{ "best_global_step": null, "best_metric": null, "best_model_checkpoint": null, "epoch": 3.0, "eval_steps": 1094, "global_step": 6561, "is_hyper_param_search": false, "is_local_process_zero": true, "is_world_process_zero": true, "log_history": [ { "epoch": 0.00457351932311914, "grad_norm": 1.0297006685314725, "learning_rate": 1.3636363636363636e-06, "loss": 0.3944, "step": 10 }, { "epoch": 0.00914703864623828, "grad_norm": 0.4697933986094453, "learning_rate": 2.8787878787878793e-06, "loss": 0.3665, "step": 20 }, { "epoch": 0.01372055796935742, "grad_norm": 0.41363751514598424, "learning_rate": 4.393939393939394e-06, "loss": 0.3485, "step": 30 }, { "epoch": 0.01829407729247656, "grad_norm": 0.40183509516926424, "learning_rate": 5.90909090909091e-06, "loss": 0.3401, "step": 40 }, { "epoch": 0.0228675966155957, "grad_norm": 0.3959965499431158, "learning_rate": 7.424242424242425e-06, "loss": 0.3405, "step": 50 }, { "epoch": 0.02744111593871484, "grad_norm": 0.47650848978419785, "learning_rate": 8.93939393939394e-06, "loss": 0.3331, "step": 60 }, { "epoch": 0.032014635261833985, "grad_norm": 0.4821536325822335, "learning_rate": 9.999994735903083e-06, "loss": 0.3368, "step": 70 }, { "epoch": 0.03658815458495312, "grad_norm": 0.44254681642187266, "learning_rate": 9.99990115226624e-06, "loss": 0.3358, "step": 80 }, { "epoch": 0.04116167390807226, "grad_norm": 0.4118339912139314, "learning_rate": 9.999690591218082e-06, "loss": 0.3298, "step": 90 }, { "epoch": 0.0457351932311914, "grad_norm": 0.4644781637544726, "learning_rate": 9.999363057684885e-06, "loss": 0.3276, "step": 100 }, { "epoch": 0.05030871255431054, "grad_norm": 0.4868967956683681, "learning_rate": 9.998918559329605e-06, "loss": 0.3317, "step": 110 }, { "epoch": 0.05488223187742968, "grad_norm": 0.4067180181197714, "learning_rate": 9.9983571065517e-06, "loss": 0.3277, "step": 120 }, { "epoch": 0.05945575120054882, "grad_norm": 0.3887129604433554, "learning_rate": 9.997678712486889e-06, "loss": 0.3245, "step": 130 }, { "epoch": 0.06402927052366797, "grad_norm": 0.4190824529383804, "learning_rate": 9.996883393006842e-06, "loss": 0.3225, "step": 140 }, { "epoch": 0.0686027898467871, "grad_norm": 0.46285614798818914, "learning_rate": 9.995971166718811e-06, "loss": 0.3209, "step": 150 }, { "epoch": 0.07317630916990624, "grad_norm": 0.43391229591936903, "learning_rate": 9.99494205496519e-06, "loss": 0.3166, "step": 160 }, { "epoch": 0.07774982849302538, "grad_norm": 0.4484295522068169, "learning_rate": 9.993796081823022e-06, "loss": 0.3217, "step": 170 }, { "epoch": 0.08232334781614452, "grad_norm": 0.46709073084466995, "learning_rate": 9.992533274103431e-06, "loss": 0.3137, "step": 180 }, { "epoch": 0.08689686713926366, "grad_norm": 0.42471375608096795, "learning_rate": 9.991153661350996e-06, "loss": 0.3183, "step": 190 }, { "epoch": 0.0914703864623828, "grad_norm": 0.43157184881337574, "learning_rate": 9.989657275843062e-06, "loss": 0.3196, "step": 200 }, { "epoch": 0.09604390578550194, "grad_norm": 0.4334744851865263, "learning_rate": 9.988044152588981e-06, "loss": 0.3083, "step": 210 }, { "epoch": 0.10061742510862108, "grad_norm": 0.4310254269812107, "learning_rate": 9.986314329329294e-06, "loss": 0.3101, "step": 220 }, { "epoch": 0.10519094443174022, "grad_norm": 0.4450111962613986, "learning_rate": 9.984467846534853e-06, "loss": 0.3101, "step": 230 }, { "epoch": 0.10976446375485936, "grad_norm": 0.4312673688389689, "learning_rate": 9.982504747405868e-06, "loss": 0.3074, "step": 240 }, { "epoch": 0.1143379830779785, "grad_norm": 0.454445728169558, "learning_rate": 9.980425077870895e-06, "loss": 0.3121, "step": 250 }, { "epoch": 0.11891150240109764, "grad_norm": 0.46058870008281705, "learning_rate": 9.97822888658577e-06, "loss": 0.3106, "step": 260 }, { "epoch": 0.12348502172421678, "grad_norm": 0.4962197299975268, "learning_rate": 9.97591622493246e-06, "loss": 0.31, "step": 270 }, { "epoch": 0.12805854104733594, "grad_norm": 0.49107419730770024, "learning_rate": 9.973487147017874e-06, "loss": 0.3088, "step": 280 }, { "epoch": 0.13263206037045505, "grad_norm": 0.5024851065875913, "learning_rate": 9.97094170967258e-06, "loss": 0.3101, "step": 290 }, { "epoch": 0.1372055796935742, "grad_norm": 0.4853763738056611, "learning_rate": 9.968279972449496e-06, "loss": 0.3043, "step": 300 }, { "epoch": 0.14177909901669333, "grad_norm": 0.5684883765333758, "learning_rate": 9.96550199762247e-06, "loss": 0.313, "step": 310 }, { "epoch": 0.14635261833981247, "grad_norm": 0.4520885476393148, "learning_rate": 9.962607850184856e-06, "loss": 0.3071, "step": 320 }, { "epoch": 0.15092613766293161, "grad_norm": 0.45438693004444647, "learning_rate": 9.959597597847964e-06, "loss": 0.304, "step": 330 }, { "epoch": 0.15549965698605075, "grad_norm": 0.4703285571119386, "learning_rate": 9.956471311039491e-06, "loss": 0.3066, "step": 340 }, { "epoch": 0.1600731763091699, "grad_norm": 0.4874627838545944, "learning_rate": 9.95322906290187e-06, "loss": 0.3017, "step": 350 }, { "epoch": 0.16464669563228904, "grad_norm": 0.46754437548693234, "learning_rate": 9.949870929290564e-06, "loss": 0.3021, "step": 360 }, { "epoch": 0.16922021495540818, "grad_norm": 0.47386492916476153, "learning_rate": 9.946396988772275e-06, "loss": 0.3036, "step": 370 }, { "epoch": 0.17379373427852732, "grad_norm": 0.48929628955352256, "learning_rate": 9.94280732262313e-06, "loss": 0.3001, "step": 380 }, { "epoch": 0.17836725360164646, "grad_norm": 0.47627090410834627, "learning_rate": 9.939102014826756e-06, "loss": 0.3028, "step": 390 }, { "epoch": 0.1829407729247656, "grad_norm": 0.49037392434461513, "learning_rate": 9.935281152072329e-06, "loss": 0.3004, "step": 400 }, { "epoch": 0.18751429224788474, "grad_norm": 0.4590894128287286, "learning_rate": 9.931344823752546e-06, "loss": 0.3004, "step": 410 }, { "epoch": 0.19208781157100388, "grad_norm": 0.4478768517416793, "learning_rate": 9.927293121961526e-06, "loss": 0.3017, "step": 420 }, { "epoch": 0.19666133089412302, "grad_norm": 0.510693329409755, "learning_rate": 9.92312614149266e-06, "loss": 0.3011, "step": 430 }, { "epoch": 0.20123485021724216, "grad_norm": 0.4495404384247314, "learning_rate": 9.918843979836394e-06, "loss": 0.2938, "step": 440 }, { "epoch": 0.2058083695403613, "grad_norm": 0.4270307652401074, "learning_rate": 9.914446737177943e-06, "loss": 0.2934, "step": 450 }, { "epoch": 0.21038188886348044, "grad_norm": 0.4605064480258557, "learning_rate": 9.909934516394957e-06, "loss": 0.2947, "step": 460 }, { "epoch": 0.21495540818659958, "grad_norm": 0.4695590327414231, "learning_rate": 9.905307423055102e-06, "loss": 0.2984, "step": 470 }, { "epoch": 0.21952892750971872, "grad_norm": 0.46288121338187016, "learning_rate": 9.900565565413599e-06, "loss": 0.293, "step": 480 }, { "epoch": 0.22410244683283786, "grad_norm": 0.4819422427494635, "learning_rate": 9.895709054410686e-06, "loss": 0.2964, "step": 490 }, { "epoch": 0.228675966155957, "grad_norm": 0.4985399192845818, "learning_rate": 9.890738003669029e-06, "loss": 0.2919, "step": 500 }, { "epoch": 0.23324948547907615, "grad_norm": 0.4714703898674373, "learning_rate": 9.885652529491056e-06, "loss": 0.2872, "step": 510 }, { "epoch": 0.2378230048021953, "grad_norm": 0.4521793827451988, "learning_rate": 9.880452750856239e-06, "loss": 0.2939, "step": 520 }, { "epoch": 0.24239652412531443, "grad_norm": 0.47247438270774933, "learning_rate": 9.875138789418312e-06, "loss": 0.2943, "step": 530 }, { "epoch": 0.24697004344843357, "grad_norm": 0.496789279665293, "learning_rate": 9.869710769502424e-06, "loss": 0.2976, "step": 540 }, { "epoch": 0.2515435627715527, "grad_norm": 0.4858550194774636, "learning_rate": 9.86416881810223e-06, "loss": 0.2895, "step": 550 }, { "epoch": 0.2561170820946719, "grad_norm": 0.4827250972662044, "learning_rate": 9.858513064876918e-06, "loss": 0.2894, "step": 560 }, { "epoch": 0.260690601417791, "grad_norm": 0.4650043048753059, "learning_rate": 9.852743642148179e-06, "loss": 0.2913, "step": 570 }, { "epoch": 0.2652641207409101, "grad_norm": 0.4705514011452597, "learning_rate": 9.846860684897107e-06, "loss": 0.2903, "step": 580 }, { "epoch": 0.26983764006402927, "grad_norm": 0.5974749261711809, "learning_rate": 9.840864330761046e-06, "loss": 0.2827, "step": 590 }, { "epoch": 0.2744111593871484, "grad_norm": 0.46109027153504417, "learning_rate": 9.834754720030366e-06, "loss": 0.2869, "step": 600 }, { "epoch": 0.27898467871026755, "grad_norm": 0.4881554173210495, "learning_rate": 9.828531995645183e-06, "loss": 0.29, "step": 610 }, { "epoch": 0.28355819803338667, "grad_norm": 0.5291232720754163, "learning_rate": 9.822196303192013e-06, "loss": 0.2844, "step": 620 }, { "epoch": 0.28813171735650583, "grad_norm": 0.4594884964530799, "learning_rate": 9.815747790900367e-06, "loss": 0.2829, "step": 630 }, { "epoch": 0.29270523667962495, "grad_norm": 0.4735548479130588, "learning_rate": 9.809186609639281e-06, "loss": 0.2924, "step": 640 }, { "epoch": 0.2972787560027441, "grad_norm": 0.5914546930717978, "learning_rate": 9.802512912913792e-06, "loss": 0.2863, "step": 650 }, { "epoch": 0.30185227532586323, "grad_norm": 0.5384740659682604, "learning_rate": 9.795726856861338e-06, "loss": 0.2871, "step": 660 }, { "epoch": 0.3064257946489824, "grad_norm": 0.49984399569519306, "learning_rate": 9.788828600248114e-06, "loss": 0.2832, "step": 670 }, { "epoch": 0.3109993139721015, "grad_norm": 0.45146450026765916, "learning_rate": 9.781818304465348e-06, "loss": 0.2873, "step": 680 }, { "epoch": 0.3155728332952207, "grad_norm": 0.5361788208447497, "learning_rate": 9.774696133525536e-06, "loss": 0.2799, "step": 690 }, { "epoch": 0.3201463526183398, "grad_norm": 0.5164956049232827, "learning_rate": 9.767462254058593e-06, "loss": 0.2814, "step": 700 }, { "epoch": 0.32471987194145896, "grad_norm": 0.4656363998697899, "learning_rate": 9.760116835307969e-06, "loss": 0.2882, "step": 710 }, { "epoch": 0.3292933912645781, "grad_norm": 0.5305718724694726, "learning_rate": 9.752660049126667e-06, "loss": 0.2874, "step": 720 }, { "epoch": 0.33386691058769724, "grad_norm": 0.47057272986697707, "learning_rate": 9.745092069973254e-06, "loss": 0.2873, "step": 730 }, { "epoch": 0.33844042991081635, "grad_norm": 0.5167611767666651, "learning_rate": 9.737413074907745e-06, "loss": 0.2858, "step": 740 }, { "epoch": 0.3430139492339355, "grad_norm": 0.4701241013280666, "learning_rate": 9.729623243587489e-06, "loss": 0.2805, "step": 750 }, { "epoch": 0.34758746855705464, "grad_norm": 0.4817388932064019, "learning_rate": 9.721722758262948e-06, "loss": 0.28, "step": 760 }, { "epoch": 0.3521609878801738, "grad_norm": 0.5160334006984353, "learning_rate": 9.713711803773441e-06, "loss": 0.2773, "step": 770 }, { "epoch": 0.3567345072032929, "grad_norm": 0.5256154889860845, "learning_rate": 9.705590567542817e-06, "loss": 0.2813, "step": 780 }, { "epoch": 0.3613080265264121, "grad_norm": 0.53796805087917, "learning_rate": 9.697359239575069e-06, "loss": 0.2756, "step": 790 }, { "epoch": 0.3658815458495312, "grad_norm": 0.5178274890535031, "learning_rate": 9.689018012449893e-06, "loss": 0.2786, "step": 800 }, { "epoch": 0.37045506517265037, "grad_norm": 0.5052674849650001, "learning_rate": 9.68056708131818e-06, "loss": 0.2754, "step": 810 }, { "epoch": 0.3750285844957695, "grad_norm": 0.5201655369171125, "learning_rate": 9.672006643897444e-06, "loss": 0.2756, "step": 820 }, { "epoch": 0.37960210381888865, "grad_norm": 