diff options
| author | YurenHao0426 <Blackhao0426@gmail.com> | 2026-06-14 04:06:32 -0500 |
|---|---|---|
| committer | YurenHao0426 <Blackhao0426@gmail.com> | 2026-06-14 04:06:32 -0500 |
| commit | aa73718eb6427d7da3b9cb416275802d90c4b2ed (patch) | |
| tree | b68b0a664fb650744ef934a1c22abd740a7b62a6 /results/snapshot_evolution_v2 | |
| parent | 827c658fa9a750f3c6ebdb87703762f10f69f6ff (diff) | |
Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Diffstat (limited to 'results/snapshot_evolution_v2')
| -rw-r--r-- | results/snapshot_evolution_v2/run_s123.log | 210 | ||||
| -rw-r--r-- | results/snapshot_evolution_v2/run_s42.log | 210 | ||||
| -rw-r--r-- | results/snapshot_evolution_v2/run_s456.log | 210 | ||||
| -rw-r--r-- | results/snapshot_evolution_v2/snapshot_evolution_s123.json | 5473 | ||||
| -rw-r--r-- | results/snapshot_evolution_v2/snapshot_evolution_s42.json | 5473 | ||||
| -rw-r--r-- | results/snapshot_evolution_v2/snapshot_evolution_s456.json | 5473 | ||||
| -rw-r--r-- | results/snapshot_evolution_v2/snapshot_fa_canonical_s42.json | 2734 | ||||
| -rw-r--r-- | results/snapshot_evolution_v2/snapshot_fa_s42.json | 2734 |
8 files changed, 22517 insertions, 0 deletions
diff --git a/results/snapshot_evolution_v2/run_s123.log b/results/snapshot_evolution_v2/run_s123.log new file mode 100644 index 0000000..c95da11 --- /dev/null +++ b/results/snapshot_evolution_v2/run_s123.log @@ -0,0 +1,210 @@ +device=cuda:0, depth=4, d_hidden=256, epochs=100, seed=123 +eval buffer: torch.Size([1024, 3072]) + +=== BP training === + [BP] Ep 0: ||h||_med=[8.690464973449707, 8.719978332519531, 8.73093318939209, 8.75937271118164, 8.790542602539062] ||g||_med=[0.0009977436857298017, 0.000992251792922616, 0.0009848480112850666, 0.0009761073160916567, 0.0009735578205436468] acc=0.0957 + [BP] Ep 1: ||h_L||=9.978e+01 ||g_2||=4.975e-05 acc=0.3965 + [BP] Ep 2: ||h_L||=1.189e+02 ||g_2||=4.264e-05 acc=0.4541 + [BP] Ep 3: ||h_L||=1.329e+02 ||g_2||=4.065e-05 acc=0.4814 + [BP] Ep 4: ||h_L||=1.422e+02 ||g_2||=4.144e-05 acc=0.4971 + [BP] Ep 5: ||h_L||=1.551e+02 ||g_2||=3.945e-05 acc=0.5078 + [BP] Ep 6: ||h_L||=1.671e+02 ||g_2||=3.917e-05 acc=0.5254 + [BP] Ep 7: ||h_L||=1.775e+02 ||g_2||=3.777e-05 acc=0.5156 + [BP] Ep 8: ||h_L||=1.825e+02 ||g_2||=3.695e-05 acc=0.5488 + [BP] Ep 9: ||h_L||=1.888e+02 ||g_2||=3.632e-05 acc=0.5430 + [BP] Ep 10: ||h_L||=1.954e+02 ||g_2||=3.601e-05 acc=0.5449 + [BP] Ep 11: ||h_L||=2.015e+02 ||g_2||=3.606e-05 acc=0.5557 + [BP] Ep 12: ||h_L||=2.081e+02 ||g_2||=3.473e-05 acc=0.5469 + [BP] Ep 13: ||h_L||=2.115e+02 ||g_2||=3.705e-05 acc=0.5332 + [BP] Ep 14: ||h_L||=2.165e+02 ||g_2||=3.531e-05 acc=0.5596 + [BP] Ep 15: ||h_L||=2.242e+02 ||g_2||=3.455e-05 acc=0.5615 + [BP] Ep 16: ||h_L||=2.239e+02 ||g_2||=3.574e-05 acc=0.5479 + [BP] Ep 17: ||h_L||=2.286e+02 ||g_2||=3.507e-05 acc=0.5498 + [BP] Ep 18: ||h_L||=2.338e+02 ||g_2||=3.669e-05 acc=0.5596 + [BP] Ep 19: ||h_L||=2.371e+02 ||g_2||=3.521e-05 acc=0.5605 + [BP] Ep 20: ||h_L||=2.396e+02 ||g_2||=3.595e-05 acc=0.5605 + [BP] Ep 21: ||h_L||=2.379e+02 ||g_2||=3.605e-05 acc=0.5781 + [BP] Ep 22: ||h_L||=2.430e+02 ||g_2||=3.571e-05 acc=0.5713 + [BP] Ep 23: ||h_L||=2.449e+02 ||g_2||=3.564e-05 acc=0.5586 + [BP] Ep 24: ||h_L||=2.481e+02 ||g_2||=3.683e-05 acc=0.5732 + [BP] Ep 25: ||h_L||=2.495e+02 ||g_2||=3.712e-05 acc=0.5771 + [BP] Ep 26: ||h_L||=2.474e+02 ||g_2||=3.760e-05 acc=0.5576 + [BP] Ep 27: ||h_L||=2.503e+02 ||g_2||=3.739e-05 acc=0.5635 + [BP] Ep 28: ||h_L||=2.517e+02 ||g_2||=3.695e-05 acc=0.5820 + [BP] Ep 29: ||h_L||=2.538e+02 ||g_2||=3.717e-05 acc=0.5771 + [BP] Ep 30: ||h_L||=2.501e+02 ||g_2||=3.752e-05 acc=0.5859 + [BP] Ep 31: ||h_L||=2.547e+02 ||g_2||=3.807e-05 acc=0.5850 + [BP] Ep 32: ||h_L||=2.542e+02 ||g_2||=3.772e-05 acc=0.5781 + [BP] Ep 33: ||h_L||=2.557e+02 ||g_2||=3.840e-05 acc=0.5713 + [BP] Ep 34: ||h_L||=2.555e+02 ||g_2||=3.730e-05 acc=0.5840 + [BP] Ep 35: ||h_L||=2.532e+02 ||g_2||=3.882e-05 acc=0.5986 + [BP] Ep 36: ||h_L||=2.578e+02 ||g_2||=3.771e-05 acc=0.5938 + [BP] Ep 37: ||h_L||=2.531e+02 ||g_2||=3.926e-05 acc=0.5869 + [BP] Ep 38: ||h_L||=2.559e+02 ||g_2||=3.848e-05 acc=0.5918 + [BP] Ep 39: ||h_L||=2.531e+02 ||g_2||=3.982e-05 acc=0.5762 + [BP] Ep 40: ||h_L||=2.516e+02 ||g_2||=3.895e-05 acc=0.5967 + [BP] Ep 41: ||h_L||=2.502e+02 ||g_2||=4.041e-05 acc=0.5967 + [BP] Ep 42: ||h_L||=2.541e+02 ||g_2||=4.035e-05 acc=0.5830 + [BP] Ep 43: ||h_L||=2.494e+02 ||g_2||=4.104e-05 acc=0.5830 + [BP] Ep 44: ||h_L||=2.519e+02 ||g_2||=4.092e-05 acc=0.5957 + [BP] Ep 45: ||h_L||=2.514e+02 ||g_2||=4.053e-05 acc=0.5732 + [BP] Ep 46: ||h_L||=2.462e+02 ||g_2||=4.221e-05 acc=0.6006 + [BP] Ep 47: ||h_L||=2.463e+02 ||g_2||=4.183e-05 acc=0.5996 + [BP] Ep 48: ||h_L||=2.455e+02 ||g_2||=4.155e-05 acc=0.6123 + [BP] Ep 49: ||h_L||=2.429e+02 ||g_2||=4.260e-05 acc=0.6006 + [BP] Ep 50: ||h_L||=2.424e+02 ||g_2||=4.242e-05 acc=0.5996 + [BP] Ep 51: ||h_L||=2.412e+02 ||g_2||=4.383e-05 acc=0.6104 + [BP] Ep 52: ||h_L||=2.402e+02 ||g_2||=4.466e-05 acc=0.6143 + [BP] Ep 53: ||h_L||=2.413e+02 ||g_2||=4.137e-05 acc=0.6152 + [BP] Ep 54: ||h_L||=2.388e+02 ||g_2||=4.456e-05 acc=0.6045 + [BP] Ep 55: ||h_L||=2.360e+02 ||g_2||=4.567e-05 acc=0.6025 + [BP] Ep 56: ||h_L||=2.356e+02 ||g_2||=4.226e-05 acc=0.6016 + [BP] Ep 57: ||h_L||=2.325e+02 ||g_2||=4.733e-05 acc=0.6016 + [BP] Ep 58: ||h_L||=2.322e+02 ||g_2||=4.771e-05 acc=0.6064 + [BP] Ep 59: ||h_L||=2.294e+02 ||g_2||=4.590e-05 acc=0.6143 + [BP] Ep 60: ||h_L||=2.299e+02 ||g_2||=4.810e-05 acc=0.6094 + [BP] Ep 61: ||h_L||=2.278e+02 ||g_2||=4.667e-05 acc=0.6045 + [BP] Ep 62: ||h_L||=2.248e+02 ||g_2||=4.761e-05 acc=0.6250 + [BP] Ep 63: ||h_L||=2.245e+02 ||g_2||=4.977e-05 acc=0.6162 + [BP] Ep 64: ||h_L||=2.243e+02 ||g_2||=4.972e-05 acc=0.6172 + [BP] Ep 65: ||h_L||=2.235e+02 ||g_2||=5.078e-05 acc=0.6113 + [BP] Ep 66: ||h_L||=2.205e+02 ||g_2||=4.961e-05 acc=0.6152 + [BP] Ep 67: ||h_L||=2.206e+02 ||g_2||=5.191e-05 acc=0.6104 + [BP] Ep 68: ||h_L||=2.189e+02 ||g_2||=5.423e-05 acc=0.6201 + [BP] Ep 69: ||h_L||=2.179e+02 ||g_2||=5.145e-05 acc=0.6152 + [BP] Ep 70: ||h_L||=2.163e+02 ||g_2||=5.319e-05 acc=0.6240 + [BP] Ep 71: ||h_L||=2.147e+02 ||g_2||=5.234e-05 acc=0.6094 + [BP] Ep 72: ||h_L||=2.142e+02 ||g_2||=5.207e-05 acc=0.6152 + [BP] Ep 73: ||h_L||=2.134e+02 ||g_2||=5.509e-05 acc=0.6152 + [BP] Ep 74: ||h_L||=2.122e+02 ||g_2||=5.333e-05 acc=0.6240 + [BP] Ep 75: ||h_L||=2.118e+02 ||g_2||=5.394e-05 acc=0.6143 + [BP] Ep 76: ||h_L||=2.104e+02 ||g_2||=5.282e-05 acc=0.6152 + [BP] Ep 77: ||h_L||=2.102e+02 ||g_2||=5.367e-05 acc=0.6182 + [BP] Ep 78: ||h_L||=2.086e+02 ||g_2||=5.341e-05 acc=0.6152 + [BP] Ep 79: ||h_L||=2.073e+02 ||g_2||=5.365e-05 acc=0.6113 + [BP] Ep 80: ||h_L||=2.062e+02 ||g_2||=5.509e-05 acc=0.6250 + [BP] Ep 81: ||h_L||=2.058e+02 ||g_2||=5.582e-05 acc=0.6221 + [BP] Ep 82: ||h_L||=2.049e+02 ||g_2||=5.464e-05 acc=0.6211 + [BP] Ep 83: ||h_L||=2.050e+02 ||g_2||=5.509e-05 acc=0.6143 + [BP] Ep 84: ||h_L||=2.043e+02 ||g_2||=5.509e-05 acc=0.6172 + [BP] Ep 85: ||h_L||=2.038e+02 ||g_2||=5.545e-05 acc=0.6191 + [BP] Ep 86: ||h_L||=2.032e+02 ||g_2||=5.440e-05 acc=0.6182 + [BP] Ep 87: ||h_L||=2.028e+02 ||g_2||=5.385e-05 acc=0.6221 + [BP] Ep 88: ||h_L||=2.031e+02 ||g_2||=5.425e-05 acc=0.6133 + [BP] Ep 89: ||h_L||=2.028e+02 ||g_2||=5.523e-05 acc=0.6162 + [BP] Ep 90: ||h_L||=2.022e+02 ||g_2||=5.593e-05 acc=0.6143 + [BP] Ep 91: ||h_L||=2.018e+02 ||g_2||=5.575e-05 acc=0.6191 + [BP] Ep 92: ||h_L||=2.019e+02 ||g_2||=5.540e-05 acc=0.6260 + [BP] Ep 93: ||h_L||=2.017e+02 ||g_2||=5.625e-05 acc=0.6211 + [BP] Ep 94: ||h_L||=2.018e+02 ||g_2||=5.597e-05 acc=0.6191 + [BP] Ep 95: ||h_L||=2.015e+02 ||g_2||=5.523e-05 acc=0.6201 + [BP] Ep 96: ||h_L||=2.015e+02 ||g_2||=5.587e-05 acc=0.6201 + [BP] Ep 97: ||h_L||=2.013e+02 ||g_2||=5.605e-05 acc=0.6191 + [BP] Ep 98: ||h_L||=2.013e+02 ||g_2||=5.599e-05 acc=0.6182 + [BP] Ep 99: ||h_L||=2.013e+02 ||g_2||=5.587e-05 acc=0.6182 + [BP] Ep 100: ||h_L||=2.013e+02 ||g_2||=5.585e-05 acc=0.6182 + +=== DFA training === + [DFA] Ep 0: ||h||_med=[8.690464973449707, 8.719978332519531, 8.73093318939209, 8.75937271118164, 8.790542602539062] ||g||_med=[0.0009977436857298017, 0.000992251792922616, 0.0009848480112850666, 0.0009761073160916567, 0.0009735578205436468] acc=0.0957 + [DFA] Ep 1: ||h_L||=6.963e+03 ||g_2||=1.315e-06 acc=0.3037 gamma_dfa=0.1066 + [DFA] Ep 2: ||h_L||=4.538e+04 ||g_2||=2.744e-07 acc=0.2930 gamma_dfa=0.1079 + [DFA] Ep 3: ||h_L||=1.244e+05 ||g_2||=1.145e-07 acc=0.3184 gamma_dfa=0.1085 + [DFA] Ep 4: ||h_L||=2.575e+05 ||g_2||=6.308e-08 acc=0.3174 gamma_dfa=0.1099 + [DFA] Ep 5: ||h_L||=4.700e+05 ||g_2||=3.892e-08 acc=0.3018 gamma_dfa=0.1084 + [DFA] Ep 6: ||h_L||=7.631e+05 ||g_2||=2.684e-08 acc=0.3096 gamma_dfa=0.1113 + [DFA] Ep 7: ||h_L||=1.163e+06 ||g_2||=1.897e-08 acc=0.3076 gamma_dfa=0.1105 + [DFA] Ep 8: ||h_L||=1.694e+06 ||g_2||=1.405e-08 acc=0.2979 gamma_dfa=0.1125 + [DFA] Ep 9: ||h_L||=2.356e+06 ||g_2||=1.107e-08 acc=0.3008 gamma_dfa=0.1132 + [DFA] Ep 10: ||h_L||=3.180e+06 ||g_2||=8.519e-09 acc=0.3086 gamma_dfa=0.1150 + [DFA] Ep 11: ||h_L||=4.215e+06 ||g_2||=7.026e-09 acc=0.3086 gamma_dfa=0.1165 + [DFA] Ep 12: ||h_L||=5.347e+06 ||g_2||=5.839e-09 acc=0.3047 gamma_dfa=0.1147 + [DFA] Ep 13: ||h_L||=6.688e+06 ||g_2||=4.985e-09 acc=0.2910 gamma_dfa=0.1138 + [DFA] Ep 14: ||h_L||=8.171e+06 ||g_2||=4.288e-09 acc=0.3086 gamma_dfa=0.1140 + [DFA] Ep 15: ||h_L||=9.891e+06 ||g_2||=3.811e-09 acc=0.3193 gamma_dfa=0.1132 + [DFA] Ep 16: ||h_L||=1.192e+07 ||g_2||=3.389e-09 acc=0.3066 gamma_dfa=0.1128 + [DFA] Ep 17: ||h_L||=1.402e+07 ||g_2||=3.017e-09 acc=0.3203 gamma_dfa=0.1126 + [DFA] Ep 18: ||h_L||=1.638e+07 ||g_2||=2.734e-09 acc=0.3154 gamma_dfa=0.1125 + [DFA] Ep 19: ||h_L||=1.904e+07 ||g_2||=2.466e-09 acc=0.3125 gamma_dfa=0.1116 + [DFA] Ep 20: ||h_L||=2.215e+07 ||g_2||=2.232e-09 acc=0.3096 gamma_dfa=0.1132 + [DFA] Ep 21: ||h_L||=2.519e+07 ||g_2||=2.051e-09 acc=0.3105 gamma_dfa=0.1127 + [DFA] Ep 22: ||h_L||=2.868e+07 ||g_2||=1.908e-09 acc=0.3105 gamma_dfa=0.1127 + [DFA] Ep 23: ||h_L||=3.257e+07 ||g_2||=1.751e-09 acc=0.3115 gamma_dfa=0.1119 + [DFA] Ep 24: ||h_L||=3.662e+07 ||g_2||=1.659e-09 acc=0.2979 gamma_dfa=0.1130 + [DFA] Ep 25: ||h_L||=4.115e+07 ||g_2||=1.500e-09 acc=0.2988 gamma_dfa=0.1120 + [DFA] Ep 26: ||h_L||=4.573e+07 ||g_2||=1.410e-09 acc=0.3115 gamma_dfa=0.1125 + [DFA] Ep 27: ||h_L||=5.048e+07 ||g_2||=1.332e-09 acc=0.2979 gamma_dfa=0.1115 + [DFA] Ep 28: ||h_L||=5.555e+07 ||g_2||=1.238e-09 acc=0.3193 gamma_dfa=0.1129 + [DFA] Ep 29: ||h_L||=6.069e+07 ||g_2||=1.189e-09 acc=0.3115 gamma_dfa=0.1119 + [DFA] Ep 30: ||h_L||=6.627e+07 ||g_2||=1.112e-09 acc=0.3125 gamma_dfa=0.1132 + [DFA] Ep 31: ||h_L||=7.196e+07 ||g_2||=1.059e-09 acc=0.3164 gamma_dfa=0.1123 + [DFA] Ep 32: ||h_L||=7.798e+07 ||g_2||=1.019e-09 acc=0.3145 gamma_dfa=0.1149 + [DFA] Ep 33: ||h_L||=8.452e+07 ||g_2||=9.572e-10 acc=0.3125 gamma_dfa=0.1125 + [DFA] Ep 34: ||h_L||=9.103e+07 ||g_2||=9.307e-10 acc=0.3125 gamma_dfa=0.1113 + [DFA] Ep 35: ||h_L||=9.773e+07 ||g_2||=8.747e-10 acc=0.3145 gamma_dfa=0.1136 + [DFA] Ep 36: ||h_L||=1.044e+08 ||g_2||=8.507e-10 acc=0.3291 gamma_dfa=0.1130 + [DFA] Ep 37: ||h_L||=1.115e+08 ||g_2||=8.077e-10 acc=0.3223 gamma_dfa=0.1129 + [DFA] Ep 38: ||h_L||=1.189e+08 ||g_2||=7.702e-10 acc=0.3047 gamma_dfa=0.1119 + [DFA] Ep 39: ||h_L||=1.265e+08 ||g_2||=7.407e-10 acc=0.3076 gamma_dfa=0.1119 + [DFA] Ep 40: ||h_L||=1.339e+08 ||g_2||=7.212e-10 acc=0.3291 gamma_dfa=0.1127 + [DFA] Ep 41: ||h_L||=1.419e+08 ||g_2||=7.015e-10 acc=0.3135 gamma_dfa=0.1130 + [DFA] Ep 42: ||h_L||=1.489e+08 ||g_2||=6.850e-10 acc=0.3145 gamma_dfa=0.1141 + [DFA] Ep 43: ||h_L||=1.566e+08 ||g_2||=6.554e-10 acc=0.3037 gamma_dfa=0.1121 + [DFA] Ep 44: ||h_L||=1.642e+08 ||g_2||=6.278e-10 acc=0.3174 gamma_dfa=0.1122 + [DFA] Ep 45: ||h_L||=1.716e+08 ||g_2||=6.384e-10 acc=0.3135 gamma_dfa=0.1135 + [DFA] Ep 46: ||h_L||=1.797e+08 ||g_2||=6.071e-10 acc=0.3213 gamma_dfa=0.1132 + [DFA] Ep 47: ||h_L||=1.878e+08 ||g_2||=5.891e-10 acc=0.3193 gamma_dfa=0.1123 + [DFA] Ep 48: ||h_L||=1.953e+08 ||g_2||=5.688e-10 acc=0.3252 gamma_dfa=0.1129 + [DFA] Ep 49: ||h_L||=2.031e+08 ||g_2||=5.532e-10 acc=0.3125 gamma_dfa=0.1124 + [DFA] Ep 50: ||h_L||=2.104e+08 ||g_2||=5.600e-10 acc=0.3301 gamma_dfa=0.1123 + [DFA] Ep 51: ||h_L||=2.182e+08 ||g_2||=5.298e-10 acc=0.3281 gamma_dfa=0.1118 + [DFA] Ep 52: ||h_L||=2.259e+08 ||g_2||=5.174e-10 acc=0.3301 gamma_dfa=0.1122 + [DFA] Ep 53: ||h_L||=2.336e+08 ||g_2||=5.160e-10 acc=0.3145 gamma_dfa=0.1123 + [DFA] Ep 54: ||h_L||=2.410e+08 ||g_2||=4.967e-10 acc=0.3320 gamma_dfa=0.1130 + [DFA] Ep 55: ||h_L||=2.485e+08 ||g_2||=4.874e-10 acc=0.3135 gamma_dfa=0.1126 + [DFA] Ep 56: ||h_L||=2.558e+08 ||g_2||=4.891e-10 acc=0.3145 gamma_dfa=0.1122 + [DFA] Ep 57: ||h_L||=2.627e+08 ||g_2||=4.827e-10 acc=0.3086 gamma_dfa=0.1129 + [DFA] Ep 58: ||h_L||=2.695e+08 ||g_2||=4.729e-10 acc=0.3301 gamma_dfa=0.1125 + [DFA] Ep 59: ||h_L||=2.762e+08 ||g_2||=4.574e-10 acc=0.3184 gamma_dfa=0.1127 + [DFA] Ep 60: ||h_L||=2.830e+08 ||g_2||=4.554e-10 acc=0.3125 gamma_dfa=0.1119 + [DFA] Ep 61: ||h_L||=2.893e+08 ||g_2||=4.464e-10 acc=0.3281 gamma_dfa=0.1126 + [DFA] Ep 62: ||h_L||=2.955e+08 ||g_2||=4.414e-10 acc=0.3145 gamma_dfa=0.1131 + [DFA] Ep 63: ||h_L||=3.015e+08 ||g_2||=4.405e-10 acc=0.3096 gamma_dfa=0.1133 + [DFA] Ep 64: ||h_L||=3.073e+08 ||g_2||=4.325e-10 acc=0.3125 gamma_dfa=0.1122 + [DFA] Ep 65: ||h_L||=3.130e+08 ||g_2||=4.353e-10 acc=0.3096 gamma_dfa=0.1132 + [DFA] Ep 66: ||h_L||=3.181e+08 ||g_2||=4.252e-10 acc=0.3164 gamma_dfa=0.1139 + [DFA] Ep 67: ||h_L||=3.232e+08 ||g_2||=4.156e-10 acc=0.3252 gamma_dfa=0.1130 + [DFA] Ep 68: ||h_L||=3.281e+08 ||g_2||=4.153e-10 acc=0.3223 gamma_dfa=0.1126 + [DFA] Ep 69: ||h_L||=3.328e+08 ||g_2||=4.167e-10 acc=0.3135 gamma_dfa=0.1135 + [DFA] Ep 70: ||h_L||=3.372e+08 ||g_2||=4.081e-10 acc=0.3115 gamma_dfa=0.1126 + [DFA] Ep 71: ||h_L||=3.416e+08 ||g_2||=4.053e-10 acc=0.3291 gamma_dfa=0.1126 + [DFA] Ep 72: ||h_L||=3.459e+08 ||g_2||=4.016e-10 acc=0.3096 gamma_dfa=0.1123 + [DFA] Ep 73: ||h_L||=3.497e+08 ||g_2||=3.978e-10 acc=0.3271 gamma_dfa=0.1132 + [DFA] Ep 74: ||h_L||=3.534e+08 ||g_2||=3.977e-10 acc=0.3057 gamma_dfa=0.1133 + [DFA] Ep 75: ||h_L||=3.566e+08 ||g_2||=3.958e-10 acc=0.3164 gamma_dfa=0.1127 + [DFA] Ep 76: ||h_L||=3.599e+08 ||g_2||=3.876e-10 acc=0.3242 gamma_dfa=0.1130 + [DFA] Ep 77: ||h_L||=3.629e+08 ||g_2||=3.873e-10 acc=0.3203 gamma_dfa=0.1131 + [DFA] Ep 78: ||h_L||=3.655e+08 ||g_2||=3.905e-10 acc=0.3145 gamma_dfa=0.1124 + [DFA] Ep 79: ||h_L||=3.682e+08 ||g_2||=3.866e-10 acc=0.3223 gamma_dfa=0.1131 + [DFA] Ep 80: ||h_L||=3.704e+08 ||g_2||=3.851e-10 acc=0.3252 gamma_dfa=0.1127 + [DFA] Ep 81: ||h_L||=3.724e+08 ||g_2||=3.836e-10 acc=0.3193 gamma_dfa=0.1128 + [DFA] Ep 82: ||h_L||=3.744e+08 ||g_2||=3.824e-10 acc=0.3262 gamma_dfa=0.1131 + [DFA] Ep 83: ||h_L||=3.761e+08 ||g_2||=3.835e-10 acc=0.3213 gamma_dfa=0.1129 + [DFA] Ep 84: ||h_L||=3.776e+08 ||g_2||=3.808e-10 acc=0.3184 gamma_dfa=0.1127 + [DFA] Ep 85: ||h_L||=3.790e+08 ||g_2||=3.804e-10 acc=0.3281 gamma_dfa=0.1129 + [DFA] Ep 86: ||h_L||=3.802e+08 ||g_2||=3.807e-10 acc=0.3154 gamma_dfa=0.1129 + [DFA] Ep 87: ||h_L||=3.812e+08 ||g_2||=3.793e-10 acc=0.3223 gamma_dfa=0.1128 + [DFA] Ep 88: ||h_L||=3.822e+08 ||g_2||=3.791e-10 acc=0.3184 gamma_dfa=0.1127 + [DFA] Ep 89: ||h_L||=3.830e+08 ||g_2||=3.775e-10 acc=0.3184 gamma_dfa=0.1129 + [DFA] Ep 90: ||h_L||=3.836e+08 ||g_2||=3.760e-10 acc=0.3262 gamma_dfa=0.1129 + [DFA] Ep 91: ||h_L||=3.841e+08 ||g_2||=3.769e-10 acc=0.3223 gamma_dfa=0.1128 + [DFA] Ep 92: ||h_L||=3.846e+08 ||g_2||=3.753e-10 acc=0.3184 gamma_dfa=0.1129 + [DFA] Ep 93: ||h_L||=3.849e+08 ||g_2||=3.765e-10 acc=0.3223 gamma_dfa=0.1129 + [DFA] Ep 94: ||h_L||=3.852e+08 ||g_2||=3.751e-10 acc=0.3184 gamma_dfa=0.1128 + [DFA] Ep 95: ||h_L||=3.854e+08 ||g_2||=3.753e-10 acc=0.3242 gamma_dfa=0.1129 + [DFA] Ep 96: ||h_L||=3.855e+08 ||g_2||=3.759e-10 acc=0.3203 gamma_dfa=0.1129 + [DFA] Ep 97: ||h_L||=3.856e+08 ||g_2||=3.758e-10 acc=0.3213 gamma_dfa=0.1129 + [DFA] Ep 98: ||h_L||=3.856e+08 ||g_2||=3.755e-10 acc=0.3203 gamma_dfa=0.1128 + [DFA] Ep 99: ||h_L||=3.857e+08 ||g_2||=3.755e-10 acc=0.3203 gamma_dfa=0.1128 + [DFA] Ep 100: ||h_L||=3.857e+08 ||g_2||=3.755e-10 acc=0.3203 gamma_dfa=0.1128 + +Saved results/snapshot_evolution_v2/snapshot_evolution_s123.json diff --git a/results/snapshot_evolution_v2/run_s42.log b/results/snapshot_evolution_v2/run_s42.log new file mode 100644 index 0000000..c764a8a --- /dev/null +++ b/results/snapshot_evolution_v2/run_s42.log @@ -0,0 +1,210 @@ +device=cuda:0, depth=4, d_hidden=256, epochs=100, seed=42 +eval buffer: torch.Size([1024, 3072]) + +=== BP training === + [BP] Ep 0: ||h||_med=[8.718070030212402, 8.780964851379395, 8.812002182006836, 8.828720092773438, 8.893179893493652] ||g||_med=[0.0010068593546748161, 0.000997236929833889, 0.0009934091940522194, 0.0009870644425973296, 0.0009829498594626784] acc=0.1152 + [BP] Ep 1: ||h_L||=9.381e+01 ||g_2||=5.500e-05 acc=0.3799 + [BP] Ep 2: ||h_L||=1.127e+02 ||g_2||=4.286e-05 acc=0.4541 + [BP] Ep 3: ||h_L||=1.245e+02 ||g_2||=4.270e-05 acc=0.4697 + [BP] Ep 4: ||h_L||=1.381e+02 ||g_2||=3.917e-05 acc=0.5039 + [BP] Ep 5: ||h_L||=1.489e+02 ||g_2||=3.916e-05 acc=0.4980 + [BP] Ep 6: ||h_L||=1.566e+02 ||g_2||=3.970e-05 acc=0.5234 + [BP] Ep 7: ||h_L||=1.659e+02 ||g_2||=3.728e-05 acc=0.5283 + [BP] Ep 8: ||h_L||=1.769e+02 ||g_2||=3.801e-05 acc=0.5410 + [BP] Ep 9: ||h_L||=1.814e+02 ||g_2||=3.649e-05 acc=0.5479 + [BP] Ep 10: ||h_L||=1.932e+02 ||g_2||=3.432e-05 acc=0.5664 + [BP] Ep 11: ||h_L||=1.958e+02 ||g_2||=3.381e-05 acc=0.5605 + [BP] Ep 12: ||h_L||=2.009e+02 ||g_2||=3.576e-05 acc=0.5459 + [BP] Ep 13: ||h_L||=2.073e+02 ||g_2||=3.389e-05 acc=0.5664 + [BP] Ep 14: ||h_L||=2.116e+02 ||g_2||=3.384e-05 acc=0.5664 + [BP] Ep 15: ||h_L||=2.138e+02 ||g_2||=3.492e-05 acc=0.5645 + [BP] Ep 16: ||h_L||=2.191e+02 ||g_2||=3.474e-05 acc=0.5742 + [BP] Ep 17: ||h_L||=2.209e+02 ||g_2||=3.491e-05 acc=0.5605 + [BP] Ep 18: ||h_L||=2.229e+02 ||g_2||=3.655e-05 acc=0.5557 + [BP] Ep 19: ||h_L||=2.255e+02 ||g_2||=3.763e-05 acc=0.5576 + [BP] Ep 20: ||h_L||=2.298e+02 ||g_2||=3.622e-05 acc=0.5635 + [BP] Ep 21: ||h_L||=2.360e+02 ||g_2||=3.521e-05 acc=0.5762 + [BP] Ep 22: ||h_L||=2.371e+02 ||g_2||=3.405e-05 acc=0.5879 + [BP] Ep 23: ||h_L||=2.383e+02 ||g_2||=3.356e-05 acc=0.5732 + [BP] Ep 24: ||h_L||=2.395e+02 ||g_2||=3.540e-05 acc=0.5723 + [BP] Ep 25: ||h_L||=2.381e+02 ||g_2||=3.570e-05 acc=0.5703 + [BP] Ep 26: ||h_L||=2.415e+02 ||g_2||=3.454e-05 acc=0.5791 + [BP] Ep 27: ||h_L||=2.422e+02 ||g_2||=3.610e-05 acc=0.5859 + [BP] Ep 28: ||h_L||=2.482e+02 ||g_2||=3.596e-05 acc=0.5811 + [BP] Ep 29: ||h_L||=2.454e+02 ||g_2||=3.742e-05 acc=0.5889 + [BP] Ep 30: ||h_L||=2.486e+02 ||g_2||=3.551e-05 acc=0.5742 + [BP] Ep 31: ||h_L||=2.451e+02 ||g_2||=3.772e-05 acc=0.5723 + [BP] Ep 32: ||h_L||=2.486e+02 ||g_2||=3.777e-05 acc=0.5762 + [BP] Ep 33: ||h_L||=2.471e+02 ||g_2||=3.571e-05 acc=0.5947 + [BP] Ep 34: ||h_L||=2.466e+02 ||g_2||=3.732e-05 acc=0.5996 + [BP] Ep 35: ||h_L||=2.483e+02 ||g_2||=3.716e-05 acc=0.5830 + [BP] Ep 36: ||h_L||=2.487e+02 ||g_2||=3.804e-05 acc=0.5820 + [BP] Ep 37: ||h_L||=2.517e+02 ||g_2||=3.812e-05 acc=0.5850 + [BP] Ep 38: ||h_L||=2.503e+02 ||g_2||=3.821e-05 acc=0.5938 + [BP] Ep 39: ||h_L||=2.469e+02 ||g_2||=3.820e-05 acc=0.5898 + [BP] Ep 40: ||h_L||=2.493e+02 ||g_2||=3.795e-05 acc=0.5918 + [BP] Ep 41: ||h_L||=2.472e+02 ||g_2||=3.859e-05 acc=0.5889 + [BP] Ep 42: ||h_L||=2.489e+02 ||g_2||=4.015e-05 acc=0.5898 + [BP] Ep 43: ||h_L||=2.455e+02 ||g_2||=3.909e-05 acc=0.5986 + [BP] Ep 44: ||h_L||=2.441e+02 ||g_2||=4.048e-05 acc=0.6025 + [BP] Ep 45: ||h_L||=2.469e+02 ||g_2||=3.789e-05 acc=0.6113 + [BP] Ep 46: ||h_L||=2.438e+02 ||g_2||=4.078e-05 acc=0.5947 + [BP] Ep 47: ||h_L||=2.418e+02 ||g_2||=4.107e-05 acc=0.6064 + [BP] Ep 48: ||h_L||=2.448e+02 ||g_2||=4.023e-05 acc=0.6025 + [BP] Ep 49: ||h_L||=2.422e+02 ||g_2||=4.135e-05 acc=0.5938 + [BP] Ep 50: ||h_L||=2.404e+02 ||g_2||=4.054e-05 acc=0.5938 + [BP] Ep 51: ||h_L||=2.403e+02 ||g_2||=4.141e-05 acc=0.6055 + [BP] Ep 52: ||h_L||=2.374e+02 ||g_2||=4.512e-05 acc=0.6035 + [BP] Ep 53: ||h_L||=2.367e+02 ||g_2||=4.363e-05 acc=0.5977 + [BP] Ep 54: ||h_L||=2.351e+02 ||g_2||=4.296e-05 acc=0.6172 + [BP] Ep 55: ||h_L||=2.323e+02 ||g_2||=4.397e-05 acc=0.6006 + [BP] Ep 56: ||h_L||=2.336e+02 ||g_2||=4.373e-05 acc=0.6113 + [BP] Ep 57: ||h_L||=2.309e+02 ||g_2||=4.451e-05 acc=0.6143 + [BP] Ep 58: ||h_L||=2.279e+02 ||g_2||=4.527e-05 acc=0.6113 + [BP] Ep 59: ||h_L||=2.295e+02 ||g_2||=4.667e-05 acc=0.5928 + [BP] Ep 60: ||h_L||=2.257e+02 ||g_2||=4.720e-05 acc=0.6094 + [BP] Ep 61: ||h_L||=2.266e+02 ||g_2||=4.760e-05 acc=0.6045 + [BP] Ep 62: ||h_L||=2.230e+02 ||g_2||=4.822e-05 acc=0.6035 + [BP] Ep 63: ||h_L||=2.228e+02 ||g_2||=4.939e-05 acc=0.6094 + [BP] Ep 64: ||h_L||=2.222e+02 ||g_2||=4.937e-05 acc=0.5986 + [BP] Ep 65: ||h_L||=2.203e+02 ||g_2||=4.871e-05 acc=0.6123 + [BP] Ep 66: ||h_L||=2.196e+02 ||g_2||=5.067e-05 acc=0.6113 + [BP] Ep 67: ||h_L||=2.173e+02 ||g_2||=5.006e-05 