{ "model.layers.0.mlp.down_proj": { "Wscale": 0.015026, "bpw": 3.0, "proxy_loss": 1e-05, "shape": [ 2048, 11008 ], "sqnr": 15.742874 }, "model.layers.0.mlp.gate_proj": { "Wscale": 0.015899, "bpw": 3.0, "proxy_loss": 0.000243, "shape": [ 11008, 2048 ], "sqnr": 16.083892 }, "model.layers.0.mlp.up_proj": { "Wscale": 0.015257, "bpw": 3.0, "proxy_loss": 0.000223, "shape": [ 11008, 2048 ], "sqnr": 16.088734 }, "model.layers.0.self_attn.k_proj": { "Wscale": 0.041056, "bpw": 3.0, "proxy_loss": 0.000401, "shape": [ 512, 2048 ], "sqnr": 15.944422 }, "model.layers.0.self_attn.o_proj": { "Wscale": 0.011211, "bpw": 3.0, "proxy_loss": 1e-06, "shape": [ 2048, 2048 ], "sqnr": 15.519045 }, "model.layers.0.self_attn.q_proj": { "Wscale": 0.029598, "bpw": 3.0, "proxy_loss": 0.000245, "shape": [ 2048, 2048 ], "sqnr": 15.544462 }, "model.layers.0.self_attn.v_proj": { "Wscale": 0.011852, "bpw": 3.0, "proxy_loss": 3.2e-05, "shape": [ 512, 2048 ], "sqnr": 16.116248 }, "model.layers.1.mlp.down_proj": { "Wscale": 0.014973, "bpw": 3.0, "proxy_loss": 1.4e-05, "shape": [ 2048, 11008 ], "sqnr": 16.926612 }, "model.layers.1.mlp.gate_proj": { "Wscale": 0.015828, "bpw": 3.0, "proxy_loss": 0.000317, "shape": [ 11008, 2048 ], "sqnr": 16.156193 }, "model.layers.1.mlp.up_proj": { "Wscale": 0.015044, "bpw": 3.0, "proxy_loss": 0.000286, "shape": [ 11008, 2048 ], "sqnr": 16.159583 }, "model.layers.1.self_attn.k_proj": { "Wscale": 0.034614, "bpw": 3.0, "proxy_loss": 0.000469, "shape": [ 512, 2048 ], "sqnr": 15.936944 }, "model.layers.1.self_attn.o_proj": { "Wscale": 0.012089, "bpw": 3.0, "proxy_loss": 3e-06, "shape": [ 2048, 2048 ], "sqnr": 14.959725 }, "model.layers.1.self_attn.q_proj": { "Wscale": 0.025205, "bpw": 3.0, "proxy_loss": 0.000251, "shape": [ 2048, 2048 ], "sqnr": 15.957773 }, "model.layers.1.self_attn.v_proj": { "Wscale": 0.012444, "bpw": 3.0, "proxy_loss": 6e-05, "shape": [ 512, 2048 ], "sqnr": 15.977445 }, "model.layers.10.mlp.down_proj": { "Wscale": 0.014414, "bpw": 3.0, "proxy_loss": 3.3e-05, "shape": [ 2048, 11008 ], "sqnr": 16.327248 }, "model.layers.10.mlp.gate_proj": { "Wscale": 0.017192, "bpw": 3.0, "proxy_loss": 0.000628, "shape": [ 11008, 2048 ], "sqnr": 16.471498 }, "model.layers.10.mlp.up_proj": { "Wscale": 0.014329, "bpw": 3.0, "proxy_loss": 0.000436, "shape": [ 11008, 2048 ], "sqnr": 16.476899 }, "model.layers.10.self_attn.k_proj": { "Wscale": 0.035175, "bpw": 3.0, "proxy_loss": 0.002989, "shape": [ 512, 2048 ], "sqnr": 16.282854 }, "model.layers.10.self_attn.o_proj": { "Wscale": 0.013609, "bpw": 3.0, "proxy_loss": 1.4e-05, "shape": [ 2048, 2048 ], "sqnr": 15.832544 }, "model.layers.10.self_attn.q_proj": { "Wscale": 0.02423, "bpw": 3.0, "proxy_loss": 0.001415, "shape": [ 2048, 2048 ], "sqnr": 16.287183 }, "model.layers.10.self_attn.v_proj": { "Wscale": 0.013487, "bpw": 3.0, "proxy_loss": 0.000435, "shape": [ 512, 2048 ], "sqnr": 16.306866 }, "model.layers.11.mlp.down_proj": { "Wscale": 0.014639, "bpw": 3.0, "proxy_loss": 3.3e-05, "shape": [ 2048, 11008 ], "sqnr": 16.448595 }, "model.layers.11.mlp.gate_proj": { "Wscale": 0.016712, "bpw": 3.0, "proxy_loss": 0.000603, "shape": [ 11008, 2048 ], "sqnr": 16.4504 }, "model.layers.11.mlp.up_proj": { "Wscale": 0.014575, "bpw": 3.0, "proxy_loss": 0.000458, "shape": [ 11008, 2048 ], "sqnr": 16.454905 }, "model.layers.11.self_attn.k_proj": { "Wscale": 0.035617, "bpw": 3.0, "proxy_loss": 0.007305, "shape": [ 512, 2048 ], "sqnr": 16.17017 }, "model.layers.11.self_attn.o_proj": { "Wscale": 0.012348, "bpw": 3.0, "proxy_loss": 1e-05, "shape": [ 2048, 2048 ], "sqnr": 15.444257 }, "model.layers.11.self_attn.q_proj": { "Wscale": 0.023042, "bpw": 3.0, "proxy_loss": 0.003047, "shape": [ 2048, 2048 ], "sqnr": 16.170305 }, "model.layers.11.self_attn.v_proj": { "Wscale": 0.009609, "bpw": 3.0, "proxy_loss": 0.000529, "shape": [ 512, 2048 ], "sqnr": 16.167441 }, "model.layers.12.mlp.down_proj": { "Wscale": 0.014597, "bpw": 3.0, "proxy_loss": 3.2e-05, "shape": [ 2048, 11008 ], "sqnr": 16.454224 }, "model.layers.12.mlp.gate_proj": { "Wscale": 0.016719, "bpw": 3.0, "proxy_loss": 0.000589, "shape": [ 11008, 2048 ], "sqnr": 16.445474 }, "model.layers.12.mlp.up_proj": { "Wscale": 0.014522, "bpw": 3.0, "proxy_loss": 0.000444, "shape": [ 11008, 2048 ], "sqnr": 16.44998 }, "model.layers.12.self_attn.k_proj": { "Wscale": 0.03448, "bpw": 3.0, "proxy_loss": 