diff options
Diffstat (limited to 'results/snapshot_vit_v1')
| -rw-r--r-- | results/snapshot_vit_v1/run_s123.log | 36 | ||||
| -rw-r--r-- | results/snapshot_vit_v1/run_s42.log | 36 | ||||
| -rw-r--r-- | results/snapshot_vit_v1/run_s456.log | 36 | ||||
| -rw-r--r-- | results/snapshot_vit_v1/snapshot_fa_canonical_s42.json | 1104 | ||||
| -rw-r--r-- | results/snapshot_vit_v1/snapshot_fa_vit_s42.json | 1104 | ||||
| -rw-r--r-- | results/snapshot_vit_v1/snapshot_vit_s123.json | 4594 | ||||
| -rw-r--r-- | results/snapshot_vit_v1/snapshot_vit_s42.json | 4594 | ||||
| -rw-r--r-- | results/snapshot_vit_v1/snapshot_vit_s456.json | 4594 |
8 files changed, 16098 insertions, 0 deletions
diff --git a/results/snapshot_vit_v1/run_s123.log b/results/snapshot_vit_v1/run_s123.log new file mode 100644 index 0000000..e23f876 --- /dev/null +++ b/results/snapshot_vit_v1/run_s123.log @@ -0,0 +1,36 @@ +ViT-MINI: depth=4, d_model=128, n_heads=4, epochs=60, seed=123 + +=== BP training (ViT-Mini) === + n_params=809354 + [BP-vit] Ep 0: ||h_L_cls||=7.591e+00 ||g_2||=9.667e-04 acc=0.1250 + [BP-vit] Ep 1: ||h_L_cls||=3.242e+01 ||g_2||=1.457e-04 acc=0.4697 + [BP-vit] Ep 5: ||h_L_cls||=3.030e+01 ||g_2||=1.731e-04 acc=0.5957 + [BP-vit] Ep 10: ||h_L_cls||=2.876e+01 ||g_2||=1.742e-04 acc=0.6768 + [BP-vit] Ep 15: ||h_L_cls||=2.787e+01 ||g_2||=1.661e-04 acc=0.7129 + [BP-vit] Ep 20: ||h_L_cls||=2.435e+01 ||g_2||=1.685e-04 acc=0.7461 + [BP-vit] Ep 25: ||h_L_cls||=2.108e+01 ||g_2||=1.419e-04 acc=0.7627 + [BP-vit] Ep 30: ||h_L_cls||=1.855e+01 ||g_2||=1.543e-04 acc=0.7783 + [BP-vit] Ep 35: ||h_L_cls||=1.614e+01 ||g_2||=1.360e-04 acc=0.7939 + [BP-vit] Ep 40: ||h_L_cls||=1.470e+01 ||g_2||=1.014e-04 acc=0.7822 + [BP-vit] Ep 45: ||h_L_cls||=1.305e+01 ||g_2||=9.261e-05 acc=0.7822 + [BP-vit] Ep 50: ||h_L_cls||=1.282e+01 ||g_2||=4.534e-05 acc=0.7822 + [BP-vit] Ep 55: ||h_L_cls||=1.249e+01 ||g_2||=4.168e-05 acc=0.7910 + [BP-vit] Ep 60: ||h_L_cls||=1.250e+01 ||g_2||=3.572e-05 acc=0.7871 + +=== DFA training (ViT-Mini, block-level DFA) === + [DFA-vit] Ep 0: ||h_L_cls||=7.591e+00 ||g_2||=9.667e-04 acc=0.1250 + [DFA-vit] Ep 1: ||h_L_cls||=1.532e+04 ||g_2||=4.354e-07 acc=0.2158 γ=0.0090 + [DFA-vit] Ep 5: ||h_L_cls||=7.063e+05 ||g_2||=1.320e-08 acc=0.2197 γ=0.0009 + [DFA-vit] Ep 10: ||h_L_cls||=4.235e+06 ||g_2||=2.736e-09 acc=0.2373 γ=0.0020 + [DFA-vit] Ep 15: ||h_L_cls||=7.955e+06 ||g_2||=1.467e-09 acc=0.2266 γ=0.0020 + [DFA-vit] Ep 20: ||h_L_cls||=1.460e+07 ||g_2||=9.774e-10 acc=0.2139 γ=0.0018 + [DFA-vit] Ep 25: ||h_L_cls||=2.068e+07 ||g_2||=7.712e-10 acc=0.2080 γ=0.0013 + [DFA-vit] Ep 30: ||h_L_cls||=3.012e+07 ||g_2||=5.962e-10 acc=0.2139 γ=0.0010 + [DFA-vit] Ep 35: ||h_L_cls||=4.466e+07 ||g_2||=4.717e-10 acc=0.1396 γ=0.0040 + [DFA-vit] Ep 40: ||h_L_cls||=4.770e+07 ||g_2||=4.407e-10 acc=0.1963 γ=0.0032 + [DFA-vit] Ep 45: ||h_L_cls||=4.536e+07 ||g_2||=3.493e-10 acc=0.2031 γ=0.0021 + [DFA-vit] Ep 50: ||h_L_cls||=4.727e+07 ||g_2||=3.342e-10 acc=0.1904 γ=0.0017 + [DFA-vit] Ep 55: ||h_L_cls||=4.865e+07 ||g_2||=3.534e-10 acc=0.1982 γ=0.0019 + [DFA-vit] Ep 60: ||h_L_cls||=5.080e+07 ||g_2||=3.359e-10 acc=0.2021 γ=0.0017 + +Saved results/snapshot_vit_v1/snapshot_vit_s123.json diff --git a/results/snapshot_vit_v1/run_s42.log b/results/snapshot_vit_v1/run_s42.log new file mode 100644 index 0000000..2bd76dd --- /dev/null +++ b/results/snapshot_vit_v1/run_s42.log @@ -0,0 +1,36 @@ +ViT-MINI: depth=4, d_model=128, n_heads=4, epochs=60, seed=42 + +=== BP training (ViT-Mini) === + n_params=809354 + [BP-vit] Ep 0: ||h_L_cls||=7.124e+00 ||g_2||=9.976e-04 acc=0.1377 + [BP-vit] Ep 1: ||h_L_cls||=3.181e+01 ||g_2||=1.631e-04 acc=0.4502 + [BP-vit] Ep 5: ||h_L_cls||=3.236e+01 ||g_2||=1.881e-04 acc=0.6035 + [BP-vit] Ep 10: ||h_L_cls||=3.166e+01 ||g_2||=1.970e-04 acc=0.6846 + [BP-vit] Ep 15: ||h_L_cls||=2.804e+01 ||g_2||=1.690e-04 acc=0.7334 + [BP-vit] Ep 20: ||h_L_cls||=2.521e+01 ||g_2||=1.525e-04 acc=0.7637 + [BP-vit] Ep 25: ||h_L_cls||=2.196e+01 ||g_2||=1.524e-04 acc=0.7568 + [BP-vit] Ep 30: ||h_L_cls||=1.903e+01 ||g_2||=1.451e-04 acc=0.7822 + [BP-vit] Ep 35: ||h_L_cls||=1.669e+01 ||g_2||=1.225e-04 acc=0.7852 + [BP-vit] Ep 40: ||h_L_cls||=1.509e+01 ||g_2||=9.628e-05 acc=0.7998 + [BP-vit] Ep 45: ||h_L_cls||=1.395e+01 ||g_2||=5.592e-05 acc=0.8066 + [BP-vit] Ep 50: ||h_L_cls||=1.327e+01 ||g_2||=4.421e-05 acc=0.7998 + [BP-vit] Ep 55: ||h_L_cls||=1.307e+01 ||g_2||=3.098e-05 acc=0.7959 + [BP-vit] Ep 60: ||h_L_cls||=1.297e+01 ||g_2||=3.051e-05 acc=0.7988 + +=== DFA training (ViT-Mini, block-level DFA) === + [DFA-vit] Ep 0: ||h_L_cls||=7.124e+00 ||g_2||=9.976e-04 acc=0.1377 + [DFA-vit] Ep 1: ||h_L_cls||=1.008e+04 ||g_2||=6.152e-07 acc=0.2402 γ=0.0062 + [DFA-vit] Ep 5: ||h_L_cls||=3.966e+05 ||g_2||=1.957e-08 acc=0.2266 γ=0.0066 + [DFA-vit] Ep 10: ||h_L_cls||=1.798e+06 ||g_2||=4.677e-09 acc=0.2266 γ=0.0045 + [DFA-vit] Ep 15: ||h_L_cls||=7.772e+06 ||g_2||=1.681e-09 acc=0.2334 γ=0.0003 + [DFA-vit] Ep 20: ||h_L_cls||=1.204e+07 ||g_2||=1.003e-09 acc=0.2402 γ=-0.0002 + [DFA-vit] Ep 25: ||h_L_cls||=1.835e+07 ||g_2||=6.058e-10 acc=0.2451 γ=-0.0014 + [DFA-vit] Ep 30: ||h_L_cls||=2.502e+07 ||g_2||=4.425e-10 acc=0.2266 γ=-0.0011 + [DFA-vit] Ep 35: ||h_L_cls||=3.150e+07 ||g_2||=3.412e-10 acc=0.2344 γ=-0.0003 + [DFA-vit] Ep 40: ||h_L_cls||=3.778e+07 ||g_2||=3.106e-10 acc=0.2549 γ=-0.0003 + [DFA-vit] Ep 45: ||h_L_cls||=4.166e+07 ||g_2||=2.891e-10 acc=0.2344 γ=-0.0004 + [DFA-vit] Ep 50: ||h_L_cls||=4.633e+07 ||g_2||=2.688e-10 acc=0.2549 γ=0.0004 + [DFA-vit] Ep 55: ||h_L_cls||=4.796e+07 ||g_2||=2.616e-10 acc=0.2510 γ=0.0005 + [DFA-vit] Ep 60: ||h_L_cls||=4.761e+07 ||g_2||=2.622e-10 acc=0.2559 γ=0.0007 + +Saved results/snapshot_vit_v1/snapshot_vit_s42.json diff --git a/results/snapshot_vit_v1/run_s456.log b/results/snapshot_vit_v1/run_s456.log new file mode 100644 index 0000000..ba19529 --- /dev/null +++ b/results/snapshot_vit_v1/run_s456.log @@ -0,0 +1,36 @@ +ViT-MINI: depth=4, d_model=128, n_heads=4, epochs=60, seed=456 + +=== BP training (ViT-Mini) === + n_params=809354 + [BP-vit] Ep 0: ||h_L_cls||=6.653e+00 ||g_2||=1.039e-03 acc=0.0791 + [BP-vit] Ep 1: ||h_L_cls||=2.995e+01 ||g_2||=1.710e-04 acc=0.4590 + [BP-vit] Ep 5: ||h_L_cls||=3.062e+01 ||g_2||=1.845e-04 acc=0.6113 + [BP-vit] Ep 10: ||h_L_cls||=3.123e+01 ||g_2||=1.730e-04 acc=0.6709 + [BP-vit] Ep 15: ||h_L_cls||=2.697e+01 ||g_2||=1.889e-04 acc=0.7012 + [BP-vit] Ep 20: ||h_L_cls||=2.387e+01 ||g_2||=1.801e-04 acc=0.7295 + [BP-vit] Ep 25: ||h_L_cls||=2.080e+01 ||g_2||=1.754e-04 acc=0.7578 + [BP-vit] Ep 30: ||h_L_cls||=1.773e+01 ||g_2||=1.553e-04 acc=0.7666 + [BP-vit] Ep 35: ||h_L_cls||=1.553e+01 ||g_2||=1.576e-04 acc=0.7705 + [BP-vit] Ep 40: ||h_L_cls||=1.420e+01 ||g_2||=1.193e-04 acc=0.7812 + [BP-vit] Ep 45: ||h_L_cls||=1.271e+01 ||g_2||=9.615e-05 acc=0.7773 + [BP-vit] Ep 50: ||h_L_cls||=1.230e+01 ||g_2||=7.114e-05 acc=0.8008 + [BP-vit] Ep 55: ||h_L_cls||=1.201e+01 ||g_2||=6.104e-05 acc=0.7920 + [BP-vit] Ep 60: ||h_L_cls||=1.197e+01 ||g_2||=5.866e-05 acc=0.7910 + +=== DFA training (ViT-Mini, block-level DFA) === + [DFA-vit] Ep 0: ||h_L_cls||=6.653e+00 ||g_2||=1.039e-03 acc=0.0791 + [DFA-vit] Ep 1: ||h_L_cls||=6.750e+03 ||g_2||=9.114e-07 acc=0.2334 γ=0.0073 + [DFA-vit] Ep 5: ||h_L_cls||=2.891e+05 ||g_2||=3.080e-08 acc=0.1963 γ=0.0068 + [DFA-vit] Ep 10: ||h_L_cls||=1.709e+06 ||g_2||=4.513e-09 acc=0.1973 γ=0.0061 + [DFA-vit] Ep 15: ||h_L_cls||=5.106e+06 ||g_2||=1.561e-09 acc=0.2363 γ=0.0023 + [DFA-vit] Ep 20: ||h_L_cls||=1.160e+07 ||g_2||=6.526e-10 acc=0.2559 γ=0.0012 + [DFA-vit] Ep 25: ||h_L_cls||=2.239e+07 ||g_2||=4.283e-10 acc=0.2568 γ=0.0006 + [DFA-vit] Ep 30: ||h_L_cls||=3.290e+07 ||g_2||=3.264e-10 acc=0.2656 γ=0.0005 + [DFA-vit] Ep 35: ||h_L_cls||=4.443e+07 ||g_2||=3.018e-10 acc=0.2354 γ=0.0008 + [DFA-vit] Ep 40: ||h_L_cls||=5.315e+07 ||g_2||=2.841e-10 acc=0.2559 γ=0.0004 + [DFA-vit] Ep 45: ||h_L_cls||=5.912e+07 ||g_2||=3.038e-10 acc=0.2441 γ=0.0004 + [DFA-vit] Ep 50: ||h_L_cls||=6.210e+07 ||g_2||=3.006e-10 acc=0.2578 γ=0.0002 + [DFA-vit] Ep 55: ||h_L_cls||=6.344e+07 ||g_2||=3.079e-10 acc=0.2529 γ=0.0001 + [DFA-vit] Ep 60: ||h_L_cls||=6.367e+07 ||g_2||=3.069e-10 acc=0.2529 γ=0.0000 + +Saved results/snapshot_vit_v1/snapshot_vit_s456.json diff --git a/results/snapshot_vit_v1/snapshot_fa_canonical_s42.json b/results/snapshot_vit_v1/snapshot_fa_canonical_s42.json new file mode 100644 index 0000000..87675ca --- /dev/null +++ b/results/snapshot_vit_v1/snapshot_fa_canonical_s42.json @@ -0,0 +1,1104 @@ +{ + "fa_log": [ + { + "hidden_norms_cls": [ + 0.31052109599113464, + 3.2536301612854004, + 4.7681732177734375, + 6.129599571228027, + 7.124356746673584 + ], + "bp_grad_per_sample_l2_med": [ + 0.006941193714737892, + 0.001723448047414422, + 0.0008084969012998044, + 0.00040042176260612905, + 0.00019880379841197282 + ], + "acc_eval": 0.123046875, + "epoch": 0 + }, + { + "hidden_norms_cls": [ + 0.5545730590820312, + 1099.6082763671875, + 1397.880126953125, + 1557.6058349609375, + 1625.5528564453125 + ], + "bp_grad_per_sample_l2_med": [ + 0.003943991847336292, + 4.105754669581074e-06, + 2.034018280028249e-06, + 1.0202892326560686e-06, + 5.029408498558041e-07 + ], + "acc_eval": 0.25, + "epoch": 1 + }, + { + "hidden_norms_cls": [ + 0.5561315417289734, + 1825.42822265625, + 3141.504638671875, + 5072.111328125, + 5524.6123046875 + ], + "bp_grad_per_sample_l2_med": [ + 0.006327113136649132, + 1.8617935211295844e-06, + 9.007959533846588e-07, + 4.491473930556822e-07, + 2.2173806257796969e-07 + ], + "acc_eval": 0.236328125, + "epoch": 2 + }, + { + "hidden_norms_cls": [ + 0.538809597492218, + 4488.783203125, + 6229.931640625, + 13038.5927734375, + 14604.98828125 + ], + "bp_grad_per_sample_l2_med": [ + 0.00814820732921362, + 7.318813572965155e-07, + 3.619582003011601e-07, + 1.8043300542558427e-07, + 9.003725409684193e-08 + ], + "acc_eval": 0.24609375, + "epoch": 3 + }, + { + "hidden_norms_cls": [ + 0.5289841294288635, + 11559.1318359375, + 21324.697265625, + 33157.46484375, + 34513.00390625 + ], + "bp_grad_per_sample_l2_med": [ + 0.004759137984365225, + 2.922441524333408e-07, + 1.4508636070331704e-07, + 7.107593802402334e-08, + 3.5740285397878324e-08 + ], + "acc_eval": 0.1953125, + "epoch": 4 + }, + { + "hidden_norms_cls": [ + 0.5187227725982666, + 13567.609375, + 31900.357421875, + 41906.99609375, + 44904.33984375 + ], + "bp_grad_per_sample_l2_med": [ + 0.008981491439044476, + 2.3057019404859602e-07, + 1.1323756154979492e-07, + 5.645526357511699e-08, + 2.834152468267348e-08 + ], + "acc_eval": 0.2373046875, + "epoch": 5 + }, + { + "hidden_norms_cls": [ + 0.5072101950645447, + 19264.873046875, + 40929.52734375, + 53545.33203125, + 56968.43359375 + ], + "bp_grad_per_sample_l2_med": [ + 0.011568969115614891, + 1.827240510010597e-07, + 9.02516177347934e-08, + 4.502405914763585e-08, + 2.257244524628277e-08 + ], + "acc_eval": 0.2333984375, + "epoch": 6 + }, + { + "hidden_norms_cls": [ + 0.49512261152267456, + 17510.943359375, + 48323.71875, + 50008.49609375, + 51787.2734375 + ], + "bp_grad_per_sample_l2_med": [ + 0.010693231597542763, + 1.6911585021262e-07, + 8.260560235839876e-08, + 4.064630942934855e-08, + 2.0301804681821523e-08 + ], + "acc_eval": 0.2158203125, + "epoch": 7 + }, + { + "hidden_norms_cls": [ + 0.48422321677207947, + 41544.78515625, + 71321.0625, + 150438.3125, + 155715.5625 + ], + "bp_grad_per_sample_l2_med": [ + 0.03921247273683548, + 1.1485686712831011e-07, + 5.667784108709384e-08, + 2.8099817583893127e-08, + 1.4131681602691515e-08 + ], + "acc_eval": 0.2109375, + "epoch": 8 + }, + { + "hidden_norms_cls": [ + 0.47471165657043457, + 61683.44921875, + 81806.375, + 149131.84375, + 154680.453125 + ], + "bp_grad_per_sample_l2_med": [ + 0.0444764718413353, + 8.208751012261928e-08, + 4.092110117426273e-08, + 2.041711866240803e-08, + 1.0249735282741312e-08 + ], + "acc_eval": 0.24609375, + "epoch": 9 + }, + { + "hidden_norms_cls": [ + 0.4715467393398285, + 100863.953125, + 117647.9453125, + 198225.390625, + 210929.75 + ], + "bp_grad_per_sample_l2_med": [ + 0.018913062289357185, + 4.7114983914298136e-08, + 2.3594729725573416e-08, + 1.1833483526402233e-08, + 5.916069856226613e-09 + ], + "acc_eval": 0.2021484375, + "epoch": 10 + }, + { + "hidden_norms_cls": [ + 0.4661370813846588, + 155109.703125, + 179093.8125, + 259543.703125, + 261068.453125 + ], + "bp_grad_per_sample_l2_med": [ + 0.014048838056623936, + 2.8752721092928368e-08, + 1.4348634280736405e-08, + 7.019417935794081e-09, + 3.5147973420635026e-09 + ], + "acc_eval": 0.205078125, + "epoch": 11 + }, + { + "hidden_norms_cls": [ + 0.45670604705810547, + 139276.359375, + 179938.421875, + 286296.46875, + 287724.21875 + ], + "bp_grad_per_sample_l2_med": [ + 0.01842617429792881, + 3.2641882086181795e-08, + 1.6271862079975108e-08, + 8.13431366708528e-09, + 4.065131786745724e-09 + ], + "acc_eval": 0.19140625, + "epoch": 12 + }, + { + "hidden_norms_cls": [ + 0.4475501775741577, + 243322.703125, + 291619.53125, + 393543.1875, + 453741.9375 + ], + "bp_grad_per_sample_l2_med": [ + 0.017274944111704826, + 2.0982721338214105e-08, + 1.0508765413419496e-08, + 5.2403605899087324e-09, + 2.6231921079755693e-09 + ], + "acc_eval": 0.216796875, + "epoch": 13 + }, + { + "hidden_norms_cls": [ + 0.44052183628082275, + 378946.46875, + 454064.40625, + 486236.65625, + 545435.0625 + ], + "bp_grad_per_sample_l2_med": [ + 0.011560702696442604, + 1.302930296986915e-08, + 6.4620948592164495e-09, + 3.2205806910212686e-09, + 1.6108331335473736e-09 + ], + "acc_eval": 0.177734375, + "epoch": 14 + }, + { + "hidden_norms_cls": [ + 0.43271052837371826, + 603185.0, + 706685.875, + 754365.6875, + 781386.0625 + ], + "bp_grad_per_sample_l2_med": [ + 0.027027767151594162, + 8.183165256525626e-09, + 4.085152216504184e-09, + 2.0333199568511873e-09, + 1.0129016514426326e-09 + ], + "acc_eval": 0.1923828125, + "epoch": 15 + }, + { + "hidden_norms_cls": [ + 0.42688167095184326, + 897163.875, + 1069202.125, + 1126727.125, + 1146658.875 + ], + "bp_grad_per_sample_l2_med": [ + 0.013954665511846542, + 5.21436360756411e-09, + 2.6102418004825267e-09, + 1.3076516536258964e-09, + 6.521915230095487e-10 + ], + "acc_eval": 0.17578125, + "epoch": 16 + }, + { + "hidden_norms_cls": [ + 0.419918030500412, + 1210681.5, + 1492631.25, + 1582945.75, + 1604470.375 + ], + "bp_grad_per_sample_l2_med": [ + 0.016081038862466812, + 3.6842373596357447e-09, + 1.8416679292698745e-09, + 9.214698448722913e-10, + 4.6121051422431947e-10 + ], + "acc_eval": 0.205078125, + "epoch": 17 + }, + { + "hidden_norms_cls": [ + 0.41367214918136597, + 1474524.0, + 1896333.625, + 2048261.125, + 2075013.625 + ], + "bp_grad_per_sample_l2_med": [ + 0.015561909414827824, + 2.9148632396669427e-09, + 1.4711201146155872e-09, + 7.326692585962746e-10, + 3.6670635972235743e-10 + ], + "acc_eval": 0.1748046875, + "epoch": 18 + }, + { + "hidden_norms_cls": [ + 0.4075075685977936, + 1721647.75, + 2392353.5, + 2664490.25, + 2676219.75 + ], + "bp_grad_per_sample_l2_med": [ + 0.013573882170021534, + 2.4936914755357975e-09, + 1.2590603004625223e-09, + 6.301172361666829e-10, + 3.1474217676574767e-10 + ], + "acc_eval": 0.171875, + "epoch": 19 + }, + { + "hidden_norms_cls": [ + 0.40185868740081787, + 2061658.75, + 3070974.0, + 3363148.5, + 3355996.5 + ], + "bp_grad_per_sample_l2_med": [ + 0.01332809403538704, + 1.9996668765287495e-09, + 9.928408095660757e-10, + 4.968849776076922e-10, + 2.4808827214783946e-10 + ], + "acc_eval": 0.1728515625, + "epoch": 20 + }, + { + "hidden_norms_cls": [ + 0.3959394693374634, + 2567158.5, + 4022831.25, + 4366507.5, + 4360674.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.012877865694463253, + 1.5711227874248834e-09, + 7.865745832447146e-10, + 3.9232733728411517e-10, + 1.9626268665806634e-10 + ], + "acc_eval": 0.1689453125, + "epoch": 21 + }, + { + "hidden_norms_cls": [ + 0.3908536732196808, + 2855278.0, + 4821230.0, + 5318397.5, + 5261785.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.01991885155439377, + 1.3769393403251229e-09, + 6.908043581610457e-10, + 3.4423255912408024e-10, + 1.725469211955044e-10 + ], + "acc_eval": 0.1748046875, + "epoch": 22 + }, + { + "hidden_norms_cls": [ + 0.38593825697898865, + 3174507.75, + 5821231.0, + 6397560.5, + 6354021.5 + ], + "bp_grad_per_sample_l2_med": [ + 0.011731060221791267, + 1.1907397290400468e-09, + 6.002150998440925e-10, + 2.995812764527983e-10, + 1.4982258489393274e-10 + ], + "acc_eval": 0.1953125, + "epoch": 23 + }, + { + "hidden_norms_cls": [ + 0.38131335377693176, + 4292486.5, + 7334872.5, + 8481275.0, + 7987194.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.015600440092384815, + 9.097935183000061e-10, + 4.627094263298659e-10, + 2.3154314576778745e-10, + 1.1568761226765645e-10 + ], + "acc_eval": 0.1650390625, + "epoch": 24 + }, + { + "hidden_norms_cls": [ + 0.3768307864665985, + 4066144.5, + 8089341.0, + 8840942.0, + 8750187.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.01041770912706852, + 9.491230024138986e-10, + 4.764759142794617e-10, + 2.383013231188613e-10, + 1.1919720765973807e-10 + ], + "acc_eval": 0.158203125, + "epoch": 25 + }, + { + "hidden_norms_cls": [ + 0.3725484013557434, + 4660335.5, + 9398058.0, + 10277586.0, + 10159446.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.012610528618097305, + 8.221311964540234e-10, + 4.1212994128514424e-10, + 2.0628378172293793e-10, + 1.0312824205715998e-10 + ], + "acc_eval": 0.162109375, + "epoch": 26 + }, + { + "hidden_norms_cls": [ + 0.3689132332801819, + 5871612.5, + 11057573.0, + 12074060.0, + 11979049.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.022938355803489685, + 6.959858245281225e-10, + 3.601695330868182e-10, + 1.795542742266676e-10, + 9.003973955312716e-11 + ], + "acc_eval": 0.1748046875, + "epoch": 27 + }, + { + "hidden_norms_cls": [ + 0.3653584122657776, + 6549866.5, + 12457973.0, + 13667217.0, + 13488831.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.01968899369239807, + 5.967032423725982e-10, + 3.0872890355304605e-10, + 1.5443717138463597e-10, + 7.714055783036855e-11 + ], + "acc_eval": 0.1669921875, + "epoch": 28 + }, + { + "hidden_norms_cls": [ + 0.36252257227897644, + 7083078.0, + 13728239.0, + 15048567.0, + 14800666.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.024753496050834656, + 5.776352174358124e-10, + 2.965032941393275e-10, + 1.4841036732882174e-10, + 7.41542729998379e-11 + ], + "acc_eval": 0.1728515625, + "epoch": 29 + }, + { + "hidden_norms_cls": [ + 0.3597593307495117, + 7156219.5, + 14613827.0, + 16177184.0, + 15867089.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.012417457066476345, + 5.718481799199537e-10, + 2.867320547661478e-10, + 1.4365150735606846e-10, + 7.184561279238721e-11 + ], + "acc_eval": 0.1650390625, + "epoch": 30 + }, + { + "hidden_norms_cls": [ + 0.35805046558380127, + 8087894.5, + 16279937.0, + 18019618.0, + 17642490.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.017159853130578995, + 5.130134095310268e-10, + 2.612176586147541e-10, + 1.3114326291585598e-10, + 6.563268678538847e-11 + ], + "acc_eval": 0.1796875, + "epoch": 31 + }, + { + "hidden_norms_cls": [ + 0.35514453053474426, + 8415584.0, + 17452890.0, + 19574252.0, + 19044988.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.017893649637699127, + 5.072835485009364e-10, + 2.5544158455126365e-10, + 1.2795793591369176e-10, + 6.410054431693624e-11 + ], + "acc_eval": 0.1962890625, + "epoch": 32 + }, + { + "hidden_norms_cls": [ + 0.35366830229759216, + 9040861.0, + 18928378.0, + 21235278.0, + 20689608.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.01662450097501278, + 4.760899452449507e-10, + 2.391463138629035e-10, + 1.19767168405005e-10, + 5.980976131025884e-11 + ], + "acc_eval": 0.177734375, + "epoch": 33 + }, + { + "hidden_norms_cls": [ + 0.35274940729141235, + 10105909.0, + 20575994.0, + 22869280.0, + 22381648.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.04045082628726959, + 4.454907553963494e-10, + 2.224223166757966e-10, + 1.1117023274165305e-10, + 5.573364339483966e-11 + ], + "acc_eval": 0.173828125, + "epoch": 34 + }, + { + "hidden_norms_cls": [ + 0.35230201482772827, + 10649243.0, + 21849520.0, + 24383386.0, + 23809486.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.03625870868563652, + 4.062269409743635e-10, + 2.0919388443729758e-10, + 1.0517077486671411e-10, + 5.271013833740845e-11 + ], + "acc_eval": 0.158203125, + "epoch": 35 + }, + { + "hidden_norms_cls": [ + 0.35188645124435425, + 11225758.0, + 23225336.0, + 25744754.0, + 25098318.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.04868545010685921, + 4.2478476292018286e-10, + 2.1696269780768773e-10, + 1.0849423442982342e-10, + 5.422649829167625e-11 + ], + "acc_eval": 0.181640625, + "epoch": 36 + }, + { + "hidden_norms_cls": [ + 0.35190674662590027, + 11552205.0, + 24186280.0, + 27053836.0, + 26295722.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.04053955897688866, + 4.0922154553868495e-10, + 2.0564389080490741e-10, + 1.0180161574835367e-10, + 5.100607436414606e-11 + ], + "acc_eval": 0.1708984375, + "epoch": 37 + }, + { + "hidden_norms_cls": [ + 0.3533709645271301, + 12511109.0, + 25674238.0, + 28429464.0, + 27729346.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.04322395101189613, + 3.725502961682281e-10, + 1.895077428315517e-10, + 9.475340650988429e-11, + 4.7370347228126164e-11 + ], + "acc_eval": 0.1943359375, + "epoch": 38 + }, + { + "hidden_norms_cls": [ + 0.35590660572052, + 12274717.0, + 26252202.0, + 29145476.0, + 28407090.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0458325557410717, + 3.9338646229403196e-10, + 1.9832839537325953e-10, + 9.937214245914205e-11, + 4.97785458686284e-11 + ], + "acc_eval": 0.162109375, + "epoch": 39 + }, + { + "hidden_norms_cls": [ + 0.35598739981651306, + 13010948.0, + 27253844.0, + 30244608.0, + 29436354.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.050111040472984314, + 3.8191338980198e-10, + 1.9492148173316792e-10, + 9.78909453497323e-11, + 4.898272759623623e-11 + ], + "acc_eval": 0.17578125, + "epoch": 40 + }, + { + "hidden_norms_cls": [ + 0.359886109828949, + 12949905.0, + 27785034.0, + 30987020.0, + 30162430.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.04287939891219139, + 3.569295692340546e-10, + 1.7953295794459478e-10, + 8.970056641910418e-11, + 4.4947719157750754e-11 + ], + "acc_eval": 0.1669921875, + "epoch": 41 + }, + { + "hidden_norms_cls": [ + 0.3610597252845764, + 13297263.0, + 28490094.0, + 31717372.0, + 30903264.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.059325557202100754, + 3.555903349550249e-10, + 1.8439673399317513e-10, + 9.228186964582719e-11, + 4.604394990281868e-11 + ], + "acc_eval": 0.1630859375, + "epoch": 42 + }, + { + "hidden_norms_cls": [ + 0.36371150612831116, + 13593509.0, + 29189264.0, + 32363910.0, + 31583456.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.08067209273576736, + 3.748426291583229e-10, + 1.9037708909319662e-10, + 9.613144308362465e-11, + 4.7860448243453035e-11 + ], + "acc_eval": 0.1669921875, + "epoch": 43 + }, + { + "hidden_norms_cls": [ + 0.3678953945636749, + 13002685.0, + 29272840.0, + 32833962.0, + 31996856.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.046668969094753265, + 3.5862279812448605e-10, + 1.8444731852973462e-10, + 9.241421516925641e-11, + 4.614601409325125e-11 + ], + "acc_eval": 0.162109375, + "epoch": 44 + }, + { + "hidden_norms_cls": [ + 0.3722097873687744, + 13026701.0, + 29583132.0, + 33140148.0, + 32223962.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.05778568983078003, + 3.8914271804912914e-10, + 1.9742661672150774e-10, + 9.861946675959743e-11, + 4.930081343168524e-11 + ], + "acc_eval": 0.1640625, + "epoch": 45 + }, + { + "hidden_norms_cls": [ + 0.3784724175930023, + 13060913.0, + 29952154.0, + 33663012.0, + 32716300.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.058499205857515335, + 3.8859623852083303e-10, + 1.9628629277512744e-10, + 9.875016082627752e-11, + 4.9333759299940994e-11 + ], + "acc_eval": 0.16015625, + "epoch": 46 + }, + { + "hidden_norms_cls": [ + 0.3851682245731354, + 13061307.0, + 30283974.0, + 33786364.0, + 32963536.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.05969119817018509, + 4.036749268188089e-10, + 2.051936537350585e-10, + 1.0281481221952049e-10, + 5.1408790419094075e-11 + ], + "acc_eval": 0.1591796875, + "epoch": 47 + }, + { + "hidden_norms_cls": [ + 0.38858678936958313, + 13053480.0, + 30258368.0, + 34267304.0, + 33358036.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.05216536298394203, + 3.8322026107984186e-10, + 1.942517396935628e-10, + 9.742346512853217e-11, + 4.861935506972337e-11 + ], + "acc_eval": 0.1630859375, + "epoch": 48 + }, + { + "hidden_norms_cls": [ + 0.3970814645290375, + 12785361.0, + 30453750.0, + 34537440.0, + 33524112.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0512184239923954, + 3.618580712849706e-10, + 1.8165123572000397e-10, + 9.080205337630431e-11, + 4.5350091737450526e-11 + ], + "acc_eval": 0.1611328125, + "epoch": 49 + }, + { + "hidden_norms_cls": [ + 0.41041070222854614, + 12378014.0, + 30515276.0, + 34276956.0, + 33306668.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.04064331203699112, + 4.035453082806839e-10, + 2.059644121921167e-10, + 1.030020096992601e-10, + 5.144335651907639e-11 + ], + "acc_eval": 0.1591796875, + "epoch": 50 + }, + { + "hidden_norms_cls": [ + 0.4208085834980011, + 12229696.0, + 30525590.0, + 34345932.0, + 33357842.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.04530888423323631, + 4.160819744303268e-10, + 2.0900184360961305e-10, + 1.0429850733073565e-10, + 5.2114305926220794e-11 + ], + "acc_eval": 0.166015625, + "epoch": 51 + }, + { + "hidden_norms_cls": [ + 0.4249289929866791, + 12130639.0, + 30354396.0, + 34678540.0, + 33531180.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.04047567769885063, + 4.068767545106766e-10, + 2.0768058106579446e-10, + 1.0380588749692166e-10, + 5.188600937788834e-11 + ], + "acc_eval": 0.162109375, + "epoch": 52 + }, + { + "hidden_norms_cls": [ + 0.4318743348121643, + 11880541.0, + 30505546.0, + 34958524.0, + 33653512.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.03571970760822296, + 4.0090192277020265e-10, + 2.011875527285767e-10, + 1.0030835884133893e-10, + 5.012859571884576e-11 + ], + "acc_eval": 0.1630859375, + "epoch": 53 + }, + { + "hidden_norms_cls": [ + 0.43913060426712036, + 11708532.0, + 30355660.0, + 34920048.0, + 33497518.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0396106131374836, + 4.1890893531792983e-10, + 2.084224598464246e-10, + 1.0319496646093995e-10, + 5.16515094584058e-11 + ], + "acc_eval": 0.1611328125, + "epoch": 54 + }, + { + "hidden_norms_cls": [ + 0.44122788310050964, + 11552111.0, + 30335972.0, + 35070540.0, + 33779876.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.03454306349158287, + 3.9193837064743775e-10, + 1.9659923689019365e-10, + 9.868480338459662e-11, + 4.934370273490529e-11 + ], + "acc_eval": 0.1552734375, + "epoch": 55 + }, + { + "hidden_norms_cls": [ + 0.4431604743003845, + 11393902.0, + 30399024.0, + 35148924.0, + 33730668.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.03607407212257385, + 4.01606164990298e-10, + 2.01549305023363e-10, + 1.0008825018781309e-10, + 5.001985631247763e-11 + ], + "acc_eval": 0.1611328125, + "epoch": 56 + }, + { + "hidden_norms_cls": [ + 0.44458895921707153, + 11456393.0, + 30366940.0, + 35126152.0, + 33815292.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.03673577681183815, + 3.9196515477790683e-10, + 1.9733305267610746e-10, + 9.858801275353102e-11, + 4.931045849421167e-11 + ], + "acc_eval": 0.16015625, + "epoch": 57 + }, + { + "hidden_norms_cls": [ + 0.44609925150871277, + 11285367.0, + 30391162.0, + 35039688.0, + 33859112.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.03655927628278732, + 3.98514693467078e-10, + 1.9965135100719067e-10, + 9.972099535016099e-11, + 4.983467111197015e-11 + ], + "acc_eval": 0.1630859375, + "epoch": 58 + }, + { + "hidden_norms_cls": [ + 0.4466340243816376, + 11309850.0, + 30354550.0, + 35045136.0, + 33748968.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.03679952770471573, + 4.077362614207658e-10, + 2.0270646272635418e-10, + 1.015226652745227e-10, + 5.069205818997169e-11 + ], + "acc_eval": 0.1630859375, + "epoch": 59 + }, + { + "hidden_norms_cls": [ + 0.4467651844024658, + 11222298.0, + 30325882.0, + 35035856.0, + 33763052.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.03729122504591942, + 4.0734565720512705e-10, + 2.0311738402334356e-10, + 1.015596079456671e-10, + 5.085605200849663e-11 + ], + "acc_eval": 0.1630859375, + "epoch": 60 + } + ], + "arch": "vit", + "seed": 42 +}
\ No newline at end of file diff --git a/results/snapshot_vit_v1/snapshot_fa_vit_s42.json b/results/snapshot_vit_v1/snapshot_fa_vit_s42.json new file mode 100644 index 0000000..450351d --- /dev/null +++ b/results/snapshot_vit_v1/snapshot_fa_vit_s42.json @@ -0,0 +1,1104 @@ +{ + "fa_log": [ + { + "hidden_norms_cls": [ + 0.31052109599113464, + 3.2536301612854004, + 4.7681732177734375, + 6.129599571228027, + 7.124356746673584 + ], + "bp_grad_per_sample_l2_med": [ + 0.006941193714737892, + 0.001723448047414422, + 0.0008084969012998044, + 0.00040042176260612905, + 0.00019880379841197282 + ], + "acc_eval": 0.123046875, + "epoch": 0 + }, + { + "hidden_norms_cls": [ + 1.6919748783111572, + 3842.78515625, + 9341.6923828125, + 9920.3046875, + 9136.099609375 + ], + "bp_grad_per_sample_l2_med": [ + 0.0017624060856178403, + 1.1595898286032025e-06, + 5.700691758647736e-07, + 2.8373631266731536e-07, + 1.4022531047430675e-07 + ], + "acc_eval": 0.1962890625, + "epoch": 1 + }, + { + "hidden_norms_cls": [ + 2.213810443878174, + 13173.2275390625, + 20675.927734375, + 21090.544921875, + 20256.06640625 + ], + "bp_grad_per_sample_l2_med": [ + 0.00036340864608064294, + 4.792477170667553e-07, + 2.2975714841777517e-07, + 1.1404723920804827e-07, + 5.58470958367252e-08 + ], + "acc_eval": 0.130859375, + "epoch": 2 + }, + { + "hidden_norms_cls": [ + 2.4401588439941406, + 27950.775390625, + 42258.1171875, + 42646.40625, + 40386.3671875 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006889981450513005, + 2.183849545644989e-07, + 1.0780762949025302e-07, + 5.330540986392407e-08, + 2.567051815560717e-08 + ], + "acc_eval": 0.1259765625, + "epoch": 3 + }, + { + "hidden_norms_cls": [ + 2.4965271949768066, + 35298.37890625, + 71921.4921875, + 73662.3828125, + 71502.1171875 + ], + "bp_grad_per_sample_l2_med": [ + 0.00035679180291481316, + 1.280830304040137e-07, + 5.992792750930676e-08, + 3.021023786686783e-08, + 1.4932888703356184e-08 + ], + "acc_eval": 0.1474609375, + "epoch": 4 + }, + { + "hidden_norms_cls": [ + 2.542451858520508, + 596.9271240234375, + 15138.552734375, + 20439.001953125, + 63235.2109375 + ], + "bp_grad_per_sample_l2_med": [ + 0.0007700499845668674, + 4.321308608723484e-07, + 1.9291219643946533e-07, + 9.676254819623864e-08, + 4.7916582701645893e-08 + ], + "acc_eval": 0.12890625, + "epoch": 5 + }, + { + "hidden_norms_cls": [ + 2.6066854000091553, + 172636.59375, + 290436.34375, + 356619.71875, + 363143.5 + ], + "bp_grad_per_sample_l2_med": [ + 0.000482373550767079, + 2.2872455929245916e-08, + 1.1479056816199318e-08, + 5.759448917785903e-09, + 2.8778499583381745e-09 + ], + "acc_eval": 0.1201171875, + "epoch": 6 + }, + { + "hidden_norms_cls": [ + 2.549955129623413, + 247359.046875, + 456810.46875, + 542983.1875, + 536014.5625 + ], + "bp_grad_per_sample_l2_med": [ + 0.0003394677478354424, + 1.5884587867276423e-08, + 8.104618309801026e-09, + 4.073737347454198e-09, + 2.0375214848655787e-09 + ], + "acc_eval": 0.125, + "epoch": 7 + }, + { + "hidden_norms_cls": [ + 2.4600400924682617, + 507.6882629394531, + 55412.9140625, + 58117.875, + 71533.109375 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005215128185227513, + 1.720908642255381e-07, + 7.895193476770146e-08, + 3.936664327852668e-08, + 1.950460415400812e-08 + ], + "acc_eval": 0.166015625, + "epoch": 8 + }, + { + "hidden_norms_cls": [ + 2.377673387527466, + 492.4173583984375, + 72010.8515625, + 80733.3828125, + 88499.03125 + ], + "bp_grad_per_sample_l2_med": [ + 0.000300612737191841, + 1.303562839893857e-07, + 5.4381189329433255e-08, + 2.712311264474465e-08, + 1.3531170850455965e-08 + ], + "acc_eval": 0.1953125, + "epoch": 9 + }, + { + "hidden_norms_cls": [ + 2.2829232215881348, + 70101.53125, + 273366.84375, + 380567.40625, + 436148.25 + ], + "bp_grad_per_sample_l2_med": [ + 0.0003461186133790761, + 2.5325679331444917e-08, + 1.3804578813392254e-08, + 7.191788053972914e-09, + 3.5467115910847724e-09 + ], + "acc_eval": 0.1982421875, + "epoch": 10 + }, + { + "hidden_norms_cls": [ + 2.2401251792907715, + 95090.734375, + 434796.5, + 535611.6875, + 555035.75 + ], + "bp_grad_per_sample_l2_med": [ + 0.0003575912269297987, + 2.0430407587923582e-08, + 1.013873163202561e-08, + 5.240056832889195e-09, + 2.62253840865867e-09 + ], + "acc_eval": 0.19140625, + "epoch": 11 + }, + { + "hidden_norms_cls": [ + 2.213266134262085, + 371645.75, + 623563.875, + 694904.125, + 745774.625 + ], + "bp_grad_per_sample_l2_med": [ + 0.0003485995694063604, + 1.2734847842921226e-08, + 6.490221693411513e-09, + 3.23543081215405e-09, + 1.6088937959679583e-09 + ], + "acc_eval": 0.185546875, + "epoch": 12 + }, + { + "hidden_norms_cls": [ + 2.3490853309631348, + 1521.5958251953125, + 117621.15625, + 153339.328125, + 234005.4375 + ], + "bp_grad_per_sample_l2_med": [ + 0.0001710709184408188, + 8.239747018023991e-08, + 2.854269354202188e-08, + 1.4295378214512766e-08, + 7.217904940404196e-09 + ], + "acc_eval": 0.181640625, + "epoch": 13 + }, + { + "hidden_norms_cls": [ + 2.568831205368042, + 461180.78125, + 1827924.375, + 2293298.0, + 2130660.5 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002101986901834607, + 7.077860963988769e-09, + 3.564110562237488e-09, + 1.7704799848416997e-09, + 8.853479060988434e-10 + ], + "acc_eval": 0.201171875, + "epoch": 14 + }, + { + "hidden_norms_cls": [ + 2.6825993061065674, + 2726.72314453125, + 97672.5703125, + 471438.28125, + 779993.8125 + ], + "bp_grad_per_sample_l2_med": [ + 9.469804354012012e-05, + 6.38537613895096e-08, + 7.950252012278725e-09, + 3.422149452703138e-09, + 1.7092232074134017e-09 + ], + "acc_eval": 0.201171875, + "epoch": 15 + }, + { + "hidden_norms_cls": [ + 2.7971737384796143, + 313802.15625, + 3608552.75, + 4391356.0, + 4212764.5 + ], + "bp_grad_per_sample_l2_med": [ + 0.00015007780166342854, + 3.938802173308886e-09, + 1.8008210478370756e-09, + 9.022776970013524e-10, + 4.517709262241709e-10 + ], + "acc_eval": 0.203125, + "epoch": 16 + }, + { + "hidden_norms_cls": [ + 2.8100268840789795, + 2875.376708984375, + 96749.3671875, + 983606.125, + 1150911.625 + ], + "bp_grad_per_sample_l2_med": [ + 0.001228083623573184, + 6.225434390216833e-08, + 7.879610741667875e-09, + 3.8225227427801656e-09, + 1.923271097936663e-09 + ], + "acc_eval": 0.1787109375, + "epoch": 17 + }, + { + "hidden_norms_cls": [ + 2.8279852867126465, + 2183.419677734375, + 65646.7734375, + 320722.65625, + 442094.75 + ], + "bp_grad_per_sample_l2_med": [ + 0.00120502186473459, + 8.017538988269735e-08, + 2.5245032730936146e-08, + 1.1776578823230466e-08, + 5.894425836316941e-09 + ], + "acc_eval": 0.1044921875, + "epoch": 18 + }, + { + "hidden_norms_cls": [ + 2.8702030181884766, + 2139.65234375, + 95514.8359375, + 324949.6875, + 462647.84375 + ], + "bp_grad_per_sample_l2_med": [ + 0.0014194594696164131, + 5.5903008444602165e-08, + 2.0797312316744865e-08, + 9.60299928465247e-09, + 4.690376087523873e-09 + ], + "acc_eval": 0.1103515625, + "epoch": 19 + }, + { + "hidden_norms_cls": [ + 2.7991764545440674, + 15302.53125, + 190705.796875, + 494117.90625, + 543885.625 + ], + "bp_grad_per_sample_l2_med": [ + 0.00047307764180004597, + 2.4012980048837562e-08, + 1.1153974632804875e-08, + 5.289668258967595e-09, + 2.6598685476386663e-09 + ], + "acc_eval": 0.0888671875, + "epoch": 20 + }, + { + "hidden_norms_cls": [ + 2.8112618923187256, + 534165.625, + 9562196.0, + 11356111.0, + 11261478.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002835187187884003, + 1.1553467071934165e-09, + 5.117212764638168e-10, + 2.5636526235217616e-10, + 1.2813658467614175e-10 + ], + "acc_eval": 0.107421875, + "epoch": 21 + }, + { + "hidden_norms_cls": [ + 2.7986021041870117, + 1302347.875, + 12959677.0, + 15286166.0, + 15075608.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.000172777334228158, + 7.537985791117308e-10, + 3.5877870119271904e-10, + 1.7971832355634376e-10, + 8.955045038838705e-11 + ], + "acc_eval": 0.126953125, + "epoch": 22 + }, + { + "hidden_norms_cls": [ + 2.756831169128418, + 422432.40625, + 12218335.0, + 15267431.0, + 14942667.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00015324499690905213, + 1.1144170031229805e-09, + 3.578760898736988e-10, + 1.796912063589673e-10, + 8.906290982491072e-11 + ], + "acc_eval": 0.166015625, + "epoch": 23 + }, + { + "hidden_norms_cls": [ + 2.713092088699341, + 441085.75, + 13975295.0, + 17594034.0, + 17150196.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0001009844709187746, + 9.29383403569517e-10, + 2.9464700124215426e-10, + 1.481769429378943e-10, + 7.335923535300992e-11 + ], + "acc_eval": 0.1826171875, + "epoch": 24 + }, + { + "hidden_norms_cls": [ + 2.677187204360962, + 329657.125, + 11149057.0, + 16110063.0, + 15466966.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00021013693185523152, + 1.6009201742051005e-09, + 3.1313171500180204e-10, + 1.5797228802849617e-10, + 7.763964471330098e-11 + ], + "acc_eval": 0.1796875, + "epoch": 25 + }, + { + "hidden_norms_cls": [ + 2.6366331577301025, + 5142.62060546875, + 367599.71875, + 988076.75, + 1002302.5 + ], + "bp_grad_per_sample_l2_med": [ + 0.00011194300896022469, + 9.242073772952608e-09, + 4.35089919648135e-09, + 2.032314094790877e-09, + 1.016630779560046e-09 + ], + "acc_eval": 0.181640625, + "epoch": 26 + }, + { + "hidden_norms_cls": [ + 2.599846601486206, + 5072.9931640625, + 354873.9375, + 1104508.0, + 1055959.75 + ], + "bp_grad_per_sample_l2_med": [ + 0.00014896050561219454, + 1.0939302796941774e-08, + 4.676385056967547e-09, + 2.02762540091328e-09, + 1.0192051647095468e-09 + ], + "acc_eval": 0.1806640625, + "epoch": 27 + }, + { + "hidden_norms_cls": [ + 2.598066568374634, + 116168.1875, + 10078028.0, + 18893916.0, + 17647230.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0001877687027445063, + 2.095827511539028e-09, + 2.405041166220201e-10, + 1.2486357781060775e-10, + 6.205166935835393e-11 + ], + "acc_eval": 0.1962890625, + "epoch": 28 + }, + { + "hidden_norms_cls": [ + 2.6015524864196777, + 740986.875, + 24883748.0, + 31021810.0, + 30329238.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00018684851238504052, + 3.8257660928131543e-10, + 1.594476078947693e-10, + 8.13158151924398e-11, + 4.050663623966777e-11 + ], + "acc_eval": 0.1845703125, + "epoch": 29 + }, + { + "hidden_norms_cls": [ + 2.589724063873291, + 2141793.5, + 28624410.0, + 34863924.0, + 34267120.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002886186121031642, + 3.2890931644935506e-10, + 1.5246036377813965e-10, + 7.833261816969639e-11, + 3.898681030789497e-11 + ], + "acc_eval": 0.1748046875, + "epoch": 30 + }, + { + "hidden_norms_cls": [ + 2.5866215229034424, + 2836275.25, + 30209988.0, + 36742252.0, + 36121420.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0004315464466344565, + 3.6725425478500995e-10, + 1.698625962109901e-10, + 8.663138262310355e-11, + 4.2701918390175564e-11 + ], + "acc_eval": 0.1787109375, + "epoch": 31 + }, + { + "hidden_norms_cls": [ + 2.581477642059326, + 3276328.25, + 31851982.0, + 38643076.0, + 37988372.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006185675738379359, + 3.898423250880967e-10, + 1.803390908827751e-10, + 9.224612046443426e-11, + 4.599535682880962e-11 + ], + "acc_eval": 0.181640625, + "epoch": 32 + }, + { + "hidden_norms_cls": [ + 2.6230173110961914, + 3459700.75, + 33572172.0, + 40661440.0, + 39924348.