729012588551817e-08, "loss": 0.0089, "step": 22047 }, { "epoch": 8.69, "learning_rate": 1.7718504103428323e-08, "loss": 0.0143, "step": 22048 }, { "epoch": 8.69, "learning_rate": 1.7707998589216256e-08, "loss": 0.0157, "step": 22049 }, { "epoch": 8.69, "learning_rate": 1.7697496046086724e-08, "loss": 0.011, "step": 22050 }, { "epoch": 8.69, "learning_rate": 1.7686996474210992e-08, "loss": 0.0111, "step": 22051 }, { "epoch": 8.69, "learning_rate": 1.7676499873760075e-08, "loss": 0.0093, "step": 22052 }, { "epoch": 8.69, "learning_rate": 1.766600624490513e-08, "loss": 0.01, "step": 22053 }, { "epoch": 8.69, "learning_rate": 1.765551558781715e-08, "loss": 0.0109, "step": 22054 }, { "epoch": 8.69, "learning_rate": 1.7645027902667175e-08, "loss": 0.0095, "step": 22055 }, { "epoch": 8.69, "learning_rate": 1.763454318962605e-08, "loss": 0.0105, "step": 22056 }, { "epoch": 8.69, "learning_rate": 1.7624061448864724e-08, "loss": 0.0086, "step": 22057 }, { "epoch": 8.69, "learning_rate": 1.761358268055406e-08, "loss": 0.0097, "step": 22058 }, { "epoch": 8.69, "learning_rate": 1.760310688486475e-08, "loss": 0.0106, "step": 22059 }, { "epoch": 8.69, "learning_rate": 1.7592634061967627e-08, "loss": 0.0096, "step": 22060 }, { "epoch": 8.69, "learning_rate": 1.758216421203327e-08, "loss": 0.0106, "step": 22061 }, { "epoch": 8.69, "learning_rate": 1.7571697335232472e-08, "loss": 0.0105, "step": 22062 }, { "epoch": 8.7, "learning_rate": 1.75612334317357e-08, "loss": 0.0107, "step": 22063 }, { "epoch": 8.7, "learning_rate": 1.7550772501713595e-08, "loss": 0.0106, "step": 22064 }, { "epoch": 8.7, "learning_rate": 1.7540314545336555e-08, "loss": 0.0084, "step": 22065 }, { "epoch": 8.7, "learning_rate": 1.7529859562775106e-08, "loss": 0.0108, "step": 22066 }, { "epoch": 8.7, "learning_rate": 1.7519407554199606e-08, "loss": 0.0113, "step": 22067 }, { "epoch": 8.7, "learning_rate": 1.750895851978047e-08, "loss": 0.0116, "step": 22068 }, { "epoch": 8.7, "learning_rate": 1.7498512459687897e-08, "loss": 0.0104, "step": 22069 }, { "epoch": 8.7, "learning_rate": 1.7488069374092217e-08, "loss": 0.0103, "step": 22070 }, { "epoch": 8.7, "learning_rate": 1.7477629263163628e-08, "loss": 0.0111, "step": 22071 }, { "epoch": 8.7, "learning_rate": 1.7467192127072305e-08, "loss": 0.0076, "step": 22072 }, { "epoch": 8.7, "learning_rate": 1.7456757965988313e-08, "loss": 0.012, "step": 22073 }, { "epoch": 8.7, "learning_rate": 1.744632678008171e-08, "loss": 0.0084, "step": 22074 }, { "epoch": 8.7, "learning_rate": 1.743589856952259e-08, "loss": 0.0102, "step": 22075 }, { "epoch": 8.7, "learning_rate": 1.7425473334480788e-08, "loss": 0.0114, "step": 22076 }, { "epoch": 8.7, "learning_rate": 1.741505107512633e-08, "loss": 0.0101, "step": 22077 }, { "epoch": 8.7, "learning_rate": 1.7404631791629056e-08, "loss": 0.0116, "step": 22078 }, { "epoch": 8.7, "learning_rate": 1.7394215484158738e-08, "loss": 0.009, "step": 22079 }, { "epoch": 8.7, "learning_rate": 1.7383802152885174e-08, "loss": 0.0109, "step": 22080 }, { "epoch": 8.7, "learning_rate": 1.737339179797812e-08, "loss": 0.0101, "step": 22081 }, { "epoch": 8.7, "learning_rate": 1.7362984419607198e-08, "loss": 0.0102, "step": 22082 }, { "epoch": 8.7, "learning_rate": 1.7352580017942042e-08, "loss": 0.0111, "step": 22083 }, { "epoch": 8.7, "learning_rate": 1.7342178593152256e-08, "loss": 0.0091, "step": 22084 }, { "epoch": 8.7, "learning_rate": 1.7331780145407372e-08, "loss": 0.0129, "step": 22085 }, { "epoch": 8.7, "learning_rate": 1.7321384674876826e-08, "loss": 0.0095, "step": 22086 }, { "epoch": 8.7, "learning_rate": 1.7310992181730088e-08, "loss": 0.0121, "step": 22087 }, { "epoch": 8.71, "learning_rate": 1.7300602666136464e-08, "loss": 0.0096, "step": 22088 }, { "epoch": 8.71, "learning_rate": 1.72902161282654e-08, "loss": 0.013, "step": 22089 }, { "epoch": 8.71, "learning_rate": 1.7279832568286113e-08, "loss": 0.0124, "step": 22090 }, { "epoch": 8.71, "learning_rate": 1.7269451986367887e-08, "loss": 0.0128, "step": 22091 }, { "epoch": 8.71, "learning_rate": 1.7259074382679794e-08, "loss": 0.0106, "step": 22092 }, { "epoch": 8.71, "learning_rate": 1.7248699757391117e-08, "loss": 0.0112, "step": 22093 }, { "epoch": 8.71, "learning_rate": 1.7238328110670874e-08, "loss": 0.0108, "step": 22094 }, { "epoch": 8.71, "learning_rate": 1.7227959442688132e-08, "loss": 0.0112, "step": 22095 }, { "epoch": 8.71, "learning_rate": 1.7217593753611826e-08, "loss": 0.0095, "step": 22096 }, { "epoch": 8.71, "learning_rate": 1.720723104361097e-08, "loss": 0.0118, "step": 22097 }, { "epoch": 8.71, "learning_rate": 1.7196871312854412e-08, "loss": 0.0115, "step": 22098 }, { "epoch": 8.71, "learning_rate": 1.718651456151108e-08, "loss": 0.0104, "step": 22099 }, { "epoch": 8.71, "learning_rate": 1.717616078974966e-08, "loss": 0.009, "step": 22100 }, { "epoch": 8.71, "learning_rate": 1.7165809997738977e-08, "loss": 0.0112, "step": 22101 }, { "epoch": 8.71, "learning_rate": 1.715546218564774e-08, "loss": 0.0099, "step": 22102 }, { "epoch": 8.71, "learning_rate": 1.7145117353644544e-08, "loss": 0.0125, "step": 22103 }, { "epoch": 8.71, "learning_rate": 1.713477550189806e-08, "loss": 0.0088, "step": 22104 }, { "epoch": 8.71, "learning_rate": 1.712443663057679e-08, "loss": 0.0084, "step": 22105 }, { "epoch": 8.71, "learning_rate": 1.7114100739849245e-08, "loss": 0.0078, "step": 22106 }, { "epoch": 8.71, "learning_rate": 1.710376782988392e-08, "loss": 0.011, "step": 22107 }, { "epoch": 8.71, "learning_rate": 1.709343790084923e-08, "loss": 0.0152, "step": 22108 }, { "epoch": 8.71, "learning_rate": 1.7083110952913505e-08, "loss": 0.0103, "step": 22109 }, { "epoch": 8.71, "learning_rate": 1.7072786986245037e-08, "loss": 0.0105, "step": 22110 }, { "epoch": 8.71, "learning_rate": 1.7062466001012155e-08, "loss": 0.0094, "step": 22111 }, { "epoch": 8.71, "learning_rate": 1.705214799738306e-08, "loss": 0.0093, "step": 22112 }, { "epoch": 8.71, "learning_rate": 1.704183297552586e-08, "loss": 0.0113, "step": 22113 }, { "epoch": 8.72, "learning_rate": 1.7031520935608733e-08, "loss": 0.0078, "step": 22114 }, { "epoch": 8.72, "learning_rate": 1.7021211877799747e-08, "loss": 0.0102, "step": 22115 }, { "epoch": 8.72, "learning_rate": 1.7010905802266917e-08, "loss": 0.0133, "step": 22116 }, { "epoch": 8.72, "learning_rate": 1.7000602709178203e-08, "loss": 0.0099, "step": 22117 }, { "epoch": 8.72, "learning_rate": 1.6990302598701557e-08, "loss": 0.0079, "step": 22118 }, { "epoch": 8.72, "learning_rate": 1.69800054710048e-08, "loss": 0.0099, "step": 22119 }, { "epoch": 8.72, "learning_rate": 1.69697113262558e-08, "loss": 0.01, "step": 22120 }, { "epoch": 8.72, "learning_rate": 1.695942016462235e-08, "loss": 0.0085, "step": 22121 }, { "epoch": 8.72, "learning_rate": 1.694913198627217e-08, "loss": 0.0092, "step": 22122 }, { "epoch": 8.72, "learning_rate": 1.6938846791372918e-08, "loss": 0.0119, "step": 22123 }, { "epoch": 8.72, "learning_rate": 1.692856458009222e-08, "loss": 0.0092, "step": 22124 }, { "epoch": 8.72, "learning_rate": 1.6918285352597738e-08, "loss": 0.011, "step": 22125 }, { "epoch": 8.72, "learning_rate": 1.69080091090569e-08, "loss": 0.0093, "step": 22126 }, { "epoch": 8.72, "learning_rate": 1.6897735849637297e-08, "loss": 0.0087, "step": 22127 }, { "epoch": 8.72, "learning_rate": 1.6887465574506244e-08, "loss": 0.0093, "step": 22128 }, { "epoch": 8.72, "learning_rate": 1.687719828383125e-08, "loss": 0.0108, "step": 22129 }, { "epoch": 8.72, "learning_rate": 1.6866933977779606e-08, "loss": 0.0119, "step": 22130 }, { "epoch": 8.72, "learning_rate": 1.6856672656518622e-08, "loss": 0.0095, "step": 22131 }, { "epoch": 8.72, "learning_rate": 1.6846414320215472e-08, "loss": 0.0107, "step": 22132 }, { "epoch": 8.72, "learning_rate": 1.6836158969037472e-08, "loss": 0.0123, "step": 22133 }, { "epoch": 8.72, "learning_rate": 1.6825906603151662e-08, "loss": 0.0112, "step": 22134 }, { "epoch": 8.72, "learning_rate": 1.6815657222725222e-08, "loss": 0.0092, "step": 22135 }, { "epoch": 8.72, "learning_rate": 1.6805410827925127e-08, "loss": 0.0102, "step": 22136 }, { "epoch": 8.72, "learning_rate": 1.679516741891842e-08, "loss": 0.0096, "step": 22137 }, { "epoch": 8.72, "learning_rate": 1.678492699587204e-08, "loss": 0.0098, "step": 22138 }, { "epoch": 8.73, "learning_rate": 1.6774689558952916e-08, "loss": 0.011, "step": 22139 }, { "epoch": 8.73, "learning_rate": 1.6764455108327867e-08, "loss": 0.013, "step": 22140 }, { "epoch": 8.73, "learning_rate": 1.67542236441637e-08, "loss": 0.0116, "step": 22141 }, { "epoch": 8.73, "learning_rate": 1.6743995166627235e-08, "loss": 0.0115, "step": 22142 }, { "epoch": 8.73, "learning_rate": 1.673376967588507e-08, "loss": 0.01, "step": 22143 }, { "epoch": 8.73, "learning_rate": 1.672354717210396e-08, "loss": 0.0126, "step": 22144 }, { "epoch": 8.73, "learning_rate": 1.6713327655450483e-08, "loss": 0.0105, "step": 22145 }, { "epoch": 8.73, "learning_rate": 1.6703111126091197e-08, "loss": 0.0106, "step": 22146 }, { "epoch": 8.73, "learning_rate": 1.669289758419261e-08, "loss": 0.0098, "step": 22147 }, { "epoch": 8.73, "learning_rate": 1.6682687029921238e-08, "loss": 0.0101, "step": 22148 }, { "epoch": 8.73, "learning_rate": 1.6672479463443412e-08, "loss": 0.0117, "step": 22149 }, { "epoch": 8.73, "learning_rate": 1.6662274884925554e-08, "loss": 0.0087, "step": 22150 }, { "epoch": 8.73, "learning_rate": 1.6652073294533952e-08, "loss": 0.0099, "step": 22151 }, { "epoch": 8.73, "learning_rate": 1.6641874692434966e-08, "loss": 0.0121, "step": 22152 }, { "epoch": 8.73, "learning_rate": 1.663167907879468e-08, "loss": 0.0141, "step": 22153 }, { "epoch": 8.73, "learning_rate": 1.6621486453779366e-08, "loss": 0.0102, "step": 22154 }, { "epoch": 8.73, "learning_rate": 1.661129681755511e-08, "loss": 0.0092, "step": 22155 }, { "epoch": 8.73, "learning_rate": 1.6601110170288023e-08, "loss": 0.0099, "step": 22156 }, { "epoch": 8.73, "learning_rate": 1.6590926512144088e-08, "loss": 0.0109, "step": 22157 }, { "epoch": 8.73, "learning_rate": 1.6580745843289323e-08, "loss": 0.0085, "step": 22158 }, { "epoch": 8.73, "learning_rate": 1.6570568163889577e-08, "loss": 0.0096, "step": 22159 }, { "epoch": 8.73, "learning_rate": 1.6560393474110867e-08, "loss": 0.0099, "step": 22160 }, { "epoch": 8.73, "learning_rate": 1.655022177411889e-08, "loss": 0.0092, "step": 22161 }, { "epoch": 8.73, "learning_rate": 1.654005306407953e-08, "loss": 0.014, "step": 22162 }, { "epoch": 8.73, "learning_rate": 1.6529887344158457e-08, "loss": 0.0081, "step": 22163 }, { "epoch": 8.74, "learning_rate": 1.6519724614521357e-08, "loss": 0.0091, "step": 22164 }, { "epoch": 8.74, "learning_rate": 1.6509564875333926e-08, "loss": 0.0114, "step": 22165 }, { "epoch": 8.74, "learning_rate": 1.6499408126761692e-08, "loss": 0.0102, "step": 22166 }, { "epoch": 8.74, "learning_rate": 1.6489254368970197e-08, "loss": 0.0097, "step": 22167 }, { "epoch": 8.74, "learning_rate": 1.6479103602124966e-08, "loss": 0.0092, "step": 22168 }, { "epoch": 8.74, "learning_rate": 1.646895582639145e-08, "loss": 0.0096, "step": 22169 }, { "epoch": 8.74, "learning_rate": 1.6458811041935004e-08, "loss": 0.0109, "step": 22170 }, { "epoch": 8.74, "learning_rate": 1.644866924892101e-08, "loss": 0.0126, "step": 22171 }, { "epoch": 8.74, "learning_rate": 1.6438530447514666e-08, "loss": 0.0109, "step": 22172 }, { "epoch": 8.74, "learning_rate": 1.642839463788137e-08, "loss": 0.0104, "step": 22173 }, { "epoch": 8.74, "learning_rate": 1.641826182018622e-08, "loss": 0.0116, "step": 22174 }, { "epoch": 8.74, "learning_rate": 1.640813199459441e-08, "loss": 0.0113, "step": 22175 }, { "epoch": 8.74, "learning_rate": 1.6398005161270987e-08, "loss": 0.0104, "step": 22176 }, { "epoch": 8.74, "learning_rate": 1.6387881320381046e-08, "loss": 0.0118, "step": 22177 }, { "epoch": 8.74, "learning_rate": 1.6377760472089587e-08, "loss": 0.0107, "step": 22178 }, { "epoch": 8.74, "learning_rate": 1.6367642616561582e-08, "loss": 0.0116, "step": 22179 }, { "epoch": 8.74, "learning_rate": 1.6357527753961885e-08, "loss": 0.0105, "step": 22180 }, { "epoch": 8.74, "learning_rate": 1.6347415884455363e-08, "loss": 0.0093, "step": 22181 }, { "epoch": 8.74, "learning_rate": 1.6337307008206857e-08, "loss": 0.0097, "step": 22182 }, { "epoch": 8.74, "learning_rate": 1.632720112538115e-08, "loss": 0.0096, "step": 22183 }, { "epoch": 8.74, "learning_rate": 1.6317098236142867e-08, "loss": 0.0123, "step": 22184 }, { "epoch": 8.74, "learning_rate": 1.6306998340656764e-08, "loss": 0.0103, "step": 22185 }, { "epoch": 8.74, "learning_rate": 1.6296901439087373e-08, "loss": 0.0119, "step": 22186 }, { "epoch": 8.74, "learning_rate": 1.6286807531599277e-08, "loss": 0.0101, "step": 22187 }, { "epoch": 8.74, "learning_rate": 1.6276716618357057e-08, "loss": 0.0117, "step": 22188 }, { "epoch": 8.74, "learning_rate": 1.6266628699525088e-08, "loss": 0.0101, "step": 22189 }, { "epoch": 8.75, "learning_rate": 1.625654377526784e-08, "loss": 0.0104, "step": 22190 }, { "epoch": 8.75, "learning_rate": 1.624646184574965e-08, "loss": 0.0089, "step": 22191 }, { "epoch": 8.75, "learning_rate": 1.6236382911134895e-08, "loss": 0.0087, "step": 22192 }, { "epoch": 8.75, "learning_rate": 1.62263069715878e-08, "loss": 0.01, "step": 22193 }, { "epoch": 8.75, "learning_rate": 1.6216234027272568e-08, "loss": 0.0082, "step": 22194 }, { "epoch": 8.75, "learning_rate": 1.6206164078353425e-08, "loss": 0.0072, "step": 22195 }, { "epoch": 8.75, "learning_rate": 1.6196097124994478e-08, "loss": 0.0133, "step": 22196 }, { "epoch": 8.75, "learning_rate": 1.61860331673598e-08, "loss": 0.0088, "step": 22197 }, { "epoch": 8.75, "learning_rate": 1.617597220561342e-08, "loss": 0.0101, "step": 22198 }, { "epoch": 8.75, "learning_rate": 1.6165914239919266e-08, "loss": 0.0102, "step": 22199 }, { "epoch": 8.75, "learning_rate": 1.615585927044136e-08, "loss": 0.0103, "step": 22200 }, { "epoch": 8.75, "learning_rate": 1.614580729734354e-08, "loss": 0.0108, "step": 22201 }, { "epoch": 8.75, "learning_rate": 1.613575832078964e-08, "loss": 0.009, "step": 22202 }, { "epoch": 8.75, "learning_rate": 1.61257123409434e-08, "loss": 0.0125, "step": 22203 }, { "epoch": 8.75, "learning_rate": 1.611566935796862e-08, "loss": 0.0132, "step": 22204 }, { "epoch": 8.75, "learning_rate": 1.6105629372028927e-08, "loss": 0.0114, "step": 22205 }, { "epoch": 8.75, "learning_rate": 1.6095592383288036e-08, "loss": 0.0092, "step": 22206 }, { "epoch": 8.75, "learning_rate": 1.608555839190946e-08, "loss": 0.0097, "step": 22207 }, { "epoch": 8.75, "learning_rate": 1.6075527398056754e-08, "loss": 0.0096, "step": 22208 }, { "epoch": 8.75, "learning_rate": 1.6065499401893455e-08, "loss": 0.0114, "step": 22209 }, { "epoch": 8.75, "learning_rate": 1.6055474403582926e-08, "loss": 0.0094, "step": 22210 }, { "epoch": 8.75, "learning_rate": 1.604545240328865e-08, "loss": 0.0114, "step": 22211 }, { "epoch": 8.75, "learning_rate": 1.6035433401173836e-08, "loss": 0.0115, "step": 22212 }, { "epoch": 8.75, "learning_rate": 1.6025417397401932e-08, "loss": 0.0083, "step": 22213 }, { "epoch": 8.75, "learning_rate": 1.6015404392136088e-08, "loss": 0.0117, "step": 22214 }, { "epoch": 8.76, "learning_rate": 1.6005394385539583e-08, "loss": 0.0096, "step": 22215 }, { "epoch": 8.76, "learning_rate": 1.5995387377775437e-08, "loss": 0.0099, "step": 22216 }, { "epoch": 8.76, "learning_rate": 1.5985383369006856e-08, "loss": 0.0123, "step": 22217 }, { "epoch": 8.76, "learning_rate": 1.597538235939684e-08, "loss": 0.0101, "step": 22218 }, { "epoch": 8.76, "learning_rate": 1.5965384349108435e-08, "loss": 0.0092, "step": 22219 }, { "epoch": 8.76, "learning_rate": 1.5955389338304537e-08, "loss": 0.0112, "step": 22220 }, { "epoch": 8.76, "learning_rate": 1.5945397327148082e-08, "loss": 0.0114, "step": 22221 }, { "epoch": 8.76, "learning_rate": 1.5935408315801912e-08, "loss": 0.0073, "step": 22222 }, { "epoch": 8.76, "learning_rate": 1.5925422304428883e-08, "loss": 0.0095, "step": 22223 }, { "epoch": 8.76, "learning_rate": 1.5915439293191678e-08, "loss": 0.0112, "step": 22224 }, { "epoch": 8.76, "learning_rate": 1.5905459282253063e-08, "loss": 0.0112, "step": 22225 }, { "epoch": 8.76, "learning_rate": 1.5895482271775596e-08, "loss": 0.0095, "step": 22226 }, { "epoch": 8.76, "learning_rate": 1.5885508261922032e-08, "loss": 0.0102, "step": 22227 }, { "epoch": 8.76, "learning_rate": 1.5875537252854843e-08, "loss": 0.0095, "step": 22228 }, { "epoch": 8.76, "learning_rate": 1.5865569244736564e-08, "loss": 0.01, "step": 22229 }, { "epoch": 8.76, "learning_rate": 1.5855604237729646e-08, "loss": 0.0108, "step": 22230 }, { "epoch": 8.76, "learning_rate": 1.584564223199649e-08, "loss": 0.0119, "step": 22231 }, { "epoch": 8.76, "learning_rate": 1.5835683227699548e-08, "loss": 0.0112, "step": 22232 }, { "epoch": 8.76, "learning_rate": 1.582572722500102e-08, "loss": 0.011, "step": 22233 }, { "epoch": 8.76, "learning_rate": 1.5815774224063215e-08, "loss": 0.0123, "step": 22234 }, { "epoch": 8.76, "learning_rate": 1.580582422504837e-08, "loss": 0.0144, "step": 22235 }, { "epoch": 8.76, "learning_rate": 1.5795877228118703e-08, "loss": 0.0129, "step": 22236 }, { "epoch": 8.76, "learning_rate": 1.578593323343622e-08, "loss": 0.0131, "step": 22237 }, { "epoch": 8.76, "learning_rate": 1.5775992241163084e-08, "loss": 0.0081, "step": 22238 }, { "epoch": 8.76, "learning_rate": 1.5766054251461246e-08, "loss": 0.0096, "step": 22239 }, { "epoch": 8.76, "learning_rate": 1.575611926449276e-08, "loss": 0.0135, "step": 22240 }, { "epoch": 8.77, "learning_rate": 1.57461872804195e-08, "loss": 0.0094, "step": 22241 }, { "epoch": 8.77, "learning_rate": 1.5736258299403372e-08, "loss": 0.0094, "step": 22242 }, { "epoch": 8.77, "learning_rate": 1.5726332321606116e-08, "loss": 0.0104, "step": 22243 }, { "epoch": 8.77, "learning_rate": 1.571640934718965e-08, "loss": 0.0086, "step": 22244 }, { "epoch": 8.77, "learning_rate": 1.57064893763156e-08, "loss": 0.0105, "step": 22245 }, { "epoch": 8.77, "learning_rate": 1.5696572409145703e-08, "loss": 0.0103, "step": 22246 }, { "epoch": 8.77, "learning_rate": 1.568665844584156e-08, "loss": 0.0117, "step": 22247 }, { "epoch": 8.77, "learning_rate": 1.5676747486564733e-08, "loss": 0.0096, "step": 22248 }, { "epoch": 8.77, "learning_rate": 1.5666839531476804e-08, "loss": 0.0093, "step": 22249 }, { "epoch": 8.77, "learning_rate": 1.5656934580739247e-08, "loss": 0.0116, "step": 22250 }, { "epoch": 8.77, "learning_rate": 1.564703263451348e-08, "loss": 0.011, "step": 22251 }, { "epoch": 8.77, "learning_rate": 1.563713369296087e-08, "loss": 0.0091, "step": 22252 }, { "epoch": 8.77, "learning_rate": 1.562723775624284e-08, "loss": 0.0116, "step": 22253 }, { "epoch": 8.77, "learning_rate": 1.5617344824520572e-08, "loss": 0.0088, "step": 22254 }, { "epoch": 8.77, "learning_rate": 1.5607454897955385e-08, "loss": 0.0089, "step": 22255 }, { "epoch": 8.77, "learning_rate": 1.5597567976708413e-08, "loss": 0.0095, "step": 22256 }, { "epoch": 8.77, "learning_rate": 1.558768406094082e-08, "loss": 0.0092, "step": 22257 }, { "epoch": 8.77, "learning_rate": 1.557780315081372e-08, "loss": 0.0094, "step": 22258 }, { "epoch": 8.77, "learning_rate": 1.5567925246488156e-08, "loss": 0.0101, "step": 22259 }, { "epoch": 8.77, "learning_rate": 1.5558050348125074e-08, "loss": 0.0082, "step": 22260 }, { "epoch": 8.77, "learning_rate": 1.5548178455885452e-08, "loss": 0.01, "step": 22261 }, { "epoch": 8.77, "learning_rate": 1.5538309569930187e-08, "loss": 0.0102, "step": 22262 }, { "epoch": 8.77, "learning_rate": 1.5528443690420146e-08, "loss": 0.0089, "step": 22263 }, { "epoch": 8.77, "learning_rate": 1.5518580817516092e-08, "loss": 0.0089, "step": 22264 }, { "epoch": 8.77, "learning_rate": 1.5508720951378827e-08, "loss": 0.0123, "step": 22265 }, { "epoch": 8.78, "learning_rate": 1.549886409216894e-08, "loss": 0.0108, "step": 22266 }, { "epoch": 8.78, "learning_rate": 1.548901024004723e-08, "loss": 0.01, "step": 22267 }, { "epoch": 8.78, "learning_rate": 1.547915939517419e-08, "loss": 0.0091, "step": 22268 }, { "epoch": 8.78, "learning_rate": 1.5469311557710452e-08, "loss": 0.0105, "step": 22269 }, { "epoch": 8.78, "learning_rate": 1.545946672781644e-08, "loss": 0.0076, "step": 22270 }, { "epoch": 8.78, "learning_rate": 1.5449624905652668e-08, "loss": 0.011, "step": 22271 }, { "epoch": 8.78, "learning_rate": 1.5439786091379502e-08, "loss": 0.0096, "step": 22272 }, { "epoch": 8.78, "learning_rate": 1.5429950285157368e-08, "loss": 0.0122, "step": 22273 }, { "epoch": 8.78, "learning_rate": 1.542011748714651e-08, "loss": 0.0128, "step": 22274 }, { "epoch": 8.78, "learning_rate": 1.5410287697507186e-08, "loss": 0.0102, "step": 22275 }, { "epoch": 8.78, "learning_rate": 1.540046091639968e-08, "loss": 0.0115, "step": 22276 }, { "epoch": 8.78, "learning_rate": 1.5390637143984052e-08, "loss": 0.0103, "step": 22277 }, { "epoch": 8.78, "learning_rate": 1.5380816380420504e-08, "loss": 0.0089, "step": 22278 }, { "epoch": 8.78, "learning_rate": 1.5370998625869014e-08, "loss": 0.0132, "step": 22279 }, { "epoch": 8.78, "learning_rate": 1.536118388048968e-08, "loss": 0.0134, "step": 22280 }, { "epoch": 8.78, "learning_rate": 1.5351372144442397e-08, "loss": 0.0109, "step": 22281 }, { "epoch": 8.78, "learning_rate": 1.534156341788717e-08, "loss": 0.0088, "step": 22282 }, { "epoch": 8.78, "learning_rate": 1.5331757700983738e-08, "loss": 0.0103, "step": 22283 }, { "epoch": 8.78, "learning_rate": 1.532195499389206e-08, "loss": 0.0083, "step": 22284 }, { "epoch": 8.78, "learning_rate": 1.531215529677179e-08, "loss": 0.0091, "step": 22285 }, { "epoch": 8.78, "learning_rate": 1.530235860978273e-08, "loss": 0.011, "step": 22286 }, { "epoch": 8.78, "learning_rate": 1.5292564933084507e-08, "loss": 0.0095, "step": 22287 }, { "epoch": 8.78, "learning_rate": 1.528277426683673e-08, "loss": 0.0114, "step": 22288 }, { "epoch": 8.78, "learning_rate": 1.5272986611199003e-08, "loss": 0.0107, "step": 22289 }, { "epoch": 8.78, "learning_rate": 1.5263201966330864e-08, "loss": 0.0132, "step": 22290 }, { "epoch": 