0.003744, "shape": [ 512, 2048 ], "sqnr": 16.090119 }, "model.layers.12.self_attn.o_proj": { "Wscale": 0.012221, "bpw": 3.0, "proxy_loss": 1.3e-05, "shape": [ 2048, 2048 ], "sqnr": 15.751918 }, "model.layers.12.self_attn.q_proj": { "Wscale": 0.023259, "bpw": 3.0, "proxy_loss": 0.001691, "shape": [ 2048, 2048 ], "sqnr": 16.107216 }, "model.layers.12.self_attn.v_proj": { "Wscale": 0.010754, "bpw": 3.0, "proxy_loss": 0.000359, "shape": [ 512, 2048 ], "sqnr": 16.130954 }, "model.layers.13.mlp.down_proj": { "Wscale": 0.014518, "bpw": 3.0, "proxy_loss": 3.3e-05, "shape": [ 2048, 11008 ], "sqnr": 16.426712 }, "model.layers.13.mlp.gate_proj": { "Wscale": 0.016755, "bpw": 3.0, "proxy_loss": 0.00059, "shape": [ 11008, 2048 ], "sqnr": 16.440542 }, "model.layers.13.mlp.up_proj": { "Wscale": 0.014406, "bpw": 3.0, "proxy_loss": 0.000436, "shape": [ 11008, 2048 ], "sqnr": 16.444985 }, "model.layers.13.self_attn.k_proj": { "Wscale": 0.033971, "bpw": 3.0, "proxy_loss": 0.004435, "shape": [ 512, 2048 ], "sqnr": 16.26646 }, "model.layers.13.self_attn.o_proj": { "Wscale": 0.013106, "bpw": 3.0, "proxy_loss": 1.8e-05, "shape": [ 2048, 2048 ], "sqnr": 15.615196 }, "model.layers.13.self_attn.q_proj": { "Wscale": 0.022706, "bpw": 3.0, "proxy_loss": 0.001974, "shape": [ 2048, 2048 ], "sqnr": 16.26767 }, "model.layers.13.self_attn.v_proj": { "Wscale": 0.012257, "bpw": 3.0, "proxy_loss": 0.000574, "shape": [ 512, 2048 ], "sqnr": 16.273629 }, "model.layers.14.mlp.down_proj": { "Wscale": 0.014465, "bpw": 3.0, "proxy_loss": 3e-05, "shape": [ 2048, 11008 ], "sqnr": 16.407018 }, "model.layers.14.mlp.gate_proj": { "Wscale": 0.016837, "bpw": 3.0, "proxy_loss": 0.000561, "shape": [ 11008, 2048 ], "sqnr": 16.408449 }, "model.layers.14.mlp.up_proj": { "Wscale": 0.014412, "bpw": 3.0, "proxy_loss": 0.000411, "shape": [ 11008, 2048 ], "sqnr": 16.414821 }, "model.layers.14.self_attn.k_proj": { "Wscale": 0.033685, "bpw": 3.0, "proxy_loss": 0.003209, "shape": [ 512, 2048 ], "sqnr": 16.116614 }, "model.layers.14.self_attn.o_proj": { "Wscale": 0.013699, "bpw": 3.0, "proxy_loss": 1.9e-05, "shape": [ 2048, 2048 ], "sqnr": 15.740012 }, "model.layers.14.self_attn.q_proj": { "Wscale": 0.02278, "bpw": 3.0, "proxy_loss": 0.001465, "shape": [ 2048, 2048 ], "sqnr": 16.123218 }, "model.layers.14.self_attn.v_proj": { "Wscale": 0.014, "bpw": 3.0, "proxy_loss": 0.000551, "shape": [ 512, 2048 ], "sqnr": 16.132034 }, "model.layers.15.mlp.down_proj": { "Wscale": 0.014423, "bpw": 3.0, "proxy_loss": 3.2e-05, "shape": [ 2048, 11008 ], "sqnr": 16.350258 }, "model.layers.15.mlp.gate_proj": { "Wscale": 0.016917, "bpw": 3.0, "proxy_loss": 0.000588, "shape": [ 11008, 2048 ], "sqnr": 16.421558 }, "model.layers.15.mlp.up_proj": { "Wscale": 0.014372, "bpw": 3.0, "proxy_loss": 0.000424, "shape": [ 11008, 2048 ], "sqnr": 16.425324 }, "model.layers.15.self_attn.k_proj": { "Wscale": 0.034472, "bpw": 3.0, "proxy_loss": 0.007449, "shape": [ 512, 2048 ], "sqnr": 16.258316 }, "model.layers.15.self_attn.o_proj": { "Wscale": 0.013101, "bpw": 3.0, "proxy_loss": 1.6e-05, "shape": [ 2048, 2048 ], "sqnr": 15.560248 }, "model.layers.15.self_attn.q_proj": { "Wscale": 0.023913, "bpw": 3.0, "proxy_loss": 0.003574, "shape": [ 2048, 2048 ], "sqnr": 16.263779 }, "model.layers.15.self_attn.v_proj": { "Wscale": 0.011644, "bpw": 3.0, "proxy_loss": 0.000846, "shape": [ 512, 2048 ], "sqnr": 16.262602 }, "model.layers.16.mlp.down_proj": { "Wscale": 0.014593, "bpw": 3.0, "proxy_loss": 3.2e-05, "shape": [ 2048, 11008 ], "sqnr": 16.41991 }, "model.layers.16.mlp.gate_proj": { "Wscale": 0.016583, "bpw": 3.0, "proxy_loss": 0.000557, "shape": [ 11008, 2048 ], "sqnr": 16.457335 }, "model.layers.16.mlp.up_proj": { "Wscale": 0.01453, "bpw": 3.0, "proxy_loss": 0.000428, "shape": [ 11008, 2048 ], "sqnr": 16.460988 }, "model.layers.16.self_attn.k_proj": { "Wscale": 0.034236, "bpw": 3.0, "proxy_loss": 0.003839, "shape": [ 512, 2048 ], "sqnr": 16.053092 }, "model.layers.16.self_attn.o_proj": { "Wscale": 0.013279, "bpw": 3.0, "proxy_loss": 1.8e-05, "shape": [ 2048, 2048 ], "sqnr": 15.562231 }, "model.layers.16.self_attn.q_proj": { "Wscale": 0.021995, "bpw": 3.0, "proxy_loss": 0.001579, "shape": [ 2048, 2048 ], "sqnr": 16.056988 }, "model.layers.16.self_attn.v_proj": { "Wscale": 0.012093, "bpw": 3.0, "proxy_loss": 0.000476, "shape": [ 512, 2048 ], "sqnr": 16.062365 }, "model.layers.17.mlp.down_proj": { "Wscale": 0.014538, "bpw": 3.0, "proxy_loss": 3.6e-05, "shape": [ 2048, 11008 ], "sqnr": 16.257904 }, "model.layers.17.mlp.gate_proj": { "Wscale": 0.016788, "bpw": 3.0, "proxy_loss": 0.000578, "shape": [ 11008, 2048 ], "sqnr": 16.501372 }, "model.layers.17.mlp.up_proj": { "Wscale": 0.014539, "bpw": 3.0, "proxy_loss": 0.000433, "shape": [ 11008, 2048 ], "sqnr": 16.50583 }, "model.layers.17.self_attn.k_proj": { "Wscale": 0.035708, "bpw": 3.0, "proxy_loss": 0.003885, "shape": [ 512, 2048 ], "sqnr": 16.16343 }, "model.layers.17.self_attn.o_proj": { "Wscale": 0.013605, "bpw": 3.0, "proxy_loss": 2.1e-05, "shape": [ 2048, 2048 ], "sqnr": 15.694539 }, "model.layers.17.self_attn.q_proj": { "Wscale": 0.024334, "bpw": 3.0, "proxy_loss": 0.001803, "shape": [ 2048, 2048 ], "sqnr": 16.170489 }, "model.layers.17.self_attn.v_proj": { "Wscale": 0.013277, "bpw": 3.0, "proxy_loss": 0.000534, "shape": [ 512, 2048 ], "sqnr": 16.179676 }, "model.layers.18.mlp.down_proj": { "Wscale": 0.014545, "bpw": 3.0, "proxy_loss": 3.7e-05, "shape": [ 2048, 11008 ], "sqnr": 16.25946 }, "model.layers.18.mlp.gate_proj": { "Wscale": 0.016799, "bpw": 3.0, "proxy_loss": 0.0006, "shape": [ 11008, 2048 ], "sqnr": 16.410172 }, "model.layers.18.mlp.up_proj": { "Wscale": 0.014503, "bpw": 3.0, "proxy_loss": 0.000447, "shape": [ 11008, 2048 ], "sqnr": 16.415353 }, "model.layers.18.self_attn.k_proj": { "Wscale": 0.033211, "bpw": 3.0, "proxy_loss": 0.004243, "shape": [ 512, 2048 ], "sqnr": 16.11818 }, "model.layers.18.self_attn.o_proj": { "Wscale": 0.013911, "bpw": 3.0, "proxy_loss": 2.8e-05, "shape": [ 2048, 2048 ], "sqnr": 15.566124 }, "model.layers.18.self_attn.q_proj": { "Wscale": 0.022763, "bpw": 3.0, "proxy_loss": 0.001982, "shape": [ 2048, 2048 ], "sqnr": 16.132007 }, "model.layers.18.self_attn.v_proj": { "Wscale": 0.013808, "bpw": 3.0, "proxy_loss": 0.000725, "shape": [ 512, 2048 ], "sqnr": 16.138995 }, "model.layers.19.mlp.down_proj": { "Wscale": 0.014457, "bpw": 3.0, "proxy_loss": 4.2e-05, "shape": [ 2048, 11008 ], "sqnr": 16.222117 }, "model.layers.19.mlp.gate_proj": { "Wscale": 0.01694, "bpw": 3.0, "proxy_loss": 0.000669, "shape": [ 11008, 2048 ], "sqnr": 16.420471 }, "model.layers.19.mlp.up_proj": { "Wscale": 0.014474, "bpw": 3.0, "proxy_loss": 0.000488, "shape": [ 11008, 2048 ], "sqnr": 16.424192 }, "model.layers.19.self_attn.k_proj": { "Wscale": 0.034483, "bpw": 3.0, "proxy_loss": 0.006752, "shape": [ 512, 2048 ], "sqnr": 16.225473 }, "model.layers.19.self_attn.o_proj": { "Wscale": 0.013581, "bpw": 3.0, "proxy_loss": 1.7e-05, "shape": [ 2048, 2048 ], "sqnr": 15.352335 }, "model.layers.19.self_attn.q_proj": { "Wscale": 0.022979, "bpw": 3.0, "proxy_loss": 0.002986, "shape": [ 2048, 2048 ], "sqnr": 16.239666 }, "model.layers.19.self_attn.v_proj": { "Wscale": 0.012381, "bpw": 3.0, "proxy_loss": 0.000866, "shape": [ 512, 2048 ], "sqnr": 16.240637 }, "model.layers.2.mlp.down_proj": { "Wscale": 0.014949, "bpw": 3.0, "proxy_loss": 1.5e-05, "shape": [ 2048, 11008 ], "sqnr": 16.353872 }, "model.layers.2.mlp.gate_proj": { "Wscale": 0.015961, "bpw": 3.0, "proxy_loss": 0.000406, "shape": [ 11008, 2048 ], "sqnr": 16.404128 }, "model.layers.2.mlp.up_proj": { "Wscale": 0.014871, "bpw": 3.0, "proxy_loss": 0.000352, "shape": [ 11008, 2048 ], "sqnr": 16.408822 }, "model.layers.2.self_attn.k_proj": { "Wscale": 0.031559, "bpw": 3.0, "proxy_loss": 0.00177, "shape": [ 512, 2048 ], "sqnr": 16.152846 }, "model.layers.2.self_attn.o_proj": { "Wscale": 0.01318, "bpw": 3.0, "proxy_loss": 4e-06, "shape": [ 2048, 2048 ], "sqnr": 15.293905 }, "model.layers.2.self_attn.q_proj": { "Wscale": 0.021809, "bpw": 3.0, "proxy_loss": 0.000844, "shape": [ 2048, 2048 ], "sqnr": 16.156492 }, "model.layers.2.self_attn.v_proj": { "Wscale": 0.013512, "bpw": 3.0, "proxy_loss": 0.000322, "shape": [ 512, 2048 ], "sqnr": 16.170732 }, "model.layers.20.mlp.down_proj": { "Wscale": 0.014639, "bpw": 3.0, "proxy_loss": 4.3e-05, "shape": [ 2048, 11008 ], "sqnr": 16.335384 }, "model.layers.20.mlp.gate_proj": { "Wscale": 0.016649, "bpw": 3.0, "proxy_loss": 0.000651, "shape": [ 11008, 2048 ], "sqnr": 16.391441 }, "model.layers.20.mlp.up_proj": { "Wscale": 0.014657, "bpw": 3.0, "proxy_loss": 0.000505, "shape": [ 11008, 2048 ], "sqnr": 16.393546 }, "model.layers.20.self_attn.k_proj": { "Wscale": 0.034138, "bpw": 