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0009694760083220899, + 3.8904643395731853e-10, + 1.8190765560532896e-10, + 9.24886070508002e-11, + 4.616174456573141e-11 + ], + "acc_eval": 0.189453125, + "epoch": 33 + }, + { + "hidden_norms_cls": [ + 2.9775710105895996, + 10399.8671875, + 571599.6875, + 1836598.625, + 1908883.5 + ], + "bp_grad_per_sample_l2_med": [ + 0.00016216874064411968, + 7.749568453618849e-08, + 8.983486843305855e-09, + 2.0125863198217075e-09, + 1.0150218443527592e-09 + ], + "acc_eval": 0.0986328125, + "epoch": 34 + }, + { + "hidden_norms_cls": [ + 3.015839099884033, + 12101.009765625, + 561385.0625, + 1955154.125, + 1975725.375 + ], + "bp_grad_per_sample_l2_med": [ + 0.00012338445230852813, + 6.92784638545163e-08, + 9.938768030792744e-09, + 1.8663843803778946e-09, + 9.43250033635934e-10 + ], + "acc_eval": 0.103515625, + "epoch": 35 + }, + { + "hidden_norms_cls": [ + 3.002318859100342, + 13969.6484375, + 549761.875, + 2067254.5, + 2046448.25 + ], + "bp_grad_per_sample_l2_med": [ + 0.00011628194624790922, + 5.204710973316651e-08, + 8.372143867063642e-09, + 1.751683464945586e-09, + 8.823823893777671e-10 + ], + "acc_eval": 0.1025390625, + "epoch": 36 + }, + { + "hidden_norms_cls": [ + 3.033550262451172, + 16014.7802734375, + 542334.375, + 2185907.75, + 2132718.0 + ], + "bp_grad_per_sample_l2_med": [ + 7.604974234709516e-05, + 4.623546701054693e-08, + 9.098562792075882e-09, + 1.7713268629648837e-09, + 8.977130705467573e-10 + ], + "acc_eval": 0.1025390625, + "epoch": 37 + }, + { + "hidden_norms_cls": [ + 2.9693667888641357, + 17655.8203125, + 542724.0, + 2311982.5, + 2234848.5 + ], + "bp_grad_per_sample_l2_med": [ + 0.0001330485101789236, + 3.560405303915104e-08, + 7.560966963637838e-09, + 1.5777356088264582e-09, + 7.958586012435376e-10 + ], + "acc_eval": 0.0966796875, + "epoch": 38 + }, + { + "hidden_norms_cls": [ + 2.9733641147613525, + 18918.01171875, + 540951.6875, + 2434576.5, + 2336018.25 + ], + "bp_grad_per_sample_l2_med": [ + 9.626700193621218e-05, + 3.403155446335404e-08, + 8.177446275681177e-09, + 1.5839350941959651e-09, + 8.022257858009141e-10 + ], + "acc_eval": 0.1201171875, + "epoch": 39 + }, + { + "hidden_norms_cls": [ + 3.0027060508728027, + 20108.37890625, + 544636.5, + 2550452.75, + 2438166.75 + ], + "bp_grad_per_sample_l2_med": [ + 7.119941437849775e-05, + 3.794270853063608e-08, + 9.455917826528548e-09, + 1.7259683682269156e-09, + 8.75004957379133e-10 + ], + "acc_eval": 0.1083984375, + "epoch": 40 + }, + { + "hidden_norms_cls": [ + 2.9846291542053223, + 21615.2578125, + 547112.625, + 2664214.25, + 2538743.25 + ], + "bp_grad_per_sample_l2_med": [ + 9.802342538023368e-05, + 3.710788476496418e-08, + 9.733276407075664e-09, + 1.7067438573548088e-09, + 8.65347682399431e-10 + ], + "acc_eval": 0.1083984375, + "epoch": 41 + }, + { + "hidden_norms_cls": [ + 3.009104013442993, + 22886.44921875, + 554807.0625, + 2772577.5, + 2636159.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0001160762258223258, + 3.776746737571557e-08, + 9.48633260833276e-09, + 1.7762409321164796e-09, + 8.999931910835812e-10 + ], + "acc_eval": 0.107421875, + "epoch": 42 + }, + { + "hidden_norms_cls": [ + 3.0807487964630127, + 24102.28515625, + 560303.4375, + 2871532.75, + 2723516.0 + ], + "bp_grad_per_sample_l2_med": [ + 6.469501386163756e-05, + 4.052726865211298e-08, + 6.607923985768593e-09, + 1.986373288076493e-09, + 1.005414418386863e-09 + ], + "acc_eval": 0.111328125, + "epoch": 43 + }, + { + "hidden_norms_cls": [ + 3.0981693267822266, + 25397.25390625, + 569329.25, + 2967934.75, + 2808716.5 + ], + "bp_grad_per_sample_l2_med": [ + 8.018066000659019e-05, + 3.965687156437525e-08, + 8.188467681691236e-09, + 1.9840931120285177e-09, + 1.0038238018594825e-09 + ], + "acc_eval": 0.1064453125, + "epoch": 44 + }, + { + "hidden_norms_cls": [ + 3.1128475666046143, + 26593.50390625, + 578056.75, + 3059475.5, + 2888212.0 + ], + "bp_grad_per_sample_l2_med": [ + 8.603822061559185e-05, + 3.689373073711977e-08, + 6.842784561200688e-09, + 1.961748541390307e-09, + 9.92456117288043e-10 + ], + "acc_eval": 0.1064453125, + "epoch": 45 + }, + { + "hidden_norms_cls": [ + 3.126692771911621, + 27752.5, + 586243.125, + 3147350.0, + 2962444.5 + ], + "bp_grad_per_sample_l2_med": [ + 7.914086017990485e-05, + 3.278608318169063e-08, + 4.82215689601162e-09, + 1.8859560579898016e-09, + 9.505569664725044e-10 + ], + "acc_eval": 0.107421875, + "epoch": 46 + }, + { + "hidden_norms_cls": [ + 3.1628687381744385, + 28661.537109375, + 592898.875, + 3223267.75, + 3025087.0 + ], + "bp_grad_per_sample_l2_med": [ + 7.642550917807966e-05, + 3.200472065145732e-08, + 5.030188265919833e-09, + 1.7837261667708049e-09, + 9.011961177307626e-10 + ], + "acc_eval": 0.111328125, + "epoch": 47 + }, + { + "hidden_norms_cls": [ + 3.1804893016815186, + 29497.119140625, + 601081.125, + 3294916.75, + 3083660.25 + ], + "bp_grad_per_sample_l2_med": [ + 8.01319838501513e-05, + 3.0802997486034656e-08, + 4.8026316257221424e-09, + 1.7661813123126535e-09, + 8.969202602848725e-10 + ], + "acc_eval": 0.111328125, + "epoch": 48 + }, + { + "hidden_norms_cls": [ + 3.201361894607544, + 30248.82421875, + 612577.625, + 3357942.0, + 3135999.25 + ], + "bp_grad_per_sample_l2_med": [ + 8.080735278781503e-05, + 3.106945101194469e-08, + 4.879642911959081e-09, + 1.7795440676593444e-09, + 9.0354357329403e-10 + ], + "acc_eval": 0.1103515625, + "epoch": 49 + }, + { + "hidden_norms_cls": [ + 3.2230916023254395, + 31060.8359375, + 621329.6875, + 3412872.75, + 3180218.0 + ], + "bp_grad_per_sample_l2_med": [ + 7.493209704989567e-05, + 3.203697929166083e-08, + 5.23388132833702e-09, + 1.7845065425348139e-09, + 9.061755235073576e-10 + ], + "acc_eval": 0.111328125, + "epoch": 50 + }, + { + "hidden_norms_cls": [ + 3.2431774139404297, + 31627.490234375, + 631077.4375, + 3459407.75, + 3218656.25 + ], + "bp_grad_per_sample_l2_med": [ + 6.88562577124685e-05, + 3.010745786014013e-08, + 4.799196151594742e-09, + 1.7849065558905863e-09, + 9.058673811068729e-10 + ], + "acc_eval": 0.111328125, + "epoch": 51 + }, + { + "hidden_norms_cls": [ + 3.250222682952881, + 32081.44921875, + 639738.5, + 3499162.5, + 3250836.0 + ], + "bp_grad_per_sample_l2_med": [ + 6.744447455275804e-05, + 2.8871626867044142e-08, + 4.598368352759508e-09, + 1.781051417459878e-09, + 9.045134086171913e-10 + ], + "acc_eval": 0.11328125, + "epoch": 52 + }, + { + "hidden_norms_cls": [ + 3.258018970489502, + 32488.755859375, + 647751.75, + 3532386.25, + 3277588.75 + ], + "bp_grad_per_sample_l2_med": [ + 7.57749512558803e-05, + 3.006854853992991e-08, + 4.75430272928179e-09, + 1.7687115105857742e-09, + 8.985628907609566e-10 + ], + "acc_eval": 0.1162109375, + "epoch": 53 + }, + { + "hidden_norms_cls": [ + 3.268540143966675, + 32846.26171875, + 653093.8125, + 3557156.75, + 3297180.25 + ], + "bp_grad_per_sample_l2_med": [ + 6.825588206993416e-05, + 3.074128329672021e-08, + 4.992171565021408e-09, + 1.775832703110325e-09, + 9.020287849992314e-10 + ], + "acc_eval": 0.1162109375, + "epoch": 54 + }, + { + "hidden_norms_cls": [ + 3.277456760406494, + 33144.296875, + 656811.625, + 3575374.5, + 3311610.25 + ], + "bp_grad_per_sample_l2_med": [ + 6.744571874151006e-05, + 3.093562384037796e-08, + 5.061019603402883e-09, + 1.7779399064110635e-09, + 9.031043690654883e-10 + ], + "acc_eval": 0.1181640625, + "epoch": 55 + }, + { + "hidden_norms_cls": [ + 3.2813076972961426, + 33291.59765625, + 660069.8125, + 3588556.5, + 3322374.0 + ], + "bp_grad_per_sample_l2_med": [ + 6.797127571189776e-05, + 3.067998122219251e-08, + 4.995078572989087e-09, + 1.7760940496103217e-09, + 9.021596802938348e-10 + ], + "acc_eval": 0.1162109375, + "epoch": 56 + }, + { + "hidden_norms_cls": [ + 3.283642292022705, + 33408.87109375, + 662075.8125, + 3597291.5, + 3329052.75 + ], + "bp_grad_per_sample_l2_med": [ + 6.738945376127958e-05, + 3.091497902119045e-08, + 5.170774475260487e-09, + 1.7745169778038417e-09, + 9.013288448933565e-10 + ], + "acc_eval": 0.1162109375, + "epoch": 57 + }, + { + "hidden_norms_cls": [ + 3.285024404525757, + 33476.29296875, + 663102.5, + 3601966.75, + 3333036.25 + ], + "bp_grad_per_sample_l2_med": [ + 6.544289499288425e-05, + 3.0026765074353534e-08, + 4.907802608755674e-09, + 1.7734476109865227e-09, + 9.008140344768378e-10 + ], + "acc_eval": 0.1181640625, + "epoch": 58 + }, + { + "hidden_norms_cls": [ + 3.285569667816162, + 33510.2890625, + 663440.8125, + 3604244.75, + 3334545.25 + ], + "bp_grad_per_sample_l2_med": [ + 6.654243770753965e-05, + 3.0157259800489555e-08, + 4.895804206483945e-09, + 1.773489466394551e-09, + 9.008297996437875e-10 + ], + "acc_eval": 0.1171875, + "epoch": 59 + }, + { + "hidden_norms_cls": [ + 3.2856643199920654, + 33519.1015625, + 663484.0, + 3604572.0, + 3334760.75 + ], + "bp_grad_per_sample_l2_med": [ + 6.647893314948305e-05, + 3.0106672710417115e-08, + 4.919553653337516e-09, + 1.7735447555011774e-09, + 9.008529477938509e-10 + ], + "acc_eval": 0.1171875, + "epoch": 60 + } + ], + "arch": "vit", + "seed": 42 +}
\ No newline at end of file diff --git a/results/snapshot_vit_v1/snapshot_vit_s123.json b/results/snapshot_vit_v1/snapshot_vit_s123.json new file mode 100644 index 0000000..93853cf --- /dev/null +++ b/results/snapshot_vit_v1/snapshot_vit_s123.json @@ -0,0 +1,4594 @@ +{ + "config": { + "output_dir": "results/snapshot_vit_v1", + "epochs": 60, + "lr": 0.001, + "wd": 0.05, + "seed": 123, + "depth": 4, + "d_model": 128, + "n_heads": 4 + }, + "depth": 4, + "d_model": 128, + "architecture": "ViTMini", + "bp_log": [ + { + "hidden_norms_cls": [ + 0.3210873007774353, + 3.3568029403686523, + 4.891699314117432, + 6.200707912445068, + 7.5911865234375 + ], + "hidden_norms_avg": [ + 5.7549052238464355, + 6.750590801239014, + 7.9020819664001465, + 8.925304412841797, + 10.076716423034668 + ], + "bp_grad_per_sample_l2_med": [ + 0.004074312746524811, + 0.0011155740357935429, + 0.0009667170234024525, + 0.0008621186716482043, + 0.0008127145702019334 + ], + "bp_grad_F": [ + 0.20160320401191711, + 0.039521340280771255, + 0.033098917454481125, + 0.029271923005580902, + 0.027136052027344704 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.125, + "loss_eval": 2.422111988067627, + "epoch": 0 + }, + { + "hidden_norms_cls": [ + 0.4782717227935791, + 11.874238014221191, + 20.313709259033203, + 26.86528778076172, + 32.42327880859375 + ], + "hidden_norms_avg": [ + 3.8830745220184326, + 11.266603469848633, + 17.881502151489258, + 23.234352111816406, + 28.51181411743164 + ], + "bp_grad_per_sample_l2_med": [ + 0.002827065996825695, + 0.00024586476502008736, + 0.00014567613834515214, + 0.00013882767234463245, + 0.0001400413311785087 + ], + "bp_grad_F": [ + 0.12621107697486877, + 0.009982436895370483, + 0.00507162231951952, + 0.004677945747971535, + 0.004658848978579044 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.4697265625, + "loss_eval": 1.4472862482070923, + "epoch": 1 + }, + { + "hidden_norms_cls": [ + 0.6104467511177063, + 13.787332534790039, + 22.073436737060547, + 28.480369567871094, + 33.141117095947266 + ], + "hidden_norms_avg": [ + 3.951775074005127, + 12.005932807922363, + 18.336896896362305, + 23.67597198486328, + 27.899608612060547 + ], + "bp_grad_per_sample_l2_med": [ + 0.002604336943477392, + 0.00026667251950129867, + 0.00014805165119469166, + 0.00013079405471216887, + 0.00012911780504509807 + ], + "bp_grad_F": [ + 0.12487494200468063, + 0.010650832206010818, + 0.005604646634310484, + 0.004644650500267744, + 0.00448915408924222 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.5146484375, + "loss_eval": 1.291869878768921, + "epoch": 2 + }, + { + "hidden_norms_cls": [ + 0.7104504108428955, + 13.74225902557373, + 21.90972328186035, + 28.958967208862305, + 33.262393951416016 + ], + "hidden_norms_avg": [ + 3.9180142879486084, + 13.097892761230469, + 19.570770263671875, + 24.899248123168945, + 28.65104866027832 + ], + "bp_grad_per_sample_l2_med": [ + 0.0020766372326761484, + 0.00025831969105638564, + 0.00014769421250093728, + 0.00012920337030664086, + 0.00012324318231549114 + ], + "bp_grad_F": [ + 0.09676238149404526, + 0.010603155009448528, + 0.005832126829773188, + 0.00465927692130208, + 0.004435759503394365 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.529296875, + "loss_eval": 1.2694497108459473, + "epoch": 3 + }, + { + "hidden_norms_cls": [ + 0.801612377166748, + 12.882991790771484, + 20.474157333374023, + 27.37413215637207, + 32.06389617919922 + ], + "hidden_norms_avg": [ + 3.8871910572052, + 13.333250999450684, + 19.042104721069336, + 24.029348373413086, + 27.96086311340332 + ], + "bp_grad_per_sample_l2_med": [ + 0.0020045258570462465, + 0.000285606540273875, + 0.00016217664233408868, + 0.00013080642384011298, + 0.00011968115722993389 + ], + "bp_grad_F": [ + 0.09496191889047623, + 0.011928788386285305, + 0.006509022321552038, + 0.005044482182711363, + 0.004537967965006828 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.58203125, + "loss_eval": 1.119094967842102, + "epoch": 4 + }, + { + "hidden_norms_cls": [ + 0.8653663396835327, + 12.994612693786621, + 20.49618148803711, + 26.709278106689453, + 30.303852081298828 + ], + "hidden_norms_avg": [ + 4.018242835998535, + 13.520885467529297, + 19.090721130371094, + 23.6751651763916, + 26.93212890625 + ], + "bp_grad_per_sample_l2_med": [ + 0.0018953323597088456, + 0.000296358106425032, + 0.00017310198745690286, + 0.0001364975469186902, + 0.00012098452862119302 + ], + "bp_grad_F": [ + 0.09499228745698929, + 0.012825642712414265, + 0.007407570723444223, + 0.005570532754063606, + 0.004951393231749535 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.595703125, + "loss_eval": 1.0847156047821045, + "epoch": 5 + }, + { + "hidden_norms_cls": [ + 0.9313029646873474, + 13.966381072998047, + 21.648372650146484, + 28.33364486694336, + 32.91484069824219 + ], + "hidden_norms_avg": [ + 4.053382396697998, + 15.042261123657227, + 20.74883270263672, + 25.032489776611328, + 28.55679702758789 + ], + "bp_grad_per_sample_l2_med": [ + 0.001841629622504115, + 0.0002753600711002946, + 0.0001624371943762526, + 0.0001305017649428919, + 0.00011278261081315577 + ], + "bp_grad_F": [ + 0.09068351984024048, + 0.0128449946641922, + 0.007257991470396519, + 0.005361978895962238, + 0.004660472739487886 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.6103515625, + "loss_eval": 1.095273494720459, + "epoch": 6 + }, + { + "hidden_norms_cls": [ + 0.9818395972251892, + 13.022477149963379, + 20.385391235351562, + 27.03214454650879, + 31.324769973754883 + ], + "hidden_norms_avg": [ + 4.1024041175842285, + 14.697205543518066, + 19.634239196777344, + 23.90021324157715, + 27.306678771972656 + ], + "bp_grad_per_sample_l2_med": [ + 0.0015902157174423337, + 0.00027956697158515453, + 0.0001692597143119201, + 0.0001353132538497448, + 0.00011523282591952011 + ], + "bp_grad_F": [ + 0.07967441529035568, + 0.012528380379080772, + 0.007520876359194517, + 0.00570708978921175, + 0.004875381011515856 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.62890625, + "loss_eval": 1.0069011449813843, + "epoch": 7 + }, + { + "hidden_norms_cls": [ + 1.0194361209869385, + 13.912450790405273, + 20.85824203491211, + 27.968477249145508, + 32.243431091308594 + ], + "hidden_norms_avg": [ + 4.178755283355713, + 14.875143051147461, + 20.261545181274414, + 24.636741638183594, + 27.893512725830078 + ], + "bp_grad_per_sample_l2_med": [ + 0.001506888889707625, + 0.00028421770548447967, + 0.0001727779454085976, + 0.00013076694449409842, + 0.00010860477777896449 + ], + "bp_grad_F": [ + 0.07398127764463425, + 0.013124965131282806, + 0.007798053789883852, + 0.005717460531741381, + 0.004830002784729004 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.638671875, + "loss_eval": 0.9834396839141846, + "epoch": 8 + }, + { + "hidden_norms_cls": [ + 1.077261209487915, + 12.524529457092285, + 19.766538619995117, + 25.7626895904541, + 30.03727149963379 + ], + "hidden_norms_avg": [ + 4.1816864013671875, + 14.304539680480957, + 19.208524703979492, + 23.25862693786621, + 26.313356399536133 + ], + "bp_grad_per_sample_l2_med": [ + 0.0014360174536705017, + 0.0003003668098244816, + 0.0001815372525015846, + 0.0001390709658153355, + 0.00011320491466904059 + ], + "bp_grad_F": [ + 0.0756259635090828, + 0.014865156263113022, + 0.008982912637293339, + 0.006551471073180437, + 0.005398898851126432 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.6494140625, + "loss_eval": 0.971701979637146, + "epoch": 9 + }, + { + "hidden_norms_cls": [ + 1.1200045347213745, + 12.160452842712402, + 19.172168731689453, + 25.067665100097656, + 28.764787673950195 + ], + "hidden_norms_avg": [ + 4.289422035217285, + 15.51036262512207, + 20.407268524169922, + 24.061931610107422, + 26.493688583374023 + ], + "bp_grad_per_sample_l2_med": [ + 0.0012870824430137873, + 0.00027074789977632463, + 0.0001742457679938525, + 0.0001348328951280564, + 0.000104573555290699 + ], + "bp_grad_F": [ + 0.07437644153833389, + 0.015084434300661087, + 0.009522279724478722, + 0.0069595095701515675, + 0.0055610560812056065 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.6767578125, + "loss_eval": 0.8734689354896545, + "epoch": 10 + }, + { + "hidden_norms_cls": [ + 1.1509246826171875, + 12.219743728637695, + 19.352998733520508, + 25.70184326171875, + 29.580278396606445 + ], + "hidden_norms_avg": [ + 4.315939426422119, + 15.6610689163208, + 20.42215347290039, + 24.016841888427734, + 26.62118911743164 + ], + "bp_grad_per_sample_l2_med": [ + 0.0013146374840289354, + 0.0002733681467361748, + 0.0001865291123976931, + 0.00013389455853030086, + 0.00010298956476617604 + ], + "bp_grad_F": [ + 0.07836466282606125, + 0.015104672871530056, + 0.009902945719659328, + 0.0069915857166051865, + 0.005596037954092026 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.6767578125, + "loss_eval": 0.917751669883728, + "epoch": 11 + }, + { + "hidden_norms_cls": [ + 1.2017452716827393, + 12.72107219696045, + 18.952699661254883, + 25.067615509033203, + 29.466550827026367 + ], + "hidden_norms_avg": [ + 4.358119487762451, + 15.672205924987793, + 20.140914916992188, + 24.013885498046875, + 26.668514251708984 + ], + "bp_grad_per_sample_l2_med": [ + 0.0011630960507318377, + 0.00025659528910182416, + 0.0001693729282123968, + 0.0001237613905686885, + 9.384296572534367e-05 + ], + "bp_grad_F": [ + 0.07334695011377335, + 0.015075902454555035, + 0.009827414527535439, + 0.007068648934364319, + 0.005549146793782711 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.6982421875, + "loss_eval": 0.8279671669006348, + "epoch": 12 + }, + { + "hidden_norms_cls": [ + 1.234994649887085, + 12.283885955810547, + 18.62163543701172, + 24.3900203704834, + 28.36040496826172 + ], + "hidden_norms_avg": [ + 4.369320869445801, + 15.747625350952148, + 19.76953887939453, + 23.351999282836914, + 25.69841194152832 + ], + "bp_grad_per_sample_l2_med": [ + 0.0011812442680820823, + 0.0002851892786566168, + 0.0001940266229212284, + 0.00013827465591020882, + 0.00010182376718148589 + ], + "bp_grad_F": [ + 0.07158326357603073, + 0.01604902744293213, + 0.010893523693084717, + 0.007844515144824982, + 0.005985426716506481 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.689453125, + "loss_eval": 0.8679210543632507, + "epoch": 13 + }, + { + "hidden_norms_cls": [ + 1.2675292491912842, + 11.706314086914062, + 18.347841262817383, + 24.01951789855957, + 28.1114501953125 + ], + "hidden_norms_avg": [ + 4.390969753265381, + 15.140497207641602, + 19.62596893310547, + 23.22857666015625, + 25.681791305541992 + ], + "bp_grad_per_sample_l2_med": [ + 0.0010449079563841224, + 0.0002611770760267973, + 0.00017687672516331077, + 0.00012088462244719267, + 8.996625547297299e-05 + ], + "bp_grad_F": [ + 0.06979352980852127, + 0.016320781782269478, + 0.01088168565183878, + 0.007916715927422047, + 0.006064872723072767 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.6884765625, + "loss_eval": 0.8092349767684937, + "epoch": 14 + }, + { + "hidden_norms_cls": [ + 1.2957582473754883, + 11.933137893676758, + 18.238237380981445, + 23.88395118713379, + 27.868513107299805 + ], + "hidden_norms_avg": [ + 4.4383697509765625, + 15.46267318725586, + 19.841346740722656, + 23.529712677001953, + 25.852684020996094 + ], + "bp_grad_per_sample_l2_med": [ + 0.0009947444777935743, + 0.00024384066637139767, + 0.00016608397709205747, + 0.0001234996016137302, + 9.023427992360666e-05 + ], + "bp_grad_F": [ + 0.07019422203302383, + 0.016806265339255333, + 0.011433122679591179, + 0.008135766722261906, + 0.006158903241157532 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.712890625, + "loss_eval": 0.7795700430870056, + "epoch": 15 + }, + { + "hidden_norms_cls": [ + 1.3391923904418945, + 11.329567909240723, + 17.184999465942383, + 22.353059768676758, + 26.889299392700195 + ], + "hidden_norms_avg": [ + 4.53090238571167, + 15.341407775878906, + 19.109649658203125, + 22.223899841308594, + 24.45347023010254 + ], + "bp_grad_per_sample_l2_med": [ + 0.001086357282474637, + 0.00030167572549544275, + 0.0002114812668878585, + 0.00015480270667467266, + 0.00010993685282301158 + ], + "bp_grad_F": [ + 0.06569939106702805, + 0.017201317474246025, + 0.012068319134414196, + 0.008909368887543678, + 0.006427914369851351 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7314453125, + "loss_eval": 0.7724976539611816, + "epoch": 16 + }, + { + "hidden_norms_cls": [ + 1.3643215894699097, + 10.724340438842773, + 16.144981384277344, + 21.656490325927734, + 25.777481079101562 + ], + "hidden_norms_avg": [ + 4.520992279052734, + 15.036820411682129, + 18.98927879333496, + 22.688859939575195, + 24.71843910217285 + ], + "bp_grad_per_sample_l2_med": [ + 0.000967099389526993, + 0.0002715744194574654, + 0.00019328886992298067, + 0.00013628297892864794, + 9.314809722127393e-05 + ], + "bp_grad_F": [ + 0.06768078356981277, + 0.01841692440211773, + 0.012980606406927109, + 0.009295133873820305, + 0.006698420271277428 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.73828125, + "loss_eval": 0.7337102890014648, + "epoch": 17 + }, + { + "hidden_norms_cls": [ + 1.3963993787765503, + 11.463358879089355, + 16.49429702758789, + 21.585325241088867, + 25.636695861816406 + ], + "hidden_norms_avg": [ + 4.539251327514648, + 14.975208282470703, + 18.360631942749023, + 21.586400985717773, + 23.561830520629883 + ], + "bp_grad_per_sample_l2_med": [ + 0.0008780099451541901, + 0.00024953793035820127, + 0.00017615300021134317, + 0.00012566034274641424, + 8.528940088581294e-05 + ], + "bp_grad_F": [ + 0.07198558002710342, + 0.019673649221658707, + 0.013869633898139, + 0.009656366892158985, + 0.006824122741818428 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7451171875, + "loss_eval": 0.7211208343505859, + "epoch": 18 + }, + { + "hidden_norms_cls": [ + 1.421103835105896, + 9.9105806350708, + 15.361790657043457, + 20.334768295288086, + 24.506919860839844 + ], + "hidden_norms_avg": [ + 4.622361660003662, + 14.7937650680542, + 18.042116165161133, + 21.208105087280273, + 23.21940040588379 + ], + "bp_grad_per_sample_l2_med": [ + 0.0007793674594722688, + 0.0002331113355467096, + 0.00016331924416590482, + 0.00011330798588460311, + 7.215427467599511e-05 + ], + "bp_grad_F": [ + 0.07613872736692429, + 0.022306306287646294, + 0.015858706086874008, + 0.011275039985775948, + 0.007263499312102795 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7392578125, + "loss_eval": 0.7346716523170471, + "epoch": 19 + }, + { + "hidden_norms_cls": [ + 1.4554439783096313, + 10.118419647216797, + 15.106456756591797, + 20.052860260009766, + 24.34604263305664 + ], + "hidden_norms_avg": [ + 4.584441661834717, + 14.634279251098633, + 17.785133361816406, + 21.110010147094727, + 22.89650535583496 + ], + "bp_grad_per_sample_l2_med": [ + 0.0007524031680077314, + 0.0002260785986436531, + 0.00016854463319759816, + 0.00011997718684142455, + 7.738151180092245e-05 + ], + "bp_grad_F": [ + 0.07370379567146301, + 0.021284889429807663, + 0.015382407233119011, + 0.010745048522949219, + 0.007394883316010237 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.74609375, + "loss_eval": 0.7024286389350891, + "epoch": 20 + }, + { + "hidden_norms_cls": [ + 1.4652466773986816, + 10.145299911499023, + 15.15330982208252, + 20.131061553955078, + 23.987186431884766 + ], + "hidden_norms_avg": [ + 4.602361679077148, + 14.532269477844238, + 17.825733184814453, + 21.43619728088379, + 23.12371253967285 + ], + "bp_grad_per_sample_l2_med": [ + 0.0008222013129852712, + 0.0002455596113577485, + 0.00017989490879699588, + 0.00012026322656311095, + 8.26524556032382e-05 + ], + "bp_grad_F": [ + 0.0741497054696083, + 0.020758304744958878, + 0.01482379250228405, + 0.010753357782959938, + 0.007582026533782482 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7412109375, + "loss_eval": 0.7016957402229309, + "epoch": 21 + }, + { + "hidden_norms_cls": [ + 1.480756402015686, + 10.028592109680176, + 14.573174476623535, + 19.24639320373535, + 23.3629150390625 + ], + "hidden_norms_avg": [ + 4.621757984161377, + 14.29589557647705, + 17.4040470123291, + 20.71320343017578, + 22.405963897705078 + ], + "bp_grad_per_sample_l2_med": [ + 0.000667835003696382, + 0.00020642780873458833, + 0.0001538341457489878, + 0.00010652844503056258, + 6.964046770008281e-05 + ], + "bp_grad_F": [ + 0.07375485450029373, + 0.021967202425003052, + 0.016106218099594116, + 0.011507184244692326, + 0.007651185616850853 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7607421875, + "loss_eval": 0.6544591188430786, + "epoch": 22 + }, + { + "hidden_norms_cls": [ + 1.5069013833999634, + 9.897040367126465, + 14.096169471740723, + 18.488874435424805, + 22.214311599731445 + ], + "hidden_norms_avg": [ + 4.621402263641357, + 13.978808403015137, + 16.55535888671875, + 19.729736328125, + 21.272388458251953 + ], + "bp_grad_per_sample_l2_med": [ + 0.0007426440715789795, + 0.0002414636837784201, + 0.00018161055049858987, + 0.00012697202328126878, + 8.096778037725016e-05 + ], + "bp_grad_F": [ + 0.07682260870933533, + 0.02333010919392109, + 0.017287475988268852, + 0.012499523349106312, + 0.008510846644639969 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7568359375, + "loss_eval": 0.6971542835235596, + "epoch": 23 + }, + { + "hidden_norms_cls": [ + 1.5169755220413208, + 9.5064058303833, + 13.812594413757324, + 17.95937156677246, + 21.51923370361328 + ], + "hidden_norms_avg": [ + 4.612667083740234, + 13.86723804473877, + 16.418981552124023, + 19.12198257446289, + 20.565521240234375 + ], + "bp_grad_per_sample_l2_med": [ + 0.0007257908000610769, + 0.00024720808141864836, + 0.0001871327986009419, + 0.00013439780741464347, + 8.671147952554747e-05 + ], + "bp_grad_F": [ + 0.07474599778652191, + 0.024257108569145203, + 0.01782340556383133, + 0.01281843613833189, + 0.008487668819725513 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.763671875, + "loss_eval": 0.6618016958236694, + "epoch": 24 + }, + { + "hidden_norms_cls": [ + 1.5133440494537354, + 9.023777961730957, + 13.088467597961426, + 17.324251174926758, + 21.083759307861328 + ], + "hidden_norms_avg": [ + 4.670994281768799, + 13.658943176269531, + 16.629199981689453, + 19.75893783569336, + 21.12710952758789 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005614451365545392, + 0.0001914121676236391, + 0.00014186649059411138, + 0.00010088241106132045, + 6.150588887976483e-05 + ], + "bp_grad_F": [ + 0.08172759413719177, + 0.027048049494624138, + 0.019993988797068596, + 0.014050732366740704, + 0.009047404862940311 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7626953125, + "loss_eval": 0.6708824634552002, + "epoch": 25 + }, + { + "hidden_norms_cls": [ + 1.5329865217208862, + 8.596381187438965, + 12.645307540893555, + 16.668807983398438, + 20.51462745666504 + ], + "hidden_norms_avg": [ + 4.642580986022949, + 13.140735626220703, + 15.32632064819336, + 17.985010147094727, + 19.472858428955078 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005455430946312845, + 0.00019156686903443187, + 0.00014253854169510305, + 0.00010098751954501495, + 6.338314415188506e-05 + ], + "bp_grad_F": [ + 0.08767001330852509, + 0.029545463621616364, + 0.021587952971458435, + 0.014973495155572891, + 0.009515349753201008 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.767578125, + "loss_eval": 0.6813780069351196, + "epoch": 26 + }, + { + "hidden_norms_cls": [ + 1.554770588874817, + 8.98424243927002, + 12.76674747467041, + 16.438858032226562, + 19.893146514892578 + ], + "hidden_norms_avg": [ + 4.635522365570068, + 13.039203643798828, + 15.419903755187988, + 18.176319122314453, + 19.39232635498047 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005697832675650716, + 0.0001985357521334663, + 0.0001536469644634053, + 0.00010455933079356328, + 6.821998249506578e-05 + ], + "bp_grad_F": [ + 0.07977671921253204, + 0.02706393413245678, + 0.020037846639752388, + 0.014591805636882782, + 0.009512925520539284 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.794921875, + "loss_eval": 0.628034770488739, + "epoch": 27 + }, + { + "hidden_norms_cls": [ + 1.5515426397323608, + 8.19682788848877, + 11.861184120178223, + 15.59919548034668, + 18.893400192260742 + ], + "hidden_norms_avg": [ + 4.5841450691223145, + 12.511472702026367, + 14.583609580993652, + 16.949586868286133, + 18.082292556762695 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005373116582632065, + 0.0001997901708818972, + 0.0001489907590439543, + 0.00010373342229286209, + 6.318451050901785e-05 + ], + "bp_grad_F": [ + 0.09340352565050125, + 0.03397079557180405, + 0.025108041241765022, + 0.017669958993792534, + 0.010975479148328304 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.767578125, + "loss_eval": 0.7042175531387329, + "epoch": 28 + }, + { + "hidden_norms_cls": [ + 1.5542187690734863, + 8.489326477050781, + 11.786734580993652, + 15.331045150756836, + 18.48196792602539 + ], + "hidden_norms_avg": [ + 4.615735054016113, + 12.620647430419922, + 14.822388648986816, + 17.454334259033203, + 18.677461624145508 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005812083836644888, + 0.00021461205324158072, + 0.00015997944865375757, + 0.00011693408305291086, + 7.380295573966578e-05 + ], + "bp_grad_F": [ + 0.09380948543548584, + 0.03345760330557823, + 0.025129586458206177, + 0.01789030432701111, + 0.011130682192742825 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7666015625, + "loss_eval": 0.6850110292434692, + "epoch": 29 + }, + { + "hidden_norms_cls": [ + 1.5533884763717651, + 8.01724910736084, + 11.423017501831055, + 15.046612739562988, + 18.549089431762695 + ], + "hidden_norms_avg": [ + 4.590519428253174, + 12.431540489196777, + 14.243208885192871, + 16.609176635742188, + 18.00058937072754 + ], + "bp_grad_per_sample_l2_med": [ + 0.0004959268844686449, + 0.00020078590023331344, + 0.00015426705067511648, + 0.00010096926416736096, + 5.95865749346558e-05 + ], + "bp_grad_F": [ + 0.09663982689380646, + 0.036162860691547394, + 0.02716146595776081, + 0.01926880143582821, + 0.011334826238453388 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7783203125, + "loss_eval": 0.6707749366760254, + "epoch": 30 + }, + { + "hidden_norms_cls": [ + 1.5530469417572021, + 7.959159851074219, + 11.2805757522583, + 14.567391395568848, + 17.63263511657715 + ], + "hidden_norms_avg": [ + 4.583434581756592, + 11.983453750610352, + 13.4386625289917, + 15.498583793640137, + 16.631824493408203 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005142099689692259, + 0.00020230942755006254, + 0.00015552931290585548, + 0.00011007695866283029, + 6.233472959138453e-05 + ], + "bp_grad_F": [ + 0.09787525236606598, + 0.037690743803977966, + 0.028980012983083725, + 0.02064511366188526, + 0.012224731035530567 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7763671875, + "loss_eval": 0.6748013496398926, + "epoch": 31 + }, + { + "hidden_norms_cls": [ + 1.5515272617340088, + 7.847743988037109, + 10.868992805480957, + 14.282126426696777, + 17.300424575805664 + ], + "hidden_norms_avg": [ + 4.579150199890137, + 11.951078414916992, + 13.621748924255371, + 15.971713066101074, + 17.04889488220215 + ], + "bp_grad_per_sample_l2_med": [ + 0.0004585929855238646, + 0.00017806614050641656, + 0.0001380510366288945, + 9.63187703746371e-05, + 5.829996371176094e-05 + ], + "bp_grad_F": [ + 0.10181005299091339, + 0.038575656712055206, + 0.02964734099805355, + 0.020979750901460648, + 0.01240287534892559 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.78125, + "loss_eval": 0.648021936416626, + "epoch": 32 + }, + { + "hidden_norms_cls": [ + 1.5606788396835327, + 7.502675533294678, + 10.430459022521973, + 13.642136573791504, + 16.445953369140625 + ], + "hidden_norms_avg": [ + 4.601283550262451, + 11.615697860717773, + 13.472222328186035, + 15.831238746643066, + 16.856765747070312 + ], + "bp_grad_per_sample_l2_med": [ + 0.0004074577009305358, + 0.00017814920283854008, + 0.00013730341743212193, + 9.565176151227206e-05, + 5.229458474786952e-05 + ], + "bp_grad_F": [ + 0.11506060510873795, + 0.04695325344800949, + 0.03571247681975365, + 0.025198202580213547, + 0.013851121068000793 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.755859375, + "loss_eval": 0.7299192547798157, + "epoch": 33 + }, + { + "hidden_norms_cls": [ + 1.5404834747314453, + 7.528626441955566, + 10.384718894958496, + 13.326257705688477, + 16.41958999633789 + ], + "hidden_norms_avg": [ + 4.533090591430664, + 11.6575345993042, + 13.231060028076172, + 15.337231636047363, + 16.463298797607422 + ], + "bp_grad_per_sample_l2_med": [ + 0.00042279696208424866, + 0.00016955600585788488, + 0.00012915951083414257, + 8.976108074421063e-05, + 4.930083014187403e-05 + ], + "bp_grad_F": [ + 0.11639728397130966, + 0.047109171748161316, + 0.03660130873322487, + 0.02588750049471855, + 0.013930009678006172 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7822265625, + "loss_eval": 0.7065586447715759, + "epoch": 34 + }, + { + "hidden_norms_cls": [ + 1.539638638496399, + 7.26485013961792, + 9.96584701538086, + 13.003418922424316, + 16.14008140563965 + ], + "hidden_norms_avg": [ + 4.537028789520264, + 11.146146774291992, + 12.836923599243164, + 14.992585182189941, + 16.076574325561523 + ], + "bp_grad_per_sample_l2_med": [ + 0.00039646061486564577, + 0.00017108044994529337, + 0.00013601842510979623, + 9.225023677572608e-05, + 4.893271398032084e-05 + ], + "bp_grad_F": [ + 0.11648086458444595, + 0.04862738400697708, + 0.0374443456530571, + 0.026762915775179863, + 0.014388044364750385 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7939453125, + "loss_eval": 0.6772804260253906, + "epoch": 35 + }, + { + "hidden_norms_cls": [ + 1.5278778076171875, + 7.107171058654785, + 9.836174011230469, + 12.691095352172852, + 15.726043701171875 + ], + "hidden_norms_avg": [ + 4.5205254554748535, + 10.983088493347168, + 12.684239387512207, + 14.876561164855957, + 15.817031860351562 + ], + "bp_grad_per_sample_l2_med": [ + 0.0003774319775402546, + 0.00015706375415902585, + 0.0001264828024432063, + 8.857168722897768e-05, + 4.440043630893342e-05 + ], + "bp_grad_F": [ + 0.12492093443870544, + 0.053397826850414276, + 0.040937285870313644, + 0.029012493789196014, + 0.015446576289832592 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7724609375, + "loss_eval": 0.707297146320343, + "epoch": 36 + }, + { + "hidden_norms_cls": [ + 1.51571524143219, + 7.153889179229736, + 9.856640815734863, + 12.511648178100586, + 15.47674560546875 + ], + "hidden_norms_avg": [ + 4.51829719543457, + 10.948126792907715, + 12.496319770812988, + 14.353506088256836, + 15.267280578613281 + ], + "bp_grad_per_sample_l2_med": [ + 0.0003654273459687829, + 0.00015038737910799682, + 0.00011584001913433895, + 8.340556814800948e-05, + 4.3377422116464004e-05 + ], + "bp_grad_F": [ + 0.13914552330970764, + 0.0589178130030632, + 0.04522731155157089, + 0.0323280394077301, + 0.016300896182656288 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7626953125, + "loss_eval": 0.729290246963501, + "epoch": 37 + }, + { + "hidden_norms_cls": [ + 1.5081822872161865, + 6.952159404754639, + 9.558011054992676, + 12.107258796691895, + 15.05250358581543 + ], + "hidden_norms_avg": [ + 4.481203556060791, + 10.740157127380371, + 12.00869083404541, + 13.831336975097656, + 14.76778507232666 + ], + "bp_grad_per_sample_l2_med": [ + 0.0003083393967244774, + 0.00012594371219165623, + 0.00010108885180670768, + 7.270355854416266e-05, + 3.639453643700108e-05 + ], + "bp_grad_F": [ + 0.14465992152690887, + 0.061646901071071625, + 0.047660429030656815, + 0.03381076827645302, + 0.01706051267683506 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7685546875, + "loss_eval": 0.7395438551902771, + "epoch": 38 + }, + { + "hidden_norms_cls": [ + 1.5159180164337158, + 6.878056526184082, + 9.382442474365234, + 11.969950675964355, + 15.129233360290527 + ], + "hidden_norms_avg": [ + 4.494526386260986, + 10.639691352844238, + 12.291059494018555, + 14.28632640838623, + 15.170000076293945 + ], + "bp_grad_per_sample_l2_med": [ + 0.00028951463173143566, + 0.00012502263416536152, + 0.00010040518827736378, + 7.046512473607436e-05, + 3.719307642313652e-05 + ], + "bp_grad_F": [ + 0.1406005620956421, + 0.06109313294291496, + 0.047722309827804565, + 0.033977217972278595, + 0.01711316779255867 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7802734375, + "loss_eval": 0.7057114839553833, + "epoch": 39 + }, + { + "hidden_norms_cls": [ + 1.5058479309082031, + 6.699619770050049, + 9.136896133422852, + 11.609856605529785, + 14.701480865478516 + ], + "hidden_norms_avg": [ + 4.482093811035156, + 10.446189880371094, + 12.056195259094238, + 13.87569522857666, + 14.783234596252441 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002856751380022615, + 0.00013275284436531365, + 0.00010139105870621279, + 6.966136425035074e-05, + 3.293903864687309e-05 + ], + "bp_grad_F": [ + 0.1529398262500763, + 0.06890096515417099, + 0.052683208137750626, + 0.03734185919165611, + 0.01789216138422489 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7822265625, + "loss_eval": 0.7478293180465698, + "epoch": 40 + }, + { + "hidden_norms_cls": [ + 1.490049123764038, + 6.627772331237793, + 9.019189834594727, + 11.385313987731934, + 14.173355102539062 + ], + "hidden_norms_avg": [ + 4.458098411560059, + 10.279450416564941, + 11.593936920166016, + 13.30923080444336, + 14.155406951904297 + ], + "bp_grad_per_sample_l2_med": [ + 0.00021343027765396982, + 9.837630932452157e-05, + 8.006956340977922e-05, + 5.6768003560137004e-05, + 2.7385456633055583e-05 + ], + "bp_grad_F": [ + 0.15772707760334015, + 0.07120667397975922, + 0.05564703419804573, + 0.03953033685684204, + 0.018776116892695427 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7724609375, + "loss_eval": 0.7677009105682373, + "epoch": 41 + }, + { + "hidden_norms_cls": [ + 1.4904595613479614, + 6.347933769226074, + 8.674281120300293, + 11.007567405700684, + 14.106690406799316 + ], + "hidden_norms_avg": [ + 4.430132865905762, + 10.265702247619629, + 11.588266372680664, + 13.410375595092773, + 14.36179256439209 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002224426862085238, + 0.00010000570182455704, + 7.794459816068411e-05, + 5.595289985649288e-05, + 2.7256841349299066e-05 + ], + "bp_grad_F": [ + 0.16928334534168243, + 0.07747916132211685, + 0.06056727096438408, + 0.04274875670671463, + 0.019731448963284492 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.771484375, + "loss_eval": 0.7774850726127625, + "epoch": 42 + }, + { + "hidden_norms_cls": [ + 1.4839239120483398, + 6.341518878936768, + 8.600671768188477, + 10.991724014282227, + 14.061833381652832 + ], + "hidden_norms_avg": [ + 4.403061866760254, + 10.020040512084961, + 11.371668815612793, + 13.18271255493164, + 14.007345199584961 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002152575325453654, + 0.00010309015488019213, + 7.964189717313275e-05, + 5.283725113258697e-05, + 2.479495196894277e-05 + ], + "bp_grad_F": [ + 0.17190901935100555, + 0.080362468957901, + 0.06201368197798729, + 0.042563296854496, + 0.019763268530368805 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.791015625, + "loss_eval": 0.7503129243850708, + "epoch": 43 + }, + { + "hidden_norms_cls": [ + 1.479953408241272, + 6.386821269989014, + 8.521474838256836, + 10.847907066345215, + 13.541037559509277 + ], + "hidden_norms_avg": [ + 4.399942874908447, + 9.979388236999512, + 11.251038551330566, + 12.998858451843262, + 13.76570987701416 + ], + "bp_grad_per_sample_l2_med": [ + 0.000195752116269432, + 