8.79, "learning_rate": 1.5253420332391742e-08, "loss": 0.0103, "step": 22291 }, { "epoch": 8.79, "learning_rate": 1.524364170954109e-08, "loss": 0.0095, "step": 22292 }, { "epoch": 8.79, "learning_rate": 1.523386609793831e-08, "loss": 0.0107, "step": 22293 }, { "epoch": 8.79, "learning_rate": 1.522409349774265e-08, "loss": 0.009, "step": 22294 }, { "epoch": 8.79, "learning_rate": 1.5214323909113458e-08, "loss": 0.0084, "step": 22295 }, { "epoch": 8.79, "learning_rate": 1.520455733220998e-08, "loss": 0.0087, "step": 22296 }, { "epoch": 8.79, "learning_rate": 1.519479376719133e-08, "loss": 0.0102, "step": 22297 }, { "epoch": 8.79, "learning_rate": 1.5185033214216648e-08, "loss": 0.0097, "step": 22298 }, { "epoch": 8.79, "learning_rate": 1.51752756734451e-08, "loss": 0.0119, "step": 22299 }, { "epoch": 8.79, "learning_rate": 1.5165521145035596e-08, "loss": 0.0123, "step": 22300 }, { "epoch": 8.79, "learning_rate": 1.5155769629147197e-08, "loss": 0.0109, "step": 22301 }, { "epoch": 8.79, "learning_rate": 1.5146021125938813e-08, "loss": 0.01, "step": 22302 }, { "epoch": 8.79, "learning_rate": 1.5136275635569384e-08, "loss": 0.0115, "step": 22303 }, { "epoch": 8.79, "learning_rate": 1.5126533158197674e-08, "loss": 0.0109, "step": 22304 }, { "epoch": 8.79, "learning_rate": 1.511679369398251e-08, "loss": 0.0103, "step": 22305 }, { "epoch": 8.79, "learning_rate": 1.5107057243082567e-08, "loss": 0.0093, "step": 22306 }, { "epoch": 8.79, "learning_rate": 1.5097323805656648e-08, "loss": 0.0086, "step": 22307 }, { "epoch": 8.79, "learning_rate": 1.5087593381863296e-08, "loss": 0.008, "step": 22308 }, { "epoch": 8.79, "learning_rate": 1.507786597186118e-08, "loss": 0.0095, "step": 22309 }, { "epoch": 8.79, "learning_rate": 1.5068141575808714e-08, "loss": 0.0105, "step": 22310 }, { "epoch": 8.79, "learning_rate": 1.5058420193864563e-08, "loss": 0.0098, "step": 22311 }, { "epoch": 8.79, "learning_rate": 1.5048701826187048e-08, "loss": 0.0129, "step": 22312 }, { "epoch": 8.79, "learning_rate": 1.503898647293462e-08, "loss": 0.0111, "step": 22313 }, { "epoch": 8.79, "learning_rate": 1.5029274134265578e-08, "loss": 0.0105, "step": 22314 }, { "epoch": 8.79, "learning_rate": 1.5019564810338237e-08, "loss": 0.0103, "step": 22315 }, { "epoch": 8.79, "learning_rate": 1.5009858501310912e-08, "loss": 0.0101, "step": 22316 }, { "epoch": 8.8, "learning_rate": 1.500015520734168e-08, "loss": 0.0112, "step": 22317 }, { "epoch": 8.8, "learning_rate": 1.499045492858877e-08, "loss": 0.0102, "step": 22318 }, { "epoch": 8.8, "learning_rate": 1.498075766521023e-08, "loss": 0.0094, "step": 22319 }, { "epoch": 8.8, "learning_rate": 1.4971063417364204e-08, "loss": 0.0097, "step": 22320 }, { "epoch": 8.8, "learning_rate": 1.4961372185208608e-08, "loss": 0.0131, "step": 22321 }, { "epoch": 8.8, "learning_rate": 1.495168396890143e-08, "loss": 0.01, "step": 22322 }, { "epoch": 8.8, "learning_rate": 1.4941998768600494e-08, "loss": 0.013, "step": 22323 }, { "epoch": 8.8, "learning_rate": 1.493231658446381e-08, "loss": 0.0118, "step": 22324 }, { "epoch": 8.8, "learning_rate": 1.4922637416649052e-08, "loss": 0.0099, "step": 22325 }, { "epoch": 8.8, "learning_rate": 1.491296126531405e-08, "loss": 0.0104, "step": 22326 }, { "epoch": 8.8, "learning_rate": 1.4903288130616455e-08, "loss": 0.0081, "step": 22327 }, { "epoch": 8.8, "learning_rate": 1.4893618012713938e-08, "loss": 0.0078, "step": 22328 }, { "epoch": 8.8, "learning_rate": 1.488395091176411e-08, "loss": 0.01, "step": 22329 }, { "epoch": 8.8, "learning_rate": 1.4874286827924599e-08, "loss": 0.0078, "step": 22330 }, { "epoch": 8.8, "learning_rate": 1.486462576135279e-08, "loss": 0.0109, "step": 22331 }, { "epoch": 8.8, "learning_rate": 1.4854967712206224e-08, "loss": 0.01, "step": 22332 }, { "epoch": 8.8, "learning_rate": 1.4845312680642285e-08, "loss": 0.0116, "step": 22333 }, { "epoch": 8.8, "learning_rate": 1.4835660666818406e-08, "loss": 0.0141, "step": 22334 }, { "epoch": 8.8, "learning_rate": 1.482601167089179e-08, "loss": 0.014, "step": 22335 }, { "epoch": 8.8, "learning_rate": 1.4816365693019783e-08, "loss": 0.0119, "step": 22336 }, { "epoch": 8.8, "learning_rate": 1.4806722733359545e-08, "loss": 0.0124, "step": 22337 }, { "epoch": 8.8, "learning_rate": 1.4797082792068238e-08, "loss": 0.0105, "step": 22338 }, { "epoch": 8.8, "learning_rate": 1.4787445869303073e-08, "loss": 0.0099, "step": 22339 }, { "epoch": 8.8, "learning_rate": 1.477781196522101e-08, "loss": 0.0113, "step": 22340 }, { "epoch": 8.8, "learning_rate": 1.4768181079979081e-08, "loss": 0.0084, "step": 22341 }, { "epoch": 8.81, "learning_rate": 1.4758553213734315e-08, "loss": 0.0082, "step": 22342 }, { "epoch": 8.81, "learning_rate": 1.474892836664361e-08, "loss": 0.01, "step": 22343 }, { "epoch": 8.81, "learning_rate": 1.4739306538863795e-08, "loss": 0.0095, "step": 22344 }, { "epoch": 8.81, "learning_rate": 1.4729687730551744e-08, "loss": 0.0109, "step": 22345 }, { "epoch": 8.81, "learning_rate": 1.4720071941864154e-08, "loss": 0.0093, "step": 22346 }, { "epoch": 8.81, "learning_rate": 1.4710459172957856e-08, "loss": 0.0102, "step": 22347 }, { "epoch": 8.81, "learning_rate": 1.4700849423989436e-08, "loss": 0.0106, "step": 22348 }, { "epoch": 8.81, "learning_rate": 1.4691242695115568e-08, "loss": 0.0094, "step": 22349 }, { "epoch": 8.81, "learning_rate": 1.468163898649275e-08, "loss": 0.0148, "step": 22350 }, { "epoch": 8.81, "learning_rate": 1.4672038298277634e-08, "loss": 0.0103, "step": 22351 }, { "epoch": 8.81, "learning_rate": 1.4662440630626604e-08, "loss": 0.0099, "step": 22352 }, { "epoch": 8.81, "learning_rate": 1.4652845983696116e-08, "loss": 0.011, "step": 22353 }, { "epoch": 8.81, "learning_rate": 1.4643254357642532e-08, "loss": 0.0125, "step": 22354 }, { "epoch": 8.81, "learning_rate": 1.4633665752622193e-08, "loss": 0.0112, "step": 22355 }, { "epoch": 8.81, "learning_rate": 1.4624080168791375e-08, "loss": 0.0085, "step": 22356 }, { "epoch": 8.81, "learning_rate": 1.4614497606306332e-08, "loss": 0.0089, "step": 22357 }, { "epoch": 8.81, "learning_rate": 1.4604918065323202e-08, "loss": 0.0117, "step": 22358 }, { "epoch": 8.81, "learning_rate": 1.4595341545998151e-08, "loss": 0.0091, "step": 22359 }, { "epoch": 8.81, "learning_rate": 1.4585768048487257e-08, "loss": 0.0103, "step": 22360 }, { "epoch": 8.81, "learning_rate": 1.4576197572946546e-08, "loss": 0.0098, "step": 22361 }, { "epoch": 8.81, "learning_rate": 1.456663011953201e-08, "loss": 0.0081, "step": 22362 }, { "epoch": 8.81, "learning_rate": 1.4557065688399517e-08, "loss": 0.0091, "step": 22363 }, { "epoch": 8.81, "learning_rate": 1.454750427970508e-08, "loss": 0.0088, "step": 22364 }, { "epoch": 8.81, "learning_rate": 1.4537945893604442e-08, "loss": 0.0099, "step": 22365 }, { "epoch": 8.81, "learning_rate": 1.4528390530253431e-08, "loss": 0.011, "step": 22366 }, { "epoch": 8.82, "learning_rate": 1.4518838189807747e-08, "loss": 0.0097, "step": 22367 }, { "epoch": 8.82, "learning_rate": 1.4509288872423109e-08, "loss": 0.0096, "step": 22368 }, { "epoch": 8.82, "learning_rate": 1.4499742578255147e-08, "loss": 0.0126, "step": 22369 }, { "epoch": 8.82, "learning_rate": 1.4490199307459472e-08, "loss": 0.0096, "step": 22370 }, { "epoch": 8.82, "learning_rate": 1.448065906019158e-08, "loss": 0.0127, "step": 22371 }, { "epoch": 8.82, "learning_rate": 1.4471121836606992e-08, "loss": 0.0082, "step": 22372 }, { "epoch": 8.82, "learning_rate": 1.4461587636861162e-08, "loss": 0.0091, "step": 22373 }, { "epoch": 8.82, "learning_rate": 1.4452056461109474e-08, "loss": 0.0101, "step": 22374 }, { "epoch": 8.82, "learning_rate": 1.4442528309507251e-08, "loss": 0.0097, "step": 22375 }, { "epoch": 8.82, "learning_rate": 1.4433003182209814e-08, "loss": 0.0094, "step": 22376 }, { "epoch": 8.82, "learning_rate": 1.4423481079372346e-08, "loss": 0.0118, "step": 22377 }, { "epoch": 8.82, "learning_rate": 1.4413962001150149e-08, "loss": 0.0085, "step": 22378 }, { "epoch": 8.82, "learning_rate": 1.4404445947698296e-08, "loss": 0.0076, "step": 22379 }, { "epoch": 8.82, "learning_rate": 1.4394932919171932e-08, "loss": 0.0079, "step": 22380 }, { "epoch": 8.82, "learning_rate": 1.438542291572602e-08, "loss": 0.01, "step": 22381 }, { "epoch": 8.82, "learning_rate": 1.4375915937515637e-08, "loss": 0.0108, "step": 22382 }, { "epoch": 8.82, "learning_rate": 1.4366411984695726e-08, "loss": 0.0119, "step": 22383 }, { "epoch": 8.82, "learning_rate": 1.435691105742114e-08, "loss": 0.0096, "step": 22384 }, { "epoch": 8.82, "learning_rate": 1.4347413155846755e-08, "loss": 0.0099, "step": 22385 }, { "epoch": 8.82, "learning_rate": 1.433791828012738e-08, "loss": 0.0101, "step": 22386 }, { "epoch": 8.82, "learning_rate": 1.4328426430417806e-08, "loss": 0.0098, "step": 22387 }, { "epoch": 8.82, "learning_rate": 1.4318937606872639e-08, "loss": 0.0093, "step": 22388 }, { "epoch": 8.82, "learning_rate": 1.4309451809646644e-08, "loss": 0.0087, "step": 22389 }, { "epoch": 8.82, "learning_rate": 1.4299969038894278e-08, "loss": 0.0101, "step": 22390 }, { "epoch": 8.82, "learning_rate": 1.4290489294770258e-08, "loss": 0.0113, "step": 22391 }, { "epoch": 8.82, "learning_rate": 1.4281012577428997e-08, "loss": 0.0106, "step": 22392 }, { "epoch": 8.83, "learning_rate": 1.4271538887024993e-08, "loss": 0.0089, "step": 22393 }, { "epoch": 8.83, "learning_rate": 1.4262068223712587e-08, "loss": 0.01, "step": 22394 }, { "epoch": 8.83, "learning_rate": 1.425260058764619e-08, "loss": 0.0119, "step": 22395 }, { "epoch": 8.83, "learning_rate": 1.4243135978980102e-08, "loss": 0.0091, "step": 22396 }, { "epoch": 8.83, "learning_rate": 1.4233674397868623e-08, "loss": 0.0111, "step": 22397 }, { "epoch": 8.83, "learning_rate": 1.422421584446587e-08, "loss": 0.0105, "step": 22398 }, { "epoch": 8.83, "learning_rate": 1.4214760318926056e-08, "loss": 0.0101, "step": 22399 }, { "epoch": 8.83, "learning_rate": 1.4205307821403301e-08, "loss": 0.0121, "step": 22400 }, { "epoch": 8.83, "learning_rate": 1.419585835205166e-08, "loss": 0.0095, "step": 22401 }, { "epoch": 8.83, "learning_rate": 1.4186411911025121e-08, "loss": 0.0103, "step": 22402 }, { "epoch": 8.83, "learning_rate": 1.417696849847767e-08, "loss": 0.0125, "step": 22403 }, { "epoch": 8.83, "learning_rate": 1.4167528114563233e-08, "loss": 0.0092, "step": 22404 }, { "epoch": 8.83, "learning_rate": 1.4158090759435615e-08, "loss": 0.0106, "step": 22405 }, { "epoch": 8.83, "learning_rate": 1.4148656433248718e-08, "loss": 0.0118, "step": 22406 }, { "epoch": 8.83, "learning_rate": 1.413922513615624e-08, "loss": 0.0099, "step": 22407 }, { "epoch": 8.83, "learning_rate": 1.4129796868311883e-08, "loss": 0.0125, "step": 22408 }, { "epoch": 8.83, "learning_rate": 1.4120371629869366e-08, "loss": 0.0127, "step": 22409 }, { "epoch": 8.83, "learning_rate": 1.4110949420982321e-08, "loss": 0.011, "step": 22410 }, { "epoch": 8.83, "learning_rate": 1.410153024180425e-08, "loss": 0.0113, "step": 22411 }, { "epoch": 8.83, "learning_rate": 1.4092114092488694e-08, "loss": 0.0088, "step": 22412 }, { "epoch": 8.83, "learning_rate": 1.4082700973189133e-08, "loss": 0.0083, "step": 22413 }, { "epoch": 8.83, "learning_rate": 1.407329088405902e-08, "loss": 0.0094, "step": 22414 }, { "epoch": 8.83, "learning_rate": 1.4063883825251677e-08, "loss": 0.0111, "step": 22415 }, { "epoch": 8.83, "learning_rate": 1.4054479796920448e-08, "loss": 0.0088, "step": 22416 }, { "epoch": 8.83, "learning_rate": 1.4045078799218524e-08, "loss": 0.012, "step": 22417 }, { "epoch": 8.84, "learning_rate": 1.4035680832299269e-08, "loss": 0.0115, "step": 22418 }, { "epoch": 8.84, "learning_rate": 1.4026285896315737e-08, "loss": 0.0083, "step": 22419 }, { "epoch": 8.84, "learning_rate": 1.4016893991421164e-08, "loss": 0.0119, "step": 22420 }, { "epoch": 8.84, "learning_rate": 1.4007505117768493e-08, "loss": 0.0089, "step": 22421 }, { "epoch": 8.84, "learning_rate": 1.3998119275510822e-08, "loss": 0.0117, "step": 22422 }, { "epoch": 8.84, "learning_rate": 1.3988736464801121e-08, "loss": 0.0109, "step": 22423 }, { "epoch": 8.84, "learning_rate": 1.3979356685792331e-08, "loss": 0.0102, "step": 22424 }, { "epoch": 8.84, "learning_rate": 1.396997993863729e-08, "loss": 0.012, "step": 22425 }, { "epoch": 8.84, "learning_rate": 1.3960606223488846e-08, "loss": 0.0105, "step": 22426 }, { "epoch": 8.84, "learning_rate": 1.3951235540499794e-08, "loss": 0.0127, "step": 22427 }, { "epoch": 8.84, "learning_rate": 1.394186788982281e-08, "loss": 0.0075, "step": 22428 }, { "epoch": 8.84, "learning_rate": 1.393250327161064e-08, "loss": 0.0106, "step": 22429 }, { "epoch": 8.84, "learning_rate": 1.3923141686015804e-08, "loss": 0.0098, "step": 22430 }, { "epoch": 8.84, "learning_rate": 1.3913783133191026e-08, "loss": 0.0097, "step": 22431 }, { "epoch": 8.84, "learning_rate": 1.3904427613288738e-08, "loss": 0.0079, "step": 22432 }, { "epoch": 8.84, "learning_rate": 1.3895075126461464e-08, "loss": 0.0074, "step": 22433 }, { "epoch": 8.84, "learning_rate": 1.3885725672861592e-08, "loss": 0.0091, "step": 22434 }, { "epoch": 8.84, "learning_rate": 1.3876379252641534e-08, "loss": 0.0098, "step": 22435 }, { "epoch": 8.84, "learning_rate": 1.3867035865953636e-08, "loss": 0.0102, "step": 22436 }, { "epoch": 8.84, "learning_rate": 1.3857695512950174e-08, "loss": 0.0066, "step": 22437 }, { "epoch": 8.84, "learning_rate": 1.3848358193783338e-08, "loss": 0.0081, "step": 22438 }, { "epoch": 8.84, "learning_rate": 1.3839023908605363e-08, "loss": 0.0081, "step": 22439 }, { "epoch": 8.84, "learning_rate": 1.3829692657568348e-08, "loss": 0.0111, "step": 22440 }, { "epoch": 8.84, "learning_rate": 1.382036444082444e-08, "loss": 0.0103, "step": 22441 }, { "epoch": 8.84, "learning_rate": 1.381103925852558e-08, "loss": 0.0095, "step": 22442 }, { "epoch": 8.84, "learning_rate": 1.380171711082383e-08, "loss": 0.01, "step": 22443 }, { "epoch": 8.85, "learning_rate": 1.379239799787113e-08, "loss": 0.0118, "step": 22444 }, { "epoch": 8.85, "learning_rate": 1.3783081919819295e-08, "loss": 0.0108, "step": 22445 }, { "epoch": 8.85, "learning_rate": 1.3773768876820203e-08, "loss": 0.0115, "step": 22446 }, { "epoch": 8.85, "learning_rate": 1.3764458869025708e-08, "loss": 0.0098, "step": 22447 }, { "epoch": 8.85, "learning_rate": 1.3755151896587425e-08, "loss": 0.0075, "step": 22448 }, { "epoch": 8.85, "learning_rate": 1.374584795965712e-08, "loss": 0.0094, "step": 22449 }, { "epoch": 8.85, "learning_rate": 1.373654705838645e-08, "loss": 0.0091, "step": 22450 }, { "epoch": 8.85, "learning_rate": 1.372724919292696e-08, "loss": 0.0109, "step": 22451 }, { "epoch": 8.85, "learning_rate": 1.3717954363430173e-08, "loss": 0.0107, "step": 22452 }, { "epoch": 8.85, "learning_rate": 1.3708662570047635e-08, "loss": 0.0114, "step": 22453 }, { "epoch": 8.85, "learning_rate": 1.3699373812930804e-08, "loss": 0.0099, "step": 22454 }, { "epoch": 8.85, "learning_rate": 1.3690088092231001e-08, "loss": 0.0104, "step": 22455 }, { "epoch": 8.85, "learning_rate": 1.3680805408099616e-08, "loss": 0.0114, "step": 22456 }, { "epoch": 8.85, "learning_rate": 1.3671525760687885e-08, "loss": 0.0117, "step": 22457 }, { "epoch": 8.85, "learning_rate": 1.3662249150147153e-08, "loss": 0.0081, "step": 22458 }, { "epoch": 8.85, "learning_rate": 1.3652975576628545e-08, "loss": 0.0091, "step": 22459 }, { "epoch": 8.85, "learning_rate": 1.3643705040283226e-08, "loss": 0.0098, "step": 22460 }, { "epoch": 8.85, "learning_rate": 1.3634437541262234e-08, "loss": 0.0123, "step": 22461 }, { "epoch": 8.85, "learning_rate": 1.3625173079716734e-08, "loss": 0.0072, "step": 22462 }, { "epoch": 8.85, "learning_rate": 1.361591165579763e-08, "loss": 0.0082, "step": 22463 }, { "epoch": 8.85, "learning_rate": 1.360665326965591e-08, "loss": 0.0119, "step": 22464 }, { "epoch": 8.85, "learning_rate": 1.3597397921442433e-08, "loss": 0.0101, "step": 22465 }, { "epoch": 8.85, "learning_rate": 1.3588145611308054e-08, "loss": 0.0101, "step": 22466 }, { "epoch": 8.85, "learning_rate": 1.3578896339403655e-08, "loss": 0.0111, "step": 22467 }, { "epoch": 8.85, "learning_rate": 1.3569650105879849e-08, "loss": 0.0108, "step": 22468 }, { "epoch": 8.86, "learning_rate": 1.3560406910887423e-08, "loss": 0.0116, "step": 22469 }, { "epoch": 8.86, "learning_rate": 1.3551166754577015e-08, "loss": 0.012, "step": 22470 }, { "epoch": 8.86, "learning_rate": 1.3541929637099236e-08, "loss": 0.0104, "step": 22471 }, { "epoch": 8.86, "learning_rate": 1.3532695558604612e-08, "loss": 0.0096, "step": 22472 }, { "epoch": 8.86, "learning_rate": 1.3523464519243666e-08, "loss": 0.011, "step": 22473 }, { "epoch": 8.86, "learning_rate": 1.351423651916681e-08, "loss": 0.0098, "step": 22474 }, { "epoch": 8.86, "learning_rate": 1.3505011558524459e-08, "loss": 0.0115, "step": 22475 }, { "epoch": 8.86, "learning_rate": 1.3495789637467003e-08, "loss": 0.0104, "step": 22476 }, { "epoch": 8.86, "learning_rate": 1.3486570756144744e-08, "loss": 0.0089, "step": 22477 }, { "epoch": 8.86, "learning_rate": 1.3477354914707873e-08, "loss": 0.0092, "step": 22478 }, { "epoch": 8.86, "learning_rate": 1.3468142113306647e-08, "loss": 0.0109, "step": 22479 }, { "epoch": 8.86, "learning_rate": 1.3458932352091213e-08, "loss": 0.0095, "step": 22480 }, { "epoch": 8.86, "learning_rate": 1.3449725631211695e-08, "loss": 0.0105, "step": 22481 }, { "epoch": 8.86, "learning_rate": 1.3440521950818085e-08, "loss": 0.0089, "step": 22482 }, { "epoch": 8.86, "learning_rate": 1.343132131106044e-08, "loss": 0.0092, "step": 22483 }, { "epoch": 8.86, "learning_rate": 1.3422123712088706e-08, "loss": 0.0104, "step": 22484 }, { "epoch": 8.86, "learning_rate": 1.3412929154052831e-08, "loss": 0.0122, "step": 22485 }, { "epoch": 8.86, "learning_rate": 1.3403737637102585e-08, "loss": 0.0113, "step": 22486 }, { "epoch": 8.86, "learning_rate": 1.339454916138787e-08, "loss": 0.0094, "step": 22487 }, { "epoch": 8.86, "learning_rate": 1.338536372705834e-08, "loss": 0.0097, "step": 22488 }, { "epoch": 8.86, "learning_rate": 1.337618133426377e-08, "loss": 0.0086, "step": 22489 }, { "epoch": 8.86, "learning_rate": 1.3367001983153835e-08, "loss": 0.011, "step": 22490 }, { "epoch": 8.86, "learning_rate": 1.335782567387811e-08, "loss": 0.0097, "step": 22491 }, { "epoch": 8.86, "learning_rate": 1.3348652406586135e-08, "loss": 0.0073, "step": 22492 }, { "epoch": 8.86, "learning_rate": 1.3339482181427464e-08, "loss": 0.0095, "step": 22493 }, { "epoch": 8.87, "learning_rate": 1.3330314998551572e-08, "loss": 0.0107, "step": 22494 }, { "epoch": 8.87, "learning_rate": 1.3321150858107811e-08, "loss": 0.0095, "step": 22495 }, { "epoch": 8.87, "learning_rate": 1.331198976024559e-08, "loss": 0.0084, "step": 22496 }, { "epoch": 8.87, "learning_rate": 1.3302831705114148e-08, "loss": 0.0106, "step": 22497 }, { "epoch": 8.87, "learning_rate": 1.3293676692862877e-08, "loss": 0.0093, "step": 22498 }, { "epoch": 8.87, "learning_rate": 1.3284524723640877e-08, "loss": 0.0099, "step": 22499 }, { "epoch": 8.87, "learning_rate": 1.3275375797597365e-08, "loss": 0.0085, "step": 22500 }, { "epoch": 8.87, "learning_rate": 1.32662299148814e-08, "loss": 0.0118, "step": 22501 }, { "epoch": 8.87, "learning_rate": 1.3257087075642125e-08, "loss": 0.0088, "step": 22502 }, { "epoch": 8.87, "learning_rate": 1.3247947280028494e-08, "loss": 0.0141, "step": 22503 }, { "epoch": 8.87, "learning_rate": 1.3238810528189514e-08, "loss": 0.0109, "step": 22504 }, { "epoch": 8.87, "learning_rate": 1.3229676820274072e-08, "loss": 0.0111, "step": 22505 }, { "epoch": 8.87, "learning_rate": 1.3220546156431e-08, "loss": 0.0111, "step": 22506 }, { "epoch": 8.87, "learning_rate": 1.3211418536809183e-08, "loss": 0.0124, "step": 22507 }, { "epoch": 8.87, "learning_rate": 1.3202293961557387e-08, "loss": 0.0125, "step": 22508 }, { "epoch": 8.87, "learning_rate": 1.3193172430824273e-08, "loss": 0.0094, "step": 22509 }, { "epoch": 8.87, "learning_rate": 1.3184053944758523e-08, "loss": 0.009, "step": 22510 }, { "epoch": 8.87, "learning_rate": 1.3174938503508792e-08, "loss": 0.0119, "step": 22511 }, { "epoch": 8.87, "learning_rate": 1.3165826107223587e-08, "loss": 0.0097, "step": 22512 }, { "epoch": 8.87, "learning_rate": 1.3156716756051478e-08, "loss": 0.0127, "step": 22513 }, { "epoch": 8.87, "learning_rate": 1.3147610450140945e-08, "loss": 0.013, "step": 22514 }, { "epoch": 8.87, "learning_rate": 1.3138507189640358e-08, "loss": 0.0101, "step": 22515 }, { "epoch": 8.87, "learning_rate": 1.3129406974698087e-08, "loss": 0.0119, "step": 22516 }, { "epoch": 8.87, "learning_rate": 1.3120309805462504e-08, "loss": 0.0122, "step": 22517 }, { "epoch": 8.87, "learning_rate": 1.3111215682081822e-08, "loss": 0.0097, "step": 22518 }, { "epoch": 8.87, "learning_rate": 1.310212460470428e-08, "loss": 0.0129, "step": 22519 }, { "epoch": 8.88, "learning_rate": 1.3093036573478067e-08, "loss": 0.0117, "step": 22520 }, { "epoch": 8.88, "learning_rate": 1.3083951588551312e-08, "loss": 0.0105, "step": 22521 }, { "epoch": 8.88, "learning_rate": 1.3074869650072051e-08, "loss": 0.0093, "step": 22522 }, { "epoch": 8.88, "learning_rate": 1.30657907581883e-08, "loss": 0.011, "step": 22523 }, { "epoch": 8.88, "learning_rate": 1.3056714913048072e-08, "loss": 0.0103, "step": 22524 }, { "epoch": 8.88, "learning_rate": 1.3047642114799296e-08, "loss": 0.0102, "step": 22525 }, { "epoch": 8.88, "learning_rate": 1.3038572363589784e-08, "loss": 0.0089, "step": 22526 }, { "epoch": 8.88, "learning_rate": 1.302950565956742e-08, "loss": 0.0098, "step": 22527 }, { "epoch": 8.88, "learning_rate": 1.3020442002879883e-08, "loss": 0.0104, "step": 22528 }, { "epoch": 8.88, "learning_rate": 1.3011381393675037e-08, "loss": 0.0112, "step": 22529 }, { "epoch": 8.88, "learning_rate": 1.3002323832100471e-08, "loss": 0.0106, "step": 22530 }, { "epoch": 8.88, "learning_rate": 1.2993269318303824e-08, "loss": 0.0096, "step": 22531 }, { "epoch": 8.88, "learning_rate": 1.2984217852432667e-08, "loss": 0.0099, "step": 22532 }, { "epoch": 8.88, "learning_rate": 1.2975169434634504e-08, "loss": 0.0122, "step": 22533 }, { "epoch": 8.88, "learning_rate": 1.2966124065056883e-08, "loss": 