3.0, "proxy_loss": 0.004333, "shape": [ 512, 2048 ], "sqnr": 16.186611 }, "model.layers.20.self_attn.o_proj": { "Wscale": 0.013378, "bpw": 3.0, "proxy_loss": 2e-05, "shape": [ 2048, 2048 ], "sqnr": 15.680858 }, "model.layers.20.self_attn.q_proj": { "Wscale": 0.023037, "bpw": 3.0, "proxy_loss": 0.001965, "shape": [ 2048, 2048 ], "sqnr": 16.193171 }, "model.layers.20.self_attn.v_proj": { "Wscale": 0.011907, "bpw": 3.0, "proxy_loss": 0.000524, "shape": [ 512, 2048 ], "sqnr": 16.201203 }, "model.layers.21.mlp.down_proj": { "Wscale": 0.014397, "bpw": 3.0, "proxy_loss": 5.5e-05, "shape": [ 2048, 11008 ], "sqnr": 16.193762 }, "model.layers.21.mlp.gate_proj": { "Wscale": 0.017201, "bpw": 3.0, "proxy_loss": 0.000756, "shape": [ 11008, 2048 ], "sqnr": 16.390568 }, "model.layers.21.mlp.up_proj": { "Wscale": 0.014438, "bpw": 3.0, "proxy_loss": 0.000532, "shape": [ 11008, 2048 ], "sqnr": 16.39103 }, "model.layers.21.self_attn.k_proj": { "Wscale": 0.036185, "bpw": 3.0, "proxy_loss": 0.003894, "shape": [ 512, 2048 ], "sqnr": 16.259109 }, "model.layers.21.self_attn.o_proj": { "Wscale": 0.013652, "bpw": 3.0, "proxy_loss": 1.8e-05, "shape": [ 2048, 2048 ], "sqnr": 15.640369 }, "model.layers.21.self_attn.q_proj": { "Wscale": 0.024717, "bpw": 3.0, "proxy_loss": 0.001808, "shape": [ 2048, 2048 ], "sqnr": 16.266771 }, "model.layers.21.self_attn.v_proj": { "Wscale": 0.013529, "bpw": 3.0, "proxy_loss": 0.000538, "shape": [ 512, 2048 ], "sqnr": 16.281696 }, "model.layers.22.mlp.down_proj": { "Wscale": 0.0144, "bpw": 3.0, "proxy_loss": 5.6e-05, "shape": [ 2048, 11008 ], "sqnr": 16.307859 }, "model.layers.22.mlp.gate_proj": { "Wscale": 0.017321, "bpw": 3.0, "proxy_loss": 0.000807, "shape": [ 11008, 2048 ], "sqnr": 16.333458 }, "model.layers.22.mlp.up_proj": { "Wscale": 0.014461, "bpw": 3.0, "proxy_loss": 0.000562, "shape": [ 11008, 2048 ], "sqnr": 16.337842 }, "model.layers.22.self_attn.k_proj": { "Wscale": 0.032841, "bpw": 3.0, "proxy_loss": 0.003594, "shape": [ 512, 2048 ], "sqnr": 16.078242 }, "model.layers.22.self_attn.o_proj": { "Wscale": 0.013157, "bpw": 3.0, "proxy_loss": 1.6e-05, "shape": [ 2048, 2048 ], "sqnr": 15.308209 }, "model.layers.22.self_attn.q_proj": { "Wscale": 0.022243, "bpw": 3.0, "proxy_loss": 0.001642, "shape": [ 2048, 2048 ], "sqnr": 16.076979 }, "model.layers.22.self_attn.v_proj": { "Wscale": 0.012042, "bpw": 3.0, "proxy_loss": 0.000481, "shape": [ 512, 2048 ], "sqnr": 16.08387 }, "model.layers.23.mlp.down_proj": { "Wscale": 0.014438, "bpw": 3.0, "proxy_loss": 6.6e-05, "shape": [ 2048, 11008 ], "sqnr": 16.225916 }, "model.layers.23.mlp.gate_proj": { "Wscale": 0.017227, "bpw": 3.0, "proxy_loss": 0.000891, "shape": [ 11008, 2048 ], "sqnr": 16.348396 }, "model.layers.23.mlp.up_proj": { "Wscale": 0.014516, "bpw": 3.0, "proxy_loss": 0.000632, "shape": [ 11008, 2048 ], "sqnr": 16.35246 }, "model.layers.23.self_attn.k_proj": { "Wscale": 0.033162, "bpw": 3.0, "proxy_loss": 0.007295, "shape": [ 512, 2048 ], "sqnr": 16.166098 }, "model.layers.23.self_attn.o_proj": { "Wscale": 0.013642, "bpw": 3.0, "proxy_loss": 1.3e-05, "shape": [ 2048, 2048 ], "sqnr": 15.348972 }, "model.layers.23.self_attn.q_proj": { "Wscale": 0.023855, "bpw": 3.0, "proxy_loss": 0.003763, "shape": [ 2048, 2048 ], "sqnr": 16.176726 }, "model.layers.23.self_attn.v_proj": { "Wscale": 0.012018, "bpw": 3.0, "proxy_loss": 0.000955, "shape": [ 512, 2048 ], "sqnr": 16.182205 }, "model.layers.24.mlp.down_proj": { "Wscale": 0.01422, "bpw": 3.0, "proxy_loss": 7e-05, "shape": [ 2048, 11008 ], "sqnr": 16.184447 }, "model.layers.24.mlp.gate_proj": { "Wscale": 0.017539, "bpw": 3.0, "proxy_loss": 0.000975, "shape": [ 11008, 2048 ], "sqnr": 16.349035 }, "model.layers.24.mlp.up_proj": { "Wscale": 0.01429, "bpw": 3.0, "proxy_loss": 0.000647, "shape": [ 11008, 2048 ], "sqnr": 16.354077 }, "model.layers.24.self_attn.k_proj": { "Wscale": 0.033771, "bpw": 3.0, "proxy_loss": 0.003613, "shape": [ 512, 2048 ], "sqnr": 16.159628 }, "model.layers.24.self_attn.o_proj": { "Wscale": 0.013274, "bpw": 3.0, "proxy_loss": 1e-05, "shape": [ 2048, 2048 ], "sqnr": 15.363759 }, "model.layers.24.self_attn.q_proj": { "Wscale": 0.023728, "bpw": 3.0, "proxy_loss": 0.001777, "shape": [ 2048, 2048 ], "sqnr": 16.15815 }, "model.layers.24.self_attn.v_proj": { "Wscale": 0.013162, "bpw": 3.0, "proxy_loss": 0.000546, "shape": [ 512, 2048 ], "sqnr": 