0.5194057305696513, "learning_rate": 9.66333690046721e-06, "loss": 0.2753, "step": 830 }, { "epoch": 0.38417562314200776, "grad_norm": 0.5804036954322609, "learning_rate": 9.654558053864315e-06, "loss": 0.2807, "step": 840 }, { "epoch": 0.38874914246512693, "grad_norm": 0.4831124830786409, "learning_rate": 9.64567030947817e-06, "loss": 0.277, "step": 850 }, { "epoch": 0.39332266178824604, "grad_norm": 0.5230113645818625, "learning_rate": 9.63667387524595e-06, "loss": 0.2767, "step": 860 }, { "epoch": 0.3978961811113652, "grad_norm": 0.4961675253068747, "learning_rate": 9.627568961647737e-06, "loss": 0.2804, "step": 870 }, { "epoch": 0.4024697004344843, "grad_norm": 0.47278815502099214, "learning_rate": 9.618355781701584e-06, "loss": 0.2727, "step": 880 }, { "epoch": 0.4070432197576035, "grad_norm": 0.5250109399630284, "learning_rate": 9.60903455095854e-06, "loss": 0.275, "step": 890 }, { "epoch": 0.4116167390807226, "grad_norm": 0.49998367573552066, "learning_rate": 9.59960548749761e-06, "loss": 0.2737, "step": 900 }, { "epoch": 0.4161902584038418, "grad_norm": 0.5291171018233924, "learning_rate": 9.590068811920637e-06, "loss": 0.271, "step": 910 }, { "epoch": 0.4207637777269609, "grad_norm": 0.5344279812598189, "learning_rate": 9.580424747347157e-06, "loss": 0.2694, "step": 920 }, { "epoch": 0.42533729705008005, "grad_norm": 0.49693022178068924, "learning_rate": 9.570673519409173e-06, "loss": 0.2701, "step": 930 }, { "epoch": 0.42991081637319917, "grad_norm": 0.5365080056201543, "learning_rate": 9.560815356245875e-06, "loss": 0.2696, "step": 940 }, { "epoch": 0.43448433569631834, "grad_norm": 0.4988616391428816, "learning_rate": 9.550850488498304e-06, "loss": 0.2741, "step": 950 }, { "epoch": 0.43905785501943745, "grad_norm": 0.48543230691929407, "learning_rate": 9.540779149303958e-06, "loss": 0.2681, "step": 960 }, { "epoch": 0.4436313743425566, "grad_norm": 0.5062169724520093, "learning_rate": 9.530601574291331e-06, "loss": 0.2681, "step": 970 }, { "epoch": 0.44820489366567573, "grad_norm": 0.5376419188211237, "learning_rate": 9.520318001574407e-06, "loss": 0.2751, "step": 980 }, { "epoch": 0.4527784129887949, "grad_norm": 0.517543383255577, "learning_rate": 9.509928671747082e-06, "loss": 0.2664, "step": 990 }, { "epoch": 0.457351932311914, "grad_norm": 0.5147090997784731, "learning_rate": 9.499433827877547e-06, "loss": 0.2699, "step": 1000 }, { "epoch": 0.4619254516350332, "grad_norm": 0.5884307532717357, "learning_rate": 9.48883371550259e-06, "loss": 0.2691, "step": 1010 }, { "epoch": 0.4664989709581523, "grad_norm": 0.486026674498229, "learning_rate": 9.47812858262185e-06, "loss": 0.2695, "step": 1020 }, { "epoch": 0.47107249028127146, "grad_norm": 0.4881721096224783, "learning_rate": 9.467318679692031e-06, "loss": 0.2705, "step": 1030 }, { "epoch": 0.4756460096043906, "grad_norm": 0.5329535415600695, "learning_rate": 9.456404259621022e-06, "loss": 0.2711, "step": 1040 }, { "epoch": 0.48021952892750974, "grad_norm": 0.5256753451099575, "learning_rate": 9.445385577761992e-06, "loss": 0.2676, "step": 1050 }, { "epoch": 0.48479304825062886, "grad_norm": 0.5418301058510792, "learning_rate": 9.434262891907413e-06, "loss": 0.263, "step": 1060 }, { "epoch": 0.489366567573748, "grad_norm": 0.5089113592029859, "learning_rate": 9.423036462283033e-06, "loss": 0.267, "step": 1070 }, { "epoch": 0.49394008689686714, "grad_norm": 0.5385775868474728, "learning_rate": 9.411706551541775e-06, "loss": 0.2694, "step": 1080 }, { "epoch": 0.4985136062199863, "grad_norm": 0.49327903705642273, "learning_rate": 9.400273424757607e-06, "loss": 0.2632, "step": 1090 }, { "epoch": 0.500343013949234, "eval_ood_font_family_decon_loss": 0.34991544485092163, "eval_ood_font_family_decon_runtime": 754.9057, "eval_ood_font_family_decon_samples_per_second": 41.104, "eval_ood_font_family_decon_steps_per_second": 1.285, "step": 1094 }, { "epoch": 0.5030871255431054, "grad_norm": 0.5452853793858355, "learning_rate": 9.388737349419334e-06, "loss": 0.2738, "step": 1100 }, { "epoch": 0.5076606448662245, "grad_norm": 0.5548959172426634, "learning_rate": 9.377098595424334e-06, "loss": 0.2683, "step": 1110 }, { "epoch": 0.5122341641893438, "grad_norm": 0.5293883857895816, "learning_rate": 9.365357435072255e-06, "loss": 0.2623, "step": 1120 }, { "epoch": 0.5168076835124629, "grad_norm": 0.4924213509855563, "learning_rate": 9.353514143058638e-06, "loss": 0.2646, "step": 1130 }, { "epoch": 0.521381202835582, "grad_norm": 0.5390449143407333, "learning_rate": 9.341568996468487e-06, "loss": 0.2618, "step": 1140 }, { "epoch": 0.5259547221587011, "grad_norm": 0.5434621836314503, "learning_rate": 9.329522274769791e-06, "loss": 0.2618, "step": 1150 }, { "epoch": 0.5305282414818202, "grad_norm": 0.5342880987150501, "learning_rate": 9.317374259806986e-06, "loss": 0.2712, "step": 1160 }, { "epoch": 0.5351017608049394, "grad_norm": 0.5256936744983733, "learning_rate": 9.305125235794364e-06, "loss": 0.2601, "step": 1170 }, { "epoch": 0.5396752801280585, "grad_norm": 0.545528973329234, "learning_rate": 9.292775489309409e-06, "loss": 0.2654, "step": 1180 }, { "epoch": 0.5442487994511777, "grad_norm": 0.5350306788029839, "learning_rate": 9.280325309286108e-06, "loss": 0.2662, "step": 1190 }, { "epoch": 0.5488223187742968, "grad_norm": 0.510141756278309, "learning_rate": 9.267774987008191e-06, "loss": 0.2635, "step": 1200 }, { "epoch": 0.553395838097416, "grad_norm": 0.49465157864359943, "learning_rate": 9.2551248161023e-06, "loss": 0.2589, "step": 1210 }, { "epoch": 0.5579693574205351, "grad_norm": 0.5036013435649483, "learning_rate": 9.242375092531138e-06, "loss": 0.2636, "step": 1220 }, { "epoch": 0.5625428767436542, "grad_norm": 0.492549866818152, "learning_rate": 9.22952611458653e-06, "loss": 0.2598, "step": 1230 }, { "epoch": 0.5671163960667733, "grad_norm": 0.5133142051508638, "learning_rate": 9.216578182882459e-06, "loss": 0.2581, "step": 1240 }, { "epoch": 0.5716899153898926, "grad_norm": 0.5481879721710791, "learning_rate": 9.203531600348017e-06, "loss": 0.2587, "step": 1250 }, { "epoch": 0.5762634347130117, "grad_norm": 0.5474500878629359, "learning_rate": 9.19038667222033e-06, "loss": 0.2557, "step": 1260 }, { "epoch": 0.5808369540361308, "grad_norm": 0.5687582578391408, "learning_rate": 9.177143706037411e-06, "loss": 0.256, "step": 1270 }, { "epoch": 0.5854104733592499, "grad_norm": 0.4814952171387427, "learning_rate": 9.163803011630967e-06, "loss": 0.2606, "step": 1280 }, { "epoch": 0.5899839926823691, "grad_norm": 0.5859333986945721, "learning_rate": 9.150364901119145e-06, "loss": 0.2566, "step": 1290 }, { "epoch": 0.5945575120054882, "grad_norm": 0.6027902699657951, "learning_rate": 9.136829688899236e-06, "loss": 0.2598, "step": 1300 }, { "epoch": 0.5991310313286073, "grad_norm": 0.5157928661591503, "learning_rate": 9.123197691640323e-06, "loss": 0.2637, "step": 1310 }, { "epoch": 0.6037045506517265, "grad_norm": 0.5352380381981824, "learning_rate": 9.109469228275855e-06, "loss": 0.2592, "step": 1320 }, { "epoch": 0.6082780699748457, "grad_norm": 0.5115957005763063, "learning_rate": 9.095644619996206e-06, "loss": 0.2575, "step": 1330 }, { "epoch": 0.6128515892979648, "grad_norm": 0.5646354090268043, "learning_rate": 9.08172419024115e-06, "loss": 0.2547, "step": 1340 }, { "epoch": 0.6174251086210839, "grad_norm": 0.5032175215021816, "learning_rate": 9.06770826469229e-06, "loss": 0.2529, "step": 1350 }, { "epoch": 0.621998627944203, "grad_norm": 0.5404811913939005, "learning_rate": 9.053597171265447e-06, "loss": 0.2559, "step": 1360 }, { "epoch": 0.6265721472673222, "grad_norm": 0.5306757662388518, "learning_rate": 9.039391240102982e-06, "loss": 0.2548, "step": 1370 }, { "epoch": 0.6311456665904414, "grad_norm": 0.5769792772269913, "learning_rate": 9.025090803566081e-06, "loss": 0.2579, "step": 1380 }, { "epoch": 0.6357191859135605, "grad_norm": 0.5449082047133691, "learning_rate": 9.010696196226963e-06, "loss": 0.254, "step": 1390 }, { "epoch": 0.6402927052366796, "grad_norm": 0.5848595086705417, "learning_rate": 8.99620775486107e-06, "loss": 0.2554, "step": 1400 }, { "epoch": 0.6448662245597988, "grad_norm": 0.561471395844701, "learning_rate": 8.981625818439172e-06, "loss": 0.2533, "step": 1410 }, { "epoch": 0.6494397438829179, "grad_norm": 0.5512542619938867, "learning_rate": 8.966950728119453e-06, "loss": 0.2565, "step": 1420 }, { "epoch": 0.654013263206037, "grad_norm": 0.5402124095056522, "learning_rate": 8.952182827239515e-06, "loss": 0.2534, "step": 1430 }, { "epoch": 0.6585867825291561, "grad_norm": 0.547826373486063, "learning_rate": 8.937322461308348e-06, "loss": 0.2592, "step": 1440 }, { "epoch": 0.6631603018522754, "grad_norm": 0.5409052165453111, "learning_rate": 8.922369977998257e-06, "loss": 0.2501, "step": 1450 }, { "epoch": 0.6677338211753945, "grad_norm": 0.5312842384620847, "learning_rate": 8.907325727136715e-06, "loss": 0.2541, "step": 1460 }, { "epoch": 0.6723073404985136, "grad_norm": 0.6095702477167819, "learning_rate": 8.892190060698185e-06, "loss": 0.2498, "step": 1470 }, { "epoch": 0.6768808598216327, "grad_norm": 0.6144375007468379, "learning_rate": 8.876963332795881e-06, "loss": 0.2493, "step": 1480 }, { "epoch": 0.6814543791447519, "grad_norm": 0.5161768746537917, "learning_rate": 8.86164589967349e-06, "loss": 0.249, "step": 1490 }, { "epoch": 0.686027898467871, "grad_norm": 0.5852613178582132, "learning_rate": 8.84623811969683e-06, "loss": 0.2496, "step": 1500 }, { "epoch": 0.6906014177909902, "grad_norm": 0.5597911618377422, "learning_rate": 8.830740353345475e-06, "loss": 0.2533, "step": 1510 }, { "epoch": 0.6951749371141093, "grad_norm": 0.5564144855645028, "learning_rate": 8.815152963204306e-06, "loss": 0.2526, "step": 1520 }, { "epoch": 0.6997484564372285, "grad_norm": 0.5178210255593919, "learning_rate": 8.799476313955044e-06, "loss": 0.2547, "step": 1530 }, { "epoch": 0.7043219757603476, "grad_norm": 0.49472672950764857, "learning_rate": 8.78371077236771e-06, "loss": 0.2485, "step": 1540 }, { "epoch": 0.7088954950834667, "grad_norm": 0.5168939600378544, "learning_rate": 8.767856707292046e-06, "loss": 0.2519, "step": 1550 }, { "epoch": 0.7134690144065858, "grad_norm": 0.531901496035939, "learning_rate": 8.751914489648881e-06, "loss": 0.2484, "step": 1560 }, { "epoch": 0.7180425337297051, "grad_norm": 0.5751463974916285, "learning_rate": 8.735884492421457e-06, "loss": 0.2512, "step": 1570 }, { "epoch": 0.7226160530528242, "grad_norm": 0.5414776737508032, "learning_rate": 8.719767090646707e-06, "loss": 0.2464, "step": 1580 }, { "epoch": 0.7271895723759433, "grad_norm": 0.5699719482619237, "learning_rate": 8.703562661406473e-06, "loss": 0.2471, "step": 1590 }, { "epoch": 0.7317630916990624, "grad_norm": 0.6436809168182792, "learning_rate": 8.687271583818687e-06, "loss": 0.2492, "step": 1600 }, { "epoch": 0.7363366110221816, "grad_norm": 0.635962047072558, "learning_rate": 8.670894239028497e-06, "loss": 0.252, "step": 1610 }, { "epoch": 0.7409101303453007, "grad_norm": 0.6347954425925024, "learning_rate": 8.654431010199362e-06, "loss": 0.2541, "step": 1620 }, { "epoch": 0.7454836496684198, "grad_norm": 0.6369379365174839, "learning_rate": 8.637882282504075e-06, "loss": 0.2486, "step": 1630 }, { "epoch": 0.750057168991539, "grad_norm": 0.5313644340284976, "learning_rate": 8.621248443115753e-06, "loss": 0.2512, "step": 1640 }, { "epoch": 0.7546306883146582, "grad_norm": 