0.0108, "step": 22534 }, { "epoch": 8.88, "learning_rate": 1.2957081743847131e-08, "loss": 0.0085, "step": 22535 }, { "epoch": 8.88, "learning_rate": 1.2948042471152686e-08, "loss": 0.0102, "step": 22536 }, { "epoch": 8.88, "learning_rate": 1.2939006247120854e-08, "loss": 0.0113, "step": 22537 }, { "epoch": 8.88, "learning_rate": 1.292997307189896e-08, "loss": 0.0111, "step": 22538 }, { "epoch": 8.88, "learning_rate": 1.292094294563415e-08, "loss": 0.0105, "step": 22539 }, { "epoch": 8.88, "learning_rate": 1.2911915868473688e-08, "loss": 0.0108, "step": 22540 }, { "epoch": 8.88, "learning_rate": 1.290289184056459e-08, "loss": 0.0111, "step": 22541 }, { "epoch": 8.88, "learning_rate": 1.2893870862054069e-08, "loss": 0.0107, "step": 22542 }, { "epoch": 8.88, "learning_rate": 1.288485293308903e-08, "loss": 0.0101, "step": 22543 }, { "epoch": 8.88, "learning_rate": 1.2875838053816556e-08, "loss": 0.0103, "step": 22544 }, { "epoch": 8.89, "learning_rate": 1.2866826224383487e-08, "loss": 0.0105, "step": 22545 }, { "epoch": 8.89, "learning_rate": 1.2857817444936747e-08, "loss": 0.0105, "step": 22546 }, { "epoch": 8.89, "learning_rate": 1.2848811715623153e-08, "loss": 0.0106, "step": 22547 }, { "epoch": 8.89, "learning_rate": 1.2839809036589522e-08, "loss": 0.0101, "step": 22548 }, { "epoch": 8.89, "learning_rate": 1.2830809407982534e-08, "loss": 0.0097, "step": 22549 }, { "epoch": 8.89, "learning_rate": 1.2821812829948875e-08, "loss": 0.0082, "step": 22550 }, { "epoch": 8.89, "learning_rate": 1.2812819302635181e-08, "loss": 0.0095, "step": 22551 }, { "epoch": 8.89, "learning_rate": 1.280382882618809e-08, "loss": 0.0103, "step": 22552 }, { "epoch": 8.89, "learning_rate": 1.2794841400754042e-08, "loss": 0.0093, "step": 22553 }, { "epoch": 8.89, "learning_rate": 1.2785857026479607e-08, "loss": 0.0091, "step": 22554 }, { "epoch": 8.89, "learning_rate": 1.2776875703511115e-08, "loss": 0.0119, "step": 22555 }, { "epoch": 8.89, "learning_rate": 1.2767897431995023e-08, "loss": 0.0089, "step": 22556 }, { "epoch": 8.89, "learning_rate": 1.2758922212077661e-08, "loss": 0.0098, "step": 22557 }, { "epoch": 8.89, "learning_rate": 1.2749950043905267e-08, "loss": 0.0092, "step": 22558 }, { "epoch": 8.89, "learning_rate": 1.27409809276241e-08, "loss": 0.0074, "step": 22559 }, { "epoch": 8.89, "learning_rate": 1.2732014863380358e-08, "loss": 0.0118, "step": 22560 }, { "epoch": 8.89, "learning_rate": 1.2723051851320165e-08, "loss": 0.0124, "step": 22561 }, { "epoch": 8.89, "learning_rate": 1.2714091891589585e-08, "loss": 0.0108, "step": 22562 }, { "epoch": 8.89, "learning_rate": 1.2705134984334654e-08, "loss": 0.0095, "step": 22563 }, { "epoch": 8.89, "learning_rate": 1.2696181129701367e-08, "loss": 0.0086, "step": 22564 }, { "epoch": 8.89, "learning_rate": 1.2687230327835697e-08, "loss": 0.0099, "step": 22565 }, { "epoch": 8.89, "learning_rate": 1.2678282578883459e-08, "loss": 0.0093, "step": 22566 }, { "epoch": 8.89, "learning_rate": 1.2669337882990539e-08, "loss": 0.0098, "step": 22567 }, { "epoch": 8.89, "learning_rate": 1.2660396240302663e-08, "loss": 0.0118, "step": 22568 }, { "epoch": 8.89, "learning_rate": 1.2651457650965647e-08, "loss": 0.0105, "step": 22569 }, { "epoch": 8.9, "learning_rate": 1.264252211512511e-08, "loss": 0.0123, "step": 22570 }, { "epoch": 8.9, "learning_rate": 1.2633589632926734e-08, "loss": 0.0102, "step": 22571 }, { "epoch": 8.9, "learning_rate": 1.2624660204516068e-08, "loss": 0.0077, "step": 22572 }, { "epoch": 8.9, "learning_rate": 1.2615733830038644e-08, "loss": 0.0105, "step": 22573 }, { "epoch": 8.9, "learning_rate": 1.2606810509639987e-08, "loss": 0.0096, "step": 22574 }, { "epoch": 8.9, "learning_rate": 1.2597890243465537e-08, "loss": 0.0097, "step": 22575 }, { "epoch": 8.9, "learning_rate": 1.2588973031660622e-08, "loss": 0.0085, "step": 22576 }, { "epoch": 8.9, "learning_rate": 1.2580058874370614e-08, "loss": 0.0099, "step": 22577 }, { "epoch": 8.9, "learning_rate": 1.257114777174082e-08, "loss": 0.0105, "step": 22578 }, { "epoch": 8.9, "learning_rate": 1.2562239723916457e-08, "loss": 0.0128, "step": 22579 }, { "epoch": 8.9, "learning_rate": 1.255333473104272e-08, "loss": 0.0104, "step": 22580 }, { "epoch": 8.9, "learning_rate": 1.2544432793264692e-08, "loss": 0.0093, "step": 22581 }, { "epoch": 8.9, "learning_rate": 1.2535533910727569e-08, "loss": 0.0122, "step": 22582 }, { "epoch": 8.9, "learning_rate": 1.2526638083576302e-08, "loss": 0.01, "step": 22583 }, { "epoch": 8.9, "learning_rate": 1.251774531195593e-08, "loss": 0.0103, "step": 22584 }, { "epoch": 8.9, "learning_rate": 1.2508855596011358e-08, "loss": 0.012, "step": 22585 }, { "epoch": 8.9, "learning_rate": 1.2499968935887474e-08, "loss": 0.0097, "step": 22586 }, { "epoch": 8.9, "learning_rate": 1.2491085331729111e-08, "loss": 0.0083, "step": 22587 }, { "epoch": 8.9, "learning_rate": 1.2482204783681139e-08, "loss": 0.0084, "step": 22588 }, { "epoch": 8.9, "learning_rate": 1.247332729188817e-08, "loss": 0.0079, "step": 22589 }, { "epoch": 8.9, "learning_rate": 1.2464452856494979e-08, "loss": 0.0109, "step": 22590 }, { "epoch": 8.9, "learning_rate": 1.2455581477646182e-08, "loss": 0.0122, "step": 22591 }, { "epoch": 8.9, "learning_rate": 1.2446713155486399e-08, "loss": 0.0117, "step": 22592 }, { "epoch": 8.9, "learning_rate": 1.2437847890160135e-08, "loss": 0.0095, "step": 22593 }, { "epoch": 8.9, "learning_rate": 1.2428985681811898e-08, "loss": 0.0115, "step": 22594 }, { "epoch": 8.9, "learning_rate": 1.2420126530586084e-08, "loss": 0.0101, "step": 22595 }, { "epoch": 8.91, "learning_rate": 1.241127043662713e-08, "loss": 0.0091, "step": 22596 }, { "epoch": 8.91, "learning_rate": 1.2402417400079368e-08, "loss": 0.0125, "step": 22597 }, { "epoch": 8.91, "learning_rate": 1.2393567421087103e-08, "loss": 0.0114, "step": 22598 }, { "epoch": 8.91, "learning_rate": 1.2384720499794555e-08, "loss": 0.0079, "step": 22599 }, { "epoch": 8.91, "learning_rate": 1.2375876636345894e-08, "loss": 0.0093, "step": 22600 }, { "epoch": 8.91, "learning_rate": 1.2367035830885342e-08, "loss": 0.0091, "step": 22601 }, { "epoch": 8.91, "learning_rate": 1.2358198083556893e-08, "loss": 0.0087, "step": 22602 }, { "epoch": 8.91, "learning_rate": 1.2349363394504631e-08, "loss": 0.0104, "step": 22603 }, { "epoch": 8.91, "learning_rate": 1.2340531763872553e-08, "loss": 0.0091, "step": 22604 }, { "epoch": 8.91, "learning_rate": 1.2331703191804633e-08, "loss": 0.0106, "step": 22605 }, { "epoch": 8.91, "learning_rate": 1.2322877678444687e-08, "loss": 0.0093, "step": 22606 }, { "epoch": 8.91, "learning_rate": 1.2314055223936626e-08, "loss": 0.012, "step": 22607 }, { "epoch": 8.91, "learning_rate": 1.2305235828424155e-08, "loss": 0.0101, "step": 22608 }, { "epoch": 8.91, "learning_rate": 1.2296419492051158e-08, "loss": 0.0107, "step": 22609 }, { "epoch": 8.91, "learning_rate": 1.2287606214961188e-08, "loss": 0.0111, "step": 22610 }, { "epoch": 8.91, "learning_rate": 1.2278795997297998e-08, "loss": 0.0088, "step": 22611 }, { "epoch": 8.91, "learning_rate": 1.226998883920507e-08, "loss": 0.0111, "step": 22612 }, { "epoch": 8.91, "learning_rate": 1.2261184740826069e-08, "loss": 0.0113, "step": 22613 }, { "epoch": 8.91, "learning_rate": 1.2252383702304369e-08, "loss": 0.0107, "step": 22614 }, { "epoch": 8.91, "learning_rate": 1.2243585723783523e-08, "loss": 0.013, "step": 22615 }, { "epoch": 8.91, "learning_rate": 1.2234790805406859e-08, "loss": 0.0112, "step": 22616 }, { "epoch": 8.91, "learning_rate": 1.2225998947317707e-08, "loss": 0.0126, "step": 22617 }, { "epoch": 8.91, "learning_rate": 1.2217210149659418e-08, "loss": 0.0102, "step": 22618 }, { "epoch": 8.91, "learning_rate": 1.220842441257519e-08, "loss": 0.0085, "step": 22619 }, { "epoch": 8.91, "learning_rate": 1.2199641736208243e-08, "loss": 0.0103, "step": 22620 }, { "epoch": 8.92, "learning_rate": 1.2190862120701684e-08, "loss": 0.0132, "step": 22621 }, { "epoch": 8.92, "learning_rate": 1.2182085566198686e-08, "loss": 0.0094, "step": 22622 }, { "epoch": 8.92, "learning_rate": 1.21733120728422e-08, "loss": 0.0108, "step": 22623 }, { "epoch": 8.92, "learning_rate": 1.2164541640775317e-08, "loss": 0.0117, "step": 22624 }, { "epoch": 8.92, "learning_rate": 1.2155774270140873e-08, "loss": 0.0106, "step": 22625 }, { "epoch": 8.92, "learning_rate": 1.2147009961081822e-08, "loss": 0.0076, "step": 22626 }, { "epoch": 8.92, "learning_rate": 1.2138248713741006e-08, "loss": 0.0093, "step": 22627 }, { "epoch": 8.92, "learning_rate": 1.2129490528261266e-08, "loss": 0.0091, "step": 22628 }, { "epoch": 8.92, "learning_rate": 1.2120735404785264e-08, "loss": 0.0116, "step": 22629 }, { "epoch": 8.92, "learning_rate": 1.2111983343455711e-08, "loss": 0.0096, "step": 22630 }, { "epoch": 8.92, "learning_rate": 1.2103234344415292e-08, "loss": 0.0112, "step": 22631 }, { "epoch": 8.92, "learning_rate": 1.2094488407806602e-08, "loss": 0.0114, "step": 22632 }, { "epoch": 8.92, "learning_rate": 1.2085745533772151e-08, "loss": 0.0088, "step": 22633 }, { "epoch": 8.92, "learning_rate": 1.207700572245447e-08, "loss": 0.0078, "step": 22634 }, { "epoch": 8.92, "learning_rate": 1.2068268973995932e-08, "loss": 0.0103, "step": 22635 }, { "epoch": 8.92, "learning_rate": 1.2059535288539046e-08, "loss": 0.0129, "step": 22636 }, { "epoch": 8.92, "learning_rate": 1.2050804666226056e-08, "loss": 0.0115, "step": 22637 }, { "epoch": 8.92, "learning_rate": 1.2042077107199333e-08, "loss": 0.0105, "step": 22638 }, { "epoch": 8.92, "learning_rate": 1.2033352611601078e-08, "loss": 0.011, "step": 22639 }, { "epoch": 8.92, "learning_rate": 1.2024631179573486e-08, "loss": 0.0119, "step": 22640 }, { "epoch": 8.92, "learning_rate": 1.2015912811258711e-08, "loss": 0.0107, "step": 22641 }, { "epoch": 8.92, "learning_rate": 1.2007197506798905e-08, "loss": 0.011, "step": 22642 }, { "epoch": 8.92, "learning_rate": 1.1998485266336e-08, "loss": 0.0102, "step": 22643 }, { "epoch": 8.92, "learning_rate": 1.1989776090012082e-08, "loss": 0.0102, "step": 22644 }, { "epoch": 8.92, "learning_rate": 1.1981069977969106e-08, "loss": 0.009, "step": 22645 }, { "epoch": 8.92, "learning_rate": 1.1972366930348887e-08, "loss": 0.0093, "step": 22646 }, { "epoch": 8.93, "learning_rate": 1.1963666947293339e-08, "loss": 0.0088, "step": 22647 }, { "epoch": 8.93, "learning_rate": 1.1954970028944188e-08, "loss": 0.0087, "step": 22648 }, { "epoch": 8.93, "learning_rate": 1.1946276175443304e-08, "loss": 0.0107, "step": 22649 }, { "epoch": 8.93, "learning_rate": 1.1937585386932281e-08, "loss": 0.0092, "step": 22650 }, { "epoch": 8.93, "learning_rate": 1.1928897663552805e-08, "loss": 0.0098, "step": 22651 }, { "epoch": 8.93, "learning_rate": 1.192021300544641e-08, "loss": 0.0119, "step": 22652 }, { "epoch": 8.93, "learning_rate": 1.1911531412754738e-08, "loss": 0.0114, "step": 22653 }, { "epoch": 8.93, "learning_rate": 1.190285288561923e-08, "loss": 0.0103, "step": 22654 }, { "epoch": 8.93, "learning_rate": 1.1894177424181374e-08, "loss": 0.0109, "step": 22655 }, { "epoch": 8.93, "learning_rate": 1.188550502858252e-08, "loss": 0.0093, "step": 22656 }, { "epoch": 8.93, "learning_rate": 1.1876835698964004e-08, "loss": 0.0086, "step": 22657 }, { "epoch": 8.93, "learning_rate": 1.1868169435467179e-08, "loss": 0.0101, "step": 22658 }, { "epoch": 8.93, "learning_rate": 1.1859506238233285e-08, "loss": 0.0102, "step": 22659 }, { "epoch": 8.93, "learning_rate": 1.1850846107403479e-08, "loss": 0.0093, "step": 22660 }, { "epoch": 8.93, "learning_rate": 1.184218904311891e-08, "loss": 0.0108, "step": 22661 }, { "epoch": 8.93, "learning_rate": 1.1833535045520715e-08, "loss": 0.0126, "step": 22662 }, { "epoch": 8.93, "learning_rate": 1.1824884114749912e-08, "loss": 0.0115, "step": 22663 }, { "epoch": 8.93, "learning_rate": 1.1816236250947497e-08, "loss": 0.0097, "step": 22664 }, { "epoch": 8.93, "learning_rate": 1.1807591454254451e-08, "loss": 0.01, "step": 22665 }, { "epoch": 8.93, "learning_rate": 1.1798949724811613e-08, "loss": 0.0135, "step": 22666 }, { "epoch": 8.93, "learning_rate": 1.1790311062759872e-08, "loss": 0.0106, "step": 22667 }, { "epoch": 8.93, "learning_rate": 1.1781675468240027e-08, "loss": 0.0102, "step": 22668 }, { "epoch": 8.93, "learning_rate": 1.1773042941392807e-08, "loss": 0.0121, "step": 22669 }, { "epoch": 8.93, "learning_rate": 1.1764413482358904e-08, "loss": 0.011, "step": 22670 }, { "epoch": 8.93, "learning_rate": 1.175578709127898e-08, "loss": 0.0089, "step": 22671 }, { "epoch": 8.94, "learning_rate": 1.1747163768293633e-08, "loss": 0.0123, "step": 22672 }, { "epoch": 8.94, "learning_rate": 1.1738543513543397e-08, "loss": 0.0091, "step": 22673 }, { "epoch": 8.94, "learning_rate": 1.1729926327168805e-08, "loss": 0.0087, "step": 22674 }, { "epoch": 8.94, "learning_rate": 1.172131220931023e-08, "loss": 0.0108, "step": 22675 }, { "epoch": 8.94, "learning_rate": 1.1712701160108162e-08, "loss": 0.0109, "step": 22676 }, { "epoch": 8.94, "learning_rate": 1.1704093179702867e-08, "loss": 0.0097, "step": 22677 }, { "epoch": 8.94, "learning_rate": 1.1695488268234721e-08, "loss": 0.0092, "step": 22678 }, { "epoch": 8.94, "learning_rate": 1.1686886425843856e-08, "loss": 0.0086, "step": 22679 }, { "epoch": 8.94, "learning_rate": 1.1678287652670583e-08, "loss": 0.0102, "step": 22680 }, { "epoch": 8.94, "learning_rate": 1.166969194885501e-08, "loss": 0.0101, "step": 22681 }, { "epoch": 8.94, "learning_rate": 1.1661099314537226e-08, "loss": 0.009, "step": 22682 }, { "epoch": 8.94, "learning_rate": 1.1652509749857275e-08, "loss": 0.0097, "step": 22683 }, { "epoch": 8.94, "learning_rate": 1.1643923254955157e-08, "loss": 0.0092, "step": 22684 }, { "epoch": 8.94, "learning_rate": 1.1635339829970825e-08, "loss": 0.0113, "step": 22685 }, { "epoch": 8.94, "learning_rate": 1.1626759475044168e-08, "loss": 0.0114, "step": 22686 }, { "epoch": 8.94, "learning_rate": 1.1618182190315029e-08, "loss": 0.0121, "step": 22687 }, { "epoch": 8.94, "learning_rate": 1.1609607975923231e-08, "loss": 0.0093, "step": 22688 }, { "epoch": 8.94, "learning_rate": 1.1601036832008504e-08, "loss": 0.0116, "step": 22689 }, { "epoch": 8.94, "learning_rate": 1.1592468758710538e-08, "loss": 0.0094, "step": 22690 }, { "epoch": 8.94, "learning_rate": 1.1583903756169e-08, "loss": 0.0079, "step": 22691 }, { "epoch": 8.94, "learning_rate": 1.1575341824523421e-08, "loss": 0.012, "step": 22692 }, { "epoch": 8.94, "learning_rate": 1.1566782963913446e-08, "loss": 0.0093, "step": 22693 }, { "epoch": 8.94, "learning_rate": 1.1558227174478497e-08, "loss": 0.0109, "step": 22694 }, { "epoch": 8.94, "learning_rate": 1.1549674456358083e-08, "loss": 0.0095, "step": 22695 }, { "epoch": 8.94, "learning_rate": 1.1541124809691538e-08, "loss": 0.0097, "step": 22696 }, { "epoch": 8.95, "learning_rate": 1.1532578234618218e-08, "loss": 0.0078, "step": 22697 }, { "epoch": 8.95, "learning_rate": 1.1524034731277432e-08, "loss": 0.0118, "step": 22698 }, { "epoch": 8.95, "learning_rate": 1.1515494299808471e-08, "loss": 0.0111, "step": 22699 }, { "epoch": 8.95, "learning_rate": 1.1506956940350465e-08, "loss": 0.0099, "step": 22700 }, { "epoch": 8.95, "learning_rate": 1.1498422653042572e-08, "loss": 0.0106, "step": 22701 }, { "epoch": 8.95, "learning_rate": 1.1489891438023924e-08, "loss": 0.0096, "step": 22702 }, { "epoch": 8.95, "learning_rate": 1.1481363295433543e-08, "loss": 0.0096, "step": 22703 }, { "epoch": 8.95, "learning_rate": 1.1472838225410409e-08, "loss": 0.0082, "step": 22704 }, { "epoch": 8.95, "learning_rate": 1.1464316228093518e-08, "loss": 0.0094, "step": 22705 }, { "epoch": 8.95, "learning_rate": 1.1455797303621695e-08, "loss": 0.0079, "step": 22706 }, { "epoch": 8.95, "learning_rate": 1.1447281452133828e-08, "loss": 0.0091, "step": 22707 }, { "epoch": 8.95, "learning_rate": 1.1438768673768718e-08, "loss": 0.0094, "step": 22708 }, { "epoch": 8.95, "learning_rate": 1.1430258968665097e-08, "loss": 0.0112, "step": 22709 }, { "epoch": 8.95, "learning_rate": 1.1421752336961633e-08, "loss": 0.0079, "step": 22710 }, { "epoch": 8.95, "learning_rate": 1.1413248778797013e-08, "loss": 0.0102, "step": 22711 }, { "epoch": 8.95, "learning_rate": 1.1404748294309862e-08, "loss": 0.0115, "step": 22712 }, { "epoch": 8.95, "learning_rate": 1.1396250883638624e-08, "loss": 0.0078, "step": 22713 }, { "epoch": 8.95, "learning_rate": 1.13877565469219e-08, "loss": 0.0109, "step": 22714 }, { "epoch": 8.95, "learning_rate": 1.1379265284298023e-08, "loss": 0.0137, "step": 22715 }, { "epoch": 8.95, "learning_rate": 1.1370777095905525e-08, "loss": 0.0118, "step": 22716 }, { "epoch": 8.95, "learning_rate": 1.1362291981882632e-08, "loss": 0.0101, "step": 22717 }, { "epoch": 8.95, "learning_rate": 1.135380994236772e-08, "loss": 0.0088, "step": 22718 }, { "epoch": 8.95, "learning_rate": 1.1345330977498923e-08, "loss": 0.0086, "step": 22719 }, { "epoch": 8.95, "learning_rate": 1.1336855087414577e-08, "loss": 0.01, "step": 22720 }, { "epoch": 8.95, "learning_rate": 1.1328382272252723e-08, "loss": 0.0117, "step": 22721 }, { "epoch": 8.95, "learning_rate": 1.1319912532151543e-08, "loss": 0.009, "step": 22722 }, { "epoch": 8.96, "learning_rate": 1.131144586724897e-08, "loss": 0.0087, "step": 22723 }, { "epoch": 8.96, "learning_rate": 1.130298227768307e-08, "loss": 0.0126, "step": 22724 }, { "epoch": 8.96, "learning_rate": 1.129452176359178e-08, "loss": 0.0081, "step": 22725 }, { "epoch": 8.96, "learning_rate": 1.1286064325112987e-08, "loss": 0.0121, "step": 22726 }, { "epoch": 8.96, "learning_rate": 1.1277609962384537e-08, "loss": 0.0098, "step": 22727 }, { "epoch": 8.96, "learning_rate": 1.126915867554421e-08, "loss": 0.0092, "step": 22728 }, { "epoch": 8.96, "learning_rate": 1.1260710464729784e-08, "loss": 0.0097, "step": 22729 }, { "epoch": 8.96, "learning_rate": 1.125226533007888e-08, "loss": 0.0092, "step": 22730 }, { "epoch": 8.96, "learning_rate": 1.1243823271729236e-08, "loss": 0.0133, "step": 22731 }, { "epoch": 8.96, "learning_rate": 1.1235384289818339e-08, "loss": 0.0099, "step": 22732 }, { "epoch": 8.96, "learning_rate": 1.1226948384483836e-08, "loss": 0.0125, "step": 22733 }, { "epoch": 8.96, "learning_rate": 1.121851555586315e-08, "loss": 0.0101, "step": 22734 }, { "epoch": 8.96, "learning_rate": 1.1210085804093795e-08, "loss": 0.0099, "step": 22735 }, { "epoch": 8.96, "learning_rate": 1.1201659129313056e-08, "loss": 0.0088, "step": 22736 }, { "epoch": 8.96, "learning_rate": 1.119323553165834e-08, "loss": 0.0104, "step": 22737 }, { "epoch": 8.96, "learning_rate": 1.1184815011266936e-08, "loss": 0.0098, "step": 22738 }, { "epoch": 8.96, "learning_rate": 1.1176397568276108e-08, "loss": 0.011, "step": 22739 }, { "epoch": 8.96, "learning_rate": 1.1167983202822994e-08, "loss": 0.0104, "step": 22740 }, { "epoch": 8.96, "learning_rate": 1.1159571915044752e-08, "loss": 0.0106, "step": 22741 }, { "epoch": 8.96, "learning_rate": 1.1151163705078493e-08, "loss": 0.0107, "step": 22742 }, { "epoch": 8.96, "learning_rate": 1.1142758573061284e-08, "loss": 0.0102, "step": 22743 }, { "epoch": 8.96, "learning_rate": 1.1134356519130039e-08, "loss": 0.0126, "step": 22744 }, { "epoch": 8.96, "learning_rate": 1.1125957543421782e-08, "loss": 0.0098, "step": 22745 }, { "epoch": 8.96, "learning_rate": 1.1117561646073315e-08, "loss": 0.0094, "step": 22746 }, { "epoch": 8.96, "learning_rate": 1.1109168827221526e-08, "loss": 0.0126, "step": 22747 }, { "epoch": 8.97, "learning_rate": 1.1100779087003221e-08, "loss": 0.01, "step": 22748 }, { "epoch": 8.97, "learning_rate": 1.1092392425555153e-08, "loss": 0.0118, "step": 22749 }, { "epoch": 8.97, "learning_rate": 1.1084008843013926e-08, "loss": 0.0121, "step": 22750 }, { "epoch": 8.97, "learning_rate": 1.1075628339516253e-08, "loss": 0.0132, "step": 22751 }, { "epoch": 8.97, "learning_rate": 1.1067250915198734e-08, "loss": 0.0106, "step": 22752 }, { "epoch": 8.97, "learning_rate": 1.105887657019786e-08, "loss": 0.009, "step": 22753 }, { "epoch": 8.97, "learning_rate": 1.105050530465017e-08, "loss": 0.0075, "step": 22754 }, { "epoch": 8.97, "learning_rate": 1.1042137118691997e-08, "loss": 0.0104, "step": 22755 }, { "epoch": 8.97, "learning_rate": 1.1033772012459874e-08, "loss": 0.0114, "step": 22756 }, { "epoch": 8.97, "learning_rate": 1.1025409986090051e-08, "loss": 0.0097, "step": 22757 }, { "epoch": 8.97, "learning_rate": 1.101705103971886e-08, "loss": 0.0111, "step": 22758 }, { "epoch": 8.97, "learning_rate": 1.100869517348244e-08, "loss": 0.0092, "step": 22759 }, { "epoch": 8.97, "learning_rate": 1.1000342387517147e-08, "loss": 0.0119, "step": 22760 }, { "epoch": 8.97, "learning_rate": 1.0991992681958983e-08, "loss": 0.0088, "step": 22761 }, { "epoch": 8.97, "learning_rate": 1.0983646056944085e-08, "loss": 0.0099, "step": 22762 }, { "epoch": 8.97, "learning_rate": 1.0975302512608454e-08, "loss": 0.0096, "step": 22763 }, { "epoch": 8.97, "learning_rate": 1.0966962049088158e-08, "loss": 0.0103, "step": 22764 }, { "epoch": 8.97, "learning_rate": 1.0958624666519068e-08, "loss": 0.0148, "step": 22765 }, { "epoch": 8.97, "learning_rate": 1.0950290365037096e-08, "loss": 0.0109, "step": 22766 }, { "epoch": 8.97, "learning_rate": 1.0941959144778046e-08, "loss": 0.0093, "step": 22767 }, { "epoch": 8.97, "learning_rate": 1.0933631005877719e-08, "loss": 0.0084, "step": 22768 }, { "epoch": 8.97, "learning_rate": 1.0925305948471874e-08, "loss": 0.011, "step": 22769 }, { "epoch": 8.97, "learning_rate": 1.0916983972696203e-08, "loss": 0.0081, "step": 22770 }, { "epoch": 8.97, "learning_rate": 1.0908665078686285e-08, "loss": 0.0124, "step": 22771 }, { "epoch": 8.97, "learning_rate": 1.0900349266577746e-08, "loss": 0.0091, "step": 22772 }, { "epoch": 8.98, "learning_rate": 1.0892036536506143e-08, "loss": 0.0089, "step": 22773 }, { "epoch": 8.98, "learning_rate": 1.0883726888606903e-08, "loss": 0.0126, "step": 22774 }, { "epoch": 8.98, "learning_rate": 1.0875420323015538e-08, "loss": 0.0097, "step": 22775 }, { "epoch": 8.98, "learning_rate": 1.0867116839867341e-08, "loss": 0.0115, "step": 22776 }, { "epoch": 8.98, "learning_rate": 1.085881643929769e-08, "loss": 0.0111, "step": 