acc=0.5898 + [BP] Ep 68: ||h_L||=2.179e+02 ||g_2||=4.987e-05 acc=0.6084 + [BP] Ep 69: ||h_L||=2.160e+02 ||g_2||=5.053e-05 acc=0.6152 + [BP] Ep 70: ||h_L||=2.133e+02 ||g_2||=5.005e-05 acc=0.6074 + [BP] Ep 71: ||h_L||=2.140e+02 ||g_2||=5.274e-05 acc=0.6113 + [BP] Ep 72: ||h_L||=2.128e+02 ||g_2||=5.268e-05 acc=0.6094 + [BP] Ep 73: ||h_L||=2.124e+02 ||g_2||=5.362e-05 acc=0.6113 + [BP] Ep 74: ||h_L||=2.106e+02 ||g_2||=5.307e-05 acc=0.6074 + [BP] Ep 75: ||h_L||=2.098e+02 ||g_2||=5.283e-05 acc=0.6201 + [BP] Ep 76: ||h_L||=2.092e+02 ||g_2||=5.413e-05 acc=0.6094 + [BP] Ep 77: ||h_L||=2.082e+02 ||g_2||=5.412e-05 acc=0.6143 + [BP] Ep 78: ||h_L||=2.077e+02 ||g_2||=5.385e-05 acc=0.6152 + [BP] Ep 79: ||h_L||=2.068e+02 ||g_2||=5.596e-05 acc=0.5996 + [BP] Ep 80: ||h_L||=2.056e+02 ||g_2||=5.548e-05 acc=0.6094 + [BP] Ep 81: ||h_L||=2.050e+02 ||g_2||=5.472e-05 acc=0.6104 + [BP] Ep 82: ||h_L||=2.042e+02 ||g_2||=5.538e-05 acc=0.6201 + [BP] Ep 83: ||h_L||=2.034e+02 ||g_2||=5.665e-05 acc=0.6035 + [BP] Ep 84: ||h_L||=2.034e+02 ||g_2||=5.529e-05 acc=0.6104 + [BP] Ep 85: ||h_L||=2.027e+02 ||g_2||=5.481e-05 acc=0.6152 + [BP] Ep 86: ||h_L||=2.026e+02 ||g_2||=5.497e-05 acc=0.6104 + [BP] Ep 87: ||h_L||=2.024e+02 ||g_2||=5.641e-05 acc=0.6084 + [BP] Ep 88: ||h_L||=2.023e+02 ||g_2||=5.639e-05 acc=0.6113 + [BP] Ep 89: ||h_L||=2.021e+02 ||g_2||=5.677e-05 acc=0.6084 + [BP] Ep 90: ||h_L||=2.022e+02 ||g_2||=5.742e-05 acc=0.6035 + [BP] Ep 91: ||h_L||=2.019e+02 ||g_2||=5.576e-05 acc=0.6113 + [BP] Ep 92: ||h_L||=2.015e+02 ||g_2||=5.681e-05 acc=0.6016 + [BP] Ep 93: ||h_L||=2.012e+02 ||g_2||=5.702e-05 acc=0.6104 + [BP] Ep 94: ||h_L||=2.011e+02 ||g_2||=5.615e-05 acc=0.6123 + [BP] Ep 95: ||h_L||=2.010e+02 ||g_2||=5.590e-05 acc=0.6084 + [BP] Ep 96: ||h_L||=2.010e+02 ||g_2||=5.641e-05 acc=0.6074 + [BP] Ep 97: ||h_L||=2.009e+02 ||g_2||=5.674e-05 acc=0.6094 + [BP] Ep 98: ||h_L||=2.009e+02 ||g_2||=5.674e-05 acc=0.6084 + [BP] Ep 99: ||h_L||=2.009e+02 ||g_2||=5.668e-05 acc=0.6094 + [BP] Ep 100: ||h_L||=2.009e+02 ||g_2||=5.669e-05 acc=0.6094 + +=== DFA training === + [DFA] Ep 0: ||h||_med=[8.718070030212402, 8.780964851379395, 8.812002182006836, 8.828720092773438, 8.893179893493652] ||g||_med=[0.0010068593546748161, 0.000997236929833889, 0.0009934091940522194, 0.0009870644425973296, 0.0009829498594626784] acc=0.1152 + [DFA] Ep 1: ||h_L||=7.304e+03 ||g_2||=1.375e-06 acc=0.3105 gamma_dfa=0.0899 + [DFA] Ep 2: ||h_L||=3.776e+04 ||g_2||=3.111e-07 acc=0.2881 gamma_dfa=0.0858 + [DFA] Ep 3: ||h_L||=1.046e+05 ||g_2||=1.275e-07 acc=0.3018 gamma_dfa=0.0863 + [DFA] Ep 4: ||h_L||=2.235e+05 ||g_2||=6.730e-08 acc=0.3076 gamma_dfa=0.0886 + [DFA] Ep 5: ||h_L||=4.042e+05 ||g_2||=4.211e-08 acc=0.3281 gamma_dfa=0.0876 + [DFA] Ep 6: ||h_L||=6.774e+05 ||g_2||=2.808e-08 acc=0.2920 gamma_dfa=0.0842 + [DFA] Ep 7: ||h_L||=1.027e+06 ||g_2||=2.087e-08 acc=0.2920 gamma_dfa=0.0859 + [DFA] Ep 8: ||h_L||=1.513e+06 ||g_2||=1.577e-08 acc=0.3066 gamma_dfa=0.0852 + [DFA] Ep 9: ||h_L||=2.150e+06 ||g_2||=1.248e-08 acc=0.3184 gamma_dfa=0.0876 + [DFA] Ep 10: ||h_L||=2.934e+06 ||g_2||=1.044e-08 acc=0.3037 gamma_dfa=0.0876 + [DFA] Ep 11: ||h_L||=3.849e+06 ||g_2||=8.960e-09 acc=0.3018 gamma_dfa=0.0892 + [DFA] Ep 12: ||h_L||=4.946e+06 ||g_2||=7.846e-09 acc=0.2871 gamma_dfa=0.0921 + [DFA] Ep 13: ||h_L||=6.325e+06 ||g_2||=6.816e-09 acc=0.2900 gamma_dfa=0.0919 + [DFA] Ep 14: ||h_L||=7.941e+06 ||g_2||=5.938e-09 acc=0.2979 gamma_dfa=0.0949 + [DFA] Ep 15: ||h_L||=9.822e+06 ||g_2||=5.388e-09 acc=0.2852 gamma_dfa=0.0958 + [DFA] Ep 16: ||h_L||=1.202e+07 ||g_2||=4.830e-09 acc=0.2979 gamma_dfa=0.0975 + [DFA] Ep 17: ||h_L||=1.430e+07 ||g_2||=4.410e-09 acc=0.3115 gamma_dfa=0.0999 + [DFA] Ep 18: ||h_L||=1.700e+07 ||g_2||=4.013e-09 acc=0.2969 gamma_dfa=0.0988 + [DFA] Ep 19: ||h_L||=2.002e+07 ||g_2||=3.688e-09 acc=0.2949 gamma_dfa=0.1002 + [DFA] Ep 20: ||h_L||=2.311e+07 ||g_2||=3.374e-09 acc=0.3086 gamma_dfa=0.1007 + [DFA] Ep 21: ||h_L||=2.656e+07 ||g_2||=3.083e-09 acc=0.3154 gamma_dfa=0.1018 + [DFA] Ep 22: ||h_L||=3.043e+07 ||g_2||=2.851e-09 acc=0.3096 gamma_dfa=0.1025 + [DFA] Ep 23: ||h_L||=3.467e+07 ||g_2||=2.638e-09 acc=0.3086 gamma_dfa=0.1020 + [DFA] Ep 24: ||h_L||=3.916e+07 ||g_2||=2.428e-09 acc=0.3008 gamma_dfa=0.1046 + [DFA] Ep 25: ||h_L||=4.389e+07 ||g_2||=2.233e-09 acc=0.3232 gamma_dfa=0.1029 + [DFA] Ep 26: ||h_L||=4.919e+07 ||g_2||=2.056e-09 acc=0.3037 gamma_dfa=0.1025 + [DFA] Ep 27: ||h_L||=5.469e+07 ||g_2||=1.928e-09 acc=0.3105 gamma_dfa=0.1034 + [DFA] Ep 28: ||h_L||=6.011e+07 ||g_2||=1.815e-09 acc=0.3027 gamma_dfa=0.1046 + [DFA] Ep 29: ||h_L||=6.573e+07 ||g_2||=1.725e-09 acc=0.3262 gamma_dfa=0.1051 + [DFA] Ep 30: ||h_L||=7.193e+07 ||g_2||=1.626e-09 acc=0.3018 gamma_dfa=0.1059 + [DFA] Ep 31: ||h_L||=7.819e+07 ||g_2||=1.523e-09 acc=0.3213 gamma_dfa=0.1048 + [DFA] Ep 32: ||h_L||=8.508e+07 ||g_2||=1.429e-09 acc=0.3086 gamma_dfa=0.1060 + [DFA] Ep 33: ||h_L||=9.217e+07 ||g_2||=1.367e-09 acc=0.3115 gamma_dfa=0.1053 + [DFA] Ep 34: ||h_L||=9.964e+07 ||g_2||=1.290e-09 acc=0.3008 gamma_dfa=0.1059 + [DFA] Ep 35: ||h_L||=1.069e+08 ||g_2||=1.223e-09 acc=0.3252 gamma_dfa=0.1062 + [DFA] Ep 36: ||h_L||=1.145e+08 ||g_2||=1.168e-09 acc=0.2998 gamma_dfa=0.1048 + [DFA] Ep 37: ||h_L||=1.224e+08 ||g_2||=1.128e-09 acc=0.3027 gamma_dfa=0.1052 + [DFA] Ep 38: ||h_L||=1.306e+08 ||g_2||=1.069e-09 acc=0.3193 gamma_dfa=0.1057 + [DFA] Ep 39: ||h_L||=1.385e+08 ||g_2||=1.027e-09 acc=0.3027 gamma_dfa=0.1056 + [DFA] Ep 40: ||h_L||=1.470e+08 ||g_2||=9.871e-10 acc=0.3184 gamma_dfa=0.1058 + [DFA] Ep 41: ||h_L||=1.555e+08 ||g_2||=9.513e-10 acc=0.3145 gamma_dfa=0.1057 + [DFA] Ep 42: ||h_L||=1.640e+08 ||g_2||=9.198e-10 acc=0.3057 gamma_dfa=0.1057 + [DFA] Ep 43: ||h_L||=1.727e+08 ||g_2||=8.925e-10 acc=0.3115 gamma_dfa=0.1065 + [DFA] Ep 44: ||h_L||=1.813e+08 ||g_2||=8.598e-10 acc=0.2939 gamma_dfa=0.1057 + [DFA] Ep 45: ||h_L||=1.903e+08 ||g_2||=8.300e-10 acc=0.3037 gamma_dfa=0.1056 + [DFA] Ep 46: ||h_L||=1.990e+08 ||g_2||=8.036e-10 acc=0.3115 gamma_dfa=0.1062 + [DFA] Ep 47: ||h_L||=2.081e+08 ||g_2||=7.852e-10 acc=0.3008 gamma_dfa=0.1059 + [DFA] Ep 48: ||h_L||=2.167e+08 ||g_2||=7.675e-10 acc=0.3037 gamma_dfa=0.1058 + [DFA] Ep 49: ||h_L||=2.257e+08 ||g_2||=7.444e-10 acc=0.3057 gamma_dfa=0.1067 + [DFA] Ep 50: ||h_L||=2.352e+08 ||g_2||=7.281e-10 acc=0.3076 gamma_dfa=0.1065 + [DFA] Ep 51: ||h_L||=2.442e+08 ||g_2||=7.057e-10 acc=0.3145 gamma_dfa=0.1064 + [DFA] Ep 52: ||h_L||=2.533e+08 ||g_2||=6.872e-10 acc=0.3125 gamma_dfa=0.1063 + [DFA] Ep 53: ||h_L||=2.617e+08 ||g_2||=6.789e-10 acc=0.3105 gamma_dfa=0.1070 + [DFA] Ep 54: ||h_L||=2.702e+08 ||g_2||=6.633e-10 acc=0.2998 gamma_dfa=0.1055 + [DFA] Ep 55: ||h_L||=2.789e+08 ||g_2||=6.521e-10 acc=0.2949 gamma_dfa=0.1067 + [DFA] Ep 56: ||h_L||=2.873e+08 ||g_2||=6.409e-10 acc=0.3076 gamma_dfa=0.1073 + [DFA] Ep 57: ||h_L||=2.952e+08 ||g_2||=6.233e-10 acc=0.3096 gamma_dfa=0.1072 + [DFA] Ep 58: ||h_L||=3.031e+08 ||g_2||=6.182e-10 acc=0.3027 gamma_dfa=0.1066 + [DFA] Ep 59: ||h_L||=3.109e+08 ||g_2||=6.106e-10 acc=0.3027 gamma_dfa=0.1067 + [DFA] Ep 60: ||h_L||=3.188e+08 ||g_2||=5.949e-10 acc=0.3145 gamma_dfa=0.1071 + [DFA] Ep 61: ||h_L||=3.262e+08 ||g_2||=5.864e-10 acc=0.3164 gamma_dfa=0.1067 + [DFA] Ep 62: ||h_L||=3.332e+08 ||g_2||=5.788e-10 acc=0.3105 gamma_dfa=0.1066 + [DFA] Ep 63: ||h_L||=3.400e+08 ||g_2||=5.673e-10 acc=0.3154 gamma_dfa=0.1068 + [DFA] Ep 64: ||h_L||=3.464e+08 ||g_2||=5.629e-10 acc=0.3115 gamma_dfa=0.1065 + [DFA] Ep 65: ||h_L||=3.534e+08 ||g_2||=5.576e-10 acc=0.3154 gamma_dfa=0.1068 + [DFA] Ep 66: ||h_L||=3.598e+08 ||g_2||=5.556e-10 acc=0.3125 gamma_dfa=0.1067 + [DFA] Ep 67: ||h_L||=3.659e+08 ||g_2||=5.460e-10 acc=0.3047 gamma_dfa=0.1065 + [DFA] Ep 68: ||h_L||=3.718e+08 ||g_2||=5.413e-10 acc=0.2988 gamma_dfa=0.1065 + [DFA] Ep 69: ||h_L||=3.776e+08 ||g_2||=5.308e-10 acc=0.3105 gamma_dfa=0.1071 + [DFA] Ep 70: ||h_L||=3.828e+08 ||g_2||=5.315e-10 acc=0.3076 gamma_dfa=0.1070 + [DFA] Ep 71: ||h_L||=3.877e+08 ||g_2||=5.271e-10 acc=0.3145 gamma_dfa=0.1069 + [DFA] Ep 72: ||h_L||=3.923e+08 ||g_2||=5.266e-10 acc=0.2998 gamma_dfa=0.1065 + [DFA] Ep 73: ||h_L||=3.969e+08 ||g_2||=5.205e-10 acc=0.3018 gamma_dfa=0.1070 + [DFA] Ep 74: ||h_L||=4.011e+08 ||g_2||=5.160e-10 acc=0.3076 gamma_dfa=0.1066 + [DFA] Ep 75: ||h_L||=4.053e+08 ||g_2||=5.108e-10 acc=0.3096 gamma_dfa=0.1070 + [DFA] Ep 76: ||h_L||=4.091e+08 ||g_2||=5.070e-10 acc=0.3193 gamma_dfa=0.1070 + [DFA] Ep 77: ||h_L||=4.125e+08 ||g_2||=5.052e-10 acc=0.3076 gamma_dfa=0.1069 + [DFA] Ep 78: ||h_L||=4.158e+08 ||g_2||=5.040e-10 acc=0.3066 gamma_dfa=0.1070 + [DFA] Ep 79: ||h_L||=4.187e+08 ||g_2||=5.055e-10 acc=0.3018 gamma_dfa=0.1068 + [DFA] Ep 80: ||h_L||=4.214e+08 ||g_2||=4.991e-10 acc=0.3105 gamma_dfa=0.1069 + [DFA] Ep 81: ||h_L||=4.237e+08 ||g_2||=4.971e-10 acc=0.3076 gamma_dfa=0.1070 + [DFA] Ep 82: ||h_L||=4.259e+08 ||g_2||=4.927e-10 acc=0.3057 gamma_dfa=0.1070 + [DFA] Ep 83: ||h_L||=4.278e+08 ||g_2||=4.936e-10 acc=0.3047 gamma_dfa=0.1069 + [DFA] Ep 84: ||h_L||=4.297e+08 ||g_2||=4.920e-10 acc=0.3076 gamma_dfa=0.1069 + [DFA] Ep 85: ||h_L||=4.313e+08 ||g_2||=4.936e-10 acc=0.3057 gamma_dfa=0.1069 + [DFA] Ep 86: ||h_L||=4.328e+08 ||g_2||=4.903e-10 acc=0.3145 gamma_dfa=0.1068 + [DFA] Ep 87: ||h_L||=4.341e+08 ||g_2||=4.904e-10 acc=0.3086 gamma_dfa=0.1068 + [DFA] Ep 88: ||h_L||=4.351e+08 ||g_2||=4.894e-10 acc=0.3115 gamma_dfa=0.1069 + [DFA] Ep 89: ||h_L||=4.361e+08 ||g_2||=4.855e-10 acc=0.3135 gamma_dfa=0.1070 + [DFA] Ep 90: ||h_L||=4.369e+08 ||g_2||=4.873e-10 acc=0.3096 gamma_dfa=0.1069 + [DFA] Ep 91: ||h_L||=4.375e+08 ||g_2||=4.876e-10 acc=0.3066 gamma_dfa=0.1069 + [DFA] Ep 92: ||h_L||=4.380e+08 ||g_2||=4.862e-10 acc=0.3086 gamma_dfa=0.1068 + [DFA] Ep 93: ||h_L||=4.385e+08 ||g_2||=4.860e-10 acc=0.3096 gamma_dfa=0.1069 + [DFA] Ep 94: ||h_L||=4.388e+08 ||g_2||=4.854e-10 acc=0.3076 gamma_dfa=0.1069 + [DFA] Ep 95: ||h_L||=4.390e+08 ||g_2||=4.861e-10 acc=0.3086 gamma_dfa=0.1069 + [DFA] Ep 96: ||h_L||=4.392e+08 ||g_2||=4.863e-10 acc=0.3066 gamma_dfa=0.1069 + [DFA] Ep 97: ||h_L||=4.393e+08 ||g_2||=4.859e-10 acc=0.3096 gamma_dfa=0.1069 + [DFA] Ep 98: ||h_L||=4.394e+08 ||g_2||=4.857e-10 acc=0.3066 gamma_dfa=0.1069 + [DFA] Ep 99: ||h_L||=4.394e+08 ||g_2||=4.857e-10 acc=0.3076 gamma_dfa=0.1069 + [DFA] Ep 100: ||h_L||=4.394e+08 ||g_2||=4.857e-10 acc=0.3076 gamma_dfa=0.1069 + +Saved results/snapshot_evolution_v2/snapshot_evolution_s42.json diff --git a/results/snapshot_evolution_v2/run_s456.log b/results/snapshot_evolution_v2/run_s456.log new file mode 100644 