0.5445476055051183, "learning_rate": 8.604529881198785e-06, "loss": 0.2481, "step": 1650 }, { "epoch": 0.7592042076377773, "grad_norm": 0.5718855886365716, "learning_rate": 8.58772698789972e-06, "loss": 0.2507, "step": 1660 }, { "epoch": 0.7637777269608964, "grad_norm": 0.5549098052946538, "learning_rate": 8.570840156338127e-06, "loss": 0.2481, "step": 1670 }, { "epoch": 0.7683512462840155, "grad_norm": 0.586513244479886, "learning_rate": 8.55386978159738e-06, "loss": 0.2402, "step": 1680 }, { "epoch": 0.7729247656071347, "grad_norm": 0.5609848536249358, "learning_rate": 8.536816260715433e-06, "loss": 0.2433, "step": 1690 }, { "epoch": 0.7774982849302539, "grad_norm": 0.5902179733155869, "learning_rate": 8.519679992675517e-06, "loss": 0.2512, "step": 1700 }, { "epoch": 0.782071804253373, "grad_norm": 0.5410282862801963, "learning_rate": 8.502461378396814e-06, "loss": 0.2427, "step": 1710 }, { "epoch": 0.7866453235764921, "grad_norm": 0.6206102476624096, "learning_rate": 8.485160820725073e-06, "loss": 0.2507, "step": 1720 }, { "epoch": 0.7912188428996112, "grad_norm": 0.543267722599167, "learning_rate": 8.467778724423187e-06, "loss": 0.2399, "step": 1730 }, { "epoch": 0.7957923622227304, "grad_norm": 0.5434834501740481, "learning_rate": 8.450315496161718e-06, "loss": 0.2408, "step": 1740 }, { "epoch": 0.8003658815458495, "grad_norm": 0.5209035150980255, "learning_rate": 8.432771544509395e-06, "loss": 0.2406, "step": 1750 }, { "epoch": 0.8049394008689686, "grad_norm": 0.5825233513191996, "learning_rate": 8.415147279923537e-06, "loss": 0.2391, "step": 1760 }, { "epoch": 0.8095129201920878, "grad_norm": 0.5481255147595794, "learning_rate": 8.397443114740468e-06, "loss": 0.2442, "step": 1770 }, { "epoch": 0.814086439515207, "grad_norm": 0.6383553114836046, "learning_rate": 8.37965946316586e-06, "loss": 0.2428, "step": 1780 }, { "epoch": 0.8186599588383261, "grad_norm": 0.5289327133427602, "learning_rate": 8.361796741265045e-06, "loss": 0.2469, "step": 1790 }, { "epoch": 0.8232334781614452, "grad_norm": 0.5860236705970534, "learning_rate": 8.343855366953274e-06, "loss": 0.2406, "step": 1800 }, { "epoch": 0.8278069974845643, "grad_norm": 0.579002203840679, "learning_rate": 8.325835759985951e-06, "loss": 0.2427, "step": 1810 }, { "epoch": 0.8323805168076835, "grad_norm": 0.5790553629814882, "learning_rate": 8.307738341948808e-06, "loss": 0.2409, "step": 1820 }, { "epoch": 0.8369540361308027, "grad_norm": 0.597914168089329, "learning_rate": 8.289563536248036e-06, "loss": 0.2388, "step": 1830 }, { "epoch": 0.8415275554539218, "grad_norm": 0.5541993395868017, "learning_rate": 8.271311768100386e-06, "loss": 0.2429, "step": 1840 }, { "epoch": 0.8461010747770409, "grad_norm": 0.5870358786680012, "learning_rate": 8.252983464523218e-06, "loss": 0.2452, "step": 1850 }, { "epoch": 0.8506745941001601, "grad_norm": 0.6100751944177363, "learning_rate": 8.234579054324512e-06, "loss": 0.2394, "step": 1860 }, { "epoch": 0.8552481134232792, "grad_norm": 0.5687300876471753, "learning_rate": 8.216098968092833e-06, "loss": 0.2395, "step": 1870 }, { "epoch": 0.8598216327463983, "grad_norm": 0.5858741218550929, "learning_rate": 8.197543638187257e-06, "loss": 0.2409, "step": 1880 }, { "epoch": 0.8643951520695174, "grad_norm": 0.5706662958141293, "learning_rate": 8.178913498727256e-06, "loss": 0.2434, "step": 1890 }, { "epoch": 0.8689686713926367, "grad_norm": 0.5898304202373802, "learning_rate": 8.160208985582547e-06, "loss": 0.2385, "step": 1900 }, { "epoch": 0.8735421907157558, "grad_norm": 0.6372196948869241, "learning_rate": 8.141430536362889e-06, "loss": 0.236, "step": 1910 }, { "epoch": 0.8781157100388749, "grad_norm": 0.5901657563321162, "learning_rate": 8.122578590407837e-06, "loss": 0.2392, "step": 1920 }, { "epoch": 0.882689229361994, "grad_norm": 0.6100768683532515, "learning_rate": 8.103653588776483e-06, "loss": 0.2367, "step": 1930 }, { "epoch": 0.8872627486851132, "grad_norm": 0.5620081742053228, "learning_rate": 8.084655974237117e-06, "loss": 0.2371, "step": 1940 }, { "epoch": 0.8918362680082323, "grad_norm": 0.6027836043481721, "learning_rate": 8.065586191256885e-06, "loss": 0.2401, "step": 1950 }, { "epoch": 0.8964097873313515, "grad_norm": 0.5803962392022427, "learning_rate": 8.046444685991369e-06, "loss": 0.2367, "step": 1960 }, { "epoch": 0.9009833066544706, "grad_norm": 0.5814896260843792, "learning_rate": 8.027231906274176e-06, "loss": 0.2357, "step": 1970 }, { "epoch": 0.9055568259775898, "grad_norm": 0.615956877721254, "learning_rate": 8.007948301606436e-06, "loss": 0.2378, "step": 1980 }, { "epoch": 0.9101303453007089, "grad_norm": 0.5821019285892404, "learning_rate": 7.9885943231463e-06, "loss": 0.2332, "step": 1990 }, { "epoch": 0.914703864623828, "grad_norm": 0.5968739162961995, "learning_rate": 7.96917042369838e-06, "loss": 0.2385, "step": 2000 }, { "epoch": 0.9192773839469471, "grad_norm": 0.6247052619384383, "learning_rate": 7.949677057703155e-06, "loss": 0.2384, "step": 2010 }, { "epoch": 0.9238509032700664, "grad_norm": 0.6270394642386049, "learning_rate": 7.930114681226341e-06, "loss": 0.2371, "step": 2020 }, { "epoch": 0.9284244225931855, "grad_norm": 0.5873320397582151, "learning_rate": 7.910483751948215e-06, "loss": 0.2407, "step": 2030 }, { "epoch": 0.9329979419163046, "grad_norm": 0.614039169994671, "learning_rate": 7.890784729152916e-06, "loss": 0.2338, "step": 2040 }, { "epoch": 0.9375714612394237, "grad_norm": 0.6715072397491144, "learning_rate": 7.871018073717693e-06, "loss": 0.2321, "step": 2050 }, { "epoch": 0.9421449805625429, "grad_norm": 0.5785883901363005, "learning_rate": 7.851184248102128e-06, "loss": 0.2275, "step": 2060 }, { "epoch": 0.946718499885662, "grad_norm": 0.6014023371650811, "learning_rate": 7.831283716337308e-06, "loss": 0.2332, "step": 2070 }, { "epoch": 0.9512920192087811, "grad_norm": 0.6376621220120618, "learning_rate": 7.811316944014974e-06, "loss": 0.2311, "step": 2080 }, { "epoch": 0.9558655385319003, "grad_norm": 0.601060047732152, "learning_rate": 7.791284398276632e-06, "loss": 0.2306, "step": 2090 }, { "epoch": 0.9604390578550195, "grad_norm": 0.6029896088697889, "learning_rate": 7.771186547802614e-06, "loss": 0.2318, "step": 2100 }, { "epoch": 0.9650125771781386, "grad_norm": 0.598952983912024, "learning_rate": 7.75102386280112e-06, "loss": 0.2362, "step": 2110 }, { "epoch": 0.9695860965012577, "grad_norm": 0.7077647222173107, "learning_rate": 7.730796814997216e-06, "loss": 0.2338, "step": 2120 }, { "epoch": 0.9741596158243768, "grad_norm": 0.6170110758708918, "learning_rate": 7.710505877621796e-06, "loss": 0.2313, "step": 2130 }, { "epoch": 0.978733135147496, "grad_norm": 0.6324578571750938, "learning_rate": 7.69015152540051e-06, "loss": 0.2305, "step": 2140 }, { "epoch": 0.9833066544706152, "grad_norm": 0.6064172890874445, "learning_rate": 7.669734234542657e-06, "loss": 0.2258, "step": 2150 }, { "epoch": 0.9878801737937343, "grad_norm": 0.6920282498203517, "learning_rate": 7.64925448273005e-06, "loss": 0.2302, "step": 2160 }, { "epoch": 0.9924536931168534, "grad_norm": 0.6294282211598161, "learning_rate": 7.628712749105831e-06, "loss": 0.2288, "step": 2170 }, { "epoch": 0.9970272124399726, "grad_norm": 0.6379211205965106, "learning_rate": 7.608109514263264e-06, "loss": 0.2253, "step": 2180 }, { "epoch": 1.000457351932312, "eval_ood_font_family_decon_loss": 0.3568974733352661, "eval_ood_font_family_decon_runtime": 749.8038, "eval_ood_font_family_decon_samples_per_second": 41.384, "eval_ood_font_family_decon_steps_per_second": 1.294, "step": 2188 }, { "epoch": 1.0013720557969357, "grad_norm": 0.6815633441189283, "learning_rate": 7.587445260234495e-06, "loss": 0.2129, "step": 2190 }, { "epoch": 1.0059455751200548, "grad_norm": 0.7205664647706543, "learning_rate": 7.5667204704792706e-06, "loss": 0.2056, "step": 2200 }, { "epoch": 1.0105190944431741, "grad_norm": 0.7171582271395964, "learning_rate": 7.545935629873626e-06, "loss": 0.2038, "step": 2210 }, { "epoch": 1.0150926137662932, "grad_norm": 0.6984306488103873, "learning_rate": 7.525091224698548e-06, "loss": 0.2052, "step": 2220 }, { "epoch": 1.0196661330894123, "grad_norm": 0.6831982992010361, "learning_rate": 7.5041877426285856e-06, "loss": 0.2042, "step": 2230 }, { "epoch": 1.0242396524125315, "grad_norm": 0.749243425626177, "learning_rate": 7.483225672720453e-06, "loss": 0.2063, "step": 2240 }, { "epoch": 1.0288131717356506, "grad_norm": 0.7312307841890472, "learning_rate": 7.462205505401581e-06, "loss": 0.2053, "step": 2250 }, { "epoch": 1.0333866910587697, "grad_norm": 0.6863372943708081, "learning_rate": 7.441127732458642e-06, "loss": 0.2041, "step": 2260 }, { "epoch": 1.0379602103818888, "grad_norm": 0.6591109188143379, "learning_rate": 7.41999284702605e-06, "loss": 0.1965, "step": 2270 }, { "epoch": 1.042533729705008, "grad_norm": 0.694679736735675, "learning_rate": 7.398801343574414e-06, "loss": 0.2027, "step": 2280 }, { "epoch": 1.0471072490281272, "grad_norm": 0.7241069304629978, "learning_rate": 7.377553717898983e-06, "loss": 0.2036, "step": 2290 }, { "epoch": 1.0516807683512464, "grad_norm": 0.724890937986387, "learning_rate": 7.3562504671080325e-06, "loss": 0.2033, "step": 2300 }, { "epoch": 1.0562542876743655, "grad_norm": 0.6786408527144132, "learning_rate": 7.334892089611243e-06, "loss": 0.2013, "step": 2310 }, { "epoch": 1.0608278069974846, "grad_norm": 0.6828122945169863, "learning_rate": 7.313479085108033e-06, "loss": 0.2022, "step": 2320 }, { "epoch": 1.0654013263206037, "grad_norm": 0.7357800820353438, "learning_rate": 7.292011954575875e-06, "loss": 0.1994, "step": 2330 }, { "epoch": 1.0699748456437228, "grad_norm": 0.6987114022791068, "learning_rate": 7.270491200258569e-06, "loss": 0.1986, "step": 2340 }, { "epoch": 1.074548364966842, "grad_norm": 0.731032296269081, "learning_rate": 7.2489173256544975e-06, "loss": 0.1986, "step": 2350 }, { "epoch": 1.079121884289961, "grad_norm": 0.7188922670349251, "learning_rate": 7.227290835504837e-06, "loss": 0.1996, "step": 2360 }, { "epoch": 1.0836954036130804, "grad_norm": 0.7703828376637508, "learning_rate": 7.205612235781756e-06, "loss": 0.1971, "step": 2370 }, { "epoch": 1.0882689229361995, "grad_norm": 0.7046881121065526, "learning_rate": 7.183882033676579e-06, "loss": 0.1968, "step": 2380 }, { "epoch": 1.0928424422593186, "grad_norm": 0.7137143862567326, "learning_rate": 7.162100737587915e-06, "loss": 0.1999, "step": 2390 }, { "epoch": 1.0974159615824377, "grad_norm": 0.7009472187220877, "learning_rate": 7.140268857109762e-06, "loss": 0.2018, "step": 2400 }, { "epoch": 1.1019894809055568, "grad_norm": 0.6865955407306371, "learning_rate": 7.118386903019594e-06, "loss": 0.1993, "step": 2410 }, { "epoch": 1.106563000228676, "grad_norm": 0.7040404730873923, "learning_rate": 7.0964553872663984e-06, "loss": 0.2002, "step": 2420 }, { "epoch": 1.111136519551795, "grad_norm": 0.7624913616273892, "learning_rate": 7.074474822958708e-06, "loss": 0.1981, "step": 2430 }, { "epoch": 1.1157100388749142, "grad_norm": 0.6945454515441686, "learning_rate": 7.05244572435259e-06, "loss": 0.1971, "step": 2440 }, { "epoch": 1.1202835581980333, "grad_norm": 0.7162264544488971, "learning_rate": 7.030368606839622e-06, "loss": 0.1954, "step": 2450 }, { "epoch": 1.1248570775211526, "grad_norm": 0.702400446001102, "learning_rate": 7.008243986934824e-06, "loss": 0.1989, "step": 2460 }, { "epoch": 1.1294305968442717, "grad_norm": 