22777 }, { "epoch": 8.98, "learning_rate": 1.085051912144186e-08, "loss": 0.01, "step": 22778 }, { "epoch": 8.98, "learning_rate": 1.0842224886435136e-08, "loss": 0.0105, "step": 22779 }, { "epoch": 8.98, "learning_rate": 1.0833933734412637e-08, "loss": 0.0107, "step": 22780 }, { "epoch": 8.98, "learning_rate": 1.0825645665509497e-08, "loss": 0.0138, "step": 22781 }, { "epoch": 8.98, "learning_rate": 1.0817360679860832e-08, "loss": 0.0114, "step": 22782 }, { "epoch": 8.98, "learning_rate": 1.080907877760171e-08, "loss": 0.0118, "step": 22783 }, { "epoch": 8.98, "learning_rate": 1.0800799958867024e-08, "loss": 0.0116, "step": 22784 }, { "epoch": 8.98, "learning_rate": 1.0792524223791777e-08, "loss": 0.0098, "step": 22785 }, { "epoch": 8.98, "learning_rate": 1.078425157251075e-08, "loss": 0.0111, "step": 22786 }, { "epoch": 8.98, "learning_rate": 1.0775982005158923e-08, "loss": 0.0094, "step": 22787 }, { "epoch": 8.98, "learning_rate": 1.0767715521870968e-08, "loss": 0.0083, "step": 22788 }, { "epoch": 8.98, "learning_rate": 1.0759452122781686e-08, "loss": 0.011, "step": 22789 }, { "epoch": 8.98, "learning_rate": 1.0751191808025683e-08, "loss": 0.0083, "step": 22790 }, { "epoch": 8.98, "learning_rate": 1.074293457773765e-08, "loss": 0.0093, "step": 22791 }, { "epoch": 8.98, "learning_rate": 1.0734680432052123e-08, "loss": 0.0094, "step": 22792 }, { "epoch": 8.98, "learning_rate": 1.0726429371103685e-08, "loss": 0.0109, "step": 22793 }, { "epoch": 8.98, "learning_rate": 1.0718181395026759e-08, "loss": 0.0102, "step": 22794 }, { "epoch": 8.98, "learning_rate": 1.0709936503955818e-08, "loss": 0.0098, "step": 22795 }, { "epoch": 8.98, "learning_rate": 1.0701694698025243e-08, "loss": 0.0121, "step": 22796 }, { "epoch": 8.98, "learning_rate": 1.0693455977369303e-08, "loss": 0.0077, "step": 22797 }, { "epoch": 8.98, "learning_rate": 1.0685220342122358e-08, "loss": 0.0076, "step": 22798 }, { "epoch": 8.99, "learning_rate": 1.0676987792418524e-08, "loss": 0.009, "step": 22799 }, { "epoch": 8.99, "learning_rate": 1.0668758328392136e-08, "loss": 0.0081, "step": 22800 }, { "epoch": 8.99, "learning_rate": 1.0660531950177176e-08, "loss": 0.0108, "step": 22801 }, { "epoch": 8.99, "learning_rate": 1.0652308657907827e-08, "loss": 0.0094, "step": 22802 }, { "epoch": 8.99, "learning_rate": 1.0644088451718024e-08, "loss": 0.0109, "step": 22803 }, { "epoch": 8.99, "learning_rate": 1.0635871331741842e-08, "loss": 0.0082, "step": 22804 }, { "epoch": 8.99, "learning_rate": 1.0627657298113124e-08, "loss": 0.0085, "step": 22805 }, { "epoch": 8.99, "learning_rate": 1.0619446350965811e-08, "loss": 0.0087, "step": 22806 }, { "epoch": 8.99, "learning_rate": 1.0611238490433661e-08, "loss": 0.0096, "step": 22807 }, { "epoch": 8.99, "learning_rate": 1.06030337166505e-08, "loss": 0.0136, "step": 22808 }, { "epoch": 8.99, "learning_rate": 1.0594832029750023e-08, "loss": 0.0111, "step": 22809 }, { "epoch": 8.99, "learning_rate": 1.0586633429865966e-08, "loss": 0.0094, "step": 22810 }, { "epoch": 8.99, "learning_rate": 1.0578437917131866e-08, "loss": 0.0102, "step": 22811 }, { "epoch": 8.99, "learning_rate": 1.0570245491681351e-08, "loss": 0.0129, "step": 22812 }, { "epoch": 8.99, "learning_rate": 1.0562056153647959e-08, "loss": 0.0108, "step": 22813 }, { "epoch": 8.99, "learning_rate": 1.0553869903165114e-08, "loss": 0.0076, "step": 22814 }, { "epoch": 8.99, "learning_rate": 1.0545686740366267e-08, "loss": 0.0122, "step": 22815 }, { "epoch": 8.99, "learning_rate": 1.0537506665384799e-08, "loss": 0.0102, "step": 22816 }, { "epoch": 8.99, "learning_rate": 1.0529329678354004e-08, "loss": 0.0115, "step": 22817 }, { "epoch": 8.99, "learning_rate": 1.0521155779407175e-08, "loss": 0.0092, "step": 22818 }, { "epoch": 8.99, "learning_rate": 1.0512984968677563e-08, "loss": 0.0085, "step": 22819 }, { "epoch": 8.99, "learning_rate": 1.0504817246298282e-08, "loss": 0.0101, "step": 22820 }, { "epoch": 8.99, "learning_rate": 1.049665261240249e-08, "loss": 0.0145, "step": 22821 }, { "epoch": 8.99, "learning_rate": 1.0488491067123217e-08, "loss": 0.0104, "step": 22822 }, { "epoch": 8.99, "learning_rate": 1.0480332610593578e-08, "loss": 0.0135, "step": 22823 }, { "epoch": 9.0, "learning_rate": 1.0472177242946423e-08, "loss": 0.0112, "step": 22824 }, { "epoch": 9.0, "learning_rate": 1.0464024964314777e-08, "loss": 0.0089, "step": 22825 }, { "epoch": 9.0, "learning_rate": 1.0455875774831402e-08, "loss": 0.0094, "step": 22826 }, { "epoch": 9.0, "learning_rate": 1.0447729674629236e-08, "loss": 0.0084, "step": 22827 }, { "epoch": 9.0, "learning_rate": 1.043958666384095e-08, "loss": 0.0087, "step": 22828 }, { "epoch": 9.0, "learning_rate": 1.0431446742599348e-08, "loss": 0.0109, "step": 22829 }, { "epoch": 9.0, "learning_rate": 1.0423309911036993e-08, "loss": 0.011, "step": 22830 }, { "epoch": 9.0, "learning_rate": 1.04151761692866e-08, "loss": 0.0111, "step": 22831 }, { "epoch": 9.0, "learning_rate": 1.0407045517480707e-08, "loss": 0.0101, "step": 22832 }, { "epoch": 9.0, "learning_rate": 1.0398917955751829e-08, "loss": 0.0099, "step": 22833 }, { "epoch": 9.0, "learning_rate": 1.0390793484232396e-08, "loss": 0.0124, "step": 22834 }, { "epoch": 9.0, "learning_rate": 1.0382672103054857e-08, "loss": 0.011, "step": 22835 }, { "epoch": 9.0, "learning_rate": 1.0374553812351617e-08, "loss": 0.0094, "step": 22836 }, { "epoch": 9.0, "learning_rate": 1.0366438612254925e-08, "loss": 0.0114, "step": 22837 }, { "epoch": 9.0, "learning_rate": 1.0358326502897052e-08, "loss": 0.0126, "step": 22838 }, { "epoch": 9.0, "learning_rate": 1.0350217484410251e-08, "loss": 0.0085, "step": 22839 }, { "epoch": 9.0, "learning_rate": 1.0342111556926702e-08, "loss": 0.0086, "step": 22840 }, { "epoch": 9.0, "learning_rate": 1.0334008720578458e-08, "loss": 0.0095, "step": 22841 }, { "epoch": 9.0, "learning_rate": 1.0325908975497655e-08, "loss": 0.0125, "step": 22842 }, { "epoch": 9.0, "learning_rate": 1.0317812321816189e-08, "loss": 0.0108, "step": 22843 }, { "epoch": 9.0, "learning_rate": 1.0309718759666153e-08, "loss": 0.0108, "step": 22844 }, { "epoch": 9.0, "learning_rate": 1.0301628289179377e-08, "loss": 0.0095, "step": 22845 }, { "epoch": 9.0, "learning_rate": 1.0293540910487797e-08, "loss": 0.0101, "step": 22846 }, { "epoch": 9.0, "learning_rate": 1.0285456623723132e-08, "loss": 0.0094, "step": 22847 }, { "epoch": 9.0, "learning_rate": 1.0277375429017187e-08, "loss": 0.0097, "step": 22848 }, { "epoch": 9.0, "learning_rate": 1.0269297326501702e-08, "loss": 0.0089, "step": 22849 }, { "epoch": 9.01, "learning_rate": 1.0261222316308327e-08, "loss": 0.0089, "step": 22850 }, { "epoch": 9.01, "learning_rate": 1.0253150398568621e-08, "loss": 0.0096, "step": 22851 }, { "epoch": 9.01, "learning_rate": 1.0245081573414193e-08, "loss": 0.0106, "step": 22852 }, { "epoch": 9.01, "learning_rate": 1.0237015840976537e-08, "loss": 0.009, "step": 22853 }, { "epoch": 9.01, "learning_rate": 1.0228953201387147e-08, "loss": 0.01, "step": 22854 }, { "epoch": 9.01, "learning_rate": 1.0220893654777384e-08, "loss": 0.0093, "step": 22855 }, { "epoch": 9.01, "learning_rate": 1.0212837201278634e-08, "loss": 0.0096, "step": 22856 }, { "epoch": 9.01, "learning_rate": 1.0204783841022168e-08, "loss": 0.0085, "step": 22857 }, { "epoch": 9.01, "learning_rate": 1.0196733574139283e-08, "loss": 0.0089, "step": 22858 }, { "epoch": 9.01, "learning_rate": 1.0188686400761182e-08, "loss": 0.0119, "step": 22859 }, { "epoch": 9.01, "learning_rate": 1.0180642321019006e-08, "loss": 0.0121, "step": 22860 }, { "epoch": 9.01, "learning_rate": 1.017260133504385e-08, "loss": 0.0111, "step": 22861 }, { "epoch": 9.01, "learning_rate": 1.0164563442966812e-08, "loss": 0.0119, "step": 22862 }, { "epoch": 9.01, "learning_rate": 1.0156528644918872e-08, "loss": 0.0083, "step": 22863 }, { "epoch": 9.01, "learning_rate": 1.0148496941030971e-08, "loss": 0.0109, "step": 22864 }, { "epoch": 9.01, "learning_rate": 1.0140468331434071e-08, "loss": 0.0088, "step": 22865 }, { "epoch": 9.01, "learning_rate": 1.0132442816258913e-08, "loss": 0.0109, "step": 22866 }, { "epoch": 9.01, "learning_rate": 1.0124420395636412e-08, "loss": 0.0127, "step": 22867 }, { "epoch": 9.01, "learning_rate": 1.0116401069697267e-08, "loss": 0.0116, "step": 22868 }, { "epoch": 9.01, "learning_rate": 1.0108384838572236e-08, "loss": 0.0113, "step": 22869 }, { "epoch": 9.01, "learning_rate": 1.0100371702391863e-08, "loss": 0.0108, "step": 22870 }, { "epoch": 9.01, "learning_rate": 1.0092361661286864e-08, "loss": 0.0105, "step": 22871 }, { "epoch": 9.01, "learning_rate": 1.0084354715387733e-08, "loss": 0.0105, "step": 22872 }, { "epoch": 9.01, "learning_rate": 1.0076350864825011e-08, "loss": 0.0124, "step": 22873 }, { "epoch": 9.01, "learning_rate": 1.0068350109729062e-08, "loss": 0.0092, "step": 22874 }, { "epoch": 9.02, "learning_rate": 1.006035245023038e-08, "loss": 0.0104, "step": 22875 }, { "epoch": 9.02, "learning_rate": 1.0052357886459261e-08, "loss": 0.0084, "step": 22876 }, { "epoch": 9.02, "learning_rate": 1.0044366418546046e-08, "loss": 0.0081, "step": 22877 }, { "epoch": 9.02, "learning_rate": 1.0036378046620941e-08, "loss": 0.0097, "step": 22878 }, { "epoch": 9.02, "learning_rate": 1.0028392770814153e-08, "loss": 0.0107, "step": 22879 }, { "epoch": 9.02, "learning_rate": 1.0020410591255867e-08, "loss": 0.0096, "step": 22880 }, { "epoch": 9.02, "learning_rate": 1.0012431508076136e-08, "loss": 0.01, "step": 22881 }, { "epoch": 9.02, "learning_rate": 1.0004455521405032e-08, "loss": 0.0104, "step": 22882 }, { "epoch": 9.02, "learning_rate": 9.996482631372493e-09, "loss": 0.0133, "step": 22883 }, { "epoch": 9.02, "learning_rate": 9.988512838108576e-09, "loss": 0.0106, "step": 22884 }, { "epoch": 9.02, "learning_rate": 9.980546141743085e-09, "loss": 0.0084, "step": 22885 }, { "epoch": 9.02, "learning_rate": 9.972582542405939e-09, "loss": 0.0108, "step": 22886 }, { "epoch": 9.02, "learning_rate": 9.964622040226856e-09, "loss": 0.0102, "step": 22887 }, { "epoch": 9.02, "learning_rate": 9.956664635335599e-09, "loss": 0.0103, "step": 22888 }, { "epoch": 9.02, "learning_rate": 9.948710327861887e-09, "loss": 0.0111, "step": 22889 }, { "epoch": 9.02, "learning_rate": 9.94075911793537e-09, "loss": 0.009, "step": 22890 }, { "epoch": 9.02, "learning_rate": 9.932811005685592e-09, "loss": 0.013, "step": 22891 }, { "epoch": 9.02, "learning_rate": 9.924865991242137e-09, "loss": 0.0077, "step": 22892 }, { "epoch": 9.02, "learning_rate": 9.916924074734479e-09, "loss": 0.0109, "step": 22893 }, { "epoch": 9.02, "learning_rate": 9.908985256292112e-09, "loss": 0.0095, "step": 22894 }, { "epoch": 9.02, "learning_rate": 9.901049536044337e-09, "loss": 0.0113, "step": 22895 }, { "epoch": 9.02, "learning_rate": 9.893116914120558e-09, "loss": 0.0108, "step": 22896 }, { "epoch": 9.02, "learning_rate": 9.885187390650008e-09, "loss": 0.0089, "step": 22897 }, { "epoch": 9.02, "learning_rate": 9.877260965762002e-09, "loss": 0.009, "step": 22898 }, { "epoch": 9.02, "learning_rate": 9.869337639585662e-09, "loss": 0.0118, "step": 22899 }, { "epoch": 9.03, "learning_rate": 9.861417412250195e-09, "loss": 0.0101, "step": 22900 }, { "epoch": 9.03, "learning_rate": 9.853500283884609e-09, "loss": 0.0088, "step": 22901 }, { "epoch": 9.03, "learning_rate": 9.845586254617976e-09, "loss": 0.01, "step": 22902 }, { "epoch": 9.03, "learning_rate": 9.837675324579309e-09, "loss": 0.0098, "step": 22903 }, { "epoch": 9.03, "learning_rate": 9.829767493897501e-09, "loss": 0.0127, "step": 22904 }, { "epoch": 9.03, "learning_rate": 9.821862762701471e-09, "loss": 0.0095, "step": 22905 }, { "epoch": 9.03, "learning_rate": 9.813961131120007e-09, "loss": 0.0081, "step": 22906 }, { "epoch": 9.03, "learning_rate": 9.806062599281983e-09, "loss": 0.0113, "step": 22907 }, { "epoch": 9.03, "learning_rate": 9.79816716731603e-09, "loss": 0.0089, "step": 22908 }, { "epoch": 9.03, "learning_rate": 9.790274835350909e-09, "loss": 0.0109, "step": 22909 }, { "epoch": 9.03, "learning_rate": 9.782385603515187e-09, "loss": 0.0123, "step": 22910 }, { "epoch": 9.03, "learning_rate": 9.774499471937515e-09, "loss": 0.01, "step": 22911 }, { "epoch": 9.03, "learning_rate": 9.766616440746388e-09, "loss": 0.0101, "step": 22912 }, { "epoch": 9.03, "learning_rate": 9.758736510070309e-09, "loss": 0.0132, "step": 22913 }, { "epoch": 9.03, "learning_rate": 9.75085968003766e-09, "loss": 0.012, "step": 22914 }, { "epoch": 9.03, "learning_rate": 9.74298595077685e-09, "loss": 0.0072, "step": 22915 }, { "epoch": 9.03, "learning_rate": 9.73511532241622e-09, "loss": 0.0084, "step": 22916 }, { "epoch": 9.03, "learning_rate": 9.727247795084049e-09, "loss": 0.0096, "step": 22917 }, { "epoch": 9.03, "learning_rate": 9.719383368908563e-09, "loss": 0.0105, "step": 22918 }, { "epoch": 9.03, "learning_rate": 9.711522044017906e-09, "loss": 0.0097, "step": 22919 }, { "epoch": 9.03, "learning_rate": 9.703663820540242e-09, "loss": 0.0098, "step": 22920 }, { "epoch": 9.03, "learning_rate": 9.69580869860367e-09, "loss": 0.0084, "step": 22921 }, { "epoch": 9.03, "learning_rate": 9.687956678336173e-09, "loss": 0.0122, "step": 22922 }, { "epoch": 9.03, "learning_rate": 9.680107759865719e-09, "loss": 0.0089, "step": 22923 }, { "epoch": 9.03, "learning_rate": 9.672261943320292e-09, "loss": 0.0094, "step": 22924 }, { "epoch": 9.03, "learning_rate": 9.664419228827702e-09, "loss": 0.0094, "step": 22925 }, { "epoch": 9.04, "learning_rate": 9.656579616515847e-09, "loss": 0.0119, "step": 22926 }, { "epoch": 9.04, "learning_rate": 9.648743106512403e-09, "loss": 0.0099, "step": 22927 }, { "epoch": 9.04, "learning_rate": 9.640909698945154e-09, "loss": 0.0081, "step": 22928 }, { "epoch": 9.04, "learning_rate": 9.633079393941779e-09, "loss": 0.0097, "step": 22929 }, { "epoch": 9.04, "learning_rate": 9.625252191629906e-09, "loss": 0.0117, "step": 22930 }, { "epoch": 9.04, "learning_rate": 9.617428092137036e-09, "loss": 0.0073, "step": 22931 }, { "epoch": 9.04, "learning_rate": 9.609607095590755e-09, "loss": 0.0099, "step": 22932 }, { "epoch": 9.04, "learning_rate": 9.601789202118538e-09, "loss": 0.0085, "step": 22933 }, { "epoch": 9.04, "learning_rate": 9.593974411847794e-09, "loss": 0.0103, "step": 22934 }, { "epoch": 9.04, "learning_rate": 9.586162724905844e-09, "loss": 0.0098, "step": 22935 }, { "epoch": 9.04, "learning_rate": 9.578354141420098e-09, "loss": 0.0102, "step": 22936 }, { "epoch": 9.04, "learning_rate": 9.570548661517697e-09, "loss": 0.0095, "step": 22937 }, { "epoch": 9.04, "learning_rate": 9.562746285326007e-09, "loss": 0.0093, "step": 22938 }, { "epoch": 9.04, "learning_rate": 9.554947012972081e-09, "loss": 0.0143, "step": 22939 }, { "epoch": 9.04, "learning_rate": 9.547150844583086e-09, "loss": 0.0091, "step": 22940 }, { "epoch": 9.04, "learning_rate": 9.539357780286073e-09, "loss": 0.0112, "step": 22941 }, { "epoch": 9.04, "learning_rate": 9.531567820208052e-09, "loss": 0.0109, "step": 22942 }, { "epoch": 9.04, "learning_rate": 9.523780964476013e-09, "loss": 0.0115, "step": 22943 }, { "epoch": 9.04, "learning_rate": 9.515997213216875e-09, "loss": 0.0082, "step": 22944 }, { "epoch": 9.04, "learning_rate": 9.508216566557448e-09, "loss": 0.01, "step": 22945 }, { "epoch": 9.04, "learning_rate": 9.500439024624563e-09, "loss": 0.0121, "step": 22946 }, { "epoch": 9.04, "learning_rate": 9.492664587545052e-09, "loss": 0.0105, "step": 22947 }, { "epoch": 9.04, "learning_rate": 9.484893255445525e-09, "loss": 0.0105, "step": 22948 }, { "epoch": 9.04, "learning_rate": 9.477125028452726e-09, "loss": 0.0079, "step": 22949 }, { "epoch": 9.04, "learning_rate": 9.469359906693153e-09, "loss": 0.0067, "step": 22950 }, { "epoch": 9.05, "learning_rate": 9.461597890293505e-09, "loss": 0.0106, "step": 22951 }, { "epoch": 9.05, "learning_rate": 9.453838979380214e-09, "loss": 0.01, "step": 22952 }, { "epoch": 9.05, "learning_rate": 9.446083174079755e-09, "loss": 0.0096, "step": 22953 }, { "epoch": 9.05, "learning_rate": 9.438330474518519e-09, "loss": 0.011, "step": 22954 }, { "epoch": 9.05, "learning_rate": 9.430580880822869e-09, "loss": 0.0118, "step": 22955 }, { "epoch": 9.05, "learning_rate": 9.422834393119127e-09, "loss": 0.0085, "step": 22956 }, { "epoch": 9.05, "learning_rate": 9.41509101153355e-09, "loss": 0.0087, "step": 22957 }, { "epoch": 9.05, "learning_rate": 9.40735073619232e-09, "loss": 0.009, "step": 22958 }, { "epoch": 9.05, "learning_rate": 9.399613567221609e-09, "loss": 0.0129, "step": 22959 }, { "epoch": 9.05, "learning_rate": 9.391879504747512e-09, "loss": 0.0111, "step": 22960 }, { "epoch": 9.05, "learning_rate": 9.384148548896131e-09, "loss": 0.0099, "step": 22961 }, { "epoch": 9.05, "learning_rate": 9.376420699793386e-09, "loss": 0.0099, "step": 22962 }, { "epoch": 9.05, "learning_rate": 9.368695957565264e-09, "loss": 0.0091, "step": 22963 }, { "epoch": 9.05, "learning_rate": 9.360974322337689e-09, "loss": 0.0112, "step": 22964 }, { "epoch": 9.05, "learning_rate": 9.353255794236492e-09, "loss": 0.0098, "step": 22965 }, { "epoch": 9.05, "learning_rate": 9.345540373387461e-09, "loss": 0.0089, "step": 22966 }, { "epoch": 9.05, "learning_rate": 9.337828059916386e-09, "loss": 0.0092, "step": 22967 }, { "epoch": 9.05, "learning_rate": 9.330118853948921e-09, "loss": 0.0133, "step": 22968 }, { "epoch": 9.05, "learning_rate": 9.322412755610742e-09, "loss": 0.0083, "step": 22969 }, { "epoch": 9.05, "learning_rate": 9.31470976502744e-09, "loss": 0.0105, "step": 22970 }, { "epoch": 9.05, "learning_rate": 9.307009882324557e-09, "loss": 0.0101, "step": 22971 }, { "epoch": 9.05, "learning_rate": 9.299313107627593e-09, "loss": 0.0093, "step": 22972 }, { "epoch": 9.05, "learning_rate": 9.291619441061983e-09, "loss": 0.0093, "step": 22973 }, { "epoch": 9.05, "learning_rate": 9.283928882753156e-09, "loss": 0.0121, "step": 22974 }, { "epoch": 9.05, "learning_rate": 9.276241432826415e-09, "loss": 0.0099, "step": 22975 }, { "epoch": 9.06, "learning_rate": 9.26855709140708e-09, "loss": 0.0092, "step": 22976 }, { "epoch": 9.06, "learning_rate": 9.260875858620365e-09, "loss": 0.0111, "step": 22977 }, { "epoch": 9.06, "learning_rate": 9.253197734591522e-09, "loss": 0.0091, "step": 22978 }, { "epoch": 9.06, "learning_rate": 9.24552271944563e-09, "loss": 0.0099, "step": 22979 }, { "epoch": 9.06, "learning_rate": 9.237850813307813e-09, "loss": 0.0087, "step": 22980 }, { "epoch": 9.06, "learning_rate": 9.230182016303079e-09, "loss": 0.0093, "step": 22981 }, { "epoch": 9.06, "learning_rate": 9.222516328556462e-09, "loss": 0.0117, "step": 22982 }, { "epoch": 9.06, "learning_rate": 9.214853750192886e-09, "loss": 0.0066, "step": 22983 }, { "epoch": 9.06, "learning_rate": 9.207194281337227e-09, "loss": 0.0103, "step": 22984 }, { "epoch": 9.06, "learning_rate": 9.19953792211432e-09, "loss": 0.0117, "step": 22985 }, { "epoch": 9.06, "learning_rate": 9.191884672648975e-09, "loss": 0.0104, "step": 22986 }, { "epoch": 9.06, "learning_rate": 9.184234533065938e-09, "loss": 0.0118, "step": 22987 }, { "epoch": 9.06, "learning_rate": 9.176587503489818e-09, "loss": 0.0105, "step": 22988 }, { "epoch": 9.06, "learning_rate": 9.168943584045341e-09, "loss": 0.0103, "step": 22989 }, { "epoch": 9.06, "learning_rate": 9.161302774857027e-09, "loss": 0.0115, "step": 22990 }, { "epoch": 9.06, "learning_rate": 9.153665076049466e-09, "loss": 0.0071, "step": 22991 }, { "epoch": 9.06, "learning_rate": 9.146030487747092e-09, "loss": 0.0083, "step": 22992 }, { "epoch": 9.06, "learning_rate": 9.138399010074382e-09, "loss": 0.0111, "step": 22993 }, { "epoch": 9.06, "learning_rate": 9.13077064315566e-09, "loss": 0.0111, "step": 22994 }, { "epoch": 9.06, "learning_rate": 9.123145387115294e-09, "loss": 0.011, "step": 22995 }, { "epoch": 9.06, "learning_rate": 9.11552324207756e-09, "loss": 0.0079, "step": 22996 }, { "epoch": 9.06, "learning_rate": 9.107904208166717e-09, "loss": 0.0099, "step": 22997 }, { "epoch": 9.06, "learning_rate": 9.100288285506862e-09, "loss": 0.0121, "step": 22998 }, { "epoch": 9.06, "learning_rate": 9.092675474222188e-09, "loss": 0.0082, "step": 22999 }, { "epoch": 9.06, "learning_rate": 9.085065774436773e-09, "loss": 0.0105, "step": 23000 }, { "epoch": 9.06, "learning_rate": 9.07745918627465e-09, "loss": 0.0099, "step": 23001 }, { "epoch": 9.07, "learning_rate": 9.069855709859741e-09, "loss": 0.0116, "step": 23002 }, { "epoch": 9.07, "learning_rate": 9.062255345315994e-09, "loss": 0.0135, "step": 23003 }, { "epoch": 9.07, "learning_rate": 9.054658092767309e-09, "loss": 0.0076, "step": 23004 }, { "epoch": 9.07, "learning_rate": 9.047063952337496e-09, "loss": 0.0114, "step": 23005 }, { "epoch": 9.07, "learning_rate": 9.039472924150326e-09, "loss": 0.0106, "step": 23006 }, { "epoch": 9.07, "learning_rate": 9.031885008329542e-09, "loss": 0.0091, "step": 23007 }, { "epoch": 9.07, "learning_rate": 9.024300204998758e-09, "loss": 0.0106, "step": 23008 }, { "epoch": 9.07, "learning_rate": 9.016718514281629e-09, "loss": 0.0117, "step": 23009 }, { "epoch": 9.07, "learning_rate": 9.009139936301769e-09, "loss": 0.0085, "step": 23010 }, { "epoch": 9.07, "learning_rate": 9.00156447118261e-09, "loss": 0.0103, "step": 23011 }, { "epoch": 9.07, "learning_rate": 8.993992119047677e-09, "loss": 0.0069, "step": 23012 }, { "epoch": 9.07, "learning_rate": 8.986422880020383e-09, "loss": 0.0087, "step": 23013 }, { "epoch": 9.07, "learning_rate": 8.978856754224117e-09, "loss": 0.0095, "step": 23014 }, { "epoch": 9.07, "learning_rate": 8.971293741782116e-09, "loss": 0.0089, "step": 23015 }, { "epoch": 9.07, "learning_rate": 8.963733842817745e-09, "loss": 0.0083, "step": 23016 }, { "epoch": 9.07, "learning_rate": 8.956177057454107e-09, "loss": 0.0085, "step": 23017 }, { "epoch": 9.07, "learning_rate": 8.948623385814503e-09, "loss": 0.0087, "step": 23018 }, { "epoch": 9.07, "learning_rate": 8.941072828021923e-09, "loss": 0.0106, "step": 23019 }, { "epoch": 9.07, "learning_rate": 8.933525384199514e-09, "loss": 0.0084, "step": 23020 }, { "epoch": 9.07, "learning_rate": 8.9259810544702e-09, "loss": 0.0107, "step": 23021 }, { "epoch": 9.07, "learning_rate": 