index 0000000..4ccb115 --- /dev/null +++ b/results/snapshot_evolution_v2/run_s456.log @@ -0,0 +1,210 @@ +device=cuda:0, depth=4, d_hidden=256, epochs=100, seed=456 +eval buffer: torch.Size([1024, 3072]) + +=== BP training === + [BP] Ep 0: ||h||_med=[8.696551322937012, 8.737784385681152, 8.813130378723145, 8.844642639160156, 8.889528274536133] ||g||_med=[0.0010079912608489394, 0.0010035138111561537, 0.0009970925748348236, 0.0009936393471434712, 0.000986678060144186] acc=0.0928 + [BP] Ep 1: ||h_L||=9.917e+01 ||g_2||=5.310e-05 acc=0.4004 + [BP] Ep 2: ||h_L||=1.152e+02 ||g_2||=4.300e-05 acc=0.4600 + [BP] Ep 3: ||h_L||=1.251e+02 ||g_2||=4.203e-05 acc=0.4932 + [BP] Ep 4: ||h_L||=1.367e+02 ||g_2||=4.196e-05 acc=0.4902 + [BP] Ep 5: ||h_L||=1.479e+02 ||g_2||=3.941e-05 acc=0.5039 + [BP] Ep 6: ||h_L||=1.583e+02 ||g_2||=3.835e-05 acc=0.5166 + [BP] Ep 7: ||h_L||=1.656e+02 ||g_2||=3.787e-05 acc=0.5283 + [BP] Ep 8: ||h_L||=1.764e+02 ||g_2||=3.733e-05 acc=0.5293 + [BP] Ep 9: ||h_L||=1.842e+02 ||g_2||=3.600e-05 acc=0.5391 + [BP] Ep 10: ||h_L||=1.915e+02 ||g_2||=3.614e-05 acc=0.5537 + [BP] Ep 11: ||h_L||=1.981e+02 ||g_2||=3.464e-05 acc=0.5449 + [BP] Ep 12: ||h_L||=2.000e+02 ||g_2||=3.492e-05 acc=0.5439 + [BP] Ep 13: ||h_L||=2.067e+02 ||g_2||=3.524e-05 acc=0.5518 + [BP] Ep 14: ||h_L||=2.096e+02 ||g_2||=3.480e-05 acc=0.5479 + [BP] Ep 15: ||h_L||=2.130e+02 ||g_2||=3.509e-05 acc=0.5693 + [BP] Ep 16: ||h_L||=2.143e+02 ||g_2||=3.710e-05 acc=0.5508 + [BP] Ep 17: ||h_L||=2.198e+02 ||g_2||=3.612e-05 acc=0.5840 + [BP] Ep 18: ||h_L||=2.220e+02 ||g_2||=3.491e-05 acc=0.5752 + [BP] Ep 19: ||h_L||=2.260e+02 ||g_2||=3.519e-05 acc=0.5674 + [BP] Ep 20: ||h_L||=2.310e+02 ||g_2||=3.526e-05 acc=0.5664 + [BP] Ep 21: ||h_L||=2.293e+02 ||g_2||=3.503e-05 acc=0.5957 + [BP] Ep 22: ||h_L||=2.355e+02 ||g_2||=3.525e-05 acc=0.5762 + [BP] Ep 23: ||h_L||=2.381e+02 ||g_2||=3.591e-05 acc=0.5684 + [BP] Ep 24: ||h_L||=2.411e+02 ||g_2||=3.563e-05 acc=0.5889 + [BP] Ep 25: ||h_L||=2.405e+02 ||g_2||=3.554e-05 acc=0.5820 + [BP] Ep 26: ||h_L||=2.425e+02 ||g_2||=3.670e-05 acc=0.6025 + [BP] Ep 27: ||h_L||=2.436e+02 ||g_2||=3.618e-05 acc=0.5762 + [BP] Ep 28: ||h_L||=2.474e+02 ||g_2||=3.597e-05 acc=0.5879 + [BP] Ep 29: ||h_L||=2.489e+02 ||g_2||=3.712e-05 acc=0.5938 + [BP] Ep 30: ||h_L||=2.480e+02 ||g_2||=3.656e-05 acc=0.5889 + [BP] Ep 31: ||h_L||=2.501e+02 ||g_2||=3.727e-05 acc=0.5674 + [BP] Ep 32: ||h_L||=2.484e+02 ||g_2||=3.738e-05 acc=0.5986 + [BP] Ep 33: ||h_L||=2.528e+02 ||g_2||=3.573e-05 acc=0.5898 + [BP] Ep 34: ||h_L||=2.496e+02 ||g_2||=3.846e-05 acc=0.5928 + [BP] Ep 35: ||h_L||=2.481e+02 ||g_2||=3.687e-05 acc=0.6152 + [BP] Ep 36: ||h_L||=2.507e+02 ||g_2||=3.670e-05 acc=0.6016 + [BP] Ep 37: ||h_L||=2.485e+02 ||g_2||=3.572e-05 acc=0.5996 + [BP] Ep 38: ||h_L||=2.492e+02 ||g_2||=3.780e-05 acc=0.5908 + [BP] Ep 39: ||h_L||=2.507e+02 ||g_2||=3.791e-05 acc=0.6152 + [BP] Ep 40: ||h_L||=2.484e+02 ||g_2||=3.952e-05 acc=0.6104 + [BP] Ep 41: ||h_L||=2.478e+02 ||g_2||=3.848e-05 acc=0.6084 + [BP] Ep 42: ||h_L||=2.470e+02 ||g_2||=3.780e-05 acc=0.6143 + [BP] Ep 43: ||h_L||=2.466e+02 ||g_2||=3.813e-05 acc=0.6211 + [BP] Ep 44: ||h_L||=2.432e+02 ||g_2||=3.922e-05 acc=0.6182 + [BP] Ep 45: ||h_L||=2.431e+02 ||g_2||=4.048e-05 acc=0.6133 + [BP] Ep 46: ||h_L||=2.431e+02 ||g_2||=3.831e-05 acc=0.6113 + [BP] Ep 47: ||h_L||=2.413e+02 ||g_2||=4.016e-05 acc=0.6133 + [BP] Ep 48: ||h_L||=2.407e+02 ||g_2||=3.886e-05 acc=0.6162 + [BP] Ep 49: ||h_L||=2.395e+02 ||g_2||=4.190e-05 acc=0.6104 + [BP] Ep 50: ||h_L||=2.385e+02 ||g_2||=4.175e-05 acc=0.6152 + [BP] Ep 51: ||h_L||=2.367e+02 ||g_2||=4.073e-05 acc=0.6191 + [BP] Ep 52: ||h_L||=2.373e+02 ||g_2||=4.066e-05 acc=0.6172 + [BP] Ep 53: ||h_L||=2.351e+02 ||g_2||=4.393e-05 acc=0.6104 + [BP] Ep 54: ||h_L||=2.338e+02 ||g_2||=4.099e-05 acc=0.6230 + [BP] Ep 55: ||h_L||=2.329e+02 ||g_2||=4.365e-05 acc=0.6270 + [BP] Ep 56: ||h_L||=2.333e+02 ||g_2||=4.267e-05 acc=0.6309 + [BP] Ep 57: ||h_L||=2.305e+02 ||g_2||=4.151e-05 acc=0.6309 + [BP] Ep 58: ||h_L||=2.282e+02 ||g_2||=4.239e-05 acc=0.6416 + [BP] Ep 59: ||h_L||=2.257e+02 ||g_2||=4.316e-05 acc=0.6279 + [BP] Ep 60: ||h_L||=2.256e+02 ||g_2||=4.490e-05 acc=0.6240 + [BP] Ep 61: ||h_L||=2.244e+02 ||g_2||=4.582e-05 acc=0.6289 + [BP] Ep 62: ||h_L||=2.224e+02 ||g_2||=4.648e-05 acc=0.6270 + [BP] Ep 63: ||h_L||=2.206e+02 ||g_2||=4.636e-05 acc=0.6211 + [BP] Ep 64: ||h_L||=2.205e+02 ||g_2||=4.671e-05 acc=0.6250 + [BP] Ep 65: ||h_L||=2.196e+02 ||g_2||=4.449e-05 acc=0.6270 + [BP] Ep 66: ||h_L||=2.172e+02 ||g_2||=4.745e-05 acc=0.6377 + [BP] Ep 67: ||h_L||=2.152e+02 ||g_2||=4.708e-05 acc=0.6367 + [BP] Ep 68: ||h_L||=2.151e+02 ||g_2||=4.813e-05 acc=0.6289 + [BP] Ep 69: ||h_L||=2.136e+02 ||g_2||=4.880e-05 acc=0.6328 + [BP] Ep 70: ||h_L||=2.124e+02 ||g_2||=4.823e-05 acc=0.6270 + [BP] Ep 71: ||h_L||=2.107e+02 ||g_2||=4.963e-05 acc=0.6230 + [BP] Ep 72: ||h_L||=2.113e+02 ||g_2||=4.797e-05 acc=0.6133 + [BP] Ep 73: ||h_L||=2.090e+02 ||g_2||=4.736e-05 acc=0.6318 + [BP] Ep 74: ||h_L||=2.096e+02 ||g_2||=4.832e-05 acc=0.6270 + [BP] Ep 75: ||h_L||=2.081e+02 ||g_2||=4.982e-05 acc=0.6240 + [BP] Ep 76: ||h_L||=2.075e+02 ||g_2||=4.877e-05 acc=0.6260 + [BP] Ep 77: ||h_L||=2.060e+02 ||g_2||=4.900e-05 acc=0.6221 + [BP] Ep 78: ||h_L||=2.052e+02 ||g_2||=4.906e-05 acc=0.6279 + [BP] Ep 79: ||h_L||=2.050e+02 ||g_2||=5.181e-05 acc=0.6211 + [BP] Ep 80: ||h_L||=2.045e+02 ||g_2||=4.986e-05 acc=0.6260 + [BP] Ep 81: ||h_L||=2.039e+02 ||g_2||=5.293e-05 acc=0.6289 + [BP] Ep 82: ||h_L||=2.032e+02 ||g_2||=5.169e-05 acc=0.6191 + [BP] Ep 83: ||h_L||=2.025e+02 ||g_2||=5.152e-05 acc=0.6260 + [BP] Ep 84: ||h_L||=2.019e+02 ||g_2||=5.158e-05 acc=0.6250 + [BP] Ep 85: ||h_L||=2.018e+02 ||g_2||=5.156e-05 acc=0.6250 + [BP] Ep 86: ||h_L||=2.015e+02 ||g_2||=5.196e-05 acc=0.6309 + [BP] Ep 87: ||h_L||=2.015e+02 ||g_2||=5.104e-05 acc=0.6289 + [BP] Ep 88: ||h_L||=2.017e+02 ||g_2||=5.253e-05 acc=0.6270 + [BP] Ep 89: ||h_L||=2.011e+02 ||g_2||=5.326e-05 acc=0.6299 + [BP] Ep 90: ||h_L||=2.005e+02 ||g_2||=5.368e-05 acc=0.6279 + [BP] Ep 91: ||h_L||=2.000e+02 ||g_2||=5.282e-05 acc=0.6289 + [BP] Ep 92: ||h_L||=1.999e+02 ||g_2||=5.349e-05 acc=0.6270 + [BP] Ep 93: ||h_L||=1.998e+02 ||g_2||=5.212e-05 acc=0.6289 + [BP] Ep 94: ||h_L||=1.993e+02 ||g_2||=5.377e-05 acc=0.6299 + [BP] Ep 95: ||h_L||=1.994e+02 ||g_2||=5.352e-05 acc=0.6309 + [BP] Ep 96: ||h_L||=1.994e+02 ||g_2||=5.367e-05 acc=0.6289 + [BP] Ep 97: ||h_L||=1.994e+02 ||g_2||=5.366e-05 acc=0.6289 + [BP] Ep 98: ||h_L||=1.993e+02 ||g_2||=5.374e-05 acc=0.6289 + [BP] Ep 99: ||h_L||=1.994e+02 ||g_2||=5.385e-05 acc=0.6289 + [BP] Ep 100: ||h_L||=1.994e+02 ||g_2||=5.381e-05 acc=0.6279 + +=== DFA training === + [DFA] Ep 0: ||h||_med=[8.696551322937012, 8.737784385681152, 8.813130378723145, 8.844642639160156, 8.889528274536133] ||g||_med=[0.0010079912608489394, 0.0010035138111561537, 0.0009970925748348236, 0.0009936393471434712, 0.000986678060144186] acc=0.0928 + [DFA] Ep 1: ||h_L||=1.347e+04 ||g_2||=7.784e-07 acc=0.2861 gamma_dfa=0.1029 + [DFA] Ep 2: ||h_L||=7.387e+04 ||g_2||=1.706e-07 acc=0.2988 gamma_dfa=0.0959 + [DFA] Ep 3: ||h_L||=2.014e+05 ||g_2||=7.528e-08 acc=0.2900 gamma_dfa=0.0944 + [DFA] Ep 4: ||h_L||=4.167e+05 ||g_2||=4.114e-08 acc=0.2988 gamma_dfa=0.0917 + [DFA] Ep 5: ||h_L||=7.642e+05 ||g_2||=2.491e-08 acc=0.2900 gamma_dfa=0.0899 + [DFA] Ep 6: ||h_L||=1.235e+06 ||g_2||=1.726e-08 acc=0.3057 gamma_dfa=0.0875 + [DFA] Ep 7: ||h_L||=1.913e+06 ||g_2||=1.233e-08 acc=0.2920 gamma_dfa=0.0888 + [DFA] Ep 8: ||h_L||=2.850e+06 ||g_2||=9.213e-09 acc=0.3115 gamma_dfa=0.0877 + [DFA] Ep 9: ||h_L||=4.005e+06 ||g_2||=7.223e-09 acc=0.2900 gamma_dfa=0.0894 + [DFA] Ep 10: ||h_L||=5.501e+06 ||g_2||=5.828e-09 acc=0.2979 gamma_dfa=0.0913 + [DFA] Ep 11: ||h_L||=7.181e+06 ||g_2||=4.814e-09 acc=0.2822 gamma_dfa=0.0899 + [DFA] Ep 12: ||h_L||=9.249e+06 ||g_2||=4.032e-09 acc=0.2998 gamma_dfa=0.0917 + [DFA] Ep 13: ||h_L||=1.180e+07 ||g_2||=3.444e-09 acc=0.3135 gamma_dfa=0.0932 + [DFA] Ep 14: ||h_L||=1.461e+07 ||g_2||=3.012e-09 acc=0.2861 gamma_dfa=0.0934 + [DFA] Ep 15: ||h_L||=1.793e+07 ||g_2||=2.598e-09 acc=0.2979 gamma_dfa=0.0936 + [DFA] Ep 16: ||h_L||=2.168e+07 ||g_2||=2.271e-09 acc=0.3086 gamma_dfa=0.0933 + [DFA] Ep 17: ||h_L||=2.591e+07 ||g_2||=2.026e-09 acc=0.2969 gamma_dfa=0.0945 + [DFA] Ep 18: ||h_L||=3.070e+07 ||g_2||=1.802e-09 acc=0.3086 gamma_dfa=0.0939 + [DFA] Ep 19: ||h_L||=3.588e+07 ||g_2||=1.638e-09 acc=0.3027 gamma_dfa=0.0952 + [DFA] Ep 20: ||h_L||=4.144e+07 ||g_2||=1.484e-09 acc=0.2959 gamma_dfa=0.0941 + [DFA] Ep 21: ||h_L||=4.772e+07 ||g_2||=1.368e-09 acc=0.2959 gamma_dfa=0.0949 + [DFA] Ep 22: ||h_L||=5.457e+07 ||g_2||=1.246e-09 acc=0.2900 gamma_dfa=0.0948 + [DFA] Ep 23: ||h_L||=6.201e+07 ||g_2||=1.159e-09 acc=0.2939 gamma_dfa=0.0947 + [DFA] Ep 24: ||h_L||=6.970e+07 ||g_2||=1.067e-09 acc=0.2969 gamma_dfa=0.0951 + [DFA] Ep 25: ||h_L||=7.780e+07 ||g_2||=9.980e-10 acc=0.2861 gamma_dfa=0.0948 + [DFA] Ep 26: ||h_L||=8.627e+07 ||g_2||=9.362e-10 acc=0.2979 gamma_dfa=0.0951 + [DFA] Ep 27: ||h_L||=9.569e+07 ||g_2||=8.744e-10 acc=0.2842 gamma_dfa=0.0955 + [DFA] Ep 28: ||h_L||=1.058e+08 ||g_2||=8.262e-10 acc=0.2891 gamma_dfa=0.0953 + [DFA] Ep 29: ||h_L||=1.159e+08 ||g_2||=7.846e-10 acc=0.2920 gamma_dfa=0.0960 + [DFA] Ep 30: ||h_L||=1.267e+08 ||g_2||=7.353e-10 acc=0.2959 gamma_dfa=0.0951 + [DFA] Ep 31: ||h_L||=1.380e+08 ||g_2||=6.977e-10 acc=0.3086 gamma_dfa=0.0952 + [DFA] Ep 32: ||h_L||=1.493e+08 ||g_2||=6.666e-10 acc=0.2949 gamma_dfa=0.0953 + [DFA] Ep 33: ||h_L||=1.617e+08 ||g_2||=6.391e-10 acc=0.2881 gamma_dfa=0.0947 + [DFA] Ep 34: ||h_L||=1.738e+08 ||g_2||=6.205e-10 acc=0.2822 gamma_dfa=0.0956 + [DFA] Ep 35: ||h_L||=1.866e+08 ||g_2||=5.893e-10 acc=0.2988 gamma_dfa=0.0959 + [DFA] Ep 36: ||h_L||=1.990e+08 ||g_2||=5.731e-10 acc=0.2949 gamma_dfa=0.0956 + [DFA] Ep 37: ||h_L||=2.125e+08 ||g_2||=5.475e-10 acc=0.3027 gamma_dfa=0.0953 + [DFA] Ep 38: ||h_L||=2.263e+08 ||g_2||=5.326e-10 acc=0.2891 gamma_dfa=0.0949 + [DFA] Ep 39: ||h_L||=2.401e+08 ||g_2||=5.206e-10 acc=0.3027 gamma_dfa=0.0958 + [DFA] Ep 40: ||h_L||=2.549e+08 ||g_2||=4.933e-10 acc=0.2959 gamma_dfa=0.0952 + [DFA] Ep 41: ||h_L||=2.702e+08 ||g_2||=4.775e-10 acc=0.3086 gamma_dfa=0.0962 + [DFA] Ep 42: ||h_L||=2.844e+08 ||g_2||=4.679e-10 acc=0.2871 gamma_dfa=0.0957 + [DFA] Ep 43: ||h_L||=2.990e+08 ||g_2||=4.493e-10 acc=0.3066 gamma_dfa=0.0961 + [DFA] Ep 44: ||h_L||=3.144e+08 ||g_2||=4.412e-10 acc=0.2900 gamma_dfa=0.0955 + [DFA] Ep 45: ||h_L||=3.282e+08 ||g_2||=4.348e-10 acc=0.2998 