9.365259029436857e-05, + 7.282401929842308e-05, + 5.209222945268266e-05, + 2.3047552531352267e-05 + ], + "bp_grad_F": [ + 0.16316288709640503, + 0.07674747705459595, + 0.06052006781101227, + 0.04277755320072174, + 0.020131859928369522 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7978515625, + "loss_eval": 0.732360303401947, + "epoch": 44 + }, + { + "hidden_norms_cls": [ + 1.4770045280456543, + 6.249807834625244, + 8.36440658569336, + 10.429434776306152, + 13.045490264892578 + ], + "hidden_norms_avg": [ + 4.391754627227783, + 9.821492195129395, + 10.789708137512207, + 12.3140869140625, + 13.067754745483398 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002329840644961223, + 0.00011798225023085251, + 9.261347440769896e-05, + 6.294570630416274e-05, + 2.9543562050093897e-05 + ], + "bp_grad_F": [ + 0.1812749058008194, + 0.08621949702501297, + 0.067754827439785, + 0.04831995815038681, + 0.021977350115776062 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7822265625, + "loss_eval": 0.7784227132797241, + "epoch": 45 + }, + { + "hidden_norms_cls": [ + 1.467228651046753, + 6.0983357429504395, + 8.231528282165527, + 10.333056449890137, + 13.266757011413574 + ], + "hidden_norms_avg": [ + 4.390368461608887, + 9.647501945495605, + 10.692501068115234, + 12.274677276611328, + 13.044783592224121 + ], + "bp_grad_per_sample_l2_med": [ + 0.0001511751615907997, + 7.784536137478426e-05, + 6.136627780506387e-05, + 4.35807196481619e-05, + 2.014869824051857e-05 + ], + "bp_grad_F": [ + 0.18805110454559326, + 0.09124159067869186, + 0.07224944978952408, + 0.050726212561130524, + 0.02254536561667919 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7822265625, + "loss_eval": 0.814141571521759, + "epoch": 46 + }, + { + "hidden_norms_cls": [ + 1.4649879932403564, + 6.082813739776611, + 8.131811141967773, + 10.158056259155273, + 13.049956321716309 + ], + "hidden_norms_avg": [ + 4.377193927764893, + 9.599099159240723, + 10.660999298095703, + 12.21728801727295, + 12.965741157531738 + ], + "bp_grad_per_sample_l2_med": [ + 0.0001712174853309989, + 8.400671504205093e-05, + 6.461612792918459e-05, + 4.6120971092022955e-05, + 2.0111323465243913e-05 + ], + "bp_grad_F": [ + 0.19550266861915588, + 0.09418290108442307, + 0.07462290674448013, + 0.053083814680576324, + 0.02292729914188385 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7890625, + "loss_eval": 0.7999259233474731, + "epoch": 47 + }, + { + "hidden_norms_cls": [ + 1.4617111682891846, + 6.058816432952881, + 8.100686073303223, + 10.085453987121582, + 12.920005798339844 + ], + "hidden_norms_avg": [ + 4.367196559906006, + 9.56169605255127, + 10.609153747558594, + 12.135587692260742, + 12.871316909790039 + ], + "bp_grad_per_sample_l2_med": [ + 0.00015642307698726654, + 7.727487536612898e-05, + 5.90925665164832e-05, + 4.136386633035727e-05, + 1.6983105524559505e-05 + ], + "bp_grad_F": [ + 0.202835351228714, + 0.09727983921766281, + 0.07683655619621277, + 0.05471916124224663, + 0.02326873689889908 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.783203125, + "loss_eval": 0.8152254223823547, + "epoch": 48 + }, + { + "hidden_norms_cls": [ + 1.4593255519866943, + 6.017489433288574, + 7.96034049987793, + 9.983046531677246, + 12.839889526367188 + ], + "hidden_norms_avg": [ + 4.349035739898682, + 9.508291244506836, + 10.475739479064941, + 12.025425910949707, + 12.764323234558105 + ], + "bp_grad_per_sample_l2_med": [ + 0.00013039771874900907, + 6.479993317043409e-05, + 5.060105831944384e-05, + 3.637294139480218e-05, + 1.4738084246346261e-05 + ], + "bp_grad_F": [ + 0.2170599102973938, + 0.10417382419109344, + 0.08229079097509384, + 0.05735529586672783, + 0.024290865287184715 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7841796875, + "loss_eval": 0.8357614278793335, + "epoch": 49 + }, + { + "hidden_norms_cls": [ + 1.4566220045089722, + 5.9422221183776855, + 7.888944149017334, + 9.933752059936523, + 12.817610740661621 + ], + "hidden_norms_avg": [ + 4.346043109893799, + 9.43443489074707, + 10.415618896484375, + 11.964469909667969, + 12.66396713256836 + ], + "bp_grad_per_sample_l2_med": [ + 0.00011193535465281457, + 5.694849824067205e-05, + 4.533635365078226e-05, + 3.3492226066300645e-05, + 1.4018894034961704e-05 + ], + "bp_grad_F": [ + 0.21774698793888092, + 0.1065358892083168, + 0.08431421220302582, + 0.05883141607046127, + 0.024615641683340073 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7822265625, + "loss_eval": 0.8349668979644775, + "epoch": 50 + }, + { + "hidden_norms_cls": [ + 1.4569600820541382, + 5.859008312225342, + 7.792572498321533, + 9.738831520080566, + 12.622002601623535 + ], + "hidden_norms_avg": [ + 4.331563472747803, + 9.398598670959473, + 10.273933410644531, + 11.764141082763672, + 12.436033248901367 + ], + "bp_grad_per_sample_l2_med": [ + 0.00012003595475107431, + 5.908689126954414e-05, + 4.746428385260515e-05, + 3.251975431339815e-05, + 1.3768915778200608e-05 + ], + "bp_grad_F": [ + 0.2144007831811905, + 0.10626016557216644, + 0.08404342085123062, + 0.05926896259188652, + 0.024889416992664337 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7861328125, + "loss_eval": 0.840790331363678, + "epoch": 51 + }, + { + "hidden_norms_cls": [ + 1.4559087753295898, + 5.8081440925598145, + 7.735576629638672, + 9.665096282958984, + 12.462581634521484 + ], + "hidden_norms_avg": [ + 4.331978797912598, + 9.35647964477539, + 10.206826210021973, + 11.663065910339355, + 12.307528495788574 + ], + "bp_grad_per_sample_l2_med": [ + 0.00010101464431500062, + 5.143512680660933e-05, + 4.1592236811993644e-05, + 3.0060553399380296e-05, + 1.2701412742899265e-05 + ], + "bp_grad_F": [ + 0.22189198434352875, + 0.11008423566818237, + 0.08644267171621323, + 0.0603644996881485, + 0.025003522634506226 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7900390625, + "loss_eval": 0.8574135303497314, + "epoch": 52 + }, + { + "hidden_norms_cls": [ + 1.4537159204483032, + 5.80010461807251, + 7.719310760498047, + 9.681657791137695, + 12.724189758300781 + ], + "hidden_norms_avg": [ + 4.318180084228516, + 9.302933692932129, + 10.182857513427734, + 11.675416946411133, + 12.381869316101074 + ], + "bp_grad_per_sample_l2_med": [ + 9.399914415553212e-05, + 4.920759965898469e-05, + 3.82029902539216e-05, + 2.4410484911641106e-05, + 9.912354471453e-06 + ], + "bp_grad_F": [ + 0.2312120646238327, + 0.11490234732627869, + 0.09031722694635391, + 0.06280362606048584, + 0.02546737529337406 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7900390625, + "loss_eval": 0.8825892210006714, + "epoch": 53 + }, + { + "hidden_norms_cls": [ + 1.4533940553665161, + 5.765341758728027, + 7.664762496948242, + 9.541180610656738, + 12.46136474609375 + ], + "hidden_norms_avg": [ + 4.321633815765381, + 9.284127235412598, + 10.135884284973145, + 11.603636741638184, + 12.27259635925293 + ], + "bp_grad_per_sample_l2_med": [ + 0.00010168847802560776, + 5.239037636783905e-05, + 4.169415478827432e-05, + 2.8813867174903862e-05, + 1.1632122550508939e-05 + ], + "bp_grad_F": [ + 0.23138603568077087, + 0.11581514030694962, + 0.09145721793174744, + 0.06368965655565262, + 0.025816213339567184 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7900390625, + "loss_eval": 0.8627672791481018, + "epoch": 54 + }, + { + "hidden_norms_cls": [ + 1.4520151615142822, + 5.752710342407227, + 7.6372294425964355, + 9.493330001831055, + 12.49281120300293 + ], + "hidden_norms_avg": [ + 4.314420223236084, + 9.275971412658691, + 10.131872177124023, + 11.594268798828125, + 12.271230697631836 + ], + "bp_grad_per_sample_l2_med": [ + 0.00010349372314522043, + 5.0890805141534656e-05, + 4.167861698078923e-05, + 2.9301188988029025e-05, + 1.1288396308373194e-05 + ], + "bp_grad_F": [ + 0.23472057282924652, + 0.11708012968301773, + 0.09245092421770096, + 0.06478185206651688, + 0.025885937735438347 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.791015625, + "loss_eval": 0.8611111640930176, + "epoch": 55 + }, + { + "hidden_norms_cls": [ + 1.4508332014083862, + 5.741517066955566, + 7.59091329574585, + 9.493820190429688, + 12.553393363952637 + ], + "hidden_norms_avg": [ + 4.315334320068359, + 9.272273063659668, + 10.107754707336426, + 11.5936279296875, + 12.266173362731934 + ], + "bp_grad_per_sample_l2_med": [ + 9.209421114064753e-05, + 4.9557409511180595e-05, + 3.8665639294777066e-05, + 2.703870450204704e-05, + 1.0333649697713554e-05 + ], + "bp_grad_F": [ + 0.23745231330394745, + 0.11897031217813492, + 0.09419847279787064, + 0.06559450179338455, + 0.02597379870712757 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.794921875, + "loss_eval": 0.8737660646438599, + "epoch": 56 + }, + { + "hidden_norms_cls": [ + 1.4514665603637695, + 5.724127292633057, + 7.567911624908447, + 9.464110374450684, + 12.53247356414795 + ], + "hidden_norms_avg": [ + 4.312892436981201, + 9.258728981018066, + 10.092171669006348, + 11.559837341308594, + 12.241935729980469 + ], + "bp_grad_per_sample_l2_med": [ + 8.842872193781659e-05, + 4.636628364096396e-05, + 3.5660887078847736e-05, + 2.4343042241525836e-05, + 9.563225830788724e-06 + ], + "bp_grad_F": [ + 0.23917540907859802, + 0.12006859481334686, + 0.09492167085409164, + 0.06607519090175629, + 0.02614584006369114 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7939453125, + "loss_eval": 0.8743407130241394, + "epoch": 57 + }, + { + "hidden_norms_cls": [ + 1.4513946771621704, + 5.7332282066345215, + 7.562880992889404, + 9.466405868530273, + 12.53167724609375 + ], + "hidden_norms_avg": [ + 4.3138651847839355, + 9.26041316986084, + 10.077324867248535, + 11.538521766662598, + 12.218836784362793 + ], + "bp_grad_per_sample_l2_med": [ + 9.282731480197981e-05, + 4.6137181925587356e-05, + 3.7237725337035954e-05, + 2.553310514485929e-05, + 9.964967830455862e-06 + ], + "bp_grad_F": [ + 0.24150457978248596, + 0.12129233777523041, + 0.09592130035161972, + 0.06674522906541824, + 0.026258554309606552 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7890625, + "loss_eval": 0.8750766515731812, + "epoch": 58 + }, + { + "hidden_norms_cls": [ + 1.4510550498962402, + 5.730595111846924, + 7.558650493621826, + 9.460441589355469, + 12.49626636505127 + ], + "hidden_norms_avg": [ + 4.311552047729492, + 9.2510347366333, + 10.070302963256836, + 11.530484199523926, + 12.212501525878906 + ], + "bp_grad_per_sample_l2_med": [ + 8.91910312930122e-05, + 4.6800774725852534e-05, + 3.584511432563886e-05, + 2.5313198420917615e-05, + 9.680608854978345e-06 + ], + "bp_grad_F": [ + 0.24175073206424713, + 0.12143565714359283, + 0.09590216726064682, + 0.06680992990732193, + 0.026268428191542625 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.787109375, + "loss_eval": 0.8746636509895325, + "epoch": 59 + }, + { + "hidden_norms_cls": [ + 1.450972557067871, + 5.727794647216797, + 7.5582756996154785, + 9.462045669555664, + 12.50033187866211 + ], + "hidden_norms_avg": [ + 4.311238765716553, + 9.248146057128906, + 10.065601348876953, + 11.523144721984863, + 12.206249237060547 + ], + "bp_grad_per_sample_l2_med": [ + 8.883981354301795e-05, + 4.730010550701991e-05, + 3.57187855115626e-05, + 2.4922017473727465e-05, + 9.98118775896728e-06 + ], + "bp_grad_F": [ + 0.24178682267665863, + 0.12154776602983475, + 0.09598654508590698, + 0.06685590744018555, + 0.026261666789650917 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.787109375, + "loss_eval": 0.8750864863395691, + "epoch": 60 + } + ], + "dfa_log": [ + { + "hidden_norms_cls": [ + 0.3210873007774353, + 3.3568029403686523, + 4.891699314117432, + 6.200707912445068, + 7.5911865234375 + ], + "hidden_norms_avg": [ + 5.7549052238464355, + 6.750590801239014, + 7.9020819664001465, + 8.925304412841797, + 10.076716423034668 + ], + "bp_grad_per_sample_l2_med": [ + 0.004074312746524811, + 0.0011155740357935429, + 0.0009667170234024525, + 0.0008621186716482043, + 0.0008127145702019334 + ], + "bp_grad_F": [ + 0.20160320401191711, + 0.039521340280771255, + 0.033098917454481125, + 0.029271923005580902, + 0.027136052027344704 + ], + "gamma_dfa": 0.0011031709145754576, + "gamma_dfa_per_layer": [ + -0.004096934571862221, + -0.003170878393575549, + 0.003171987598761916, + 0.008508509024977684 + ], + "acc_eval": 0.125, + "loss_eval": 2.422111988067627, + "epoch": 0 + }, + { + "hidden_norms_cls": [ + 1.7445409297943115, + 3393.872314453125, + 8096.7275390625, + 11786.4453125, + 15317.9609375 + ], + "hidden_norms_avg": [ + 46.059967041015625, + 2908.78857421875, + 9192.9140625, + 12480.814453125, + 15843.904296875 + ], + "bp_grad_per_sample_l2_med": [ + 0.0007700070273131132, + 5.177932962396881e-07, + 4.354141651674581e-07, + 4.0796010125632165e-07, + 4.096184795798763e-07 + ], + "bp_grad_F": [ + 0.14847944676876068, + 7.175570499384776e-05, + 1.7355905583826825e-05, + 1.4375820683198981e-05, + 1.4468741937889718e-05 + ], + "gamma_dfa": 0.00903075491078198, + "gamma_dfa_per_layer": [ + 0.014700280502438545, + 0.007455145940184593, + 0.009132559411227703, + 0.004835033789277077 + ], + "acc_eval": 0.2158203125, + "loss_eval": 2.072723388671875, + "epoch": 1 + }, + { + "hidden_norms_cls": [ + 3.5329439640045166, + 17943.025390625, + 46227.09765625, + 60098.9921875, + 105510.8828125 + ], + "hidden_norms_avg": [ + 95.54289245605469, + 14775.4267578125, + 44856.01953125, + 59234.35546875, + 87862.078125 + ], + "bp_grad_per_sample_l2_med": [ + 0.00029628784977830946, + 1.0676165373979529e-07, + 7.34040170868866e-08, + 6.658416396021494e-08, + 6.750254044618487e-08 + ], + "bp_grad_F": [ + 0.12055971473455429, + 8.933278149925172e-05, + 3.910079612978734e-06, + 3.3086737403209554e-06, + 3.354543650857522e-06 + ], + "gamma_dfa": 0.007886230800068006, + "gamma_dfa_per_layer": [ + 0.016902755945920944, + 0.003473796183243394, + 0.009775697253644466, + 0.0013926738174632192 + ], + "acc_eval": 0.2197265625, + "loss_eval": 2.0924925804138184, + "epoch": 2 + }, + { + "hidden_norms_cls": [ + 5.165565013885498, + 33863.33203125, + 93132.8515625, + 145828.34375, + 276496.0 + ], + "hidden_norms_avg": [ + 142.7160186767578, + 36882.9296875, + 93356.96875, + 138199.0625, + 216913.359375 + ], + "bp_grad_per_sample_l2_med": [ + 0.00022745825117453933, + 4.267175768291054e-08, + 3.181965624321492e-08, + 2.8314032007870082e-08, + 2.8843814448009653e-08 + ], + "bp_grad_F": [ + 0.06954267621040344, + 0.00012680335203185678, + 2.0300262804084923e-06, + 1.8529049157223199e-06, + 1.8672349142434541e-06 + ], + "gamma_dfa": 0.00556139531545341, + "gamma_dfa_per_layer": [ + 0.01000538095831871, + 0.0014471756294369698, + 0.011906315572559834, + -0.001113290898501873 + ], + "acc_eval": 0.2412109375, + "loss_eval": 2.0521552562713623, + "epoch": 3 + }, + { + "hidden_norms_cls": [ + 6.674264907836914, + 43835.6796875, + 150229.609375, + 253133.515625, + 481876.5 + ], + "hidden_norms_avg": [ + 186.87832641601562, + 74412.015625, + 150159.890625, + 253252.734375, + 439077.25 + ], + "bp_grad_per_sample_l2_med": [ + 0.0003350951592437923, + 2.7693221937852286e-08, + 1.989219278186738e-08, + 1.704596108709211e-08, + 1.7547710839949104e-08 + ], + "bp_grad_F": [ + 0.058595605194568634, + 0.0002476655936334282, + 1.0984595064655878e-06, + 1.0368654557169066e-06, + 1.0405622106191004e-06 + ], + "gamma_dfa": 0.002481349220033735, + "gamma_dfa_per_layer": [ + 3.730843309313059e-05, + 0.0014904557028785348, + 0.011494583450257778, + -0.0030969507060945034 + ], + "acc_eval": 0.232421875, + "loss_eval": 2.058844566345215, + "epoch": 4 + }, + { + "hidden_norms_cls": [ + 8.061405181884766, + 102865.78125, + 229004.390625, + 413580.59375, + 706253.625 + ], + "hidden_norms_avg": [ + 228.99224853515625, + 131342.546875, + 229257.390625, + 410435.25, + 727001.6875 + ], + "bp_grad_per_sample_l2_med": [ + 0.00029934433405287564, + 1.670629679040303e-08, + 1.3200744497510186e-08, + 1.1068386207568892e-08, + 1.1248447506773118e-08 + ], + "bp_grad_F": [ + 0.49810686707496643, + 0.0004038630868308246, + 1.391591922583757e-06, + 6.11680206930032e-07, + 5.242948759587307e-07 + ], + "gamma_dfa": 0.0009187236428260803, + "gamma_dfa_per_layer": [ + -0.003027769736945629, + 0.002294017933309078, + 0.00904145184904337, + -0.004632805474102497 + ], + "acc_eval": 0.2197265625, + "loss_eval": 2.091141700744629, + "epoch": 5 + }, + { + "hidden_norms_cls": [ + 9.400635719299316, + 164205.859375, + 302756.3125, + 584338.25, + 1250749.125 + ], + "hidden_norms_avg": [ + 270.8563232421875, + 190832.34375, + 307117.59375, + 620500.3125, + 1149191.375 + ], + "bp_grad_per_sample_l2_med": [ + 0.00034470504033379257, + 1.3897826889319731e-08, + 8.662226491651381e-09, + 6.850678690994982e-09, + 7.0105756755367565e-09 + ], + "bp_grad_F": [ + 0.08854538202285767, + 0.0002125167811755091, + 3.9876397295302013e-07, + 2.9177280680414697e-07, + 2.903063318626664e-07 + ], + "gamma_dfa": 0.0008902656263671815, + "gamma_dfa_per_layer": [ + -0.004116768017411232, + 0.004187953192740679, + 0.007109872996807098, + -0.003619995666667819 + ], + "acc_eval": 0.208984375, + "loss_eval": 2.09378981590271, + "epoch": 6 + }, + { + "hidden_norms_cls": [ + 10.676177978515625, + 216327.671875, + 429372.28125, + 910110.1875, + 1955787.375 + ], + "hidden_norms_avg": [ + 311.3556823730469, + 279124.65625, + 434169.5, + 917506.125, + 1714750.5 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002802011731546372, + 8.183582700382885e-09, + 5.617039278149605e-09, + 4.765932093420133e-09, + 4.884364024348997e-09 + ], + "bp_grad_F": [ + 0.2021757811307907, + 0.00017153903900180012, + 2.627296851187566e-07, + 1.9583571031489555e-07, + 1.9719628596703842e-07 + ], + "gamma_dfa": 0.00245969754178077, + "gamma_dfa_per_layer": [ + 0.0030665153171867132, + 0.0039703575894236565, + 0.005591260269284248, + -0.0027893430087715387 + ], + "acc_eval": 0.224609375, + "loss_eval": 2.0816030502319336, + "epoch": 7 + }, + { + "hidden_norms_cls": [ + 11.970849990844727, + 282019.0, + 553729.25, + 1250316.5, + 2676136.75 + ], + "hidden_norms_avg": [ + 350.0874938964844, + 367917.96875, + 563719.125, + 1231394.75, + 2379567.75 + ], + "bp_grad_per_sample_l2_med": [ + 0.00015754146443214267, + 7.837374305097455e-09, + 4.1749870227647534e-09, + 3.572480755664742e-09, + 3.6429026462059255e-09 + ], + "bp_grad_F": [ + 0.06816119700670242, + 0.0002165154874091968, + 1.8739321205885062e-07, + 1.4396131575722393e-07, + 1.449055417879208e-07 + ], + "gamma_dfa": 0.0018902571755461395, + "gamma_dfa_per_layer": [ + 0.0010774387046694756, + 0.004019944928586483, + 0.004736708011478186, + -0.0022730629425495863 + ], + "acc_eval": 0.2236328125, + "loss_eval": 2.069587469100952, + "epoch": 8 + }, + { + "hidden_norms_cls": [ + 13.234355926513672, + 498576.65625, + 747356.3125, + 1569417.0, + 3492356.0 + ], + "hidden_norms_avg": [ + 390.34356689453125, + 517596.5, + 755798.625, + 1559863.25, + 3087860.75 + ], + "bp_grad_per_sample_l2_med": [ + 0.00012954612611792982, + 6.955718223622398e-09, + 3.3238092278509157e-09, + 2.7362234700234467e-09, + 2.7891511322764018e-09 + ], + "bp_grad_F": [ + 0.03943011909723282, + 0.00019621588580776006, + 1.9796995331944345e-07, + 1.1484767270530938e-07, + 1.1313876058238748e-07 + ], + "gamma_dfa": 0.002013051969697699, + "gamma_dfa_per_layer": [ + 0.0007974229520186782, + 0.0045172227546572685, + 0.004162413999438286, + -0.0014248518273234367 + ], + "acc_eval": 0.21484375, + "loss_eval": 2.0846829414367676, + "epoch": 9 + }, + { + "hidden_norms_cls": [ + 14.412362098693848, + 653898.875, + 962527.9375, + 1974772.75, + 4235229.0 + ], + "hidden_norms_avg": [ + 426.1747741699219, + 660897.75, + 954217.625, + 1992937.5, + 3912584.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0001297950657317415, + 8.463428180505161e-09, + 2.735849546908753e-09, + 2.2942541200166033e-09, + 2.329255233135541e-09 + ], + "bp_grad_F": [ + 0.07335253059864044, + 0.0002778704511001706, + 2.1924184068211616e-07, + 1.5885484572208952e-07, + 9.551467172741468e-08 + ], + "gamma_dfa": 0.001994484744500369, + "gamma_dfa_per_layer": [ + 0.0028600264340639114, + 0.0031631651800125837, + 0.0033731083385646343, + -0.0014183609746396542 + ], + "acc_eval": 0.2373046875, + "loss_eval": 2.0768301486968994, + "epoch": 10 + }, + { + "hidden_norms_cls": [ + 15.627301216125488, + 805321.625, + 1191654.25, + 2375967.75, + 5068243.5 + ], + "hidden_norms_avg": [ + 464.92144775390625, + 870169.75, + 1209570.75, + 2342305.5, + 4797491.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00014358766202349216, + 6.655038742309216e-09, + 2.213497385383789e-09, + 1.8578552030135143e-09, + 1.8845198734851465e-09 + ], + "bp_grad_F": [ + 0.22341646254062653, + 0.0010545816039666533, + 3.022053363110899e-07, + 8.581165644727662e-08, + 7.942276880612553e-08 + ], + "gamma_dfa": 0.0012935771082993597, + "gamma_dfa_per_layer": [ + -0.00014482985716313124, + 0.0035164642613381147, + 0.002910400740802288, + -0.0011077267117798328 + ], + "acc_eval": 0.2138671875, + "loss_eval": 2.076242446899414, + "epoch": 11 + }, + { + "hidden_norms_cls": [ + 16.751590728759766, + 758638.375, + 1437438.0, + 2779040.25, + 5662257.0 + ], + "hidden_norms_avg": [ + 501.60235595703125, + 1119163.625, + 1513203.375, + 2677392.25, + 5339190.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00020390149438753724, + 5.5311413227343564e-09, + 1.8890189412701375e-09, + 1.6241474831701908e-09, + 1.6578916017806478e-09 + ], + "bp_grad_F": [ + 1.734735369682312, + 0.000374374445527792, + 1.2862761877840967e-07, + 1.1062630989044919e-07, + 6.887377423936414e-08 + ], + "gamma_dfa": 0.0020944410644005984, + "gamma_dfa_per_layer": [ + 0.0036250988487154245, + 0.0027163547929376364, + 0.0028407545760273933, + -0.0008044439600780606 + ], + "acc_eval": 0.23046875, + "loss_eval": 2.083003520965576, + "epoch": 12 + }, + { + "hidden_norms_cls": [ + 17.76032066345215, + 1093088.5, + 1739864.0, + 3080344.5, + 6632118.5 + ], + "hidden_norms_avg": [ + 539.1420288085938, + 1433551.75, + 1843404.125, + 3096921.75, + 6237853.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00014031230239197612, + 4.252878937904825e-09, + 1.704668295410272e-09, + 1.518661418842271e-09, + 1.5263605934734414e-09 + ], + "bp_grad_F": [ + 0.05247655138373375, + 0.00048432740732096136, + 1.395327586806161e-07, + 8.375257465331742e-08, + 5.970760241780226e-08 + ], + "gamma_dfa": 0.0018704376125242561, + "gamma_dfa_per_layer": [ + 0.0021162822376936674, + 0.003257386852055788, + 0.002796958899125457, + -0.0006888775387778878 + ], + "acc_eval": 0.2177734375, + "loss_eval": 2.1021971702575684, + "epoch": 13 + }, + { + "hidden_norms_cls": [ + 18.760025024414062, + 2223954.75, + 2460609.0, + 3718541.25, + 7618764.0 + ], + "hidden_norms_avg": [ + 577.8324584960938, + 1908819.625, + 2348359.5, + 3542983.0, + 6779725.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00010403544729342684, + 2.662567277766925e-09, + 1.6195218499603925e-09, + 1.422262307926303e-09, + 1.433551610752204e-09 + ], + "bp_grad_F": [ + 0.08673901855945587, + 0.0001567020226502791, + 8.41187173250546e-08, + 5.8843131256480774e-08, + 5.371895994699116e-08 + ], + "gamma_dfa": 0.003976444830186665, + "gamma_dfa_per_layer": [ + 0.010876781307160854, + 0.0029237696435302496, + 0.0026041539385914803, + -0.000498925568535924 + ], + "acc_eval": 0.2177734375, + "loss_eval": 2.138009548187256, + "epoch": 14 + }, + { + "hidden_norms_cls": [ + 19.738407135009766, + 2339586.0, + 2696652.5, + 4235438.5, + 7955372.0 + ], + "hidden_norms_avg": [ + 619.4888916015625, + 2243613.75, + 2801893.5, + 4018190.75, + 7143025.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00014718857710249722, + 2.6654392026870255e-09, + 1.4668186665289795e-09, + 1.3113429231381701e-09, + 1.3212382299343517e-09 + ], + "bp_grad_F": [ + 0.1564428061246872, + 0.00024187436792999506, + 2.621962664761668e-07, + 5.580703898999673e-08, + 5.442945649747344e-08 + ], + "gamma_dfa": 0.0019682426573126577, + "gamma_dfa_per_layer": [ + 0.003052386222407222, + 0.0027148486115038395, + 0.0025699653197079897, + -0.00046422952436842024 + ], + "acc_eval": 0.2265625, + "loss_eval": 2.145447254180908, + "epoch": 15 + }, + { + "hidden_norms_cls": [ + 20.727081298828125, + 2753822.5, + 3114285.75, + 4713040.0, + 8652256.0 + ], + "hidden_norms_avg": [ + 660.8045654296875, + 2684893.0, + 3331407.5, + 4614215.5, + 7864381.5 + ], + "bp_grad_per_sample_l2_med": [ + 0.00013858993770554662, + 2.7730009399817845e-09, + 1.344458322449782e-09, + 1.2393497339502346e-09, + 1.2431760065823028e-09 + ], + "bp_grad_F": [ + 0.1324291080236435, + 0.00022225745487958193, + 9.285694346772289e-08, + 4.7669061586930184e-08, + 4.746264181676452e-08 + ], + "gamma_dfa": 0.001855946145951748, + "gamma_dfa_per_layer": [ + 0.004218719433993101, + 0.001718730665743351, + 0.0019668969325721264, + -0.0004805624485015869 + ], + "acc_eval": 0.251953125, + "loss_eval": 2.101409673690796, + "epoch": 16 + }, + { + "hidden_norms_cls": [ + 21.603151321411133, + 3877476.75, + 4156743.75, + 5474745.5, + 10299149.0 + ], + "hidden_norms_avg": [ + 698.7645263671875, + 3523264.5, + 4085091.75, + 5354783.0, + 9094561.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.000199332891497761, + 1.9098649328697093e-09, + 1.2144363292776461e-09, + 1.1166952917918138e-09, + 1.1315383074972374e-09 + ], + "bp_grad_F": [ + 0.15142236649990082, + 4.2183397454209626e-05, + 7.309424887580462e-08, + 5.2962114693855256e-08, + 4.2337880756804225e-08 + ], + "gamma_dfa": 0.0013709207050851546, + "gamma_dfa_per_layer": [ + 0.0023175266105681658, + 0.0016730788629502058, + 0.0018372924532741308, + -0.0003442151064518839 + ], + "acc_eval": 0.220703125, + "loss_eval": 2.1280713081359863, + "epoch": 17 + }, + { + "hidden_norms_cls": [ + 22.48230743408203, + 3581296.5, + 4024996.25, + 6183271.0, + 11393130.0 + ], + "hidden_norms_avg": [ + 735.2485961914062, + 4017679.75, + 4751523.5, + 6215236.0, + 10366257.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00018803548300638795, + 1.895784862426808e-09, + 1.1237420993737146e-09, + 1.0378963244406236e-09, + 1.0573794062551656e-09 + ], + "bp_grad_F": [ + 0.12878409028053284, + 4.194497887510806e-05, + 4.941083631138099e-08, + 3.9714532107382183e-08, + 3.994472663748638e-08 + ], + "gamma_dfa": 0.0012161588310846128, + "gamma_dfa_per_layer": [ + 0.0021385664585977793, + 0.0015012789517641068, + 0.0016565187834203243, + -0.0004317288694437593 + ], + "acc_eval": 0.2158203125, + "loss_eval": 2.1137895584106445, + "epoch": 18 + }, + { + "hidden_norms_cls": [ + 23.29258155822754, + 4616424.5, + 5061994.0, + 6831782.5, + 12437102.0 + ], + "hidden_norms_avg": [ + 768.5599975585938, + 4586839.5, + 5567332.5, + 6953557.0, + 11342041.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00021542877948377281, + 1.9688994878919175e-09, + 1.0831572305747272e-09, + 9.59386015075836e-10, + 9.76972058808201e-10 + ], + "bp_grad_F": [ + 0.09346769750118256, + 4.979455843567848e-05, + 4.6927411290198506e-08, + 3.8312244043936516e-08, + 3.882623289541698e-08 + ], + "gamma_dfa": 0.0011904535058420151, + "gamma_dfa_per_layer": [ + 0.0024171490222215652, + 0.0012326111318543553, + 0.0015420420095324516, + -0.0004299881402403116 + ], + "acc_eval": 0.2333984375, + "loss_eval": 2.0970499515533447, + "epoch": 19 + }, + { + "hidden_norms_cls": [ + 24.054462432861328, + 5806660.5, + 6270370.0, + 7861086.0, + 14600824.0 + ], + "hidden_norms_avg": [ + 801.1851806640625, + 5855518.5, + 6747570.0, + 8137969.0, + 13373154.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00012799039541278034, + 1.5442244150065676e-09, + 9.774002718287989e-10, + 8.729071909741037e-10, + 8.921305361120346e-10 + ], + "bp_grad_F": [ + 0.9552626609802246, + 1.163827892014524e-05, + 2.461903250150499e-07, + 3.4675800009154045e-08, + 3.4831803219503854e-08 + ], + "gamma_dfa": 0.0018355100764892995, + "gamma_dfa_per_layer": [ + 0.004604285582900047, + 0.00167210737708956, + 0.001404301729053259, + -0.0003386543830856681 + ], + "acc_eval": 0.2138671875, + "loss_eval": 2.11769700050354, + "epoch": 20 + }, + { + "hidden_norms_cls": [ + 24.826902389526367, + 6414338.0, + 6989573.0, + 8609765.0, + 15368301.0 + ], + "hidden_norms_avg": [ + 830.1490478515625, + 6946265.5, + 7916239.0, + 9224772.0, + 14431417.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00015248128329403698, + 1.4729809594271615e-09, + 9.462167716023373e-10, + 8.629364445233989e-10, + 8.805687845558907e-10 + ], + "bp_grad_F": [ + 0.1705506592988968, + 0.00016797611897345632, + 4.810285503253908e-08, + 3.399246040203252e-08, + 3.436711182303043e-08 + ], + "gamma_dfa": 0.0015317398574552499, + "gamma_dfa_per_layer": [ + 0.004021936096251011, + 0.0012830595951527357, + 0.0012305387062951922, + -0.00040857496787793934 + ], + "acc_eval": 0.2236328125, + "loss_eval": 2.102954864501953, + "epoch": 21 + }, + { + "hidden_norms_cls": [ + 25.52900505065918, + 8152435.5, + 8736555.0, + 9881049.0, + 16997340.0 + ], + "hidden_norms_avg": [ + 858.3137817382812, + 7544203.0, + 8860779.0, + 10224814.0, + 15607154.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0001363755582133308, + 1.2969386675720784e-09, + 9.174655479782245e-10, + 8.421383035361885e-10, + 8.538599827190296e-10 + ], + "bp_grad_F": [ + 0.07549053430557251, + 2.9420058126561344e-05, + 5.8904493727141016e-08, + 3.187446395713778e-08, + 3.1860317051268794e-08 + ], + "gamma_dfa": 0.0023631651129107922, + "gamma_dfa_per_layer": [ + 0.007739294320344925, + 0.0009300304809585214, + 0.001171439653262496, + -0.00038810400292277336 + ], + "acc_eval": 0.2080078125, + "loss_eval": 2.1162400245666504, + "epoch": 22 + }, + { + "hidden_norms_cls": [ + 26.122589111328125, + 9255466.0, + 9964734.0, + 11054792.0, + 18565818.0 + ], + "hidden_norms_avg": [ + 884.7098388671875, + 8814330.0, + 10177720.0, + 11552646.0, + 17489728.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00014142012514639646, + 1.3653665975610352e-09, + 8.580440802319345e-10, + 7.922714151398225e-10, + 8.032630116616701e-10 + ], + "bp_grad_F": [ + 0.11926258355379105, + 1.2633940968953539e-05, + 3.893048727832138e-08, + 3.075604126934195e-08, + 3.043995278062539e-08 + ], + "gamma_dfa": 0.0013515961618395522, + "gamma_dfa_per_layer": [ + 0.0038036422338336706, + 0.0009011125075630844, + 0.0010650388430804014, + -0.0003634089371189475 + ], + "acc_eval": 0.201171875, + "loss_eval": 2.134256362915039, + "epoch": 23 + }, + { + "hidden_norms_cls": [ + 26.71797752380371, + 8691604.0, + 9652268.0, + 11413482.0, + 19342164.0 + ], + "hidden_norms_avg": [ + 908.3400268554688, + 9418884.0, + 11132975.0, + 12548590.0, + 18516546.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00016020517796278, + 1.5232745065318909e-09, + 8.103306359252826e-10, + 7.392123024807518e-10, + 7.529800671868259e-10 + ], + "bp_grad_F": [ + 0.08055854588747025, + 1.5452902516699396e-05, + 3.9288735820264264e-08, + 3.091117406484045e-08, + 2.9865013573271426e-08 + ], + "gamma_dfa": 0.0012649961790884845, + "gamma_dfa_per_layer": [ + 0.0036856848746538162, + 0.0008738536853343248, + 0.0009324033162556589, + -0.00043195715988986194 + ], + "acc_eval": 0.216796875, + "loss_eval": 2.1074087619781494, + "epoch": 24 + }, + { + "hidden_norms_cls": [ + 27.27407455444336, + 9931691.0, + 10836582.0, + 12350148.0, + 20677670.0 + ], + "hidden_norms_avg": [ + 931.8362426757812, + 10575293.0, + 12476721.0, + 13979366.0, + 20145220.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002644142950884998, + 1.4137412351900025e-09, + 7.711813965194381e-10, + 7.245293809354791e-10, + 7.329452600401964e-10 + ], + "bp_grad_F": [ + 0.08752908557653427, + 3.3139032893814147e-05, + 4.584432744536571e-08, + 2.996421955003825e-08, + 2.8689356668110122e-08 + ], + "gamma_dfa": 0.0013240749976830557, + "gamma_dfa_per_layer": [ + 0.003940493334084749, + 0.0008323215297423303, + 0.0009093738626688719, + -0.0003858887357637286 + ], + "acc_eval": 0.2080078125, + "loss_eval": 2.1011557579040527, + "epoch": 25 + }, + { + "hidden_norms_cls": [ + 27.817367553710938, + 12253263.0, + 13105294.0, + 14345712.0, + 22455440.0 + ], + "hidden_norms_avg": [ + 953.0517578125, + 12373978.0, + 14141729.0, + 15597335.0, + 22309104.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00014759604528080672, + 1.123480086739903e-09, + 7.102037291595309e-10, + 6.690955567378865e-10, + 6.791667228611686e-10 + ], + "bp_grad_F": [ + 0.05615593492984772, + 0.00010146480053663254, + 3.714368190799178e-08, + 2.748037175592799e-08, + 2.6385574614096186e-08 + ], + "gamma_dfa": 0.0012311612081248313, + "gamma_dfa_per_layer": [ + 0.003369898535311222, + 0.0009893677197396755, + 0.0009041082230396569, + -0.0003387296455912292 + ], + "acc_eval": 0.234375, + "loss_eval": 2.105236291885376, + "epoch": 26 + }, + { + "hidden_norms_cls": [ + 28.32236099243164, + 13643201.0, + 14611628.0, + 15764247.0, + 24160378.0 + ], + "hidden_norms_avg": [ + 973.9281616210938, + 13221320.0, + 15394035.0, + 16890642.0, + 23678466.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00016657183005008847, + 9.720990679085162e-10, + 6.881580305595492e-10, + 6.496979620962406e-10, + 6.5528305004392e-10 + ], + "bp_grad_F": [ + 0.21266740560531616, + 1.0097430276800878e-05, + 5.8009760550703504e-08, + 2.5056060337647068e-08, + 2.5318335872270836e-08 + ], + "gamma_dfa": 0.0010233484499622136, + "gamma_dfa_per_layer": [ + 0.0030805980786681175, + 0.0005203179316595197, + 0.000822446309030056, + -0.00032996851950883865 + ], + "acc_eval": 0.212890625, + "loss_eval": 2.137085437774658, + "epoch": 27 + }, + { + "hidden_norms_cls": [ + 28.79292106628418, + 13177481.0, + 14299000.0, + 15555998.0, + 23304192.0 + ], + "hidden_norms_avg": [ + 992.415771484375, + 14624837.0, + 16826644.0, + 18325902.0, + 25292966.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00020211696391925216, + 1.1425522750130312e-09, + 6.55300591567709e-10, + 6.094423854463571e-10, + 6.189129764244683e-10 + ], + "bp_grad_F": [ + 0.17359277606010437, + 1.2817296010325663e-05, + 7.396955936656013e-08, + 2.5667416636565576e-08, + 2.485118777428852e-08 + ], + "gamma_dfa": 0.0013693289074581116, + "gamma_dfa_per_layer": [ + 0.004313473589718342, + 0.0007373306434601545, + 0.000740988994948566, + -0.00031447759829461575 + ], + "acc_eval": 0.2255859375, + "loss_eval": 2.095831871032715, + "epoch": 28 + }, + { + "hidden_norms_cls": [ + 29.206546783447266, + 13026473.0, + 14086985.0, + 16207103.0, + 25752542.0 + ], + "hidden_norms_avg": [ + 1012.0676879882812, + 14977745.0, + 17788350.0, + 19459138.0, + 26297900.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002467916056048125, + 1.5415436704913077e-09, + 6.588643519656046e-10, + 6.053296752739357e-10, + 6.112412798131572e-10 + ], + "bp_grad_F": [ + 0.042935822159051895, + 9.192503966914956e-06, + 3.478140442325639e-08, + 2.366191687031005e-08, + 2.384759412166204e-08 + ], + "gamma_dfa": 0.0008732089045224711, + "gamma_dfa_per_layer": [ + 0.002771018771454692, + 0.00032788922544568777, + 0.0007068651611916721, + -0.0003129375400021672 + ], + "acc_eval": 0.2294921875, + "loss_eval": 2.0877132415771484, + "epoch": 29 + }, + { + "hidden_norms_cls": [ + 29.655010223388672, + 18104814.0, + 18959442.0, + 20352136.0, + 30115644.0 + ], + "hidden_norms_avg": [ + 1030.1336669921875, + 17704192.0, + 19945910.0, + 21540694.0, + 29199112.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00019920141494367272, + 8.472024748407136e-10, + 5.961620086480934e-10, + 5.578593698096768e-10, + 5.602495689593923e-10 + ], + "bp_grad_F": [ + 0.05658308416604996, + 2.5135868781944737e-05, + 4.3493169954444966e-08, + 2.1390142990185268e-08, + 2.1341643119399123e-08 + ], + "gamma_dfa": 0.0010463381186127663, + "gamma_dfa_per_layer": [ + 0.0027386434376239777, + 0.0009209397248923779, + 0.0007645037840120494, + -0.0002387344720773399 + ], + "acc_eval": 0.2138671875, + "loss_eval": 2.1298673152923584, + "epoch": 30 + }, + { + "hidden_norms_cls": [ + 30.063920974731445, + 20103628.0, + 20986630.0, + 22120356.0, + 31195618.0 + ], + "hidden_norms_avg": [ + 1046.1927490234375, + 18906620.0, + 21404482.0, + 23038508.0, + 30526688.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002040969266090542, + 9.528980937645315e-10, + 5.700001026731627e-10, + 5.158900528989818e-10, + 5.214851328538828e-10 + ], + "bp_grad_F": [ + 0.0816737711429596, + 5.908617822569795e-05, + 3.4592272157851767e-08, + 2.022995992945198e-08, + 2.0381063947638722e-08 + ], + "gamma_dfa": 0.001085347576008644, + "gamma_dfa_per_layer": [ + 0.0035934830084443092, + 0.0002959886915050447, + 0.0007062341319397092, + -0.0002543155278544873 + ], + "acc_eval": 0.2041015625, + "loss_eval": 2.116055965423584, + "epoch": 31 + }, + { + "hidden_norms_cls": [ + 30.392677307128906, + 21290946.0, + 22100456.0, + 23440498.0, + 32839040.0 + ], + "hidden_norms_avg": [ + 1062.797119140625, + 20072588.0, + 22866046.0, + 24608556.0, + 32259930.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00022883246128913015, + 1.3679187782500435e-09, + 5.444520945196984e-10, + 5.043216955158414e-10, + 5.114720313947885e-10 + ], + "bp_grad_F": [ + 0.2192479968070984, + 0.0001256070681847632, + 1.2757884348957305e-07, + 2.0006645229386777e-08, + 1.9971306386423748e-08 + ], + "gamma_dfa": 0.0019978012933279388, + "gamma_dfa_per_layer": [ + 0.0072806887328624725, + 0.0002412673202343285, + 0.000707695260643959, + -0.00023844614042900503 + ], + "acc_eval": 0.2197265625, + "loss_eval": 2.1034791469573975, + "epoch": 32 + }, + { + "hidden_norms_cls": [ + 30.711090087890625, + 22272344.0, + 23036462.0, + 24462584.0, + 35130192.0 + ], + "hidden_norms_avg": [ + 1082.3468017578125, + 20945452.0, + 23993758.0, + 25838460.0, + 33448582.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005148049676790833, + 2.304462176638822e-09, + 5.340937136999457e-10, + 4.950916898671665e-10, + 5.010572512453848e-10 + ], + "bp_grad_F": [ + 0.15110087394714355, + 3.209088754374534e-05, + 1.0334197497741116e-07, + 2.0131187383753968e-08, + 1.9814510920923567e-08 + ], + "gamma_dfa": 0.003046433750569122, + "gamma_dfa_per_layer": [ + 0.011593285948038101, + 0.00019314857490826398, + 0.0006439671851694584, + -0.0002446667058393359 + ], + "acc_eval": 0.208984375, + "loss_eval": 2.112250328063965, + "epoch": 33 + }, + { + "hidden_norms_cls": [ + 31.0146427154541, + 29106790.0, + 29617952.0, + 31075512.0, + 42307980.0 + ], + "hidden_norms_avg": [ + 1102.4920654296875, + 22080922.0, + 25255976.0, + 27220448.0, + 35072796.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0003266549902036786, + 3.0358568992028268e-09, + 4.893331295718895e-10, + 4.395950825575312e-10, + 4.4405815136094873e-10 + ], + "bp_grad_F": [ + 1.0139163732528687, + 1.6822159523144364e-05, + 1.2544650473955699e-07, + 1.6449900996917677e-08, + 1.5756539184508256e-08 + ], + "gamma_dfa": 0.004347974703705404, + "gamma_dfa_per_layer": [ + 0.016904333606362343, + -2.5271147023886442e-05, + 0.0006614690646529198, + -0.00014863270916976035 + ], + "acc_eval": 0.1806640625, + "loss_eval": 2.1869568824768066, + "epoch": 34 + }, + { + "hidden_norms_cls": [ + 31.282560348510742, + 30793450.0, + 31273076.0, + 33029836.0, + 44659544.0 + ], + "hidden_norms_avg": [ + 1120.8310546875, + 23223250.0, + 26597348.0, + 28740710.0, + 37219140.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0004533866886049509, + 8.621259817154225e-10, + 4.717314872060285e-10, + 4.43607900413312e-10, + 4.492739236194865e-10 + ], + "bp_grad_F": [ + 0.8242164254188538, + 9.425415555597283e-06, + 6.696246401816097e-08, + 1.4385987512355314e-08, + 1.456661191667763e-08 + ], + "gamma_dfa": 0.004019053936644923, + "gamma_dfa_per_layer": [ + 0.015497363172471523, + 0.0001274545502383262, + 0.0006036070408299565, + -0.00015220901696011424 + ], + "acc_eval": 0.1396484375, + "loss_eval": 2.199540615081787, + "epoch": 35 + }, + { + "hidden_norms_cls": [ + 31.512802124023438, + 21969868.0, + 22884238.0, + 24682478.0, + 38405980.0 + ], + "hidden_norms_avg": [ + 1138.61962890625, + 23861642.0, + 27547416.0, + 29877192.0, + 38524416.