8.918439838957059e-09, "loss": 0.0068, "step": 23022 }, { "epoch": 9.07, "learning_rate": 8.910901737782905e-09, "loss": 0.0127, "step": 23023 }, { "epoch": 9.07, "learning_rate": 8.90336675107064e-09, "loss": 0.0095, "step": 23024 }, { "epoch": 9.07, "learning_rate": 8.895834878943053e-09, "loss": 0.0105, "step": 23025 }, { "epoch": 9.07, "learning_rate": 8.888306121522915e-09, "loss": 0.0133, "step": 23026 }, { "epoch": 9.08, "learning_rate": 8.880780478932926e-09, "loss": 0.008, "step": 23027 }, { "epoch": 9.08, "learning_rate": 8.873257951295786e-09, "loss": 0.0124, "step": 23028 }, { "epoch": 9.08, "learning_rate": 8.865738538734025e-09, "loss": 0.0102, "step": 23029 }, { "epoch": 9.08, "learning_rate": 8.858222241370228e-09, "loss": 0.0129, "step": 23030 }, { "epoch": 9.08, "learning_rate": 8.850709059326945e-09, "loss": 0.0108, "step": 23031 }, { "epoch": 9.08, "learning_rate": 8.843198992726564e-09, "loss": 0.01, "step": 23032 }, { "epoch": 9.08, "learning_rate": 8.835692041691546e-09, "loss": 0.0103, "step": 23033 }, { "epoch": 9.08, "learning_rate": 8.828188206344168e-09, "loss": 0.0099, "step": 23034 }, { "epoch": 9.08, "learning_rate": 8.820687486806777e-09, "loss": 0.0085, "step": 23035 }, { "epoch": 9.08, "learning_rate": 8.813189883201633e-09, "loss": 0.0104, "step": 23036 }, { "epoch": 9.08, "learning_rate": 8.805695395650947e-09, "loss": 0.0104, "step": 23037 }, { "epoch": 9.08, "learning_rate": 8.7982040242768e-09, "loss": 0.0118, "step": 23038 }, { "epoch": 9.08, "learning_rate": 8.790715769201362e-09, "loss": 0.0087, "step": 23039 }, { "epoch": 9.08, "learning_rate": 8.783230630546623e-09, "loss": 0.0113, "step": 23040 }, { "epoch": 9.08, "learning_rate": 8.775748608434663e-09, "loss": 0.0098, "step": 23041 }, { "epoch": 9.08, "learning_rate": 8.76826970298734e-09, "loss": 0.0116, "step": 23042 }, { "epoch": 9.08, "learning_rate": 8.760793914326582e-09, "loss": 0.0097, "step": 23043 }, { "epoch": 9.08, "learning_rate": 8.753321242574219e-09, "loss": 0.0091, "step": 23044 }, { "epoch": 9.08, "learning_rate": 8.74585168785209e-09, "loss": 0.0109, "step": 23045 }, { "epoch": 9.08, "learning_rate": 8.738385250281876e-09, "loss": 0.0141, "step": 23046 }, { "epoch": 9.08, "learning_rate": 8.730921929985347e-09, "loss": 0.008, "step": 23047 }, { "epoch": 9.08, "learning_rate": 8.723461727084025e-09, "loss": 0.011, "step": 23048 }, { "epoch": 9.08, "learning_rate": 8.716004641699614e-09, "loss": 0.0093, "step": 23049 }, { "epoch": 9.08, "learning_rate": 8.708550673953596e-09, "loss": 0.0129, "step": 23050 }, { "epoch": 9.08, "learning_rate": 8.701099823967495e-09, "loss": 0.009, "step": 23051 }, { "epoch": 9.08, "learning_rate": 8.693652091862702e-09, "loss": 0.0098, "step": 23052 }, { "epoch": 9.09, "learning_rate": 8.686207477760631e-09, "loss": 0.0094, "step": 23053 }, { "epoch": 9.09, "learning_rate": 8.678765981782632e-09, "loss": 0.0079, "step": 23054 }, { "epoch": 9.09, "learning_rate": 8.671327604049961e-09, "loss": 0.0097, "step": 23055 }, { "epoch": 9.09, "learning_rate": 8.663892344683854e-09, "loss": 0.0104, "step": 23056 }, { "epoch": 9.09, "learning_rate": 8.656460203805505e-09, "loss": 0.0086, "step": 23057 }, { "epoch": 9.09, "learning_rate": 8.64903118153606e-09, "loss": 0.011, "step": 23058 }, { "epoch": 9.09, "learning_rate": 8.641605277996578e-09, "loss": 0.0094, "step": 23059 }, { "epoch": 9.09, "learning_rate": 8.634182493308117e-09, "loss": 0.0132, "step": 23060 }, { "epoch": 9.09, "learning_rate": 8.626762827591583e-09, "loss": 0.009, "step": 23061 }, { "epoch": 9.09, "learning_rate": 8.619346280968032e-09, "loss": 0.009, "step": 23062 }, { "epoch": 9.09, "learning_rate": 8.611932853558235e-09, "loss": 0.0098, "step": 23063 }, { "epoch": 9.09, "learning_rate": 8.604522545483074e-09, "loss": 0.0111, "step": 23064 }, { "epoch": 9.09, "learning_rate": 8.597115356863293e-09, "loss": 0.0086, "step": 23065 }, { "epoch": 9.09, "learning_rate": 8.589711287819623e-09, "loss": 0.0134, "step": 23066 }, { "epoch": 9.09, "learning_rate": 8.582310338472743e-09, "loss": 0.0113, "step": 23067 }, { "epoch": 9.09, "learning_rate": 8.574912508943332e-09, "loss": 0.011, "step": 23068 }, { "epoch": 9.09, "learning_rate": 8.567517799351875e-09, "loss": 0.0091, "step": 23069 }, { "epoch": 9.09, "learning_rate": 8.56012620981894e-09, "loss": 0.0098, "step": 23070 }, { "epoch": 9.09, "learning_rate": 8.552737740464989e-09, "loss": 0.0129, "step": 23071 }, { "epoch": 9.09, "learning_rate": 8.545352391410454e-09, "loss": 0.0086, "step": 23072 }, { "epoch": 9.09, "learning_rate": 8.537970162775687e-09, "loss": 0.0101, "step": 23073 }, { "epoch": 9.09, "learning_rate": 8.530591054681058e-09, "loss": 0.0113, "step": 23074 }, { "epoch": 9.09, "learning_rate": 8.523215067246758e-09, "loss": 0.0084, "step": 23075 }, { "epoch": 9.09, "learning_rate": 8.515842200593027e-09, "loss": 0.011, "step": 23076 }, { "epoch": 9.09, "learning_rate": 8.508472454840099e-09, "loss": 0.0091, "step": 23077 }, { "epoch": 9.1, "learning_rate": 8.501105830107991e-09, "loss": 0.0099, "step": 23078 }, { "epoch": 9.1, "learning_rate": 8.493742326516807e-09, "loss": 0.0099, "step": 23079 }, { "epoch": 9.1, "learning_rate": 8.486381944186582e-09, "loss": 0.0099, "step": 23080 }, { "epoch": 9.1, "learning_rate": 8.47902468323729e-09, "loss": 0.01, "step": 23081 }, { "epoch": 9.1, "learning_rate": 8.47167054378879e-09, "loss": 0.009, "step": 23082 }, { "epoch": 9.1, "learning_rate": 8.464319525960984e-09, "loss": 0.0072, "step": 23083 }, { "epoch": 9.1, "learning_rate": 8.456971629873621e-09, "loss": 0.0104, "step": 23084 }, { "epoch": 9.1, "learning_rate": 8.449626855646585e-09, "loss": 0.0089, "step": 23085 }, { "epoch": 9.1, "learning_rate": 8.442285203399446e-09, "loss": 0.0078, "step": 23086 }, { "epoch": 9.1, "learning_rate": 8.434946673251975e-09, "loss": 0.0089, "step": 23087 }, { "epoch": 9.1, "learning_rate": 8.427611265323675e-09, "loss": 0.012, "step": 23088 }, { "epoch": 9.1, "learning_rate": 8.420278979734207e-09, "loss": 0.0114, "step": 23089 }, { "epoch": 9.1, "learning_rate": 8.412949816602988e-09, "loss": 0.0087, "step": 23090 }, { "epoch": 9.1, "learning_rate": 8.405623776049541e-09, "loss": 0.0096, "step": 23091 }, { "epoch": 9.1, "learning_rate": 8.398300858193197e-09, "loss": 0.0125, "step": 23092 }, { "epoch": 9.1, "learning_rate": 8.390981063153368e-09, "loss": 0.0098, "step": 23093 }, { "epoch": 9.1, "learning_rate": 8.383664391049317e-09, "loss": 0.0109, "step": 23094 }, { "epoch": 9.1, "learning_rate": 8.376350842000346e-09, "loss": 0.0118, "step": 23095 }, { "epoch": 9.1, "learning_rate": 8.369040416125605e-09, "loss": 0.0093, "step": 23096 }, { "epoch": 9.1, "learning_rate": 8.361733113544245e-09, "loss": 0.0105, "step": 23097 }, { "epoch": 9.1, "learning_rate": 8.354428934375412e-09, "loss": 0.0092, "step": 23098 }, { "epoch": 9.1, "learning_rate": 8.347127878738103e-09, "loss": 0.0095, "step": 23099 }, { "epoch": 9.1, "learning_rate": 8.339829946751331e-09, "loss": 0.0111, "step": 23100 }, { "epoch": 9.1, "learning_rate": 8.332535138534025e-09, "loss": 0.0094, "step": 23101 }, { "epoch": 9.1, "learning_rate": 8.32524345420511e-09, "loss": 0.0095, "step": 23102 }, { "epoch": 9.11, "learning_rate": 8.317954893883406e-09, "loss": 0.0098, "step": 23103 }, { "epoch": 9.11, "learning_rate": 8.310669457687746e-09, "loss": 0.0116, "step": 23104 }, { "epoch": 9.11, "learning_rate": 8.303387145736817e-09, "loss": 0.0111, "step": 23105 }, { "epoch": 9.11, "learning_rate": 8.296107958149323e-09, "loss": 0.012, "step": 23106 }, { "epoch": 9.11, "learning_rate": 8.288831895043901e-09, "loss": 0.0092, "step": 23107 }, { "epoch": 9.11, "learning_rate": 8.281558956539192e-09, "loss": 0.0099, "step": 23108 }, { "epoch": 9.11, "learning_rate": 8.274289142753677e-09, "loss": 0.0074, "step": 23109 }, { "epoch": 9.11, "learning_rate": 8.267022453805838e-09, "loss": 0.0097, "step": 23110 }, { "epoch": 9.11, "learning_rate": 8.25975888981414e-09, "loss": 0.0107, "step": 23111 }, { "epoch": 9.11, "learning_rate": 8.252498450896994e-09, "loss": 0.0098, "step": 23112 }, { "epoch": 9.11, "learning_rate": 8.245241137172666e-09, "loss": 0.0141, "step": 23113 }, { "epoch": 9.11, "learning_rate": 8.237986948759501e-09, "loss": 0.0112, "step": 23114 }, { "epoch": 9.11, "learning_rate": 8.230735885775698e-09, "loss": 0.0089, "step": 23115 }, { "epoch": 9.11, "learning_rate": 8.223487948339424e-09, "loss": 0.011, "step": 23116 }, { "epoch": 9.11, "learning_rate": 8.216243136568835e-09, "loss": 0.0092, "step": 23117 }, { "epoch": 9.11, "learning_rate": 8.209001450582031e-09, "loss": 0.0082, "step": 23118 }, { "epoch": 9.11, "learning_rate": 8.201762890496989e-09, "loss": 0.0133, "step": 23119 }, { "epoch": 9.11, "learning_rate": 8.194527456431699e-09, "loss": 0.0115, "step": 23120 }, { "epoch": 9.11, "learning_rate": 8.187295148504136e-09, "loss": 0.0126, "step": 23121 }, { "epoch": 9.11, "learning_rate": 8.180065966832118e-09, "loss": 0.0097, "step": 23122 }, { "epoch": 9.11, "learning_rate": 8.172839911533525e-09, "loss": 0.0087, "step": 23123 }, { "epoch": 9.11, "learning_rate": 8.16561698272602e-09, "loss": 0.0096, "step": 23124 }, { "epoch": 9.11, "learning_rate": 8.158397180527487e-09, "loss": 0.0115, "step": 23125 }, { "epoch": 9.11, "learning_rate": 8.151180505055454e-09, "loss": 0.0096, "step": 23126 }, { "epoch": 9.11, "learning_rate": 8.14396695642765e-09, "loss": 0.0121, "step": 23127 }, { "epoch": 9.11, "learning_rate": 8.136756534761558e-09, "loss": 0.009, "step": 23128 }, { "epoch": 9.12, "learning_rate": 8.129549240174772e-09, "loss": 0.0104, "step": 23129 }, { "epoch": 9.12, "learning_rate": 8.122345072784687e-09, "loss": 0.016, "step": 23130 }, { "epoch": 9.12, "learning_rate": 8.115144032708809e-09, "loss": 0.0103, "step": 23131 }, { "epoch": 9.12, "learning_rate": 8.107946120064379e-09, "loss": 0.0113, "step": 23132 }, { "epoch": 9.12, "learning_rate": 8.100751334968836e-09, "loss": 0.0111, "step": 23133 }, { "epoch": 9.12, "learning_rate": 8.093559677539374e-09, "loss": 0.01, "step": 23134 }, { "epoch": 9.12, "learning_rate": 8.086371147893256e-09, "loss": 0.0109, "step": 23135 }, { "epoch": 9.12, "learning_rate": 8.079185746147588e-09, "loss": 0.0089, "step": 23136 }, { "epoch": 9.12, "learning_rate": 8.072003472419498e-09, "loss": 0.012, "step": 23137 }, { "epoch": 9.12, "learning_rate": 8.06482432682607e-09, "loss": 0.009, "step": 23138 }, { "epoch": 9.12, "learning_rate": 8.0576483094843e-09, "loss": 0.0109, "step": 23139 }, { "epoch": 9.12, "learning_rate": 8.050475420511117e-09, "loss": 0.0108, "step": 23140 }, { "epoch": 9.12, "learning_rate": 8.043305660023469e-09, "loss": 0.0092, "step": 23141 }, { "epoch": 9.12, "learning_rate": 8.036139028138222e-09, "loss": 0.0089, "step": 23142 }, { "epoch": 9.12, "learning_rate": 8.028975524972125e-09, "loss": 0.0116, "step": 23143 }, { "epoch": 9.12, "learning_rate": 8.021815150641997e-09, "loss": 0.0103, "step": 23144 }, { "epoch": 9.12, "learning_rate": 8.0146579052645e-09, "loss": 0.0092, "step": 23145 }, { "epoch": 9.12, "learning_rate": 8.00750378895627e-09, "loss": 0.0103, "step": 23146 }, { "epoch": 9.12, "learning_rate": 8.000352801833931e-09, "loss": 0.0086, "step": 23147 }, { "epoch": 9.12, "learning_rate": 7.993204944014075e-09, "loss": 0.0084, "step": 23148 }, { "epoch": 9.12, "learning_rate": 7.986060215613144e-09, "loss": 0.0081, "step": 23149 }, { "epoch": 9.12, "learning_rate": 7.978918616747599e-09, "loss": 0.0066, "step": 23150 }, { "epoch": 9.12, "learning_rate": 7.971780147533836e-09, "loss": 0.0114, "step": 23151 }, { "epoch": 9.12, "learning_rate": 7.964644808088228e-09, "loss": 0.0106, "step": 23152 }, { "epoch": 9.12, "learning_rate": 7.957512598527016e-09, "loss": 0.0105, "step": 23153 }, { "epoch": 9.13, "learning_rate": 7.95038351896653e-09, "loss": 0.0107, "step": 23154 }, { "epoch": 9.13, "learning_rate": 7.943257569522833e-09, "loss": 0.0108, "step": 23155 }, { "epoch": 9.13, "learning_rate": 7.936134750312206e-09, "loss": 0.0095, "step": 23156 }, { "epoch": 9.13, "learning_rate": 7.92901506145065e-09, "loss": 0.0094, "step": 23157 }, { "epoch": 9.13, "learning_rate": 7.921898503054269e-09, "loss": 0.0095, "step": 23158 }, { "epoch": 9.13, "learning_rate": 7.914785075238995e-09, "loss": 0.0109, "step": 23159 }, { "epoch": 9.13, "learning_rate": 7.907674778120776e-09, "loss": 0.0095, "step": 23160 }, { "epoch": 9.13, "learning_rate": 7.900567611815546e-09, "loss": 0.01, "step": 23161 }, { "epoch": 9.13, "learning_rate": 7.893463576439074e-09, "loss": 0.0086, "step": 23162 }, { "epoch": 9.13, "learning_rate": 7.886362672107184e-09, "loss": 0.0121, "step": 23163 }, { "epoch": 9.13, "learning_rate": 7.8792648989356e-09, "loss": 0.0137, "step": 23164 }, { "epoch": 9.13, "learning_rate": 7.872170257040035e-09, "loss": 0.012, "step": 23165 }, { "epoch": 9.13, "learning_rate": 7.86507874653608e-09, "loss": 0.0083, "step": 23166 }, { "epoch": 9.13, "learning_rate": 7.857990367539335e-09, "loss": 0.0111, "step": 23167 }, { "epoch": 9.13, "learning_rate": 7.850905120165285e-09, "loss": 0.0117, "step": 23168 }, { "epoch": 9.13, "learning_rate": 7.843823004529504e-09, "loss": 0.0102, "step": 23169 }, { "epoch": 9.13, "learning_rate": 7.836744020747321e-09, "loss": 0.0099, "step": 23170 }, { "epoch": 9.13, "learning_rate": 7.8296681689342e-09, "loss": 0.0096, "step": 23171 }, { "epoch": 9.13, "learning_rate": 7.82259544920536e-09, "loss": 0.0095, "step": 23172 }, { "epoch": 9.13, "learning_rate": 7.815525861676198e-09, "loss": 0.0085, "step": 23173 }, { "epoch": 9.13, "learning_rate": 7.808459406461843e-09, "loss": 0.0103, "step": 23174 }, { "epoch": 9.13, "learning_rate": 7.801396083677537e-09, "loss": 0.0086, "step": 23175 }, { "epoch": 9.13, "learning_rate": 7.79433589343832e-09, "loss": 0.0088, "step": 23176 }, { "epoch": 9.13, "learning_rate": 7.787278835859324e-09, "loss": 0.0093, "step": 23177 }, { "epoch": 9.13, "learning_rate": 7.780224911055566e-09, "loss": 0.0106, "step": 23178 }, { "epoch": 9.14, "learning_rate": 7.773174119142e-09, "loss": 0.0109, "step": 23179 }, { "epoch": 9.14, "learning_rate": 7.766126460233536e-09, "loss": 0.009, "step": 23180 }, { "epoch": 9.14, "learning_rate": 7.759081934445055e-09, "loss": 0.0137, "step": 23181 }, { "epoch": 9.14, "learning_rate": 7.752040541891381e-09, "loss": 0.0128, "step": 23182 }, { "epoch": 9.14, "learning_rate": 7.74500228268724e-09, "loss": 0.011, "step": 23183 }, { "epoch": 9.14, "learning_rate": 7.737967156947367e-09, "loss": 0.0098, "step": 23184 }, { "epoch": 9.14, "learning_rate": 7.730935164786446e-09, "loss": 0.0104, "step": 23185 }, { "epoch": 9.14, "learning_rate": 7.72390630631905e-09, "loss": 0.0109, "step": 23186 }, { "epoch": 9.14, "learning_rate": 7.716880581659757e-09, "loss": 0.0089, "step": 23187 }, { "epoch": 9.14, "learning_rate": 7.709857990923096e-09, "loss": 0.0086, "step": 23188 }, { "epoch": 9.14, "learning_rate": 7.702838534223466e-09, "loss": 0.0101, "step": 23189 }, { "epoch": 9.14, "learning_rate": 7.69582221167533e-09, "loss": 0.0105, "step": 23190 }, { "epoch": 9.14, "learning_rate": 7.688809023393017e-09, "loss": 0.0087, "step": 23191 }, { "epoch": 9.14, "learning_rate": 7.68179896949086e-09, "loss": 0.0087, "step": 23192 }, { "epoch": 9.14, "learning_rate": 7.674792050083057e-09, "loss": 0.0113, "step": 23193 }, { "epoch": 9.14, "learning_rate": 7.66778826528387e-09, "loss": 0.0099, "step": 23194 }, { "epoch": 9.14, "learning_rate": 7.660787615207365e-09, "loss": 0.0098, "step": 23195 }, { "epoch": 9.14, "learning_rate": 7.653790099967761e-09, "loss": 0.0087, "step": 23196 }, { "epoch": 9.14, "learning_rate": 7.646795719679034e-09, "loss": 0.0125, "step": 23197 }, { "epoch": 9.14, "learning_rate": 7.639804474455203e-09, "loss": 0.0103, "step": 23198 }, { "epoch": 9.14, "learning_rate": 7.632816364410154e-09, "loss": 0.0102, "step": 23199 }, { "epoch": 9.14, "learning_rate": 7.625831389657889e-09, "loss": 0.0101, "step": 23200 }, { "epoch": 9.14, "learning_rate": 7.618849550312179e-09, "loss": 0.0112, "step": 23201 }, { "epoch": 9.14, "learning_rate": 7.611870846486845e-09, "loss": 0.009, "step": 23202 }, { "epoch": 9.14, "learning_rate": 7.604895278295597e-09, "loss": 0.0119, "step": 23203 }, { "epoch": 9.14, "learning_rate": 7.597922845852166e-09, "loss": 0.0101, "step": 23204 }, { "epoch": 9.15, "learning_rate": 7.590953549270196e-09, "loss": 0.0092, "step": 23205 }, { "epoch": 9.15, "learning_rate": 7.583987388663216e-09, "loss": 0.0111, "step": 23206 }, { "epoch": 9.15, "learning_rate": 7.577024364144801e-09, "loss": 0.0082, "step": 23207 }, { "epoch": 9.15, "learning_rate": 7.570064475828464e-09, "loss": 0.0107, "step": 23208 }, { "epoch": 9.15, "learning_rate": 7.563107723827622e-09, "loss": 0.0096, "step": 23209 }, { "epoch": 9.15, "learning_rate": 7.556154108255631e-09, "loss": 0.0082, "step": 23210 }, { "epoch": 9.15, "learning_rate": 7.549203629225865e-09, "loss": 0.0127, "step": 23211 }, { "epoch": 9.15, "learning_rate": 7.542256286851545e-09, "loss": 0.0093, "step": 23212 }, { "epoch": 9.15, "learning_rate": 7.535312081246004e-09, "loss": 0.0085, "step": 23213 }, { "epoch": 9.15, "learning_rate": 7.528371012522327e-09, "loss": 0.0118, "step": 23214 }, { "epoch": 9.15, "learning_rate": 7.521433080793716e-09, "loss": 0.0101, "step": 23215 }, { "epoch": 9.15, "learning_rate": 7.514498286173188e-09, "loss": 0.0084, "step": 23216 }, { "epoch": 9.15, "learning_rate": 7.507566628773788e-09, "loss": 0.0101, "step": 23217 }, { "epoch": 9.15, "learning_rate": 7.50063810870849e-09, "loss": 0.0118, "step": 23218 }, { "epoch": 9.15, "learning_rate": 7.493712726090251e-09, "loss": 0.0097, "step": 23219 }, { "epoch": 9.15, "learning_rate": 7.48679048103189e-09, "loss": 0.01, "step": 23220 }, { "epoch": 9.15, "learning_rate": 7.479871373646274e-09, "loss": 0.0122, "step": 23221 }, { "epoch": 9.15, "learning_rate": 7.472955404046155e-09, "loss": 0.0138, "step": 23222 }, { "epoch": 9.15, "learning_rate": 7.466042572344266e-09, "loss": 0.0126, "step": 23223 }, { "epoch": 9.15, "learning_rate": 7.459132878653251e-09, "loss": 0.0094, "step": 23224 }, { "epoch": 9.15, "learning_rate": 7.452226323085776e-09, "loss": 0.0102, "step": 23225 }, { "epoch": 9.15, "learning_rate": 7.445322905754325e-09, "loss": 0.0104, "step": 23226 }, { "epoch": 9.15, "learning_rate": 7.438422626771479e-09, "loss": 0.0096, "step": 23227 }, { "epoch": 9.15, "learning_rate": 7.4315254862497015e-09, "loss": 0.0118, "step": 23228 }, { "epoch": 9.15, "learning_rate": 7.424631484301369e-09, "loss": 0.0117, "step": 23229 }, { "epoch": 9.16, "learning_rate": 7.417740621038859e-09, "loss": 0.0075, "step": 23230 }, { "epoch": 9.16, "learning_rate": 7.410852896574482e-09, "loss": 0.0104, "step": 23231 }, { "epoch": 9.16, "learning_rate": 7.403968311020525e-09, "loss": 0.0098, "step": 23232 }, { "epoch": 9.16, "learning_rate": 7.397086864489166e-09, "loss": 0.0097, "step": 23233 }, { "epoch": 9.16, "learning_rate": 7.3902085570925806e-09, "loss": 0.0095, "step": 23234 }, { "epoch": 9.16, "learning_rate": 7.383333388942814e-09, "loss": 0.0107, "step": 23235 }, { "epoch": 9.16, "learning_rate": 7.37646136015202e-09, "loss": 0.0116, "step": 23236 }, { "epoch": 9.16, "learning_rate": 7.3695924708321086e-09, "loss": 0.0127, "step": 23237 }, { "epoch": 9.16, "learning_rate": 7.362726721095125e-09, "loss": 0.01, "step": 23238 }, { "epoch": 9.16, "learning_rate": 7.355864111052867e-09, "loss": 0.0114, "step": 23239 }, { "epoch": 9.16, "learning_rate": 7.349004640817269e-09, "loss": 0.0088, "step": 23240 }, { "epoch": 9.16, "learning_rate": 7.342148310500085e-09, "loss": 0.0101, "step": 23241 }, { "epoch": 9.16, "learning_rate": 7.335295120213114e-09, "loss": 0.0106, "step": 23242 }, { "epoch": 9.16, "learning_rate": 7.328445070067979e-09, "loss": 0.0098, "step": 23243 }, { "epoch": 9.16, "learning_rate": 7.3215981601763674e-09, "loss": 0.0118, "step": 23244 }, { "epoch": 9.16, "learning_rate": 7.3147543906498795e-09, "loss": 0.0083, "step": 23245 }, { "epoch": 9.16, "learning_rate": 7.307913761600048e-09, "loss": 0.0106, "step": 23246 }, { "epoch": 9.16, "learning_rate": 7.3010762731383624e-09, "loss": 0.0078, "step": 23247 }, { "epoch": 9.16, "learning_rate": 7.2942419253762435e-09, "loss": 0.0126, "step": 23248 }, { "epoch": 9.16, "learning_rate": 7.287410718425135e-09, "loss": 0.0109, "step": 23249 }, { "epoch": 9.16, "learning_rate": 7.2805826523963275e-09, "loss": 0.0104, "step": 23250 }, { "epoch": 9.16, "learning_rate": 7.273757727401131e-09, "loss": 0.012, "step": 23251 }, { "epoch": 9.16, "learning_rate": 7.266935943550745e-09, "loss": 0.0107, "step": 23252 }, { "epoch": 9.16, "learning_rate": 7.260117300956415e-09, "loss": 0.011, "step": 23253 }, { "epoch": 9.16, "learning_rate": 7.253301799729228e-09, "loss": 0.0081, "step": 23254 }, { "epoch": 9.16, "learning_rate": 7.246489439980296e-09, "loss": 0.0088, "step": 23255 }, { "epoch": 9.17, "learning_rate": 7.239680221820621e-09, "loss": 0.0092, "step": 23256 }, { "epoch": 9.17, "learning_rate": 7.232874145361178e-09, "loss": 0.0095, "step": 23257 }, { "epoch": 9.17, "learning_rate": 7.226071210712925e-09, "loss": 0.0115, "step": 23258 }, { "epoch": 9.17, "learning_rate": 7.21927141798675e-09, "loss": 0.011, "step": 23259 }, { "epoch": 9.17, "learning_rate": 7.21247476729343e-09, "loss": 0.0098, "step": 23260 }, { "epoch": 9.17, "learning_rate": 7.20568125874379e-09, "loss": 0.0097, "step": 23261 }, { "epoch": 9.17, "learning_rate": 7.198890892448517e-09, "loss": 0.0108, "step": 23262 }, { "epoch": 9.17, "learning_rate": 7.192103668518323e-09, "loss": 0.0098, "step": 23263 }, { "epoch": 9.17, "learning_rate": 7.185319587063765e-09, "loss": 0.0095, "step": 23264 }, { "epoch": 9.17, "learning_rate": 7.178538648195509e-09, "loss": 0.0121, "step": 23265 }, { "epoch": 9.17, "learning_rate": 7.171760852023978e-09, "loss": 0.0116, "step": 23266 }, { "epoch": 9.17, "learning_rate": 7.164986198659683e-09, "loss": 0.0113, "step": 23267 }, { "epoch": 9.17, "learning_rate": 