gamma_dfa=0.0961 + [DFA] Ep 46: ||h_L||=3.423e+08 ||g_2||=4.229e-10 acc=0.3018 gamma_dfa=0.0956 + [DFA] Ep 47: ||h_L||=3.566e+08 ||g_2||=4.139e-10 acc=0.2969 gamma_dfa=0.0957 + [DFA] Ep 48: ||h_L||=3.713e+08 ||g_2||=4.000e-10 acc=0.3027 gamma_dfa=0.0961 + [DFA] Ep 49: ||h_L||=3.859e+08 ||g_2||=3.904e-10 acc=0.3105 gamma_dfa=0.0957 + [DFA] Ep 50: ||h_L||=4.007e+08 ||g_2||=3.812e-10 acc=0.3008 gamma_dfa=0.0961 + [DFA] Ep 51: ||h_L||=4.150e+08 ||g_2||=3.752e-10 acc=0.3047 gamma_dfa=0.0957 + [DFA] Ep 52: ||h_L||=4.302e+08 ||g_2||=3.718e-10 acc=0.2910 gamma_dfa=0.0954 + [DFA] Ep 53: ||h_L||=4.447e+08 ||g_2||=3.651e-10 acc=0.2881 gamma_dfa=0.0957 + [DFA] Ep 54: ||h_L||=4.586e+08 ||g_2||=3.605e-10 acc=0.2998 gamma_dfa=0.0961 + [DFA] Ep 55: ||h_L||=4.722e+08 ||g_2||=3.467e-10 acc=0.3018 gamma_dfa=0.0963 + [DFA] Ep 56: ||h_L||=4.857e+08 ||g_2||=3.460e-10 acc=0.3047 gamma_dfa=0.0954 + [DFA] Ep 57: ||h_L||=4.994e+08 ||g_2||=3.379e-10 acc=0.2949 gamma_dfa=0.0962 + [DFA] Ep 58: ||h_L||=5.127e+08 ||g_2||=3.307e-10 acc=0.3018 gamma_dfa=0.0962 + [DFA] Ep 59: ||h_L||=5.255e+08 ||g_2||=3.281e-10 acc=0.3008 gamma_dfa=0.0960 + [DFA] Ep 60: ||h_L||=5.379e+08 ||g_2||=3.225e-10 acc=0.2988 gamma_dfa=0.0961 + [DFA] Ep 61: ||h_L||=5.493e+08 ||g_2||=3.184e-10 acc=0.3047 gamma_dfa=0.0962 + [DFA] Ep 62: ||h_L||=5.612e+08 ||g_2||=3.170e-10 acc=0.3027 gamma_dfa=0.0955 + [DFA] Ep 63: ||h_L||=5.725e+08 ||g_2||=3.109e-10 acc=0.2939 gamma_dfa=0.0960 + [DFA] Ep 64: ||h_L||=5.833e+08 ||g_2||=3.055e-10 acc=0.3076 gamma_dfa=0.0956 + [DFA] Ep 65: ||h_L||=5.946e+08 ||g_2||=3.044e-10 acc=0.2979 gamma_dfa=0.0959 + [DFA] Ep 66: ||h_L||=6.045e+08 ||g_2||=2.993e-10 acc=0.3008 gamma_dfa=0.0963 + [DFA] Ep 67: ||h_L||=6.137e+08 ||g_2||=2.988e-10 acc=0.3105 gamma_dfa=0.0961 + [DFA] Ep 68: ||h_L||=6.227e+08 ||g_2||=2.978e-10 acc=0.2939 gamma_dfa=0.0962 + [DFA] Ep 69: ||h_L||=6.309e+08 ||g_2||=2.942e-10 acc=0.2979 gamma_dfa=0.0964 + [DFA] Ep 70: ||h_L||=6.393e+08 ||g_2||=2.940e-10 acc=0.2930 gamma_dfa=0.0964 + [DFA] Ep 71: ||h_L||=6.472e+08 ||g_2||=2.872e-10 acc=0.3018 gamma_dfa=0.0963 + [DFA] Ep 72: ||h_L||=6.547e+08 ||g_2||=2.899e-10 acc=0.2959 gamma_dfa=0.0961 + [DFA] Ep 73: ||h_L||=6.618e+08 ||g_2||=2.843e-10 acc=0.2959 gamma_dfa=0.0962 + [DFA] Ep 74: ||h_L||=6.688e+08 ||g_2||=2.825e-10 acc=0.2998 gamma_dfa=0.0962 + [DFA] Ep 75: ||h_L||=6.755e+08 ||g_2||=2.794e-10 acc=0.3066 gamma_dfa=0.0963 + [DFA] Ep 76: ||h_L||=6.812e+08 ||g_2||=2.790e-10 acc=0.2959 gamma_dfa=0.0962 + [DFA] Ep 77: ||h_L||=6.867e+08 ||g_2||=2.792e-10 acc=0.3066 gamma_dfa=0.0958 + [DFA] Ep 78: ||h_L||=6.919e+08 ||g_2||=2.797e-10 acc=0.2988 gamma_dfa=0.0960 + [DFA] Ep 79: ||h_L||=6.969e+08 ||g_2||=2.750e-10 acc=0.3037 gamma_dfa=0.0959 + [DFA] Ep 80: ||h_L||=7.010e+08 ||g_2||=2.738e-10 acc=0.3066 gamma_dfa=0.0961 + [DFA] Ep 81: ||h_L||=7.049e+08 ||g_2||=2.723e-10 acc=0.3008 gamma_dfa=0.0961 + [DFA] Ep 82: ||h_L||=7.085e+08 ||g_2||=2.722e-10 acc=0.3018 gamma_dfa=0.0960 + [DFA] Ep 83: ||h_L||=7.118e+08 ||g_2||=2.707e-10 acc=0.3018 gamma_dfa=0.0962 + [DFA] Ep 84: ||h_L||=7.148e+08 ||g_2||=2.702e-10 acc=0.3027 gamma_dfa=0.0962 + [DFA] Ep 85: ||h_L||=7.174e+08 ||g_2||=2.683e-10 acc=0.3096 gamma_dfa=0.0961 + [DFA] Ep 86: ||h_L||=7.197e+08 ||g_2||=2.699e-10 acc=0.2969 gamma_dfa=0.0962 + [DFA] Ep 87: ||h_L||=7.218e+08 ||g_2||=2.699e-10 acc=0.2998 gamma_dfa=0.0960 + [DFA] Ep 88: ||h_L||=7.235e+08 ||g_2||=2.703e-10 acc=0.2979 gamma_dfa=0.0960 + [DFA] Ep 89: ||h_L||=7.250e+08 ||g_2||=2.679e-10 acc=0.3018 gamma_dfa=0.0962 + [DFA] Ep 90: ||h_L||=7.262e+08 ||g_2||=2.680e-10 acc=0.2998 gamma_dfa=0.0960 + [DFA] Ep 91: ||h_L||=7.273e+08 ||g_2||=2.691e-10 acc=0.2969 gamma_dfa=0.0960 + [DFA] Ep 92: ||h_L||=7.281e+08 ||g_2||=2.671e-10 acc=0.3018 gamma_dfa=0.0961 + [DFA] Ep 93: ||h_L||=7.288e+08 ||g_2||=2.677e-10 acc=0.3008 gamma_dfa=0.0961 + [DFA] Ep 94: ||h_L||=7.293e+08 ||g_2||=2.685e-10 acc=0.2979 gamma_dfa=0.0961 + [DFA] Ep 95: ||h_L||=7.297e+08 ||g_2||=2.670e-10 acc=0.2998 gamma_dfa=0.0961 + [DFA] Ep 96: ||h_L||=7.299e+08 ||g_2||=2.674e-10 acc=0.2998 gamma_dfa=0.0961 + [DFA] Ep 97: ||h_L||=7.300e+08 ||g_2||=2.671e-10 acc=0.2988 gamma_dfa=0.0961 + [DFA] Ep 98: ||h_L||=7.301e+08 ||g_2||=2.672e-10 acc=0.2998 gamma_dfa=0.0961 + [DFA] Ep 99: ||h_L||=7.302e+08 ||g_2||=2.671e-10 acc=0.2998 gamma_dfa=0.0961 + [DFA] Ep 100: ||h_L||=7.302e+08 ||g_2||=2.672e-10 acc=0.2998 gamma_dfa=0.0961 + +Saved results/snapshot_evolution_v2/snapshot_evolution_s456.json diff --git a/results/snapshot_evolution_v2/snapshot_evolution_s123.json b/results/snapshot_evolution_v2/snapshot_evolution_s123.json new file mode 100644 index 0000000..4ccefc8 --- /dev/null +++ b/results/snapshot_evolution_v2/snapshot_evolution_s123.json @@ -0,0 +1,5473 @@ +{ + "config": { + "output_dir": "results/snapshot_evolution_v2", + "epochs": 100, + "lr": 0.001, + "wd": 0.01, + "seed": 123, + "depth": 4, + "d_hidden": 256, + "log_every": 1 + }, + "depth": 4, + "d_hidden": 256, + "num_classes": 10, + "bp_log": [ + { + "hidden_norms": [ + 8.690464973449707, + 8.719978332519531, + 8.73093318939209, + 8.75937271118164, + 8.790542602539062 + ], + "bp_grad_norms_per_sample_med": [ + 0.0009977436857298017, + 0.000992251792922616, + 0.0009848480112850666, + 0.0009761073160916567, + 0.0009735578205436468 + ], + "bp_grad_norms_F": [ + 0.035340383648872375, + 0.03499162197113037, + 0.03454425558447838, + 0.034113530069589615, + 0.033874060958623886 + ], + "gamma_dfa": NaN, + "acc_eval": 0.095703125, + "loss_eval": 2.486203908920288, + "epoch": 0 + }, + { + "hidden_norms": [ + 78.8471908569336, + 81.27874755859375, + 85.74082946777344, + 93.0085678100586, + 99.78337860107422 + ], + "bp_grad_norms_per_sample_med": [ + 5.44870927114971e-05, + 5.06363830936607e-05, + 4.9749945901567116e-05, + 5.010423774365336e-05, + 5.131187572260387e-05 + ], + "bp_grad_norms_F": [ + 0.0025542855728417635, + 0.0021727080456912518, + 0.002010831143707037, + 0.0019116367911919951, + 0.0019338354468345642 + ], + "gamma_dfa": NaN, + "acc_eval": 0.396484375, + "loss_eval": 1.6637382507324219, + "epoch": 1 + }, + { + "hidden_norms": [ + 94.79721069335938, + 98.6406021118164, + 103.38369750976562, + 111.4940185546875, + 118.88549041748047 + ], + "bp_grad_norms_per_sample_med": [ + 4.842087219003588e-05, + 4.405825529829599e-05, + 4.263824666850269e-05, + 4.2027873860206455e-05, + 4.283937960281037e-05 + ], + "bp_grad_norms_F": [ + 0.0020658739376813173, + 0.001806268934160471, + 0.001665781601332128, + 0.0015872427029535174, + 0.0015909546054899693 + ], + "gamma_dfa": NaN, + "acc_eval": 0.4541015625, + "loss_eval": 1.5350896120071411, + "epoch": 2 + }, + { + "hidden_norms": [ + 112.59109497070312, + 116.00851440429688, + 118.50328063964844, + 125.58804321289062, + 132.86842346191406 + ], + "bp_grad_norms_per_sample_med": [ + 4.3564141378737986e-05, + 4.136884308536537e-05, + 4.064830864081159e-05, + 3.9523682062281296e-05, + 3.8838741602376103e-05 + ], + "bp_grad_norms_F": [ + 0.00200729351490736, + 0.001785867614671588, + 0.0016562910750508308, + 0.0015452943043783307, + 0.0015010037459433079 + ], + "gamma_dfa": NaN, + "acc_eval": 0.4814453125, + "loss_eval": 1.449176549911499, + "epoch": 3 + }, + { + "hidden_norms": [ + 124.81511688232422, + 128.57049560546875, + 129.71421813964844, + 136.4766387939453, + 142.1854705810547 + ], + "bp_grad_norms_per_sample_med": [ + 4.382780389278196e-05, + 4.2459021642571315e-05, + 4.143684418522753e-05, + 4.0614067984279245e-05, + 3.921420648111962e-05 + ], + "bp_grad_norms_F": [ + 0.0018197334138676524, + 0.0016814853297546506, + 0.0015934663824737072, + 0.0015063106548041105, + 0.0014485444407910109 + ], + "gamma_dfa": NaN, + "acc_eval": 0.4970703125, + "loss_eval": 1.4218697547912598, + "epoch": 4 + }, + { + "hidden_norms": [ + 138.24440002441406, + 140.13018798828125, + 141.3839874267578, + 147.2488555908203, + 155.097412109375 + ], + "bp_grad_norms_per_sample_med": [ + 4.045376044814475e-05, + 4.011149940197356e-05, + 3.9452468627132475e-05, + 3.831404319498688e-05, + 3.721412576851435e-05 + ], + "bp_grad_norms_F": [ + 0.0017204430187121034, + 0.0016258049290627241, + 0.001551075023598969, + 0.0014644934562966228, + 0.0013933938462287188 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5078125, + "loss_eval": 1.3836241960525513, + "epoch": 5 + }, + { + "hidden_norms": [ + 152.11605834960938, + 153.1806182861328, + 153.2488555908203, + 159.2217559814453, + 167.07318115234375 + ], + "bp_grad_norms_per_sample_med": [ + 3.981084591941908e-05, + 3.969454701291397e-05, + 3.9165588532341644e-05, + 3.808845940511674e-05, + 3.58163051714655e-05 + ], + "bp_grad_norms_F": [ + 0.0017226944910362363, + 0.001611364888958633, + 0.0015574879944324493, + 0.001471089432016015, + 0.001374671934172511 + ], + "gamma_dfa": NaN, + "acc_eval": 0.525390625, + "loss_eval": 1.372281551361084, + "epoch": 6 + }, + { + "hidden_norms": [ + 161.75347900390625, + 163.58787536621094, + 163.35731506347656, + 170.1693115234375, + 177.47879028320312 + ], + "bp_grad_norms_per_sample_med": [ + 3.749227471416816e-05, + 3.707691575982608e-05, + 3.776738958549686e-05, + 3.692766040330753e-05, + 3.5146917070960626e-05 + ], + "bp_grad_norms_F": [ + 0.0016339722787961364, + 0.00156014587264508, + 0.0015048144850879908, + 0.00141648028511554, + 0.0013256004313006997 + ], + "gamma_dfa": NaN, + "acc_eval": 0.515625, + "loss_eval": 1.345876693725586, + "epoch": 7 + }, + { + "hidden_norms": [ + 171.66441345214844, + 171.10499572753906, + 169.24270629882812, + 175.52674865722656, + 182.4739990234375 + ], + "bp_grad_norms_per_sample_med": [ + 3.675510379252955e-05, + 3.6888646718580276e-05, + 3.695035775308497e-05, + 3.644685784820467e-05, + 3.359343099873513e-05 + ], + "bp_grad_norms_F": [ + 0.0016642085975036025, + 0.0015915494877845049, + 0.0015412021894007921, + 0.0014428672147914767, + 0.0013226643204689026 + ], + "gamma_dfa": NaN, + "acc_eval": 0.548828125, + "loss_eval": 1.323807954788208, + "epoch": 8 + }, + { + "hidden_norms": [ + 177.58766174316406, + 178.5358123779297, + 176.316650390625, + 182.18365478515625, + 188.81033325195312 + ], + "bp_grad_norms_per_sample_med": [ + 3.550657493178733e-05, + 3.561190533218905e-05, + 3.632329753600061e-05, + 3.5655833926284686e-05, + 3.342977652209811e-05 + ], + "bp_grad_norms_F": [ + 0.0016174933407455683, + 0.0015753052430227399, + 0.0015375673538073897, + 0.001461501931771636, + 0.0013270385097712278 + ], + "gamma_dfa": NaN, + "acc_eval": 