0.7554769898519392, "learning_rate": 6.9860723822645825e-06, "loss": 0.1983, "step": 2470 }, { "epoch": 1.1340041161673908, "grad_norm": 0.7364816551405189, "learning_rate": 6.963854311554537e-06, "loss": 0.1973, "step": 2480 }, { "epoch": 1.13857763549051, "grad_norm": 0.7980632859232769, "learning_rate": 6.941590294617443e-06, "loss": 0.1996, "step": 2490 }, { "epoch": 1.143151154813629, "grad_norm": 0.7781784931794704, "learning_rate": 6.919280852341011e-06, "loss": 0.2023, "step": 2500 }, { "epoch": 1.1477246741367482, "grad_norm": 0.696733209606982, "learning_rate": 6.89692650667572e-06, "loss": 0.1948, "step": 2510 }, { "epoch": 1.1522981934598673, "grad_norm": 0.7717868774849564, "learning_rate": 6.874527780622608e-06, "loss": 0.2014, "step": 2520 }, { "epoch": 1.1568717127829866, "grad_norm": 0.774844710708259, "learning_rate": 6.852085198221035e-06, "loss": 0.1953, "step": 2530 }, { "epoch": 1.1614452321061057, "grad_norm": 0.6877654774781731, "learning_rate": 6.829599284536416e-06, "loss": 0.1946, "step": 2540 }, { "epoch": 1.1660187514292248, "grad_norm": 0.7246326237313143, "learning_rate": 6.807070565647948e-06, "loss": 0.1936, "step": 2550 }, { "epoch": 1.170592270752344, "grad_norm": 0.7035896915884237, "learning_rate": 6.7844995686362955e-06, "loss": 0.1948, "step": 2560 }, { "epoch": 1.175165790075463, "grad_norm": 0.7751464210172985, "learning_rate": 6.761886821571258e-06, "loss": 0.1949, "step": 2570 }, { "epoch": 1.1797393093985822, "grad_norm": 0.7190582876048454, "learning_rate": 6.739232853499414e-06, "loss": 0.1968, "step": 2580 }, { "epoch": 1.1843128287217013, "grad_norm": 0.8028096308649686, "learning_rate": 6.716538194431754e-06, "loss": 0.1947, "step": 2590 }, { "epoch": 1.1888863480448204, "grad_norm": 0.7693363647978207, "learning_rate": 6.693803375331268e-06, "loss": 0.1973, "step": 2600 }, { "epoch": 1.1934598673679395, "grad_norm": 0.8259128300391488, "learning_rate": 6.671028928100527e-06, "loss": 0.1927, "step": 2610 }, { "epoch": 1.1980333866910589, "grad_norm": 0.7440187526802534, "learning_rate": 6.6482153855692395e-06, "loss": 0.1935, "step": 2620 }, { "epoch": 1.202606906014178, "grad_norm": 0.7676595272617575, "learning_rate": 6.625363281481792e-06, "loss": 0.1921, "step": 2630 }, { "epoch": 1.207180425337297, "grad_norm": 0.7407650033366908, "learning_rate": 6.602473150484746e-06, "loss": 0.1897, "step": 2640 }, { "epoch": 1.2117539446604162, "grad_norm": 0.709294518403159, "learning_rate": 6.579545528114344e-06, "loss": 0.1912, "step": 2650 }, { "epoch": 1.2163274639835353, "grad_norm": 0.7991512178588244, "learning_rate": 6.556580950783972e-06, "loss": 0.1943, "step": 2660 }, { "epoch": 1.2209009833066544, "grad_norm": 0.777495969024946, "learning_rate": 6.5335799557716155e-06, "loss": 0.1912, "step": 2670 }, { "epoch": 1.2254745026297735, "grad_norm": 0.77672960394565, "learning_rate": 6.510543081207281e-06, "loss": 0.1909, "step": 2680 }, { "epoch": 1.2300480219528929, "grad_norm": 0.7545305820868883, "learning_rate": 6.487470866060417e-06, "loss": 0.189, "step": 2690 }, { "epoch": 1.234621541276012, "grad_norm": 0.7375972891474601, "learning_rate": 6.464363850127292e-06, "loss": 0.1921, "step": 2700 }, { "epoch": 1.239195060599131, "grad_norm": 0.8281145828603478, "learning_rate": 6.441222574018378e-06, "loss": 0.1911, "step": 2710 }, { "epoch": 1.2437685799222502, "grad_norm": 0.7642874229638403, "learning_rate": 6.418047579145693e-06, "loss": 0.1919, "step": 2720 }, { "epoch": 1.2483420992453693, "grad_norm": 0.826032279847095, "learning_rate": 6.394839407710135e-06, "loss": 0.197, "step": 2730 }, { "epoch": 1.2529156185684884, "grad_norm": 0.7578189916801957, "learning_rate": 6.3715986026888046e-06, "loss": 0.1889, "step": 2740 }, { "epoch": 1.2574891378916075, "grad_norm": 0.7122673327338493, "learning_rate": 6.348325707822291e-06, "loss": 0.1898, "step": 2750 }, { "epoch": 1.2620626572147267, "grad_norm": 0.7387483304222602, "learning_rate": 6.325021267601958e-06, "loss": 0.193, "step": 2760 }, { "epoch": 1.2666361765378458, "grad_norm": 0.7809082128710795, "learning_rate": 6.301685827257202e-06, "loss": 0.1934, "step": 2770 }, { "epoch": 1.271209695860965, "grad_norm": 0.7416987368055662, "learning_rate": 6.278319932742699e-06, "loss": 0.1893, "step": 2780 }, { "epoch": 1.2757832151840842, "grad_norm": 0.7662362927432017, "learning_rate": 6.2549241307256244e-06, "loss": 0.188, "step": 2790 }, { "epoch": 1.2803567345072033, "grad_norm": 0.7389560569252562, "learning_rate": 6.231498968572872e-06, "loss": 0.1877, "step": 2800 }, { "epoch": 1.2849302538303224, "grad_norm": 0.8358426547156163, "learning_rate": 6.2080449943382415e-06, "loss": 0.1939, "step": 2810 }, { "epoch": 1.2895037731534416, "grad_norm": 0.776493921788284, "learning_rate": 6.18456275674962e-06, "loss": 0.187, "step": 2820 }, { "epoch": 1.2940772924765607, "grad_norm": 0.7629670951400791, "learning_rate": 6.161052805196141e-06, "loss": 0.1924, "step": 2830 }, { "epoch": 1.2986508117996798, "grad_norm": 0.8306963378544165, "learning_rate": 6.1375156897153335e-06, "loss": 0.1874, "step": 2840 }, { "epoch": 1.3032243311227991, "grad_norm": 0.7636008243237192, "learning_rate": 6.11395196098025e-06, "loss": 0.1881, "step": 2850 }, { "epoch": 1.307797850445918, "grad_norm": 0.759659299142774, "learning_rate": 6.090362170286591e-06, "loss": 0.1911, "step": 2860 }, { "epoch": 1.3123713697690373, "grad_norm": 0.6774511744348428, "learning_rate": 6.066746869539793e-06, "loss": 0.1864, "step": 2870 }, { "epoch": 1.3169448890921565, "grad_norm": 0.7951908827179025, "learning_rate": 6.043106611242131e-06, "loss": 0.1865, "step": 2880 }, { "epoch": 1.3215184084152756, "grad_norm": 0.7786344234669238, "learning_rate": 6.019441948479784e-06, "loss": 0.1858, "step": 2890 }, { "epoch": 1.3260919277383947, "grad_norm": 0.7980040185479854, "learning_rate": 5.995753434909891e-06, "loss": 0.1836, "step": 2900 }, { "epoch": 1.3306654470615138, "grad_norm": 0.7918664885583329, "learning_rate": 5.972041624747611e-06, "loss": 0.1825, "step": 2910 }, { "epoch": 1.335238966384633, "grad_norm": 0.7815480118030956, "learning_rate": 5.948307072753146e-06, "loss": 0.1869, "step": 2920 }, { "epoch": 1.339812485707752, "grad_norm": 0.7239080631668606, "learning_rate": 5.924550334218762e-06, "loss": 0.1852, "step": 2930 }, { "epoch": 1.3443860050308714, "grad_norm": 0.7792482012027093, "learning_rate": 5.900771964955803e-06, "loss": 0.1813, "step": 2940 }, { "epoch": 1.3489595243539905, "grad_norm": 0.7402858994057647, "learning_rate": 5.876972521281683e-06, "loss": 0.187, "step": 2950 }, { "epoch": 1.3535330436771096, "grad_norm": 0.7890177541677389, "learning_rate": 5.853152560006874e-06, "loss": 0.1831, "step": 2960 }, { "epoch": 1.3581065630002287, "grad_norm": 0.7636409677823, "learning_rate": 5.829312638421872e-06, "loss": 0.1828, "step": 2970 }, { "epoch": 1.3626800823233478, "grad_norm": 0.713265035638444, "learning_rate": 5.805453314284168e-06, "loss": 0.1786, "step": 2980 }, { "epoch": 1.367253601646467, "grad_norm": 0.7923096852757117, "learning_rate": 5.781575145805193e-06, "loss": 0.1799, "step": 2990 }, { "epoch": 1.371827120969586, "grad_norm": 0.7868947481545041, "learning_rate": 5.7576786916372574e-06, "loss": 0.1886, "step": 3000 }, { "epoch": 1.3764006402927054, "grad_norm": 0.7887694153707763, "learning_rate": 5.733764510860482e-06, "loss": 0.1865, "step": 3010 }, { "epoch": 1.3809741596158243, "grad_norm": 0.761292945338406, "learning_rate": 5.709833162969724e-06, "loss": 0.1807, "step": 3020 }, { "epoch": 1.3855476789389436, "grad_norm": 0.7535044748560811, "learning_rate": 5.685885207861473e-06, "loss": 0.182, "step": 3030 }, { "epoch": 1.3901211982620627, "grad_norm": 0.7792606083271629, "learning_rate": 5.661921205820767e-06, "loss": 0.177, "step": 3040 }, { "epoch": 1.3946947175851818, "grad_norm": 0.8396601720623091, "learning_rate": 5.637941717508075e-06, "loss": 0.1832, "step": 3050 }, { "epoch": 1.399268236908301, "grad_norm": 0.8044384652922599, "learning_rate": 5.61394730394618e-06, "loss": 0.1804, "step": 3060 }, { "epoch": 1.40384175623142, "grad_norm": 0.7775277584517364, "learning_rate": 5.589938526507059e-06, "loss": 0.1808, "step": 3070 }, { "epoch": 1.4084152755545392, "grad_norm": 0.8384859076249401, "learning_rate": 5.565915946898741e-06, "loss": 0.1777, "step": 3080 }, { "epoch": 1.4129887948776583, "grad_norm": 0.8224777811169652, "learning_rate": 5.541880127152175e-06, "loss": 0.1789, "step": 3090 }, { "epoch": 1.4175623142007776, "grad_norm": 0.7886140569076497, "learning_rate": 5.51783162960807e-06, "loss": 0.1768, "step": 3100 }, { "epoch": 1.4221358335238967, "grad_norm": 0.8846604786290522, "learning_rate": 5.493771016903749e-06, "loss": 0.1805, "step": 3110 }, { "epoch": 1.4267093528470158, "grad_norm": 0.7433132346237088, "learning_rate": 5.469698851959977e-06, "loss": 0.1792, "step": 3120 }, { "epoch": 1.431282872170135, "grad_norm": 0.7925722129720355, "learning_rate": 5.445615697967797e-06, "loss": 0.1797, "step": 3130 }, { "epoch": 1.435856391493254, "grad_norm": 0.7867951136877106, "learning_rate": 5.42152211837535e-06, "loss": 0.1778, "step": 3140 }, { "epoch": 1.4404299108163732, "grad_norm": 0.7772200446295263, "learning_rate": 5.3974186768746926e-06, "loss": 0.1806, "step": 3150 }, { "epoch": 1.4450034301394923, "grad_norm": 0.7800798016157442, "learning_rate": 5.373305937388613e-06, "loss": 0.1783, "step": 3160 }, { "epoch": 1.4495769494626114, "grad_norm": 0.7510530165529571, "learning_rate": 5.349184464057433e-06, "loss": 0.1772, "step": 3170 }, { "epoch": 1.4541504687857305, "grad_norm": 0.8544878974732155, "learning_rate": 5.325054821225811e-06, "loss": 0.1779, "step": 3180 }, { "epoch": 1.4587239881088498, "grad_norm": 0.8956067830558161, "learning_rate": 5.300917573429536e-06, "loss": 0.1793, "step": 3190 }, { "epoch": 1.463297507431969, "grad_norm": 0.7950477965606637, "learning_rate": 5.276773285382327e-06, "loss": 0.1784, "step": 3200 }, { "epoch": 1.467871026755088, "grad_norm": 0.8205638938207572, "learning_rate": 5.252622521962615e-06, "loss": 0.1789, "step": 3210 }, { "epoch": 1.4724445460782072, "grad_norm": 0.8207015856871955, "learning_rate": 5.228465848200327e-06, "loss": 0.1752, "step": 3220 }, { "epoch": 1.4770180654013263, "grad_norm": 0.7971420075843006, "learning_rate": 5.204303829263666e-06, "loss": 0.1791, "step": 3230 }, { "epoch": 1.4815915847244454, "grad_norm": 0.8285904975398035, "learning_rate": 5.180137030445896e-06, "loss": 0.176, "step": 3240 }, { "epoch": 1.4861651040475645, "grad_norm": 0.8093573220964169, "learning_rate": 5.155966017152108e-06, "loss": 0.1793, "step": 3250 }, { "epoch": 1.4907386233706839, "grad_norm": 0.8811000604222163, "learning_rate": 5.131791354885992e-06, "loss": 0.1712, "step": 3260 }, { "epoch": 1.4953121426938027, "grad_norm": 0.8140045627487338, "learning_rate": 5.107613609236613e-06, "loss": 0.171, "step": 3270 }, { "epoch": 1.499885662016922, "grad_norm": 0.8224221032532308, "learning_rate": 5.083433345865175e-06, "loss": 0.1733, "step": 3280 }, { "epoch": 1.5008003658815459, "eval_ood_font_family_decon_loss": 0.372624009847641, "eval_ood_font_family_decon_runtime": 746.7835, "eval_ood_font_family_decon_samples_per_second": 41.552, "eval_ood_font_family_decon_steps_per_second": 1.299, "step": 3282 }, { "epoch": 1.5044591813400412, "grad_norm": 0.8197841996584896, "learning_rate": 