7.158214688213027e-09, "loss": 0.0131, "step": 23268 }, { "epoch": 9.17, "learning_rate": 7.151446320794407e-09, "loss": 0.0083, "step": 23269 }, { "epoch": 9.17, "learning_rate": 7.144681096514116e-09, "loss": 0.0136, "step": 23270 }, { "epoch": 9.17, "learning_rate": 7.137919015482396e-09, "loss": 0.0095, "step": 23271 }, { "epoch": 9.17, "learning_rate": 7.131160077809517e-09, "loss": 0.0098, "step": 23272 }, { "epoch": 9.17, "learning_rate": 7.124404283605567e-09, "loss": 0.0102, "step": 23273 }, { "epoch": 9.17, "learning_rate": 7.117651632980726e-09, "loss": 0.0124, "step": 23274 }, { "epoch": 9.17, "learning_rate": 7.110902126044949e-09, "loss": 0.0104, "step": 23275 }, { "epoch": 9.17, "learning_rate": 7.1041557629083705e-09, "loss": 0.0097, "step": 23276 }, { "epoch": 9.17, "learning_rate": 7.0974125436808585e-09, "loss": 0.0098, "step": 23277 }, { "epoch": 9.17, "learning_rate": 7.090672468472369e-09, "loss": 0.0095, "step": 23278 }, { "epoch": 9.17, "learning_rate": 7.083935537392682e-09, "loss": 0.0093, "step": 23279 }, { "epoch": 9.17, "learning_rate": 7.0772017505516865e-09, "loss": 0.012, "step": 23280 }, { "epoch": 9.18, "learning_rate": 7.070471108059095e-09, "loss": 0.0101, "step": 23281 }, { "epoch": 9.18, "learning_rate": 7.0637436100245975e-09, "loss": 0.0096, "step": 23282 }, { "epoch": 9.18, "learning_rate": 7.057019256557839e-09, "loss": 0.0113, "step": 23283 }, { "epoch": 9.18, "learning_rate": 7.050298047768444e-09, "loss": 0.0099, "step": 23284 }, { "epoch": 9.18, "learning_rate": 7.043579983765924e-09, "loss": 0.0112, "step": 23285 }, { "epoch": 9.18, "learning_rate": 7.036865064659814e-09, "loss": 0.0113, "step": 23286 }, { "epoch": 9.18, "learning_rate": 7.030153290559515e-09, "loss": 0.013, "step": 23287 }, { "epoch": 9.18, "learning_rate": 7.0234446615744295e-09, "loss": 0.0094, "step": 23288 }, { "epoch": 9.18, "learning_rate": 7.016739177813935e-09, "loss": 0.0117, "step": 23289 }, { "epoch": 9.18, "learning_rate": 7.010036839387257e-09, "loss": 0.0119, "step": 23290 }, { "epoch": 9.18, "learning_rate": 7.003337646403684e-09, "loss": 0.0093, "step": 23291 }, { "epoch": 9.18, "learning_rate": 6.996641598972375e-09, "loss": 0.0126, "step": 23292 }, { "epoch": 9.18, "learning_rate": 6.989948697202508e-09, "loss": 0.0094, "step": 23293 }, { "epoch": 9.18, "learning_rate": 6.983258941203107e-09, "loss": 0.0112, "step": 23294 }, { "epoch": 9.18, "learning_rate": 6.976572331083242e-09, "loss": 0.0084, "step": 23295 }, { "epoch": 9.18, "learning_rate": 6.9698888669518894e-09, "loss": 0.0082, "step": 23296 }, { "epoch": 9.18, "learning_rate": 6.9632085489179646e-09, "loss": 0.0112, "step": 23297 }, { "epoch": 9.18, "learning_rate": 6.956531377090358e-09, "loss": 0.0079, "step": 23298 }, { "epoch": 9.18, "learning_rate": 6.949857351577914e-09, "loss": 0.0097, "step": 23299 }, { "epoch": 9.18, "learning_rate": 6.943186472489393e-09, "loss": 0.0097, "step": 23300 }, { "epoch": 9.18, "learning_rate": 6.9365187399335056e-09, "loss": 0.0117, "step": 23301 }, { "epoch": 9.18, "learning_rate": 6.9298541540189435e-09, "loss": 0.011, "step": 23302 }, { "epoch": 9.18, "learning_rate": 6.923192714854331e-09, "loss": 0.0102, "step": 23303 }, { "epoch": 9.18, "learning_rate": 6.916534422548226e-09, "loss": 0.0087, "step": 23304 }, { "epoch": 9.18, "learning_rate": 6.9098792772091855e-09, "loss": 0.0104, "step": 23305 }, { "epoch": 9.19, "learning_rate": 6.903227278945589e-09, "loss": 0.0117, "step": 23306 }, { "epoch": 9.19, "learning_rate": 6.896578427865974e-09, "loss": 0.0128, "step": 23307 }, { "epoch": 9.19, "learning_rate": 6.8899327240786065e-09, "loss": 0.013, "step": 23308 }, { "epoch": 9.19, "learning_rate": 6.8832901676918905e-09, "loss": 0.0091, "step": 23309 }, { "epoch": 9.19, "learning_rate": 6.876650758814007e-09, "loss": 0.0091, "step": 23310 }, { "epoch": 9.19, "learning_rate": 6.870014497553201e-09, "loss": 0.0104, "step": 23311 }, { "epoch": 9.19, "learning_rate": 6.8633813840176306e-09, "loss": 0.0081, "step": 23312 }, { "epoch": 9.19, "learning_rate": 6.8567514183154314e-09, "loss": 0.0116, "step": 23313 }, { "epoch": 9.19, "learning_rate": 6.85012460055463e-09, "loss": 0.0103, "step": 23314 }, { "epoch": 9.19, "learning_rate": 6.843500930843227e-09, "loss": 0.0103, "step": 23315 }, { "epoch": 9.19, "learning_rate": 6.836880409289225e-09, "loss": 0.0095, "step": 23316 }, { "epoch": 9.19, "learning_rate": 6.830263036000494e-09, "loss": 0.0087, "step": 23317 }, { "epoch": 9.19, "learning_rate": 6.823648811084881e-09, "loss": 0.0077, "step": 23318 }, { "epoch": 9.19, "learning_rate": 6.817037734650166e-09, "loss": 0.0111, "step": 23319 }, { "epoch": 9.19, "learning_rate": 6.810429806804197e-09, "loss": 0.0103, "step": 23320 }, { "epoch": 9.19, "learning_rate": 6.803825027654553e-09, "loss": 0.0095, "step": 23321 }, { "epoch": 9.19, "learning_rate": 6.797223397308993e-09, "loss": 0.0083, "step": 23322 }, { "epoch": 9.19, "learning_rate": 6.790624915875009e-09, "loss": 0.0113, "step": 23323 }, { "epoch": 9.19, "learning_rate": 6.7840295834602044e-09, "loss": 0.0127, "step": 23324 }, { "epoch": 9.19, "learning_rate": 6.777437400172048e-09, "loss": 0.0085, "step": 23325 }, { "epoch": 9.19, "learning_rate": 6.7708483661180315e-09, "loss": 0.0101, "step": 23326 }, { "epoch": 9.19, "learning_rate": 6.764262481405491e-09, "loss": 0.0098, "step": 23327 }, { "epoch": 9.19, "learning_rate": 6.757679746141787e-09, "loss": 0.0119, "step": 23328 }, { "epoch": 9.19, "learning_rate": 6.7511001604342085e-09, "loss": 0.0107, "step": 23329 }, { "epoch": 9.19, "learning_rate": 6.744523724390028e-09, "loss": 0.011, "step": 23330 }, { "epoch": 9.19, "learning_rate": 6.737950438116357e-09, "loss": 0.0095, "step": 23331 }, { "epoch": 9.2, "learning_rate": 6.73138030172038e-09, "loss": 0.0115, "step": 23332 }, { "epoch": 9.2, "learning_rate": 6.724813315309186e-09, "loss": 0.0109, "step": 23333 }, { "epoch": 9.2, "learning_rate": 6.718249478989757e-09, "loss": 0.0105, "step": 23334 }, { "epoch": 9.2, "learning_rate": 6.711688792869119e-09, "loss": 0.0118, "step": 23335 }, { "epoch": 9.2, "learning_rate": 6.705131257054209e-09, "loss": 0.0109, "step": 23336 }, { "epoch": 9.2, "learning_rate": 6.698576871651873e-09, "loss": 0.0124, "step": 23337 }, { "epoch": 9.2, "learning_rate": 6.692025636768939e-09, "loss": 0.0104, "step": 23338 }, { "epoch": 9.2, "learning_rate": 6.685477552512209e-09, "loss": 0.0117, "step": 23339 }, { "epoch": 9.2, "learning_rate": 6.678932618988376e-09, "loss": 0.0111, "step": 23340 }, { "epoch": 9.2, "learning_rate": 6.672390836304131e-09, "loss": 0.0103, "step": 23341 }, { "epoch": 9.2, "learning_rate": 6.665852204566102e-09, "loss": 0.0106, "step": 23342 }, { "epoch": 9.2, "learning_rate": 6.659316723880848e-09, "loss": 0.0102, "step": 23343 }, { "epoch": 9.2, "learning_rate": 6.652784394354882e-09, "loss": 0.0121, "step": 23344 }, { "epoch": 9.2, "learning_rate": 6.646255216094676e-09, "loss": 0.0091, "step": 23345 }, { "epoch": 9.2, "learning_rate": 6.639729189206611e-09, "loss": 0.0101, "step": 23346 }, { "epoch": 9.2, "learning_rate": 6.633206313797157e-09, "loss": 0.011, "step": 23347 }, { "epoch": 9.2, "learning_rate": 6.626686589972497e-09, "loss": 0.0122, "step": 23348 }, { "epoch": 9.2, "learning_rate": 6.620170017838988e-09, "loss": 0.0095, "step": 23349 }, { "epoch": 9.2, "learning_rate": 6.61365659750277e-09, "loss": 0.0082, "step": 23350 }, { "epoch": 9.2, "learning_rate": 6.607146329070068e-09, "loss": 0.0111, "step": 23351 }, { "epoch": 9.2, "learning_rate": 6.600639212646952e-09, "loss": 0.0097, "step": 23352 }, { "epoch": 9.2, "learning_rate": 6.594135248339494e-09, "loss": 0.0115, "step": 23353 }, { "epoch": 9.2, "learning_rate": 6.587634436253653e-09, "loss": 0.0121, "step": 23354 }, { "epoch": 9.2, "learning_rate": 6.581136776495433e-09, "loss": 0.0097, "step": 23355 }, { "epoch": 9.2, "learning_rate": 6.574642269170749e-09, "loss": 0.0098, "step": 23356 }, { "epoch": 9.21, "learning_rate": 6.5681509143853845e-09, "loss": 0.0096, "step": 23357 }, { "epoch": 9.21, "learning_rate": 6.561662712245186e-09, "loss": 0.0097, "step": 23358 }, { "epoch": 9.21, "learning_rate": 6.555177662855893e-09, "loss": 0.0107, "step": 23359 }, { "epoch": 9.21, "learning_rate": 6.54869576632322e-09, "loss": 0.0072, "step": 23360 }, { "epoch": 9.21, "learning_rate": 6.542217022752794e-09, "loss": 0.0083, "step": 23361 }, { "epoch": 9.21, "learning_rate": 6.535741432250197e-09, "loss": 0.0104, "step": 23362 }, { "epoch": 9.21, "learning_rate": 6.529268994920989e-09, "loss": 0.0114, "step": 23363 }, { "epoch": 9.21, "learning_rate": 6.522799710870641e-09, "loss": 0.0113, "step": 23364 }, { "epoch": 9.21, "learning_rate": 6.5163335802046025e-09, "loss": 0.0093, "step": 23365 }, { "epoch": 9.21, "learning_rate": 6.5098706030283e-09, "loss": 0.0105, "step": 23366 }, { "epoch": 9.21, "learning_rate": 6.503410779447005e-09, "loss": 0.0088, "step": 23367 }, { "epoch": 9.21, "learning_rate": 6.496954109566033e-09, "loss": 0.0106, "step": 23368 }, { "epoch": 9.21, "learning_rate": 6.490500593490611e-09, "loss": 0.0079, "step": 23369 }, { "epoch": 9.21, "learning_rate": 6.484050231325966e-09, "loss": 0.0122, "step": 23370 }, { "epoch": 9.21, "learning_rate": 6.477603023177148e-09, "loss": 0.0086, "step": 23371 }, { "epoch": 9.21, "learning_rate": 6.4711589691493154e-09, "loss": 0.009, "step": 23372 }, { "epoch": 9.21, "learning_rate": 6.46471806934743e-09, "loss": 0.0131, "step": 23373 }, { "epoch": 9.21, "learning_rate": 6.458280323876542e-09, "loss": 0.0117, "step": 23374 }, { "epoch": 9.21, "learning_rate": 6.451845732841499e-09, "loss": 0.0084, "step": 23375 }, { "epoch": 9.21, "learning_rate": 6.4454142963472404e-09, "loss": 0.0092, "step": 23376 }, { "epoch": 9.21, "learning_rate": 6.438986014498527e-09, "loss": 0.0094, "step": 23377 }, { "epoch": 9.21, "learning_rate": 6.432560887400185e-09, "loss": 0.0089, "step": 23378 }, { "epoch": 9.21, "learning_rate": 6.426138915156909e-09, "loss": 0.0105, "step": 23379 }, { "epoch": 9.21, "learning_rate": 6.4197200978733714e-09, "loss": 0.0087, "step": 23380 }, { "epoch": 9.21, "learning_rate": 6.413304435654176e-09, "loss": 0.0112, "step": 23381 }, { "epoch": 9.22, "learning_rate": 6.406891928603908e-09, "loss": 0.0129, "step": 23382 }, { "epoch": 9.22, "learning_rate": 6.400482576827082e-09, "loss": 0.0113, "step": 23383 }, { "epoch": 9.22, "learning_rate": 6.394076380428149e-09, "loss": 0.0097, "step": 23384 }, { "epoch": 9.22, "learning_rate": 6.387673339511534e-09, "loss": 0.0108, "step": 23385 }, { "epoch": 9.22, "learning_rate": 6.381273454181535e-09, "loss": 0.011, "step": 23386 }, { "epoch": 9.22, "learning_rate": 6.374876724542577e-09, "loss": 0.0081, "step": 23387 }, { "epoch": 9.22, "learning_rate": 6.3684831506988e-09, "loss": 0.0098, "step": 23388 }, { "epoch": 9.22, "learning_rate": 6.362092732754521e-09, "loss": 0.0115, "step": 23389 }, { "epoch": 9.22, "learning_rate": 6.3557054708137654e-09, "loss": 0.0102, "step": 23390 }, { "epoch": 9.22, "learning_rate": 6.349321364980742e-09, "loss": 0.0082, "step": 23391 }, { "epoch": 9.22, "learning_rate": 6.342940415359477e-09, "loss": 0.009, "step": 23392 }, { "epoch": 9.22, "learning_rate": 6.336562622053954e-09, "loss": 0.0107, "step": 23393 }, { "epoch": 9.22, "learning_rate": 6.330187985168112e-09, "loss": 0.0104, "step": 23394 }, { "epoch": 9.22, "learning_rate": 6.323816504805845e-09, "loss": 0.0099, "step": 23395 }, { "epoch": 9.22, "learning_rate": 6.31744818107105e-09, "loss": 0.0116, "step": 23396 }, { "epoch": 9.22, "learning_rate": 6.311083014067486e-09, "loss": 0.01, "step": 23397 }, { "epoch": 9.22, "learning_rate": 6.304721003898872e-09, "loss": 0.011, "step": 23398 }, { "epoch": 9.22, "learning_rate": 6.298362150668924e-09, "loss": 0.0117, "step": 23399 }, { "epoch": 9.22, "learning_rate": 6.292006454481314e-09, "loss": 0.0102, "step": 23400 }, { "epoch": 9.22, "learning_rate": 6.285653915439559e-09, "loss": 0.0109, "step": 23401 }, { "epoch": 9.22, "learning_rate": 6.279304533647267e-09, "loss": 0.0105, "step": 23402 }, { "epoch": 9.22, "learning_rate": 6.272958309207843e-09, "loss": 0.0084, "step": 23403 }, { "epoch": 9.22, "learning_rate": 6.266615242224804e-09, "loss": 0.0099, "step": 23404 }, { "epoch": 9.22, "learning_rate": 6.260275332801468e-09, "loss": 0.0106, "step": 23405 }, { "epoch": 9.22, "learning_rate": 6.2539385810412185e-09, "loss": 0.0105, "step": 23406 }, { "epoch": 9.22, "learning_rate": 6.247604987047306e-09, "loss": 0.0095, "step": 23407 }, { "epoch": 9.23, "learning_rate": 6.241274550922959e-09, "loss": 0.0092, "step": 23408 }, { "epoch": 9.23, "learning_rate": 6.234947272771341e-09, "loss": 0.0084, "step": 23409 }, { "epoch": 9.23, "learning_rate": 6.228623152695633e-09, "loss": 0.0096, "step": 23410 }, { "epoch": 9.23, "learning_rate": 6.222302190798822e-09, "loss": 0.0088, "step": 23411 }, { "epoch": 9.23, "learning_rate": 6.215984387184004e-09, "loss": 0.0084, "step": 23412 }, { "epoch": 9.23, "learning_rate": 6.2096697419541156e-09, "loss": 0.0118, "step": 23413 }, { "epoch": 9.23, "learning_rate": 6.203358255212099e-09, "loss": 0.0111, "step": 23414 }, { "epoch": 9.23, "learning_rate": 6.197049927060782e-09, "loss": 0.013, "step": 23415 }, { "epoch": 9.23, "learning_rate": 6.190744757603039e-09, "loss": 0.011, "step": 23416 }, { "epoch": 9.23, "learning_rate": 6.1844427469415425e-09, "loss": 0.0092, "step": 23417 }, { "epoch": 9.23, "learning_rate": 6.178143895179122e-09, "loss": 0.0122, "step": 23418 }, { "epoch": 9.23, "learning_rate": 6.17184820241834e-09, "loss": 0.0137, "step": 23419 }, { "epoch": 9.23, "learning_rate": 6.16555566876189e-09, "loss": 0.0099, "step": 23420 }, { "epoch": 9.23, "learning_rate": 6.159266294312249e-09, "loss": 0.0124, "step": 23421 }, { "epoch": 9.23, "learning_rate": 6.152980079171977e-09, "loss": 0.0079, "step": 23422 }, { "epoch": 9.23, "learning_rate": 6.146697023443526e-09, "loss": 0.0101, "step": 23423 }, { "epoch": 9.23, "learning_rate": 6.140417127229281e-09, "loss": 0.0096, "step": 23424 }, { "epoch": 9.23, "learning_rate": 6.134140390631626e-09, "loss": 0.0104, "step": 23425 }, { "epoch": 9.23, "learning_rate": 6.127866813752791e-09, "loss": 0.0094, "step": 23426 }, { "epoch": 9.23, "learning_rate": 6.121596396695117e-09, "loss": 0.0102, "step": 23427 }, { "epoch": 9.23, "learning_rate": 6.115329139560743e-09, "loss": 0.011, "step": 23428 }, { "epoch": 9.23, "learning_rate": 6.109065042451833e-09, "loss": 0.0084, "step": 23429 }, { "epoch": 9.23, "learning_rate": 6.10280410547046e-09, "loss": 0.0075, "step": 23430 }, { "epoch": 9.23, "learning_rate": 6.096546328718743e-09, "loss": 0.0102, "step": 23431 }, { "epoch": 9.23, "learning_rate": 6.090291712298579e-09, "loss": 0.0112, "step": 23432 }, { "epoch": 9.24, "learning_rate": 6.084040256311995e-09, "loss": 0.0101, "step": 23433 }, { "epoch": 9.24, "learning_rate": 6.077791960860801e-09, "loss": 0.0092, "step": 23434 }, { "epoch": 9.24, "learning_rate": 6.0715468260468695e-09, "loss": 0.0101, "step": 23435 }, { "epoch": 9.24, "learning_rate": 6.065304851971987e-09, "loss": 0.01, "step": 23436 }, { "epoch": 9.24, "learning_rate": 6.059066038737915e-09, "loss": 0.0106, "step": 23437 }, { "epoch": 9.24, "learning_rate": 6.0528303864462836e-09, "loss": 0.0112, "step": 23438 }, { "epoch": 9.24, "learning_rate": 6.046597895198746e-09, "loss": 0.0098, "step": 23439 }, { "epoch": 9.24, "learning_rate": 6.04036856509691e-09, "loss": 0.0118, "step": 23440 }, { "epoch": 9.24, "learning_rate": 6.0341423962422925e-09, "loss": 0.0107, "step": 23441 }, { "epoch": 9.24, "learning_rate": 6.027919388736325e-09, "loss": 0.0126, "step": 23442 }, { "epoch": 9.24, "learning_rate": 6.021699542680503e-09, "loss": 0.0088, "step": 23443 }, { "epoch": 9.24, "learning_rate": 6.0154828581761465e-09, "loss": 0.0119, "step": 23444 }, { "epoch": 9.24, "learning_rate": 6.0092693353245736e-09, "loss": 0.0116, "step": 23445 }, { "epoch": 9.24, "learning_rate": 6.0030589742271265e-09, "loss": 0.0085, "step": 23446 }, { "epoch": 9.24, "learning_rate": 5.9968517749849455e-09, "loss": 0.0104, "step": 23447 }, { "epoch": 9.24, "learning_rate": 5.990647737699239e-09, "loss": 0.0091, "step": 23448 }, { "epoch": 9.24, "learning_rate": 5.9844468624711045e-09, "loss": 0.0097, "step": 23449 }, { "epoch": 9.24, "learning_rate": 5.978249149401637e-09, "loss": 0.0101, "step": 23450 }, { "epoch": 9.24, "learning_rate": 5.9720545985918024e-09, "loss": 0.0094, "step": 23451 }, { "epoch": 9.24, "learning_rate": 5.965863210142608e-09, "loss": 0.0096, "step": 23452 }, { "epoch": 9.24, "learning_rate": 5.959674984154928e-09, "loss": 0.0082, "step": 23453 }, { "epoch": 9.24, "learning_rate": 5.953489920729659e-09, "loss": 0.0091, "step": 23454 }, { "epoch": 9.24, "learning_rate": 5.9473080199675676e-09, "loss": 0.0131, "step": 23455 }, { "epoch": 9.24, "learning_rate": 5.941129281969459e-09, "loss": 0.0087, "step": 23456 }, { "epoch": 9.24, "learning_rate": 5.934953706835965e-09, "loss": 0.0117, "step": 23457 }, { "epoch": 9.24, "learning_rate": 5.928781294667829e-09, "loss": 0.0091, "step": 23458 }, { "epoch": 9.25, "learning_rate": 5.922612045565567e-09, "loss": 0.0095, "step": 23459 }, { "epoch": 9.25, "learning_rate": 5.91644595962979e-09, "loss": 0.0101, "step": 23460 }, { "epoch": 9.25, "learning_rate": 5.91028303696095e-09, "loss": 0.0135, "step": 23461 }, { "epoch": 9.25, "learning_rate": 5.904123277659523e-09, "loss": 0.0112, "step": 23462 }, { "epoch": 9.25, "learning_rate": 5.897966681825894e-09, "loss": 0.0094, "step": 23463 }, { "epoch": 9.25, "learning_rate": 5.891813249560406e-09, "loss": 0.0123, "step": 23464 }, { "epoch": 9.25, "learning_rate": 5.885662980963357e-09, "loss": 0.0101, "step": 23465 }, { "epoch": 9.25, "learning_rate": 5.879515876134977e-09, "loss": 0.0094, "step": 23466 }, { "epoch": 9.25, "learning_rate": 5.873371935175475e-09, "loss": 0.0111, "step": 23467 }, { "epoch": 9.25, "learning_rate": 5.867231158184949e-09, "loss": 0.0089, "step": 23468 }, { "epoch": 9.25, "learning_rate": 5.861093545263518e-09, "loss": 0.0125, "step": 23469 }, { "epoch": 9.25, "learning_rate": 5.8549590965111916e-09, "loss": 0.0104, "step": 23470 }, { "epoch": 9.25, "learning_rate": 5.848827812027979e-09, "loss": 0.0097, "step": 23471 }, { "epoch": 9.25, "learning_rate": 5.842699691913799e-09, "loss": 0.0082, "step": 23472 }, { "epoch": 9.25, "learning_rate": 5.836574736268529e-09, "loss": 0.0099, "step": 23473 }, { "epoch": 9.25, "learning_rate": 5.830452945191999e-09, "loss": 0.0107, "step": 23474 }, { "epoch": 9.25, "learning_rate": 5.824334318783952e-09, "loss": 0.0093, "step": 23475 }, { "epoch": 9.25, "learning_rate": 5.818218857144175e-09, "loss": 0.0092, "step": 23476 }, { "epoch": 9.25, "learning_rate": 5.812106560372298e-09, "loss": 0.0097, "step": 23477 }, { "epoch": 9.25, "learning_rate": 5.805997428567955e-09, "loss": 0.0107, "step": 23478 }, { "epoch": 9.25, "learning_rate": 5.7998914618307085e-09, "loss": 0.0102, "step": 23479 }, { "epoch": 9.25, "learning_rate": 5.7937886602600796e-09, "loss": 0.0097, "step": 23480 }, { "epoch": 9.25, "learning_rate": 5.787689023955544e-09, "loss": 0.0107, "step": 23481 }, { "epoch": 9.25, "learning_rate": 5.781592553016512e-09, "loss": 0.0093, "step": 23482 }, { "epoch": 9.25, "learning_rate": 5.775499247542348e-09, "loss": 0.0112, "step": 23483 }, { "epoch": 9.26, "learning_rate": 5.7694091076323285e-09, "loss": 0.0126, "step": 23484 }, { "epoch": 9.26, "learning_rate": 5.763322133385773e-09, "loss": 0.0089, "step": 23485 }, { "epoch": 9.26, "learning_rate": 5.757238324901848e-09, "loss": 0.0104, "step": 23486 }, { "epoch": 9.26, "learning_rate": 5.751157682279739e-09, "loss": 0.0098, "step": 23487 }, { "epoch": 9.26, "learning_rate": 5.745080205618502e-09, "loss": 0.0097, "step": 23488 }, { "epoch": 9.26, "learning_rate": 5.739005895017257e-09, "loss": 0.0072, "step": 23489 }, { "epoch": 9.26, "learning_rate": 5.732934750574969e-09, "loss": 0.0099, "step": 23490 }, { "epoch": 9.26, "learning_rate": 5.726866772390582e-09, "loss": 0.0089, "step": 23491 }, { "epoch": 9.26, "learning_rate": 5.720801960562993e-09, "loss": 0.0084, "step": 23492 }, { "epoch": 9.26, "learning_rate": 5.71474031519108e-09, "loss": 0.0091, "step": 23493 }, { "epoch": 9.26, "learning_rate": 5.70868183637363e-09, "loss": 0.0104, "step": 23494 }, { "epoch": 9.26, "learning_rate": 5.702626524209364e-09, "loss": 0.0094, "step": 23495 }, { "epoch": 9.26, "learning_rate": 5.696574378797003e-09, "loss": 0.0101, "step": 23496 }, { "epoch": 9.26, "learning_rate": 5.690525400235113e-09, "loss": 0.0088, "step": 23497 }, { "epoch": 9.26, "learning_rate": 5.684479588622415e-09, "loss": 0.0097, "step": 23498 }, { "epoch": 9.26, "learning_rate": 5.678436944057341e-09, "loss": 0.0087, "step": 23499 }, { "epoch": 9.26, "learning_rate": 5.672397466638435e-09, "loss": 0.0092, "step": 23500 }, { "epoch": 9.26, "learning_rate": 5.666361156464061e-09, "loss": 0.0097, "step": 23501 }, { "epoch": 9.26, "learning_rate": 5.660328013632676e-09, "loss": 0.0099, "step": 23502 }, { "epoch": 9.26, "learning_rate": 5.654298038242578e-09, "loss": 0.0082, "step": 23503 }, { "epoch": 9.26, "learning_rate": 5.648271230392066e-09, "loss": 0.0093, "step": 23504 }, { "epoch": 9.26, "learning_rate": 5.642247590179328e-09, "loss": 0.011, "step": 23505 }, { "epoch": 9.26, "learning_rate": 5.6362271177025746e-09, "loss": 0.0102, "step": 23506 }, { "epoch": 9.26, "learning_rate": 5.6302098130599275e-09, "loss": 0.0117, "step": 23507 }, { "epoch": 9.26, "learning_rate": 5.624195676349441e-09, "loss": 0.0111, "step": 23508 }, { "epoch": 9.27, "learning_rate": 5.618184707669149e-09, "loss": 0.0102, "step": 23509 }, { "epoch": 9.27, "learning_rate": 5.612176907117017e-09, "loss": 0.0099, "step": 23510 }, { "epoch": 9.27, "learning_rate": 5.606172274790988e-09, "loss": 0.0067, "step": 23511 }, { "epoch": 9.27, "learning_rate": 5.600170810788896e-09, "loss": 0.0128, "step": 23512 }, { "epoch": 9.27, "learning_rate": 