0.54296875, + "loss_eval": 1.323765754699707, + "epoch": 9 + }, + { + "hidden_norms": [ + 186.0078125, + 185.2810821533203, + 182.9329833984375, + 188.10736083984375, + 195.3680877685547 + ], + "bp_grad_norms_per_sample_med": [ + 3.5615968954516575e-05, + 3.5252433008281514e-05, + 3.600740819820203e-05, + 3.5232467780588195e-05, + 3.2765579817350954e-05 + ], + "bp_grad_norms_F": [ + 0.0015362550038844347, + 0.001510681351646781, + 0.0014847773127257824, + 0.0014146054163575172, + 0.00128385319840163 + ], + "gamma_dfa": NaN, + "acc_eval": 0.544921875, + "loss_eval": 1.2933119535446167, + "epoch": 10 + }, + { + "hidden_norms": [ + 193.46343994140625, + 194.1528778076172, + 190.9546661376953, + 196.03440856933594, + 201.5336151123047 + ], + "bp_grad_norms_per_sample_med": [ + 3.449141877354123e-05, + 3.581260534701869e-05, + 3.6061541322851554e-05, + 3.5576940717874095e-05, + 3.29486902046483e-05 + ], + "bp_grad_norms_F": [ + 0.0015161881456151605, + 0.0014996343525126576, + 0.001473701442591846, + 0.0013998147333040833, + 0.0012604170478880405 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5556640625, + "loss_eval": 1.2653318643569946, + "epoch": 11 + }, + { + "hidden_norms": [ + 200.17047119140625, + 198.5863037109375, + 196.08795166015625, + 200.61297607421875, + 208.13084411621094 + ], + "bp_grad_norms_per_sample_med": [ + 3.377814937266521e-05, + 3.471539093879983e-05, + 3.4731467167148367e-05, + 3.4925535146612674e-05, + 3.2187374017667025e-05 + ], + "bp_grad_norms_F": [ + 0.0014635181287303567, + 0.0014519579708576202, + 0.001435206620953977, + 0.0013784168986603618, + 0.0012511539971455932 + ], + "gamma_dfa": NaN, + "acc_eval": 0.546875, + "loss_eval": 1.2704923152923584, + "epoch": 12 + }, + { + "hidden_norms": [ + 203.8260955810547, + 203.86688232421875, + 200.01004028320312, + 204.2325439453125, + 211.52891540527344 + ], + "bp_grad_norms_per_sample_med": [ + 3.5213015507906675e-05, + 3.635779648902826e-05, + 3.70470697816927e-05, + 3.6254354199627414e-05, + 3.324731005704962e-05 + ], + "bp_grad_norms_F": [ + 0.0014951277989894152, + 0.0014758999459445477, + 0.0014606777112931013, + 0.0014066072180867195, + 0.0012636061292141676 + ], + "gamma_dfa": NaN, + "acc_eval": 0.533203125, + "loss_eval": 1.2718024253845215, + "epoch": 13 + }, + { + "hidden_norms": [ + 211.8421630859375, + 209.4982147216797, + 206.025634765625, + 210.11019897460938, + 216.48568725585938 + ], + "bp_grad_norms_per_sample_med": [ + 3.425328395678662e-05, + 3.4357526601525024e-05, + 3.5312823456479236e-05, + 3.490734161459841e-05, + 3.1354313250631094e-05 + ], + "bp_grad_norms_F": [ + 0.0015326166758313775, + 0.0015238082269206643, + 0.0015112065011635423, + 0.0014346316456794739, + 0.0012716582277789712 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5595703125, + "loss_eval": 1.2611415386199951, + "epoch": 14 + }, + { + "hidden_norms": [ + 220.0836639404297, + 218.3101806640625, + 212.22019958496094, + 216.97576904296875, + 224.16114807128906 + ], + "bp_grad_norms_per_sample_med": [ + 3.320865289424546e-05, + 3.393836959730834e-05, + 3.4554454032331705e-05, + 3.380324415047653e-05, + 3.0850649636704475e-05 + ], + "bp_grad_norms_F": [ + 0.0015296684578061104, + 0.0015227487310767174, + 0.0015054721152409911, + 0.0014223494799807668, + 0.0012559365713968873 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5615234375, + "loss_eval": 1.2511423826217651, + "epoch": 15 + }, + { + "hidden_norms": [ + 223.3075714111328, + 219.4958953857422, + 213.75701904296875, + 219.41580200195312, + 223.92572021484375 + ], + "bp_grad_norms_per_sample_med": [ + 3.406661926419474e-05, + 3.485638080746867e-05, + 3.5736342397285625e-05, + 3.5617544199340045e-05, + 3.142367859254591e-05 + ], + "bp_grad_norms_F": [ + 0.001539028249680996, + 0.0015256714541465044, + 0.0015207449905574322, + 0.001449243980459869, + 0.0012804584112018347 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5478515625, + "loss_eval": 1.2619048357009888, + "epoch": 16 + }, + { + "hidden_norms": [ + 229.43417358398438, + 224.83290100097656, + 218.015869140625, + 221.56915283203125, + 228.59274291992188 + ], + "bp_grad_norms_per_sample_med": [ + 3.325551369925961e-05, + 3.411810757825151e-05, + 3.506721623125486e-05, + 3.451627344475128e-05, + 3.0976541893323883e-05 + ], + "bp_grad_norms_F": [ + 0.001564670237712562, + 0.0015591747360303998, + 0.0015565806534141302, + 0.001464744214899838, + 0.0012746548745781183 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5498046875, + "loss_eval": 1.2594949007034302, + "epoch": 17 + }, + { + "hidden_norms": [ + 233.65724182128906, + 228.65892028808594, + 223.14169311523438, + 227.2765350341797, + 233.75588989257812 + ], + "bp_grad_norms_per_sample_med": [ + 3.4493146813474596e-05, + 3.510132592055015e-05, + 3.669387297122739e-05, + 3.577200186555274e-05, + 3.192189251421951e-05 + ], + "bp_grad_norms_F": [ + 0.0015125939389690757, + 0.0015123466728255153, + 0.0015171029372140765, + 0.0014489478198811412, + 0.001258584321476519 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5595703125, + "loss_eval": 1.2556639909744263, + "epoch": 18 + }, + { + "hidden_norms": [ + 237.43548583984375, + 232.39987182617188, + 227.10951232910156, + 230.72291564941406, + 237.13616943359375 + ], + "bp_grad_norms_per_sample_med": [ + 3.340727198519744e-05, + 3.450348594924435e-05, + 3.5213739465689287e-05, + 3.445050970185548e-05, + 3.137766179861501e-05 + ], + "bp_grad_norms_F": [ + 0.0015336342621594667, + 0.0015343877021223307, + 0.0015299322549253702, + 0.0014644783223047853, + 0.0012699750950559974 + ], + "gamma_dfa": NaN, + "acc_eval": 0.560546875, + "loss_eval": 1.2415752410888672, + "epoch": 19 + }, + { + "hidden_norms": [ + 240.3662109375, + 235.49240112304688, + 227.54061889648438, + 233.16600036621094, + 239.60838317871094 + ], + "bp_grad_norms_per_sample_med": [ + 3.387282777111977e-05, + 3.518611265462823e-05, + 3.594699956011027e-05, + 3.514933632686734e-05, + 3.159138941555284e-05 + ], + "bp_grad_norms_F": [ + 0.0015161214396357536, + 0.0015244479291141033, + 0.0015337818767875433, + 0.0014538948889821768, + 0.0012562318006530404 + ], + "gamma_dfa": NaN, + "acc_eval": 0.560546875, + "loss_eval": 1.2436577081680298, + "epoch": 20 + }, + { + "hidden_norms": [ + 246.6840057373047, + 239.0929412841797, + 228.7230682373047, + 233.1182403564453, + 237.92440795898438 + ], + "bp_grad_norms_per_sample_med": [ + 3.365568773006089e-05, + 3.4782005968736485e-05, + 3.604988887673244e-05, + 3.48928188032005e-05, + 3.140935586998239e-05 + ], + "bp_grad_norms_F": [ + 0.0015826384769752622, + 0.0015907500637695193, + 0.0016004826175048947, + 0.0015078946016728878, + 0.0012762312544509768 + ], + "gamma_dfa": NaN, + "acc_eval": 0.578125, + "loss_eval": 1.2292213439941406, + "epoch": 21 + }, + { + "hidden_norms": [ + 248.95265197753906, + 241.59408569335938, + 233.2073211669922, + 236.39898681640625, + 243.02566528320312 + ], + "bp_grad_norms_per_sample_med": [ + 3.3556283597135916e-05, + 3.463738539721817e-05, + 3.5712662793230265e-05, + 3.4992117434740067e-05, + 3.1405961635755375e-05 + ], + "bp_grad_norms_F": [ + 0.0015417077811434865, + 0.0015624086372554302, + 0.00158181122969836, + 0.0014982762513682246, + 0.001269534695893526 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5712890625, + "loss_eval": 1.201985239982605, + "epoch": 22 + }, + { + "hidden_norms": [ + 250.6766815185547, + 243.67706298828125, + 236.90567016601562, + 240.1038360595703, + 244.89207458496094 + ], + "bp_grad_norms_per_sample_med": [ + 3.3543499739607796e-05, + 3.4865890484070405e-05, + 3.564134021871723e-05, + 3.4787921322276816e-05, + 3.112335252808407e-05 + ], + "bp_grad_norms_F": [ + 0.0015237597981467843, + 0.0015382410492748022, + 0.0015483599854633212, + 0.0014793339651077986, + 0.0012637422187253833 + ], + "gamma_dfa": NaN, + "acc_eval": 0.55859375, + "loss_eval": 1.2245514392852783, + "epoch": 23 + }, + { + "hidden_norms": [ + 256.1240539550781, + 248.83609008789062, + 240.2549285888672, + 243.1914825439453, + 248.11964416503906 + ], + "bp_grad_norms_per_sample_med": [ + 3.322990596643649e-05, + 3.523347186273895e-05, + 3.6828067095484585e-05, + 3.5215769457863644e-05, + 3.133456266368739e-05 + ], + "bp_grad_norms_F": [ + 0.0015543467598035932, + 0.0015818241517990828, + 0.0015992531552910805, + 0.001527936663478613, + 0.0012819069670513272 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5732421875, + "loss_eval": 1.2261133193969727, + "epoch": 24 + }, + { + "hidden_norms": [ + 256.9990234375, + 248.78497314453125, + 239.55239868164062, + 243.13818359375, + 249.48153686523438 + ], + "bp_grad_norms_per_sample_med": [ + 3.39320904458873e-05, + 3.600086711230688e-05, + 3.7119287298992276e-05, + 3.6292254662839696e-05, + 3.193991506122984e-05 + ], + "bp_grad_norms_F": [ + 0.0015332578914240003, + 0.00156042305752635, + 0.001574728754349053, + 0.0015015782555565238, + 0.0012670031283050776 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5771484375, + "loss_eval": 1.2080682516098022, + "epoch": 25 + }, + { + "hidden_norms": [ + 260.4493103027344, + 251.83595275878906, + 240.95025634765625, + 242.58055114746094, + 247.42816162109375 + ], + "bp_grad_norms_per_sample_med": [ + 3.4877586585935205e-05, + 3.68267938029021e-05, + 3.760240360861644e-05, + 3.5770553949987516e-05, + 3.202233710908331e-05 + ], + "bp_grad_norms_F": [ + 0.0016203002305701375, + 0.0016500651836395264, + 0.0016584799159318209, + 0.0015617734752595425, + 0.00129653827752918 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5576171875, + "loss_eval": 1.2154628038406372, + "epoch": 26 + }, + { + "hidden_norms": [ + 263.1241760253906, + 253.4596405029297, + 243.6095733642578, + 245.4304656982422, + 250.32362365722656 + ], + "bp_grad_norms_per_sample_med": [ + 3.480416853562929e-05, + 3.640647264546715e-05, + 3.738802479347214e-05, + 3.5808730899589136e-05, + 3.25721557601355e-05 + ], + "bp_grad_norms_F": [ + 0.0015986696816980839, + 0.0016261462587863207, + 0.0016554947942495346, + 0.0015633245930075645, + 0.0013003128115087748 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5634765625, + "loss_eval": 1.2160842418670654, + "epoch": 27 + }, + { + "hidden_norms": [ + 266.5273132324219, + 256.0182189941406, + 245.50253295898438, + 246.18280029296875, + 251.716064453125 + ], + "bp_grad_norms_per_sample_med": [ + 3.359848051331937e-05, + 3.572113564587198e-05, + 