5.059251130491784e-06, "loss": 0.1744, "step": 3290 }, { "epoch": 1.5090327006631603, "grad_norm": 0.9145784591909878, "learning_rate": 5.035067528882215e-06, "loss": 0.1797, "step": 3300 }, { "epoch": 1.5136062199862794, "grad_norm": 0.8284458776892752, "learning_rate": 5.010883106834676e-06, "loss": 0.1765, "step": 3310 }, { "epoch": 1.5181797393093985, "grad_norm": 0.8423517434051327, "learning_rate": 4.9866984301665695e-06, "loss": 0.1729, "step": 3320 }, { "epoch": 1.5227532586325179, "grad_norm": 0.8682905493267522, "learning_rate": 4.962514064701257e-06, "loss": 0.172, "step": 3330 }, { "epoch": 1.5273267779556368, "grad_norm": 0.8998491356884444, "learning_rate": 4.938330576254817e-06, "loss": 0.1737, "step": 3340 }, { "epoch": 1.531900297278756, "grad_norm": 0.8332404290293175, "learning_rate": 4.9141485306228095e-06, "loss": 0.1732, "step": 3350 }, { "epoch": 1.536473816601875, "grad_norm": 0.8102682295821275, "learning_rate": 4.889968493567041e-06, "loss": 0.1718, "step": 3360 }, { "epoch": 1.5410473359249943, "grad_norm": 0.7673672884921555, "learning_rate": 4.8657910308023205e-06, "loss": 0.1644, "step": 3370 }, { "epoch": 1.5456208552481134, "grad_norm": 0.8169999820852306, "learning_rate": 4.841616707983237e-06, "loss": 0.1672, "step": 3380 }, { "epoch": 1.5501943745712325, "grad_norm": 0.8833831939516149, "learning_rate": 4.817446090690907e-06, "loss": 0.1719, "step": 3390 }, { "epoch": 1.5547678938943517, "grad_norm": 0.8558906889434255, "learning_rate": 4.7932797444197604e-06, "loss": 0.1729, "step": 3400 }, { "epoch": 1.5593414132174708, "grad_norm": 0.876562761803751, "learning_rate": 4.7691182345643e-06, "loss": 0.1718, "step": 3410 }, { "epoch": 1.56391493254059, "grad_norm": 0.8634146400078141, "learning_rate": 4.744962126405877e-06, "loss": 0.1701, "step": 3420 }, { "epoch": 1.568488451863709, "grad_norm": 0.8212184566232759, "learning_rate": 4.720811985099464e-06, "loss": 0.1705, "step": 3430 }, { "epoch": 1.5730619711868283, "grad_norm": 0.9333234613526586, "learning_rate": 4.696668375660434e-06, "loss": 0.1712, "step": 3440 }, { "epoch": 1.5776354905099474, "grad_norm": 0.8058227316715402, "learning_rate": 4.672531862951342e-06, "loss": 0.169, "step": 3450 }, { "epoch": 1.5822090098330666, "grad_norm": 0.9293432568343326, "learning_rate": 4.6484030116687014e-06, "loss": 0.1718, "step": 3460 }, { "epoch": 1.5867825291561857, "grad_norm": 0.8215724066831221, "learning_rate": 4.62428238632979e-06, "loss": 0.1663, "step": 3470 }, { "epoch": 1.5913560484793048, "grad_norm": 0.8000701432638395, "learning_rate": 4.600170551259424e-06, "loss": 0.1719, "step": 3480 }, { "epoch": 1.5959295678024241, "grad_norm": 0.8119876180605534, "learning_rate": 4.5760680705767665e-06, "loss": 0.1657, "step": 3490 }, { "epoch": 1.600503087125543, "grad_norm": 0.8136454309242126, "learning_rate": 4.551975508182123e-06, "loss": 0.1725, "step": 3500 }, { "epoch": 1.6050766064486623, "grad_norm": 0.8439575671483175, "learning_rate": 4.527893427743757e-06, "loss": 0.1633, "step": 3510 }, { "epoch": 1.6096501257717812, "grad_norm": 0.9524066574065465, "learning_rate": 4.5038223926846905e-06, "loss": 0.1703, "step": 3520 }, { "epoch": 1.6142236450949006, "grad_norm": 0.8577766429700633, "learning_rate": 4.479762966169531e-06, "loss": 0.165, "step": 3530 }, { "epoch": 1.6187971644180197, "grad_norm": 0.8382093463636254, "learning_rate": 4.455715711091291e-06, "loss": 0.1658, "step": 3540 }, { "epoch": 1.6233706837411388, "grad_norm": 0.883262297823796, "learning_rate": 4.431681190058224e-06, "loss": 0.1682, "step": 3550 }, { "epoch": 1.627944203064258, "grad_norm": 0.8747400480833419, "learning_rate": 4.40765996538066e-06, "loss": 0.1696, "step": 3560 }, { "epoch": 1.632517722387377, "grad_norm": 0.8992821865626748, "learning_rate": 4.38365259905784e-06, "loss": 0.1677, "step": 3570 }, { "epoch": 1.6370912417104964, "grad_norm": 0.8427359332210314, "learning_rate": 4.359659652764786e-06, "loss": 0.1611, "step": 3580 }, { "epoch": 1.6416647610336152, "grad_norm": 0.8864317670972008, "learning_rate": 4.335681687839143e-06, "loss": 0.1574, "step": 3590 }, { "epoch": 1.6462382803567346, "grad_norm": 0.8402998912917335, "learning_rate": 4.311719265268054e-06, "loss": 0.1623, "step": 3600 }, { "epoch": 1.6508117996798537, "grad_norm": 0.8906323031144928, "learning_rate": 4.287772945675035e-06, "loss": 0.1669, "step": 3610 }, { "epoch": 1.6553853190029728, "grad_norm": 0.8022548208757941, "learning_rate": 4.263843289306856e-06, "loss": 0.1569, "step": 3620 }, { "epoch": 1.659958838326092, "grad_norm": 0.8520148413610501, "learning_rate": 4.239930856020437e-06, "loss": 0.1644, "step": 3630 }, { "epoch": 1.664532357649211, "grad_norm": 0.9114171777586244, "learning_rate": 4.216036205269748e-06, "loss": 0.1686, "step": 3640 }, { "epoch": 1.6691058769723304, "grad_norm": 0.8989660056443852, "learning_rate": 4.1921598960927176e-06, "loss": 0.1628, "step": 3650 }, { "epoch": 1.6736793962954493, "grad_norm": 0.8572031955674774, "learning_rate": 4.168302487098158e-06, "loss": 0.1622, "step": 3660 }, { "epoch": 1.6782529156185686, "grad_norm": 0.9195827783415927, "learning_rate": 4.144464536452693e-06, "loss": 0.1627, "step": 3670 }, { "epoch": 1.6828264349416875, "grad_norm": 0.8970543339314866, "learning_rate": 4.120646601867698e-06, "loss": 0.1648, "step": 3680 }, { "epoch": 1.6873999542648068, "grad_norm": 0.8549835197238776, "learning_rate": 4.096849240586256e-06, "loss": 0.164, "step": 3690 }, { "epoch": 1.691973473587926, "grad_norm": 0.8456074107159487, "learning_rate": 4.0730730093701185e-06, "loss": 0.1606, "step": 3700 }, { "epoch": 1.696546992911045, "grad_norm": 0.926097468465809, "learning_rate": 4.0493184644866784e-06, "loss": 0.158, "step": 3710 }, { "epoch": 1.7011205122341642, "grad_norm": 0.9191243633568398, "learning_rate": 4.025586161695956e-06, "loss": 0.1621, "step": 3720 }, { "epoch": 1.7056940315572833, "grad_norm": 1.0525809458003057, "learning_rate": 4.0018766562375984e-06, "loss": 0.1577, "step": 3730 }, { "epoch": 1.7102675508804026, "grad_norm": 0.8329749900477065, "learning_rate": 3.978190502817886e-06, "loss": 0.1584, "step": 3740 }, { "epoch": 1.7148410702035215, "grad_norm": 0.8532879582212549, "learning_rate": 3.954528255596754e-06, "loss": 0.1591, "step": 3750 }, { "epoch": 1.7194145895266408, "grad_norm": 0.8529356288421338, "learning_rate": 3.930890468174833e-06, "loss": 0.1587, "step": 3760 }, { "epoch": 1.7239881088497597, "grad_norm": 0.9065173051949476, "learning_rate": 3.907277693580493e-06, "loss": 0.1547, "step": 3770 }, { "epoch": 1.728561628172879, "grad_norm": 0.8577730693913589, "learning_rate": 3.8836904842569055e-06, "loss": 0.153, "step": 3780 }, { "epoch": 1.7331351474959982, "grad_norm": 0.8844608250121475, "learning_rate": 3.8601293920491165e-06, "loss": 0.1634, "step": 3790 }, { "epoch": 1.7377086668191173, "grad_norm": 0.978954885991917, "learning_rate": 3.83659496819114e-06, "loss": 0.156, "step": 3800 }, { "epoch": 1.7422821861422364, "grad_norm": 0.8622196060764703, "learning_rate": 3.8130877632930553e-06, "loss": 0.1592, "step": 3810 }, { "epoch": 1.7468557054653555, "grad_norm": 0.8742828857709949, "learning_rate": 3.7896083273281324e-06, "loss": 0.1547, "step": 3820 }, { "epoch": 1.7514292247884748, "grad_norm": 0.8802753411403145, "learning_rate": 3.7661572096199567e-06, "loss": 0.1552, "step": 3830 }, { "epoch": 1.7560027441115937, "grad_norm": 0.9122018453250995, "learning_rate": 3.7427349588295842e-06, "loss": 0.1531, "step": 3840 }, { "epoch": 1.760576263434713, "grad_norm": 0.9219011118492411, "learning_rate": 3.7193421229427017e-06, "loss": 0.1583, "step": 3850 }, { "epoch": 1.7651497827578322, "grad_norm": 1.019257866929455, "learning_rate": 3.6959792492568054e-06, "loss": 0.1584, "step": 3860 }, { "epoch": 1.7697233020809513, "grad_norm": 0.9638988264774837, "learning_rate": 3.6726468843683992e-06, "loss": 0.1558, "step": 3870 }, { "epoch": 1.7742968214040704, "grad_norm": 0.9630409979747435, "learning_rate": 3.6493455741602035e-06, "loss": 0.1582, "step": 3880 }, { "epoch": 1.7788703407271895, "grad_norm": 1.0740891540624595, "learning_rate": 3.6260758637883853e-06, "loss": 0.1525, "step": 3890 }, { "epoch": 1.7834438600503089, "grad_norm": 1.0049597156017656, "learning_rate": 3.6028382976698027e-06, "loss": 0.153, "step": 3900 }, { "epoch": 1.7880173793734278, "grad_norm": 0.9408157847189943, "learning_rate": 3.5796334194692704e-06, "loss": 0.1466, "step": 3910 }, { "epoch": 1.792590898696547, "grad_norm": 0.9052852408187713, "learning_rate": 3.5564617720868382e-06, "loss": 0.152, "step": 3920 }, { "epoch": 1.797164418019666, "grad_norm": 0.9215909670938703, "learning_rate": 3.533323897645089e-06, "loss": 0.1517, "step": 3930 }, { "epoch": 1.8017379373427853, "grad_norm": 0.8984634496383483, "learning_rate": 3.5102203374764555e-06, "loss": 0.1564, "step": 3940 }, { "epoch": 1.8063114566659044, "grad_norm": 0.9217354188233934, "learning_rate": 3.4871516321105553e-06, "loss": 0.1538, "step": 3950 }, { "epoch": 1.8108849759890235, "grad_norm": 0.9609580906401771, "learning_rate": 3.464118321261547e-06, "loss": 0.1496, "step": 3960 }, { "epoch": 1.8154584953121427, "grad_norm": 0.867198591716571, "learning_rate": 3.441120943815497e-06, "loss": 0.1475, "step": 3970 }, { "epoch": 1.8200320146352618, "grad_norm": 1.0250004819635927, "learning_rate": 3.4181600378177806e-06, "loss": 0.1566, "step": 3980 }, { "epoch": 1.824605533958381, "grad_norm": 0.8815355121482957, "learning_rate": 3.395236140460486e-06, "loss": 0.1517, "step": 3990 }, { "epoch": 1.8291790532815, "grad_norm": 1.0146920230720184, "learning_rate": 3.37234978806985e-06, "loss": 0.1491, "step": 4000 }, { "epoch": 1.8337525726046193, "grad_norm": 0.9520772056618201, "learning_rate": 3.3495015160937105e-06, "loss": 0.1483, "step": 4010 }, { "epoch": 1.8383260919277384, "grad_norm": 0.9515376909523551, "learning_rate": 3.3266918590889762e-06, "loss": 0.1462, "step": 4020 }, { "epoch": 1.8428996112508576, "grad_norm": 1.0151323493239146, "learning_rate": 3.303921350709124e-06, "loss": 0.1489, "step": 4030 }, { "epoch": 1.8474731305739767, "grad_norm": 0.969394079902149, "learning_rate": 3.2811905236917096e-06, "loss": 0.1507, "step": 4040 }, { "epoch": 1.8520466498970958, "grad_norm": 0.8805485332639026, "learning_rate": 3.258499909845907e-06, "loss": 0.1492, "step": 4050 }, { "epoch": 1.856620169220215, "grad_norm": 0.9700676081741338, "learning_rate": 3.235850040040066e-06, "loss": 0.1454, "step": 4060 }, { "epoch": 1.861193688543334, "grad_norm": 1.036147454260966, "learning_rate": 3.2132414441892894e-06, "loss": 0.1495, "step": 4070 }, { "epoch": 1.8657672078664533, "grad_norm": 0.9744940875072222, "learning_rate": 3.1906746512430386e-06, "loss": 0.1484, "step": 4080 }, { "epoch": 1.8703407271895722, "grad_norm": 0.9472605821855146, "learning_rate": 3.168150189172754e-06, "loss": 0.149, "step": 4090 }, { "epoch": 1.8749142465126916, "grad_norm": 1.0784513773809161, "learning_rate": 3.145668584959507e-06, "loss": 0.1446, "step": 4100 }, { "epoch": 1.8794877658358107, "grad_norm": 0.9937236059488981, "learning_rate": 3.12323036458167e-06, "loss": 0.1511, "step": 4110 }, { "epoch": 1.8840612851589298, "grad_norm": 0.8975385713545477, "learning_rate": 3.1008360530026053e-06, "loss": 0.1435, "step": 4120 }, { "epoch": 1.888634804482049, "grad_norm": 