5.5941725152085724e-09, "loss": 0.0124, "step": 23513 }, { "epoch": 9.27, "learning_rate": 5.588177388147741e-09, "loss": 0.0111, "step": 23514 }, { "epoch": 9.27, "learning_rate": 5.582185429704167e-09, "loss": 0.0094, "step": 23515 }, { "epoch": 9.27, "learning_rate": 5.576196639975484e-09, "loss": 0.0094, "step": 23516 }, { "epoch": 9.27, "learning_rate": 5.570211019059323e-09, "loss": 0.0125, "step": 23517 }, { "epoch": 9.27, "learning_rate": 5.564228567053209e-09, "loss": 0.0095, "step": 23518 }, { "epoch": 9.27, "learning_rate": 5.558249284054661e-09, "loss": 0.009, "step": 23519 }, { "epoch": 9.27, "learning_rate": 5.552273170161137e-09, "loss": 0.0081, "step": 23520 }, { "epoch": 9.27, "learning_rate": 5.546300225470046e-09, "loss": 0.0085, "step": 23521 }, { "epoch": 9.27, "learning_rate": 5.5403304500786895e-09, "loss": 0.0082, "step": 23522 }, { "epoch": 9.27, "learning_rate": 5.534363844084455e-09, "loss": 0.0101, "step": 23523 }, { "epoch": 9.27, "learning_rate": 5.5284004075844655e-09, "loss": 0.0106, "step": 23524 }, { "epoch": 9.27, "learning_rate": 5.5224401406760436e-09, "loss": 0.01, "step": 23525 }, { "epoch": 9.27, "learning_rate": 5.516483043456243e-09, "loss": 0.01, "step": 23526 }, { "epoch": 9.27, "learning_rate": 5.510529116022233e-09, "loss": 0.0084, "step": 23527 }, { "epoch": 9.27, "learning_rate": 5.504578358470957e-09, "loss": 0.0129, "step": 23528 }, { "epoch": 9.27, "learning_rate": 5.498630770899471e-09, "loss": 0.0116, "step": 23529 }, { "epoch": 9.27, "learning_rate": 5.4926863534047185e-09, "loss": 0.0099, "step": 23530 }, { "epoch": 9.27, "learning_rate": 5.486745106083535e-09, "loss": 0.0076, "step": 23531 }, { "epoch": 9.27, "learning_rate": 5.480807029032797e-09, "loss": 0.0094, "step": 23532 }, { "epoch": 9.27, "learning_rate": 5.47487212234925e-09, "loss": 0.0125, "step": 23533 }, { "epoch": 9.27, "learning_rate": 5.468940386129661e-09, "loss": 0.0136, "step": 23534 }, { "epoch": 9.28, "learning_rate": 5.4630118204706864e-09, "loss": 0.0108, "step": 23535 }, { "epoch": 9.28, "learning_rate": 5.45708642546896e-09, "loss": 0.01, "step": 23536 }, { "epoch": 9.28, "learning_rate": 5.451164201221026e-09, "loss": 0.0076, "step": 23537 }, { "epoch": 9.28, "learning_rate": 5.445245147823474e-09, "loss": 0.009, "step": 23538 }, { "epoch": 9.28, "learning_rate": 5.4393292653727164e-09, "loss": 0.012, "step": 23539 }, { "epoch": 9.28, "learning_rate": 5.433416553965209e-09, "loss": 0.0104, "step": 23540 }, { "epoch": 9.28, "learning_rate": 5.4275070136972745e-09, "loss": 0.0098, "step": 23541 }, { "epoch": 9.28, "learning_rate": 5.42160064466528e-09, "loss": 0.0099, "step": 23542 }, { "epoch": 9.28, "learning_rate": 5.415697446965484e-09, "loss": 0.0095, "step": 23543 }, { "epoch": 9.28, "learning_rate": 5.409797420694073e-09, "loss": 0.0122, "step": 23544 }, { "epoch": 9.28, "learning_rate": 5.403900565947217e-09, "loss": 0.0073, "step": 23545 }, { "epoch": 9.28, "learning_rate": 5.398006882821016e-09, "loss": 0.0082, "step": 23546 }, { "epoch": 9.28, "learning_rate": 5.392116371411548e-09, "loss": 0.0135, "step": 23547 }, { "epoch": 9.28, "learning_rate": 5.386229031814848e-09, "loss": 0.0105, "step": 23548 }, { "epoch": 9.28, "learning_rate": 5.380344864126796e-09, "loss": 0.0071, "step": 23549 }, { "epoch": 9.28, "learning_rate": 5.3744638684433356e-09, "loss": 0.0081, "step": 23550 }, { "epoch": 9.28, "learning_rate": 5.368586044860346e-09, "loss": 0.0105, "step": 23551 }, { "epoch": 9.28, "learning_rate": 5.362711393473596e-09, "loss": 0.0094, "step": 23552 }, { "epoch": 9.28, "learning_rate": 5.35683991437883e-09, "loss": 0.0091, "step": 23553 }, { "epoch": 9.28, "learning_rate": 5.350971607671728e-09, "loss": 0.0099, "step": 23554 }, { "epoch": 9.28, "learning_rate": 5.345106473447969e-09, "loss": 0.0103, "step": 23555 }, { "epoch": 9.28, "learning_rate": 5.3392445118031204e-09, "loss": 0.0116, "step": 23556 }, { "epoch": 9.28, "learning_rate": 5.3333857228327725e-09, "loss": 0.0084, "step": 23557 }, { "epoch": 9.28, "learning_rate": 5.327530106632361e-09, "loss": 0.0099, "step": 23558 }, { "epoch": 9.28, "learning_rate": 5.321677663297319e-09, "loss": 0.0091, "step": 23559 }, { "epoch": 9.29, "learning_rate": 5.3158283929230606e-09, "loss": 0.0082, "step": 23560 }, { "epoch": 9.29, "learning_rate": 5.309982295604953e-09, "loss": 0.0117, "step": 23561 }, { "epoch": 9.29, "learning_rate": 5.304139371438188e-09, "loss": 0.011, "step": 23562 }, { "epoch": 9.29, "learning_rate": 5.298299620518088e-09, "loss": 0.0097, "step": 23563 }, { "epoch": 9.29, "learning_rate": 5.2924630429397544e-09, "loss": 0.0091, "step": 23564 }, { "epoch": 9.29, "learning_rate": 5.286629638798379e-09, "loss": 0.0093, "step": 23565 }, { "epoch": 9.29, "learning_rate": 5.280799408188996e-09, "loss": 0.012, "step": 23566 }, { "epoch": 9.29, "learning_rate": 5.274972351206663e-09, "loss": 0.0099, "step": 23567 }, { "epoch": 9.29, "learning_rate": 5.269148467946305e-09, "loss": 0.0101, "step": 23568 }, { "epoch": 9.29, "learning_rate": 5.263327758502889e-09, "loss": 0.0107, "step": 23569 }, { "epoch": 9.29, "learning_rate": 5.257510222971251e-09, "loss": 0.0105, "step": 23570 }, { "epoch": 9.29, "learning_rate": 5.2516958614462485e-09, "loss": 0.0085, "step": 23571 }, { "epoch": 9.29, "learning_rate": 5.2458846740226046e-09, "loss": 0.0108, "step": 23572 }, { "epoch": 9.29, "learning_rate": 5.240076660795045e-09, "loss": 0.0109, "step": 23573 }, { "epoch": 9.29, "learning_rate": 5.234271821858249e-09, "loss": 0.0104, "step": 23574 }, { "epoch": 9.29, "learning_rate": 5.228470157306786e-09, "loss": 0.0101, "step": 23575 }, { "epoch": 9.29, "learning_rate": 5.222671667235268e-09, "loss": 0.0095, "step": 23576 }, { "epoch": 9.29, "learning_rate": 5.216876351738131e-09, "loss": 0.0134, "step": 23577 }, { "epoch": 9.29, "learning_rate": 5.2110842109099215e-09, "loss": 0.01, "step": 23578 }, { "epoch": 9.29, "learning_rate": 5.205295244844987e-09, "loss": 0.0115, "step": 23579 }, { "epoch": 9.29, "learning_rate": 5.199509453637674e-09, "loss": 0.0111, "step": 23580 }, { "epoch": 9.29, "learning_rate": 5.193726837382284e-09, "loss": 0.0095, "step": 23581 }, { "epoch": 9.29, "learning_rate": 5.187947396173098e-09, "loss": 0.0078, "step": 23582 }, { "epoch": 9.29, "learning_rate": 5.182171130104263e-09, "loss": 0.0104, "step": 23583 }, { "epoch": 9.29, "learning_rate": 5.176398039269992e-09, "loss": 0.0102, "step": 23584 }, { "epoch": 9.3, "learning_rate": 5.170628123764298e-09, "loss": 0.0123, "step": 23585 }, { "epoch": 9.3, "learning_rate": 5.1648613836812625e-09, "loss": 0.0087, "step": 23586 }, { "epoch": 9.3, "learning_rate": 5.159097819114877e-09, "loss": 0.0107, "step": 23587 }, { "epoch": 9.3, "learning_rate": 5.1533374301590886e-09, "loss": 0.0069, "step": 23588 }, { "epoch": 9.3, "learning_rate": 5.147580216907732e-09, "loss": 0.0099, "step": 23589 }, { "epoch": 9.3, "learning_rate": 5.141826179454689e-09, "loss": 0.0116, "step": 23590 }, { "epoch": 9.3, "learning_rate": 5.136075317893751e-09, "loss": 0.0074, "step": 23591 }, { "epoch": 9.3, "learning_rate": 5.13032763231862e-09, "loss": 0.009, "step": 23592 }, { "epoch": 9.3, "learning_rate": 5.124583122822956e-09, "loss": 0.0093, "step": 23593 }, { "epoch": 9.3, "learning_rate": 5.118841789500439e-09, "loss": 0.0099, "step": 23594 }, { "epoch": 9.3, "learning_rate": 5.113103632444615e-09, "loss": 0.0082, "step": 23595 }, { "epoch": 9.3, "learning_rate": 5.107368651748989e-09, "loss": 0.0099, "step": 23596 }, { "epoch": 9.3, "learning_rate": 5.101636847507063e-09, "loss": 0.01, "step": 23597 }, { "epoch": 9.3, "learning_rate": 5.095908219812251e-09, "loss": 0.007, "step": 23598 }, { "epoch": 9.3, "learning_rate": 5.0901827687579e-09, "loss": 0.0095, "step": 23599 }, { "epoch": 9.3, "learning_rate": 5.084460494437337e-09, "loss": 0.0126, "step": 23600 }, { "epoch": 9.3, "learning_rate": 5.078741396943842e-09, "loss": 0.0102, "step": 23601 }, { "epoch": 9.3, "learning_rate": 5.073025476370607e-09, "loss": 0.0118, "step": 23602 }, { "epoch": 9.3, "learning_rate": 5.067312732810824e-09, "loss": 0.0107, "step": 23603 }, { "epoch": 9.3, "learning_rate": 5.061603166357553e-09, "loss": 0.0096, "step": 23604 }, { "epoch": 9.3, "learning_rate": 5.055896777103896e-09, "loss": 0.0104, "step": 23605 }, { "epoch": 9.3, "learning_rate": 5.050193565142824e-09, "loss": 0.0116, "step": 23606 }, { "epoch": 9.3, "learning_rate": 5.0444935305673285e-09, "loss": 0.0116, "step": 23607 }, { "epoch": 9.3, "learning_rate": 5.038796673470269e-09, "loss": 0.0099, "step": 23608 }, { "epoch": 9.3, "learning_rate": 5.033102993944527e-09, "loss": 0.0127, "step": 23609 }, { "epoch": 9.3, "learning_rate": 5.027412492082894e-09, "loss": 0.0117, "step": 23610 }, { "epoch": 9.31, "learning_rate": 5.021725167978119e-09, "loss": 0.0097, "step": 23611 }, { "epoch": 9.31, "learning_rate": 5.016041021722884e-09, "loss": 0.0094, "step": 23612 }, { "epoch": 9.31, "learning_rate": 5.0103600534098234e-09, "loss": 0.0104, "step": 23613 }, { "epoch": 9.31, "learning_rate": 5.004682263131555e-09, "loss": 0.0091, "step": 23614 }, { "epoch": 9.31, "learning_rate": 4.999007650980625e-09, "loss": 0.0118, "step": 23615 }, { "epoch": 9.31, "learning_rate": 4.9933362170494705e-09, "loss": 0.0094, "step": 23616 }, { "epoch": 9.31, "learning_rate": 4.987667961430575e-09, "loss": 0.0093, "step": 23617 }, { "epoch": 9.31, "learning_rate": 4.98200288421633e-09, "loss": 0.0144, "step": 23618 }, { "epoch": 9.31, "learning_rate": 4.9763409854990165e-09, "loss": 0.01, "step": 23619 }, { "epoch": 9.31, "learning_rate": 4.970682265370962e-09, "loss": 0.0094, "step": 23620 }, { "epoch": 9.31, "learning_rate": 4.965026723924337e-09, "loss": 0.0087, "step": 23621 }, { "epoch": 9.31, "learning_rate": 4.9593743612514e-09, "loss": 0.0112, "step": 23622 }, { "epoch": 9.31, "learning_rate": 4.9537251774442125e-09, "loss": 0.0132, "step": 23623 }, { "epoch": 9.31, "learning_rate": 4.948079172594877e-09, "loss": 0.0085, "step": 23624 }, { "epoch": 9.31, "learning_rate": 4.942436346795409e-09, "loss": 0.0079, "step": 23625 }, { "epoch": 9.31, "learning_rate": 4.936796700137757e-09, "loss": 0.0098, "step": 23626 }, { "epoch": 9.31, "learning_rate": 4.931160232713871e-09, "loss": 0.0107, "step": 23627 }, { "epoch": 9.31, "learning_rate": 4.92552694461561e-09, "loss": 0.0122, "step": 23628 }, { "epoch": 9.31, "learning_rate": 4.919896835934767e-09, "loss": 0.0087, "step": 23629 }, { "epoch": 9.31, "learning_rate": 4.914269906763113e-09, "loss": 0.0084, "step": 23630 }, { "epoch": 9.31, "learning_rate": 4.908646157192375e-09, "loss": 0.0098, "step": 23631 }, { "epoch": 9.31, "learning_rate": 4.903025587314213e-09, "loss": 0.0079, "step": 23632 }, { "epoch": 9.31, "learning_rate": 4.8974081972201985e-09, "loss": 0.0119, "step": 23633 }, { "epoch": 9.31, "learning_rate": 4.891793987001924e-09, "loss": 0.0105, "step": 23634 }, { "epoch": 9.31, "learning_rate": 4.886182956750873e-09, "loss": 0.0091, "step": 23635 }, { "epoch": 9.32, "learning_rate": 4.880575106558482e-09, "loss": 0.0093, "step": 23636 }, { "epoch": 9.32, "learning_rate": 4.87497043651619e-09, "loss": 0.0115, "step": 23637 }, { "epoch": 9.32, "learning_rate": 4.869368946715324e-09, "loss": 0.0107, "step": 23638 }, { "epoch": 9.32, "learning_rate": 4.8637706372471665e-09, "loss": 0.009, "step": 23639 }, { "epoch": 9.32, "learning_rate": 4.858175508202977e-09, "loss": 0.0117, "step": 23640 }, { "epoch": 9.32, "learning_rate": 4.8525835596739725e-09, "loss": 0.01, "step": 23641 }, { "epoch": 9.32, "learning_rate": 4.846994791751235e-09, "loss": 0.0144, "step": 23642 }, { "epoch": 9.32, "learning_rate": 4.8414092045259145e-09, "loss": 0.0099, "step": 23643 }, { "epoch": 9.32, "learning_rate": 4.83582679808896e-09, "loss": 0.0097, "step": 23644 }, { "epoch": 9.32, "learning_rate": 4.830247572531454e-09, "loss": 0.0101, "step": 23645 }, { "epoch": 9.32, "learning_rate": 4.824671527944279e-09, "loss": 0.0099, "step": 23646 }, { "epoch": 9.32, "learning_rate": 4.819098664418364e-09, "loss": 0.0087, "step": 23647 }, { "epoch": 9.32, "learning_rate": 4.813528982044435e-09, "loss": 0.0102, "step": 23648 }, { "epoch": 9.32, "learning_rate": 4.807962480913374e-09, "loss": 0.0082, "step": 23649 }, { "epoch": 9.32, "learning_rate": 4.802399161115867e-09, "loss": 0.0104, "step": 23650 }, { "epoch": 9.32, "learning_rate": 4.7968390227425714e-09, "loss": 0.0128, "step": 23651 }, { "epoch": 9.32, "learning_rate": 4.791282065884128e-09, "loss": 0.0115, "step": 23652 }, { "epoch": 9.32, "learning_rate": 4.7857282906311084e-09, "loss": 0.0125, "step": 23653 }, { "epoch": 9.32, "learning_rate": 4.780177697074017e-09, "loss": 0.0088, "step": 23654 }, { "epoch": 9.32, "learning_rate": 4.774630285303361e-09, "loss": 0.012, "step": 23655 }, { "epoch": 9.32, "learning_rate": 4.769086055409488e-09, "loss": 0.0109, "step": 23656 }, { "epoch": 9.32, "learning_rate": 4.763545007482817e-09, "loss": 0.0098, "step": 23657 }, { "epoch": 9.32, "learning_rate": 4.7580071416136295e-09, "loss": 0.0123, "step": 23658 }, { "epoch": 9.32, "learning_rate": 4.7524724578921875e-09, "loss": 0.01, "step": 23659 }, { "epoch": 9.32, "learning_rate": 4.746940956408707e-09, "loss": 0.0091, "step": 23660 }, { "epoch": 9.32, "learning_rate": 4.7414126372533396e-09, "loss": 0.0104, "step": 23661 }, { "epoch": 9.33, "learning_rate": 4.735887500516189e-09, "loss": 0.0096, "step": 23662 }, { "epoch": 9.33, "learning_rate": 4.730365546287296e-09, "loss": 0.0084, "step": 23663 }, { "epoch": 9.33, "learning_rate": 4.724846774656677e-09, "loss": 0.0063, "step": 23664 }, { "epoch": 9.33, "learning_rate": 4.719331185714281e-09, "loss": 0.0097, "step": 23665 }, { "epoch": 9.33, "learning_rate": 4.713818779549972e-09, "loss": 0.01, "step": 23666 }, { "epoch": 9.33, "learning_rate": 4.708309556253609e-09, "loss": 0.0092, "step": 23667 }, { "epoch": 9.33, "learning_rate": 4.702803515915033e-09, "loss": 0.0094, "step": 23668 }, { "epoch": 9.33, "learning_rate": 4.6973006586239035e-09, "loss": 0.0088, "step": 23669 }, { "epoch": 9.33, "learning_rate": 4.691800984469951e-09, "loss": 0.0109, "step": 23670 }, { "epoch": 9.33, "learning_rate": 4.686304493542814e-09, "loss": 0.0101, "step": 23671 }, { "epoch": 9.33, "learning_rate": 4.680811185932087e-09, "loss": 0.0095, "step": 23672 }, { "epoch": 9.33, "learning_rate": 4.675321061727277e-09, "loss": 0.0123, "step": 23673 }, { "epoch": 9.33, "learning_rate": 4.669834121017891e-09, "loss": 0.009, "step": 23674 }, { "epoch": 9.33, "learning_rate": 4.664350363893299e-09, "loss": 0.0076, "step": 23675 }, { "epoch": 9.33, "learning_rate": 4.658869790442943e-09, "loss": 0.0105, "step": 23676 }, { "epoch": 9.33, "learning_rate": 4.653392400756129e-09, "loss": 0.0089, "step": 23677 }, { "epoch": 9.33, "learning_rate": 4.64791819492214e-09, "loss": 0.0092, "step": 23678 }, { "epoch": 9.33, "learning_rate": 4.642447173030151e-09, "loss": 0.0129, "step": 23679 }, { "epoch": 9.33, "learning_rate": 4.6369793351693775e-09, "loss": 0.0089, "step": 23680 }, { "epoch": 9.33, "learning_rate": 4.6315146814289496e-09, "loss": 0.0103, "step": 23681 }, { "epoch": 9.33, "learning_rate": 4.626053211897862e-09, "loss": 0.0095, "step": 23682 }, { "epoch": 9.33, "learning_rate": 4.6205949266652e-09, "loss": 0.0087, "step": 23683 }, { "epoch": 9.33, "learning_rate": 4.615139825819869e-09, "loss": 0.0121, "step": 23684 }, { "epoch": 9.33, "learning_rate": 4.609687909450843e-09, "loss": 0.0085, "step": 23685 }, { "epoch": 9.33, "learning_rate": 4.60423917764694e-09, "loss": 0.0107, "step": 23686 }, { "epoch": 9.34, "learning_rate": 4.598793630496978e-09, "loss": 0.0111, "step": 23687 }, { "epoch": 9.34, "learning_rate": 4.593351268089662e-09, "loss": 0.0092, "step": 23688 }, { "epoch": 9.34, "learning_rate": 4.587912090513768e-09, "loss": 0.0115, "step": 23689 }, { "epoch": 9.34, "learning_rate": 4.582476097857912e-09, "loss": 0.0099, "step": 23690 }, { "epoch": 9.34, "learning_rate": 4.577043290210714e-09, "loss": 0.0104, "step": 23691 }, { "epoch": 9.34, "learning_rate": 4.571613667660656e-09, "loss": 0.0109, "step": 23692 }, { "epoch": 9.34, "learning_rate": 4.566187230296314e-09, "loss": 0.012, "step": 23693 }, { "epoch": 9.34, "learning_rate": 4.560763978206083e-09, "loss": 0.0105, "step": 23694 }, { "epoch": 9.34, "learning_rate": 4.555343911478382e-09, "loss": 0.0094, "step": 23695 }, { "epoch": 9.34, "learning_rate": 4.5499270302015165e-09, "loss": 0.0115, "step": 23696 }, { "epoch": 9.34, "learning_rate": 4.544513334463773e-09, "loss": 0.0077, "step": 23697 }, { "epoch": 9.34, "learning_rate": 4.539102824353435e-09, "loss": 0.0084, "step": 23698 }, { "epoch": 9.34, "learning_rate": 4.533695499958657e-09, "loss": 0.0111, "step": 23699 }, { "epoch": 9.34, "learning_rate": 4.528291361367542e-09, "loss": 0.0101, "step": 23700 }, { "epoch": 9.34, "learning_rate": 4.522890408668201e-09, "loss": 0.0103, "step": 23701 }, { "epoch": 9.34, "learning_rate": 4.517492641948672e-09, "loss": 0.0084, "step": 23702 }, { "epoch": 9.34, "learning_rate": 4.51209806129691e-09, "loss": 0.0097, "step": 23703 }, { "epoch": 9.34, "learning_rate": 4.506706666800842e-09, "loss": 0.0123, "step": 23704 }, { "epoch": 9.34, "learning_rate": 4.501318458548331e-09, "loss": 0.0098, "step": 23705 }, { "epoch": 9.34, "learning_rate": 4.495933436627197e-09, "loss": 0.0091, "step": 23706 }, { "epoch": 9.34, "learning_rate": 4.490551601125214e-09, "loss": 0.0099, "step": 23707 }, { "epoch": 9.34, "learning_rate": 4.485172952130134e-09, "loss": 0.0114, "step": 23708 }, { "epoch": 9.34, "learning_rate": 4.479797489729553e-09, "loss": 0.0086, "step": 23709 }, { "epoch": 9.34, "learning_rate": 4.474425214011135e-09, "loss": 0.0104, "step": 23710 }, { "epoch": 9.34, "learning_rate": 4.469056125062409e-09, "loss": 0.0109, "step": 23711 }, { "epoch": 9.35, "learning_rate": 4.4636902229709065e-09, "loss": 0.0093, "step": 23712 }, { "epoch": 9.35, "learning_rate": 4.458327507824067e-09, "loss": 0.0126, "step": 23713 }, { "epoch": 9.35, "learning_rate": 4.45296797970931e-09, "loss": 0.013, "step": 23714 }, { "epoch": 9.35, "learning_rate": 4.447611638713944e-09, "loss": 0.0083, "step": 23715 }, { "epoch": 9.35, "learning_rate": 4.442258484925343e-09, "loss": 0.0092, "step": 23716 }, { "epoch": 9.35, "learning_rate": 4.436908518430704e-09, "loss": 0.0109, "step": 23717 }, { "epoch": 9.35, "learning_rate": 4.431561739317247e-09, "loss": 0.0137, "step": 23718 }, { "epoch": 9.35, "learning_rate": 4.426218147672056e-09, "loss": 0.0121, "step": 23719 }, { "epoch": 9.35, "learning_rate": 4.420877743582307e-09, "loss": 0.0097, "step": 23720 }, { "epoch": 9.35, "learning_rate": 4.415540527134998e-09, "loss": 0.0079, "step": 23721 }, { "epoch": 9.35, "learning_rate": 4.410206498417146e-09, "loss": 0.0103, "step": 23722 }, { "epoch": 9.35, "learning_rate": 4.404875657515639e-09, "loss": 0.0125, "step": 23723 }, { "epoch": 9.35, "learning_rate": 4.399548004517384e-09, "loss": 0.01, "step": 23724 }, { "epoch": 9.35, "learning_rate": 4.394223539509245e-09, "loss": 0.0101, "step": 23725 }, { "epoch": 9.35, "learning_rate": 4.388902262577954e-09, "loss": 0.0128, "step": 23726 }, { "epoch": 9.35, "learning_rate": 4.383584173810262e-09, "loss": 0.009, "step": 23727 }, { "epoch": 9.35, "learning_rate": 4.378269273292856e-09, "loss": 0.011, "step": 23728 }, { "epoch": 9.35, "learning_rate": 4.372957561112356e-09, "loss": 0.0128, "step": 23729 }, { "epoch": 9.35, "learning_rate": 4.367649037355314e-09, "loss": 0.008, "step": 23730 }, { "epoch": 9.35, "learning_rate": 4.362343702108284e-09, "loss": 0.0097, "step": 23731 }, { "epoch": 9.35, "learning_rate": 4.357041555457663e-09, "loss": 0.0113, "step": 23732 }, { "epoch": 9.35, "learning_rate": 4.351742597489982e-09, "loss": 0.011, "step": 23733 }, { "epoch": 9.35, "learning_rate": 4.346446828291528e-09, "loss": 0.0094, "step": 23734 }, { "epoch": 9.35, "learning_rate": 4.341154247948631e-09, "loss": 0.0092, "step": 23735 }, { "epoch": 9.35, "learning_rate": 4.3358648565475554e-09, "loss": 0.0097, "step": 23736 }, { "epoch": 9.35, "learning_rate": 4.330578654174499e-09, "loss": 0.0077, "step": 23737 }, { "epoch": 9.36, "learning_rate": 4.325295640915616e-09, "loss": 0.0105, "step": 23738 }, { "epoch": 9.36, "learning_rate": 4.320015816857059e-09, "loss": 0.0101, "step": 23739 }, { "epoch": 9.36, "learning_rate": 4.314739182084804e-09, "loss": 0.0117, "step": 23740 }, { "epoch": 9.36, "learning_rate": 4.309465736684914e-09, "loss": 0.0121, "step": 23741 }, { "epoch": 9.36, "learning_rate": 4.3041954807433e-09, "loss": 0.0089, "step": 23742 }, { "epoch": 9.36, "learning_rate": 4.298928414345915e-09, "loss": 0.0119, "step": 23743 }, { "epoch": 9.36, "learning_rate": 4.293664537578534e-09, "loss": 0.008, "step": 23744 }, { "epoch": 9.36, "learning_rate": 4.288403850527001e-09, "loss": 0.0132, "step": 23745 }, { "epoch": 9.36, "learning_rate": 4.283146353277023e-09, "loss": 0.0095, "step": 23746 }, { "epoch": 9.36, "learning_rate": 4.277892045914311e-09, "loss": 0.0125, "step": 23747 }, { "epoch": 9.36, "learning_rate": 4.272640928524485e-09, "loss": 0.011, "step": 23748 }, { "epoch": 9.36, "learning_rate": 4.2673930011931425e-09, "loss": 0.0098, "step": 23749 }, { "epoch": 9.36, "learning_rate": 4.262148264005816e-09, "loss": 0.0095, "step": 23750 }, { "epoch": 9.36, "learning_rate": 4.256906717047992e-09, "loss": 0.0101, "step": 23751 }, { "epoch": 9.36, "learning_rate": 4.251668360405092e-09, "loss": 0.0095, "step": 23752 }, { "epoch": 9.36, "learning_rate": 4.246433194162491e-09, "loss": 0.0078, "step": 23753 }, { "epoch": 9.36, "learning_rate": 4.241201218405544e-09, "loss": 0.01, "step": 23754 }, { "epoch": 9.36, "learning_rate": 4.235972433219448e-09, "loss": 0.0093, "step": 23755 }, { "epoch": 9.36, "learning_rate": 4.230746838689514e-09, "loss": 0.009, "step": 23756 }, { "epoch": 9.36, "learning_rate": 4.225524434900851e-09, "loss": 0.0079, "step": 23757 }, { "epoch": 9.36, "learning_rate": 4.220305221938614e-09, "loss": 