3.695064515341073e-05, + 3.506165376165882e-05, + 3.114769424428232e-05 + ], + "bp_grad_norms_F": [ + 0.0016088238917291164, + 0.0016375494888052344, + 0.0016567507991567254, + 0.0015764845302328467, + 0.001285399659536779 + ], + "gamma_dfa": NaN, + "acc_eval": 0.58203125, + "loss_eval": 1.2021496295928955, + "epoch": 28 + }, + { + "hidden_norms": [ + 267.4769592285156, + 258.21368408203125, + 248.0418701171875, + 248.66949462890625, + 253.77452087402344 + ], + "bp_grad_norms_per_sample_med": [ + 3.400501736905426e-05, + 3.611011925386265e-05, + 3.7168374547036365e-05, + 3.569516775314696e-05, + 3.031879896298051e-05 + ], + "bp_grad_norms_F": [ + 0.0015880028950050473, + 0.0016197538934648037, + 0.001652923645451665, + 0.0015723761171102524, + 0.00128253607545048 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5771484375, + "loss_eval": 1.2029783725738525, + "epoch": 29 + }, + { + "hidden_norms": [ + 268.6047668457031, + 256.4200744628906, + 246.27796936035156, + 246.51229858398438, + 250.08482360839844 + ], + "bp_grad_norms_per_sample_med": [ + 3.425245813559741e-05, + 3.6405861465027556e-05, + 3.752295742742717e-05, + 3.6318160709924996e-05, + 3.103434937656857e-05 + ], + "bp_grad_norms_F": [ + 0.0017358324257656932, + 0.0017647893400862813, + 0.0017767423996701837, + 0.0016745994798839092, + 0.0013456137385219336 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5859375, + "loss_eval": 1.2118260860443115, + "epoch": 30 + }, + { + "hidden_norms": [ + 270.9348449707031, + 260.40045166015625, + 250.72512817382812, + 251.9379425048828, + 254.6614227294922 + ], + "bp_grad_norms_per_sample_med": [ + 3.491883762762882e-05, + 3.647417179308832e-05, + 3.8068057619966567e-05, + 3.653564635897055e-05, + 3.172537253703922e-05 + ], + "bp_grad_norms_F": [ + 0.0015994011191651225, + 0.0016339519061148167, + 0.0016624036943539977, + 0.0015806201845407486, + 0.0013034256407991052 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5849609375, + "loss_eval": 1.1830923557281494, + "epoch": 31 + }, + { + "hidden_norms": [ + 274.6016845703125, + 262.9326477050781, + 250.766357421875, + 252.1786346435547, + 254.2251739501953 + ], + "bp_grad_norms_per_sample_med": [ + 3.4794014936778694e-05, + 3.6746245314134285e-05, + 3.771794217755087e-05, + 3.6664638173533604e-05, + 3.1814517569728196e-05 + ], + "bp_grad_norms_F": [ + 0.0016584239201620221, + 0.0016917918110266328, + 0.001726189162582159, + 0.0016276866663247347, + 0.0013194811763241887 + ], + "gamma_dfa": NaN, + "acc_eval": 0.578125, + "loss_eval": 1.1851024627685547, + "epoch": 32 + }, + { + "hidden_norms": [ + 273.59783935546875, + 261.92218017578125, + 250.27320861816406, + 251.19146728515625, + 255.70233154296875 + ], + "bp_grad_norms_per_sample_med": [ + 3.505383210722357e-05, + 3.6710382119053975e-05, + 3.839766577584669e-05, + 3.708387885126285e-05, + 3.1251351174432784e-05 + ], + "bp_grad_norms_F": [ + 0.0016781548038125038, + 0.0017169808270409703, + 0.0017399545758962631, + 0.0016386040952056646, + 0.0013246783055365086 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5712890625, + "loss_eval": 1.1789460182189941, + "epoch": 33 + }, + { + "hidden_norms": [ + 274.9112854003906, + 262.3539123535156, + 251.67474365234375, + 253.08824157714844, + 255.4647979736328 + ], + "bp_grad_norms_per_sample_med": [ + 3.4545508242445067e-05, + 3.622366421041079e-05, + 3.7300182157196105e-05, + 3.594972440623678e-05, + 3.1557658076053485e-05 + ], + "bp_grad_norms_F": [ + 0.001666227588430047, + 0.0016955926548689604, + 0.0017251977697014809, + 0.0016294753877446055, + 0.001327820005826652 + ], + "gamma_dfa": NaN, + "acc_eval": 0.583984375, + "loss_eval": 1.1757192611694336, + "epoch": 34 + }, + { + "hidden_norms": [ + 273.9092712402344, + 261.26507568359375, + 250.60923767089844, + 250.5943145751953, + 253.19862365722656 + ], + "bp_grad_norms_per_sample_med": [ + 3.532320988597348e-05, + 3.7557252653641626e-05, + 3.882365126628429e-05, + 3.676790583995171e-05, + 3.2376574381487444e-05 + ], + "bp_grad_norms_F": [ + 0.001686039031483233, + 0.0017200085567310452, + 0.0017563666915521026, + 0.0016565105179324746, + 0.0013397492002695799 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5986328125, + "loss_eval": 1.1719229221343994, + "epoch": 35 + }, + { + "hidden_norms": [ + 275.9344177246094, + 265.1002502441406, + 253.2598114013672, + 253.86024475097656, + 257.8154296875 + ], + "bp_grad_norms_per_sample_med": [ + 3.419027416384779e-05, + 3.54782423528377e-05, + 3.770734474528581e-05, + 3.707608630065806e-05, + 3.1706651498097926e-05 + ], + "bp_grad_norms_F": [ + 0.0016564616234973073, + 0.001694328966550529, + 0.001729298266582191, + 0.0016409243689849973, + 0.0013242866843938828 + ], + "gamma_dfa": NaN, + "acc_eval": 0.59375, + "loss_eval": 1.169456958770752, + "epoch": 36 + }, + { + "hidden_norms": [ + 277.0843200683594, + 263.2900390625, + 250.48553466796875, + 251.51341247558594, + 253.14413452148438 + ], + "bp_grad_norms_per_sample_med": [ + 3.5850312997354195e-05, + 3.7839381548110396e-05, + 3.925598502974026e-05, + 3.7184508983045816e-05, + 3.174091762048192e-05 + ], + "bp_grad_norms_F": [ + 0.0017717990558594465, + 0.0018099851440638304, + 0.0018417692044749856, + 0.0017209915677085519, + 0.0013732420047745109 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5869140625, + "loss_eval": 1.1764099597930908, + "epoch": 37 + }, + { + "hidden_norms": [ + 280.2084655761719, + 265.3043518066406, + 254.26673889160156, + 253.3932342529297, + 255.91488647460938 + ], + "bp_grad_norms_per_sample_med": [ + 3.4716471418505535e-05, + 3.675218249554746e-05, + 3.84828781534452e-05, + 3.653839303296991e-05, + 3.1539821065962315e-05 + ], + "bp_grad_norms_F": [ + 0.00171388138551265, + 0.0017648075008764863, + 0.00180675252340734, + 0.0016973463352769613, + 0.0013539392966777086 + ], + "gamma_dfa": NaN, + "acc_eval": 0.591796875, + "loss_eval": 1.1670037508010864, + "epoch": 38 + }, + { + "hidden_norms": [ + 278.0702209472656, + 264.4510498046875, + 250.75350952148438, + 248.73284912109375, + 253.1072540283203 + ], + "bp_grad_norms_per_sample_med": [ + 3.694290717248805e-05, + 3.843691956717521e-05, + 3.982238922617398e-05, + 3.8299614971037954e-05, + 3.253731847507879e-05 + ], + "bp_grad_norms_F": [ + 0.0017997974064201117, + 0.0018521619495004416, + 0.0018853276269510388, + 0.001754313474521041, + 0.0013789198128506541 + ], + "gamma_dfa": NaN, + "acc_eval": 0.576171875, + "loss_eval": 1.168372631072998, + "epoch": 39 + }, + { + "hidden_norms": [ + 279.5745849609375, + 265.3549499511719, + 252.89109802246094, + 250.0029296875, + 251.57737731933594 + ], + "bp_grad_norms_per_sample_med": [ + 3.5274497349746525e-05, + 3.741122054634616e-05, + 3.8946731365285814e-05, + 3.7627731217071414e-05, + 3.1920495530357584e-05 + ], + "bp_grad_norms_F": [ + 0.0017959685064852238, + 0.0018515808042138815, + 0.001886399113573134, + 0.0017732164124026895, + 0.001389715587720275 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5966796875, + "loss_eval": 1.1735193729400635, + "epoch": 40 + }, + { + "hidden_norms": [ + 280.56402587890625, + 265.1068115234375, + 251.912353515625, + 248.08627319335938, + 250.21177673339844 + ], + "bp_grad_norms_per_sample_med": [ + 3.672741513582878e-05, + 3.8823065551696345e-05, + 4.041267675347626e-05, + 3.835778989014216e-05, + 3.2424144592368975e-05 + ], + "bp_grad_norms_F": [ + 0.0018404219299554825, + 0.0018969813827425241, + 0.0019448500825092196, + 0.0018052044324576855, + 0.0014045372372493148 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5966796875, + "loss_eval": 1.163725733757019, + "epoch": 41 + }, + { + "hidden_norms": [ + 281.2812194824219, + 266.5726623535156, + 253.18124389648438, + 252.3292694091797, + 254.1314697265625 + ], + "bp_grad_norms_per_sample_med": [ + 3.6245146475266665e-05, + 3.849809218081646e-05, + 4.034954326925799e-05, + 3.857027695630677e-05, + 3.267046849941835e-05 + ], + "bp_grad_norms_F": [ + 0.0018043599557131529, + 0.0018654613522812724, + 0.0018976799910888076, + 0.0017763315699994564, + 0.0013986497651785612 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5830078125, + "loss_eval": 1.172363519668579, + "epoch": 42 + }, + { + "hidden_norms": [ + 278.3638916015625, + 264.4734802246094, + 250.60198974609375, + 248.59568786621094, + 249.39353942871094 + ], + "bp_grad_norms_per_sample_med": [ + 3.7236037314869463e-05, + 3.946686410927214e-05, + 4.103714309167117e-05, + 3.970286707044579e-05, + 3.281752287875861e-05 + ], + "bp_grad_norms_F": [ + 0.0018646781099960208, + 0.0019272645004093647, + 0.0019672266207635403, + 0.001839098404161632, + 0.0014395405305549502 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5830078125, + "loss_eval": 1.1776624917984009, + "epoch": 43 + }, + { + "hidden_norms": [ + 280.8514099121094, + 269.1718444824219, + 253.03790283203125, + 251.60220336914062, + 251.88168334960938 + ], + "bp_grad_norms_per_sample_med": [ + 3.689880759338848e-05, + 3.937961082556285e-05, + 4.092283779755235e-05, + 3.866894985549152e-05, + 3.262344398535788e-05 + ], + "bp_grad_norms_F": [ + 0.0018725661793723702, + 0.0019201217219233513, + 0.0019532060250639915, + 0.0018317755311727524, + 0.001439971849322319 + ], + "gamma_dfa": NaN, + "acc_eval": 0.595703125, + "loss_eval": 1.1693906784057617, + "epoch": 44 + }, + { + "hidden_norms": [ + 280.3335266113281, + 263.89471435546875, + 253.01638793945312, + 250.5576171875, + 251.4005889892578 + ], + "bp_grad_norms_per_sample_med": [ + 3.6666475352831185e-05, + 3.875952461385168e-05, + 4.0531358536100015e-05, + 3.9068203477654606e-05, + 3.450641088420525e-05 + ], + "bp_grad_norms_F": [ + 0.0018578553572297096, + 0.0019184533739462495, + 0.001965942559763789, + 0.0018550093518570065, + 0.0014542803401127458 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5732421875, + "loss_eval": 1.1839377880096436, + "epoch": 45 + }, + { + "hidden_norms": [ + 280.3553771972656, + 263.3617248535156, + 248.94859313964844, + 246.9832000732422, + 246.24644470214844 + ], + "bp_grad_norms_per_sample_med": [ + 3.8412734284065664e-05, + 4.001844354206696e-05, + 4.220755727146752e-05, + 4.0393635572399944e-05, + 3.391467180335894e-05 + ], + "bp_grad_norms_F": [ + 0.001917686895467341, + 0.0019766101613640785, + 0.0020170181524008512, + 0.0018900329014286399, + 0.001476101577281952 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6005859375, + "loss_eval": 