1.0009763404147596, "learning_rate": 3.0784861741583925e-06, "loss": 0.1453, "step": 4130 }, { "epoch": 1.893208323805168, "grad_norm": 0.9596483229481366, "learning_rate": 3.0561812509455624e-06, "loss": 0.1455, "step": 4140 }, { "epoch": 1.8977818431282873, "grad_norm": 1.033716885249794, "learning_rate": 3.033921805208867e-06, "loss": 0.1467, "step": 4150 }, { "epoch": 1.9023553624514062, "grad_norm": 1.0473617860119497, "learning_rate": 3.01170835772907e-06, "loss": 0.1471, "step": 4160 }, { "epoch": 1.9069288817745256, "grad_norm": 0.961067600916794, "learning_rate": 2.9895414282107616e-06, "loss": 0.1506, "step": 4170 }, { "epoch": 1.9115024010976447, "grad_norm": 1.0167513304853366, "learning_rate": 2.967421535270203e-06, "loss": 0.1422, "step": 4180 }, { "epoch": 1.9160759204207638, "grad_norm": 1.0191511524310475, "learning_rate": 2.945349196423185e-06, "loss": 0.1467, "step": 4190 }, { "epoch": 1.920649439743883, "grad_norm": 1.061740372307547, "learning_rate": 2.923324928072931e-06, "loss": 0.1435, "step": 4200 }, { "epoch": 1.925222959067002, "grad_norm": 1.0753532128754477, "learning_rate": 2.9013492454980074e-06, "loss": 0.1414, "step": 4210 }, { "epoch": 1.9297964783901214, "grad_norm": 1.0132158849557145, "learning_rate": 2.879422662840268e-06, "loss": 0.143, "step": 4220 }, { "epoch": 1.9343699977132403, "grad_norm": 1.0681270915992913, "learning_rate": 2.857545693092833e-06, "loss": 0.1395, "step": 4230 }, { "epoch": 1.9389435170363596, "grad_norm": 1.008136275548994, "learning_rate": 2.835718848088076e-06, "loss": 0.1449, "step": 4240 }, { "epoch": 1.9435170363594785, "grad_norm": 0.9735711764156847, "learning_rate": 2.813942638485664e-06, "loss": 0.1392, "step": 4250 }, { "epoch": 1.9480905556825978, "grad_norm": 1.0789582806730293, "learning_rate": 2.792217573760585e-06, "loss": 0.1413, "step": 4260 }, { "epoch": 1.952664075005717, "grad_norm": 1.0899246182469222, "learning_rate": 2.770544162191261e-06, "loss": 0.1442, "step": 4270 }, { "epoch": 1.957237594328836, "grad_norm": 0.9447108283460726, "learning_rate": 2.7489229108476288e-06, "loss": 0.1422, "step": 4280 }, { "epoch": 1.9618111136519552, "grad_norm": 1.091174024276081, "learning_rate": 2.7273543255792945e-06, "loss": 0.1427, "step": 4290 }, { "epoch": 1.9663846329750743, "grad_norm": 1.097498133935512, "learning_rate": 2.7058389110036835e-06, "loss": 0.1378, "step": 4300 }, { "epoch": 1.9709581522981936, "grad_norm": 1.109480545961287, "learning_rate": 2.6843771704942527e-06, "loss": 0.1369, "step": 4310 }, { "epoch": 1.9755316716213125, "grad_norm": 1.0224217884201956, "learning_rate": 2.662969606168696e-06, "loss": 0.1363, "step": 4320 }, { "epoch": 1.9801051909444318, "grad_norm": 0.9388762721436228, "learning_rate": 2.6416167188772052e-06, "loss": 0.1411, "step": 4330 }, { "epoch": 1.9846787102675507, "grad_norm": 1.057455785488616, "learning_rate": 2.620319008190752e-06, "loss": 0.1381, "step": 4340 }, { "epoch": 1.98925222959067, "grad_norm": 1.0123934951246434, "learning_rate": 2.5990769723894028e-06, "loss": 0.1374, "step": 4350 }, { "epoch": 1.9938257489137892, "grad_norm": 0.994918537286174, "learning_rate": 2.577891108450651e-06, "loss": 0.1374, "step": 4360 }, { "epoch": 1.9983992682369083, "grad_norm": 1.0904640268650352, "learning_rate": 2.556761912037803e-06, "loss": 0.1313, "step": 4370 }, { "epoch": 2.000914703864624, "eval_ood_font_family_decon_loss": 0.41473495960235596, "eval_ood_font_family_decon_runtime": 748.4306, "eval_ood_font_family_decon_samples_per_second": 41.46, "eval_ood_font_family_decon_steps_per_second": 1.296, "step": 4376 }, { "epoch": 2.0027441115938713, "grad_norm": 1.077769914169416, "learning_rate": 2.535689877488369e-06, "loss": 0.102, "step": 4380 }, { "epoch": 2.0073176309169907, "grad_norm": 1.228800406844578, "learning_rate": 2.514675497802508e-06, "loss": 0.0969, "step": 4390 }, { "epoch": 2.0118911502401096, "grad_norm": 1.097130289981402, "learning_rate": 2.4937192646314855e-06, "loss": 0.0911, "step": 4400 }, { "epoch": 2.016464669563229, "grad_norm": 1.269351418913015, "learning_rate": 2.4728216682661764e-06, "loss": 0.0923, "step": 4410 }, { "epoch": 2.0210381888863482, "grad_norm": 1.1053772963178847, "learning_rate": 2.4519831976255892e-06, "loss": 0.09, "step": 4420 }, { "epoch": 2.025611708209467, "grad_norm": 1.2709581231354397, "learning_rate": 2.4312043402454365e-06, "loss": 0.0903, "step": 4430 }, { "epoch": 2.0301852275325865, "grad_norm": 1.0230237831171516, "learning_rate": 2.410485582266715e-06, "loss": 0.0932, "step": 4440 }, { "epoch": 2.0347587468557053, "grad_norm": 1.0914375922455184, "learning_rate": 2.389827408424345e-06, "loss": 0.0904, "step": 4450 }, { "epoch": 2.0393322661788247, "grad_norm": 1.056633082616366, "learning_rate": 2.3692303020358186e-06, "loss": 0.0915, "step": 4460 }, { "epoch": 2.0439057855019436, "grad_norm": 1.1780389128015578, "learning_rate": 2.3486947449899046e-06, "loss": 0.0919, "step": 4470 }, { "epoch": 2.048479304825063, "grad_norm": 1.1626812446744246, "learning_rate": 2.328221217735355e-06, "loss": 0.0947, "step": 4480 }, { "epoch": 2.0530528241481822, "grad_norm": 1.2120052079097106, "learning_rate": 2.3078101992696877e-06, "loss": 0.0912, "step": 4490 }, { "epoch": 2.057626343471301, "grad_norm": 1.103358158704817, "learning_rate": 2.287462167127962e-06, "loss": 0.0877, "step": 4500 }, { "epoch": 2.0621998627944205, "grad_norm": 1.0477004112849235, "learning_rate": 2.267177597371616e-06, "loss": 0.0909, "step": 4510 }, { "epoch": 2.0667733821175394, "grad_norm": 1.117237940674604, "learning_rate": 2.246956964577321e-06, "loss": 0.0919, "step": 4520 }, { "epoch": 2.0713469014406587, "grad_norm": 1.1463332188144963, "learning_rate": 2.2268007418258898e-06, "loss": 0.093, "step": 4530 }, { "epoch": 2.0759204207637776, "grad_norm": 1.2437379477439128, "learning_rate": 2.2067094006911943e-06, "loss": 0.0916, "step": 4540 }, { "epoch": 2.080493940086897, "grad_norm": 1.1343182422756457, "learning_rate": 2.186683411229143e-06, "loss": 0.0895, "step": 4550 }, { "epoch": 2.085067459410016, "grad_norm": 1.1459243493332056, "learning_rate": 2.166723241966678e-06, "loss": 0.091, "step": 4560 }, { "epoch": 2.089640978733135, "grad_norm": 1.1793232002449703, "learning_rate": 2.14682935989082e-06, "loss": 0.0864, "step": 4570 }, { "epoch": 2.0942144980562545, "grad_norm": 1.1356758760502546, "learning_rate": 2.1270022304377335e-06, "loss": 0.0891, "step": 4580 }, { "epoch": 2.0987880173793734, "grad_norm": 1.2238618947022213, "learning_rate": 2.107242317481847e-06, "loss": 0.0886, "step": 4590 }, { "epoch": 2.1033615367024927, "grad_norm": 1.0369021390545177, "learning_rate": 2.08755008332499e-06, "loss": 0.0885, "step": 4600 }, { "epoch": 2.1079350560256116, "grad_norm": 1.1666993818658289, "learning_rate": 2.0679259886855914e-06, "loss": 0.0855, "step": 4610 }, { "epoch": 2.112508575348731, "grad_norm": 1.2169827831787545, "learning_rate": 2.048370492687878e-06, "loss": 0.0865, "step": 4620 }, { "epoch": 2.11708209467185, "grad_norm": 1.1721983619011584, "learning_rate": 2.02888405285116e-06, "loss": 0.0888, "step": 4630 }, { "epoch": 2.121655613994969, "grad_norm": 1.2282433851448065, "learning_rate": 2.009467125079105e-06, "loss": 0.0845, "step": 4640 }, { "epoch": 2.1262291333180885, "grad_norm": 1.2226053300463011, "learning_rate": 1.9901201636490858e-06, "loss": 0.0885, "step": 4650 }, { "epoch": 2.1308026526412074, "grad_norm": 1.2621266589861262, "learning_rate": 1.970843621201541e-06, "loss": 0.0856, "step": 4660 }, { "epoch": 2.1353761719643267, "grad_norm": 1.182299682702729, "learning_rate": 1.9516379487293967e-06, "loss": 0.0868, "step": 4670 }, { "epoch": 2.1399496912874456, "grad_norm": 1.1289233180634102, "learning_rate": 1.9325035955675025e-06, "loss": 0.0874, "step": 4680 }, { "epoch": 2.144523210610565, "grad_norm": 1.2400158445190055, "learning_rate": 1.913441009382133e-06, "loss": 0.0874, "step": 4690 }, { "epoch": 2.149096729933684, "grad_norm": 1.0807483402857705, "learning_rate": 1.894450636160497e-06, "loss": 0.0823, "step": 4700 }, { "epoch": 2.153670249256803, "grad_norm": 1.20151443393666, "learning_rate": 1.8755329202003208e-06, "loss": 0.0847, "step": 4710 }, { "epoch": 2.158243768579922, "grad_norm": 1.249742742218928, "learning_rate": 1.8566883040994411e-06, "loss": 0.0865, "step": 4720 }, { "epoch": 2.1628172879030414, "grad_norm": 1.1698993423851831, "learning_rate": 1.8379172287454579e-06, "loss": 0.0827, "step": 4730 }, { "epoch": 2.1673908072261607, "grad_norm": 1.1963924999810194, "learning_rate": 1.8192201333054104e-06, "loss": 0.0853, "step": 4740 }, { "epoch": 2.1719643265492796, "grad_norm": 1.1476768209494772, "learning_rate": 1.8005974552155158e-06, "loss": 0.084, "step": 4750 }, { "epoch": 2.176537845872399, "grad_norm": 1.1760016868081118, "learning_rate": 1.7820496301709201e-06, "loss": 0.0839, "step": 4760 }, { "epoch": 2.181111365195518, "grad_norm": 1.230750914876748, "learning_rate": 1.7635770921155138e-06, "loss": 0.084, "step": 4770 }, { "epoch": 2.185684884518637, "grad_norm": 1.1447317209073706, "learning_rate": 1.7451802732317763e-06, "loss": 0.0851, "step": 4780 }, { "epoch": 2.190258403841756, "grad_norm": 1.2295730418302155, "learning_rate": 1.726859603930669e-06, "loss": 0.0841, "step": 4790 }, { "epoch": 2.1948319231648754, "grad_norm": 1.2200989835002083, "learning_rate": 1.708615512841557e-06, "loss": 0.0806, "step": 4800 }, { "epoch": 2.1994054424879943, "grad_norm": 1.2463059217931394, "learning_rate": 1.6904484268021915e-06, "loss": 0.0857, "step": 4810 }, { "epoch": 2.2039789618111136, "grad_norm": 1.3421421738801107, "learning_rate": 1.6723587708487099e-06, "loss": 0.0825, "step": 4820 }, { "epoch": 2.208552481134233, "grad_norm": 1.2261466658072597, "learning_rate": 1.6543469682057105e-06, "loss": 0.0845, "step": 4830 }, { "epoch": 2.213126000457352, "grad_norm": 1.1891825626221917, "learning_rate": 1.636413440276326e-06, "loss": 0.0797, "step": 4840 }, { "epoch": 2.217699519780471, "grad_norm": 1.2253821320075657, "learning_rate": 1.6185586066323917e-06, "loss": 0.0797, "step": 4850 }, { "epoch": 2.22227303910359, "grad_norm": 1.1083778704112748, "learning_rate": 1.6007828850046054e-06, "loss": 0.0818, "step": 4860 }, { "epoch": 2.2268465584267094, "grad_norm": 1.2901602921782305, "learning_rate": 1.5830866912727722e-06, "loss": 0.0813, "step": 4870 }, { "epoch": 2.2314200777498283, "grad_norm": 1.1449550587844235, "learning_rate": 1.565470439456061e-06, "loss": 0.0811, "step": 4880 }, { "epoch": 2.2359935970729476, "grad_norm": 1.161570984539211, "learning_rate": 1.5479345417033303e-06, "loss": 0.0809, "step": 4890 }, { "epoch": 2.2405671163960665, "grad_norm": 1.2549755914956502, "learning_rate": 1.5304794082834713e-06, "loss": 0.0832, "step": 4900 }, { "epoch": 2.245140635719186, "grad_norm": 1.2554510766891818, "learning_rate": 1.5131054475758277e-06, "loss": 0.0823, "step": 4910 }, { "epoch": 2.249714155042305, "grad_norm": 1.2344309261956705, "learning_rate": 1.495813066060618e-06, "loss": 0.0811, "step": 4920 }, { "epoch": 2.254287674365424, "grad_norm": 1.2150405513338038, "learning_rate": 1.47860266830945e-06, "loss": 0.0846, "step": 4930 }, { "epoch": 2.2588611936885434, "grad_norm": 1.205579274804413, "learning_rate": 1.461474656975837e-06, "loss": 0.078, "step": 4940 }, { "epoch": 2.2634347130116623, "grad_norm": 