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00065120670478791, + 1.8789774181016128e-09, + 5.075246334307337e-10, + 4.637440431665141e-10, + 4.687980004192127e-10 + ], + "bp_grad_F": [ + 0.22404567897319794, + 6.978048531891545e-06, + 5.974419536869391e-08, + 1.854446907145757e-08, + 1.8726453276940447e-08 + ], + "gamma_dfa": 0.0030916051182430238, + "gamma_dfa_per_layer": [ + 0.011011095717549324, + 0.0010498585179448128, + 0.0005736061139032245, + -0.00026813987642526627 + ], + "acc_eval": 0.20703125, + "loss_eval": 2.1297407150268555, + "epoch": 36 + }, + { + "hidden_norms_cls": [ + 31.728811264038086, + 29299954.0, + 29843046.0, + 31522240.0, + 44597740.0 + ], + "hidden_norms_avg": [ + 1154.2742919921875, + 26176784.0, + 29361794.0, + 31622036.0, + 40631524.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00057179294526577, + 8.620974489836897e-10, + 4.3859474385676833e-10, + 4.131678610352907e-10, + 4.158901834028228e-10 + ], + "bp_grad_F": [ + 0.1988120824098587, + 6.200312782311812e-06, + 7.633398269035752e-08, + 1.800971105581084e-08, + 1.6420651505200112e-08 + ], + "gamma_dfa": 0.0036353242212499026, + "gamma_dfa_per_layer": [ + 0.013258620165288448, + 0.0009053392568603158, + 0.0005595044931396842, + -0.00018216703028883785 + ], + "acc_eval": 0.193359375, + "loss_eval": 2.162503719329834, + "epoch": 37 + }, + { + "hidden_norms_cls": [ + 31.93093490600586, + 24153264.0, + 24958016.0, + 27164164.0, + 41821856.0 + ], + "hidden_norms_avg": [ + 1168.3504638671875, + 25251302.0, + 29331440.0, + 31888322.0, + 40732864.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0007007869426161051, + 3.803806603031035e-09, + 4.6435830181046356e-10, + 4.2823591894780577e-10, + 4.335505288111108e-10 + ], + "bp_grad_F": [ + 0.23370306193828583, + 5.834654530190164e-06, + 4.249230300956697e-08, + 1.778762737103534e-08, + 1.7575523258983594e-08 + ], + "gamma_dfa": 0.003036663241800852, + "gamma_dfa_per_layer": [ + 0.010693107731640339, + 0.001166085829026997, + 0.0005278221797198057, + -0.00024036277318373322 + ], + "acc_eval": 0.2060546875, + "loss_eval": 2.145090103149414, + "epoch": 38 + }, + { + "hidden_norms_cls": [ + 32.11568069458008, + 30941254.0, + 31479742.0, + 33270594.0, + 46886708.0 + ], + "hidden_norms_avg": [ + 1180.89794921875, + 26732550.0, + 30676362.0, + 33269292.0, + 42382468.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006622264045290649, + 1.5328843749884413e-09, + 4.1302217201888425e-10, + 3.8878930630481534e-10, + 3.9369618676232676e-10 + ], + "bp_grad_F": [ + 0.7251601815223694, + 2.9320523026399314e-05, + 1.0578004605577007e-07, + 1.5190060764780355e-08, + 1.538120386612718e-08 + ], + "gamma_dfa": 0.0034242832334712148, + "gamma_dfa_per_layer": [ + 0.012806901708245277, + 0.0005313451401889324, + 0.0005365218385122716, + -0.00017763575306162238 + ], + "acc_eval": 0.1904296875, + "loss_eval": 2.1417269706726074, + "epoch": 39 + }, + { + "hidden_norms_cls": [ + 32.295955657958984, + 33935232.0, + 34368956.0, + 35934572.0, + 47697056.0 + ], + "hidden_norms_avg": [ + 1192.405517578125, + 27784236.0, + 31654758.0, + 34253712.0, + 43239588.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.000609845039434731, + 1.1788049425476288e-09, + 4.4074988103659507e-10, + 3.883143528948807e-10, + 3.8990835560248627e-10 + ], + "bp_grad_F": [ + 0.3570167124271393, + 5.2556606533471495e-06, + 6.658066098452764e-08, + 1.6441132899558397e-08, + 1.5662063646004754e-08 + ], + "gamma_dfa": 0.0031971330136002507, + "gamma_dfa_per_layer": [ + 0.011760826222598553, + 0.000679977354593575, + 0.0005443703266791999, + -0.00019664184947032481 + ], + "acc_eval": 0.1962890625, + "loss_eval": 2.159468173980713, + "epoch": 40 + }, + { + "hidden_norms_cls": [ + 32.446258544921875, + 19652562.0, + 21036314.0, + 25921266.0, + 40927384.0 + ], + "hidden_norms_avg": [ + 1203.144287109375, + 31749304.0, + 34556376.0, + 36976868.0, + 47059032.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006137214368209243, + 8.001962981118993e-10, + 4.3686718131930036e-10, + 4.163595301864831e-10, + 4.1693556940280985e-10 + ], + "bp_grad_F": [ + 0.12317141890525818, + 1.0874373401748016e-05, + 3.419825134187704e-08, + 1.7081790204542813e-08, + 1.7078438219186864e-08 + ], + "gamma_dfa": 0.002385696723649744, + "gamma_dfa_per_layer": [ + 0.008561503142118454, + 0.000761325703933835, + 0.00044924806570634246, + -0.0002292900171596557 + ], + "acc_eval": 0.1943359375, + "loss_eval": 2.1486196517944336, + "epoch": 41 + }, + { + "hidden_norms_cls": [ + 32.65601348876953, + 18852588.0, + 20806824.0, + 26129350.0, + 41854136.0 + ], + "hidden_norms_avg": [ + 1211.216064453125, + 33512262.0, + 36150576.0, + 38488828.0, + 48876332.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00041886867256835103, + 6.960994003435417e-10, + 3.895795075425923e-10, + 3.7316269518861134e-10, + 3.8092326515304364e-10 + ], + "bp_grad_F": [ + 0.14164908230304718, + 1.2379082363622729e-06, + 2.6289937338219715e-08, + 1.6158612226035984e-08, + 1.6345055087185756e-08 + ], + "gamma_dfa": 0.0026270662965544034, + "gamma_dfa_per_layer": [ + 0.009110232815146446, + 0.0011993220541626215, + 0.00041866287938319147, + -0.00021995256247464567 + ], + "acc_eval": 0.19140625, + "loss_eval": 2.1592655181884766, + "epoch": 42 + }, + { + "hidden_norms_cls": [ + 32.8077507019043, + 13604255.0, + 20737208.0, + 26201736.0, + 38498212.0 + ], + "hidden_norms_avg": [ + 1218.5357666015625, + 34337856.0, + 37186496.0, + 39492900.0, + 49837356.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00024631666019558907, + 7.821958081244418e-10, + 3.8121927836698433e-10, + 3.6691791271969976e-10, + 3.7013728193535655e-10 + ], + "bp_grad_F": [ + 0.16077305376529694, + 4.323472239775583e-06, + 3.772007772795405e-08, + 1.5986831414238623e-08, + 1.5948634413121e-08 + ], + "gamma_dfa": 0.002168099657865241, + "gamma_dfa_per_layer": [ + 0.007105600088834763, + 0.0013786349445581436, + 0.00041434637387283146, + -0.00022618277580477297 + ], + "acc_eval": 0.193359375, + "loss_eval": 2.1537227630615234, + "epoch": 43 + }, + { + "hidden_norms_cls": [ + 32.95870590209961, + 17278438.0, + 21010234.0, + 26451948.0, + 41976576.0 + ], + "hidden_norms_avg": [ + 1225.39501953125, + 36196112.0, + 38833272.0, + 41042520.0, + 51415488.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00017274360288865864, + 7.227461407133262e-10, + 3.619011756939017e-10, + 3.5793992769761473e-10, + 3.6092046018509905e-10 + ], + "bp_grad_F": [ + 0.9672663807868958, + 3.41643476531317e-06, + 5.25712664511957e-08, + 1.5481989024124232e-08, + 1.548605688128646e-08 + ], + "gamma_dfa": 0.001979457345441915, + "gamma_dfa_per_layer": [ + 0.006772718392312527, + 0.0009743775590322912, + 0.00037397228879854083, + -0.00020323885837569833 + ], + "acc_eval": 0.208984375, + "loss_eval": 2.1646997928619385, + "epoch": 44 + }, + { + "hidden_norms_cls": [ + 33.06986999511719, + 23368844.0, + 24521804.0, + 27230594.0, + 45358652.0 + ], + "hidden_norms_avg": [ + 1231.394287109375, + 38020996.0, + 40414464.0, + 42538228.0, + 53283596.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00015840897685848176, + 6.067092384043349e-10, + 3.4933173020945674e-10, + 3.3962907486362326e-10, + 3.42745692938351e-10 + ], + "bp_grad_F": [ + 0.3695656955242157, + 2.3579739263368538e-06, + 5.5092176154403205e-08, + 1.777137192959799e-08, + 1.4559810246339566e-08 + ], + "gamma_dfa": 0.0020704490307252854, + "gamma_dfa_per_layer": [ + 0.007164421491324902, + 0.0009055124246515334, + 0.0003939281450584531, + -0.00018206593813374639 + ], + "acc_eval": 0.203125, + "loss_eval": 2.1666717529296875, + "epoch": 45 + }, + { + "hidden_norms_cls": [ + 33.164432525634766, + 29174184.0, + 30089798.0, + 32367896.0, + 48892740.0 + ], + "hidden_norms_avg": [ + 1237.1142578125, + 37470168.0, + 40219512.0, + 42356984.0, + 52772660.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00024464691523462534, + 5.441591066634999e-10, + 3.6414088411262924e-10, + 3.320065611323031e-10, + 3.387692348866267e-10 + ], + "bp_grad_F": [ + 0.20688536763191223, + 4.435638402355835e-06, + 8.288016317692382e-08, + 1.4215274291018432e-08, + 1.4331217101926086e-08 + ], + "gamma_dfa": 0.0017712319058773573, + "gamma_dfa_per_layer": [ + 0.006079941056668758, + 0.0008057263912633061, + 0.00038248312193900347, + -0.0001832229463616386 + ], + "acc_eval": 0.1806640625, + "loss_eval": 2.173765182495117, + "epoch": 46 + }, + { + "hidden_norms_cls": [ + 33.24842071533203, + 33830332.0, + 34499304.0, + 36673048.0, + 51963528.0 + ], + "hidden_norms_avg": [ + 1242.226318359375, + 38073352.0, + 40806512.0, + 42901500.0, + 53437180.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002524294250179082, + 5.388652302151797e-10, + 3.6845529405304944e-10, + 3.3760005901939394e-10, + 3.4174474361492457e-10 + ], + "bp_grad_F": [ + 0.12659014761447906, + 2.532647158659529e-06, + 5.97612910269163e-08, + 1.3431267653629675e-08, + 1.3499856343912597e-08 + ], + "gamma_dfa": 0.0021579467684205156, + "gamma_dfa_per_layer": [ + 0.007928929291665554, + 0.00047530903248116374, + 0.00039780279621481895, + -0.0001702540466794744 + ], + "acc_eval": 0.2099609375, + "loss_eval": 2.161896228790283, + "epoch": 47 + }, + { + "hidden_norms_cls": [ + 33.338741302490234, + 19795148.0, + 21916948.0, + 27189936.0, + 43588712.0 + ], + "hidden_norms_avg": [ + 1246.7552490234375, + 38983196.0, + 41658504.0, + 43769460.0, + 54177164.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002599447325337678, + 6.2289540192495e-10, + 3.6807956682594067e-10, + 3.4037242468976103e-10, + 3.472495624379235e-10 + ], + "bp_grad_F": [ + 0.11130323261022568, + 6.972830760787474e-06, + 5.447330764241087e-08, + 1.4756569299834155e-08, + 1.4917180379825368e-08 + ], + "gamma_dfa": 0.0021847522366442718, + "gamma_dfa_per_layer": [ + 0.007594278547912836, + 0.0009666476398706436, + 0.0003712670295499265, + -0.00019318427075631917 + ], + "acc_eval": 0.1904296875, + "loss_eval": 2.1626296043395996, + "epoch": 48 + }, + { + "hidden_norms_cls": [ + 33.42045974731445, + 24898126.0, + 26009376.0, + 28516464.0, + 47394980.0 + ], + "hidden_norms_avg": [ + 1250.1884765625, + 39632580.0, + 42285708.0, + 44390100.0, + 55233620.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0001829452667152509, + 5.726033536213038e-10, + 3.5659752928296484e-10, + 3.292066619309253e-10, + 3.31846911061362e-10 + ], + "bp_grad_F": [ + 0.23442988097667694, + 3.2595114589639707e-06, + 1.0605135969399271e-07, + 1.4677630666426467e-08, + 1.445219677265186e-08 + ], + "gamma_dfa": 0.0021932553718215786, + "gamma_dfa_per_layer": [ + 0.007973091676831245, + 0.0006458763382397592, + 0.00032728860969655216, + -0.00017323513748124242 + ], + "acc_eval": 0.1923828125, + "loss_eval": 2.175835609436035, + "epoch": 49 + }, + { + "hidden_norms_cls": [ + 33.47613525390625, + 25550002.0, + 26776384.0, + 28507352.0, + 47268464.0 + ], + "hidden_norms_avg": [ + 1253.279541015625, + 39494300.0, + 42348356.0, + 44479368.0, + 55121100.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00012893354869447649, + 5.787402224122218e-10, + 3.341840137949248e-10, + 3.2919431069977634e-10, + 3.3201591476128556e-10 + ], + "bp_grad_F": [ + 0.10876749455928802, + 4.772132342623081e-06, + 7.90316647680811e-08, + 1.4042406348835357e-08, + 1.4153799909877307e-08 + ], + "gamma_dfa": 0.001727514507365413, + "gamma_dfa_per_layer": [ + 0.0062596434727311134, + 0.0004772312822751701, + 0.000357049866579473, + -0.0001838665921241045 + ], + "acc_eval": 0.1904296875, + "loss_eval": 2.164583444595337, + "epoch": 50 + }, + { + "hidden_norms_cls": [ + 33.52043914794922, + 28050636.0, + 29099278.0, + 31227182.0, + 49045872.0 + ], + "hidden_norms_avg": [ + 1255.7454833984375, + 40231008.0, + 42960368.0, + 45079784.0, + 55989944.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00015386792074423283, + 5.41802436249128e-10, + 3.373940293815991e-10, + 3.3321326253776817e-10, + 3.361276257329848e-10 + ], + "bp_grad_F": [ + 0.2656819522380829, + 2.482481932020164e-06, + 8.287289432473699e-08, + 1.3792345932017724e-08, + 1.3917338392843703e-08 + ], + "gamma_dfa": 0.001938080065883696, + "gamma_dfa_per_layer": [ + 0.006978346034884453, + 0.0005946386372670531, + 0.00036033656215295196, + -0.00018100097076967359 + ], + "acc_eval": 0.1962890625, + "loss_eval": 2.176473379135132, + "epoch": 51 + }, + { + "hidden_norms_cls": [ + 33.55419921875, + 25508434.0, + 26505134.0, + 29051658.0, + 47339916.0 + ], + "hidden_norms_avg": [ + 1257.7822265625, + 40388772.0, + 43162480.0, + 45217960.0, + 55966100.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00018102869216818362, + 5.443374639924059e-10, + 3.9548594954474936e-10, + 3.679410942591943e-10, + 3.717116892065775e-10 + ], + "bp_grad_F": [ + 0.28429684042930603, + 3.3983885714405915e-06, + 8.993236377818903e-08, + 1.4510244561449781e-08, + 1.4562592021150067e-08 + ], + "gamma_dfa": 0.001642326584260445, + "gamma_dfa_per_layer": [ + 0.006192460190504789, + 0.00019942648941650987, + 0.00037004181649535894, + -0.0001926221593748778 + ], + "acc_eval": 0.2001953125, + "loss_eval": 2.1743006706237793, + "epoch": 52 + }, + { + "hidden_norms_cls": [ + 33.58127212524414, + 28540952.0, + 29607362.0, + 31472670.0, + 49191068.0 + ], + "hidden_norms_avg": [ + 1259.4189453125, + 40550800.0, + 43364984.0, + 45506960.0, + 56297104.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00018638238543644547, + 5.299162220140374e-10, + 3.2437005859087265e-10, + 3.1624741714253446e-10, + 3.1920088794379353e-10 + ], + "bp_grad_F": [ + 0.21056434512138367, + 3.217361381757655e-06, + 5.1724125427199397e-08, + 1.390105897058902e-08, + 1.3961146017038573e-08 + ], + "gamma_dfa": 0.0018532263711676933, + "gamma_dfa_per_layer": [ + 0.006584585644304752, + 0.0006806463352404535, + 0.00032758695306256413, + -0.00017991344793699682 + ], + "acc_eval": 0.19921875, + "loss_eval": 2.167957067489624, + "epoch": 53 + }, + { + "hidden_norms_cls": [ + 33.60611343383789, + 27949698.0, + 28918304.0, + 31366946.0, + 48876736.0 + ], + "hidden_norms_avg": [ + 1260.694091796875, + 41001392.0, + 43744060.0, + 45853688.0, + 56636716.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00021861983987037092, + 5.396287861003657e-10, + 3.377726154329963e-10, + 3.201852394330018e-10, + 3.237338452866112e-10 + ], + "bp_grad_F": [ + 0.12415990233421326, + 1.6796100226201816e-06, + 6.823196940786147e-08, + 1.384469427989643e-08, + 1.3965705036866893e-08 + ], + "gamma_dfa": 0.0017868579998321366, + "gamma_dfa_per_layer": [ + 0.006317156832665205, + 0.0006519090966321528, + 0.0003622731601353735, + -0.00018390709010418504 + ], + "acc_eval": 0.1953125, + "loss_eval": 2.178144931793213, + "epoch": 54 + }, + { + "hidden_norms_cls": [ + 33.620914459228516, + 27607036.0, + 28622492.0, + 31370470.0, + 48645068.0 + ], + "hidden_norms_avg": [ + 1261.64208984375, + 40948476.0, + 43773536.0, + 45888992.0, + 56623536.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002112231741193682, + 5.336073249928575e-10, + 3.533912051878474e-10, + 3.186636232666018e-10, + 3.226934830458106e-10 + ], + "bp_grad_F": [ + 0.1486925333738327, + 3.3336793876515003e-06, + 9.466339889740993e-08, + 1.4568017014937595e-08, + 1.4240884915750485e-08 + ], + "gamma_dfa": 0.0018792969822243322, + "gamma_dfa_per_layer": [ + 0.006730419583618641, + 0.000605170032940805, + 0.0003661802038550377, + -0.00018458189151715487 + ], + "acc_eval": 0.1982421875, + "loss_eval": 2.172356367111206, + "epoch": 55 + }, + { + "hidden_norms_cls": [ + 33.63143539428711, + 28410040.0, + 29532286.0, + 31697578.0, + 48914624.0 + ], + "hidden_norms_avg": [ + 1262.2918701171875, + 40698620.0, + 43622904.0, + 45764712.0, + 56531748.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002270409750053659, + 5.479600106994553e-10, + 3.2429942065093087e-10, + 3.150284755282229e-10, + 3.179212726411862e-10 + ], + "bp_grad_F": [ + 0.10118122398853302, + 2.5195840862579644e-06, + 8.798109263352671e-08, + 1.3570585544186997e-08, + 1.3736383586149259e-08 + ], + "gamma_dfa": 0.0017859039653558284, + "gamma_dfa_per_layer": [ + 0.006382801569998264, + 0.0005453285994008183, + 0.00039217708399519324, + -0.00017669139197096229 + ], + "acc_eval": 0.1943359375, + "loss_eval": 2.1648902893066406, + "epoch": 56 + }, + { + "hidden_norms_cls": [ + 33.63875961303711, + 31114778.0, + 32184654.0, + 34278688.0, + 50894352.0 + ], + "hidden_norms_avg": [ + 1262.7132568359375, + 40792528.0, + 43713536.0, + 45850224.0, + 56615416.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00019923689251299948, + 5.368823718931992e-10, + 3.306654117185559e-10, + 3.113070912164062e-10, + 3.1328556415743947e-10 + ], + "bp_grad_F": [ + 0.3573400378227234, + 1.249598790309392e-05, + 1.6106491784739774e-07, + 3.301076745287901e-08, + 1.3328084413899433e-08 + ], + "gamma_dfa": 0.0016940671957854647, + "gamma_dfa_per_layer": [ + 0.006114102900028229, + 0.0004918627091683447, + 0.00032346430816687644, + -0.00015316113422159106 + ], + "acc_eval": 0.201171875, + "loss_eval": 2.1634349822998047, + "epoch": 57 + }, + { + "hidden_norms_cls": [ + 33.642520904541016, + 31868976.0, + 32779660.0, + 35005740.0, + 51305220.0 + ], + "hidden_norms_avg": [ + 1262.9461669921875, + 41087592.0, + 43928832.0, + 46043268.0, + 56839576.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002181618328904733, + 5.162476557352136e-10, + 3.3206276617292474e-10, + 3.1130611977125966e-10, + 3.1441516057384433e-10 + ], + "bp_grad_F": [ + 0.12289460748434067, + 1.5301338862627745e-05, + 8.779379356838035e-08, + 1.4331889452989799e-08, + 1.318165843144925e-08 + ], + "gamma_dfa": 0.001529279034002684, + "gamma_dfa_per_layer": [ + 0.00539398193359375, + 0.0005496743833646178, + 0.00034435215638950467, + -0.00017089233733713627 + ], + "acc_eval": 0.2021484375, + "loss_eval": 2.164745807647705, + "epoch": 58 + }, + { + "hidden_norms_cls": [ + 33.6442756652832, + 31315430.0, + 32160978.0, + 34340796.0, + 51077484.0 + ], + "hidden_norms_avg": [ + 1263.062744140625, + 41015680.0, + 43882708.0, + 46002096.0, + 56770896.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00021884885791223496, + 5.177114292820306e-10, + 3.350683897007656e-10, + 3.101042200803761e-10, + 3.12167347527037e-10 + ], + "bp_grad_F": [ + 0.11127890646457672, + 1.1515544429130387e-05, + 8.542571805492116e-08, + 1.319963871537766e-08, + 1.3264477516372608e-08 + ], + "gamma_dfa": 0.0016308917693095282, + "gamma_dfa_per_layer": [ + 0.005676542408764362, + 0.0006629846757277846, + 0.0003523063496686518, + -0.0001682663569226861 + ], + "acc_eval": 0.19921875, + "loss_eval": 2.1635866165161133, + "epoch": 59 + }, + { + "hidden_norms_cls": [ + 33.643985748291016, + 31097308.0, + 32052318.0, + 34155804.0, + 50803736.0 + ], + "hidden_norms_avg": [ + 1263.064208984375, + 40997444.0, + 43870400.0, + 45990872.0, + 56755540.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00021703331731259823, + 5.1921633659191e-10, + 3.35882821556055e-10, + 3.102589296588576e-10, + 3.1300673164480486e-10 + ], + "bp_grad_F": [ + 0.1889268010854721, + 1.0328590178687591e-05, + 1.1096560825762936e-07, + 1.3215729843807367e-08, + 1.3306935997547953e-08 + ], + "gamma_dfa": 0.0016646812255203258, + "gamma_dfa_per_layer": [ + 0.005869406741112471, + 0.0006016636616550386, + 0.0003579354961402714, + -0.00017028099682647735 + ], + "acc_eval": 0.2021484375, + "loss_eval": 2.163691520690918, + "epoch": 60 + } + ] +}
\ No newline at end of file diff --git a/results/snapshot_vit_v1/snapshot_vit_s42.json b/results/snapshot_vit_v1/snapshot_vit_s42.json new file mode 100644 index 0000000..569af46 --- /dev/null +++ b/results/snapshot_vit_v1/snapshot_vit_s42.json @@ -0,0 +1,4594 @@ +{ + "config": { + "output_dir": "results/snapshot_vit_v1", + "epochs": 60, + "lr": 0.001, + "wd": 0.05, + "seed": 42, + "depth": 4, + "d_model": 128, + "n_heads": 4 + }, + "depth": 4, + "d_model": 128, + "architecture": "ViTMini", + "bp_log": [ + { + "hidden_norms_cls": [ + 0.31052109599113464, + 3.2536301612854004, + 4.7681732177734375, + 6.129599571228027, + 7.124356746673584 + ], + "hidden_norms_avg": [ + 6.008513927459717, + 7.034290313720703, + 7.942849636077881, + 8.915729522705078, + 9.635361671447754 + ], + "bp_grad_per_sample_l2_med": [ + 0.004442029166966677, + 0.0012008437188342214, + 0.0009976272704079747, + 0.0009104337659664452, + 0.0008558662957511842 + ], + "bp_grad_F": [ + 0.18535660207271576, + 0.03968670591711998, + 0.032604776322841644, + 0.029427478089928627, + 0.02746850810945034 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.1376953125, + "loss_eval": 2.35188627243042, + "epoch": 0 + }, + { + "hidden_norms_cls": [ + 0.511958122253418, + 12.453446388244629, + 18.366989135742188, + 25.54116439819336, + 31.80699348449707 + ], + "hidden_norms_avg": [ + 4.10694694519043, + 10.166566848754883, + 14.66998291015625, + 19.93342399597168, + 25.32619857788086 + ], + "bp_grad_per_sample_l2_med": [ + 0.0019771556835621595, + 0.00022324378369376063, + 0.00016308785416185856, + 0.00014621105219703168, + 0.0001451292191632092 + ], + "bp_grad_F": [ + 0.11066891252994537, + 0.01034498494118452, + 0.006372526753693819, + 0.0052248528227210045, + 0.005017552524805069 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.4501953125, + "loss_eval": 1.5154438018798828, + "epoch": 1 + }, + { + "hidden_norms_cls": [ + 0.6518111824989319, + 13.549692153930664, + 20.607715606689453, + 27.90418243408203, + 32.70498275756836 + ], + "hidden_norms_avg": [ + 4.0909905433654785, + 11.942686080932617, + 17.39116096496582, + 22.200014114379883, + 26.230010986328125 + ], + "bp_grad_per_sample_l2_med": [ + 0.0017800139030441642, + 0.00026728332159109414, + 0.0001720335567370057, + 0.0001423172070644796, + 0.00013423117343336344 + ], + "bp_grad_F": [ + 0.08781559765338898, + 0.010987765155732632, + 0.0064147054217755795, + 0.0050079175271093845, + 0.004637593869119883 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.494140625, + "loss_eval": 1.3824832439422607, + "epoch": 2 + }, + { + "hidden_norms_cls": [ + 0.755299985408783, + 11.224863052368164, + 19.053363800048828, + 27.303081512451172, + 33.078182220458984 + ], + "hidden_norms_avg": [ + 4.027151584625244, + 12.449769973754883, + 17.769508361816406, + 22.97556495666504, + 27.065759658813477 + ], + "bp_grad_per_sample_l2_med": [ + 0.0018998769810423255, + 0.00030996452551335096, + 0.00018013815861195326, + 0.0001340982416877523, + 0.00012205552775412798 + ], + "bp_grad_F": [ + 0.09037011116743088, + 0.013488140888512135, + 0.007346263621002436, + 0.0051797907799482346, + 0.004589984193444252 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.5439453125, + "loss_eval": 1.2540550231933594, + "epoch": 3 + }, + { + "hidden_norms_cls": [ + 0.844818651676178, + 13.648336410522461, + 20.277976989746094, + 28.240915298461914, + 33.095726013183594 + ], + "hidden_norms_avg": [ + 4.036877632141113, + 12.003630638122559, + 17.05451011657715, + 22.15869903564453, + 25.917491912841797 + ], + "bp_grad_per_sample_l2_med": [ + 0.0016167981084436178, + 0.00028610811568796635, + 0.0001777650322765112, + 0.00012878721463494003, + 0.0001112063619075343 + ], + "bp_grad_F": [ + 0.07410340011119843, + 0.012637714855372906, + 0.007633093744516373, + 0.005242784973233938, + 0.004431390203535557 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.603515625, + "loss_eval": 1.1045944690704346, + "epoch": 4 + }, + { + "hidden_norms_cls": [ + 0.9051892161369324, + 12.476008415222168, + 19.24710464477539, + 27.6123046875, + 32.35737609863281 + ], + "hidden_norms_avg": [ + 4.138596057891846, + 12.409324645996094, + 17.427003860473633, + 22.245147705078125, + 25.471385955810547 + ], + "bp_grad_per_sample_l2_med": [ + 0.0016044765943661332, + 0.00030248358962126076, + 0.00018807635933626443, + 0.00013626758300233632, + 0.00011398641800042242 + ], + "bp_grad_F": [ + 0.08322811126708984, + 0.01390017382800579, + 0.008505746722221375, + 0.005747759714722633, + 0.004776888992637396 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.603515625, + "loss_eval": 1.1296963691711426, + "epoch": 5 + }, + { + "hidden_norms_cls": [ + 0.9687634110450745, + 11.625859260559082, + 18.38014793395996, + 26.869504928588867, + 31.455507278442383 + ], + "hidden_norms_avg": [ + 4.263001441955566, + 13.1234769821167, + 17.853057861328125, + 22.476547241210938, + 25.29656410217285 + ], + "bp_grad_per_sample_l2_med": [ + 0.0015340449754148722, + 0.00033552697277627885, + 0.00021536082203965634, + 0.00015067037020344287, + 0.00012220896314829588 + ], + "bp_grad_F": [ + 0.07571946829557419, + 0.015132292173802853, + 0.009512492455542088, + 0.006373303942382336, + 0.005029390100389719 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.603515625, + "loss_eval": 1.0963702201843262, + "epoch": 6 + }, + { + "hidden_norms_cls": [ + 1.0029383897781372, + 12.687585830688477, + 19.734830856323242, + 28.174379348754883, + 32.5982780456543 + ], + "hidden_norms_avg": [ + 4.357974052429199, + 13.182960510253906, + 17.720989227294922, + 21.832895278930664, + 24.934776306152344 + ], + "bp_grad_per_sample_l2_med": [ + 0.0013370191445574164, + 0.00026788320974446833, + 0.00018144956266041845, + 0.00013219293032307178, + 0.00010196284711128101 + ], + "bp_grad_F": [ + 0.06996501982212067, + 0.013071151450276375, + 0.008477694354951382, + 0.0061036222614347935, + 0.004761465825140476 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.6416015625, + "loss_eval": 0.9737913012504578, + "epoch": 7 + }, + { + "hidden_norms_cls": [ + 1.043707013130188, + 12.487404823303223, + 18.845643997192383, + 26.870864868164062, + 31.145292282104492 + ], + "hidden_norms_avg": [ + 4.454991340637207, + 13.503378868103027, + 18.694488525390625, + 23.029804229736328, + 25.810880661010742 + ], + "bp_grad_per_sample_l2_med": [ + 0.0013130803126841784, + 0.00029930632445029914, + 0.0002022052649408579, + 0.0001383959170198068, + 0.00010439592006150633 + ], + "bp_grad_F": [ + 0.0743240937590599, + 0.015402271412312984, + 0.009977605193853378, + 0.006877230014652014, + 0.005202346481382847 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.666015625, + "loss_eval": 0.9317076206207275, + "epoch": 8 + }, + { + "hidden_norms_cls": [ + 1.0887501239776611, + 12.662396430969238, + 18.622724533081055, + 26.599454879760742, + 31.788663864135742 + ], + "hidden_norms_avg": [ + 4.529710292816162, + 13.683384895324707, + 18.47463607788086, + 22.395898818969727, + 25.244558334350586 + ], + "bp_grad_per_sample_l2_med": [ + 0.0012017677072435617, + 0.0002972102665808052, + 0.00019775003602262586, + 0.00013734954700339586, + 9.880279685603455e-05 + ], + "bp_grad_F": [ + 0.0661192312836647, + 0.01516024675220251, + 0.010001113638281822, + 0.006893018260598183, + 0.0051134442910552025 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.6669921875, + "loss_eval": 0.9026263952255249, + "epoch": 9 + }, + { + "hidden_norms_cls": [ + 1.1163448095321655, + 12.711283683776855, + 18.790128707885742, + 26.846359252929688, + 31.663158416748047 + ], + "hidden_norms_avg": [ + 4.502295970916748, + 13.801639556884766, + 18.488975524902344, + 22.378704071044922, + 25.22056007385254 + ], + "bp_grad_per_sample_l2_med": [ + 0.0011062510311603546, + 0.00029147896566428244, + 0.0001970102748600766, + 0.00013174059859011322, + 0.0001037597976392135 + ], + "bp_grad_F": [ + 0.06217500939965248, + 0.014721231535077095, + 0.00993143580853939, + 0.006723989732563496, + 0.005166618153452873 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.6845703125, + "loss_eval": 0.8660677671432495, + "epoch": 10 + }, + { + "hidden_norms_cls": [ + 1.1505078077316284, + 12.160571098327637, + 18.15353012084961, + 25.762012481689453, + 31.478300094604492 + ], + "hidden_norms_avg": [ + 4.5605149269104, + 13.918573379516602, + 18.88881492614746, + 22.79344940185547, + 25.800094604492188 + ], + "bp_grad_per_sample_l2_med": [ + 0.0009689349099062383, + 0.0002486997109372169, + 0.0001736732665449381, + 0.00011473154881969094, + 8.573567902203649e-05 + ], + "bp_grad_F": [ + 0.06455213576555252, + 0.01547171175479889, + 0.010573537088930607, + 0.0071961241774261, + 0.005212985444813967 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.71484375, + "loss_eval": 0.812907874584198, + "epoch": 11 + }, + { + "hidden_norms_cls": [ + 1.1844152212142944, + 11.332664489746094, + 16.57625961303711, + 23.82307243347168, + 28.492321014404297 + ], + "hidden_norms_avg": [ + 4.602160453796387, + 13.876313209533691, + 17.891036987304688, + 21.410982131958008, + 23.97108268737793 + ], + "bp_grad_per_sample_l2_med": [ + 0.0009290817542932928, + 0.0002613075776025653, + 0.0001877332106232643, + 0.0001262558507733047, + 9.168039832729846e-05 + ], + "bp_grad_F": [ + 0.06392678618431091, + 0.01674485206604004, + 0.011727217584848404, + 0.007738997228443623, + 0.005730107892304659 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7197265625, + "loss_eval": 0.7641609907150269, + "epoch": 12 + }, + { + "hidden_norms_cls": [ + 1.2184560298919678, + 12.450925827026367, + 18.076860427856445, + 25.162586212158203, + 30.365047454833984 + ], + "hidden_norms_avg": [ + 4.7080278396606445, + 13.851634979248047, + 18.517131805419922, + 22.074214935302734, + 24.72054100036621 + ], + "bp_grad_per_sample_l2_med": [ + 0.0009277731878682971, + 0.00024563330225646496, + 0.00017736255540512502, + 0.00012202116340631619, + 8.748921391088516e-05 + ], + "bp_grad_F": [ + 0.06389915198087692, + 0.016115156933665276, + 0.0110858753323555, + 0.0076499818824231625, + 0.005541808437556028 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.697265625, + "loss_eval": 0.8037601113319397, + "epoch": 13 + }, + { + "hidden_norms_cls": [ + 1.243064045906067, + 11.932015419006348, + 17.215856552124023, + 24.764291763305664, + 29.49314308166504 + ], + "hidden_norms_avg": [ + 4.651937961578369, + 13.890728950500488, + 18.2476749420166, + 21.80103302001953, + 24.145692825317383 + ], + "bp_grad_per_sample_l2_med": [ + 0.0009111211402341723, + 0.000253814272582531, + 0.00018247529806103557, + 0.00011639296280918643, + 8.364167297258973e-05 + ], + "bp_grad_F": [ + 0.06324896216392517, + 0.016394753009080887, + 0.01153184287250042, + 0.007614497095346451, + 0.0056136660277843475 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.728515625, + "loss_eval": 0.7384279370307922, + "epoch": 14 + }, + { + "hidden_norms_cls": [ + 1.2652677297592163, + 11.532588005065918, + 16.7224178314209, + 23.326444625854492, + 28.03766441345215 + ], + "hidden_norms_avg": [ + 4.652615547180176, + 13.981542587280273, + 17.984512329101562, + 21.317432403564453, + 23.233760833740234 + ], + "bp_grad_per_sample_l2_med": [ + 0.0008032417972572148, + 0.00022395161795429885, + 0.00016904002404771745, + 0.0001078559216693975, + 7.841538899810985e-05 + ], + "bp_grad_F": [ + 0.06386830657720566, + 0.016731705516576767, + 0.012135310098528862, + 0.008108492009341717, + 0.006046997383236885 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7333984375, + "loss_eval": 0.7387630939483643, + "epoch": 15 + }, + { + "hidden_norms_cls": [ + 1.2944419384002686, + 11.34481430053711, + 16.45362091064453, + 22.63861846923828, + 28.011232376098633 + ], + "hidden_norms_avg": [ + 4.738389015197754, + 13.769956588745117, + 17.63072395324707, + 21.14189338684082, + 23.455045700073242 + ], + "bp_grad_per_sample_l2_med": [ + 0.0007927753031253815, + 0.0002345799293834716, + 0.00016934113227762282, + 0.00011508075840538368, + 7.966109114931896e-05 + ], + "bp_grad_F": [ + 0.0655454769730568, + 0.01815187744796276, + 0.013005951419472694, + 0.008820701390504837, + 0.006184164900332689 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7255859375, + "loss_eval": 0.7302451133728027, + "epoch": 16 + }, + { + "hidden_norms_cls": [ + 1.3085625171661377, + 11.323317527770996, + 15.866230964660645, + 21.746976852416992, + 27.081127166748047 + ], + "hidden_norms_avg": [ + 4.687533855438232, + 13.471115112304688, + 17.396968841552734, + 20.327377319335938, + 22.58539581298828 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006625292589887977, + 0.00019486426026560366, + 0.00014659677981399, + 9.787825547391549e-05, + 6.321907130768523e-05 + ], + "bp_grad_F": [ + 0.07393868267536163, + 0.020178642123937607, + 0.014442694373428822, + 0.009784683585166931, + 0.00656729843467474 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7412109375, + "loss_eval": 0.7063955068588257, + "epoch": 17 + }, + { + "hidden_norms_cls": [ + 1.3433176279067993, + 10.762125968933105, + 15.129413604736328, + 21.06867790222168, + 26.1944522857666 + ], + "hidden_norms_avg": [ + 4.709607124328613, + 13.69739055633545, + 17.006061553955078, + 19.97213363647461, + 22.212177276611328 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006978239398449659, + 0.00021396303782239556, + 0.00016030693950597197, + 0.00010856003791559488, + 6.917239079484716e-05 + ], + "bp_grad_F": [ + 0.06653428077697754, + 0.019508732482790947, + 0.014456766657531261, + 0.010107609443366528, + 0.006681134924292564 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.75, + "loss_eval": 0.6990396976470947, + "epoch": 18 + }, + { + "hidden_norms_cls": [ + 1.3607734441757202, + 10.718966484069824, + 15.383923530578613, + 21.34387969970703, + 26.49329376220703 + ], + "hidden_norms_avg": [ + 4.7122907638549805, + 13.804581642150879, + 17.532636642456055, + 20.843929290771484, + 23.01592254638672 + ], + "bp_grad_per_sample_l2_med": [ + 0.0007053805165924132, + 0.0002122409496223554, + 0.00016182110994122922, + 0.00011085433652624488, + 7.204789289971814e-05 + ], + "bp_grad_F": [ + 0.06105736270546913, + 0.017779415473341942, + 0.013269647024571896, + 0.009227588772773743, + 0.006410240661352873 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7763671875, + "loss_eval": 0.6498311758041382, + "epoch": 19 + }, + { + "hidden_norms_cls": [ + 1.3752704858779907, + 10.530582427978516, + 14.803340911865234, + 20.217910766601562, + 25.20522689819336 + ], + "hidden_norms_avg": [ + 4.7595720291137695, + 13.601245880126953, + 17.491943359375, + 20.960437774658203, + 22.910301208496094 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006476694834418595, + 0.00020479912927839905, + 0.0001524801627965644, + 0.00010630583710735664, + 6.906194903422147e-05 + ], + "bp_grad_F": [ + 0.06572327762842178, + 0.019566379487514496, + 0.01484121847897768, + 0.010426007211208344, + 0.006943442393094301 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.763671875, + "loss_eval": 0.6451045274734497, + "epoch": 20 + }, + { + "hidden_norms_cls": [ + 1.394100546836853, + 9.805717468261719, + 14.102628707885742, + 19.18745231628418, + 24.205224990844727 + ], + "hidden_norms_avg": [ + 4.746951580047607, + 13.343804359436035, + 16.609134674072266, + 19.829851150512695, + 21.580671310424805 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006192427244968712, + 0.0002108582411892712, + 0.0001636892993701622, + 0.0001098570428439416, + 6.864540046080947e-05 + ], + "bp_grad_F": [ + 0.06809430569410324, + 0.02151457779109478, + 0.01606965810060501, + 0.011247958056628704, + 0.007375919260084629 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7705078125, + "loss_eval": 0.6522111892700195, + "epoch": 21 + }, + { + "hidden_norms_cls": [ + 1.4160058498382568, + 9.728543281555176, + 13.389554977416992, + 18.542705535888672, + 23.791994094848633 + ], + "hidden_norms_avg": [ + 4.738738536834717, + 13.15449333190918, + 16.28533172607422, + 19.581254959106445, + 21.488550186157227 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006278376094996929, + 0.00020051853789482266, + 0.00015803576388861984, + 0.0001043542506522499, + 6.271857273532078e-05 + ], + "bp_grad_F": [ + 0.07535671442747116, + 0.02472539246082306, + 0.01880054548382759, + 0.012977317906916142, + 0.007879889570176601 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.767578125, + "loss_eval": 0.6732022762298584, + "epoch": 22 + }, + { + "hidden_norms_cls": [ + 1.4301073551177979, + 9.656251907348633, + 13.501574516296387, + 18.518707275390625, + 23.49370574951172 + ], + "hidden_norms_avg": [ + 4.7338972091674805, + 12.90231704711914, + 15.899849891662598, + 18.982860565185547, + 20.785310745239258 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005697127780877054, + 0.00018775691569317132, + 0.00014216902491170913, + 9.182754729408771e-05, + 5.984708695905283e-05 + ], + "bp_grad_F": [ + 0.07130222022533417, + 0.023520665243268013, + 0.017786188051104546, + 0.012143870815634727, + 0.007813607342541218 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7763671875, + "loss_eval": 0.6380050182342529, + "epoch": 23 + }, + { + "hidden_norms_cls": [ + 1.4623057842254639, + 10.028787612915039, + 13.687016487121582, + 18.426584243774414, + 23.602413177490234 + ], + "hidden_norms_avg": [ + 4.807516574859619, + 12.739899635314941, + 16.150466918945312, + 19.439754486083984, + 21.316791534423828 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005416463245637715, + 0.00019103426893707365, + 0.00014568559709005058, + 9.899997530737892e-05, + 6.29485584795475e-05 + ], + "bp_grad_F": [ + 0.07191846519708633, + 0.023787740617990494, + 0.01782786287367344, + 0.012350695207715034, + 0.007955120876431465 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7841796875, + "loss_eval": 0.6439570784568787, + "epoch": 24 + }, + { + "hidden_norms_cls": [ + 1.4565976858139038, + 9.294268608093262, + 12.656200408935547, + 17.385971069335938, + 21.959299087524414 + ], + "hidden_norms_avg": [ + 4.774696350097656, + 12.394432067871094, + 15.230389595031738, + 18.10796356201172, + 19.88006591796875 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005462650442495942, + 0.00019751323270611465, + 0.00015244862879626453, + 0.00010562407260295004, + 6.40779035165906e-05 + ], + "bp_grad_F": [ + 0.07698184996843338, + 0.026412304490804672, + 0.02022000029683113, + 0.013991317711770535, + 0.008763212710618973 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7568359375, + "loss_eval": 0.6531677842140198, + "epoch": 25 + }, + { + "hidden_norms_cls": [ + 1.4813296794891357, + 9.043482780456543, + 12.308734893798828, + 16.513948440551758, + 21.061283111572266 + ], + "hidden_norms_avg": [ + 4.7356462478637695, + 12.33294677734375, + 15.104657173156738, + 17.771398544311523, + 19.526002883911133 + ], + "bp_grad_per_sample_l2_med": [ + 0.000557436200324446, + 0.0002021614054683596, + 0.0001567570143379271, + 0.00010984922846546397, + 6.634221062995493e-05 + ], + "bp_grad_F": [ + 0.07777652144432068, + 0.027695877477526665, + 0.021396547555923462, + 0.015052770264446735, + 0.009372876025736332 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7744140625, + "loss_eval": 0.6539995074272156, + "epoch": 26 + }, + { + "hidden_norms_cls": [ + 1.4828969240188599, + 8.677760124206543, + 12.009501457214355, + 16.262908935546875, + 20.79404067993164 + ], + "hidden_norms_avg": [ + 4.744379043579102, + 12.125540733337402, + 14.89413833618164, + 17.47431182861328, + 18.953710556030273 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005026232684031129, + 0.00019721532589755952, + 0.00014348917466122657, + 0.00010225686855847016, + 5.8645866374718025e-05 + ], + "bp_grad_F": [ + 0.0830698236823082, + 0.03011341392993927, + 0.023107260465621948, + 0.016188694164156914, + 0.00968033168464899 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7744140625, + "loss_eval": 0.6444424390792847, + "epoch": 27 + }, + { + "hidden_norms_cls": [ + 1.5019822120666504, + 9.1107816696167, + 12.172623634338379, + 16.5975341796875, + 20.453947067260742 + ], + "hidden_norms_avg": [ + 4.741573333740234, + 12.045503616333008, + 14.836026191711426, + 17.666833877563477, + 19.175966262817383 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005040269461460412, + 0.00019225970027036965, + 0.0001496450131526217, + 0.00010478955664439127, + 6.137428135843948e-05 + ], + "bp_grad_F": [ + 0.07822537422180176, + 0.02863621897995472, + 0.022284477949142456, + 0.01552330981940031, + 0.009474868886172771 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7822265625, + "loss_eval": 0.6074927449226379, + "epoch": 28 + }, + { + "hidden_norms_cls": [ + 1.5009087324142456, + 8.223862648010254, + 11.225082397460938, + 15.531359672546387, + 19.848068237304688 + ], + "hidden_norms_avg": [ + 4.69788122177124, + 