0.0114, "step": 23758 }, { "epoch": 9.36, "learning_rate": 4.215089199887822e-09, "loss": 0.0091, "step": 23759 }, { "epoch": 9.36, "learning_rate": 4.209876368833542e-09, "loss": 0.01, "step": 23760 }, { "epoch": 9.36, "learning_rate": 4.204666728860706e-09, "loss": 0.0116, "step": 23761 }, { "epoch": 9.36, "learning_rate": 4.1994602800542456e-09, "loss": 0.0106, "step": 23762 }, { "epoch": 9.37, "learning_rate": 4.19425702249896e-09, "loss": 0.0111, "step": 23763 }, { "epoch": 9.37, "learning_rate": 4.189056956279713e-09, "loss": 0.01, "step": 23764 }, { "epoch": 9.37, "learning_rate": 4.183860081481239e-09, "loss": 0.0121, "step": 23765 }, { "epoch": 9.37, "learning_rate": 4.178666398188246e-09, "loss": 0.0105, "step": 23766 }, { "epoch": 9.37, "learning_rate": 4.17347590648538e-09, "loss": 0.0079, "step": 23767 }, { "epoch": 9.37, "learning_rate": 4.168288606457216e-09, "loss": 0.0072, "step": 23768 }, { "epoch": 9.37, "learning_rate": 4.163104498188352e-09, "loss": 0.0077, "step": 23769 }, { "epoch": 9.37, "learning_rate": 4.157923581763212e-09, "loss": 0.01, "step": 23770 }, { "epoch": 9.37, "learning_rate": 4.1527458572663045e-09, "loss": 0.0116, "step": 23771 }, { "epoch": 9.37, "learning_rate": 4.14757132478194e-09, "loss": 0.0107, "step": 23772 }, { "epoch": 9.37, "learning_rate": 4.14239998439454e-09, "loss": 0.0083, "step": 23773 }, { "epoch": 9.37, "learning_rate": 4.137231836188327e-09, "loss": 0.0139, "step": 23774 }, { "epoch": 9.37, "learning_rate": 4.132066880247586e-09, "loss": 0.011, "step": 23775 }, { "epoch": 9.37, "learning_rate": 4.126905116656432e-09, "loss": 0.0106, "step": 23776 }, { "epoch": 9.37, "learning_rate": 4.12174654549906e-09, "loss": 0.0102, "step": 23777 }, { "epoch": 9.37, "learning_rate": 4.116591166859495e-09, "loss": 0.0082, "step": 23778 }, { "epoch": 9.37, "learning_rate": 4.111438980821824e-09, "loss": 0.0084, "step": 23779 }, { "epoch": 9.37, "learning_rate": 4.106289987469958e-09, "loss": 0.0105, "step": 23780 }, { "epoch": 9.37, "learning_rate": 4.1011441868878285e-09, "loss": 0.0087, "step": 23781 }, { "epoch": 9.37, "learning_rate": 4.096001579159325e-09, "loss": 0.0098, "step": 23782 }, { "epoch": 9.37, "learning_rate": 4.0908621643682695e-09, "loss": 0.01, "step": 23783 }, { "epoch": 9.37, "learning_rate": 4.085725942598395e-09, "loss": 0.0103, "step": 23784 }, { "epoch": 9.37, "learning_rate": 4.080592913933456e-09, "loss": 0.0116, "step": 23785 }, { "epoch": 9.37, "learning_rate": 4.075463078457076e-09, "loss": 0.0105, "step": 23786 }, { "epoch": 9.37, "learning_rate": 4.070336436252874e-09, "loss": 0.0095, "step": 23787 }, { "epoch": 9.38, "learning_rate": 4.0652129874044095e-09, "loss": 0.0085, "step": 23788 }, { "epoch": 9.38, "learning_rate": 4.060092731995191e-09, "loss": 0.0152, "step": 23789 }, { "epoch": 9.38, "learning_rate": 4.0549756701086625e-09, "loss": 0.0095, "step": 23790 }, { "epoch": 9.38, "learning_rate": 4.049861801828225e-09, "loss": 0.0088, "step": 23791 }, { "epoch": 9.38, "learning_rate": 4.044751127237256e-09, "loss": 0.0094, "step": 23792 }, { "epoch": 9.38, "learning_rate": 4.039643646419e-09, "loss": 0.0104, "step": 23793 }, { "epoch": 9.38, "learning_rate": 4.0345393594567455e-09, "loss": 0.0099, "step": 23794 }, { "epoch": 9.38, "learning_rate": 4.029438266433627e-09, "loss": 0.008, "step": 23795 }, { "epoch": 9.38, "learning_rate": 4.024340367432866e-09, "loss": 0.0107, "step": 23796 }, { "epoch": 9.38, "learning_rate": 4.019245662537485e-09, "loss": 0.009, "step": 23797 }, { "epoch": 9.38, "learning_rate": 4.014154151830573e-09, "loss": 0.0098, "step": 23798 }, { "epoch": 9.38, "learning_rate": 4.00906583539502e-09, "loss": 0.0093, "step": 23799 }, { "epoch": 9.38, "learning_rate": 4.0039807133138925e-09, "loss": 0.0095, "step": 23800 }, { "epoch": 9.38, "learning_rate": 3.9988987856699685e-09, "loss": 0.0097, "step": 23801 }, { "epoch": 9.38, "learning_rate": 3.9938200525461155e-09, "loss": 0.0101, "step": 23802 }, { "epoch": 9.38, "learning_rate": 3.988744514025089e-09, "loss": 0.0099, "step": 23803 }, { "epoch": 9.38, "learning_rate": 3.9836721701896226e-09, "loss": 0.0089, "step": 23804 }, { "epoch": 9.38, "learning_rate": 3.978603021122384e-09, "loss": 0.0083, "step": 23805 }, { "epoch": 9.38, "learning_rate": 3.973537066906041e-09, "loss": 0.0103, "step": 23806 }, { "epoch": 9.38, "learning_rate": 3.968474307623082e-09, "loss": 0.0108, "step": 23807 }, { "epoch": 9.38, "learning_rate": 3.963414743356042e-09, "loss": 0.0103, "step": 23808 }, { "epoch": 9.38, "learning_rate": 3.958358374187454e-09, "loss": 0.0097, "step": 23809 }, { "epoch": 9.38, "learning_rate": 3.9533052001996305e-09, "loss": 0.009, "step": 23810 }, { "epoch": 9.38, "learning_rate": 3.9482552214749945e-09, "loss": 0.0105, "step": 23811 }, { "epoch": 9.38, "learning_rate": 3.9432084380958134e-09, "loss": 0.0111, "step": 23812 }, { "epoch": 9.38, "learning_rate": 3.938164850144398e-09, "loss": 0.0094, "step": 23813 }, { "epoch": 9.39, "learning_rate": 3.933124457702886e-09, "loss": 0.0097, "step": 23814 }, { "epoch": 9.39, "learning_rate": 3.928087260853474e-09, "loss": 0.0095, "step": 23815 }, { "epoch": 9.39, "learning_rate": 3.923053259678233e-09, "loss": 0.0099, "step": 23816 }, { "epoch": 9.39, "learning_rate": 3.918022454259229e-09, "loss": 0.0114, "step": 23817 }, { "epoch": 9.39, "learning_rate": 3.912994844678441e-09, "loss": 0.0095, "step": 23818 }, { "epoch": 9.39, "learning_rate": 3.907970431017826e-09, "loss": 0.0095, "step": 23819 }, { "epoch": 9.39, "learning_rate": 3.902949213359274e-09, "loss": 0.012, "step": 23820 }, { "epoch": 9.39, "learning_rate": 3.897931191784587e-09, "loss": 0.0075, "step": 23821 }, { "epoch": 9.39, "learning_rate": 3.892916366375609e-09, "loss": 0.0092, "step": 23822 }, { "epoch": 9.39, "learning_rate": 3.887904737214054e-09, "loss": 0.0095, "step": 23823 }, { "epoch": 9.39, "learning_rate": 3.8828963043815666e-09, "loss": 0.0087, "step": 23824 }, { "epoch": 9.39, "learning_rate": 3.877891067959837e-09, "loss": 0.0104, "step": 23825 }, { "epoch": 9.39, "learning_rate": 3.8728890280303794e-09, "loss": 0.0106, "step": 23826 }, { "epoch": 9.39, "learning_rate": 3.867890184674771e-09, "loss": 0.0105, "step": 23827 }, { "epoch": 9.39, "learning_rate": 3.862894537974459e-09, "loss": 0.0116, "step": 23828 }, { "epoch": 9.39, "learning_rate": 3.85790208801089e-09, "loss": 0.0111, "step": 23829 }, { "epoch": 9.39, "learning_rate": 3.852912834865396e-09, "loss": 0.0116, "step": 23830 }, { "epoch": 9.39, "learning_rate": 3.847926778619315e-09, "loss": 0.0112, "step": 23831 }, { "epoch": 9.39, "learning_rate": 3.842943919353914e-09, "loss": 0.0104, "step": 23832 }, { "epoch": 9.39, "learning_rate": 3.837964257150417e-09, "loss": 0.0116, "step": 23833 }, { "epoch": 9.39, "learning_rate": 3.832987792089937e-09, "loss": 0.0121, "step": 23834 }, { "epoch": 9.39, "learning_rate": 3.828014524253631e-09, "loss": 0.0098, "step": 23835 }, { "epoch": 9.39, "learning_rate": 3.823044453722568e-09, "loss": 0.0084, "step": 23836 }, { "epoch": 9.39, "learning_rate": 3.818077580577683e-09, "loss": 0.0093, "step": 23837 }, { "epoch": 9.39, "learning_rate": 3.813113904900023e-09, "loss": 0.01, "step": 23838 }, { "epoch": 9.4, "learning_rate": 3.8081534267703665e-09, "loss": 0.0096, "step": 23839 }, { "epoch": 9.4, "learning_rate": 3.803196146269694e-09, "loss": 0.0078, "step": 23840 }, { "epoch": 9.4, "learning_rate": 3.798242063478718e-09, "loss": 0.0091, "step": 23841 }, { "epoch": 9.4, "learning_rate": 3.793291178478197e-09, "loss": 0.0068, "step": 23842 }, { "epoch": 9.4, "learning_rate": 3.788343491348844e-09, "loss": 0.0115, "step": 23843 }, { "epoch": 9.4, "learning_rate": 3.783399002171261e-09, "loss": 0.0123, "step": 23844 }, { "epoch": 9.4, "learning_rate": 3.778457711026073e-09, "loss": 0.009, "step": 23845 }, { "epoch": 9.4, "learning_rate": 3.773519617993815e-09, "loss": 0.0086, "step": 23846 }, { "epoch": 9.4, "learning_rate": 3.768584723154955e-09, "loss": 0.0085, "step": 23847 }, { "epoch": 9.4, "learning_rate": 3.763653026589897e-09, "loss": 0.0112, "step": 23848 }, { "epoch": 9.4, "learning_rate": 3.7587245283790645e-09, "loss": 0.0089, "step": 23849 }, { "epoch": 9.4, "learning_rate": 3.7537992286027944e-09, "loss": 0.0077, "step": 23850 }, { "epoch": 9.4, "learning_rate": 3.74887712734131e-09, "loss": 0.0092, "step": 23851 }, { "epoch": 9.4, "learning_rate": 3.7439582246748814e-09, "loss": 0.0093, "step": 23852 }, { "epoch": 9.4, "learning_rate": 3.739042520683644e-09, "loss": 0.0096, "step": 23853 }, { "epoch": 9.4, "learning_rate": 3.734130015447734e-09, "loss": 0.0111, "step": 23854 }, { "epoch": 9.4, "learning_rate": 3.729220709047221e-09, "loss": 0.01, "step": 23855 }, { "epoch": 9.4, "learning_rate": 3.7243146015621063e-09, "loss": 0.0112, "step": 23856 }, { "epoch": 9.4, "learning_rate": 3.7194116930723498e-09, "loss": 0.0078, "step": 23857 }, { "epoch": 9.4, "learning_rate": 3.7145119836578865e-09, "loss": 0.0086, "step": 23858 }, { "epoch": 9.4, "learning_rate": 3.709615473398564e-09, "loss": 0.0088, "step": 23859 }, { "epoch": 9.4, "learning_rate": 3.704722162374163e-09, "loss": 0.0074, "step": 23860 }, { "epoch": 9.4, "learning_rate": 3.699832050664464e-09, "loss": 0.0117, "step": 23861 }, { "epoch": 9.4, "learning_rate": 3.694945138349159e-09, "loss": 0.0098, "step": 23862 }, { "epoch": 9.4, "learning_rate": 3.690061425507918e-09, "loss": 0.0105, "step": 23863 }, { "epoch": 9.4, "learning_rate": 3.685180912220276e-09, "loss": 0.0096, "step": 23864 }, { "epoch": 9.41, "learning_rate": 3.680303598565859e-09, "loss": 0.0097, "step": 23865 }, { "epoch": 9.41, "learning_rate": 3.675429484624093e-09, "loss": 0.0101, "step": 23866 }, { "epoch": 9.41, "learning_rate": 3.6705585704744913e-09, "loss": 0.0097, "step": 23867 }, { "epoch": 9.41, "learning_rate": 3.665690856196346e-09, "loss": 0.0114, "step": 23868 }, { "epoch": 9.41, "learning_rate": 3.6608263418691055e-09, "loss": 0.01, "step": 23869 }, { "epoch": 9.41, "learning_rate": 3.6559650275719277e-09, "loss": 0.0106, "step": 23870 }, { "epoch": 9.41, "learning_rate": 3.65110691338415e-09, "loss": 0.0085, "step": 23871 }, { "epoch": 9.41, "learning_rate": 3.646251999384908e-09, "loss": 0.0106, "step": 23872 }, { "epoch": 9.41, "learning_rate": 3.6414002856533397e-09, "loss": 0.0097, "step": 23873 }, { "epoch": 9.41, "learning_rate": 3.6365517722685148e-09, "loss": 0.008, "step": 23874 }, { "epoch": 9.41, "learning_rate": 3.631706459309458e-09, "loss": 0.0131, "step": 23875 }, { "epoch": 9.41, "learning_rate": 3.6268643468551517e-09, "loss": 0.0101, "step": 23876 }, { "epoch": 9.41, "learning_rate": 3.6220254349844657e-09, "loss": 0.0127, "step": 23877 }, { "epoch": 9.41, "learning_rate": 3.617189723776337e-09, "loss": 0.0098, "step": 23878 }, { "epoch": 9.41, "learning_rate": 3.6123572133095247e-09, "loss": 0.0088, "step": 23879 }, { "epoch": 9.41, "learning_rate": 3.6075279036628546e-09, "loss": 0.0101, "step": 23880 }, { "epoch": 9.41, "learning_rate": 3.602701794914953e-09, "loss": 0.0107, "step": 23881 }, { "epoch": 9.41, "learning_rate": 3.5978788871445566e-09, "loss": 0.0104, "step": 23882 }, { "epoch": 9.41, "learning_rate": 3.5930591804302246e-09, "loss": 0.0117, "step": 23883 }, { "epoch": 9.41, "learning_rate": 3.588242674850517e-09, "loss": 0.0094, "step": 23884 }, { "epoch": 9.41, "learning_rate": 3.583429370483948e-09, "loss": 0.0098, "step": 23885 }, { "epoch": 9.41, "learning_rate": 3.5786192674089664e-09, "loss": 0.0093, "step": 23886 }, { "epoch": 9.41, "learning_rate": 3.5738123657039545e-09, "loss": 0.0104, "step": 23887 }, { "epoch": 9.41, "learning_rate": 3.5690086654472706e-09, "loss": 0.012, "step": 23888 }, { "epoch": 9.41, "learning_rate": 3.5642081667172088e-09, "loss": 0.0118, "step": 23889 }, { "epoch": 9.42, "learning_rate": 3.5594108695920165e-09, "loss": 0.0094, "step": 23890 }, { "epoch": 9.42, "learning_rate": 3.5546167741498545e-09, "loss": 0.0104, "step": 23891 }, { "epoch": 9.42, "learning_rate": 3.5498258804688817e-09, "loss": 0.0118, "step": 23892 }, { "epoch": 9.42, "learning_rate": 3.5450381886271474e-09, "loss": 0.0099, "step": 23893 }, { "epoch": 9.42, "learning_rate": 3.5402536987027666e-09, "loss": 0.009, "step": 23894 }, { "epoch": 9.42, "learning_rate": 3.5354724107736324e-09, "loss": 0.0081, "step": 23895 }, { "epoch": 9.42, "learning_rate": 3.5306943249177266e-09, "loss": 0.0111, "step": 23896 }, { "epoch": 9.42, "learning_rate": 3.525919441212899e-09, "loss": 0.0076, "step": 23897 }, { "epoch": 9.42, "learning_rate": 3.521147759736953e-09, "loss": 0.0096, "step": 23898 }, { "epoch": 9.42, "learning_rate": 3.5163792805677382e-09, "loss": 0.0124, "step": 23899 }, { "epoch": 9.42, "learning_rate": 3.511614003782881e-09, "loss": 0.0088, "step": 23900 }, { "epoch": 9.42, "learning_rate": 3.506851929460075e-09, "loss": 0.01, "step": 23901 }, { "epoch": 9.42, "learning_rate": 3.5020930576769694e-09, "loss": 0.0102, "step": 23902 }, { "epoch": 9.42, "learning_rate": 3.4973373885111014e-09, "loss": 0.0113, "step": 23903 }, { "epoch": 9.42, "learning_rate": 3.4925849220399874e-09, "loss": 0.0093, "step": 23904 }, { "epoch": 9.42, "learning_rate": 3.4878356583410983e-09, "loss": 0.0094, "step": 23905 }, { "epoch": 9.42, "learning_rate": 3.483089597491773e-09, "loss": 0.0088, "step": 23906 }, { "epoch": 9.42, "learning_rate": 3.47834673956946e-09, "loss": 0.0109, "step": 23907 }, { "epoch": 9.42, "learning_rate": 3.473607084651409e-09, "loss": 0.0138, "step": 23908 }, { "epoch": 9.42, "learning_rate": 3.468870632814891e-09, "loss": 0.0101, "step": 23909 }, { "epoch": 9.42, "learning_rate": 3.464137384137045e-09, "loss": 0.0079, "step": 23910 }, { "epoch": 9.42, "learning_rate": 3.4594073386950972e-09, "loss": 0.0079, "step": 23911 }, { "epoch": 9.42, "learning_rate": 3.454680496566076e-09, "loss": 0.0091, "step": 23912 }, { "epoch": 9.42, "learning_rate": 3.4499568578270743e-09, "loss": 0.0077, "step": 23913 }, { "epoch": 9.42, "learning_rate": 3.4452364225550312e-09, "loss": 0.0113, "step": 23914 }, { "epoch": 9.43, "learning_rate": 3.4405191908269295e-09, "loss": 0.0121, "step": 23915 }, { "epoch": 9.43, "learning_rate": 3.435805162719596e-09, "loss": 0.0116, "step": 23916 }, { "epoch": 9.43, "learning_rate": 3.4310943383099255e-09, "loss": 0.0075, "step": 23917 }, { "epoch": 9.43, "learning_rate": 3.4263867176746564e-09, "loss": 0.0099, "step": 23918 }, { "epoch": 9.43, "learning_rate": 3.421682300890527e-09, "loss": 0.0112, "step": 23919 }, { "epoch": 9.43, "learning_rate": 3.416981088034232e-09, "loss": 0.0095, "step": 23920 }, { "epoch": 9.43, "learning_rate": 3.412283079182332e-09, "loss": 0.0093, "step": 23921 }, { "epoch": 9.43, "learning_rate": 3.407588274411477e-09, "loss": 0.0114, "step": 23922 }, { "epoch": 9.43, "learning_rate": 3.402896673798117e-09, "loss": 0.0109, "step": 23923 }, { "epoch": 9.43, "learning_rate": 3.398208277418768e-09, "loss": 0.0082, "step": 23924 }, { "epoch": 9.43, "learning_rate": 3.3935230853498143e-09, "loss": 0.0107, "step": 23925 }, { "epoch": 9.43, "learning_rate": 3.3888410976676385e-09, "loss": 0.0114, "step": 23926 }, { "epoch": 9.43, "learning_rate": 3.384162314448513e-09, "loss": 0.0109, "step": 23927 }, { "epoch": 9.43, "learning_rate": 3.3794867357687328e-09, "loss": 0.0085, "step": 23928 }, { "epoch": 9.43, "learning_rate": 3.374814361704503e-09, "loss": 0.0108, "step": 23929 }, { "epoch": 9.43, "learning_rate": 3.3701451923319635e-09, "loss": 0.0115, "step": 23930 }, { "epoch": 9.43, "learning_rate": 3.365479227727186e-09, "loss": 0.0109, "step": 23931 }, { "epoch": 9.43, "learning_rate": 3.3608164679662877e-09, "loss": 0.0093, "step": 23932 }, { "epoch": 9.43, "learning_rate": 3.3561569131251853e-09, "loss": 0.0074, "step": 23933 }, { "epoch": 9.43, "learning_rate": 3.3515005632798856e-09, "loss": 0.0107, "step": 23934 }, { "epoch": 9.43, "learning_rate": 3.346847418506238e-09, "loss": 0.0103, "step": 23935 }, { "epoch": 9.43, "learning_rate": 3.3421974788801156e-09, "loss": 0.0099, "step": 23936 }, { "epoch": 9.43, "learning_rate": 3.33755074447728e-09, "loss": 0.0114, "step": 23937 }, { "epoch": 9.43, "learning_rate": 3.3329072153734707e-09, "loss": 0.0103, "step": 23938 }, { "epoch": 9.43, "learning_rate": 3.3282668916443822e-09, "loss": 0.0101, "step": 23939 }, { "epoch": 9.43, "learning_rate": 3.3236297733656437e-09, "loss": 0.0083, "step": 23940 }, { "epoch": 9.44, "learning_rate": 3.3189958606128167e-09, "loss": 0.0093, "step": 23941 }, { "epoch": 9.44, "learning_rate": 3.3143651534614404e-09, "loss": 0.01, "step": 23942 }, { "epoch": 9.44, "learning_rate": 3.3097376519869878e-09, "loss": 0.0103, "step": 23943 }, { "epoch": 9.44, "learning_rate": 3.305113356264866e-09, "loss": 0.0088, "step": 23944 }, { "epoch": 9.44, "learning_rate": 3.300492266370458e-09, "loss": 0.0083, "step": 23945 }, { "epoch": 9.44, "learning_rate": 3.29587438237906e-09, "loss": 0.0079, "step": 23946 }, { "epoch": 9.44, "learning_rate": 3.2912597043659895e-09, "loss": 0.009, "step": 23947 }, { "epoch": 9.44, "learning_rate": 3.286648232406386e-09, "loss": 0.0112, "step": 23948 }, { "epoch": 9.44, "learning_rate": 3.2820399665754785e-09, "loss": 0.0095, "step": 23949 }, { "epoch": 9.44, "learning_rate": 3.2774349069483174e-09, "loss": 0.0106, "step": 23950 }, { "epoch": 9.44, "learning_rate": 3.272833053599999e-09, "loss": 0.0103, "step": 23951 }, { "epoch": 9.44, "learning_rate": 3.268234406605508e-09, "loss": 0.0106, "step": 23952 }, { "epoch": 9.44, "learning_rate": 3.2636389660398057e-09, "loss": 0.0089, "step": 23953 }, { "epoch": 9.44, "learning_rate": 3.259046731977788e-09, "loss": 0.01, "step": 23954 }, { "epoch": 9.44, "learning_rate": 3.254457704494262e-09, "loss": 0.0072, "step": 23955 }, { "epoch": 9.44, "learning_rate": 3.2498718836640794e-09, "loss": 0.0102, "step": 23956 }, { "epoch": 9.44, "learning_rate": 3.2452892695619792e-09, "loss": 0.01, "step": 23957 }, { "epoch": 9.44, "learning_rate": 3.240709862262614e-09, "loss": 0.0083, "step": 23958 }, { "epoch": 9.44, "learning_rate": 3.236133661840612e-09, "loss": 0.0111, "step": 23959 }, { "epoch": 9.44, "learning_rate": 3.2315606683706033e-09, "loss": 0.0118, "step": 23960 }, { "epoch": 9.44, "learning_rate": 3.2269908819271274e-09, "loss": 0.0114, "step": 23961 }, { "epoch": 9.44, "learning_rate": 3.222424302584614e-09, "loss": 0.0119, "step": 23962 }, { "epoch": 9.44, "learning_rate": 3.2178609304175373e-09, "loss": 0.0114, "step": 23963 }, { "epoch": 9.44, "learning_rate": 3.2133007655002375e-09, "loss": 0.0102, "step": 23964 }, { "epoch": 9.44, "learning_rate": 3.208743807907055e-09, "loss": 0.0106, "step": 23965 }, { "epoch": 9.45, "learning_rate": 3.204190057712264e-09, "loss": 0.009, "step": 23966 }, { "epoch": 9.45, "learning_rate": 3.1996395149900713e-09, "loss": 0.0134, "step": 23967 }, { "epoch": 9.45, "learning_rate": 3.19509217981464e-09, "loss": 0.0101, "step": 23968 }, { "epoch": 9.45, "learning_rate": 3.1905480522601115e-09, "loss": 0.0072, "step": 23969 }, { "epoch": 9.45, "learning_rate": 3.186007132400559e-09, "loss": 0.0086, "step": 23970 }, { "epoch": 9.45, "learning_rate": 3.181469420309946e-09, "loss": 0.0118, "step": 23971 }, { "epoch": 9.45, "learning_rate": 3.1769349160622573e-09, "loss": 0.009, "step": 23972 }, { "epoch": 9.45, "learning_rate": 3.1724036197313676e-09, "loss": 0.0101, "step": 23973 }, { "epoch": 9.45, "learning_rate": 3.167875531391173e-09, "loss": 0.0097, "step": 23974 }, { "epoch": 9.45, "learning_rate": 3.1633506511154597e-09, "loss": 0.0109, "step": 23975 }, { "epoch": 9.45, "learning_rate": 3.1588289789779898e-09, "loss": 0.0109, "step": 23976 }, { "epoch": 9.45, "learning_rate": 3.154310515052394e-09, "loss": 0.0111, "step": 23977 }, { "epoch": 9.45, "learning_rate": 3.1497952594124135e-09, "loss": 0.0103, "step": 23978 }, { "epoch": 9.45, "learning_rate": 3.1452832121315664e-09, "loss": 0.0105, "step": 23979 }, { "epoch": 9.45, "learning_rate": 3.140774373283439e-09, "loss": 0.0099, "step": 23980 }, { "epoch": 9.45, "learning_rate": 3.1362687429414838e-09, "loss": 0.0088, "step": 23981 }, { "epoch": 9.45, "learning_rate": 3.131766321179152e-09, "loss": 0.0106, "step": 23982 }, { "epoch": 9.45, "learning_rate": 3.1272671080698086e-09, "loss": 0.0122, "step": 23983 }, { "epoch": 9.45, "learning_rate": 3.1227711036868386e-09, "loss": 0.0105, "step": 23984 }, { "epoch": 9.45, "learning_rate": 3.1182783081034505e-09, "loss": 0.0088, "step": 23985 }, { "epoch": 9.45, "learning_rate": 3.113788721392896e-09, "loss": 0.0102, "step": 23986 }, { "epoch": 9.45, "learning_rate": 3.109302343628384e-09, "loss": 0.012, "step": 23987 }, { "epoch": 9.45, "learning_rate": 3.1048191748829665e-09, "loss": 0.0101, "step": 23988 }, { "epoch": 9.45, "learning_rate": 3.100339215229786e-09, "loss": 0.0104, "step": 23989 }, { "epoch": 9.45, "learning_rate": 3.095862464741783e-09, "loss": 0.0086, "step": 23990 }, { "epoch": 9.46, "learning_rate": 3.0913889234919886e-09, "loss": 0.0073, "step": 23991 }, { "epoch": 9.46, "learning_rate": 3.0869185915533e-09, "loss": 0.0089, "step": 23992 }, { "epoch": 9.46, "learning_rate": 3.082451468998548e-09, "loss": 0.0092, "step": 23993 }, { "epoch": 9.46, "learning_rate": 3.077987555900563e-09, "loss": 0.0102, "step": 23994 }, { "epoch": 9.46, "learning_rate": 3.073526852332087e-09, "loss": 0.0102, "step": 23995 }, { "epoch": 9.46, "learning_rate": 3.0690693583658167e-09, "loss": 0.0094, "step": 23996 }, { "epoch": 9.46, "learning_rate": 3.0646150740744504e-09, "loss": 0.0082, "step": 23997 }, { "epoch": 9.46, "learning_rate": 3.060163999530507e-09, "loss": 0.012, "step": 23998 }, { "epoch": 9.46, "learning_rate": 3.055716134806596e-09, "loss": 0.0079, "step": 23999 }, { "epoch": 9.46, "learning_rate": 3.0512714799751928e-09, "loss": 0.0089, "step": 24000 } ], "logging_steps": 1.0, "max_steps": 25370, "num_input_tokens_seen": 0, "num_train_epochs": 10, "save_steps": 500, "total_flos": 1.3918848139224045e+22, "train_batch_size": 2, "trial_name": null, "trial_params": null }