16.177525 }, "model.layers.25.mlp.down_proj": { "Wscale": 0.01435, "bpw": 3.0, "proxy_loss": 6.7e-05, "shape": [ 2048, 11008 ], "sqnr": 16.314953 }, "model.layers.25.mlp.gate_proj": { "Wscale": 0.017328, "bpw": 3.0, "proxy_loss": 0.000962, "shape": [ 11008, 2048 ], "sqnr": 16.369565 }, "model.layers.25.mlp.up_proj": { "Wscale": 0.014426, "bpw": 3.0, "proxy_loss": 0.000667, "shape": [ 11008, 2048 ], "sqnr": 16.376393 }, "model.layers.25.self_attn.k_proj": { "Wscale": 0.032293, "bpw": 3.0, "proxy_loss": 0.004474, "shape": [ 512, 2048 ], "sqnr": 16.09733 }, "model.layers.25.self_attn.o_proj": { "Wscale": 0.013674, "bpw": 3.0, "proxy_loss": 1.1e-05, "shape": [ 2048, 2048 ], "sqnr": 15.472283 }, "model.layers.25.self_attn.q_proj": { "Wscale": 0.022724, "bpw": 3.0, "proxy_loss": 0.002193, "shape": [ 2048, 2048 ], "sqnr": 16.107278 }, "model.layers.25.self_attn.v_proj": { "Wscale": 0.012828, "bpw": 3.0, "proxy_loss": 0.000704, "shape": [ 512, 2048 ], "sqnr": 16.111837 }, "model.layers.26.mlp.down_proj": { "Wscale": 0.014444, "bpw": 3.0, "proxy_loss": 6.8e-05, "shape": [ 2048, 11008 ], "sqnr": 16.33089 }, "model.layers.26.mlp.gate_proj": { "Wscale": 0.017209, "bpw": 3.0, "proxy_loss": 0.000953, "shape": [ 11008, 2048 ], "sqnr": 16.392397 }, "model.layers.26.mlp.up_proj": { "Wscale": 0.014529, "bpw": 3.0, "proxy_loss": 0.000679, "shape": [ 11008, 2048 ], "sqnr": 16.396896 }, "model.layers.26.self_attn.k_proj": { "Wscale": 0.030631, "bpw": 3.0, "proxy_loss": 0.00331, "shape": [ 512, 2048 ], "sqnr": 16.094051 }, "model.layers.26.self_attn.o_proj": { "Wscale": 0.013788, "bpw": 3.0, "proxy_loss": 1e-05, "shape": [ 2048, 2048 ], "sqnr": 15.544758 }, "model.layers.26.self_attn.q_proj": { "Wscale": 0.022249, "bpw": 3.0, "proxy_loss": 0.001734, "shape": [ 2048, 2048 ], "sqnr": 16.098124 }, "model.layers.26.self_attn.v_proj": { "Wscale": 0.013756, "bpw": 3.0, "proxy_loss": 0.000667, "shape": [ 512, 2048 ], "sqnr": 16.099781 }, "model.layers.27.mlp.down_proj": { "Wscale": 0.014627, "bpw": 3.0, "proxy_loss": 6.8e-05, "shape": [ 2048, 11008 ], "sqnr": 16.500582 }, "model.layers.27.mlp.gate_proj": { "Wscale": 0.016948, "bpw": 3.0, "proxy_loss": 0.000949, "shape": [ 11008, 2048 ], "sqnr": 16.385253 }, "model.layers.27.mlp.up_proj": { "Wscale": 0.014689, "bpw": 3.0, "proxy_loss": 0.000713, "shape": [ 11008, 2048 ], "sqnr": 16.384162 }, "model.layers.27.self_attn.k_proj": { "Wscale": 0.031096, "bpw": 3.0, "proxy_loss": 0.004905, "shape": [ 512, 2048 ], "sqnr": 16.09483 }, "model.layers.27.self_attn.o_proj": { "Wscale": 0.013563, "bpw": 3.0, "proxy_loss": 8e-06, "shape": [ 2048, 2048 ], "sqnr": 15.586515 }, "model.layers.27.self_attn.q_proj": { "Wscale": 0.023844, "bpw": 3.0, "proxy_loss": 0.002868, "shape": [ 2048, 2048 ], "sqnr": 16.1028 }, "model.layers.27.self_attn.v_proj": { "Wscale": 0.012426, "bpw": 3.0, "proxy_loss": 0.000782, "shape": [ 512, 2048 ], "sqnr": 16.103762 }, "model.layers.28.mlp.down_proj": { "Wscale": 0.014632, "bpw": 3.0, "proxy_loss": 7.6e-05, "shape": [ 2048, 11008 ], "sqnr": 16.503192 }, "model.layers.28.mlp.gate_proj": { "Wscale": 0.017053, "bpw": 3.0, "proxy_loss": 0.001039, "shape": [ 11008, 2048 ], "sqnr": 16.428011 }, "model.layers.28.mlp.up_proj": { "Wscale": 0.014685, "bpw": 3.0, "proxy_loss": 0.00077, "shape": [ 11008, 2048 ], "sqnr": 16.431559 }, "model.layers.28.self_attn.k_proj": { "Wscale": 0.030122, "bpw": 3.0, "proxy_loss": 0.003625, "shape": [ 512, 2048 ], "sqnr": 16.15488 }, "model.layers.28.self_attn.o_proj": { "Wscale": 0.014081, "bpw": 3.0, "proxy_loss": 1.2e-05, "shape": [ 2048, 2048 ], "sqnr": 15.796622 }, "model.layers.28.self_attn.q_proj": { "Wscale": 0.021981, "bpw": 3.0, "proxy_loss": 0.00192, "shape": [ 2048, 2048 ], "sqnr": 16.156195 }, "model.layers.28.self_attn.v_proj": { "Wscale": 0.014247, "bpw": 3.0, "proxy_loss": 0.000808, "shape": [ 512, 2048 ], "sqnr": 16.162043 }, "model.layers.29.mlp.down_proj": { "Wscale": 0.014736, "bpw": 3.0, "proxy_loss": 9.6e-05, "shape": [ 2048, 11008 ], "sqnr": 16.465403 }, "model.layers.29.mlp.gate_proj": { "Wscale": 0.017071, "bpw": 3.0, "proxy_loss": 0.001151, "shape": [ 11008, 2048 ], "sqnr": 16.442001 }, "model.layers.29.mlp.up_proj": { "Wscale": 0.014774, "bpw": 3.0, "proxy_loss": 0.000862, "shape": [ 11008, 2048 ], "sqnr": 16.44553 }, "model.layers.29.self_attn.k_proj": { "Wscale": 