1.168546332886758, "learning_rate": 1.4444294327857894e-06, "loss": 0.0828, "step": 4950 }, { "epoch": 2.2680082323347817, "grad_norm": 1.1854383943209412, "learning_rate": 1.4274673945284278e-06, "loss": 0.0802, "step": 4960 }, { "epoch": 2.272581751657901, "grad_norm": 1.1827808686152284, "learning_rate": 1.410588939046666e-06, "loss": 0.0797, "step": 4970 }, { "epoch": 2.27715527098102, "grad_norm": 1.1721110884649917, "learning_rate": 1.3937944612279148e-06, "loss": 0.0791, "step": 4980 }, { "epoch": 2.281728790304139, "grad_norm": 1.164399016063137, "learning_rate": 1.3770843539948508e-06, "loss": 0.0828, "step": 4990 }, { "epoch": 2.286302309627258, "grad_norm": 1.2651868100759887, "learning_rate": 1.3604590082962177e-06, "loss": 0.0785, "step": 5000 }, { "epoch": 2.2908758289503774, "grad_norm": 1.252820037126073, "learning_rate": 1.34391881309769e-06, "loss": 0.0795, "step": 5010 }, { "epoch": 2.2954493482734963, "grad_norm": 1.1680970570445361, "learning_rate": 1.3274641553727568e-06, "loss": 0.0801, "step": 5020 }, { "epoch": 2.3000228675966157, "grad_norm": 1.260058172862296, "learning_rate": 1.311095420093686e-06, "loss": 0.0835, "step": 5030 }, { "epoch": 2.3045963869197346, "grad_norm": 1.2298581356534384, "learning_rate": 1.2948129902225005e-06, "loss": 0.0811, "step": 5040 }, { "epoch": 2.309169906242854, "grad_norm": 1.3236937161045836, "learning_rate": 1.2786172467020357e-06, "loss": 0.0795, "step": 5050 }, { "epoch": 2.3137434255659732, "grad_norm": 1.1492049996206348, "learning_rate": 1.2625085684470074e-06, "loss": 0.0776, "step": 5060 }, { "epoch": 2.318316944889092, "grad_norm": 1.1576310528043592, "learning_rate": 1.2464873323351657e-06, "loss": 0.0772, "step": 5070 }, { "epoch": 2.3228904642122115, "grad_norm": 1.2413337800281323, "learning_rate": 1.2305539131984646e-06, "loss": 0.0781, "step": 5080 }, { "epoch": 2.3274639835353303, "grad_norm": 1.2093092757178983, "learning_rate": 1.2147086838143002e-06, "loss": 0.0769, "step": 5090 }, { "epoch": 2.3320375028584497, "grad_norm": 1.3487866110656108, "learning_rate": 1.1989520148967827e-06, "loss": 0.0754, "step": 5100 }, { "epoch": 2.3366110221815686, "grad_norm": 1.3073322844593334, "learning_rate": 1.1832842750880702e-06, "loss": 0.0756, "step": 5110 }, { "epoch": 2.341184541504688, "grad_norm": 1.2318698640979664, "learning_rate": 1.1677058309497374e-06, "loss": 0.0766, "step": 5120 }, { "epoch": 2.345758060827807, "grad_norm": 1.2015349258625985, "learning_rate": 1.1522170469542027e-06, "loss": 0.0757, "step": 5130 }, { "epoch": 2.350331580150926, "grad_norm": 1.1855975489354469, "learning_rate": 1.1368182854762005e-06, "loss": 0.0757, "step": 5140 }, { "epoch": 2.3549050994740455, "grad_norm": 1.1300232279444944, "learning_rate": 1.1215099067843067e-06, "loss": 0.0754, "step": 5150 }, { "epoch": 2.3594786187971644, "grad_norm": 1.2496116002791076, "learning_rate": 1.1062922690325006e-06, "loss": 0.0776, "step": 5160 }, { "epoch": 2.3640521381202837, "grad_norm": 1.1185177869926426, "learning_rate": 1.091165728251799e-06, "loss": 0.0738, "step": 5170 }, { "epoch": 2.3686256574434026, "grad_norm": 1.1557068105495683, "learning_rate": 1.0761306383419107e-06, "loss": 0.0787, "step": 5180 }, { "epoch": 2.373199176766522, "grad_norm": 1.0871650657798408, "learning_rate": 1.0611873510629727e-06, "loss": 0.0765, "step": 5190 }, { "epoch": 2.377772696089641, "grad_norm": 1.1603851214780856, "learning_rate": 1.0463362160273076e-06, "loss": 0.0769, "step": 5200 }, { "epoch": 2.38234621541276, "grad_norm": 1.1187255971218357, "learning_rate": 1.0315775806912515e-06, "loss": 0.0758, "step": 5210 }, { "epoch": 2.386919734735879, "grad_norm": 1.1898271554795894, "learning_rate": 1.0169117903470188e-06, "loss": 0.0738, "step": 5220 }, { "epoch": 2.3914932540589984, "grad_norm": 1.3992377782070355, "learning_rate": 1.0023391881146349e-06, "loss": 0.0737, "step": 5230 }, { "epoch": 2.3960667733821177, "grad_norm": 1.2021049780855009, "learning_rate": 9.87860114933894e-07, "loss": 0.074, "step": 5240 }, { "epoch": 2.4006402927052366, "grad_norm": 1.1884377855025092, "learning_rate": 9.734749095563956e-07, "loss": 0.0758, "step": 5250 }, { "epoch": 2.405213812028356, "grad_norm": 1.3176770044095036, "learning_rate": 9.59183908537607e-07, "loss": 0.0735, "step": 5260 }, { "epoch": 2.409787331351475, "grad_norm": 1.1327872849085017, "learning_rate": 9.449874462290043e-07, "loss": 0.0739, "step": 5270 }, { "epoch": 2.414360850674594, "grad_norm": 1.2623454047105018, "learning_rate": 9.308858547702298e-07, "loss": 0.0756, "step": 5280 }, { "epoch": 2.418934369997713, "grad_norm": 1.2157161510793102, "learning_rate": 9.168794640813428e-07, "loss": 0.0747, "step": 5290 }, { "epoch": 2.4235078893208324, "grad_norm": 1.2026772420377383, "learning_rate": 9.029686018550843e-07, "loss": 0.0741, "step": 5300 }, { "epoch": 2.4280814086439513, "grad_norm": 1.2935714413046169, "learning_rate": 8.891535935492207e-07, "loss": 0.0737, "step": 5310 }, { "epoch": 2.4326549279670706, "grad_norm": 1.2170663592360373, "learning_rate": 8.754347623789222e-07, "loss": 0.077, "step": 5320 }, { "epoch": 2.43722844729019, "grad_norm": 1.2153798387901247, "learning_rate": 8.618124293092073e-07, "loss": 0.0747, "step": 5330 }, { "epoch": 2.441801966613309, "grad_norm": 1.2479747761427689, "learning_rate": 8.482869130474269e-07, "loss": 0.071, "step": 5340 }, { "epoch": 2.446375485936428, "grad_norm": 1.2010560524270188, "learning_rate": 8.34858530035813e-07, "loss": 0.0706, "step": 5350 }, { "epoch": 2.450949005259547, "grad_norm": 1.1785691349170302, "learning_rate": 8.215275944440721e-07, "loss": 0.075, "step": 5360 }, { "epoch": 2.4555225245826664, "grad_norm": 1.2240982744001758, "learning_rate": 8.082944181620394e-07, "loss": 0.073, "step": 5370 }, { "epoch": 2.4600960439057857, "grad_norm": 1.0892646704287998, "learning_rate": 7.951593107923744e-07, "loss": 0.0683, "step": 5380 }, { "epoch": 2.4646695632289046, "grad_norm": 1.0550582843246485, "learning_rate": 7.821225796433262e-07, "loss": 0.0695, "step": 5390 }, { "epoch": 2.469243082552024, "grad_norm": 1.257436440542298, "learning_rate": 7.69184529721535e-07, "loss": 0.0702, "step": 5400 }, { "epoch": 2.473816601875143, "grad_norm": 1.3131986333254564, "learning_rate": 7.563454637249056e-07, "loss": 0.0705, "step": 5410 }, { "epoch": 2.478390121198262, "grad_norm": 1.2233304106332723, "learning_rate": 7.436056820355125e-07, "loss": 0.0723, "step": 5420 }, { "epoch": 2.482963640521381, "grad_norm": 1.1706796221669824, "learning_rate": 7.309654827125889e-07, "loss": 0.0697, "step": 5430 }, { "epoch": 2.4875371598445004, "grad_norm": 1.218992944017262, "learning_rate": 7.184251614855369e-07, "loss": 0.0707, "step": 5440 }, { "epoch": 2.4921106791676193, "grad_norm": 1.1713205258775266, "learning_rate": 7.05985011747023e-07, "loss": 0.0702, "step": 5450 }, { "epoch": 2.4966841984907386, "grad_norm": 1.361204798714588, "learning_rate": 6.936453245461006e-07, "loss": 0.07, "step": 5460 }, { "epoch": 2.501257717813858, "grad_norm": 1.1874028612748257, "learning_rate": 6.814063885814127e-07, "loss": 0.0653, "step": 5470 }, { "epoch": 2.501257717813858, "eval_ood_font_family_decon_loss": 0.5047458410263062, "eval_ood_font_family_decon_runtime": 750.7381, "eval_ood_font_family_decon_samples_per_second": 41.333, "eval_ood_font_family_decon_steps_per_second": 1.292, "step": 5470 }, { "epoch": 2.505831237136977, "grad_norm": 1.2444614315528253, "learning_rate": 6.692684901944268e-07, "loss": 0.0714, "step": 5480 }, { "epoch": 2.510404756460096, "grad_norm": 1.2930112939593634, "learning_rate": 6.572319133627469e-07, "loss": 0.0716, "step": 5490 }, { "epoch": 2.514978275783215, "grad_norm": 1.2174396535574574, "learning_rate": 6.452969396934567e-07, "loss": 0.0719, "step": 5500 }, { "epoch": 2.5195517951063344, "grad_norm": 1.211234582740258, "learning_rate": 6.334638484165428e-07, "loss": 0.0711, "step": 5510 }, { "epoch": 2.5241253144294533, "grad_norm": 1.2638636584777687, "learning_rate": 6.217329163783548e-07, "loss": 0.0677, "step": 5520 }, { "epoch": 2.5286988337525727, "grad_norm": 1.0890916357754288, "learning_rate": 6.101044180351318e-07, "loss": 0.0712, "step": 5530 }, { "epoch": 2.5332723530756915, "grad_norm": 1.2691847187450922, "learning_rate": 5.98578625446578e-07, "loss": 0.0692, "step": 5540 }, { "epoch": 2.537845872398811, "grad_norm": 1.199689543916915, "learning_rate": 5.871558082695017e-07, "loss": 0.0701, "step": 5550 }, { "epoch": 2.54241939172193, "grad_norm": 1.1783662054710138, "learning_rate": 5.758362337515028e-07, "loss": 0.0688, "step": 5560 }, { "epoch": 2.546992911045049, "grad_norm": 1.2270374581758479, "learning_rate": 5.646201667247209e-07, "loss": 0.0686, "step": 5570 }, { "epoch": 2.5515664303681684, "grad_norm": 1.4557355828149834, "learning_rate": 5.535078695996399e-07, "loss": 0.0705, "step": 5580 }, { "epoch": 2.5561399496912873, "grad_norm": 1.229503016327861, "learning_rate": 5.424996023589524e-07, "loss": 0.0686, "step": 5590 }, { "epoch": 2.5607134690144067, "grad_norm": 1.1650658463461216, "learning_rate": 5.31595622551469e-07, "loss": 0.0688, "step": 5600 }, { "epoch": 2.565286988337526, "grad_norm": 1.2252797738295218, "learning_rate": 5.20796185286101e-07, "loss": 0.0709, "step": 5610 }, { "epoch": 2.569860507660645, "grad_norm": 1.1699907296153322, "learning_rate": 5.101015432258843e-07, "loss": 0.0691, "step": 5620 }, { "epoch": 2.574434026983764, "grad_norm": 1.1519768176577354, "learning_rate": 4.995119465820775e-07, "loss": 0.0687, "step": 5630 }, { "epoch": 2.579007546306883, "grad_norm": 1.2964700842760177, "learning_rate": 4.890276431082946e-07, "loss": 0.0687, "step": 5640 }, { "epoch": 2.5835810656300024, "grad_norm": 1.2835300839036692, "learning_rate": 4.786488780947246e-07, "loss": 0.065, "step": 5650 }, { "epoch": 2.5881545849531213, "grad_norm": 1.1754554238435015, "learning_rate": 4.6837589436237784e-07, "loss": 0.0726, "step": 5660 }, { "epoch": 2.5927281042762407, "grad_norm": 1.267541442731523, "learning_rate": 4.5820893225741624e-07, "loss": 0.0695, "step": 5670 }, { "epoch": 2.5973016235993596, "grad_norm": 1.2768658967632311, "learning_rate": 4.4814822964552363e-07, "loss": 0.0709, "step": 5680 }, { "epoch": 2.601875142922479, "grad_norm": 1.2365173256249191, "learning_rate": 4.3819402190634294e-07, "loss": 0.0711, "step": 5690 }, { "epoch": 2.6064486622455982, "grad_norm": 1.2468806490273114, "learning_rate": 4.283465419279692e-07, "loss": 0.0687, "step": 5700 }, { "epoch": 2.611022181568717, "grad_norm": 1.2751097650873808, "learning_rate": 4.186060201014991e-07, "loss": 0.0685, "step": 5710 }, { "epoch": 2.615595700891836, "grad_norm": 1.2271132958558133, "learning_rate": 4.0897268431564385e-07, "loss": 0.068, "step": 5720 }, { "epoch": 2.6201692202149554, "grad_norm": 1.1553637144793711, "learning_rate": 3.994467599513968e-07, "loss": 0.0676, "step": 5730 }, { "epoch": 2.6247427395380747, "grad_norm": 1.090394685497435, "learning_rate": 3.9002846987675704e-07, "loss": 0.068, "step": 5740 }, { "epoch": 2.6293162588611936, "grad_norm": 1.1774794035588032, "learning_rate": 3.807180344415201e-07, "loss": 0.0686, "step": 5750 }, { "epoch": 2.633889778184313, "grad_norm": 1.2168238277226393, "learning_rate": 3.7151567147211907e-07, "loss": 0.0659, "step": 5760 }, { "epoch": 2.638463297507432, "grad_norm": 1.1716148636013919, "learning_rate": 