11.848491668701172, + 14.342337608337402, + 16.882020950317383, + 18.197385787963867 + ], + "bp_grad_per_sample_l2_med": [ + 0.0004685261519625783, + 0.00017912790644913912, + 0.00013862684136256576, + 9.223359666066244e-05, + 5.140554640092887e-05 + ], + "bp_grad_F": [ + 0.09127916395664215, + 0.03443564847111702, + 0.026544470340013504, + 0.01809101179242134, + 0.010493002831935883 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7783203125, + "loss_eval": 0.6329880952835083, + "epoch": 29 + }, + { + "hidden_norms_cls": [ + 1.509100079536438, + 8.238924980163574, + 10.995248794555664, + 14.927000045776367, + 19.032428741455078 + ], + "hidden_norms_avg": [ + 4.678654193878174, + 11.513160705566406, + 14.12202262878418, + 16.704364776611328, + 18.131078720092773 + ], + "bp_grad_per_sample_l2_med": [ + 0.0004594121128320694, + 0.00018182865460403264, + 0.000145129015436396, + 0.00010142508108401671, + 5.208289803704247e-05 + ], + "bp_grad_F": [ + 0.09080124646425247, + 0.03489048779010773, + 0.027099598199129105, + 0.018760183826088905, + 0.010796112939715385 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7822265625, + "loss_eval": 0.625268816947937, + "epoch": 30 + }, + { + "hidden_norms_cls": [ + 1.503785252571106, + 7.896090030670166, + 10.495330810546875, + 14.530640602111816, + 18.728647232055664 + ], + "hidden_norms_avg": [ + 4.70166540145874, + 11.420204162597656, + 13.774767875671387, + 16.2652530670166, + 17.631481170654297 + ], + "bp_grad_per_sample_l2_med": [ + 0.00043488750816322863, + 0.00017315168224740773, + 0.0001335645210929215, + 8.99006481631659e-05, + 5.126056930748746e-05 + ], + "bp_grad_F": [ + 0.10076858103275299, + 0.039271026849746704, + 0.0300181582570076, + 0.020545953884720802, + 0.011564387008547783 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.771484375, + "loss_eval": 0.6479744911193848, + "epoch": 31 + }, + { + "hidden_norms_cls": [ + 1.505173921585083, + 7.728392124176025, + 10.420252799987793, + 14.133453369140625, + 18.110719680786133 + ], + "hidden_norms_avg": [ + 4.7109880447387695, + 11.242951393127441, + 13.997397422790527, + 16.60598373413086, + 17.915313720703125 + ], + "bp_grad_per_sample_l2_med": [ + 0.00040974540752358735, + 0.0001792571711121127, + 0.00014464404375758022, + 9.883226448437199e-05, + 5.038536983192898e-05 + ], + "bp_grad_F": [ + 0.09923475980758667, + 0.039516862481832504, + 0.03056514263153076, + 0.021311543881893158, + 0.011838918551802635 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7880859375, + "loss_eval": 0.6011433601379395, + "epoch": 32 + }, + { + "hidden_norms_cls": [ + 1.5013800859451294, + 7.8755106925964355, + 10.452071189880371, + 14.013816833496094, + 17.622900009155273 + ], + "hidden_norms_avg": [ + 4.662457466125488, + 11.018245697021484, + 13.179399490356445, + 15.56103515625, + 16.739452362060547 + ], + "bp_grad_per_sample_l2_med": [ + 0.00033098450512625277, + 0.0001421417691744864, + 0.00011021740647265688, + 7.645406731171533e-05, + 4.017115134047344e-05 + ], + "bp_grad_F": [ + 0.09463275969028473, + 0.03948793560266495, + 0.03067062795162201, + 0.021538278087973595, + 0.012101472355425358 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.8076171875, + "loss_eval": 0.5846524238586426, + "epoch": 33 + }, + { + "hidden_norms_cls": [ + 1.5032685995101929, + 7.549060344696045, + 10.04562759399414, + 13.609556198120117, + 17.36180877685547 + ], + "hidden_norms_avg": [ + 4.688360214233398, + 11.03510856628418, + 13.660865783691406, + 16.024188995361328, + 17.242145538330078 + ], + "bp_grad_per_sample_l2_med": [ + 0.0003352685598656535, + 0.0001500560319982469, + 0.000116633229481522, + 8.394405449507758e-05, + 4.0652244933880866e-05 + ], + "bp_grad_F": [ + 0.10540100187063217, + 0.044482193887233734, + 0.03421863541007042, + 0.023781608790159225, + 0.012753129936754704 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7861328125, + "loss_eval": 0.5960580110549927, + "epoch": 34 + }, + { + "hidden_norms_cls": [ + 1.5032752752304077, + 7.387601375579834, + 9.652260780334473, + 13.042936325073242, + 16.688335418701172 + ], + "hidden_norms_avg": [ + 4.644099235534668, + 10.739524841308594, + 13.165095329284668, + 15.435381889343262, + 16.59172248840332 + ], + "bp_grad_per_sample_l2_med": [ + 0.0003447248600423336, + 0.0001518868375569582, + 0.00012246235564816743, + 8.41514120111242e-05, + 4.193704080535099e-05 + ], + "bp_grad_F": [ + 0.11063266545534134, + 0.04730832949280739, + 0.03743334487080574, + 0.02577192150056362, + 0.01377912238240242 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.78515625, + "loss_eval": 0.652167797088623, + "epoch": 35 + }, + { + "hidden_norms_cls": [ + 1.4935848712921143, + 7.262828826904297, + 9.431328773498535, + 12.742668151855469, + 16.348758697509766 + ], + "hidden_norms_avg": [ + 4.6145453453063965, + 10.670759201049805, + 12.641580581665039, + 14.64714241027832, + 15.685794830322266 + ], + "bp_grad_per_sample_l2_med": [ + 0.0003325030847918242, + 0.00014367286348715425, + 0.00011324919614708051, + 7.871988782426342e-05, + 4.058879494550638e-05 + ], + "bp_grad_F": [ + 0.11436768621206284, + 0.04893247410655022, + 0.03840672969818115, + 0.02707456797361374, + 0.014087510295212269 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7978515625, + "loss_eval": 0.6156834363937378, + "epoch": 36 + }, + { + "hidden_norms_cls": [ + 1.4897189140319824, + 7.156031608581543, + 9.365946769714355, + 12.499292373657227, + 15.942588806152344 + ], + "hidden_norms_avg": [ + 4.635710716247559, + 10.602685928344727, + 12.637743949890137, + 14.74621868133545, + 15.834792137145996 + ], + "bp_grad_per_sample_l2_med": [ + 0.000299928622553125, + 0.00013245265290606767, + 0.00010241761629004031, + 7.247657049447298e-05, + 3.780105180339888e-05 + ], + "bp_grad_F": [ + 0.11267601698637009, + 0.0501401461660862, + 0.03941003605723381, + 0.0279407799243927, + 0.014723201282322407 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.80078125, + "loss_eval": 0.5903886556625366, + "epoch": 37 + }, + { + "hidden_norms_cls": [ + 1.4791479110717773, + 7.027911186218262, + 9.0321044921875, + 12.123476028442383, + 15.987238883972168 + ], + "hidden_norms_avg": [ + 4.5436177253723145, + 10.157598495483398, + 12.115620613098145, + 14.023022651672363, + 15.130277633666992 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002464659046381712, + 0.00011603213351918384, + 9.401622082805261e-05, + 6.44488463876769e-05, + 2.8129303245805204e-05 + ], + "bp_grad_F": [ + 0.1319781094789505, + 0.05854209512472153, + 0.04639327526092529, + 0.03260229155421257, + 0.01578359492123127 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7841796875, + "loss_eval": 0.6532921195030212, + "epoch": 38 + }, + { + "hidden_norms_cls": [ + 1.4722236394882202, + 6.847982406616211, + 9.039284706115723, + 11.94421672821045, + 15.479789733886719 + ], + "hidden_norms_avg": [ + 4.536457538604736, + 10.0848388671875, + 11.944480895996094, + 13.931117057800293, + 15.018562316894531 + ], + "bp_grad_per_sample_l2_med": [ + 0.00023636725381948054, + 0.00011084756260970607, + 8.837366476655006e-05, + 6.370043411152437e-05, + 3.0101380616542883e-05 + ], + "bp_grad_F": [ + 0.1352969855070114, + 0.06069062277674675, + 0.04737559333443642, + 0.03317096084356308, + 0.016367904841899872 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7783203125, + "loss_eval": 0.6675163507461548, + "epoch": 39 + }, + { + "hidden_norms_cls": [ + 1.4741134643554688, + 6.7887864112854, + 8.833945274353027, + 11.777705192565918, + 15.091635704040527 + ], + "hidden_norms_avg": [ + 4.546427249908447, + 10.082904815673828, + 11.992964744567871, + 14.035684585571289, + 15.085047721862793 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002497230307199061, + 0.00011889061715919524, + 9.627552208257839e-05, + 6.80557859595865e-05, + 3.187589391018264e-05 + ], + "bp_grad_F": [ + 0.13283543288707733, + 0.06097453087568283, + 0.047521013766527176, + 0.033583469688892365, + 0.016502397134900093 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7998046875, + "loss_eval": 0.6332878470420837, + "epoch": 40 + }, + { + "hidden_norms_cls": [ + 1.4732073545455933, + 6.690948009490967, + 8.623305320739746, + 11.383458137512207, + 14.868409156799316 + ], + "hidden_norms_avg": [ + 4.510478973388672, + 9.772372245788574, + 11.55829906463623, + 13.379657745361328, + 14.343344688415527 + ], + "bp_grad_per_sample_l2_med": [ + 0.00019870084361173213, + 0.00010116927296621725, + 8.329773845616728e-05, + 6.15280878264457e-05, + 2.8435944841476157e-05 + ], + "bp_grad_F": [ + 0.13917136192321777, + 0.06636927276849747, + 0.05231080949306488, + 0.0366261750459671, + 0.017451239749789238 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.8017578125, + "loss_eval": 0.6271109580993652, + "epoch": 41 + }, + { + "hidden_norms_cls": [ + 1.4620546102523804, + 6.489831924438477, + 8.487679481506348, + 11.202116966247559, + 14.761962890625 + ], + "hidden_norms_avg": [ + 4.5009870529174805, + 9.717012405395508, + 11.480990409851074, + 13.235471725463867, + 14.178875923156738 + ], + "bp_grad_per_sample_l2_med": [ + 0.00015383858408313245, + 8.131976937875152e-05, + 6.472255336120725e-05, + 4.378501034807414e-05, + 1.9541592337191105e-05 + ], + "bp_grad_F": [ + 0.1546960175037384, + 0.07402726262807846, + 0.058267489075660706, + 0.041084613651037216, + 0.018521569669246674 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.794921875, + "loss_eval": 0.6767467260360718, + "epoch": 42 + }, + { + "hidden_norms_cls": [ + 1.4651520252227783, + 6.564362525939941, + 8.296653747558594, + 11.014183044433594, + 14.418519973754883 + ], + "hidden_norms_avg": [ + 4.483984470367432, + 9.593335151672363, + 11.23758316040039, + 13.036303520202637, + 14.015531539916992 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002037636877503246, + 0.00010235400259261951, + 8.395169425057247e-05, + 5.674947897205129e-05, + 2.5295201339758933e-05 + ], + "bp_grad_F": [ + 0.15471766889095306, + 0.07422170042991638, + 0.05941145122051239, + 0.04111991077661514, + 0.018912144005298615 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.8037109375, + "loss_eval": 0.6682149171829224, + "epoch": 43 + }, + { + "hidden_norms_cls": [ + 1.457137107849121, + 6.368557929992676, + 8.152215957641602, + 10.783031463623047, + 14.167640686035156 + ], + "hidden_norms_avg": [ + 4.462830066680908, + 9.585762977600098, + 11.208064079284668, + 13.065739631652832, + 13.993027687072754 + ], + "bp_grad_per_sample_l2_med": [ + 0.00016671593766659498, + 8.65811962285079e-05, + 7.217681559268385e-05, + 5.042032353230752e-05, + 2.1260022549540736e-05 + ], + "bp_grad_F": [ + 0.15730538964271545, + 0.07605564594268799, + 0.06052647903561592, + 0.04247846081852913, + 0.01923321560025215 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.8134765625, + "loss_eval": 0.6757298111915588, + "epoch": 44 + }, + { + "hidden_norms_cls": [ + 1.4529122114181519, + 6.260828018188477, + 8.032843589782715, + 10.586301803588867, + 13.950445175170898 + ], + "hidden_norms_avg": [ + 4.4613800048828125, + 9.4591646194458, + 11.012088775634766, + 12.760184288024902, + 13.692852973937988 + ], + "bp_grad_per_sample_l2_med": [ + 0.0001344321499345824, + 6.65406696498394e-05, + 5.5920852901181206e-05, + 3.9107719203457236e-05, + 1.5611329217790626e-05 + ], + "bp_grad_F": [ + 0.1660272479057312, + 0.08182978630065918, + 0.06507853418588638, + 0.046549297869205475, + 0.020592277869582176 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.806640625, + "loss_eval": 0.685518205165863, + "epoch": 45 + }, + { + "hidden_norms_cls": [ + 1.4509673118591309, + 6.163156032562256, + 7.82518196105957, + 10.378498077392578, + 13.96407699584961 + ], + "hidden_norms_avg": [ + 4.434635162353516, + 9.335474967956543, + 10.857998847961426, + 12.551763534545898, + 13.511585235595703 + ], + "bp_grad_per_sample_l2_med": [ + 0.00011230952804908156, + 5.540825441130437e-05, + 4.5063996367389336e-05, + 2.9814285881002434e-05, + 1.1903831364179496e-05 + ], + "bp_grad_F": [ + 0.1772848516702652, + 0.08942516893148422, + 0.07126298546791077, + 0.04956020042300224, + 0.021146513521671295 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7958984375, + "loss_eval": 0.69295334815979, + "epoch": 46 + }, + { + "hidden_norms_cls": [ + 1.4527415037155151, + 6.165233612060547, + 7.756027698516846, + 10.242467880249023, + 13.45471477508545 + ], + "hidden_norms_avg": [ + 4.438976287841797, + 9.343517303466797, + 10.836493492126465, + 12.435514450073242, + 13.287426948547363 + ], + "bp_grad_per_sample_l2_med": [ + 0.0001074640967999585, + 5.268112363410182e-05, + 4.352951145847328e-05, + 3.164535519317724e-05, + 1.2792614143108949e-05 + ], + "bp_grad_F": [ + 0.17606250941753387, + 0.08888891339302063, + 0.07090400159358978, + 0.05056491121649742, + 0.021766021847724915 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.806640625, + "loss_eval": 0.6969807147979736, + "epoch": 47 + }, + { + "hidden_norms_cls": [ + 1.4455187320709229, + 6.104230880737305, + 7.606966972351074, + 10.056215286254883, + 13.531522750854492 + ], + "hidden_norms_avg": [ + 4.394144535064697, + 9.197455406188965, + 10.598797798156738, + 12.198134422302246, + 13.1181001663208 + ], + "bp_grad_per_sample_l2_med": [ + 9.084033081308007e-05, + 4.605063077178784e-05, + 3.957438457291573e-05, + 2.727917853917461e-05, + 1.0920393833657727e-05 + ], + "bp_grad_F": [ + 0.1919114887714386, + 0.09742862731218338, + 0.07786751538515091, + 0.05414625629782677, + 0.022070109844207764 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.810546875, + "loss_eval": 0.713995635509491, + "epoch": 48 + }, + { + "hidden_norms_cls": [ + 1.4383388757705688, + 6.067695617675781, + 7.595578670501709, + 9.960418701171875, + 13.367838859558105 + ], + "hidden_norms_avg": [ + 4.4121479988098145, + 9.207226753234863, + 10.62590217590332, + 12.186077117919922, + 13.054762840270996 + ], + "bp_grad_per_sample_l2_med": [ + 0.00010577002831269056, + 5.2172203140798956e-05, + 4.119141885894351e-05, + 2.997498268086929e-05, + 1.1721556802513078e-05 + ], + "bp_grad_F": [ + 0.19711492955684662, + 0.09961695969104767, + 0.07941404730081558, + 0.05621125549077988, + 0.02285071834921837 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7998046875, + "loss_eval": 0.7457866668701172, + "epoch": 49 + }, + { + "hidden_norms_cls": [ + 1.438454270362854, + 6.097946643829346, + 7.5858659744262695, + 9.956851959228516, + 13.270756721496582 + ], + "hidden_norms_avg": [ + 4.392654895782471, + 9.144302368164062, + 10.536030769348145, + 12.092294692993164, + 12.985077857971191 + ], + "bp_grad_per_sample_l2_med": [ + 0.00010758326970972121, + 5.274294744594954e-05, + 4.421125413500704e-05, + 3.0753104510949925e-05, + 1.1889996130776126e-05 + ], + "bp_grad_F": [ + 0.19883227348327637, + 0.1017213836312294, + 0.0807848572731018, + 0.056292060762643814, + 0.022987989708781242 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7998046875, + "loss_eval": 0.7402137517929077, + "epoch": 50 + }, + { + "hidden_norms_cls": [ + 1.4367005825042725, + 6.004332542419434, + 7.504194259643555, + 9.782310485839844, + 13.125389099121094 + ], + "hidden_norms_avg": [ + 4.3845720291137695, + 9.06988525390625, + 10.41169548034668, + 11.905806541442871, + 12.784433364868164 + ], + "bp_grad_per_sample_l2_med": [ + 9.236831829184666e-05, + 4.803440242540091e-05, + 3.901722448063083e-05, + 2.752364525804296e-05, + 1.1449340490798932e-05 + ], + "bp_grad_F": [ + 0.19664740562438965, + 0.10213448107242584, + 0.08124449104070663, + 0.05709528177976608, + 0.023260990157723427 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.8017578125, + "loss_eval": 0.732274055480957, + "epoch": 51 + }, + { + "hidden_norms_cls": [ + 1.4371269941329956, + 5.965888500213623, + 7.401495933532715, + 9.748833656311035, + 13.274619102478027 + ], + "hidden_norms_avg": [ + 4.380629539489746, + 9.019668579101562, + 10.396828651428223, + 11.923394203186035, + 12.77461051940918 + ], + "bp_grad_per_sample_l2_med": [ + 7.760760490782559e-05, + 4.048245318699628e-05, + 3.216555342078209e-05, + 2.2310470740194432e-05, + 8.838631401886232e-06 + ], + "bp_grad_F": [ + 0.20731569826602936, + 0.10790340602397919, + 0.08557390421628952, + 0.059937626123428345, + 0.023827020078897476 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.8046875, + "loss_eval": 0.7751725316047668, + "epoch": 52 + }, + { + "hidden_norms_cls": [ + 1.4341260194778442, + 5.9393768310546875, + 7.411903381347656, + 9.739734649658203, + 13.19786548614502 + ], + "hidden_norms_avg": [ + 4.378067493438721, + 8.986294746398926, + 10.316237449645996, + 11.792922019958496, + 12.673585891723633 + ], + "bp_grad_per_sample_l2_med": [ + 6.968516390770674e-05, + 3.471451054792851e-05, + 2.8814916731789708e-05, + 1.9318980776006356e-05, + 7.639268915227149e-06 + ], + "bp_grad_F": [ + 0.2137812376022339, + 0.11110929399728775, + 0.08847261965274811, + 0.06186935305595398, + 0.024307044222950935 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.8056640625, + "loss_eval": 0.764312744140625, + "epoch": 53 + }, + { + "hidden_norms_cls": [ + 1.4324616193771362, + 5.883492469787598, + 7.319663047790527, + 9.547243118286133, + 13.008822441101074 + ], + "hidden_norms_avg": [ + 4.376597881317139, + 8.99113941192627, + 10.321625709533691, + 11.771256446838379, + 12.628817558288574 + ], + "bp_grad_per_sample_l2_med": [ + 7.972177991177887e-05, + 3.871036460623145e-05, + 3.142197601846419e-05, + 2.2371379600372165e-05, + 8.571933904022444e-06 + ], + "bp_grad_F": [ + 0.21534259617328644, + 0.11241323500871658, + 0.08915204554796219, + 0.06287811696529388, + 0.024627922102808952 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7958984375, + "loss_eval": 0.7731261253356934, + "epoch": 54 + }, + { + "hidden_norms_cls": [ + 1.4321078062057495, + 5.858248710632324, + 7.306819915771484, + 9.56387996673584, + 13.071084022521973 + ], + "hidden_norms_avg": [ + 4.371452808380127, + 8.94082260131836, + 10.26811695098877, + 11.765360832214355, + 12.617104530334473 + ], + "bp_grad_per_sample_l2_med": [ + 7.236401870613918e-05, + 3.8528516597580165e-05, + 3.0984694603830576e-05, + 2.1565809220192023e-05, + 7.989257937879302e-06 + ], + "bp_grad_F": [ + 0.2197490781545639, + 0.1155213937163353, + 0.09188564866781235, + 0.06453806161880493, + 0.024861471727490425 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7958984375, + "loss_eval": 0.7778797149658203, + "epoch": 55 + }, + { + "hidden_norms_cls": [ + 1.4319349527359009, + 5.846956253051758, + 7.283481121063232, + 9.520010948181152, + 13.023626327514648 + ], + "hidden_norms_avg": [ + 4.372097969055176, + 8.950923919677734, + 10.25277328491211, + 11.715353012084961, + 12.572748184204102 + ], + "bp_grad_per_sample_l2_med": [ + 6.399707490345463e-05, + 3.384953743079677e-05, + 2.7512824090081267e-05, + 1.8538266886025667e-05, + 7.2009356699709315e-06 + ], + "bp_grad_F": [ + 0.21942120790481567, + 0.11553706228733063, + 0.09221186488866806, + 0.06456339359283447, + 0.024819573387503624 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.798828125, + "loss_eval": 0.7808230519294739, + "epoch": 56 + }, + { + "hidden_norms_cls": [ + 1.4311723709106445, + 5.84969425201416, + 7.275974273681641, + 9.513206481933594, + 12.99256420135498 + ], + "hidden_norms_avg": [ + 4.37282133102417, + 8.946002006530762, + 10.248116493225098, + 11.702008247375488, + 12.54873275756836 + ], + "bp_grad_per_sample_l2_med": [ + 6.404191663023084e-05, + 3.509471571305767e-05, + 2.9476050258381292e-05, + 1.9310447896714322e-05, + 7.2306984293391e-06 + ], + "bp_grad_F": [ + 0.22379544377326965, + 0.11763197183609009, + 0.09364764392375946, + 0.06539658457040787, + 0.02503892220556736 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.80078125, + "loss_eval": 0.7862963080406189, + "epoch": 57 + }, + { + "hidden_norms_cls": [ + 1.4306355714797974, + 5.837501049041748, + 7.250905990600586, + 9.502866744995117, + 13.022948265075684 + ], + "hidden_norms_avg": [ + 4.368107795715332, + 8.927281379699707, + 10.20923900604248, + 11.650206565856934, + 12.504077911376953 + ], + "bp_grad_per_sample_l2_med": [ + 7.216054655145854e-05, + 3.7455520214280114e-05, + 3.0207094823708758e-05, + 2.1436455426737666e-05, + 7.401154562103329e-06 + ], + "bp_grad_F": [ + 0.22603340446949005, + 0.11918352544307709, + 0.0947541669011116, + 0.06624340265989304, + 0.02514214999973774 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.8017578125, + "loss_eval": 0.7883187532424927, + "epoch": 58 + }, + { + "hidden_norms_cls": [ + 1.4303507804870605, + 5.833686351776123, + 7.24592399597168, + 9.48628044128418, + 12.962425231933594 + ], + "hidden_norms_avg": [ + 4.368690013885498, + 8.926249504089355, + 10.21507453918457, + 11.65561294555664, + 12.507142066955566 + ], + "bp_grad_per_sample_l2_med": [ + 6.914411642355844e-05, + 3.7666370189981535e-05, + 3.0613842682214454e-05, + 2.0578616386046633e-05, + 7.153785645641619e-06 + ], + "bp_grad_F": [ + 0.22665317356586456, + 0.1197197362780571, + 0.09526386857032776, + 0.06651060283184052, + 0.02521701157093048 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.798828125, + "loss_eval": 0.7879265546798706, + "epoch": 59 + }, + { + "hidden_norms_cls": [ + 1.430293083190918, + 5.832674026489258, + 7.243373394012451, + 9.484408378601074, + 12.968917846679688 + ], + "hidden_norms_avg": [ + 4.368374347686768, + 8.92542552947998, + 10.214056015014648, + 11.65438175201416, + 12.507084846496582 + ], + "bp_grad_per_sample_l2_med": [ + 6.862521695438772e-05, + 3.740868851309642e-05, + 3.050716986763291e-05, + 2.0533103452180512e-05, + 7.137502961995779e-06 + ], + "bp_grad_F": [ + 0.22697713971138, + 0.11988355964422226, + 0.09540446847677231, + 0.06659507006406784, + 0.025233659893274307 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.798828125, + "loss_eval": 0.7890625, + "epoch": 60 + } + ], + "dfa_log": [ + { + "hidden_norms_cls": [ + 0.31052109599113464, + 3.2536301612854004, + 4.7681732177734375, + 6.129599571228027, + 7.124356746673584 + ], + "hidden_norms_avg": [ + 6.008513927459717, + 7.034290313720703, + 7.942849636077881, + 8.915729522705078, + 9.635361671447754 + ], + "bp_grad_per_sample_l2_med": [ + 0.004442029166966677, + 0.0012008437188342214, + 0.0009976272704079747, + 0.0009104337659664452, + 0.0008558662957511842 + ], + "bp_grad_F": [ + 0.18535660207271576, + 0.03968670591711998, + 0.032604776322841644, + 0.029427478089928627, + 0.02746850810945034 + ], + "gamma_dfa": 0.003905070887412876, + "gamma_dfa_per_layer": [ + 0.004920602310448885, + 0.005249542184174061, + 0.0038580193649977446, + 0.0015921196900308132 + ], + "acc_eval": 0.1376953125, + "loss_eval": 2.35188627243042, + "epoch": 0 + }, + { + "hidden_norms_cls": [ + 1.969543218612671, + 2409.528564453125, + 2965.8515625, + 5388.57568359375, + 10078.4892578125 + ], + "hidden_norms_avg": [ + 45.02196502685547, + 4042.294189453125, + 10780.892578125, + 14336.607421875, + 18185.720703125 + ], + "bp_grad_per_sample_l2_med": [ + 0.00021220046619419008, + 6.587009124814358e-07, + 6.152115474833408e-07, + 6.009134381201875e-07, + 6.002246095704322e-07 + ], + "bp_grad_F": [ + 0.327708899974823, + 5.584151222137734e-05, + 3.1912604754325e-05, + 2.0850193322985433e-05, + 2.061383202089928e-05 + ], + "gamma_dfa": 0.0062168055737856776, + "gamma_dfa_per_layer": [ + 0.0034891567192971706, + 0.014096655882894993, + 0.008750341832637787, + -0.0014689321396872401 + ], + "acc_eval": 0.240234375, + "loss_eval": 2.0569005012512207, + "epoch": 1 + }, + { + "hidden_norms_cls": [ + 4.063174247741699, + 6453.43212890625, + 10849.3359375, + 13374.7021484375, + 36052.8046875 + ], + "hidden_norms_avg": [ + 93.43009185791016, + 12578.36328125, + 48372.984375, + 57469.9609375, + 73801.421875 + ], + "bp_grad_per_sample_l2_med": [ + 0.00015539559535682201, + 2.588524239399703e-07, + 1.9553621655177267e-07, + 1.8433813409046707e-07, + 1.781664309419284e-07 + ], + "bp_grad_F": [ + 0.09239349514245987, + 5.2396142564248294e-05, + 9.398144356964622e-06, + 6.586112704098923e-06, + 6.009819117025472e-06 + ], + "gamma_dfa": 0.009027999331010506, + "gamma_dfa_per_layer": [ + 0.010615326464176178, + 0.014681536704301834, + 0.012467417865991592, + -0.0016522837104275823 + ], + "acc_eval": 0.2236328125, + "loss_eval": 2.0732030868530273, + "epoch": 2 + }, + { + "hidden_norms_cls": [ + 6.139724254608154, + 13514.7626953125, + 22634.615234375, + 30011.625, + 105646.9140625 + ], + "hidden_norms_avg": [ + 140.93157958984375, + 27337.47265625, + 105659.2734375, + 131533.03125, + 178079.78125 + ], + "bp_grad_per_sample_l2_med": [ + 0.00013191952893976122, + 1.5711624712366756e-07, + 6.783211148331247e-08, + 6.273136676782087e-08, + 6.216072989673194e-08 + ], + "bp_grad_F": [ + 0.15658611059188843, + 2.8997201297897846e-05, + 2.6110699309356278e-06, + 2.0865154510829598e-06, + 2.029548568316386e-06 + ], + "gamma_dfa": 0.0068359210272319615, + "gamma_dfa_per_layer": [ + 0.01239929161965847, + 0.008297638036310673, + 0.009697571396827698, + -0.0030508169438689947 + ], + "acc_eval": 0.2255859375, + "loss_eval": 2.0859451293945312, + "epoch": 3 + }, + { + "hidden_norms_cls": [ + 8.060498237609863, + 26771.85546875, + 47700.99609375, + 65194.9375, + 243467.953125 + ], + "hidden_norms_avg": [ + 189.60569763183594, + 53537.2734375, + 224677.375, + 271932.0625, + 380665.78125 + ], + "bp_grad_per_sample_l2_med": [ + 6.88996515236795e-05, + 7.635610188572173e-08, + 3.178359264666142e-08, + 2.8310823907418126e-08, + 2.8113991135114702e-08 + ], + "bp_grad_F": [ + 0.18565885722637177, + 2.3117625460145064e-05, + 1.2200183618915617e-06, + 1.0046638863059343e-06, + 9.67514324656804e-07 + ], + "gamma_dfa": 0.00588334770873189, + "gamma_dfa_per_layer": [ + 0.012929551303386688, + 0.005936518311500549, + 0.009927853010594845, + -0.0052605317905545235 + ], + "acc_eval": 0.216796875, + "loss_eval": 2.1008124351501465, + "epoch": 4 + }, + { + "hidden_norms_cls": [ + 9.924980163574219, + 45836.29296875, + 86015.5546875, + 115240.734375, + 396649.5625 + ], + "hidden_norms_avg": [ + 236.08885192871094, + 94806.203125, + 447709.03125, + 539570.125, + 680795.4375 + ], + "bp_grad_per_sample_l2_med": [ + 0.00014524323341902345, + 3.066869780354864e-08, + 1.9569522891060842e-08, + 1.7725259482404e-08, + 1.7704602228718613e-08 + ], + "bp_grad_F": [ + 0.21425659954547882, + 1.3653657333634328e-05, + 6.662682494606997e-07, + 5.818570798510336e-07, + 5.813048460368009e-07 + ], + "gamma_dfa": 0.006635856116190553, + "gamma_dfa_per_layer": [ + 0.010879420675337315, + 0.013957185670733452, + 0.00779206957668066, + -0.006085251457989216 + ], + "acc_eval": 0.2265625, + "loss_eval": 2.1039814949035645, + "epoch": 5 + }, + { + "hidden_norms_cls": [ + 11.746355056762695, + 70552.8046875, + 256103.15625, + 694951.625, + 808366.5625 + ], + "hidden_norms_avg": [ + 279.4807434082031, + 128343.046875, + 645095.375, + 810123.8125, + 959871.0 + ], + "bp_grad_per_sample_l2_med": [ + 6.159865733934566e-05, + 2.1975946395969004e-08, + 1.1060357074654803e-08, + 9.967390468546e-09, + 1.011228256686536e-08 + ], + "bp_grad_F": [ + 0.09021499752998352, + 5.0024723350361455e-06, + 4.539916176327097e-07, + 3.891582309734076e-07, + 3.9641980720261927e-07 + ], + "gamma_dfa": 0.003665695490781218, + "gamma_dfa_per_layer": [ + 0.009222292341291904, + 0.0031545532401651144, + 0.00775552075356245, + -0.005469584371894598 + ], + "acc_eval": 0.220703125, + "loss_eval": 2.1214356422424316, + "epoch": 6 + }, + { + "hidden_norms_cls": [ + 13.500020980834961, + 102720.734375, + 194937.703125, + 291956.1875, + 801362.9375 + ], + "hidden_norms_avg": [ + 321.6004943847656, + 208601.421875, + 1002364.1875, + 1189027.25, + 1448220.5 + ], + "bp_grad_per_sample_l2_med": [ + 0.00014696457947138697, + 1.642234792598174e-08, + 8.907985460382406e-09, + 8.127726935924784e-09, + 8.014287011803845e-09 + ], + "bp_grad_F": [ + 0.06841801106929779, + 4.044012712256517e-06, + 3.4213746857858496e-07, + 2.990170742123155e-07, + 2.988945198012516e-07 + ], + "gamma_dfa": 0.004651258233934641, + "gamma_dfa_per_layer": [ + 0.007132797501981258, + 0.010231958702206612, + 0.006907162256538868, + -0.0056668855249881744 + ], + "acc_eval": 0.2099609375, + "loss_eval": 2.0963404178619385, + "epoch": 7 + }, + { + "hidden_norms_cls": [ + 15.302024841308594, + 160872.96875, + 604043.4375, + 1296474.125, + 1546227.0 + ], + "hidden_norms_avg": [ + 359.47259521484375, + 298263.5, + 1317573.75, + 1553171.25, + 1878602.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.000136460512294434, + 1.0422226637274434e-08, + 5.898865396147812e-09, + 5.4856741371622775e-09, + 5.4926698744850455e-09 + ], + "bp_grad_F": [ + 0.07932842522859573, + 2.1081664272060152e-06, + 2.560300345066935e-07, + 2.2075181504987995e-07, + 2.2043255398784822e-07 + ], + "gamma_dfa": 0.004627774120308459, + "gamma_dfa_per_layer": [ + 0.006024104077368975, + 0.01116865873336792, + 0.005747191607952118, + -0.004428857937455177 + ], + "acc_eval": 0.220703125, + "loss_eval": 2.0744481086730957, + "epoch": 8 + }, + { + "hidden_norms_cls": [ + 17.009462356567383, + 204948.671875, + 405065.21875, + 549855.3125, + 1400973.875 + ], + "hidden_norms_avg": [ + 396.66558837890625, + 480022.3125, + 2019487.75, + 2350966.75, + 2708745.25 + ], + "bp_grad_per_sample_l2_med": [ + 0.00026586122112348676, + 1.0135083883255902e-08, + 6.120911777429683e-09, + 5.8359819199438334e-09, + 5.662275093243352e-09 + ], + "bp_grad_F": [ + 0.37594228982925415, + 1.2430107744876295e-05, + 2.3164905371686473e-07, + 2.0265188993562333e-07, + 2.001279000296563e-07 + ], + "gamma_dfa": 0.004862090456299484, + "gamma_dfa_per_layer": [ + 0.006278482731431723, + 0.012495312839746475, + 0.005351733416318893, + -0.004677167162299156 + ], + "acc_eval": 0.2265625, + "loss_eval": 2.0641043186187744, + "epoch": 9 + }, + { + "hidden_norms_cls": [ + 18.669273376464844, + 316404.0625, + 619048.375, + 804163.5, + 1798441.0 + ], + "hidden_norms_avg": [ + 434.9389343261719, + 650998.25, + 2437815.25, + 2802479.0, + 3269810.75 + ], + "bp_grad_per_sample_l2_med": [ + 0.0003605277743190527, + 6.009571951182124e-09, + 4.676837139783174e-09, + 4.657723096102018e-09, + 4.49635484400801e-09 + ], + "bp_grad_F": [ + 0.14846132695674896, + 4.3979816837236285e-06, + 1.7215310776919068e-07, + 1.581869213396203e-07, + 1.561998885790672e-07 + ], + "gamma_dfa": 0.004480502451770008, + "gamma_dfa_per_layer": [ + 0.006086423061788082, + 0.010957473888993263, + 0.004586817231029272, + -0.003708704374730587 + ], + "acc_eval": 0.2265625, + "loss_eval": 2.0591208934783936, + "epoch": 10 + }, + { + "hidden_norms_cls": [ + 20.326766967773438, + 443508.46875, + 899668.0625, + 1131299.375, + 2182177.5 + ], + "hidden_norms_avg": [ + 473.4957275390625, + 854388.9375, + 2926402.0, + 3295286.75, + 3825241.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005711699486710131, + 4.8561932253221585e-09, + 3.871160725310574e-09, + 3.833432238309342e-09, + 3.74674646863582e-09 + ], + "bp_grad_F": [ + 1.0882306098937988, + 3.257442813264788e-06, + 1.6311048511852277e-07, + 1.3397652764979284e-07, + 1.326638283671855e-07 + ], + "gamma_dfa": 0.00462920154677704, + "gamma_dfa_per_layer": [ + 0.008616073988378048, + 0.009468363597989082, + 0.003528183326125145, + -0.003095814725384116 + ], + "acc_eval": 0.234375, + "loss_eval": 2.0471031665802, + "epoch": 11 + }, + { + "hidden_norms_cls": [ + 21.944751739501953, + 581616.1875, + 1205317.625, + 1456984.375, + 2587496.75 + ], + "hidden_norms_avg": [ + 511.203125, + 1119566.25, + 3562080.25, + 3992681.5, + 4538977.5 + ], + "bp_grad_per_sample_l2_med": [ + 0.0009346234728582203, + 4.117307383921798e-09, + 3.257990321969828e-09, + 3.2349201095627222e-09, + 3.1638915931608835e-09 + ], + "bp_grad_F": [ + 0.42184242606163025, + 2.2249487301451154e-06, + 1.2586481545895367e-07, + 1.1919513553948491e-07, + 1.181974553787768e-07 + ], + "gamma_dfa": 0.0034259117674082518, + "gamma_dfa_per_layer": [ + 0.00526365265250206, + 0.008116653189063072, + 0.0030386103317141533, + -0.0027152691036462784 + ], + "acc_eval": 0.2255859375, + "loss_eval": 2.0867815017700195, + "epoch": 12 + }, + { + "hidden_norms_cls": [ + 23.46192741394043, + 733909.875, + 1508464.125, + 1841074.875, + 3055228.75 + ], + "hidden_norms_avg": [ + 546.8953247070312, + 1235037.0, + 3742642.75, + 4200378.0, + 4892382.5 + ], + "bp_grad_per_sample_l2_med": [ + 0.0009773626225069165, + 3.5754739169391314e-09, + 2.9525657474493983e-09, + 2.9297144710227485e-09, + 2.8955173814182444e-09 + ], + "bp_grad_F": [ + 0.15538839995861053, + 2.0067516288690967e-06, + 1.1871279781416888e-07, + 1.046255491132797e-07, + 1.0416235340926505e-07 + ], + "gamma_dfa": 0.002340688748518005, + "gamma_dfa_per_layer": [ + 0.0019294173689559102, + 0.007100729271769524, + 0.0026949853636324406, + -0.0023623770102858543 + ], + "acc_eval": 0.232421875, + "loss_eval": 2.088564872741699, + "epoch": 13 + }, + { + "hidden_norms_cls": [ + 24.94734764099121, + 1151802.75, + 3671404.25, + 4625593.5, + 5052590.0 + ], + "hidden_norms_avg": [ + 582.0977172851562, + 1475600.75, + 4403784.0, + 5008330.0, + 5801902.5 + ], + "bp_grad_per_sample_l2_med": [ + 0.0007562927203252912, + 2.676908916754428e-09, + 2.2055020032496486e-09, + 2.179777247590664e-09, + 2.173634161550808e-09 + ], + "bp_grad_F": [ + 0.36642539501190186, + 3.967601060139714e-06, + 8.444904864290947e-08, + 7.993684647544796e-08, + 7.900475651467787e-08 + ], + "gamma_dfa": 0.0017957038653548807, + "gamma_dfa_per_layer": [ + 0.0019046596717089415, + 0.005139546003192663, + 0.001759346341714263, + -0.0016207365551963449 + ], + "acc_eval": 0.2109375, + "loss_eval": 2.08038330078125, + "epoch": 14 + }, + { + "hidden_norms_cls": [ + 26.36701011657715, + 2210167.25, + 6804633.5, + 7592304.0, + 7772230.0 + ], + "hidden_norms_avg": [ + 615.4749145507812, + 1852637.875, + 5297662.0, + 6054509.0, + 6940769.5 + ], + "bp_grad_per_sample_l2_med": [ + 0.0009053644607774913, + 2.014937106054049e-09, + 1.6813730407960747e-09, + 1.63718916201816e-09, + 1.6135126568173064e-09 + ], + "bp_grad_F": [ + 0.15154746174812317, + 3.760401341423858e-06, + 6.479213254806382e-08, + 6.350511227992683e-08, + 6.00040195308793e-08 + ], + "gamma_dfa": 0.00026495405472815037, + "gamma_dfa_per_layer": [ + -0.0023696955759078264, + 0.0033954931423068047, + 0.0012687842827290297, + -0.0012347656302154064 + ], + "acc_eval": 0.2333984375, + "loss_eval": 2.0685791969299316, + "epoch": 15 + }, + { + "hidden_norms_cls": [ + 27.747814178466797, + 2020373.25, + 6352088.0, + 7372759.0, + 7751557.0 + ], + "hidden_norms_avg": [ + 647.8715209960938, + 2166255.0, + 5964871.0, + 6820316.0, + 7888183.5 + ], + "bp_grad_per_sample_l2_med": [ + 0.000535959261469543, + 1.8833379300531305e-09, + 1.6251041623505103e-09, + 1.5936774122593533e-09, + 1.5920619267362213e-09 + ], + "bp_grad_F": [ + 0.06977186352014542, + 4.2289293560315855e-06, + 5.94431135425566e-08, + 5.634814925770115e-08, + 5.626874255426628e-08 + ], + "gamma_dfa": 0.0009004889943753369, + "gamma_dfa_per_layer": [ + 1.3074517482891679e-05, + 0.003544538514688611, + 0.0012792267370969057, + -0.0012348837917670608 + ], + "acc_eval": 0.234375, + "loss_eval": 2.0676896572113037, + "epoch": 16 + }, + { + "hidden_norms_cls": [ + 29.053590774536133, + 3083884.0, + 8692576.0, + 10024582.0, + 10278143.0 + ], + "hidden_norms_avg": [ + 679.1322021484375, + 2750560.0, + 7205648.0, + 8204248.5, + 9336888.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006835769745521247, + 1.5781399520520267e-09, + 1.3378861352109084e-09, + 1.2995797771253592e-09, + 1.2820219330578198e-09 + ], + "bp_grad_F": [ + 0.25767308473587036, + 1.6246455061263987e-06, + 6.349461756371966e-08, + 6.242527206268278e-08, + 4.5665522918625356e-08 + ], + "gamma_dfa": 0.0003357882087584585, + "gamma_dfa_per_layer": [ + -0.001319598057307303, + 0.002733789850026369, + 0.0008947808528319001, + -0.0009658198105171323 + ], + "acc_eval": 0.23046875, + "loss_eval": 2.0777204036712646, + "epoch": 17 + }, + { + "hidden_norms_cls": [ + 30.246440887451172, + 3200014.0, + 8626485.0, + 9743865.0, + 10146681.0 + ], + "hidden_norms_avg": [ + 708.5135498046875, + 2989787.0, + 7458187.5, + 8607358.0, + 10070516.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005841738311573863, + 1.4247033552905464e-09, + 1.2406603522308046e-09, + 1.21434362565509e-09, + 1.217411615961339e-09 + ], + "bp_grad_F": [ + 0.08999939262866974, + 1.7538721976961824e-06, + 4.803559150445835e-08, + 4.36282370230856e-08, + 4.3068951072200434e-08 + ], + "gamma_dfa": -2.2506676032207906e-05, + "gamma_dfa_per_layer": [ + -0.002581980312243104, + 0.002518225461244583, + 0.0009317069780081511, + -0.0009579788311384618 + ], + "acc_eval": 0.2392578125, + "loss_eval": 2.0627658367156982, + "epoch": 18 + }, + { + "hidden_norms_cls": [ + 31.42615509033203, + 4556513.5, + 10711725.0, + 11956874.0, + 12318100.0 + ], + "hidden_norms_avg": [ + 737.157958984375, + 3520006.5, + 8303534.5, + 9571905.0, + 11293369.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0007374613196589053, + 1.2111865954622658e-09, + 1.0633514069269268e-09, + 1.0194807220642588e-09, + 1.0034837405470398e-09 + ], + "bp_grad_F": [ + 0.12247106432914734, + 2.740941681622644e-06, + 4.6019909660799385e-08, + 4.372383344275477e-08, + 3.6865596797497346e-08 + ], + "gamma_dfa": 0.0005736843086197041, + "gamma_dfa_per_layer": [ + 0.0002251334663014859, + 0.0021776340436190367, + 0.0006795942317694426, + -0.0007876245072111487 + ], + "acc_eval": 0.22265625, + "loss_eval": 2.062939167022705, + "epoch": 19 + }, + { + "hidden_norms_cls": [ + 32.559165954589844, + 4224958.5, + 10227904.0, + 11507743.0, + 12041642.0 + ], + "hidden_norms_avg": [ + 765.5013427734375, + 3988012.5, + 8953060.0, + 10431809.0, + 12380471.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006474430556409061, + 1.179957687114097e-09, + 1.0031043773395254e-09, + 9.744786089171953e-10, + 9.782840093564005e-10 + ], + "bp_grad_F": [ + 0.12209221720695496, + 1.616509621271689e-06, + 3.8821241332698264e-08, + 3.6648025059093925e-08, + 3.527756931021031e-08 + ], + "gamma_dfa": -0.00019280995184089988, + "gamma_dfa_per_layer": [ + -0.002911232877522707, + 0.0022857212461531162, + 0.0006594176520593464, + -0.0008051458280533552 + ], + "acc_eval": 0.240234375, + "loss_eval": 2.0555338859558105, + "epoch": 20 + }, + { + "hidden_norms_cls": [ + 33.59370040893555, + 5323277.5, + 11855673.0, + 13697022.0, + 14246667.0 + ], + "hidden_norms_avg": [ + 789.9629516601562, + 4634178.0, + 10205192.0, + 11930021.0, + 13993152.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006352875498123467, + 1.0036302899862903e-09, + 8.794808215029093e-10, + 8.579849608558732e-10, + 8.574586596310496e-10 + ], + "bp_grad_F": [ + 0.09745635837316513, + 5.127232043378172e-07, + 3.342180221466151e-08, + 3.173540363832217e-08, + 3.091372491326183e-08 + ], + "gamma_dfa": -0.0006959605379961431, + "gamma_dfa_per_layer": [ + -0.004454955458641052, + 0.0018251645378768444, + 0.0005447655566968024, + -0.000698816787917167 + ], + "acc_eval": 0.2265625, + "loss_eval": 2.056840181350708, + "epoch": 21 + }, + { + "hidden_norms_cls": [ + 34.583858489990234, + 6778143.0, + 13401385.0, + 15847366.0, + 16344266.0 + ], + "hidden_norms_avg": [ + 814.3038330078125, + 5395352.5, + 11137891.0, + 13094203.0, + 15406614.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006492820102721453, + 8.925201688825268e-10, + 7.777718469270667e-10, + 7.627384279729199e-10, + 7.601627105557895e-10 + ], + "bp_grad_F": [ + 0.08801906555891037, + 1.8660392697711359e-06, + 3.472371190582635e-08, + 2.7845679539950652e-08, + 2.7135749647300145e-08 + ], + "gamma_dfa": -0.0012016263353871182, + "gamma_dfa_per_layer": [ + -0.006280785892158747, + 0.001651690574362874, + 0.0004266214673407376, + -0.0006040314910933375 + ], + "acc_eval": 0.2255859375, + "loss_eval": 2.0746021270751953, + "epoch": 22 + }, + { + "hidden_norms_cls": [ + 35.516666412353516, + 6985935.5, + 14020116.0, + 16575473.0, + 17109272.0 + ], + "hidden_norms_avg": [ + 838.5614013671875, + 6365095.0, + 12448551.0, + 14566203.0, + 16943832.