0.031774, "bpw": 3.0, "proxy_loss": 0.004058, "shape": [ 512, 2048 ], "sqnr": 16.233772 }, "model.layers.29.self_attn.o_proj": { "Wscale": 0.013524, "bpw": 3.0, "proxy_loss": 1e-05, "shape": [ 2048, 2048 ], "sqnr": 15.71085 }, "model.layers.29.self_attn.q_proj": { "Wscale": 0.02245, "bpw": 3.0, "proxy_loss": 0.002017, "shape": [ 2048, 2048 ], "sqnr": 16.244967 }, "model.layers.29.self_attn.v_proj": { "Wscale": 0.014011, "bpw": 3.0, "proxy_loss": 0.000787, "shape": [ 512, 2048 ], "sqnr": 16.239334 }, "model.layers.3.mlp.down_proj": { "Wscale": 0.014403, "bpw": 3.0, "proxy_loss": 1.5e-05, "shape": [ 2048, 11008 ], "sqnr": 16.060005 }, "model.layers.3.mlp.gate_proj": { "Wscale": 0.016393, "bpw": 3.0, "proxy_loss": 0.000415, "shape": [ 11008, 2048 ], "sqnr": 16.488256 }, "model.layers.3.mlp.up_proj": { "Wscale": 0.014421, "bpw": 3.0, "proxy_loss": 0.000321, "shape": [ 11008, 2048 ], "sqnr": 16.492353 }, "model.layers.3.self_attn.k_proj": { "Wscale": 0.033231, "bpw": 3.0, "proxy_loss": 0.003659, "shape": [ 512, 2048 ], "sqnr": 16.388597 }, "model.layers.3.self_attn.o_proj": { "Wscale": 0.013614, "bpw": 3.0, "proxy_loss": 1e-06, "shape": [ 2048, 2048 ], "sqnr": 15.563037 }, "model.layers.3.self_attn.q_proj": { "Wscale": 0.022703, "bpw": 3.0, "proxy_loss": 0.001708, "shape": [ 2048, 2048 ], "sqnr": 16.381277 }, "model.layers.3.self_attn.v_proj": { "Wscale": 0.01401, "bpw": 3.0, "proxy_loss": 0.00065, "shape": [ 512, 2048 ], "sqnr": 16.388187 }, "model.layers.30.mlp.down_proj": { "Wscale": 0.014914, "bpw": 3.0, "proxy_loss": 0.000119, "shape": [ 2048, 11008 ], "sqnr": 16.376814 }, "model.layers.30.mlp.gate_proj": { "Wscale": 0.016989, "bpw": 3.0, "proxy_loss": 0.001136, "shape": [ 11008, 2048 ], "sqnr": 16.426447 }, "model.layers.30.mlp.up_proj": { "Wscale": 0.014984, "bpw": 3.0, "proxy_loss": 0.000884, "shape": [ 11008, 2048 ], "sqnr": 16.429016 }, "model.layers.30.self_attn.k_proj": { "Wscale": 0.032257, "bpw": 3.0, "proxy_loss": 0.003839, "shape": [ 512, 2048 ], "sqnr": 16.293894 }, "model.layers.30.self_attn.o_proj": { "Wscale": 0.014285, "bpw": 3.0, "proxy_loss": 1.7e-05, "shape": [ 2048, 2048 ], "sqnr": 15.695345 }, "model.layers.30.self_attn.q_proj": { "Wscale": 0.023235, "bpw": 3.0, "proxy_loss": 0.001991, "shape": [ 2048, 2048 ], "sqnr": 16.294406 }, "model.layers.30.self_attn.v_proj": { "Wscale": 0.015631, "bpw": 3.0, "proxy_loss": 0.000896, "shape": [ 512, 2048 ], "sqnr": 16.308414 }, "model.layers.31.mlp.down_proj": { "Wscale": 0.015023, "bpw": 3.0, "proxy_loss": 0.000129, "shape": [ 2048, 11008 ], "sqnr": 16.513153 }, "model.layers.31.mlp.gate_proj": { "Wscale": 0.016889, "bpw": 3.0, "proxy_loss": 0.001205, "shape": [ 11008, 2048 ], "sqnr": 16.436233 }, "model.layers.31.mlp.up_proj": { "Wscale": 0.015085, "bpw": 3.0, "proxy_loss": 0.000961, "shape": [ 11008, 2048 ], "sqnr": 16.438225 }, "model.layers.31.self_attn.k_proj": { "Wscale": 0.028371, "bpw": 3.0, "proxy_loss": 0.004192, "shape": [ 512, 2048 ], "sqnr": 16.102209 }, "model.layers.31.self_attn.o_proj": { "Wscale": 0.01563, "bpw": 3.0, "proxy_loss": 2.3e-05, "shape": [ 2048, 2048 ], "sqnr": 15.871659 }, "model.layers.31.self_attn.q_proj": { "Wscale": 0.021253, "bpw": 3.0, "proxy_loss": 0.002336, "shape": [ 2048, 2048 ], "sqnr": 16.111998 }, "model.layers.31.self_attn.v_proj": { "Wscale": 0.016801, "bpw": 3.0, "proxy_loss": 0.00146, "shape": [ 512, 2048 ], "sqnr": 16.123512 }, "model.layers.32.mlp.down_proj": { "Wscale": 0.015064, "bpw": 3.0, "proxy_loss": 0.000146, "shape": [ 2048, 11008 ], "sqnr": 16.556373 }, "model.layers.32.mlp.gate_proj": { "Wscale": 0.016951, "bpw": 3.0, "proxy_loss": 0.001306, "shape": [ 11008, 2048 ], "sqnr": 16.471859 }, "model.layers.32.mlp.up_proj": { "Wscale": 0.015133, "bpw": 3.0, "proxy_loss": 0.001041, "shape": [ 11008, 2048 ], "sqnr": 16.472679 }, "model.layers.32.self_attn.k_proj": { "Wscale": 0.029694, "bpw": 3.0, "proxy_loss": 0.003571, "shape": [ 512, 2048 ], "sqnr": 16.273612 }, "model.layers.32.self_attn.o_proj": { "Wscale": 0.014921, "bpw": 3.0, "proxy_loss": 1.9e-05, "shape": [ 2048, 2048 ], "sqnr": 15.93343 }, "model.layers.32.self_attn.q_proj": { "Wscale": 0.021476, "bpw": 3.0, "proxy_loss": 0.001862, "shape": [ 2048, 2048 ], "sqnr": 16.273438 }, "model.layers.32.self_attn.v_proj": { "Wscale": 0.0166, "bpw": 3.0, "proxy_loss": 0.001111, "shape": [ 512, 2048 ], "sqnr": 16.289821 }, "model.layers.33.mlp.down_proj": { "Wscale": 0.01498, "bpw": 3.0, "proxy_loss": 0.000173, "shape": [ 2048, 11008 ], "sqnr": 16.537255 }, "model.layers.33.mlp.gate_proj": { "Wscale": 0.0174, "bpw": 3.0, "proxy_loss": 0.001513, "shape": [ 11008, 2048 ], "sqnr": 16.516341 }, "model.layers.33.mlp.up_proj": { "Wscale": 0.015263, "bpw": 3.0, "proxy_loss": 0.001163, "shape": [ 11008, 2048 ], "sqnr": 16.518073 }, "model.layers.33.self_attn.k_proj": { "Wscale": 0.028366, "bpw": 3.0, "proxy_loss": 0.003468, "shape": [ 512, 2048 ], "sqnr": 16.266572 }, "model.layers.33.self_attn.o_proj": { "Wscale": 0.016032, "bpw": 3.0, "proxy_loss": 2.9e-05, "shape": [ 2048, 2048 ], "sqnr": 15.892727 }, "model.layers.33.self_attn.q_proj": { "Wscale": 0.020934, "bpw": 3.0, "proxy_loss": 0.001878, "shape": [ 2048, 2048 ], "sqnr": 16.284419 }, "model.layers.33.self_attn.v_proj": { "Wscale": 0.018881, "bpw": 3.0, "proxy_loss": 0.001524, "shape": [ 512, 2048 ], "sqnr": 16.295207 }, "model.layers.34.mlp.down_proj": { "Wscale": 0.014884, "bpw": 3.0, "proxy_loss": 0.000181, "shape": [ 2048, 11008 ], "sqnr": 16.348543 }, "model.layers.34.mlp.gate_proj": { "Wscale": 0.0178, "bpw": 3.0, "proxy_loss": 0.001456, "shape": [ 11008, 2048 ], "sqnr": 16.47947 }, "model.layers.34.mlp.up_proj": { "Wscale": 0.015258, "bpw": 3.0, "proxy_loss": 0.001069, "shape": [ 11008, 2048 ], "sqnr": 16.482276 }, "model.layers.34.self_attn.k_proj": { "Wscale": 0.03412, "bpw": 3.0, "proxy_loss": 0.003231, "shape": [ 512, 2048 ], "sqnr": 16.344725 }, "model.layers.34.self_attn.o_proj": { "Wscale": 0.015533, "bpw": 3.0, "proxy_loss": 3.7e-05, "shape": [ 2048, 2048 ], "sqnr": 15.679779 }, "model.layers.34.self_attn.q_proj": { "Wscale": 0.025445, "bpw": 3.0, "proxy_loss": 0.001802, "shape": [ 2048, 2048 ], "sqnr": 16.339902 }, "model.layers.34.self_attn.v_proj": { "Wscale": 0.017775, "bpw": 3.0, "proxy_loss": 0.000871, "shape": [ 512, 2048 ], "sqnr": 16.366227 }, "model.layers.35.mlp.down_proj": { "Wscale": 0.014823, "bpw": 3.0, "proxy_loss": 0.000221, "shape": [ 2048, 11008 ], "sqnr": 16.240917 }, "model.layers.35.mlp.gate_proj": { "Wscale": 0.019276, "bpw": 3.0, "proxy_loss": 0.001233, "shape": [ 11008, 2048 ], "sqnr": 16.295885 }, "model.layers.35.mlp.up_proj": { "Wscale": 0.016341, "bpw": 3.0, "proxy_loss": 0.000885, "shape": [ 11008, 2048 ], "sqnr": 16.305791 }, "model.layers.35.self_attn.k_proj": { "Wscale": 0.032016, "bpw": 3.0, "proxy_loss": 0.004511, "shape": [ 512, 2048 ], "sqnr": 16.22501 }, "model.layers.35.self_attn.o_proj": { "Wscale": 0.013245, "bpw": 3.0, "proxy_loss": 8e-06, "shape": [ 2048, 2048 ], "sqnr": 15.45814 }, "model.layers.35.self_attn.q_proj": { "Wscale": 0.02146, "bpw": 3.0, "proxy_loss": 0.002022, "shape": [ 2048, 2048 ], "sqnr": 16.222849 }, "model.layers.35.self_attn.v_proj": { "Wscale": 0.011698, "bpw": 3.0, "proxy_loss": 0.000599, "shape": [ 512, 2048 ], "sqnr": 16.221338 }, "model.layers.4.mlp.down_proj": { "Wscale": 0.014383, "bpw": 3.0, "proxy_loss": 1.5e-05, "shape": [ 2048, 11008 ], "sqnr": 16.141519 }, "model.layers.4.mlp.gate_proj": { "Wscale": 0.016389, "bpw": 3.0, "proxy_loss": 0.000401, "shape": [ 11008, 2048 ], "sqnr": 16.426074 }, "model.layers.4.mlp.up_proj": { "Wscale": 0.014414, "bpw": 3.0, "proxy_loss": 0.00031, "shape": [ 11008, 2048 ], "sqnr": 16.432068 }, "model.layers.4.self_attn.k_proj": { "Wscale": 0.031686, "bpw": 3.0, "proxy_loss": 0.002198, "shape": [ 512, 2048 ], "sqnr": 16.094149 }, "model.layers.4.self_attn.o_proj": { "Wscale": 0.012773, "bpw": 3.0, "proxy_loss": 5e-06, "shape": [ 2048, 2048 ], "sqnr": 15.647254 }, "model.layers.4.self_attn.q_proj": { "Wscale": 0.022577, "bpw": 3.0, "proxy_loss": 0.001114, "shape": [ 2048, 2048 ], "sqnr": 16.097742 }, "model.layers.4.self_attn.v_proj": { "Wscale": 0.012647, "bpw": 3.0, "proxy_loss": 0.000346, "shape": [ 512, 2048 ], "sqnr": 16.107205 }, "model.layers.5.mlp.down_proj": { "Wscale": 0.014986, "bpw": 3.0, "proxy_loss": 1.7e-05, "shape": [ 2048, 11008 ], "sqnr": 16.491439 }, "model.layers.5.mlp.gate_proj": { "Wscale": 0.015789, "bpw": 3.0, "proxy_loss": 0.000407, "shape": [ 11008, 2048 ], "sqnr": 16.3911 }, "model.layers.5.mlp.up_proj": { "Wscale": 0.014979, "bpw": 3.0, "proxy_loss": 0.000366, "shape": [ 11008, 2048 ], "sqnr": 16.396702 }, "model.layers.5.self_attn.k_proj": { "Wscale": 