3.6242159626653004e-07, "loss": 0.0666, "step": 5770 }, { "epoch": 2.643036816830551, "grad_norm": 1.2138094797188868, "learning_rate": 3.5343602158923464e-07, "loss": 0.0676, "step": 5780 }, { "epoch": 2.6476103361536705, "grad_norm": 1.204548223696344, "learning_rate": 3.4455915766624157e-07, "loss": 0.0677, "step": 5790 }, { "epoch": 2.6521838554767894, "grad_norm": 1.2977699180958662, "learning_rate": 3.357912121801682e-07, "loss": 0.0637, "step": 5800 }, { "epoch": 2.6567573747999083, "grad_norm": 1.2063249090641337, "learning_rate": 3.2713239026538504e-07, "loss": 0.0651, "step": 5810 }, { "epoch": 2.6613308941230276, "grad_norm": 1.255493154511345, "learning_rate": 3.1858289450320936e-07, "loss": 0.0654, "step": 5820 }, { "epoch": 2.665904413446147, "grad_norm": 1.1764266643207675, "learning_rate": 3.1014292491717444e-07, "loss": 0.0673, "step": 5830 }, { "epoch": 2.670477932769266, "grad_norm": 1.2720827559481798, "learning_rate": 3.018126789683412e-07, "loss": 0.0642, "step": 5840 }, { "epoch": 2.675051452092385, "grad_norm": 1.1540513971797848, "learning_rate": 2.935923515506839e-07, "loss": 0.0656, "step": 5850 }, { "epoch": 2.679624971415504, "grad_norm": 1.3786242538002051, "learning_rate": 2.854821349865289e-07, "loss": 0.0685, "step": 5860 }, { "epoch": 2.6841984907386234, "grad_norm": 1.3523288016403008, "learning_rate": 2.7748221902205607e-07, "loss": 0.0663, "step": 5870 }, { "epoch": 2.6887720100617427, "grad_norm": 1.2534561260833565, "learning_rate": 2.695927908228563e-07, "loss": 0.0677, "step": 5880 }, { "epoch": 2.6933455293848616, "grad_norm": 1.3694930580842588, "learning_rate": 2.618140349695575e-07, "loss": 0.0673, "step": 5890 }, { "epoch": 2.697919048707981, "grad_norm": 1.2332828750907012, "learning_rate": 2.5414613345350116e-07, "loss": 0.0662, "step": 5900 }, { "epoch": 2.7024925680311, "grad_norm": 1.2217935230493537, "learning_rate": 2.4658926567248886e-07, "loss": 0.0624, "step": 5910 }, { "epoch": 2.707066087354219, "grad_norm": 1.2044558275872, "learning_rate": 2.391436084265814e-07, "loss": 0.0654, "step": 5920 }, { "epoch": 2.711639606677338, "grad_norm": 1.302279110221505, "learning_rate": 2.318093359139645e-07, "loss": 0.0669, "step": 5930 }, { "epoch": 2.7162131260004574, "grad_norm": 1.2630194748124701, "learning_rate": 2.2458661972687258e-07, "loss": 0.0662, "step": 5940 }, { "epoch": 2.7207866453235763, "grad_norm": 1.157025703309838, "learning_rate": 2.17475628847576e-07, "loss": 0.0651, "step": 5950 }, { "epoch": 2.7253601646466956, "grad_norm": 1.195864213542712, "learning_rate": 2.1047652964442277e-07, "loss": 0.0658, "step": 5960 }, { "epoch": 2.729933683969815, "grad_norm": 1.3231565748566925, "learning_rate": 2.0358948586795378e-07, "loss": 0.067, "step": 5970 }, { "epoch": 2.734507203292934, "grad_norm": 1.1639966157706896, "learning_rate": 1.9681465864706372e-07, "loss": 0.0633, "step": 5980 }, { "epoch": 2.739080722616053, "grad_norm": 1.3385422642171627, "learning_rate": 1.901522064852368e-07, "loss": 0.0638, "step": 5990 }, { "epoch": 2.743654241939172, "grad_norm": 1.2156766745023297, "learning_rate": 1.8360228525683477e-07, "loss": 0.0619, "step": 6000 }, { "epoch": 2.7482277612622914, "grad_norm": 1.1995199604965106, "learning_rate": 1.7716504820345427e-07, "loss": 0.0658, "step": 6010 }, { "epoch": 2.7528012805854107, "grad_norm": 1.1625948507142823, "learning_rate": 1.708406459303369e-07, "loss": 0.0625, "step": 6020 }, { "epoch": 2.7573747999085296, "grad_norm": 1.1059621540906581, "learning_rate": 1.646292264028504e-07, "loss": 0.0634, "step": 6030 }, { "epoch": 2.7619483192316485, "grad_norm": 1.2690798798687575, "learning_rate": 1.5853093494302195e-07, "loss": 0.0641, "step": 6040 }, { "epoch": 2.766521838554768, "grad_norm": 1.1759672927812295, "learning_rate": 1.5254591422614319e-07, "loss": 0.0636, "step": 6050 }, { "epoch": 2.771095357877887, "grad_norm": 1.1738431525914035, "learning_rate": 1.4667430427742714e-07, "loss": 0.0655, "step": 6060 }, { "epoch": 2.775668877201006, "grad_norm": 1.263500982599949, "learning_rate": 1.409162424687366e-07, "loss": 0.063, "step": 6070 }, { "epoch": 2.7802423965241254, "grad_norm": 1.256262397514191, "learning_rate": 1.3527186351536658e-07, "loss": 0.0634, "step": 6080 }, { "epoch": 2.7848159158472443, "grad_norm": 1.3065509184073028, "learning_rate": 1.2974129947289748e-07, "loss": 0.0664, "step": 6090 }, { "epoch": 2.7893894351703636, "grad_norm": 1.212213159092694, "learning_rate": 1.2432467973409857e-07, "loss": 0.0649, "step": 6100 }, { "epoch": 2.793962954493483, "grad_norm": 1.2525434057981755, "learning_rate": 1.1902213102590709e-07, "loss": 0.0643, "step": 6110 }, { "epoch": 2.798536473816602, "grad_norm": 1.1654106104389876, "learning_rate": 1.1383377740646018e-07, "loss": 0.063, "step": 6120 }, { "epoch": 2.8031099931397208, "grad_norm": 1.2556909714998763, "learning_rate": 1.0875974026219149e-07, "loss": 0.0646, "step": 6130 }, { "epoch": 2.80768351246284, "grad_norm": 1.294053806396921, "learning_rate": 1.0380013830499469e-07, "loss": 0.0644, "step": 6140 }, { "epoch": 2.8122570317859594, "grad_norm": 1.2262670140274397, "learning_rate": 9.895508756944339e-08, "loss": 0.0649, "step": 6150 }, { "epoch": 2.8168305511090783, "grad_norm": 1.3248987737608768, "learning_rate": 9.422470141007667e-08, "loss": 0.0636, "step": 6160 }, { "epoch": 2.8214040704321977, "grad_norm": 1.2320588187162067, "learning_rate": 8.960909049874845e-08, "loss": 0.0613, "step": 6170 }, { "epoch": 2.8259775897553165, "grad_norm": 1.240393510360426, "learning_rate": 8.510836282203616e-08, "loss": 0.0624, "step": 6180 }, { "epoch": 2.830551109078436, "grad_norm": 1.2059226513528654, "learning_rate": 8.072262367871675e-08, "loss": 0.0638, "step": 6190 }, { "epoch": 2.835124628401555, "grad_norm": 1.2561973558543666, "learning_rate": 7.645197567730078e-08, "loss": 0.0611, "step": 6200 }, { "epoch": 2.839698147724674, "grad_norm": 1.3347348874923324, "learning_rate": 7.22965187336333e-08, "loss": 0.0635, "step": 6210 }, { "epoch": 2.8442716670477934, "grad_norm": 1.2397033297336804, "learning_rate": 6.825635006855458e-08, "loss": 0.0626, "step": 6220 }, { "epoch": 2.8488451863709123, "grad_norm": 1.250766076599714, "learning_rate": 6.433156420562914e-08, "loss": 0.0623, "step": 6230 }, { "epoch": 2.8534187056940317, "grad_norm": 1.2420172257095343, "learning_rate": 6.052225296892866e-08, "loss": 0.0634, "step": 6240 }, { "epoch": 2.8579922250171506, "grad_norm": 1.2575995597478875, "learning_rate": 5.682850548089036e-08, "loss": 0.0649, "step": 6250 }, { "epoch": 2.86256574434027, "grad_norm": 1.2129956492898422, "learning_rate": 5.325040816022531e-08, "loss": 0.0655, "step": 6260 }, { "epoch": 2.867139263663389, "grad_norm": 1.373678315507613, "learning_rate": 4.978804471990117e-08, "loss": 0.0663, "step": 6270 }, { "epoch": 2.871712782986508, "grad_norm": 1.377918142457103, "learning_rate": 4.644149616518212e-08, "loss": 0.0646, "step": 6280 }, { "epoch": 2.8762863023096275, "grad_norm": 1.2511597168900848, "learning_rate": 4.32108407917331e-08, "loss": 0.0623, "step": 6290 }, { "epoch": 2.8808598216327463, "grad_norm": 1.2146912003470003, "learning_rate": 4.009615418378854e-08, "loss": 0.0614, "step": 6300 }, { "epoch": 2.8854333409558657, "grad_norm": 1.4155952713620488, "learning_rate": 3.709750921238486e-08, "loss": 0.0651, "step": 6310 }, { "epoch": 2.8900068602789846, "grad_norm": 1.2196196689451506, "learning_rate": 3.421497603365354e-08, "loss": 0.0625, "step": 6320 }, { "epoch": 2.894580379602104, "grad_norm": 1.2350648894955072, "learning_rate": 3.1448622087183465e-08, "loss": 0.0654, "step": 6330 }, { "epoch": 2.899153898925223, "grad_norm": 1.1140967950730598, "learning_rate": 2.8798512094436738e-08, "loss": 0.062, "step": 6340 }, { "epoch": 2.903727418248342, "grad_norm": 1.1496479230728414, "learning_rate": 2.626470805724146e-08, "loss": 0.064, "step": 6350 }, { "epoch": 2.908300937571461, "grad_norm": 1.25893477590324, "learning_rate": 2.3847269256336848e-08, "loss": 0.0646, "step": 6360 }, { "epoch": 2.9128744568945804, "grad_norm": 1.2339732408820683, "learning_rate": 2.1546252249988186e-08, "loss": 0.0631, "step": 6370 }, { "epoch": 2.9174479762176997, "grad_norm": 1.211280537309632, "learning_rate": 1.9361710872661254e-08, "loss": 0.0629, "step": 6380 }, { "epoch": 2.9220214955408186, "grad_norm": 1.1918209881447972, "learning_rate": 1.729369623376609e-08, "loss": 0.0625, "step": 6390 }, { "epoch": 2.926595014863938, "grad_norm": 1.2248314679504857, "learning_rate": 1.5342256716459058e-08, "loss": 0.0623, "step": 6400 }, { "epoch": 2.931168534187057, "grad_norm": 1.105245617566905, "learning_rate": 1.3507437976510996e-08, "loss": 0.0638, "step": 6410 }, { "epoch": 2.935742053510176, "grad_norm": 1.2867136648836608, "learning_rate": 1.178928294123971e-08, "loss": 0.0647, "step": 6420 }, { "epoch": 2.9403155728332955, "grad_norm": 1.2742658419239792, "learning_rate": 1.018783180850691e-08, "loss": 0.0613, "step": 6430 }, { "epoch": 2.9448890921564144, "grad_norm": 1.2804173454573406, "learning_rate": 8.703122045772838e-09, "loss": 0.0638, "step": 6440 }, { "epoch": 2.9494626114795333, "grad_norm": 1.2776703711119592, "learning_rate": 7.335188389226977e-09, "loss": 0.0612, "step": 6450 }, { "epoch": 2.9540361308026526, "grad_norm": 1.3273671539717513, "learning_rate": 6.084062842968696e-09, "loss": 0.0635, "step": 6460 }, { "epoch": 2.958609650125772, "grad_norm": 1.2233821775713383, "learning_rate": 4.949774678262298e-09, "loss": 0.0626, "step": 6470 }, { "epoch": 2.963183169448891, "grad_norm": 1.2623910317397984, "learning_rate": 3.932350432850896e-09, "loss": 0.0637, "step": 6480 }, { "epoch": 2.96775668877201, "grad_norm": 1.1940768995412179, "learning_rate": 3.0318139103363564e-09, "loss": 0.0623, "step": 6490 }, { "epoch": 2.972330208095129, "grad_norm": 1.3133288588159049, "learning_rate": 2.2481861796214145e-09, "loss": 0.0633, "step": 6500 }, { "epoch": 2.9769037274182484, "grad_norm": 1.1759574740768084, "learning_rate": 1.58148557441673e-09, "loss": 0.0612, "step": 6510 }, { "epoch": 2.9814772467413677, "grad_norm": 1.219126694517436, "learning_rate": 1.0317276928134557e-09, "loss": 0.0608, "step": 6520 }, { "epoch": 2.9860507660644866, "grad_norm": 1.3473048904965068, "learning_rate": 5.989253969163056e-10, "loss": 0.0599, "step": 6530 }, { "epoch": 2.9906242853876055, "grad_norm": 1.2526336921803998, "learning_rate": 2.830888125437969e-10, "loss": 0.0623, "step": 6540 }, { "epoch": 2.995197804710725, "grad_norm": 1.160512322538154, "learning_rate": 8.422532899121649e-11, "loss": 0.0627, "step": 6550 }, { "epoch": 2.999771324033844, "grad_norm": 1.2574664439964043, "learning_rate": 2.3395988579810735e-12, "loss": 0.0616, "step": 6560 }, { "epoch": 3.0, "step": 6561, "total_flos": 1.4987532993298432e+16, "train_loss": 0.1726040010526606, "train_runtime": 67874.2861, "train_samples_per_second": 12.369, "train_steps_per_second": 0.097 } ], "logging_steps": 10, "max_steps": 6561, "num_input_tokens_seen": 0, "num_train_epochs": 3, "save_steps": 1094, "stateful_callbacks": { "TrainerControl": { "args": { "should_epoch_stop": false, "should_evaluate": false, "should_log": false, "should_save": true, "should_training_stop": true }, "attributes": {} } }, "total_flos": 1.4987532993298432e+16, "train_batch_size": 2, "trial_name": null, "trial_params": null }