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0007437304593622684, + 8.387662786546457e-10, + 7.299132409599451e-10, + 7.170505300635455e-10, + 7.115786293532267e-10 + ], + "bp_grad_F": [ + 0.1540054827928543, + 1.4262270724429982e-06, + 2.9031822279534936e-08, + 2.776059560005706e-08, + 2.5270722403547552e-08 + ], + "gamma_dfa": -0.00010950746218441054, + "gamma_dfa_per_layer": [ + -0.0017721566837280989, + 0.0015054477844387293, + 0.0003872170636896044, + -0.000558538013137877 + ], + "acc_eval": 0.2392578125, + "loss_eval": 2.065904140472412, + "epoch": 23 + }, + { + "hidden_norms_cls": [ + 36.39850997924805, + 7861434.0, + 14415042.0, + 16660478.0, + 17496562.0 + ], + "hidden_norms_avg": [ + 862.4805297851562, + 6600421.5, + 12694272.0, + 15043836.0, + 17968980.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005834798794239759, + 7.558619286029966e-10, + 6.530426754913776e-10, + 6.420492471015393e-10, + 6.389526685524061e-10 + ], + "bp_grad_F": [ + 0.09943251311779022, + 1.4788431599299656e-06, + 2.582997993272329e-08, + 2.3754548550414256e-08, + 2.322415859623561e-08 + ], + "gamma_dfa": -0.0009345750440843403, + "gamma_dfa_per_layer": [ + -0.005123761482536793, + 0.0015550897223874927, + 0.0003447675844654441, + -0.0005143960006535053 + ], + "acc_eval": 0.2529296875, + "loss_eval": 2.041043758392334, + "epoch": 24 + }, + { + "hidden_norms_cls": [ + 37.283199310302734, + 8933519.0, + 15359962.0, + 17441364.0, + 18346258.0 + ], + "hidden_norms_avg": [ + 883.0538330078125, + 7354756.0, + 13721222.0, + 16105006.0, + 19368310.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00026283934130333364, + 6.873029922971341e-10, + 6.058342161274766e-10, + 6.000055452481945e-10, + 5.98545990548871e-10 + ], + "bp_grad_F": [ + 0.11474847048521042, + 2.4628958271932788e-05, + 1.3788282160476228e-07, + 2.199209525599599e-08, + 2.15279509774291e-08 + ], + "gamma_dfa": -0.0014451614988502115, + "gamma_dfa_per_layer": [ + -0.006978219375014305, + 0.0013526652473956347, + 0.0002992081572301686, + -0.0004543000250123441 + ], + "acc_eval": 0.2451171875, + "loss_eval": 2.0492172241210938, + "epoch": 25 + }, + { + "hidden_norms_cls": [ + 38.11500930786133, + 10836846.0, + 17507178.0, + 19929002.0, + 20726542.0 + ], + "hidden_norms_avg": [ + 901.960205078125, + 8138403.5, + 14552406.0, + 17157852.0, + 20773122.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005291851703077555, + 6.35470620569123e-10, + 5.6208532273061e-10, + 5.496996191567405e-10, + 5.433966054901873e-10 + ], + "bp_grad_F": [ + 0.07972504198551178, + 1.2489309710872476e-06, + 2.2593511062041216e-08, + 2.1460177634935462e-08, + 1.941634231172884e-08 + ], + "gamma_dfa": -0.000788729106716346, + "gamma_dfa_per_layer": [ + -0.00451657734811306, + 0.0015372345224022865, + 0.00024382121046073735, + -0.00041939481161534786 + ], + "acc_eval": 0.2265625, + "loss_eval": 2.0669429302215576, + "epoch": 26 + }, + { + "hidden_norms_cls": [ + 38.87166213989258, + 11480826.0, + 18077316.0, + 20884150.0, + 22098916.0 + ], + "hidden_norms_avg": [ + 920.577880859375, + 8893063.0, + 15579717.0, + 18401958.0, + 22225722.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005104477168060839, + 5.858571405781277e-10, + 5.198571018105724e-10, + 5.114682011253535e-10, + 5.120092128052534e-10 + ], + "bp_grad_F": [ + 0.13577552139759064, + 2.0482397303567268e-05, + 2.1394010119024642e-08, + 2.0649670418038113e-08, + 1.8286096192809964e-08 + ], + "gamma_dfa": -0.0004262418224243447, + "gamma_dfa_per_layer": [ + -0.0029134145006537437, + 0.0013950758147984743, + 0.00018188939429819584, + -0.0003685179981403053 + ], + "acc_eval": 0.2490234375, + "loss_eval": 2.053382396697998, + "epoch": 27 + }, + { + "hidden_norms_cls": [ + 39.6179313659668, + 10591964.0, + 16346992.0, + 18847678.0, + 21847726.0 + ], + "hidden_norms_avg": [ + 937.9719848632812, + 9576851.0, + 16784906.0, + 19828342.0, + 23673520.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00023470228188671172, + 5.629120503058971e-10, + 5.04818964408571e-10, + 4.974847200855947e-10, + 4.968530586957343e-10 + ], + "bp_grad_F": [ + 0.03459252789616585, + 8.028884622035548e-06, + 1.9716548393944322e-08, + 1.807922522800709e-08, + 1.7913645677936074e-08 + ], + "gamma_dfa": -0.0018089709192281589, + "gamma_dfa_per_layer": [ + -0.008484212681651115, + 0.0014374498277902603, + 0.00020184949971735477, + -0.00039097032276913524 + ], + "acc_eval": 0.224609375, + "loss_eval": 2.0604443550109863, + "epoch": 28 + }, + { + "hidden_norms_cls": [ + 40.344242095947266, + 10467030.0, + 15235544.0, + 17027748.0, + 23281082.0 + ], + "hidden_norms_avg": [ + 954.3844604492188, + 10038487.0, + 16584430.0, + 19613710.0, + 24505476.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00015729073493275791, + 5.36418076624301e-10, + 4.5378717450361705e-10, + 4.4796202858243817e-10, + 4.477457571372412e-10 + ], + "bp_grad_F": [ + 0.04820869117975235, + 2.3596087430632906e-06, + 1.7855482425943592e-08, + 1.7340139990551506e-08, + 1.672622929049794e-08 + ], + "gamma_dfa": -0.0013171448226785287, + "gamma_dfa_per_layer": [ + -0.006438364274799824, + 0.0013229507021605968, + 0.00020032597240060568, + -0.00035349169047549367 + ], + "acc_eval": 0.236328125, + "loss_eval": 2.061581611633301, + "epoch": 29 + }, + { + "hidden_norms_cls": [ + 41.04903793334961, + 12235056.0, + 17721728.0, + 19784938.0, + 25024272.0 + ], + "hidden_norms_avg": [ + 969.5825805664062, + 11940616.0, + 19401350.0, + 22468846.0, + 26879212.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00035265026963315904, + 5.423251847602728e-10, + 4.4247547292819434e-10, + 4.358849947649901e-10, + 4.3593659238005955e-10 + ], + "bp_grad_F": [ + 0.0735621377825737, + 1.917086819958058e-06, + 1.9346130031294706e-08, + 1.5936171493535767e-08, + 1.5501758099389917e-08 + ], + "gamma_dfa": -0.001081134716514498, + "gamma_dfa_per_layer": [ + -0.005434883758425713, + 0.0012318368535488844, + 0.00019993717432953417, + -0.00032142913551069796 + ], + "acc_eval": 0.2265625, + "loss_eval": 2.0760293006896973, + "epoch": 30 + }, + { + "hidden_norms_cls": [ + 41.723567962646484, + 10562926.0, + 13746881.0, + 15736975.0, + 26354068.0 + ], + "hidden_norms_avg": [ + 984.3485717773438, + 12533570.0, + 19717132.0, + 22961588.0, + 28083302.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002546848263591528, + 4.777176432213537e-10, + 4.0483683072523036e-10, + 4.0072575813177025e-10, + 4.013349652609577e-10 + ], + "bp_grad_F": [ + 0.06879052519798279, + 1.7057992636182462e-06, + 1.6271217262442406e-08, + 1.4877675091895526e-08, + 1.484982270483215e-08 + ], + "gamma_dfa": -0.0015064112594700418, + "gamma_dfa_per_layer": [ + -0.007336798589676619, + 0.0013772943057119846, + 0.00022022609482519329, + -0.0002863668487407267 + ], + "acc_eval": 0.20703125, + "loss_eval": 2.078207492828369, + "epoch": 31 + }, + { + "hidden_norms_cls": [ + 42.342227935791016, + 13283171.0, + 17422292.0, + 19770980.0, + 28086972.0 + ], + "hidden_norms_avg": [ + 999.26123046875, + 13875387.0, + 21404830.0, + 24833206.0, + 29923332.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0003442879824433476, + 4.4142753341525065e-10, + 3.8908712363117104e-10, + 3.845144203040718e-10, + 3.8369810106964053e-10 + ], + "bp_grad_F": [ + 0.10482411086559296, + 2.8695783385046525e-06, + 1.7273855235089286e-08, + 1.4195452813225984e-08, + 1.3857171410336377e-08 + ], + "gamma_dfa": -0.0009248298447346315, + "gamma_dfa_per_layer": [ + -0.004888010676950216, + 0.0012753453338518739, + 0.00017224671319127083, + -0.0002589007490314543 + ], + "acc_eval": 0.2373046875, + "loss_eval": 2.078000068664551, + "epoch": 32 + }, + { + "hidden_norms_cls": [ + 42.90097427368164, + 16870614.0, + 22116480.0, + 24442270.0, + 29665536.0 + ], + "hidden_norms_avg": [ + 1013.56396484375, + 14764026.0, + 21856036.0, + 25334308.0, + 30474444.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00047380090109072626, + 4.22005735911668e-10, + 3.5831293487831317e-10, + 3.5505443030103834e-10, + 3.5614466931122024e-10 + ], + "bp_grad_F": [ + 0.1035437062382698, + 1.2711110457530594e-06, + 1.418432571398398e-08, + 1.2978671470875724e-08, + 1.2937439564097986e-08 + ], + "gamma_dfa": -0.00034118785697501153, + "gamma_dfa_per_layer": [ + -0.002568737603724003, + 0.001293556997552514, + 0.000137164635816589, + -0.00022673545754514635 + ], + "acc_eval": 0.2265625, + "loss_eval": 2.0730109214782715, + "epoch": 33 + }, + { + "hidden_norms_cls": [ + 43.419898986816406, + 18221210.0, + 22675590.0, + 24728546.0, + 30520888.0 + ], + "hidden_norms_avg": [ + 1027.6595458984375, + 16337822.0, + 23378412.0, + 26921720.0, + 31927994.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0004629126051440835, + 4.106864293085266e-10, + 3.557905081663648e-10, + 3.536362591649578e-10, + 3.5421768296295397e-10 + ], + "bp_grad_F": [ + 0.08281872421503067, + 1.4263683851822861e-06, + 1.3245164076636229e-08, + 1.2740572152836194e-08, + 1.2752362721357713e-08 + ], + "gamma_dfa": -0.0006677984529233072, + "gamma_dfa_per_layer": [ + -0.0036979843862354755, + 0.0011370242573320866, + 0.00012192466238047928, + -0.00023215834517031908 + ], + "acc_eval": 0.23828125, + "loss_eval": 2.0669164657592773, + "epoch": 34 + }, + { + "hidden_norms_cls": [ + 43.868682861328125, + 18758150.0, + 23109030.0, + 25432404.0, + 31503354.0 + ], + "hidden_norms_avg": [ + 1040.644287109375, + 15512033.0, + 21965288.0, + 25882228.0, + 31813648.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00048007385339587927, + 4.271450415593847e-10, + 3.4119135294830016e-10, + 3.391489034054729e-10, + 3.387232438978316e-10 + ], + "bp_grad_F": [ + 0.36019906401634216, + 1.9220671674702317e-05, + 1.358929768713324e-08, + 1.2455363851415768e-08, + 1.2401786264604198e-08 + ], + "gamma_dfa": -0.00031488661625189707, + "gamma_dfa_per_layer": [ + -0.002551896497607231, + 0.001431336859241128, + 9.301750105805695e-05, + -0.00023200432769954205 + ], + "acc_eval": 0.234375, + "loss_eval": 2.0609660148620605, + "epoch": 35 + }, + { + "hidden_norms_cls": [ + 44.293697357177734, + 21659484.0, + 26491946.0, + 29332332.0, + 32793040.0 + ], + "hidden_norms_avg": [ + 1051.83740234375, + 17567000.0, + 24049368.0, + 28083622.0, + 33467380.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00043948061647824943, + 4.0486294872188466e-10, + 3.3493480211532756e-10, + 3.3453453895937457e-10, + 3.3604671823006527e-10 + ], + "bp_grad_F": [ + 0.10878865420818329, + 3.037727083210484e-06, + 1.2592222375928941e-08, + 1.2103843261002112e-08, + 1.2022392859023512e-08 + ], + "gamma_dfa": -0.00015630935013177805, + "gamma_dfa_per_layer": [ + -0.0019119374919682741, + 0.0014281735057011247, + 7.57485831854865e-05, + -0.00021722199744544923 + ], + "acc_eval": 0.2431640625, + "loss_eval": 2.0587897300720215, + "epoch": 36 + }, + { + "hidden_norms_cls": [ + 44.68658447265625, + 21628572.0, + 26077032.0, + 28336790.0, + 33305650.0 + ], + "hidden_norms_avg": [ + 1062.067138671875, + 18163602.0, + 24404446.0, + 28478594.0, + 34156064.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005865619168616831, + 4.0249623078914e-10, + 3.3121710929506776e-10, + 3.310392793220984e-10, + 3.318675056984688e-10 + ], + "bp_grad_F": [ + 0.20061102509498596, + 1.4544431223839638e-06, + 1.3441372459510603e-08, + 1.2418687411752671e-08, + 1.2016870165609816e-08 + ], + "gamma_dfa": -0.00039686575473751873, + "gamma_dfa_per_layer": [ + -0.0029668076895177364, + 0.0015295962803065777, + 6.985102663747966e-05, + -0.00022010263637639582 + ], + "acc_eval": 0.2275390625, + "loss_eval": 2.0569839477539062, + "epoch": 37 + }, + { + "hidden_norms_cls": [ + 45.062259674072266, + 25376026.0, + 30203382.0, + 32772746.0, + 35119040.0 + ], + "hidden_norms_avg": [ + 1071.6630859375, + 18637428.0, + 24671676.0, + 28819074.0, + 34739232.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005594053072854877, + 3.9023292930373543e-10, + 3.2509483993692356e-10, + 3.23913590394298e-10, + 3.250162639023557e-10 + ], + "bp_grad_F": [ + 0.13359014689922333, + 4.247889592079446e-06, + 1.4051211749688264e-08, + 1.2317144637563615e-08, + 1.1531787968976914e-08 + ], + "gamma_dfa": -0.0005283378432068275, + "gamma_dfa_per_layer": [ + -0.0037097278982400894, + 0.0017695487476885319, + 4.308431380195543e-05, + -0.000216256536077708 + ], + "acc_eval": 0.2666015625, + "loss_eval": 2.0476396083831787, + "epoch": 38 + }, + { + "hidden_norms_cls": [ + 45.38731384277344, + 24903830.0, + 30026926.0, + 33563200.0, + 35539932.0 + ], + "hidden_norms_avg": [ + 1080.0860595703125, + 20496828.0, + 26709690.0, + 30932660.0, + 36357072.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005004233680665493, + 3.8017103354270887e-10, + 3.1730867933177365e-10, + 3.167333340048373e-10, + 3.1768215835725755e-10 + ], + "bp_grad_F": [ + 0.24717594683170319, + 2.9436714612529613e-06, + 1.1934382371237007e-08, + 1.1459838411553847e-08, + 1.1365100860416533e-08 + ], + "gamma_dfa": -0.00022031009211787023, + "gamma_dfa_per_layer": [ + -0.0022991469595581293, + 0.0015725505072623491, + 4.821528273168951e-05, + -0.00020285919890739024 + ], + "acc_eval": 0.244140625, + "loss_eval": 2.0488319396972656, + "epoch": 39 + }, + { + "hidden_norms_cls": [ + 45.68830490112305, + 27470426.0, + 32688434.0, + 35418348.0, + 37778420.0 + ], + "hidden_norms_avg": [ + 1088.195068359375, + 19636778.0, + 25748976.0, + 30137794.0, + 36341748.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.000507087679579854, + 3.7039282752004965e-10, + 3.106248036566228e-10, + 3.096240208666501e-10, + 3.10577202844442e-10 + ], + "bp_grad_F": [ + 0.247142493724823, + 5.3006801863375586e-06, + 1.5794508811950436e-08, + 1.5706817180216603e-08, + 1.0953104201405495e-08 + ], + "gamma_dfa": -0.00026929086834570626, + "gamma_dfa_per_layer": [ + -0.0025479549076408148, + 0.0016400209860876203, + 2.4549921363359317e-05, + -0.00019377947319298983 + ], + "acc_eval": 0.2548828125, + "loss_eval": 2.0380632877349854, + "epoch": 40 + }, + { + "hidden_norms_cls": [ + 45.95172119140625, + 29219790.0, + 34751968.0, + 37645720.0, + 39363420.0 + ], + "hidden_norms_avg": [ + 1095.55908203125, + 20059094.0, + 26161532.0, + 30539128.0, + 36948852.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0004415555449668318, + 3.578430884942918e-10, + 3.05192260352527e-10, + 3.046965735276075e-10, + 3.0560776131949297e-10 + ], + "bp_grad_F": [ + 0.1088763028383255, + 4.900910880678566e-06, + 1.1387745857405207e-08, + 1.0883239198733463e-08, + 1.0722531307294503e-08 + ], + "gamma_dfa": -0.0006201292017067317, + "gamma_dfa_per_layer": [ + -0.004000760614871979, + 0.0016785510815680027, + 3.265401755925268e-05, + -0.0001909612910822034 + ], + "acc_eval": 0.25, + "loss_eval": 2.041342258453369, + "epoch": 41 + }, + { + "hidden_norms_cls": [ + 46.1973876953125, + 29375584.0, + 34745100.0, + 38109944.0, + 40454424.0 + ], + "hidden_norms_avg": [ + 1102.340087890625, + 21066706.0, + 27296236.0, + 31874516.0, + 38108092.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00040148120024241507, + 3.579030682931972e-10, + 3.0047467292071417e-10, + 2.979275159908923e-10, + 2.9703536852387913e-10 + ], + "bp_grad_F": [ + 0.2525903582572937, + 8.925781003199518e-06, + 1.1219404072448924e-08, + 1.084880185686643e-08, + 1.0492501978376367e-08 + ], + "gamma_dfa": -0.00034770679030771134, + "gamma_dfa_per_layer": [ + -0.002933461917564273, + 0.001698363572359085, + 3.748677045223303e-05, + -0.0001932155864778906 + ], + "acc_eval": 0.244140625, + "loss_eval": 2.0345935821533203, + "epoch": 42 + }, + { + "hidden_norms_cls": [ + 46.42340087890625, + 27399134.0, + 31821602.0, + 34474508.0, + 38109560.0 + ], + "hidden_norms_avg": [ + 1108.236328125, + 21543940.0, + 27684760.0, + 32082502.0, + 38718660.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.000570801377762109, + 3.5802838471710174e-10, + 2.990230563160168e-10, + 2.9852947891484405e-10, + 2.9978877713610075e-10 + ], + "bp_grad_F": [ + 0.173945352435112, + 2.9223572255432373e-06, + 1.2221713419080515e-08, + 1.1539666999738074e-08, + 1.0805790928714032e-08 + ], + "gamma_dfa": -5.9806044191645924e-05, + "gamma_dfa_per_layer": [ + -0.001857165014371276, + 0.0017762510105967522, + 3.49261645169463e-05, + -0.00019323633750900626 + ], + "acc_eval": 0.2412109375, + "loss_eval": 2.043308734893799, + "epoch": 43 + }, + { + "hidden_norms_cls": [ + 46.619728088378906, + 31052948.0, + 36542220.0, + 39935176.0, + 42316204.0 + ], + "hidden_norms_avg": [ + 1113.68896484375, + 22679808.0, + 29079456.0, + 33751472.0, + 40006012.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00046165595995262265, + 3.339039600369631e-10, + 2.9132621315319795e-10, + 2.9080904351275194e-10, + 2.9134980539247124e-10 + ], + "bp_grad_F": [ + 0.08871550858020782, + 1.195927779917838e-05, + 1.2490571243972681e-08, + 1.2371740965022582e-08, + 1.015614703447909e-08 + ], + "gamma_dfa": -0.0003672020420708577, + "gamma_dfa_per_layer": [ + -0.002894133096560836, + 0.0016025153454393148, + 2.5333010853501037e-05, + -0.0002025234280154109 + ], + "acc_eval": 0.244140625, + "loss_eval": 2.029773473739624, + "epoch": 44 + }, + { + "hidden_norms_cls": [ + 46.796348571777344, + 30912490.0, + 36281496.0, + 39245376.0, + 41656108.0 + ], + "hidden_norms_avg": [ + 1118.38134765625, + 23560890.0, + 29646840.0, + 34192144.0, + 40581760.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005251869442872703, + 3.3097524720915317e-10, + 2.891357153700369e-10, + 2.8932781170887267e-10, + 2.8904145743524623e-10 + ], + "bp_grad_F": [ + 0.09888254106044769, + 4.3903874029638246e-06, + 1.0815308648659538e-08, + 1.040402519691952e-08, + 1.0228043301196976e-08 + ], + "gamma_dfa": -0.00037046667421236634, + "gamma_dfa_per_layer": [ + -0.002757731592282653, + 0.0014231146778911352, + 3.9225356886163354e-05, + -0.00018647513934411108 + ], + "acc_eval": 0.234375, + "loss_eval": 2.0383992195129395, + "epoch": 45 + }, + { + "hidden_norms_cls": [ + 46.95427322387695, + 31720426.0, + 37234788.0, + 40129424.0, + 42928428.0 + ], + "hidden_norms_avg": [ + 1122.57470703125, + 22583276.0, + 28617936.0, + 33269996.0, + 40462568.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0004295881371945143, + 3.419305394380956e-10, + 2.8222821302215095e-10, + 2.8184321543278656e-10, + 2.8312005517783234e-10 + ], + "bp_grad_F": [ + 0.5366736650466919, + 2.256676998513285e-05, + 1.1914533359913548e-08, + 1.1219233542192342e-08, + 1.004730165732326e-08 + ], + "gamma_dfa": 0.00015619578334735706, + "gamma_dfa_per_layer": [ + -0.0010500183561816812, + 0.001834970898926258, + 3.652184386737645e-05, + -0.00019669125322252512 + ], + "acc_eval": 0.2431640625, + "loss_eval": 2.039337396621704, + "epoch": 46 + }, + { + "hidden_norms_cls": [ + 47.09259796142578, + 32824944.0, + 38341332.0, + 41754692.0, + 44005196.0 + ], + "hidden_norms_avg": [ + 1126.26953125, + 24313264.0, + 30669532.0, + 35293632.0, + 41785680.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005083036958239973, + 3.2607763711389737e-10, + 2.8033902976787317e-10, + 2.7943025671106625e-10, + 2.7980692762774595e-10 + ], + "bp_grad_F": [ + 0.06971772015094757, + 2.262569068989251e-06, + 1.0069017619684928e-08, + 9.845756210324907e-09, + 9.862499261714675e-09 + ], + "gamma_dfa": 0.0007457349984179018, + "gamma_dfa_per_layer": [ + 0.0015201317146420479, + 0.0016060702037066221, + 3.652479063021019e-05, + -0.00017978671530727297 + ], + "acc_eval": 0.25, + "loss_eval": 2.0368857383728027, + "epoch": 47 + }, + { + "hidden_norms_cls": [ + 47.21082305908203, + 33649040.0, + 39036836.0, + 42500816.0, + 44653764.0 + ], + "hidden_norms_avg": [ + 1129.5389404296875, + 24447482.0, + 30689206.0, + 35330028.0, + 42064012.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00046678923536092043, + 3.204189136241098e-10, + 2.755325689829391e-10, + 2.7554050707756517e-10, + 2.7654081802275243e-10 + ], + "bp_grad_F": [ + 0.07496139407157898, + 2.2861270281282486e-06, + 9.997683569906712e-09, + 9.715622084627284e-09, + 9.736355721656764e-09 + ], + "gamma_dfa": 0.0006021281524226652, + "gamma_dfa_per_layer": [ + 0.0007927162805572152, + 0.001763419946655631, + 3.104796996922232e-05, + -0.000178671587491408 + ], + "acc_eval": 0.259765625, + "loss_eval": 2.0324196815490723, + "epoch": 48 + }, + { + "hidden_norms_cls": [ + 47.315826416015625, + 34514608.0, + 40222940.0, + 43606628.0, + 45508540.0 + ], + "hidden_norms_avg": [ + 1132.157958984375, + 24487910.0, + 30821460.0, + 35505456.0, + 42327908.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00041785358916968107, + 3.3263317100740153e-10, + 2.696754763942266e-10, + 2.701733836651954e-10, + 2.71275196750409e-10 + ], + "bp_grad_F": [ + 0.08130628615617752, + 3.3661247016425477e-06, + 1.078094857831502e-08, + 9.643851939244996e-09, + 9.617946439277603e-09 + ], + "gamma_dfa": 0.0005021913366363151, + "gamma_dfa_per_layer": [ + 0.00036606384674087167, + 0.0017899572849273682, + 2.7483671146910638e-05, + -0.00017473945626989007 + ], + "acc_eval": 0.2626953125, + "loss_eval": 2.028841495513916, + "epoch": 49 + }, + { + "hidden_norms_cls": [ + 47.388946533203125, + 35118772.0, + 40722008.0, + 44347232.0, + 46331384.0 + ], + "hidden_norms_avg": [ + 1134.3621826171875, + 24821028.0, + 31074444.0, + 35754528.0, + 42627356.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00043533919961191714, + 3.173189211391758e-10, + 2.687766953446413e-10, + 2.68161992611482e-10, + 2.690903055935223e-10 + ], + "bp_grad_F": [ + 0.09458157420158386, + 1.9965789306297665e-06, + 1.025628915130028e-08, + 9.58252854843522e-09, + 9.523047239667903e-09 + ], + "gamma_dfa": 0.00038969498109509004, + "gamma_dfa_per_layer": [ + -0.00010247333557344973, + 0.0018051008228212595, + 3.0877545214025304e-05, + -0.0001747251080814749 + ], + "acc_eval": 0.2548828125, + "loss_eval": 2.0298049449920654, + "epoch": 50 + }, + { + "hidden_norms_cls": [ + 47.4621696472168, + 34836160.0, + 40469796.0, + 44120564.0, + 46361016.0 + ], + "hidden_norms_avg": [ + 1136.3079833984375, + 24152832.0, + 30436140.0, + 35273740.0, + 42621312.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.000491869286634028, + 3.2774055691575654e-10, + 2.667974174919152e-10, + 2.669577336966711e-10, + 2.680060617876734e-10 + ], + "bp_grad_F": [ + 0.0854402706027031, + 1.6183460047614062e-06, + 9.980491100236577e-09, + 9.461099459429079e-09, + 9.473674289495193e-09 + ], + "gamma_dfa": 0.0005695745421689935, + "gamma_dfa_per_layer": [ + 0.0004214277723804116, + 0.001994946738705039, + 3.653779276646674e-05, + -0.00017461413517594337 + ], + "acc_eval": 0.23828125, + "loss_eval": 2.024928092956543, + "epoch": 51 + }, + { + "hidden_norms_cls": [ + 47.52342224121094, + 36446320.0, + 42468064.0, + 46175108.0, + 47947040.0 + ], + "hidden_norms_avg": [ + 1137.8057861328125, + 25400616.0, + 31703746.0, + 36432956.0, + 43307828.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00034005154157057405, + 3.121023717245208e-10, + 2.6523216956064743e-10, + 2.6489818671926457e-10, + 2.649658548126155e-10 + ], + "bp_grad_F": [ + 0.10125189274549484, + 4.0649438233231194e-06, + 9.830144698241838e-09, + 9.345456852827283e-09, + 9.317212779080819e-09 + ], + "gamma_dfa": 0.0005080709906906122, + "gamma_dfa_per_layer": [ + 0.00023581110872328281, + 0.0019333910895511508, + 3.0449831683654338e-05, + -0.00016736806719563901 + ], + "acc_eval": 0.251953125, + "loss_eval": 2.0343496799468994, + "epoch": 52 + }, + { + "hidden_norms_cls": [ + 47.567466735839844, + 35483484.0, + 41150516.0, + 44742840.0, + 47102380.0 + ], + "hidden_norms_avg": [ + 1139.05029296875, + 25225282.0, + 31585082.0, + 36363108.0, + 43411416.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00046872245729900897, + 3.1678357159670156e-10, + 2.6331647973165673e-10, + 2.64141680750285e-10, + 2.646937946604311e-10 + ], + "bp_grad_F": [ + 0.08037562668323517, + 2.9532295684475685e-06, + 9.44631484145475e-09, + 9.393840372240447e-09, + 9.41009226096412e-09 + ], + "gamma_dfa": 0.0005205844290685491, + "gamma_dfa_per_layer": [ + 0.0003642886003945023, + 0.0018537556752562523, + 3.766161171370186e-05, + -0.00017336817109026015 + ], + "acc_eval": 0.2392578125, + "loss_eval": 2.02734375, + "epoch": 53 + }, + { + "hidden_norms_cls": [ + 47.602928161621094, + 35582280.0, + 41200288.0, + 44686488.0, + 47126900.0 + ], + "hidden_norms_avg": [ + 1139.965087890625, + 25657812.0, + 31984420.0, + 36722060.0, + 43681628.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00047245126916095614, + 3.1708738412739024e-10, + 2.6278745846042284e-10, + 2.6274540876336516e-10, + 2.6386243190401615e-10 + ], + "bp_grad_F": [ + 0.08141150325536728, + 1.0596501851978246e-05, + 1.0370135861137442e-08, + 1.0129613592368969e-08, + 9.407171930320146e-09 + ], + "gamma_dfa": 0.00048223476824205136, + "gamma_dfa_per_layer": [ + 0.00032963097328320146, + 0.00173661345615983, + 3.3767075365176424e-05, + -0.00017107243184000254 + ], + "acc_eval": 0.251953125, + "loss_eval": 2.0290825366973877, + "epoch": 54 + }, + { + "hidden_norms_cls": [ + 47.62870788574219, + 36373308.0, + 42049216.0, + 45803868.0, + 47955172.0 + ], + "hidden_norms_avg": [ + 1140.6492919921875, + 25698864.0, + 31990548.0, + 36707920.0, + 43700336.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0004157091898377985, + 3.1451533044624114e-10, + 2.6160756894100246e-10, + 2.62937283057596e-10, + 2.633904483406724e-10 + ], + "bp_grad_F": [ + 0.28710466623306274, + 2.876764028769685e-06, + 9.405439982401731e-09, + 9.308096515781017e-09, + 9.319982119393444e-09 + ], + "gamma_dfa": 0.00045457817577698734, + "gamma_dfa_per_layer": [ + 0.00011597995762713253, + 0.0018383568385615945, + 3.561873018043116e-05, + -0.00017164282326120883 + ], + "acc_eval": 0.2509765625, + "loss_eval": 2.0309882164001465, + "epoch": 55 + }, + { + "hidden_norms_cls": [ + 47.646034240722656, + 35748556.0, + 41284136.0, + 44861788.0, + 47367128.0 + ], + "hidden_norms_avg": [ + 1141.1492919921875, + 25812252.0, + 32106642.0, + 36789100.0, + 43809728.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0004877470782957971, + 3.1724206595029614e-10, + 2.6251881224403917e-10, + 2.627319750647672e-10, + 2.634734097561875e-10 + ], + "bp_grad_F": [ + 0.08355043828487396, + 2.428784682706464e-06, + 9.579466109244095e-09, + 9.378172016738517e-09, + 9.3862775329967e-09 + ], + "gamma_dfa": 0.0006408758436009521, + "gamma_dfa_per_layer": [ + 0.0009491340606473386, + 0.0017479720991104841, + 4.02485384256579e-05, + -0.00017385132377967238 + ], + "acc_eval": 0.2451171875, + "loss_eval": 2.030231475830078, + "epoch": 56 + }, + { + "hidden_norms_cls": [ + 47.65734100341797, + 35952124.0, + 41526152.0, + 45192104.0, + 47596588.0 + ], + "hidden_norms_avg": [ + 1141.4708251953125, + 25858530.0, + 32145604.0, + 36816776.0, + 43830904.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0004684069426730275, + 3.1406693912217065e-10, + 2.6245111639511265e-10, + 2.628100792545496e-10, + 2.632017381820617e-10 + ], + "bp_grad_F": [ + 0.16165603697299957, + 3.1715710520074936e-06, + 1.017021755700398e-08, + 9.52006651289139e-09, + 9.360165087457517e-09 + ], + "gamma_dfa": 0.0005993541035422822, + "gamma_dfa_per_layer": [ + 0.0007396150031127036, + 0.0017993964720517397, + 3.345730510773137e-05, + -0.00017505236610304564 + ], + "acc_eval": 0.2529296875, + "loss_eval": 2.030412197113037, + "epoch": 57 + }, + { + "hidden_norms_cls": [ + 47.6636962890625, + 36249912.0, + 41976852.0, + 45623612.0, + 47785064.0 + ], + "hidden_norms_avg": [ + 1141.6495361328125, + 25785932.0, + 32065610.0, + 36745760.0, + 43805624.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00047133295447565615, + 3.155541938859585e-10, + 2.617887573386213e-10, + 2.615481442536094e-10, + 2.6276933406954583e-10 + ], + "bp_grad_F": [ + 0.09146469831466675, + 1.8705788988881977e-06, + 9.432087999527994e-09, + 9.323818161988129e-09, + 9.325629157785897e-09 + ], + "gamma_dfa": 0.0005754136018367717, + "gamma_dfa_per_layer": [ + 0.0005892712506465614, + 0.0018461447907611728, + 3.762807318707928e-05, + -0.00017138970724772662 + ], + "acc_eval": 0.251953125, + "loss_eval": 2.031130790710449, + "epoch": 58 + }, + { + "hidden_norms_cls": [ + 47.66678237915039, + 36057328.0, + 41647260.0, + 45391196.0, + 47626604.0 + ], + "hidden_norms_avg": [ + 1141.7391357421875, + 25717830.0, + 31995718.0, + 36684084.0, + 43781820.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.00047832890413701534, + 3.149952798597866e-10, + 2.614737593109595e-10, + 2.6173804790197153e-10, + 2.6257329643897265e-10 + ], + "bp_grad_F": [ + 0.16335515677928925, + 1.6483251101817586e-06, + 9.493151154060797e-09, + 9.393547273361946e-09, + 9.338456230523207e-09 + ], + "gamma_dfa": 0.0005906115802645218, + "gamma_dfa_per_layer": [ + 0.0007219060207717121, + 0.0017734188586473465, + 3.73886723536998e-05, + -0.00017026723071467131 + ], + "acc_eval": 0.255859375, + "loss_eval": 2.0293946266174316, + "epoch": 59 + }, + { + "hidden_norms_cls": [ + 47.666595458984375, + 36029164.0, + 41621316.0, + 45328004.0, + 47606828.0 + ], + "hidden_norms_avg": [ + 1141.736328125, + 25716872.0, + 31991962.0, + 36680384.0, + 43781988.0 + ], + "bp_grad_per_sample_l2_med": [ + 0.0004811616672668606, + 3.1537170097628575e-10, + 2.6223428983840336e-10, + 2.6155200227861997e-10, + 2.6258059615535956e-10 + ], + "bp_grad_F": [ + 0.0951727032661438, + 1.623439857212361e-06, + 9.415242807619961e-09, + 9.343901652414388e-09, + 9.342033813197759e-09 + ], + "gamma_dfa": 0.000660182828141842, + "gamma_dfa_per_layer": [ + 0.0009655870962888002, + 0.0018112185643985868, + 3.598399052862078e-05, + -0.00017205833864863962 + ], + "acc_eval": 0.255859375, + "loss_eval": 2.029409408569336, + "epoch": 60 + } + ] +}
\ No newline at end of file diff --git a/results/snapshot_vit_v1/snapshot_vit_s456.json b/results/snapshot_vit_v1/snapshot_vit_s456.json new file mode 100644 index 0000000..1b43b92 --- /dev/null +++ b/results/snapshot_vit_v1/snapshot_vit_s456.json @@ -0,0 +1,4594 @@ +{ + "config": { + "output_dir": "results/snapshot_vit_v1", + "epochs": 60, + "lr": 0.001, + "wd": 0.05, + "seed": 456, + "depth": 4, + "d_model": 128, + "n_heads": 4 + }, + "depth": 4, + "d_model": 128, + "architecture": "ViTMini", + "bp_log": [ + { + "hidden_norms_cls": [ + 0.3042769432067871, + 3.2273340225219727, + 4.6926493644714355, + 6.003997802734375, + 6.652705669403076 + ], + "hidden_norms_avg": [ + 6.027750492095947, + 7.1044697761535645, + 7.910475730895996, + 8.79189682006836, + 9.458833694458008 + ], + "bp_grad_per_sample_l2_med": [ + 0.005248428788036108, + 0.0012287507997825742, + 0.0010387187357991934, + 0.0009678652859292924, + 0.0009174557635560632 + ], + "bp_grad_F": [ + 0.248850479722023, + 0.0424945093691349, + 0.0348997563123703, + 0.03202275559306145, + 0.030239813029766083 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.0791015625, + "loss_eval": 2.544933319091797, + "epoch": 0 + }, + { + "hidden_norms_cls": [ + 0.47296053171157837, + 11.65749454498291, + 18.3483943939209, + 24.296405792236328, + 29.94684410095215 + ], + "hidden_norms_avg": [ + 4.04442834854126, + 9.927255630493164, + 14.762720108032227, + 19.817049026489258, + 24.69188690185547 + ], + "bp_grad_per_sample_l2_med": [ + 0.0025219416711479425, + 0.00024655580637045205, + 0.00017099862452596426, + 0.0001493149611633271, + 0.00015047143097035587 + ], + "bp_grad_F": [ + 0.11354690045118332, + 0.00995445717126131, + 0.006139951758086681, + 0.005140931811183691, + 0.005088545847684145 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.458984375, + "loss_eval": 1.4817876815795898, + "epoch": 1 + }, + { + "hidden_norms_cls": [ + 0.6174129843711853, + 12.388140678405762, + 20.656904220581055, + 27.073535919189453, + 32.43647003173828 + ], + "hidden_norms_avg": [ + 3.8196825981140137, + 10.982677459716797, + 17.020437240600586, + 22.471168518066406, + 27.064218521118164 + ], + "bp_grad_per_sample_l2_med": [ + 0.0020673249382525682, + 0.0002501676499377936, + 0.00015905944746918976, + 0.00013718422269448638, + 0.0001328626531176269 + ], + "bp_grad_F": [ + 0.09556808322668076, + 0.01051106583327055, + 0.005822391714900732, + 0.004766201134771109, + 0.004577947780489922 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.498046875, + "loss_eval": 1.356250286102295, + "epoch": 2 + }, + { + "hidden_norms_cls": [ + 0.7216586470603943, + 12.527997016906738, + 19.815393447875977, + 26.452791213989258, + 30.71564483642578 + ], + "hidden_norms_avg": [ + 3.8907485008239746, + 11.261369705200195, + 16.61492156982422, + 21.57110023498535, + 24.978900909423828 + ], + "bp_grad_per_sample_l2_med": [ + 0.0021546317730098963, + 0.00030062923906371, + 0.00018509272194933146, + 0.00013865686196368188, + 0.00012920796871185303 + ], + "bp_grad_F": [ + 0.10040326416492462, + 0.012564396485686302, + 0.007171053439378738, + 0.005316051654517651, + 0.004880763124674559 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.5263671875, + "loss_eval": 1.2655537128448486, + "epoch": 3 + }, + { + "hidden_norms_cls": [ + 0.8026084899902344, + 12.023736953735352, + 20.35123062133789, + 26.937381744384766, + 31.122047424316406 + ], + "hidden_norms_avg": [ + 3.9065797328948975, + 11.544925689697266, + 17.131011962890625, + 22.292007446289062, + 25.61225128173828 + ], + "bp_grad_per_sample_l2_med": [ + 0.0019258656539022923, + 0.00027825392317026854, + 0.00017385024693794549, + 0.0001323536125710234, + 0.00012172113929409534 + ], + "bp_grad_F": [ + 0.09238748997449875, + 0.01207020040601492, + 0.007036368828266859, + 0.005208354443311691, + 0.0047498126514256 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.5673828125, + "loss_eval": 1.1375834941864014, + "epoch": 4 + }, + { + "hidden_norms_cls": [ + 0.8784403204917908, + 11.689644813537598, + 20.318578720092773, + 27.47210121154785, + 30.616111755371094 + ], + "hidden_norms_avg": [ + 3.991969585418701, + 11.499222755432129, + 17.010013580322266, + 21.869060516357422, + 24.784456253051758 + ], + "bp_grad_per_sample_l2_med": [ + 0.0017466156277805567, + 0.00030884172883816063, + 0.00018451173673383892, + 0.00013519970525521785, + 0.00011779358464991674 + ], + "bp_grad_F": [ + 0.0881652981042862, + 0.014197942800819874, + 0.008167672902345657, + 0.00570154283195734, + 0.0048713223077356815 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.611328125, + "loss_eval": 1.078152060508728, + "epoch": 5 + }, + { + "hidden_norms_cls": [ + 0.9501535892486572, + 12.59860610961914, + 20.491111755371094, + 27.280519485473633, + 30.350204467773438 + ], + "hidden_norms_avg": [ + 4.104201793670654, + 12.098063468933105, + 17.173715591430664, + 21.416263580322266, + 24.0885066986084 + ], + "bp_grad_per_sample_l2_med": [ + 0.001603498705662787, + 0.0003018827992491424, + 0.00017963100981432945, + 0.00013513887824956328, + 0.00010907831892836839 + ], + "bp_grad_F": [ + 0.08621370792388916, + 0.01517424639314413, + 0.008794959634542465, + 0.006117440760135651, + 0.0050094337202608585 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.6337890625, + "loss_eval": 1.0198562145233154, + "epoch": 6 + }, + { + "hidden_norms_cls": [ + 1.000899076461792, + 12.766633987426758, + 21.096908569335938, + 28.377235412597656, + 31.630756378173828 + ], + "hidden_norms_avg": [ + 4.143934726715088, + 13.046825408935547, + 18.295841217041016, + 23.147098541259766, + 25.70033836364746 + ], + "bp_grad_per_sample_l2_med": [ + 0.0015043691964820027, + 0.00027919537387788296, + 0.0001850408734753728, + 0.0001339056616416201, + 0.00011227549111936241 + ], + "bp_grad_F": [ + 0.07296615093946457, + 0.012746848165988922, + 0.00809843186289072, + 0.005660817492753267, + 0.00492064468562603 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.6357421875, + "loss_eval": 1.0125277042388916, + "epoch": 7 + }, + { + "hidden_norms_cls": [ + 1.0544841289520264, + 10.223775863647461, + 18.59557342529297, + 26.47129249572754, + 29.77008819580078 + ], + "hidden_norms_avg": [ + 4.213057994842529, + 13.021050453186035, + 18.172080993652344, + 22.743032455444336, + 25.225244522094727 + ], + "bp_grad_per_sample_l2_med": [ + 0.001364147407002747, + 0.0002890345931518823, + 0.00017917664081323892, + 0.000133156034280546, + 0.00011372177687007934 + ], + "bp_grad_F": [ + 0.07409710437059402, + 0.015113108791410923, + 0.00921319704502821, + 0.006366679444909096, + 0.005320113617926836 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.6376953125, + "loss_eval": 0.9630928039550781, + "epoch": 8 + }, + { + "hidden_norms_cls": [ + 1.1071640253067017, + 12.073644638061523, + 19.524917602539062, + 26.858781814575195, + 30.318849563598633 + ], + "hidden_norms_avg": [ + 4.280055522918701, + 13.520170211791992, + 18.620624542236328, + 22.658164978027344, + 25.25646209716797 + ], + "bp_grad_per_sample_l2_med": [ + 0.0012990576215088367, + 0.00027366873109713197, + 0.00016998387582134455, + 0.00012336595682427287, + 9.77109739324078e-05 + ], + "bp_grad_F": [ + 0.07662060856819153, + 0.015166237019002438, + 0.0096206646412611, + 0.006477049086242914, + 0.005294579081237316 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.654296875, + "loss_eval": 0.9390516877174377, + "epoch": 9 + }, + { + "hidden_norms_cls": [ + 1.1426520347595215, + 12.005794525146484, + 20.33642578125, + 27.80829429626465, + 31.22698974609375 + ], + "hidden_norms_avg": [ + 4.328567981719971, + 13.725446701049805, + 19.644866943359375, + 23.91346549987793, + 26.409910202026367 + ], + "bp_grad_per_sample_l2_med": [ + 0.0012125002685934305, + 0.00027867915923707187, + 0.00017298922466579825, + 0.0001231397909577936, + 0.00010014672443503514 + ], + "bp_grad_F": [ + 0.06728038191795349, + 0.014830472879111767, + 0.008772061206400394, + 0.006222804542630911, + 0.00518346531316638 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.6708984375, + "loss_eval": 0.8987849950790405, + "epoch": 10 + }, + { + "hidden_norms_cls": [ + 1.182861089706421, + 11.343650817871094, + 19.158103942871094, + 26.537256240844727, + 29.248126983642578 + ], + "hidden_norms_avg": [ + 4.343320846557617, + 13.940053939819336, + 19.19382667541504, + 23.443157196044922, + 25.7342529296875 + ], + "bp_grad_per_sample_l2_med": [ + 0.0012841055868193507, + 0.00029811219428665936, + 0.00019176788919139653, + 0.0001351374085061252, + 0.00010611475590849295 + ], + "bp_grad_F": [ + 0.07373103499412537, + 0.015750709921121597, + 0.010071114636957645, + 0.007257349323481321, + 0.00578929902985692 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.6708984375, + "loss_eval": 0.9137783646583557, + "epoch": 11 + }, + { + "hidden_norms_cls": [ + 1.2363859415054321, + 11.583894729614258, + 18.828346252441406, + 26.30917739868164, + 29.363861083984375 + ], + "hidden_norms_avg": [ + 4.441896915435791, + 13.568662643432617, + 18.774181365966797, + 22.73027801513672, + 24.84675407409668 + ], + "bp_grad_per_sample_l2_med": [ + 0.0011289563262835145, + 0.00029387036920525134, + 0.0001886928512249142, + 0.00013024669897276908, + 9.712533210404217e-05 + ], + "bp_grad_F": [ + 0.06850193440914154, + 0.016727343201637268, + 0.010432050563395023, + 0.00744031835347414, + 0.0057919141836464405 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.6865234375, + "loss_eval": 0.8655319809913635, + "epoch": 12 + }, + { + "hidden_norms_cls": [ + 1.276246428489685, + 11.161674499511719, + 18.18111801147461, + 24.498512268066406, + 28.472366333007812 + ], + "hidden_norms_avg": [ + 4.438092231750488, + 13.71449089050293, + 18.358896255493164, + 21.998794555664062, + 24.44672966003418 + ], + "bp_grad_per_sample_l2_med": [ + 0.001100392546504736, + 0.0002873367629945278, + 0.00018384194117970765, + 0.0001337265275651589, + 9.951705578714609e-05 + ], + "bp_grad_F": [ + 0.06749056279659271, + 0.01716373674571514, + 0.010762249119579792, + 0.007881316356360912, + 0.005892573855817318 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.6982421875, + "loss_eval": 0.8163608908653259, + "epoch": 13 + }, + { + "hidden_norms_cls": [ + 1.3131792545318604, + 11.034884452819824, + 17.936412811279297, + 24.541194915771484, + 28.02472496032715 + ], + "hidden_norms_avg": [ + 4.504117012023926, + 13.86273193359375, + 18.456012725830078, + 22.296085357666016, + 24.57513427734375 + ], + "bp_grad_per_sample_l2_med": [ + 0.0010489820269867778, + 0.0002784461830742657, + 0.00018129698582924902, + 0.00013628850865643471, + 0.00010334269609302282 + ], + "bp_grad_F": [ + 0.06553377211093903, + 0.016712775453925133, + 0.010981550440192223, + 0.007675213273614645, + 0.006061363499611616 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.69921875, + "loss_eval": 0.8088906407356262, + "epoch": 14 + }, + { + "hidden_norms_cls": [ + 1.3631325960159302, + 10.72367000579834, + 17.147600173950195, + 23.750171661376953, + 26.97018051147461 + ], + "hidden_norms_avg": [ + 4.5338335037231445, + 13.55031967163086, + 17.503305435180664, + 21.003847122192383, + 23.168630599975586 + ], + "bp_grad_per_sample_l2_med": [ + 0.0010022176429629326, + 0.0002914569922722876, + 0.00018890199135057628, + 0.00013593978655990213, + 0.00010251985804643482 + ], + "bp_grad_F": [ + 0.06372744590044022, + 0.017576243728399277, + 0.011715950444340706, + 0.008481587283313274, + 0.006441249046474695 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.701171875, + "loss_eval": 0.815929651260376, + "epoch": 15 + }, + { + "hidden_norms_cls": [ + 1.3926054239273071, + 10.609240531921387, + 17.234729766845703, + 23.659292221069336, + 27.24931526184082 + ], + "hidden_norms_avg": [ + 4.513574600219727, + 13.48448371887207, + 17.66766929626465, + 21.098644256591797, + 23.213909149169922 + ], + "bp_grad_per_sample_l2_med": [ + 0.0008784966776147485, + 0.0002452125190757215, + 0.0001629967155167833, + 0.00012193062138976529, + 8.820713992463425e-05 + ], + "bp_grad_F": [ + 0.06515656411647797, + 0.017664309591054916, + 0.011493389494717121, + 0.008405888453125954, + 0.006338838487863541 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.71875, + "loss_eval": 0.7755489349365234, + "epoch": 16 + }, + { + "hidden_norms_cls": [ + 1.4288005828857422, + 9.991934776306152, + 16.30902671813965, + 22.53181266784668, + 25.98328971862793 + ], + "hidden_norms_avg": [ + 4.606818199157715, + 13.761014938354492, + 18.061887741088867, + 21.805465698242188, + 23.856918334960938 + ], + "bp_grad_per_sample_l2_med": [ + 0.0008885160204954445, + 0.00026409278507344425, + 0.0001749310758896172, + 0.00012756467913277447, + 8.965048618847504e-05 + ], + "bp_grad_F": [ + 0.07283030450344086, + 0.01973560079932213, + 0.013354410417377949, + 0.009586230851709843, + 0.006897584535181522 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.693359375, + "loss_eval": 0.8082709908485413, + "epoch": 17 + }, + { + "hidden_norms_cls": [ + 1.4623849391937256, + 9.794211387634277, + 16.102649688720703, + 21.62982940673828, + 25.186201095581055 + ], + "hidden_norms_avg": [ + 4.6317901611328125, + 13.527764320373535, + 17.6087589263916, + 21.295202255249023, + 23.361501693725586 + ], + "bp_grad_per_sample_l2_med": [ + 0.0008634055848233402, + 0.0002727442770265043, + 0.00018466924666427076, + 0.00013278424739837646, + 8.995016105473042e-05 + ], + "bp_grad_F": [ + 0.06913765519857407, + 0.02126000076532364, + 0.014007972553372383, + 0.010056250728666782, + 0.00732502993196249 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.720703125, + "loss_eval": 0.7786685824394226, + "epoch": 18 + }, + { + "hidden_norms_cls": [ + 1.4771180152893066, + 9.776408195495605, + 15.485087394714355, + 21.054513931274414, + 25.07387351989746 + ], + "hidden_norms_avg": [ + 4.570744514465332, + 13.13935375213623, + 16.70624351501465, + 20.002471923828125, + 22.400487899780273 + ], + "bp_grad_per_sample_l2_med": [ + 0.0008809147984720767, + 0.00026235057157464325, + 0.0001801249454729259, + 0.0001285924663534388, + 8.227767830248922e-05 + ], + "bp_grad_F": [ + 0.07356031239032745, + 0.02142982743680477, + 0.014684871770441532, + 0.010477974079549313, + 0.007170504424721003 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7412109375, + "loss_eval": 0.7548146843910217, + "epoch": 19 + }, + { + "hidden_norms_cls": [ + 1.4932866096496582, + 9.586905479431152, + 15.190006256103516, + 20.53044891357422, + 23.871917724609375 + ], + "hidden_norms_avg": [ + 4.640237808227539, + 13.333914756774902, + 16.603134155273438, + 19.70735740661621, + 21.712968826293945 + ], + "bp_grad_per_sample_l2_med": [ + 0.0008201555465348065, + 0.00025074987206608057, + 0.00018013773660641164, + 0.0001373633131152019, + 8.871335012372583e-05 + ], + "bp_grad_F": [ + 0.07453557848930359, + 0.02180560864508152, + 0.015117364004254341, + 0.010920760221779346, + 0.007648298051208258 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7294921875, + "loss_eval": 0.7338082790374756, + "epoch": 20 + }, + { + "hidden_norms_cls": [ + 1.5380369424819946, + 8.941458702087402, + 14.383026123046875, + 19.64396858215332, + 23.333805084228516 + ], + "hidden_norms_avg": [ + 4.57804012298584, + 12.777068138122559, + 15.852535247802734, + 19.008466720581055, + 21.123043060302734 + ], + "bp_grad_per_sample_l2_med": [ + 0.0007403296767733991, + 0.00023140048142522573, + 0.0001578826195327565, + 0.00011260491737630218, + 7.754044054308906e-05 + ], + "bp_grad_F": [ + 0.07555028051137924, + 0.024111615493893623, + 0.01656532846391201, + 0.01178746297955513, + 0.007888367399573326 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.744140625, + "loss_eval": 0.7344021201133728, + "epoch": 21 + }, + { + "hidden_norms_cls": [ + 1.5561290979385376, + 9.454068183898926, + 14.400370597839355, + 19.253843307495117, + 22.27718734741211 + ], + "hidden_norms_avg": [ + 4.687749862670898, + 12.994861602783203, + 15.969095230102539, + 18.976381301879883, + 20.873851776123047 + ], + "bp_grad_per_sample_l2_med": [ + 0.0008383935783058405, + 0.0002680755569599569, + 0.000193025975022465, + 0.00013935549941379577, + 9.83258432825096e-05 + ], + "bp_grad_F": [ + 0.06634547561407089, + 0.021603766828775406, + 0.01579204760491848, + 0.011702721007168293, + 0.008362310007214546 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.734375, + "loss_eval": 0.7299590110778809, + "epoch": 22 + }, + { + "hidden_norms_cls": [ + 1.5774884223937988, + 8.741745948791504, + 13.803518295288086, + 18.90570831298828, + 22.218149185180664 + ], + "hidden_norms_avg": [ + 4.594933032989502, + 12.66419506072998, + 15.714310646057129, + 19.085254669189453, + 21.135629653930664 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006682098028250039, + 0.00022848948719911277, + 0.00016414522542618215, + 0.0001115989507525228, + 7.056116737658158e-05 + ], + "bp_grad_F": [ + 0.07456765323877335, + 0.024261610582470894, + 0.01710386760532856, + 0.012147205881774426, + 0.008177526295185089 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7763671875, + "loss_eval": 0.699420154094696, + "epoch": 23 + }, + { + "hidden_norms_cls": [ + 1.576693058013916, + 8.762345314025879, + 13.443113327026367, + 18.230920791625977, + 22.07720375061035 + ], + "hidden_norms_avg": [ + 4.667433738708496, + 12.606098175048828, + 15.544108390808105, + 18.674053192138672, + 20.65712547302246 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006520473980344832, + 0.00022687950695399195, + 0.0001629517209948972, + 0.00011739470210159197, + 7.377319707302377e-05 + ], + "bp_grad_F": [ + 0.08037886768579483, + 0.026545122265815735, + 0.0190938338637352, + 0.013439938426017761, + 0.00869175884872675 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7353515625, + "loss_eval": 0.733207106590271, + "epoch": 24 + }, + { + "hidden_norms_cls": [ + 1.5915056467056274, + 8.431538581848145, + 13.075739860534668, + 17.782840728759766, + 20.799991607666016 + ], + "hidden_norms_avg": [ + 4.626980304718018, + 12.152971267700195, + 14.936114311218262, + 18.13477325439453, + 19.853097915649414 + ], + "bp_grad_per_sample_l2_med": [ + 0.0007118558278307319, + 0.00024262107035610825, + 0.0001753592659952119, + 0.00012127246736781672, + 7.87421886343509e-05 + ], + "bp_grad_F": [ + 0.0734892264008522, + 0.025602566078305244, + 0.018658706918358803, + 0.013666817918419838, + 0.009108936414122581 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7578125, + "loss_eval": 0.6893049478530884, + "epoch": 25 + }, + { + "hidden_norms_cls": [ + 1.589570164680481, + 8.064777374267578, + 12.609880447387695, + 17.195837020874023, + 20.27288818359375 + ], + "hidden_norms_avg": [ + 4.651576995849609, + 12.02392578125, + 14.5856351852417, + 17.578563690185547, + 19.327190399169922 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006660808576270938, + 0.0002373493480263278, + 0.0001695255923550576, + 0.00012064678594470024, + 7.89113255450502e-05 + ], + "bp_grad_F": [ + 0.08072065562009811, + 0.02827218733727932, + 0.020335549488663673, + 0.014427169226109982, + 0.009556911885738373 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.755859375, + "loss_eval": 0.7047942876815796, + "epoch": 26 + }, + { + "hidden_norms_cls": [ + 1.6077263355255127, + 7.734760284423828, + 11.882841110229492, + 15.991426467895508, + 19.111549377441406 + ], + "hidden_norms_avg": [ + 4.672310829162598, + 11.994673728942871, + 14.48837661743164, + 16.922487258911133, + 18.498437881469727 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005716760642826557, + 0.00020494087948463857, + 0.0001548439176985994, + 0.00011144367454107851, + 6.537805165862665e-05 + ], + "bp_grad_F": [ + 0.08237622678279877, + 0.03036416694521904, + 0.022357620298862457, + 0.01651921682059765, + 0.010484294034540653 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7548828125, + "loss_eval": 0.6863161325454712, + "epoch": 27 + }, + { + "hidden_norms_cls": [ + 1.6167913675308228, + 7.661223411560059, + 11.857555389404297, + 16.11384391784668, + 19.207155227661133 + ], + "hidden_norms_avg": [ + 4.6568603515625, + 11.916447639465332, + 14.127670288085938, + 16.765464782714844, + 18.548715591430664 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006015675025992095, + 0.00022187062131706625, + 0.00015665855607949197, + 0.00011372221342753619, + 6.673130701528862e-05 + ], + "bp_grad_F": [ + 0.08522997796535492, + 0.03222256526350975, + 0.023626696318387985, + 0.01693013124167919, + 0.010611210018396378 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.75390625, + "loss_eval": 0.6998279094696045, + "epoch": 28 + }, + { + "hidden_norms_cls": [ + 1.6032800674438477, + 7.3315229415893555, + 11.458590507507324, + 15.827820777893066, + 19.03597640991211 + ], + "hidden_norms_avg": [ + 4.592748641967773, + 11.431793212890625, + 13.694733619689941, + 16.380725860595703, + 17.913341522216797 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006015695980750024, + 0.00023376141325570643, + 0.00015724304830655456, + 0.00010976113844662905, + 6.412796210497618e-05 + ], + "bp_grad_F": [ + 0.09064542502164841, + 0.03456525132060051, + 0.02491987682878971, + 0.01721413992345333, + 0.010797572322189808 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7734375, + "loss_eval": 0.6845672130584717, + "epoch": 29 + }, + { + "hidden_norms_cls": [ + 1.6067036390304565, + 7.343894004821777, + 11.072548866271973, + 15.027584075927734, + 17.731834411621094 + ], + "hidden_norms_avg": [ + 4.656601428985596, + 11.393535614013672, + 13.465471267700195, + 15.975980758666992, + 17.53518295288086 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005384382093325257, + 0.00021189975086599588, + 0.0001552984758745879, + 0.00010658086830517277, + 6.326750008156523e-05 + ], + "bp_grad_F": [ + 0.08886278420686722, + 0.03500310331583023, + 0.025932280346751213, + 0.018638672307133675, + 0.011571940965950489 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7666015625, + "loss_eval": 0.6816176772117615, + "epoch": 30 + }, + { + "hidden_norms_cls": [ + 1.6034201383590698, + 7.28795051574707, + 10.777915000915527, + 14.885525703430176, + 17.61635398864746 + ], + "hidden_norms_avg": [ + 4.614856719970703, + 11.156272888183594, + 13.103291511535645, + 15.514464378356934, + 16.799636840820312 + ], + "bp_grad_per_sample_l2_med": [ + 0.0005928887985646725, + 0.0002398774668108672, + 0.00018353872292209417, + 0.00012629387492779642, + 8.196983981179073e-05 + ], + "bp_grad_F": [ + 0.09026212245225906, + 0.03624382242560387, + 0.027555860579013824, + 0.019604889675974846, + 0.012266743928194046 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7578125, + "loss_eval": 0.6996170878410339, + "epoch": 31 + }, + { + "hidden_norms_cls": [ + 1.611714243888855, + 7.183743953704834, + 10.724074363708496, + 14.5570650100708, + 17.022544860839844 + ], + "hidden_norms_avg": [ + 4.623075485229492, + 11.19589614868164, + 13.009132385253906, + 15.365489959716797, + 16.62962532043457 + ], + "bp_grad_per_sample_l2_med": [ + 0.00036223753704689443, + 0.00014417112106457353, + 0.00010984936670865864, + 8.023829286685213e-05, + 4.624977009370923e-05 + ], + "bp_grad_F": [ + 0.09605100005865097, + 0.03848516568541527, + 0.029046298936009407, + 0.021008647978305817, + 0.012804670259356499 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7666015625, + "loss_eval": 0.6707476377487183, + "epoch": 32 + }, + { + "hidden_norms_cls": [ + 1.6100214719772339, + 6.872183322906494, + 10.294036865234375, + 14.205705642700195, + 17.145418167114258 + ], + "hidden_norms_avg": [ + 4.6112518310546875, + 11.086106300354004, + 12.928976058959961, + 15.350056648254395, + 16.69057846069336 + ], + "bp_grad_per_sample_l2_med": [ + 0.00041789902024902403, + 0.00017518684035167098, + 0.0001297849084949121, + 8.867425640346482e-05, + 4.886053648078814e-05 + ], + "bp_grad_F": [ + 0.10439282655715942, + 0.04290134832262993, + 0.03223341330885887, + 0.02269606664776802, + 0.013486144132912159 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.759765625, + "loss_eval": 0.7303465604782104, + "epoch": 33 + }, + { + "hidden_norms_cls": [ + 1.5996708869934082, + 6.756699085235596, + 9.930717468261719, + 13.55428695678711, + 16.25560188293457 + ], + "hidden_norms_avg": [ + 4.561036109924316, + 10.72261905670166, + 12.331817626953125, + 14.622406005859375, + 16.10407829284668 + ], + "bp_grad_per_sample_l2_med": [ + 0.0004882134380750358, + 0.0002138838026439771, + 0.00016901774506550282, + 0.0001125320850405842, + 6.579834007425234e-05 + ], + "bp_grad_F": [ + 0.10555483400821686, + 0.043922487646341324, + 0.03336327522993088, + 0.02399253286421299, + 0.013897648081183434 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7744140625, + "loss_eval": 0.71100252866745, + "epoch": 34 + }, + { + "hidden_norms_cls": [ + 1.5904232263565063, + 6.662050247192383, + 9.66409683227539, + 13.113885879516602, + 15.527644157409668 + ], + "hidden_norms_avg": [ + 4.5454936027526855, + 10.621905326843262, + 12.306397438049316, + 14.576898574829102, + 15.954529762268066 + ], + "bp_grad_per_sample_l2_med": [ + 0.0004819612659048289, + 0.00020982844580430537, + 0.00015760491078253835, + 0.00010895934246946126, + 5.958566543995403e-05 + ], + "bp_grad_F": [ + 0.10991083830595016, + 0.047212228178977966, + 0.03634340688586235, + 0.02575613744556904, + 0.014786512590944767 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7705078125, + "loss_eval": 0.7015272378921509, + "epoch": 35 + }, + { + "hidden_norms_cls": [ + 1.5926766395568848, + 6.752022743225098, + 9.809683799743652, + 13.14870548248291, + 15.66618537902832 + ], + "hidden_norms_avg": [ + 4.554390907287598, + 10.59433650970459, + 12.38070011138916, + 14.597192764282227, + 15.870081901550293 + ], + "bp_grad_per_sample_l2_med": [ + 0.00043442551395855844, + 0.0001767174544511363, + 0.0001353291590930894, + 9.88640749710612e-05, + 5.115106250741519e-05 + ], + "bp_grad_F": [ + 0.11412353813648224, + 0.04936130344867706, + 0.03786878660321236, + 0.027230119332671165, + 0.015144454315304756 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7734375, + "loss_eval": 0.7246569395065308, + "epoch": 36 + }, + { + "hidden_norms_cls": [ + 1.5873360633850098, + 6.424077987670898, + 9.43139934539795, + 12.677597999572754, + 15.172491073608398 + ], + "hidden_norms_avg": [ + 4.4964070320129395, + 10.287262916564941, + 11.751147270202637, + 13.867583274841309, + 15.24795150756836 + ], + "bp_grad_per_sample_l2_med": [ + 0.0004166838189121336, + 0.0001985515555134043, + 0.00014922211994417012, + 0.0001073441599146463, + 5.765057721873745e-05 + ], + "bp_grad_F": [ + 0.11522030830383301, + 0.05126715078949928, + 0.039791759103536606, + 0.028239542618393898, + 0.015849631279706955 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.76953125, + "loss_eval": 0.7268451452255249, + "epoch": 37 + }, + { + "hidden_norms_cls": [ + 1.5890607833862305, + 6.295929431915283, + 9.119135856628418, + 12.20608901977539, + 14.642122268676758 + ], + "hidden_norms_avg": [ + 4.5021796226501465, + 10.095436096191406, + 11.7933988571167, + 13.924290657043457, + 15.25149917602539 + ], + "bp_grad_per_sample_l2_med": [ + 0.0004105751868337393, + 0.00017681249300949275, + 0.00014306302182376385, + 9.77686358965002e-05, + 5.147796036908403e-05 + ], + "bp_grad_F": [ + 0.12662801146507263, + 0.057947222143411636, + 0.044759348034858704, + 0.03212600201368332, + 0.017064588144421577 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7587890625, + "loss_eval": 0.7464388012886047, + "epoch": 38 + }, + { + "hidden_norms_cls": [ + 1.5761758089065552, + 6.18004846572876, + 9.069403648376465, + 12.180702209472656, + 14.516327857971191 + ], + "hidden_norms_avg": [ + 4.492326736450195, + 9.971968650817871, + 11.50478744506836, + 13.5086669921875, + 14.760706901550293 + ], + "bp_grad_per_sample_l2_med": [ + 0.0003273288893979043, + 0.00015071693633217365, + 0.00011665018246276304, + 8.30927092465572e-05, + 4.371673639980145e-05 + ], + "bp_grad_F": [ + 0.1212363913655281, + 0.055609479546546936, + 0.04309426248073578, + 0.0311796423047781, + 0.016882745549082756 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7783203125, + "loss_eval": 0.7176246643066406, + "epoch": 39 + }, + { + "hidden_norms_cls": [ + 1.5679142475128174, + 6.109372615814209, + 8.818140029907227, + 11.853668212890625, + 14.200343132019043 + ], + "hidden_norms_avg": [ + 4.478363037109375, + 9.777668952941895, + 11.217857360839844, + 13.170299530029297, + 14.35566520690918 + ], + "bp_grad_per_sample_l2_med": [ + 0.0003435175749473274, + 0.00015563883061986417, + 0.00011927648301934823, + 8.298752072732896e-05, + 4.157157673034817e-05 + ], + "bp_grad_F": [ + 0.13136376440525055, + 0.06076597049832344, + 0.046719472855329514, + 0.033368490636348724, + 0.01777712069451809 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.78125, + "loss_eval": 0.7149413228034973, + "epoch": 40 + }, + { + "hidden_norms_cls": [ + 1.5576080083847046, + 6.149768829345703, + 8.711091041564941, + 11.498035430908203, + 13.903903007507324 + ], + "hidden_norms_avg": [ + 4.435859680175781, + 9.742182731628418, + 10.966800689697266, + 12.826833724975586, + 13.987249374389648 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002820935333147645, + 0.00013546310947276652, + 0.00010691316856537014, + 7.935341272968799e-05, + 4.027710747322999e-05 + ], + "bp_grad_F": [ + 0.1424364149570465, + 0.0660286620259285, + 0.05190045386552811, + 0.037452246993780136, + 0.019488148391246796 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.759765625, + "loss_eval": 0.7794969081878662, + "epoch": 41 + }, + { + "hidden_norms_cls": [ + 1.55929696559906, + 6.027406215667725, + 8.56254768371582, + 11.315313339233398, + 13.816405296325684 + ], + "hidden_norms_avg": [ + 4.444095611572266, + 9.598461151123047, + 11.017721176147461, + 12.939472198486328, + 14.150469779968262 + ], + "bp_grad_per_sample_l2_med": [ + 0.00028881963226012886, + 0.00014081982953939587, + 0.00010864014620892704, + 7.903494406491518e-05, + 3.460505831753835e-05 + ], + "bp_grad_F": [ + 0.14435029029846191, + 0.06792720407247543, + 0.05288391932845116, + 0.03809889778494835, + 0.0192741546779871 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7783203125, + "loss_eval": 0.7500836253166199, + "epoch": 42 + }, + { + "hidden_norms_cls": [ + 1.5569288730621338, + 5.83992338180542, + 8.285759925842285, + 11.15902328491211, + 13.44404125213623 + ], + "hidden_norms_avg": [ + 4.416918754577637, + 9.525116920471191, + 10.900503158569336, + 12.753186225891113, + 13.810358047485352 + ], + "bp_grad_per_sample_l2_med": [ + 0.0003584402147680521, + 0.00016757070261519402, + 0.000135006892378442, + 8.858168439473957e-05, + 4.690655987360515e-05 + ], + "bp_grad_F": [ + 0.14751866459846497, + 0.07057010382413864, + 0.05518123507499695, + 0.039941366761922836, + 0.019994540140032768 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.783203125, + "loss_eval": 0.7740926146507263, + "epoch": 43 + }, + { + "hidden_norms_cls": [ + 1.5507482290267944, + 5.7182297706604, + 8.143424034118652, + 10.900334358215332, + 13.179159164428711 + ], + "hidden_norms_avg": [ + 4.415921688079834, + 9.42380142211914, + 10.672327995300293, + 12.439417839050293, + 13.481410026550293 + ], + "bp_grad_per_sample_l2_med": [ + 0.0002743537479545921, + 0.00013802653120364994, + 0.00010440178448334336, + 7.395831926260144e-05, + 3.546295192791149e-05 + ], + "bp_grad_F": [ + 0.15799099206924438, + 0.0763181671500206, + 0.060047682374715805, + 0.043460771441459656, + 0.02125636488199234 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.78125, + "loss_eval": 0.7693630456924438, + "epoch": 44 + }, + { + "hidden_norms_cls": [ + 1.5424573421478271, + 5.688177585601807, + 7.995815753936768, + 10.642090797424316, + 12.71364688873291 + ], + "hidden_norms_avg": [ + 4.399099349975586, + 9.228434562683105, + 10.541167259216309, + 12.30521297454834, + 13.311983108520508 + ], + "bp_grad_per_sample_l2_med": [ + 0.00024789522285573184, + 0.0001290303043788299, + 9.614775626687333e-05, + 6.680067599518225e-05, + 3.041914533241652e-05 + ], + "bp_grad_F": [ + 0.16617916524410248, + 0.08274275809526443, + 0.06457101553678513, + 0.04608331620693207, + 0.02214660868048668 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.77734375, + "loss_eval": 0.7963705062866211, + "epoch": 45 + }, + { + "hidden_norms_cls": [ + 1.5373201370239258, + 5.651702880859375, + 7.870075702667236, + 10.383352279663086, + 12.742950439453125 + ], + "hidden_norms_avg": [ + 4.359274387359619, + 9.120989799499512, + 10.328361511230469, + 11.979691505432129, + 13.047332763671875 + ], + "bp_grad_per_sample_l2_med": [ + 0.00020789829432033002, + 0.00010236673551844433, + 8.31241559353657e-05, + 5.5589163821423426e-05, + 2.5046059818123467e-05 + ], + "bp_grad_F": [ + 0.1711728274822235, + 0.08452030271291733, + 0.06641850620508194, + 0.048034753650426865, + 0.022233933210372925 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.796875, + "loss_eval": 0.787158727645874, + "epoch": 46 + }, + { + "hidden_norms_cls": [ + 1.534119725227356, + 5.64213752746582, + 7.87826681137085, + 10.351583480834961, + 12.802285194396973 + ], + "hidden_norms_avg": [ + 4.381091594696045, + 9.210761070251465, + 10.374320030212402, + 12.017292976379395, + 13.094472885131836 + ], + "bp_grad_per_sample_l2_med": [ + 0.00020434240286704153, + 9.69645261648111e-05, + 7.39863608032465e-05, + 5.592328670900315e-05, + 2.489626240276266e-05 + ], + "bp_grad_F": [ + 0.17339713871479034, + 0.08632633090019226, + 0.06861955672502518, + 0.05006933584809303, + 0.022428762167692184 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7919921875, + "loss_eval": 0.7851632833480835, + "epoch": 47 + }, + { + "hidden_norms_cls": [ + 1.5289297103881836, + 5.52672004699707, + 7.675796985626221, + 10.204718589782715, + 12.873387336730957 + ], + "hidden_norms_avg": [ + 4.3382954597473145, + 8.994754791259766, + 10.112909317016602, + 11.741432189941406, + 12.863862037658691 + ], + "bp_grad_per_sample_l2_med": [ + 0.0001689869095571339, + 8.723067730898038e-05, + 6.962454062886536e-05, + 4.9918842705665156e-05, + 2.0077559383935295e-05 + ], + "bp_grad_F": [ + 0.1893138736486435, + 0.09529414772987366, + 0.07472710311412811, + 0.053626082837581635, + 0.023708311840891838 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7802734375, + "loss_eval": 0.8306612968444824, + "epoch": 48 + }, + { + "hidden_norms_cls": [ + 1.5282790660858154, + 5.412533760070801, + 7.559310436248779, + 10.040712356567383, + 12.344849586486816 + ], + "hidden_norms_avg": [ + 4.344850063323975, + 8.943357467651367, + 10.036619186401367, + 11.63318157196045, + 12.690427780151367 + ], + "bp_grad_per_sample_l2_med": [ + 0.00015994974819477648, + 8.589095523348078e-05, + 6.632162694586441e-05, + 4.7776728024473414e-05, + 1.923311901919078e-05 + ], + "bp_grad_F": [ + 0.19004860520362854, + 0.09673094749450684, + 0.07585947215557098, + 0.05484768748283386, + 0.02464146539568901 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7880859375, + "loss_eval": 0.8193543553352356, + "epoch": 49 + }, + { + "hidden_norms_cls": [ + 1.5247719287872314, + 5.414742469787598, + 7.496610164642334, + 9.884673118591309, + 12.301740646362305 + ], + "hidden_norms_avg": [ + 4.334397792816162, + 8.873777389526367, + 9.937097549438477, + 11.524014472961426, + 12.564128875732422 + ], + "bp_grad_per_sample_l2_med": [ + 0.00016137116472236812, + 8.459146192763001e-05, + 7.113611354725435e-05, + 4.750298467115499e-05, + 1.9900418919860385e-05 + ], + "bp_grad_F": [ + 0.1871533840894699, + 0.09687218815088272, + 0.0768747329711914, + 0.05509059876203537, + 0.024593451991677284 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.80078125, + "loss_eval": 0.810278058052063, + "epoch": 50 + }, + { + "hidden_norms_cls": [ + 1.5175530910491943, + 5.423762798309326, + 7.493518352508545, + 9.848480224609375, + 12.152653694152832 + ], + "hidden_norms_avg": [ + 4.333771228790283, + 8.846830368041992, + 9.895172119140625, + 11.470880508422852, + 12.51230239868164 + ], + "bp_grad_per_sample_l2_med": [ + 0.00018688049749471247, + 9.502686589257792e-05, + 7.833576091798022e-05, + 5.505367516889237e-05, + 2.1109613953740336e-05 + ], + "bp_grad_F": [ + 0.19925038516521454, + 0.10273440927267075, + 0.08146088570356369, + 0.05800405889749527, + 0.025326546281576157 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.79296875, + "loss_eval": 0.8315975069999695, + "epoch": 51 + }, + { + "hidden_norms_cls": [ + 1.5154390335083008, + 5.332769870758057, + 7.351593971252441, + 9.726226806640625, + 12.217824935913086 + ], + "hidden_norms_avg": [ + 4.312089443206787, + 8.77189826965332, + 9.80828857421875, + 11.376079559326172, + 12.424320220947266 + ], + "bp_grad_per_sample_l2_med": [ + 0.0001625730947125703, + 8.300685294670984e-05, + 6.064521585358307e-05, + 4.498598718782887e-05, + 1.878884904726874e-05 + ], + "bp_grad_F": [ + 0.20003639161586761, + 0.10495735704898834, + 0.08297032862901688, + 0.059033025056123734, + 0.02547283098101616 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7890625, + "loss_eval": 0.8281288146972656, + "epoch": 52 + }, + { + "hidden_norms_cls": [ + 1.5136713981628418, + 5.367674827575684, + 7.355134963989258, + 9.682082176208496, + 12.047696113586426 + ], + "hidden_norms_avg": [ + 4.316784381866455, + 8.783827781677246, + 9.812792778015137, + 11.395760536193848, + 12.420063018798828 + ], + "bp_grad_per_sample_l2_med": [ + 0.00016071743448264897, + 7.887884567026049e-05, + 6.097723598941229e-05, + 4.243743751430884e-05, + 1.8102018657373264e-05 + ], + "bp_grad_F": [ + 0.20199808478355408, + 0.10541712492704391, + 0.0835103914141655, + 0.05967256799340248, + 0.025771932676434517 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7890625, + "loss_eval": 0.8389135599136353, + "epoch": 53 + }, + { + "hidden_norms_cls": [ + 1.5129119157791138, + 5.326069355010986, + 7.275172233581543, + 9.633310317993164, + 12.064473152160645 + ], + "hidden_norms_avg": [ + 4.308472633361816, + 8.717856407165527, + 9.717677116394043, + 11.28734016418457, + 12.3176851272583 + ], + "bp_grad_per_sample_l2_med": [ + 0.00015608215471729636, + 8.108095789793879e-05, + 6.294465129030868e-05, + 4.2116138502024114e-05, + 1.7733293134369887e-05 + ], + "bp_grad_F": [ + 0.20727074146270752, + 0.10958535224199295, + 0.08667060732841492, + 0.06113011762499809, + 0.026037005707621574 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.787109375, + "loss_eval": 0.8504288792610168, + "epoch": 54 + }, + { + "hidden_norms_cls": [ + 1.5120813846588135, + 5.313214302062988, + 7.253427982330322, + 9.583664894104004, + 12.01283073425293 + ], + "hidden_norms_avg": [ + 4.303354263305664, + 8.709893226623535, + 9.714727401733398, + 11.287162780761719, + 12.301621437072754 + ], + "bp_grad_per_sample_l2_med": [ + 0.00014261712203733623, + 8.006545249372721e-05, + 6.104232306824997e-05, + 4.1361305193277076e-05, + 1.670202073000837e-05 + ], + "bp_grad_F": [ + 0.20587928593158722, + 0.10970292240381241, + 0.0869779959321022, + 0.06163511052727699, + 0.02630491554737091 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7919921875, + "loss_eval": 0.8500803709030151, + "epoch": 55 + }, + { + "hidden_norms_cls": [ + 1.5113301277160645, + 5.33270263671875, + 7.255039215087891, + 9.560588836669922, + 11.935233116149902 + ], + "hidden_norms_avg": [ + 4.311485767364502, + 8.737537384033203, + 9.710383415222168, + 11.27385425567627, + 12.309149742126465 + ], + "bp_grad_per_sample_l2_med": [ + 0.00014448253205046058, + 7.629570609424263e-05, + 6.040706284693442e-05, + 4.236707536620088e-05, + 1.7641770682530478e-05 + ], + "bp_grad_F": [ + 0.20638784766197205, + 0.10957971960306168, + 0.08719436824321747, + 0.06181200593709946, + 0.026369713246822357 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7900390625, + "loss_eval": 0.8525989055633545, + "epoch": 56 + }, + { + "hidden_norms_cls": [ + 1.5115892887115479, + 5.313480377197266, + 7.247244834899902, + 9.550230979919434, + 11.989883422851562 + ], + "hidden_norms_avg": [ + 4.3022141456604, + 8.688987731933594, + 9.685582160949707, + 11.234566688537598, + 12.270132064819336 + ], + "bp_grad_per_sample_l2_med": [ + 0.0001483379164710641, + 7.948598067741841e-05, + 6.1946535424795e-05, + 4.4680935388896614e-05, + 1.6852744010975584e-05 + ], + "bp_grad_F": [ + 0.21088635921478271, + 0.11181779950857162, + 0.08855349570512772, + 0.06269123405218124, + 0.026410162448883057 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.7890625, + "loss_eval": 0.8474454879760742, + "epoch": 57 + }, + { + "hidden_norms_cls": [ + 1.5110703706741333, + 5.306856632232666, + 7.230412483215332, + 9.542105674743652, + 11.942556381225586 + ], + "hidden_norms_avg": [ + 4.305404186248779, + 8.697504997253418, + 9.679413795471191, + 11.226194381713867, + 12.25047492980957 + ], + "bp_grad_per_sample_l2_med": [ + 0.00014003549586050212, + 7.460260530933738e-05, + 6.204132660059258e-05, + 3.912465035682544e-05, + 1.6441039406345226e-05 + ], + "bp_grad_F": [ + 0.20991134643554688, + 0.11143218725919724, + 0.08832813799381256, + 0.06269684433937073, + 0.026480402797460556 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.791015625, + "loss_eval": 0.8496487140655518, + "epoch": 58 + }, + { + "hidden_norms_cls": [ + 1.5111491680145264, + 5.304075717926025, + 7.234795093536377, + 9.53318977355957, + 11.964473724365234 + ], + "hidden_norms_avg": [ + 4.304405689239502, + 8.697833061218262, + 9.683843612670898, + 11.22890853881836, + 12.259684562683105 + ], + "bp_grad_per_sample_l2_med": [ + 0.00014237761206459254, + 7.472094148397446e-05, + 5.683211929863319e-05, + 4.049929339089431e-05, + 1.4964447473175824e-05 + ], + "bp_grad_F": [ + 0.21009324491024017, + 0.11155029386281967, + 0.08846566081047058, + 0.06290413439273834, + 0.026488499715924263 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.794921875, + "loss_eval": 0.8498167395591736, + "epoch": 59 + }, + { + "hidden_norms_cls": [ + 1.5111573934555054, + 5.304897785186768, + 7.23397970199585, + 9.53254508972168, + 11.969621658325195 + ], + "hidden_norms_avg": [ + 4.304327964782715, + 8.696623802185059, + 9.681411743164062, + 11.225934982299805, + 12.255182266235352 + ], + "bp_grad_per_sample_l2_med": [ + 0.0001376860454911366, + 7.4876043072436e-05, + 5.8659916248871014e-05, + 4.048796836286783e-05, + 1.5446972611243837e-05 + ], + "bp_grad_F": [ + 0.210032656788826, + 0.11159469932317734, + 0.08847951143980026, + 0.06288634985685349, + 0.02649829164147377 + ], + "gamma_dfa": NaN, + "gamma_dfa_per_layer": [], + "acc_eval": 0.791015625, + "loss_eval": 0.8496516942977905, + "epoch": 60 + } + ], + "dfa_log": [ + { + "hidden_norms_cls": [ + 0.3042769432067871, + 3.2273340225219727, + 4.6926493644714355, + 6.003997802734375, + 6.652705669403076 + ], + "hidden_norms_avg": [ + 6.027750492095947, + 7.1044697761535645, + 7.910475730895996, + 8.79189682006836, + 9.458833694458008 + ], + "bp_grad_per_sample_l2_med": [ + 0.005248428788036108, + 0.0012287507997825742, + 0.0010387187357991934, + 0.0009678652859292924, + 0.0009174557635560632 + ], + "bp_grad_F": [ + 0.248850479722023, + 0.0424945093691349, + 0.0348997563123703, + 0.03202275559306145, + 0.030239813029766083 + ], + "gamma_dfa": -0.0027123973850393668, + "gamma_dfa_per_layer": [ + -0.0010305923642590642, + -0.01111956313252449, + 0.0014823779929429293, + -0.00018181203631684184 + ], + "acc_eval": 0.0791015625, + "loss_eval": 2.544933319091797, + "epoch": 0 + }, + { + "hidden_norms_cls": [ + 1.8898234367370605, + 1755.5985107421875, + 3812.865966796875, + 4063.967529296875, + 6749.62646484375 + ], + "hidden_norms_avg": [ + 48.02373504638672, + 2610.970458984375, + 7442.5693359375, + 13001.9443359375, + 19569.59765625 + ], + "bp_grad_per_sample_l2_med": [ + 0.0006367825553752482, + 1.061867237694969e-06, + 9.113553005590802e-07, + 8.042399599617056e-07, + 7.901451226643985e-07 + ], + "bp_grad_F": [ + 0.11358296871185303, + 5.570299981627613e-05, + 4.605366484611295e-05, + 3.752685370272957e-05, + 3.027659113286063e-05 + ], + "gamma_dfa": 0.007312511210329831, + "gamma_dfa_per_layer": [ + 0.01007317565381527, + 0.008024577051401138, + 0.008248038589954376, + 0.0029042535461485386 + ], + "acc_eval": 0.2333984375, + "loss_eval": 2.0879125595092773, + "epoch": 1 + }, + { + "hidden_norms_cls": [ + 3.704786777496338, + 7219.806640625, + 19592.4140625, + 21808.44921875, + 46117.11328125 + ], + "hidden_norms_avg": [ + 97.33866882324219, + 9787.240234375, + 35216.66015625, + 64377.67578125, + 97116.8125 + ], + "bp_grad_per_sample_l2_med": [ + 9.637367475079373e-05, + 1.6794896851024532e-07, + 1.5954641696680483e-07, + 1.4061814113119908e-07, + 1.445507677999558e-07 + ], + "bp_grad_F": [ + 0.04642009362578392, + 2.128358573827427e-05, + 2.0123250578762963e-05, + 6.922710781509522e-06, + 6.6381039687257726e-06 + ], + "gamma_dfa": 0.007922766148112714, + "gamma_dfa_per_layer": [ + 0.016379408538341522, + 0.005931380204856396, + 0.007020577322691679, + 0.0023596985265612602 + ], + "acc_eval": 0.205078125, + "loss_eval": 2.1287283897399902, + "epoch": 2 + }, + { + "hidden_norms_cls": [ + 5.554904937744141, + 16081.3388671875, + 48028.96875, + 51717.93359375, + 100197.7265625 + ], + "hidden_norms_avg": [ + 147.4190216064453, + 22860.939453125, + 90670.5078125, + 172520.15625, + 247624.09375 + ], + "bp_grad_per_sample_l2_med": [ + 3.051399289688561e-05, + 9.248702070863146e-08, + 7.303499671706959e-08, + 6.487525183729304e-08, + 6.620962977876843e-08 + ], + "bp_grad_F": [ + 0.034720927476882935, + 1.6823112673591822e-05, + 1.1384844583517406e-05, + 3.0318508379423292e-06, + 2.624829903652426e-06 + ], + "gamma_dfa": 0.0067759081866825, + "gamma_dfa_per_layer": [ + 0.014651123434305191, + 0.004187515936791897, + 0.007594288792461157, + 0.0006707045831717551 + ], + "acc_eval": 0.20703125, + "loss_eval": 2.107574462890625, + "epoch": 3 + }, + { + "hidden_norms_cls": [ + 7.38655424118042, + 30113.763671875, + 92815.4375, + 102848.4453125, + 183389.40625 + ], + "hidden_norms_avg": [ + 194.20396423339844, + 39156.78515625, + 166144.453125, + 360300.25, + 500574.96875 + ], + "bp_grad_per_sample_l2_med": [ + 2.7757007046602666e-05, + 6.177719313882335e-08, + 4.3474216226968565e-08, + 3.8972562066419414e-08, + 3.9463689205376795e-08 + ], + "bp_grad_F": [ + 0.03289186581969261, + 5.563175363931805e-06, + 2.151713033526903e-06, + 1.463158696424216e-06, + 1.3795681752526434e-06 + ], + "gamma_dfa": 0.006396490931365406, + "gamma_dfa_per_layer": [ + 0.015158161520957947, + 0.0035972977057099342, + 0.006714952643960714, + 0.00011555185483302921 + ], + "acc_eval": 0.203125, + "loss_eval": 2.1379737854003906, + "epoch": 4 + }, + { + "hidden_norms_cls": [ + 9.15835189819336, + 53591.8984375, + 170615.734375, + 181427.171875, + 289063.5625 + ], + "hidden_norms_avg": [ + 238.28347778320312, + 61062.31640625, + 250187.6875, + 548101.3125, + 724058.1875 + ], + "bp_grad_per_sample_l2_med": [ + 2.960537312901579e-05, + 4.670241438020639e-08, + 3.080446120407032e-08, + 2.670985388419922e-08, + 2.6331489877406966e-08 + ], + "bp_grad_F": [ + 0.016260992735624313, + 4.766869551531272e-06, + 2.067620698653627e-06, + 9.787106591829797e-07, + 8.912123234949831e-07 + ], + "gamma_dfa": 0.0068261599517427385, + "gamma_dfa_per_layer": [ + 0.014116348698735237, + 0.003242844482883811, + 0.008411802351474762, + 0.0015336442738771439 + ], + "acc_eval": 0.1962890625, + "loss_eval": 2.1309733390808105, + "epoch": 5 + }, + { + "hidden_norms_cls": [ + 10.888343811035156, + 91755.1171875, + 249917.8125, + 274005.3125, + 444277.53125 + ], + "hidden_norms_avg": [ + 276.9202880859375, + 104021.7109375, + 349872.5, + 931915.875, + 1222273.75 + ], + "bp_grad_per_sample_l2_med": [ + 2.7342781322658993e-05, + 3.6854661544794e-08, + 2.4993127567540796e-08, + 1.8662358769461207e-08, + 1.7811659702715588e-08 + ], + "bp_grad_F": [ + 0.022468894720077515, + 3.493201802484691e-06, + 1.1520894531713566e-06, + 7.113425795068906e-07, + 6.108040793151304e-07 + ], + "gamma_dfa": 0.007193240569904447, + "gamma_dfa_per_layer": [ + 0.017048656940460205, + 0.002665129955857992, + 0.007390158250927925, + 0.001669017132371664 + ], + "acc_eval": 0.2109375, + "loss_eval": 2.1020355224609375, + "epoch": 6 + }, + { + "hidden_norms_cls": [ + 12.623483657836914, + 158915.890625, + 360070.6875, + 406906.3125, + 630641.3125 + ], + "hidden_norms_avg": [ + 315.0859680175781, + 172244.328125, + 490065.59375, + 1510240.375, + 1952914.5 + ], + "bp_grad_per_sample_l2_med": [ + 3.403881783015095e-05, + 2.7625405962794503e-08, + 1.651102543576144e-08, + 1.1742161909467086e-08, + 1.0927361238088906e-08 + ], + "bp_grad_F": [ + 0.024832597002387047, + 1.8080392010233481e-06, + 7.246239306368807e-07, + 4.936595132676302e-07, + 4.0145107504940825e-07 + ], + "gamma_dfa": 0.006596001767320558, + "gamma_dfa_per_layer": [ + 0.016851693391799927, + 0.001910479972139001, + 0.0058369822800159454, + 0.0017848514253273606 + ], + "acc_eval": 0.2138671875, + "loss_eval": 2.1312928199768066, + "epoch": 7 + }, + { + "hidden_norms_cls": [ + 14.368335723876953, + 243332.8125, + 525771.0, + 593291.75, + 875288.5 + ], + "hidden_norms_avg": [ + 351.16015625, + 247222.375, + 640658.75, + 1855662.75, + 2391194.25 + ], + "bp_grad_per_sample_l2_med": [ + 2.87416060018586e-05, + 1.8018790015617014e-08, + 1.2651691250198382e-08, + 9.33115984480537e-09, + 8.563736386690834e-09 + ], + "bp_grad_F": [ + 0.017871394753456116, + 9.426115070709784e-07, + 4.7217255882969766e-07, + 3.564711050785263e-07, + 3.2240865266430774e-07 + ], + "gamma_dfa": 0.006289168493822217, + "gamma_dfa_per_layer": [ + 0.013373943977057934, + 0.0037157258484512568, + 0.00672483816742897, + 0.001342165982350707 + ], + "acc_eval": 0.220703125, + "loss_eval": 2.0829837322235107, + "epoch": 8 + }, + { + "hidden_norms_cls": [ + 16.041412353515625, + 385486.09375, + 776263.875, + 869037.0, + 1191002.25 + ], + "hidden_norms_avg": [ + 383.5213317871094, + 383708.40625, + 941498.0, + 2793651.25, + 3510556.5 + ], + "bp_grad_per_sample_l2_med": [ + 2.8917142117279582e-05, + 1.1363525231899985e-08, + 8.616331648170217e-09, + 5.602694308493028e-09, + 5.260086588521062e-09 + ], + "bp_grad_F": [ + 0.024791762232780457, + 6.833873840150773e-07, + 3.100701633229619e-07, + 2.26619619070334e-07, + 2.1342954426017968e-07 + ], + "gamma_dfa": 0.00652129843365401, + "gamma_dfa_per_layer": [ + 0.014406262896955013, + 0.004345592111349106, + 0.006529162637889385, + 0.0008041760884225368 + ], + "acc_eval": 0.2197265625, + "loss_eval": 2.1437277793884277, + "epoch": 9 + }, + { + "hidden_norms_cls": [ + 17.677919387817383, + 582942.5, + 1148260.75, + 1279834.25, + 1709270.625 + ], + "hidden_norms_avg": [ + 416.1814880371094, + 530246.375, + 1201761.5, + 2944394.0, + 3732565.75 + ], + "bp_grad_per_sample_l2_med": [ + 3.369058686075732e-05, + 6.623859238885643e-09, + 4.513381224313662e-09, + 3.933114278709127e-09, + 3.794405678547719e-09 + ], + "bp_grad_F": [ + 0.01620429754257202, + 3.022730652446626e-07, + 1.9097409165169665e-07, + 1.773983058228623e-07, + 1.593236333974346e-07 + ], + "gamma_dfa": 0.006082151841837913, + "gamma_dfa_per_layer": [ + 0.01578153669834137, + 0.002655572956427932, + 0.005509376525878906, + 0.00038212118670344353 + ], + "acc_eval": 0.197265625, + "loss_eval": 2.089698314666748, + "epoch": 10 + }, + { + "hidden_norms_cls": [ + 19.31073760986328, + 843820.0625, + 1587423.25, + 1759006.0, + 2222149.5 + ], + "hidden_norms_avg": [ + 448.93133544921875, + 796172.1875, + 1678321.375, + 3888912.0, + 4810418.0 + ], + "bp_grad_per_sample_l2_med": [ + 4.445635568117723e-05, + 3.5140981236025937e-09, + 2.8926525619255017e-09, + 2.65975419466713e-09, + 2.6499111793754082e-09 + ], + "bp_grad_F": [ + 0.020245229825377464, + 2.6993967594535206e-07, + 1.701987457636278e-07, + 1.3198047099649557e-07, + 1.2469256205349666e-07 + ], + "gamma_dfa": 0.005559869911849091, + "gamma_dfa_per_layer": [ + 0.015185668133199215, + 0.0025689376052469015, + 0.004468783736228943, + 1.6090172721305862e-05 + ], + "acc_eval": 0.2314453125, + "loss_eval": 2.0838193893432617, + "epoch": 11 + }, + { + "hidden_norms_cls": [ + 20.861183166503906, + 1157778.75, + 2054207.5, + 2289084.5, + 2841995.0 + ], + "hidden_norms_avg": [ + 479.8387145996094, + 981138.875, + 2112497.25, + 4439588.0, + 5517614.5 + ], + "bp_grad_per_sample_l2_med": [ + 3.176981772412546e-05, + 3.1782512177613853e-09, + 2.5456772245746606e-09, + 2.4393245201537184e-09, + 2.414718425214346e-09 + ], + "bp_grad_F": [ + 0.09368129819631577, + 9.166008680949744e-07, + 2.675243706562469e-07, + 1.0984689424731187e-07, + 1.0444787079677553e-07 + ], + "gamma_dfa": 0.004390636342577636, + "gamma_dfa_per_layer": [ + 0.011994147673249245, + 0.001997251994907856, + 0.003865728620439768, + -0.00029458291828632355 + ], + "acc_eval": 0.224609375, + "loss_eval": 2.0896079540252686, + "epoch": 12 + }, + { + "hidden_norms_cls": [ + 22.362810134887695, + 1504072.375, + 2527410.75, + 2870361.25, + 3557771.0 + ], + "hidden_norms_avg": [ + 510.81561279296875, + 1188843.375, + 2574796.25, + 5296089.0, + 6640932.5 + ], + "bp_grad_per_sample_l2_med": [ + 2.8154454412288032e-05, + 3.057065711686846e-09, + 2.121731013104977e-09, + 2.0999049166192663e-09, + 2.0892536589656174e-09 + ], + "bp_grad_F": [ + 0.011068353429436684, + 1.3467251847032458e-06, + 1.0581060649883511e-07, + 9.427417069218791e-08, + 9.197237460512042e-08 + ], + "gamma_dfa": 0.003989392673247494, + "gamma_dfa_per_layer": [ + 0.011822337284684181, + 0.001384879695251584, + 0.00329465395770967, + -0.0005443002446554601 + ], + "acc_eval": 0.244140625, + "loss_eval": 2.077056407928467, + "epoch": 13 + }, + { + "hidden_norms_cls": [ + 23.812599182128906, + 1930070.0, + 2995981.75, + 3469627.0, + 4303896.0 + ], + "hidden_norms_avg": [ + 539.9255981445312, + 1589889.5, + 3066154.25, + 6015238.0, + 7540605.5 + ], + "bp_grad_per_sample_l2_med": [ + 3.170168929500505e-05, + 2.518652619798445e-09, + 1.924201020742089e-09, + 1.8520206479522017e-09, + 1.8417810609960839e-09 + ], + "bp_grad_F": [ + 0.032089244574308395, + 1.0142422297576559e-06, + 1.6566445992793888e-07, + 8.33219857554468e-08, + 7.640721122470495e-08 + ], + "gamma_dfa": 0.0029267592035466805, + "gamma_dfa_per_layer": [ + 0.008105762302875519, + 0.001629034522920847, + 0.0022919299080967903, + -0.000319689919706434 + ], + "acc_eval": 0.224609375, + "loss_eval": 2.0748229026794434, + "epoch": 14 + }, + { + "hidden_norms_cls": [ + 25.248563766479492, + 2452306.5, + 3664022.75, + 4168536.0, + 5106242.5 + ], + "hidden_norms_avg": [ + 570.19384765625, + 1965615.5, + 3710471.5, + 7372995.0, + 9239893.