0.032522, "bpw": 3.0, "proxy_loss": 0.002068, "shape": [ 512, 2048 ], "sqnr": 16.035451 }, "model.layers.5.self_attn.o_proj": { "Wscale": 0.012526, "bpw": 3.0, "proxy_loss": 4e-06, "shape": [ 2048, 2048 ], "sqnr": 15.739542 }, "model.layers.5.self_attn.q_proj": { "Wscale": 0.023167, "bpw": 3.0, "proxy_loss": 0.001046, "shape": [ 2048, 2048 ], "sqnr": 16.042417 }, "model.layers.5.self_attn.v_proj": { "Wscale": 0.011546, "bpw": 3.0, "proxy_loss": 0.000258, "shape": [ 512, 2048 ], "sqnr": 16.050871 }, "model.layers.6.mlp.down_proj": { "Wscale": 0.014807, "bpw": 3.0, "proxy_loss": 2.3e-05, "shape": [ 2048, 11008 ], "sqnr": 16.529375 }, "model.layers.6.mlp.gate_proj": { "Wscale": 0.016142, "bpw": 3.0, "proxy_loss": 0.000499, "shape": [ 11008, 2048 ], "sqnr": 16.491004 }, "model.layers.6.mlp.up_proj": { "Wscale": 0.014734, "bpw": 3.0, "proxy_loss": 0.000415, "shape": [ 11008, 2048 ], "sqnr": 16.494976 }, "model.layers.6.self_attn.k_proj": { "Wscale": 0.033402, "bpw": 3.0, "proxy_loss": 0.002684, "shape": [ 512, 2048 ], "sqnr": 16.294591 }, "model.layers.6.self_attn.o_proj": { "Wscale": 0.012967, "bpw": 3.0, "proxy_loss": 7e-06, "shape": [ 2048, 2048 ], "sqnr": 15.80415 }, "model.layers.6.self_attn.q_proj": { "Wscale": 0.023263, "bpw": 3.0, "proxy_loss": 0.001301, "shape": [ 2048, 2048 ], "sqnr": 16.298821 }, "model.layers.6.self_attn.v_proj": { "Wscale": 0.012316, "bpw": 3.0, "proxy_loss": 0.000363, "shape": [ 512, 2048 ], "sqnr": 16.313414 }, "model.layers.7.mlp.down_proj": { "Wscale": 0.014645, "bpw": 3.0, "proxy_loss": 2.9e-05, "shape": [ 2048, 11008 ], "sqnr": 16.46818 }, "model.layers.7.mlp.gate_proj": { "Wscale": 0.016595, "bpw": 3.0, "proxy_loss": 0.000627, "shape": [ 11008, 2048 ], "sqnr": 16.591195 }, "model.layers.7.mlp.up_proj": { "Wscale": 0.014585, "bpw": 3.0, "proxy_loss": 0.000484, "shape": [ 11008, 2048 ], "sqnr": 16.595986 }, "model.layers.7.self_attn.k_proj": { "Wscale": 0.033219, "bpw": 3.0, "proxy_loss": 0.006656, "shape": [ 512, 2048 ], "sqnr": 16.357637 }, "model.layers.7.self_attn.o_proj": { "Wscale": 0.012888, "bpw": 3.0, "proxy_loss": 3e-06, "shape": [ 2048, 2048 ], "sqnr": 15.228898 }, "model.layers.7.self_attn.q_proj": { "Wscale": 0.023099, "bpw": 3.0, "proxy_loss": 0.003221, "shape": [ 2048, 2048 ], "sqnr": 16.35449 }, "model.layers.7.self_attn.v_proj": { "Wscale": 0.011746, "bpw": 3.0, "proxy_loss": 0.000831, "shape": [ 512, 2048 ], "sqnr": 16.362304 }, "model.layers.8.mlp.down_proj": { "Wscale": 0.014334, "bpw": 3.0, "proxy_loss": 3.2e-05, "shape": [ 2048, 11008 ], "sqnr": 16.29709 }, "model.layers.8.mlp.gate_proj": { "Wscale": 0.01718, "bpw": 3.0, "proxy_loss": 0.000662, "shape": [ 11008, 2048 ], "sqnr": 16.594251 }, "model.layers.8.mlp.up_proj": { "Wscale": 0.014272, "bpw": 3.0, "proxy_loss": 0.000456, "shape": [ 11008, 2048 ], "sqnr": 16.598699 }, "model.layers.8.self_attn.k_proj": { "Wscale": 0.031551, "bpw": 3.0, "proxy_loss": 0.002005, "shape": [ 512, 2048 ], "sqnr": 16.324565 }, "model.layers.8.self_attn.o_proj": { "Wscale": 0.013474, "bpw": 3.0, "proxy_loss": 1.1e-05, "shape": [ 2048, 2048 ], "sqnr": 15.686315 }, "model.layers.8.self_attn.q_proj": { "Wscale": 0.023131, "bpw": 3.0, "proxy_loss": 0.001078, "shape": [ 2048, 2048 ], "sqnr": 16.336801 }, "model.layers.8.self_attn.v_proj": { "Wscale": 0.013682, "bpw": 3.0, "proxy_loss": 0.000374, "shape": [ 512, 2048 ], "sqnr": 16.342537 }, "model.layers.9.mlp.down_proj": { "Wscale": 0.014342, "bpw": 3.0, "proxy_loss": 3.3e-05, "shape": [ 2048, 11008 ], "sqnr": 16.361882 }, "model.layers.9.mlp.gate_proj": { "Wscale": 0.017038, "bpw": 3.0, "proxy_loss": 0.000646, "shape": [ 11008, 2048 ], "sqnr": 16.559933 }, "model.layers.9.mlp.up_proj": { "Wscale": 0.014321, "bpw": 3.0, "proxy_loss": 0.000456, "shape": [ 11008, 2048 ], "sqnr": 16.563313 }, "model.layers.9.self_attn.k_proj": { "Wscale": 0.03412, "bpw": 3.0, "proxy_loss": 0.003453, "shape": [ 512, 2048 ], "sqnr": 16.265587 }, "model.layers.9.self_attn.o_proj": { "Wscale": 0.013133, "bpw": 3.0, "proxy_loss": 1.5e-05, "shape": [ 2048, 2048 ], "sqnr": 15.87207 }, "model.layers.9.self_attn.q_proj": { "Wscale": 0.023516, "bpw": 3.0, "proxy_loss": 0.001634, "shape": [ 2048, 2048 ], "sqnr": 16.281388 }, "model.layers.9.self_attn.v_proj": { "Wscale": 0.012032, "bpw": 3.0, "proxy_loss": 0.000426, "shape": [ 512, 2048 ], "sqnr": 16.293029 } }