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.4190969270421192e-05, + 1.8390389211475622e-09, + 1.5613663695290825e-09, + 1.5359991056840272e-09, + 1.5388512686342892e-09 + ], + "bp_grad_F": [ + 0.015056902542710304, + 6.693913292110665e-06, + 1.241161982079575e-07, + 7.622224984515924e-08, + 6.743793790064956e-08 + ], + "gamma_dfa": 0.0022661721595795825, + "gamma_dfa_per_layer": [ + 0.006208081729710102, + 0.001149914925917983, + 0.0020575951784849167, + -0.0003509031957946718 + ], + "acc_eval": 0.236328125, + "loss_eval": 2.0835204124450684, + "epoch": 15 + }, + { + "hidden_norms_cls": [ + 26.6426944732666, + 3107169.0, + 4381372.0, + 5027998.5, + 6065136.0 + ], + "hidden_norms_avg": [ + 596.7793579101562, + 2542559.0, + 4400625.5, + 8374831.5, + 10510773.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.787545781757217e-05, + 1.5701163702530607e-09, + 1.2841328000945396e-09, + 1.2669904014828148e-09, + 1.2700215323846464e-09 + ], + "bp_grad_F": [ + 0.006971500348299742, + 1.8619180991663598e-05, + 6.465826629664662e-08, + 5.84209978171657e-08, + 5.694714744208795e-08 + ], + "gamma_dfa": 0.0030232197314035147, + "gamma_dfa_per_layer": [ + 0.009539565071463585, + 0.0012945058988407254, + 0.0015697493217885494, + -0.0003109413664788008 + ], + "acc_eval": 0.236328125, + "loss_eval": 2.058485507965088, + "epoch": 16 + }, + { + "hidden_norms_cls": [ + 27.95879364013672, + 3613966.25, + 5044349.0, + 5874924.0, + 7080756.0 + ], + "hidden_norms_avg": [ + 623.2052612304688, + 2844551.5, + 4942977.0, + 8742455.0, + 11094594.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.8442762413760647e-05, + 1.629689605486817e-09, + 1.115437520127216e-09, + 1.0797144289753646e-09, + 1.0779978021346892e-09 + ], + "bp_grad_F": [ + 0.0072508929297327995, + 1.4132167507341364e-06, + 6.271753250075562e-08, + 4.991998281411725e-08, + 4.950436149897541e-08 + ], + "gamma_dfa": 0.001284227124415338, + "gamma_dfa_per_layer": [ + 0.00256139412522316, + 0.0014944188296794891, + 0.0013982560485601425, + -0.0003171605058014393 + ], + "acc_eval": 0.240234375, + "loss_eval": 2.0549392700195312, + "epoch": 17 + }, + { + "hidden_norms_cls": [ + 29.22184181213379, + 4211867.5, + 5636704.5, + 6891397.5, + 8559809.0 + ], + "hidden_norms_avg": [ + 647.4288330078125, + 3423908.25, + 5695376.5, + 10367815.0, + 13188250.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.8086895326850936e-05, + 1.1501447572115353e-09, + 8.738501033889179e-10, + 8.528160400089746e-10, + 8.584906674435899e-10 + ], + "bp_grad_F": [ + 0.012267161160707474, + 1.591232035025314e-06, + 4.697067268466526e-08, + 4.271150899626264e-08, + 4.017679700041299e-08 + ], + "gamma_dfa": 0.0022953646985115483, + "gamma_dfa_per_layer": [ + 0.007007642649114132, + 0.0012370320037007332, + 0.0011846421984955668, + -0.0002478580572642386 + ], + "acc_eval": 0.228515625, + "loss_eval": 2.053128242492676, + "epoch": 18 + }, + { + "hidden_norms_cls": [ + 30.412078857421875, + 4986411.0, + 6504586.0, + 7889628.0, + 9783147.0 + ], + "hidden_norms_avg": [ + 670.3391723632812, + 3921667.0, + 6270645.0, + 10589857.0, + 13580617.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.768085505522322e-05, + 1.3132780418700918e-09, + 7.742924079678914e-10, + 7.418777814294231e-10, + 7.448673344789825e-10 + ], + "bp_grad_F": [ + 0.007575999945402145, + 1.7771724287740653e-06, + 5.529824420591467e-08, + 3.790939828718365e-08, + 3.6307092443621514e-08 + ], + "gamma_dfa": 0.0009071348831639625, + "gamma_dfa_per_layer": [ + 0.001273356145247817, + 0.001484773471020162, + 0.0010661636479198933, + -0.00019575373153202236 + ], + "acc_eval": 0.25390625, + "loss_eval": 2.0479414463043213, + "epoch": 19 + }, + { + "hidden_norms_cls": [ + 31.54771614074707, + 5502896.5, + 7070197.5, + 8994800.0, + 11604998.0 + ], + "hidden_norms_avg": [ + 692.4578247070312, + 4249610.0, + 6837255.0, + 11617683.0, + 15223152.0 + ], + "bp_grad_per_sample_l2_med": [ + 3.0367684303200804e-05, + 1.325711207478264e-09, + 6.525959217462685e-10, + 6.324714640904006e-10, + 6.32227936669949e-10 + ], + "bp_grad_F": [ + 0.0062190573662519455, + 1.129235442931531e-06, + 3.317861896334762e-08, + 3.131979653403505e-08, + 3.109861168582029e-08 + ], + "gamma_dfa": 0.0011510169642861001, + "gamma_dfa_per_layer": [ + 0.002568549243733287, + 0.0013005147920921445, + 0.0009535959688946605, + -0.00021859214757569134 + ], + "acc_eval": 0.255859375, + "loss_eval": 2.0359444618225098, + "epoch": 20 + }, + { + "hidden_norms_cls": [ + 32.638641357421875, + 6320697.0, + 7897073.5, + 10165307.0, + 13313912.0 + ], + "hidden_norms_avg": [ + 713.3525390625, + 4812640.5, + 7533297.5, + 12847435.0, + 16993420.0 + ], + "bp_grad_per_sample_l2_med": [ + 3.871564695145935e-05, + 1.2221273992807369e-09, + 6.080098646776833e-10, + 5.917044632042234e-10, + 5.89539916884263e-10 + ], + "bp_grad_F": [ + 0.011915000155568123, + 6.25122083874885e-06, + 3.6603957198622084e-08, + 2.876341298474472e-08, + 2.8581927935533713e-08 + ], + "gamma_dfa": 0.0002986440376844257, + "gamma_dfa_per_layer": [ + -0.0004627959569916129, + 0.0009951044339686632, + 0.0008257199660874903, + -0.00016345229232683778 + ], + "acc_eval": 0.2626953125, + "loss_eval": 2.059875965118408, + "epoch": 21 + }, + { + "hidden_norms_cls": [ + 33.6330451965332, + 7264995.5, + 8758988.0, + 11785280.0, + 15874132.0 + ], + "hidden_norms_avg": [ + 732.8869018554688, + 5982206.0, + 8454064.0, + 14846262.0, + 19834080.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.5420635211048648e-05, + 7.56205986718328e-10, + 5.169706884800007e-10, + 5.026724592127607e-10, + 5.024383686880185e-10 + ], + "bp_grad_F": [ + 0.014803220517933369, + 4.478873961488716e-06, + 4.4555019229619575e-08, + 2.4910459472948787e-08, + 2.4745437698925343e-08 + ], + "gamma_dfa": 0.001046390665578656, + "gamma_dfa_per_layer": [ + 0.0027164756320416927, + 0.0009058329742401838, + 0.0007253356743603945, + -0.00016208161832764745 + ], + "acc_eval": 0.255859375, + "loss_eval": 2.0538365840911865, + "epoch": 22 + }, + { + "hidden_norms_cls": [ + 34.621524810791016, + 8089090.5, + 9671455.0, + 13086530.0, + 17616980.0 + ], + "hidden_norms_avg": [ + 751.2442016601562, + 6699752.5, + 9159771.0, + 15622865.0, + 21214146.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.7939373467233963e-05, + 7.656756340068682e-10, + 5.061054908495066e-10, + 4.837483191799663e-10, + 4.820439047925618e-10 + ], + "bp_grad_F": [ + 0.008210284635424614, + 3.91981393477181e-06, + 2.727725245676993e-08, + 2.349457695061119e-08, + 2.2768137597495297e-08 + ], + "gamma_dfa": 0.0013028495595790446, + "gamma_dfa_per_layer": [ + 0.0034610098227858543, + 0.0012297509238123894, + 0.0006837932160124183, + -0.00016315572429448366 + ], + "acc_eval": 0.2646484375, + "loss_eval": 2.0430214405059814, + "epoch": 23 + }, + { + "hidden_norms_cls": [ + 35.54478454589844, + 9145767.0, + 10558942.0, + 14636200.0, + 19721632.0 + ], + "hidden_norms_avg": [ + 768.5211791992188, + 7473006.0, + 10093548.0, + 17775052.0, + 24159262.0 + ], + "bp_grad_per_sample_l2_med": [ + 3.364618532941677e-05, + 8.066202705769854e-10, + 4.5619852340195166e-10, + 4.4121445386124947e-10, + 4.323804370098827e-10 + ], + "bp_grad_F": [ + 0.09167197346687317, + 4.855019597016508e-06, + 2.689265521382822e-08, + 2.1627007740221416e-08, + 2.0867716443717654e-08 + ], + "gamma_dfa": 0.0001647558601689525, + "gamma_dfa_per_layer": [ + -0.0005364040262065828, + 0.0007433416321873665, + 0.0005910456529818475, + -0.00013895981828682125 + ], + "acc_eval": 0.23046875, + "loss_eval": 2.0427517890930176, + "epoch": 24 + }, + { + "hidden_norms_cls": [ + 36.43881607055664, + 10216266.0, + 11478190.0, + 16476424.0, + 22393992.0 + ], + "hidden_norms_avg": [ + 785.1990966796875, + 8300483.0, + 10872433.0, + 18667924.0, + 25447926.0 + ], + "bp_grad_per_sample_l2_med": [ + 3.0938736017560586e-05, + 7.561202219896757e-10, + 4.2832748459176173e-10, + 4.0923736621678586e-10, + 4.0218126051705383e-10 + ], + "bp_grad_F": [ + 0.013463743962347507, + 1.9222785340389237e-05, + 2.2825457079989064e-08, + 1.9977358434175585e-08, + 1.9259800865256693e-08 + ], + "gamma_dfa": 0.0005702937960450072, + "gamma_dfa_per_layer": [ + 0.0011116554960608482, + 0.0007594762137159705, + 0.0005430461023934186, + -0.00013300262799020857 + ], + "acc_eval": 0.2568359375, + "loss_eval": 2.0594615936279297, + "epoch": 25 + }, + { + "hidden_norms_cls": [ + 37.30263137817383, + 11623684.0, + 12776919.0, + 18306868.0, + 24740858.0 + ], + "hidden_norms_avg": [ + 800.9972534179688, + 9603253.0, + 12104963.0, + 20835068.0, + 28405318.0 + ], + "bp_grad_per_sample_l2_med": [ + 3.261065648985095e-05, + 6.65525801135658e-10, + 3.838940831890625e-10, + 3.742672838313865e-10, + 3.7395189722566613e-10 + ], + "bp_grad_F": [ + 0.007853930816054344, + 1.2899076864414383e-06, + 1.9278010299217385e-08, + 1.8154999281705386e-08, + 1.7368931182204506e-08 + ], + "gamma_dfa": 0.00021361159815569408, + "gamma_dfa_per_layer": [ + -0.00034064118517562747, + 0.0008438127697445452, + 0.0004718025156762451, + -0.00012052770762238652 + ], + "acc_eval": 0.255859375, + "loss_eval": 2.0433781147003174, + "epoch": 26 + }, + { + "hidden_norms_cls": [ + 38.06776428222656, + 12599130.0, + 13703165.0, + 19834172.0, + 26867468.0 + ], + "hidden_norms_avg": [ + 816.2463989257812, + 10421772.0, + 12873138.0, + 21321398.0, + 29452538.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.774638414848596e-05, + 6.227232618449818e-10, + 3.566741624272396e-10, + 3.4781866276034634e-10, + 3.45437983773067e-10 + ], + "bp_grad_F": [ + 0.013920681551098824, + 3.4208371744171018e-06, + 2.1883508338760294e-08, + 1.7128741092165e-08, + 1.6603545205384762e-08 + ], + "gamma_dfa": 0.0008690126524015795, + "gamma_dfa_per_layer": [ + 0.0021882583387196064, + 0.0009064457844942808, + 0.0004841286572627723, + -0.00010278217087034136 + ], + "acc_eval": 0.2529296875, + "loss_eval": 2.0433974266052246, + "epoch": 27 + }, + { + "hidden_norms_cls": [ + 38.7847785949707, + 14429584.0, + 15438086.0, + 21566880.0, + 29267340.0 + ], + "hidden_norms_avg": [ + 830.9344482421875, + 11436777.0, + 13832175.0, + 22913616.0, + 32026754.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.8682481570285745e-05, + 5.758302168423768e-10, + 3.3607527871737375e-10, + 3.304233830991876e-10, + 3.2846284026000205e-10 + ], + "bp_grad_F": [ + 0.012251574546098709, + 2.996052717207931e-05, + 2.7452134787608884e-08, + 1.577506480998636e-08, + 1.5141546683139495e-08 + ], + "gamma_dfa": 0.0004505918041104451, + "gamma_dfa_per_layer": [ + 0.00026967719895765185, + 0.001212525530718267, + 0.0004131481982767582, + -9.298371151089668e-05 + ], + "acc_eval": 0.2607421875, + "loss_eval": 2.039923667907715, + "epoch": 28 + }, + { + "hidden_norms_cls": [ + 39.49647521972656, + 15523604.0, + 16504434.0, + 23236370.0, + 31306890.0 + ], + "hidden_norms_avg": [ + 844.8562622070312, + 12856661.0, + 15124535.0, + 23633704.0, + 32995118.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.6219468054478057e-05, + 5.519696366640403e-10, + 3.2825916984613457e-10, + 3.235848811122821e-10, + 3.2190172749579915e-10 + ], + "bp_grad_F": [ + 0.012884745374321938, + 1.8329246813664213e-05, + 1.8323355277516384e-08, + 1.5584015855552025e-08, + 1.4755748622974352e-08 + ], + "gamma_dfa": 0.0006857650478195865, + "gamma_dfa_per_layer": [ + 0.0016966589028015733, + 0.0007126724813133478, + 0.0003994358121417463, + -6.570700497832149e-05 + ], + "acc_eval": 0.2451171875, + "loss_eval": 2.0533735752105713, + "epoch": 29 + }, + { + "hidden_norms_cls": [ + 40.16881561279297, + 16141660.0, + 17137856.0, + 24286464.0, + 32903152.0 + ], + "hidden_norms_avg": [ + 857.8116455078125, + 12895693.0, + 15492559.0, + 23745768.0, + 33843860.0 + ], + "bp_grad_per_sample_l2_med": [ + 3.2289473892888054e-05, + 6.340856728570543e-10, + 3.264067627295475e-10, + 3.149370209065694e-10, + 3.097411216401724e-10 + ], + "bp_grad_F": [ + 0.006337009835988283, + 1.7065680367522873e-05, + 5.785592804841144e-08, + 1.519365788738014e-08, + 1.464850551968766e-08 + ], + "gamma_dfa": 0.0005419648223323748, + "gamma_dfa_per_layer": [ + 0.00038546844734810293, + 0.0014093497302383184, + 0.0004054713645018637, + -3.243025275878608e-05 + ], + "acc_eval": 0.265625, + "loss_eval": 2.0390102863311768, + "epoch": 30 + }, + { + "hidden_norms_cls": [ + 40.790164947509766, + 18768648.0, + 19610716.0, + 26926266.0, + 36060360.0 + ], + "hidden_norms_avg": [ + 869.6182861328125, + 15248417.0, + 17480158.0, + 26928640.0, + 37555876.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.939497062470764e-05, + 5.127394064885493e-10, + 3.1137872835707014e-10, + 3.074558663218596e-10, + 3.04990144250894e-10 + ], + "bp_grad_F": [ + 0.03832927718758583, + 8.881174835551064e-06, + 2.5206949416656244e-08, + 1.4185034480362901e-08, + 1.3504863893842867e-08 + ], + "gamma_dfa": 0.0005767397606177838, + "gamma_dfa_per_layer": [ + 0.00058704090770334, + 0.0013887862442061305, + 0.0003486467176117003, + -1.751482705003582e-05 + ], + "acc_eval": 0.232421875, + "loss_eval": 2.055915117263794, + "epoch": 31 + }, + { + "hidden_norms_cls": [ + 41.36415481567383, + 20590018.0, + 21418164.0, + 28674892.0, + 38102536.0 + ], + "hidden_norms_avg": [ + 880.8012084960938, + 15857472.0, + 18200580.0, + 27607590.0, + 38905496.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.6930878448183648e-05, + 5.041319584009329e-10, + 3.0716035270828e-10, + 2.9963459491355593e-10, + 2.9755845010193127e-10 + ], + "bp_grad_F": [ + 0.01052916795015335, + 1.7662874597590417e-05, + 1.8223250464188823e-08, + 1.4229329714510186e-08, + 1.2928855319671584e-08 + ], + "gamma_dfa": 0.0005623552362976625, + "gamma_dfa_per_layer": [ + 0.0005056762602180243, + 0.00140343327075243, + 0.0003372218052390963, + 3.0896089810994454e-06 + ], + "acc_eval": 0.2451171875, + "loss_eval": 2.054579734802246, + "epoch": 32 + }, + { + "hidden_norms_cls": [ + 41.9267578125, + 21995458.0, + 22701258.0, + 30227660.0, + 40381468.0 + ], + "hidden_norms_avg": [ + 891.6727294921875, + 17925388.0, + 19924592.0, + 29375846.0, + 41265952.0 + ], + "bp_grad_per_sample_l2_med": [ + 3.4433691325830296e-05, + 4.786913088139499e-10, + 3.083476807219654e-10, + 3.030005690796145e-10, + 3.0202479406327143e-10 + ], + "bp_grad_F": [ + 0.006410651374608278, + 9.426087672181893e-06, + 1.748733602369157e-08, + 1.330249954634155e-08, + 1.2548162509062877e-08 + ], + "gamma_dfa": 0.0005204869294175296, + "gamma_dfa_per_layer": [ + 0.0004088875139132142, + 0.001322226831689477, + 0.00032363145146518946, + 2.7201920602237806e-05 + ], + "acc_eval": 0.2568359375, + "loss_eval": 2.0504493713378906, + "epoch": 33 + }, + { + "hidden_norms_cls": [ + 42.43289566040039, + 23788478.0, + 24547006.0, + 32122702.0, + 42524392.0 + ], + "hidden_norms_avg": [ + 902.5420532226562, + 19126440.0, + 21210522.0, + 30872922.0, + 43207040.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.8793765523005277e-05, + 4.69506655775831e-10, + 2.9882180063722785e-10, + 2.9027436010409247e-10, + 2.88086998700976e-10 + ], + "bp_grad_F": [ + 0.02046363614499569, + 1.535968840471469e-05, + 1.9181511490273806e-08, + 1.3211375105015577e-08, + 1.2052867148781843e-08 + ], + "gamma_dfa": 0.0005800028211524477, + "gamma_dfa_per_layer": [ + 0.0005789051647298038, + 0.001365642063319683, + 0.0003222030645702034, + 5.326099199010059e-05 + ], + "acc_eval": 0.236328125, + "loss_eval": 2.056929349899292, + "epoch": 34 + }, + { + "hidden_norms_cls": [ + 42.88957595825195, + 24876160.0, + 25665632.0, + 33698024.0, + 44433192.0 + ], + "hidden_norms_avg": [ + 912.5694580078125, + 20388924.0, + 22492538.0, + 32293406.0, + 44758584.0 + ], + "bp_grad_per_sample_l2_med": [ + 3.043079595954623e-05, + 4.6556239419182077e-10, + 3.0177585430557485e-10, + 2.9259758504984745e-10, + 2.8850585809259144e-10 + ], + "bp_grad_F": [ + 0.018651673570275307, + 8.0938043538481e-06, + 2.600621051840335e-08, + 1.3419185762586494e-08, + 1.200509913701353e-08 + ], + "gamma_dfa": 0.0008372370848519495, + "gamma_dfa_per_layer": [ + 0.001413366524502635, + 0.0015261531807482243, + 0.0003260627272538841, + 8.336590690305457e-05 + ], + "acc_eval": 0.2353515625, + "loss_eval": 2.060340404510498, + "epoch": 35 + }, + { + "hidden_norms_cls": [ + 43.32978057861328, + 25917158.0, + 26720118.0, + 34948196.0, + 46313048.0 + ], + "hidden_norms_avg": [ + 921.58642578125, + 21474648.0, + 23549474.0, + 33572188.0, + 46615396.0 + ], + "bp_grad_per_sample_l2_med": [ + 3.262757672928274e-05, + 4.5062673037499223e-10, + 2.984587021970242e-10, + 2.923088993078693e-10, + 2.888300987269332e-10 + ], + "bp_grad_F": [ + 0.01846941001713276, + 2.4291131921927445e-05, + 2.994345393858566e-08, + 1.4564162320596097e-08, + 1.1765412644137996e-08 + ], + "gamma_dfa": 0.0007057133989292197, + "gamma_dfa_per_layer": [ + 0.0011925556464120746, + 0.0012432597577571869, + 0.0002897845406550914, + 9.725365089252591e-05 + ], + "acc_eval": 0.2451171875, + "loss_eval": 2.0591180324554443, + "epoch": 36 + }, + { + "hidden_norms_cls": [ + 43.74128341674805, + 27200966.0, + 27911578.0, + 36161972.0, + 47854788.0 + ], + "hidden_norms_avg": [ + 930.2008666992188, + 22529414.0, + 24601434.0, + 34829948.0, + 48238220.0 + ], + "bp_grad_per_sample_l2_med": [ + 3.328157981741242e-05, + 4.659617691693541e-10, + 2.8960928100119077e-10, + 2.8687860420539835e-10, + 2.800103759970085e-10 + ], + "bp_grad_F": [ + 0.004556785803288221, + 9.644008059694897e-06, + 1.5685751364458156e-08, + 1.2929571191477862e-08, + 1.1707988356590704e-08 + ], + "gamma_dfa": 0.0005315981106832623, + "gamma_dfa_per_layer": [ + 0.0004304055473767221, + 0.0012520912569016218, + 0.0003118191962130368, + 0.0001320764422416687 + ], + "acc_eval": 0.255859375, + "loss_eval": 2.049736738204956, + "epoch": 37 + }, + { + "hidden_norms_cls": [ + 44.10750198364258, + 28860724.0, + 29566448.0, + 37605440.0, + 49767188.0 + ], + "hidden_norms_avg": [ + 938.7098999023438, + 24297102.0, + 26192510.0, + 36561148.0, + 49957968.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.900460640375968e-05, + 4.445998014190877e-10, + 2.888302930159625e-10, + 2.855987946137617e-10, + 2.818917876901139e-10 + ], + "bp_grad_F": [ + 0.013354518450796604, + 6.135221156000625e-06, + 1.780158420672251e-08, + 1.3947193622243503e-08, + 1.161484952660885e-08 + ], + "gamma_dfa": 0.0005519518890650943, + "gamma_dfa_per_layer": [ + 0.0005505518638528883, + 0.001156628131866455, + 0.00033354441984556615, + 0.0001670831406954676 + ], + "acc_eval": 0.2412109375, + "loss_eval": 2.0527162551879883, + "epoch": 38 + }, + { + "hidden_norms_cls": [ + 44.42500305175781, + 30085758.0, + 30793664.0, + 39317268.0, + 51773680.0 + ], + "hidden_norms_avg": [ + 946.3699340820312, + 24723960.0, + 26763662.0, + 37607688.0, + 51575988.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.9910466764704324e-05, + 4.753269444712771e-10, + 2.936461906966059e-10, + 2.873846438600225e-10, + 2.8520716344182517e-10 + ], + "bp_grad_F": [ + 0.00976789090782404, + 1.76885150722228e-05, + 1.7970489096796882e-08, + 1.3952313970833075e-08, + 1.130456972475713e-08 + ], + "gamma_dfa": 0.0009108289959840477, + "gamma_dfa_per_layer": [ + 0.0019160081865265965, + 0.0012680008076131344, + 0.00030145354685373604, + 0.00015785344294272363 + ], + "acc_eval": 0.2412109375, + "loss_eval": 2.0550765991210938, + "epoch": 39 + }, + { + "hidden_norms_cls": [ + 44.73127746582031, + 30730856.0, + 31431762.0, + 40187116.0, + 53148352.0 + ], + "hidden_norms_avg": [ + 953.6856079101562, + 24778752.0, + 27014248.0, + 38275640.0, + 52795456.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.8919872420374304e-05, + 4.755910110176842e-10, + 2.84101492331601e-10, + 2.7933383384137755e-10, + 2.7627339305169585e-10 + ], + "bp_grad_F": [ + 0.009574169293045998, + 4.757252463605255e-05, + 1.5709005651842745e-08, + 1.3778580054690792e-08, + 1.1154800638735196e-08 + ], + "gamma_dfa": 0.0003838329284917563, + "gamma_dfa_per_layer": [ + -0.0004134870832785964, + 0.001480170525610447, + 0.0002943193248938769, + 0.00017432894674129784 + ], + "acc_eval": 0.255859375, + "loss_eval": 2.0383803844451904, + "epoch": 40 + }, + { + "hidden_norms_cls": [ + 44.999107360839844, + 32617970.0, + 33381886.0, + 41663220.0, + 54853168.0 + ], + "hidden_norms_avg": [ + 960.5386962890625, + 27910628.0, + 29647900.0, + 40418716.0, + 54716684.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.9992028430569917e-05, + 4.221052118946744e-10, + 2.9753777219809763e-10, + 2.8370578109004896e-10, + 2.813366484222257e-10 + ], + "bp_grad_F": [ + 0.015446176752448082, + 3.8060017686802894e-05, + 5.211516906911129e-08, + 1.3235438522940512e-08, + 1.116207481999254e-08 + ], + "gamma_dfa": 0.0005705733274226077, + "gamma_dfa_per_layer": [ + 0.0006168894469738007, + 0.0011725829681381583, + 0.00029753358103334904, + 0.00019528731354512274 + ], + "acc_eval": 0.240234375, + "loss_eval": 2.0557656288146973, + "epoch": 41 + }, + { + "hidden_norms_cls": [ + 45.25053024291992, + 33936992.0, + 34621240.0, + 42946580.0, + 56031144.0 + ], + "hidden_norms_avg": [ + 966.2922973632812, + 28239400.0, + 30055502.0, + 41338012.0, + 56048388.0 + ], + "bp_grad_per_sample_l2_med": [ + 3.0256345780799165e-05, + 4.424338395647709e-10, + 2.9891861208497517e-10, + 2.8403482343897224e-10, + 2.798661857816853e-10 + ], + "bp_grad_F": [ + 0.006179355084896088, + 8.527667887392454e-06, + 2.116273023489157e-08, + 1.2261468285146293e-08, + 1.1058569171495947e-08 + ], + "gamma_dfa": 0.0009438929373573046, + "gamma_dfa_per_layer": [ + 0.00216879416257143, + 0.0010916339233517647, + 0.0003159396001137793, + 0.00019920406339224428 + ], + "acc_eval": 0.2470703125, + "loss_eval": 2.0517942905426025, + "epoch": 42 + }, + { + "hidden_norms_cls": [ + 45.481727600097656, + 34746476.0, + 35397348.0, + 43639664.0, + 57119836.0 + ], + "hidden_norms_avg": [ + 971.5364379882812, + 28891962.0, + 30705540.0, + 42078288.0, + 57065612.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.9764614737359807e-05, + 4.5060349895820195e-10, + 3.0148228358228835e-10, + 2.912529661891483e-10, + 2.8715479993834947e-10 + ], + "bp_grad_F": [ + 0.02291245572268963, + 1.3926567589805927e-05, + 3.118159241921603e-08, + 2.1897225366274142e-08, + 1.1063916893760961e-08 + ], + "gamma_dfa": 0.00046937728620832786, + "gamma_dfa_per_layer": [ + -0.00010078266495838761, + 0.0014219002332538366, + 0.0003067174111492932, + 0.00024967416538856924 + ], + "acc_eval": 0.2451171875, + "loss_eval": 2.054462432861328, + "epoch": 43 + }, + { + "hidden_norms_cls": [ + 45.68215560913086, + 35169816.0, + 35827116.0, + 44266788.0, + 58219984.0 + ], + "hidden_norms_avg": [ + 976.61279296875, + 29302304.0, + 31207594.0, + 42541932.0, + 57918596.0 + ], + "bp_grad_per_sample_l2_med": [ + 3.18697166221682e-05, + 4.4465031656670817e-10, + 3.0559865749069104e-10, + 2.891223926937414e-10, + 2.8127278284273416e-10 + ], + "bp_grad_F": [ + 0.02285478264093399, + 5.0576767534948885e-05, + 2.966898726697309e-08, + 1.374073210769211e-08, + 1.1011787037773502e-08 + ], + "gamma_dfa": 0.0006015256294631399, + "gamma_dfa_per_layer": [ + 0.0005435256171040237, + 0.0013653798960149288, + 0.000284912355709821, + 0.0002122846490237862 + ], + "acc_eval": 0.244140625, + "loss_eval": 2.0539493560791016, + "epoch": 44 + }, + { + "hidden_norms_cls": [ + 45.86240005493164, + 36071220.0, + 36702996.0, + 45105948.0, + 59117940.0 + ], + "hidden_norms_avg": [ + 980.9766845703125, + 29250424.0, + 31315982.0, + 42895240.0, + 58632240.0 + ], + "bp_grad_per_sample_l2_med": [ + 3.0838917155051604e-05, + 4.517226037670241e-10, + 3.037594897836726e-10, + 2.95300450758873e-10, + 2.8246391337027887e-10 + ], + "bp_grad_F": [ + 0.009441024623811245, + 2.5670700779301114e-05, + 2.221704953342396e-08, + 1.750507294673298e-08, + 1.1021731971538884e-08 + ], + "gamma_dfa": 0.00038505904376506805, + "gamma_dfa_per_layer": [ + -0.0006303560803644359, + 0.0016098625492304564, + 0.00032150247716344893, + 0.00023922722903080285 + ], + "acc_eval": 0.244140625, + "loss_eval": 2.0488905906677246, + "epoch": 45 + }, + { + "hidden_norms_cls": [ + 46.00979995727539, + 36876176.0, + 37547776.0, + 45586304.0, + 59641656.0 + ], + "hidden_norms_avg": [ + 984.9158935546875, + 29720264.0, + 31840778.0, + 43341952.0, + 59202960.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.8270042093936354e-05, + 4.981586254615422e-10, + 3.011032534416813e-10, + 2.830454204350019e-10, + 2.763113071679868e-10 + ], + "bp_grad_F": [ + 0.01947774738073349, + 4.3539188482100144e-05, + 2.1426325602647012e-08, + 1.4225844502391283e-08, + 1.0958768115187922e-08 + ], + "gamma_dfa": 0.0003295415808679536, + "gamma_dfa_per_layer": [ + -0.0008024196140468121, + 0.0015490652294829488, + 0.0003235205076634884, + 0.0002480002003721893 + ], + "acc_eval": 0.2568359375, + "loss_eval": 2.0367085933685303, + "epoch": 46 + }, + { + "hidden_norms_cls": [ + 46.14581298828125, + 37542628.0, + 38211032.0, + 46396536.0, + 60543468.0 + ], + "hidden_norms_avg": [ + 988.5308837890625, + 31589698.0, + 33421126.0, + 44806960.0, + 60355708.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.7766556740971282e-05, + 4.28473617697378e-10, + 3.07322972625812e-10, + 2.9381372335102185e-10, + 2.797302944834712e-10 + ], + "bp_grad_F": [ + 0.06864868849515915, + 3.788271715166047e-05, + 2.1370066605186366e-08, + 1.4025233419090455e-08, + 1.1044066994259083e-08 + ], + "gamma_dfa": 9.462468733545393e-05, + "gamma_dfa_per_layer": [ + -0.0015880623832345009, + 0.0014036648208275437, + 0.0003099621389992535, + 0.00025293417274951935 + ], + "acc_eval": 0.244140625, + "loss_eval": 2.0631895065307617, + "epoch": 47 + }, + { + "hidden_norms_cls": [ + 46.26278305053711, + 38372568.0, + 39046204.0, + 47148944.0, + 61308416.0 + ], + "hidden_norms_avg": [ + 991.4658813476562, + 31279712.0, + 33238512.0, + 44902484.0, + 60826544.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.812536695273593e-05, + 4.3919237691092405e-10, + 3.053821084897379e-10, + 2.9203767182295337e-10, + 2.833127066281804e-10 + ], + "bp_grad_F": [ + 0.011117708869278431, + 2.0018645955133252e-05, + 1.9244726701117543e-08, + 1.4165241424279884e-08, + 1.090406964721069e-08 + ], + "gamma_dfa": 0.00016991631855489686, + "gamma_dfa_per_layer": [ + -0.0015358870150521398, + 0.0016542371595278382, + 0.0003059905138798058, + 0.00025532461586408317 + ], + "acc_eval": 0.2548828125, + "loss_eval": 2.0491740703582764, + "epoch": 48 + }, + { + "hidden_norms_cls": [ + 46.36857223510742, + 38788812.0, + 39461172.0, + 47582644.0, + 61831224.0 + ], + "hidden_norms_avg": [ + 994.04541015625, + 32215950.0, + 34096260.0, + 45599356.0, + 61425516.0 + ], + "bp_grad_per_sample_l2_med": [ + 3.0041363061172888e-05, + 4.480187609789965e-10, + 3.0562316566395964e-10, + 2.9085803210371353e-10, + 2.820148004012424e-10 + ], + "bp_grad_F": [ + 0.03149434179067612, + 6.209969433257356e-05, + 2.1469357847081483e-08, + 1.3228349970972886e-08, + 1.092844392758252e-08 + ], + "gamma_dfa": 0.00010758559074020013, + "gamma_dfa_per_layer": [ + -0.0014505982398986816, + 0.001327117788605392, + 0.0002981229918077588, + 0.0002556998224463314 + ], + "acc_eval": 0.244140625, + "loss_eval": 2.0554656982421875, + "epoch": 49 + }, + { + "hidden_norms_cls": [ + 46.4505729675293, + 38845108.0, + 39504860.0, + 47660992.0, + 62097052.0 + ], + "hidden_norms_avg": [ + 996.2877807617188, + 32178676.0, + 34106748.0, + 45494312.0, + 61553340.0 + ], + "bp_grad_per_sample_l2_med": [ + 3.0276962206698954e-05, + 4.5564105266571175e-10, + 3.0059965627771135e-10, + 2.8785077099691136e-10, + 2.794499076586021e-10 + ], + "bp_grad_F": [ + 0.011724308133125305, + 1.9755616449401714e-05, + 2.1592299503936374e-08, + 1.3796129572085647e-08, + 1.0886693324607677e-08 + ], + "gamma_dfa": 0.0002330350544070825, + "gamma_dfa_per_layer": [ + -0.0009828422917053103, + 0.0013922813814133406, + 0.00027237151516601443, + 0.00025032961275428534 + ], + "acc_eval": 0.2578125, + "loss_eval": 2.0504729747772217, + "epoch": 50 + }, + { + "hidden_norms_cls": [ + 46.521934509277344, + 39260604.0, + 39906660.0, + 48073644.0, + 62524844.0 + ], + "hidden_norms_avg": [ + 998.1359252929688, + 32138106.0, + 34129788.0, + 45757544.0, + 61958428.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.9659293431905098e-05, + 4.597033309572396e-10, + 3.0463259692581346e-10, + 2.893623118893629e-10, + 2.780784769118583e-10 + ], + "bp_grad_F": [ + 0.009033501148223877, + 1.7998809198616073e-05, + 2.5067617315244206e-08, + 1.8738653295713448e-08, + 1.087139978039886e-08 + ], + "gamma_dfa": 0.00016801172751002014, + "gamma_dfa_per_layer": [ + -0.0013465819647535682, + 0.0014174225507304072, + 0.00031502131605520844, + 0.00028618500800803304 + ], + "acc_eval": 0.255859375, + "loss_eval": 2.0473151206970215, + "epoch": 51 + }, + { + "hidden_norms_cls": [ + 46.57882308959961, + 39427768.0, + 40090548.0, + 48369740.0, + 62859588.0 + ], + "hidden_norms_avg": [ + 999.5881958007812, + 32546802.0, + 34502940.0, + 46179212.0, + 62426560.0 + ], + "bp_grad_per_sample_l2_med": [ + 3.0221139240893535e-05, + 4.495410987903625e-10, + 3.065779852207129e-10, + 2.9148336522233365e-10, + 2.8107866034687845e-10 + ], + "bp_grad_F": [ + 0.016372976824641228, + 2.868929186661262e-05, + 2.4158071099122935e-08, + 2.1377838166358742e-08, + 1.0851396226030374e-08 + ], + "gamma_dfa": 0.00012949823576491326, + "gamma_dfa_per_layer": [ + -0.0014246907085180283, + 0.0013606694992631674, + 0.0003105594660155475, + 0.0002714546862989664 + ], + "acc_eval": 0.2568359375, + "loss_eval": 2.0468568801879883, + "epoch": 52 + }, + { + "hidden_norms_cls": [ + 46.621891021728516, + 39709320.0, + 40367452.0, + 48619224.0, + 63258040.0 + ], + "hidden_norms_avg": [ + 1000.7721557617188, + 33325650.0, + 35170832.0, + 47006208.0, + 63117408.0 + ], + "bp_grad_per_sample_l2_med": [ + 3.0193185011739843e-05, + 4.388156504830931e-10, + 3.137757831339627e-10, + 2.8996427481331466e-10, + 2.7836044580453745e-10 + ], + "bp_grad_F": [ + 0.020163528621196747, + 3.054905027966015e-05, + 2.3064796295102497e-08, + 1.4526934322134366e-08, + 1.0852998499899513e-08 + ], + "gamma_dfa": 0.00017183857562486082, + "gamma_dfa_per_layer": [ + -0.0013185564894229174, + 0.0014156483812257648, + 0.00031098624458536506, + 0.00027927616611123085 + ], + "acc_eval": 0.2529296875, + "loss_eval": 2.054586410522461, + "epoch": 53 + }, + { + "hidden_norms_cls": [ + 46.65864562988281, + 39798476.0, + 40463920.0, + 48764504.0, + 63346472.0 + ], + "hidden_norms_avg": [ + 1001.6641845703125, + 33379128.0, + 35239452.0, + 46955488.0, + 63122060.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.965362000395544e-05, + 4.4001230437018535e-10, + 3.0546970508638083e-10, + 2.8820767994375274e-10, + 2.777780505613947e-10 + ], + "bp_grad_F": [ + 0.048469845205545425, + 4.790218736161478e-05, + 2.6222814142329298e-08, + 2.4079582772174035e-08, + 1.0861339383438917e-08 + ], + "gamma_dfa": -6.432890222640708e-05, + "gamma_dfa_per_layer": [ + -0.002126228529959917, + 0.0012497524730861187, + 0.00034036440774798393, + 0.0002787960402201861 + ], + "acc_eval": 0.2509765625, + "loss_eval": 2.0531256198883057, + "epoch": 54 + }, + { + "hidden_norms_cls": [ + 46.683528900146484, + 39922192.0, + 40544356.0, + 48855360.0, + 63435588.0 + ], + "hidden_norms_avg": [ + 1002.31494140625, + 33437468.0, + 35316604.0, + 47039952.0, + 63236136.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.79724281426752e-05, + 4.357196548010478e-10, + 3.0789479299464517e-10, + 2.8999944112761966e-10, + 2.7672003577450255e-10 + ], + "bp_grad_F": [ + 0.009082098491489887, + 6.448025487770792e-06, + 2.3372145108169207e-08, + 1.58603672417712e-08, + 1.0855457865943663e-08 + ], + "gamma_dfa": 8.819025242701173e-05, + "gamma_dfa_per_layer": [ + -0.0015805144794285297, + 0.0013432127889245749, + 0.000321696512401104, + 0.00026836618781089783 + ], + "acc_eval": 0.2529296875, + "loss_eval": 2.0517051219940186, + "epoch": 55 + }, + { + "hidden_norms_cls": [ + 46.70249938964844, + 40094040.0, + 40695052.0, + 48969004.0, + 63542232.0 + ], + "hidden_norms_avg": [ + 1002.7830810546875, + 33826664.0, + 35653116.0, + 47279168.0, + 63409756.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.8208709409227595e-05, + 4.3868975119210063e-10, + 3.102354484418868e-10, + 2.8893865078316594e-10, + 2.794520170823489e-10 + ], + "bp_grad_F": [ + 0.012775770388543606, + 3.5907680285163224e-05, + 3.276610627267473e-08, + 1.9733393585852355e-08, + 1.0837335473468102e-08 + ], + "gamma_dfa": -7.604052370879799e-06, + "gamma_dfa_per_layer": [ + -0.0020841513760387897, + 0.001464233035221696, + 0.0003138828033115715, + 0.0002756193280220032 + ], + "acc_eval": 0.2509765625, + "loss_eval": 2.053762912750244, + "epoch": 56 + }, + { + "hidden_norms_cls": [ + 46.71390914916992, + 39990840.0, + 40613704.0, + 48947088.0, + 63626668.0 + ], + "hidden_norms_avg": [ + 1003.0780639648438, + 33512400.0, + 35398544.0, + 47052320.0, + 63319172.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.8852005925728008e-05, + 4.408986231663192e-10, + 3.0767774439333095e-10, + 2.8811911190196327e-10, + 2.780060626150771e-10 + ], + "bp_grad_F": [ + 0.017775027081370354, + 2.9518716473830864e-05, + 1.6785485001946654e-08, + 1.3991450664718741e-08, + 1.082214762249123e-08 + ], + "gamma_dfa": 0.0001292051310883835, + "gamma_dfa_per_layer": [ + -0.001528067048639059, + 0.0014583747833967209, + 0.0003044582554139197, + 0.0002820545341819525 + ], + "acc_eval": 0.2548828125, + "loss_eval": 2.0512585639953613, + "epoch": 57 + }, + { + "hidden_norms_cls": [ + 46.72029495239258, + 40048976.0, + 40673908.0, + 49005780.0, + 63650516.0 + ], + "hidden_norms_avg": [ + 1003.2393188476562, + 33518072.0, + 35408260.0, + 47059300.0, + 63337588.0 + ], + "bp_grad_per_sample_l2_med": [ + 3.040391493414063e-05, + 4.422071042675668e-10, + 3.0922153726464785e-10, + 2.885988392709038e-10, + 2.7767882437856883e-10 + ], + "bp_grad_F": [ + 0.05996337905526161, + 4.702213846030645e-05, + 2.0060964445178797e-08, + 1.6814832193290385e-08, + 1.0825355722943186e-08 + ], + "gamma_dfa": 6.977042357902974e-05, + "gamma_dfa_per_layer": [ + -0.0016858375165611506, + 0.001342209754511714, + 0.00032920570811256766, + 0.00029350374825298786 + ], + "acc_eval": 0.25390625, + "loss_eval": 2.0525753498077393, + "epoch": 58 + }, + { + "hidden_norms_cls": [ + 46.723426818847656, + 40078224.0, + 40696452.0, + 49005116.0, + 63676392.0 + ], + "hidden_norms_avg": [ + 1003.3213500976562, + 33528454.0, + 35420932.0, + 47084140.0, + 63362884.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.9969949537189677e-05, + 4.438247824811725e-10, + 3.0574925924398144e-10, + 2.876155424935689e-10, + 2.772651552795935e-10 + ], + "bp_grad_F": [ + 0.015555496327579021, + 1.9468383470666595e-05, + 2.2645199493354085e-08, + 2.0153352764396004e-08, + 1.082723954937137e-08 + ], + "gamma_dfa": -2.766123361652717e-05, + "gamma_dfa_per_layer": [ + -0.0021022262517362833, + 0.0013793597463518381, + 0.0003187692491337657, + 0.0002934523217845708 + ], + "acc_eval": 0.2529296875, + "loss_eval": 2.05129337310791, + "epoch": 59 + }, + { + "hidden_norms_cls": [ + 46.72323226928711, + 40073516.0, + 40694276.0, + 48970848.0, + 63674696.0 + ], + "hidden_norms_avg": [ + 1003.3206176757812, + 33517796.0, + 35411552.0, + 47074216.0, + 63356572.0 + ], + "bp_grad_per_sample_l2_med": [ + 2.986433173646219e-05, + 4.4209438887499175e-10, + 3.069269838285038e-10, + 2.8717661582078335e-10, + 2.7721389073143143e-10 + ], + "bp_grad_F": [ + 0.011835500597953796, + 1.2977254300494678e-05, + 2.7325663509714104e-08, + 2.23609397664859e-08, + 1.0828303587118171e-08 + ], + "gamma_dfa": 4.879416519543156e-05, + "gamma_dfa_per_layer": [ + -0.0018596879672259092, + 0.0014319919282570481, + 0.0003201703366357833, + 0.00030270236311480403 + ], + "acc_eval": 0.2529296875, + "loss_eval": 2.051326274871826, + "epoch": 60 + } + ] +}
\ No newline at end of file |