1.1689397096633911, + "epoch": 46 + }, + { + "hidden_norms": [ + 278.8861389160156, + 263.3866882324219, + 249.24301147460938, + 246.43963623046875, + 246.34503173828125 + ], + "bp_grad_norms_per_sample_med": [ + 3.77853139070794e-05, + 4.044857996632345e-05, + 4.182702468824573e-05, + 4.050548523082398e-05, + 3.3606509532546625e-05 + ], + "bp_grad_norms_F": [ + 0.0019284948939457536, + 0.001992136472836137, + 0.0020461773965507746, + 0.001900508883409202, + 0.001466717105358839 + ], + "gamma_dfa": NaN, + "acc_eval": 0.599609375, + "loss_eval": 1.1585946083068848, + "epoch": 47 + }, + { + "hidden_norms": [ + 279.963623046875, + 263.52496337890625, + 249.6890869140625, + 245.60646057128906, + 245.52540588378906 + ], + "bp_grad_norms_per_sample_med": [ + 3.724947964656167e-05, + 3.964625648222864e-05, + 4.1545175918145105e-05, + 3.953047780669294e-05, + 3.275729977758601e-05 + ], + "bp_grad_norms_F": [ + 0.0019616533536463976, + 0.002027863636612892, + 0.0020759752951562405, + 0.0019283785950392485, + 0.001484199776314199 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6123046875, + "loss_eval": 1.157920002937317, + "epoch": 48 + }, + { + "hidden_norms": [ + 279.7983093261719, + 263.42352294921875, + 247.6257781982422, + 244.5139923095703, + 242.93429565429688 + ], + "bp_grad_norms_per_sample_med": [ + 3.738125451491214e-05, + 4.007351162726991e-05, + 4.2600284359650686e-05, + 4.0704209823161364e-05, + 3.302833283669315e-05 + ], + "bp_grad_norms_F": [ + 0.001973477192223072, + 0.002032764023169875, + 0.0020790304988622665, + 0.0019371822709217668, + 0.0014914683997631073 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6005859375, + "loss_eval": 1.1545627117156982, + "epoch": 49 + }, + { + "hidden_norms": [ + 278.16461181640625, + 259.95208740234375, + 246.212890625, + 243.02810668945312, + 242.42227172851562 + ], + "bp_grad_norms_per_sample_med": [ + 3.896220732713118e-05, + 4.0713552152737975e-05, + 4.242185241309926e-05, + 4.1096616769209504e-05, + 3.4584638342494145e-05 + ], + "bp_grad_norms_F": [ + 0.0020273446571081877, + 0.0020888603758066893, + 0.0021337512880563736, + 0.0019914479926228523, + 0.0015125740319490433 + ], + "gamma_dfa": NaN, + "acc_eval": 0.599609375, + "loss_eval": 1.1588966846466064, + "epoch": 50 + }, + { + "hidden_norms": [ + 278.74609375, + 259.810302734375, + 244.75640869140625, + 241.40191650390625, + 241.2429962158203 + ], + "bp_grad_norms_per_sample_med": [ + 4.005424489150755e-05, + 4.289989738026634e-05, + 4.383309351396747e-05, + 4.207885649520904e-05, + 3.527112494339235e-05 + ], + "bp_grad_norms_F": [ + 0.002059010788798332, + 0.002119551645591855, + 0.002172222128137946, + 0.0020189452916383743, + 0.0015335450880229473 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6103515625, + "loss_eval": 1.1555505990982056, + "epoch": 51 + }, + { + "hidden_norms": [ + 277.6565246582031, + 262.0436096191406, + 246.2357635498047, + 241.67991638183594, + 240.23500061035156 + ], + "bp_grad_norms_per_sample_med": [ + 4.1031005821423605e-05, + 4.313238605391234e-05, + 4.465782694751397e-05, + 4.2896612285403535e-05, + 3.4334370866417885e-05 + ], + "bp_grad_norms_F": [ + 0.002025892725214362, + 0.0020990471821278334, + 0.0021509609650820494, + 0.002011507283896208, + 0.0015318752266466618 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6142578125, + "loss_eval": 1.1416809558868408, + "epoch": 52 + }, + { + "hidden_norms": [ + 277.4100646972656, + 260.2519836425781, + 245.48947143554688, + 242.5662384033203, + 241.2648162841797 + ], + "bp_grad_norms_per_sample_med": [ + 3.732202821993269e-05, + 3.9125570765463635e-05, + 4.136607458349317e-05, + 3.977569213020615e-05, + 3.3676675229799e-05 + ], + "bp_grad_norms_F": [ + 0.002009452786296606, + 0.002076543401926756, + 0.002131648361682892, + 0.00199576816521585, + 0.0015338497469201684 + ], + "gamma_dfa": NaN, + "acc_eval": 0.615234375, + "loss_eval": 1.143639087677002, + "epoch": 53 + }, + { + "hidden_norms": [ + 277.3269348144531, + 259.49072265625, + 243.7945098876953, + 240.342041015625, + 238.76239013671875 + ], + "bp_grad_norms_per_sample_med": [ + 3.978172389906831e-05, + 4.20100441260729e-05, + 4.455841553863138e-05, + 4.2333795136073604e-05, + 3.435015969444066e-05 + ], + "bp_grad_norms_F": [ + 0.0020786102395504713, + 0.0021364488638937473, + 0.0021890606731176376, + 0.0020492339972406626, + 0.0015625122468918562 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6044921875, + "loss_eval": 1.1515132188796997, + "epoch": 54 + }, + { + "hidden_norms": [ + 276.1414489746094, + 257.6651306152344, + 242.12867736816406, + 238.3046875, + 235.9711151123047 + ], + "bp_grad_norms_per_sample_med": [ + 4.158447700319812e-05, + 4.40051153418608e-05, + 4.566523784887977e-05, + 4.3018761061830446e-05, + 3.456107515376061e-05 + ], + "bp_grad_norms_F": [ + 0.0021895321551710367, + 0.0022726275492459536, + 0.00232017133384943, + 0.002137792995199561, + 0.0015979791060090065 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6025390625, + "loss_eval": 1.148828148841858, + "epoch": 55 + }, + { + "hidden_norms": [ + 276.29547119140625, + 257.9006652832031, + 242.25363159179688, + 237.8865966796875, + 235.57620239257812 + ], + "bp_grad_norms_per_sample_med": [ + 3.9041460695443675e-05, + 4.0436327253701165e-05, + 4.2260016925865784e-05, + 4.071593502885662e-05, + 3.461613232502714e-05 + ], + "bp_grad_norms_F": [ + 0.0021726686973124743, + 0.002256714040413499, + 0.0023172625806182623, + 0.0021478317212313414, + 0.0016112272860482335 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6015625, + "loss_eval": 1.165194034576416, + "epoch": 56 + }, + { + "hidden_norms": [ + 274.73480224609375, + 257.3844909667969, + 239.9276580810547, + 236.4628448486328, + 232.49490356445312 + ], + "bp_grad_norms_per_sample_med": [ + 4.182914926786907e-05, + 4.482160511543043e-05, + 4.732825618702918e-05, + 4.5475029764929786e-05, + 3.673886021715589e-05 + ], + "bp_grad_norms_F": [ + 0.002222857205197215, + 0.0022976247128099203, + 0.002366685541346669, + 0.002194431144744158, + 0.0016426561633124948 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6015625, + "loss_eval": 1.150635004043579, + "epoch": 57 + }, + { + "hidden_norms": [ + 273.9837341308594, + 255.2301025390625, + 239.6470184326172, + 235.07806396484375, + 232.15016174316406 + ], + "bp_grad_norms_per_sample_med": [ + 4.28417552029714e-05, + 4.5315191528061405e-05, + 4.770928717334755e-05, + 4.653819996747188e-05, + 3.727963485289365e-05 + ], + "bp_grad_norms_F": [ + 0.002205377910286188, + 0.00228099524974823, + 0.002334901597350836, + 0.0021908036433160305, + 0.001635397202335298 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6064453125, + "loss_eval": 1.150489330291748, + "epoch": 58 + }, + { + "hidden_norms": [ + 273.4665222167969, + 255.21783447265625, + 238.12306213378906, + 232.99537658691406, + 229.35629272460938 + ], + "bp_grad_norms_per_sample_med": [ + 4.174295099801384e-05, + 4.412833004607819e-05, + 4.589854142977856e-05, + 4.366271969047375e-05, + 3.4910473914351314e-05 + ], + "bp_grad_norms_F": [ + 0.0022425123024731874, + 0.002337042009457946, + 0.0024111224338412285, + 0.002237174427136779, + 0.0016597777139395475 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6142578125, + "loss_eval": 1.1445674896240234, + "epoch": 59 + }, + { + "hidden_norms": [ + 272.8526306152344, + 254.31373596191406, + 238.0008087158203, + 233.6641082763672, + 229.88113403320312 + ], + "bp_grad_norms_per_sample_med": [ + 4.274978709872812e-05, + 4.589447416947223e-05, + 4.8100573621923104e-05, + 4.519677167991176e-05, + 3.628300692071207e-05 + ], + "bp_grad_norms_F": [ + 0.0022435274440795183, + 0.0023362115025520325, + 0.0024012199137359858, + 0.00223422609269619, + 0.0016617706278339028 + ], + "gamma_dfa": NaN, + "acc_eval": 0.609375, + "loss_eval": 1.1417796611785889, + "epoch": 60 + }, + { + "hidden_norms": [ + 272.5745849609375, + 253.39877319335938, + 236.8953094482422, + 231.5461883544922, + 227.8491668701172 + ], + "bp_grad_norms_per_sample_med": [ + 4.257826367393136e-05, + 4.485245153773576e-05, + 4.667421308113262e-05, + 4.5228414819575846e-05, + 3.560770346666686e-05 + ], + "bp_grad_norms_F": [ + 0.0022986247204244137, + 0.002396916039288044, + 0.0024696297477930784, + 0.0022973858285695314, + 0.0016907128738239408 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6044921875, + "loss_eval": 1.1428865194320679, + "epoch": 61 + }, + { + "hidden_norms": [ + 269.6819763183594, + 250.9730682373047, + 234.21188354492188, + 228.83775329589844, + 224.82958984375 + ], + "bp_grad_norms_per_sample_med": [ + 4.29281426477246e-05, + 4.5681605115532875e-05, + 4.7611269110348076e-05, + 4.553339022095315e-05, + 3.683664544951171e-05 + ], + "bp_grad_norms_F": [ + 0.0023400019854307175, + 0.002444769022986293, + 0.0025141045916825533, + 0.002344615990296006, + 0.0017126062884926796 + ], + "gamma_dfa": NaN, + "acc_eval": 0.625, + "loss_eval": 1.1239444017410278, + "epoch": 62 + }, + { + "hidden_norms": [ + 269.7929992675781, + 250.5220947265625, + 233.3997802734375, + 228.2130584716797, + 224.5063018798828 + ], + "bp_grad_norms_per_sample_med": [ + 4.4715372496284544e-05, + 4.6482971811201423e-05, + 4.9765483709052205e-05, + 4.759176226798445e-05, + 3.8208585465326905e-05 + ], + "bp_grad_norms_F": [ + 0.0023639060091227293, + 0.002476333873346448, + 0.0025423571933060884, + 0.00236364989541471, + 0.0017294714925810695 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6162109375, + "loss_eval": 1.1502506732940674, + "epoch": 63 + }, + { + "hidden_norms": [ + 269.24493408203125, + 251.6195831298828, + 234.21717834472656, + 228.9242401123047, + 224.33628845214844 + ], + "bp_grad_norms_per_sample_med": [ + 4.4276617700234056e-05, + 4.620026084012352e-05, + 4.972432361682877e-05, + 4.657481986214407e-05, + 3.729217132786289e-05 + ], + "bp_grad_norms_F": [ + 0.0023896305356174707, + 0.002498477231711149, + 0.0025652945041656494, + 0.002390124835073948, + 0.0017562232678756118 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6171875, + "loss_eval": 1.146531581878662, + "epoch": 64 + }, + { + "hidden_norms": [ + 268.0794982910156, + 250.7041778564453, + 233.16966247558594, + 227.9508056640625, + 223.48045349121094 + ], + "bp_grad_norms_per_sample_med": [ + 4.518003697739914e-05, + 4.807085497304797e-05, + 5.077984678791836e-05, + 4.9026388296624646e-05, + 3.8487425626954064e-05 + ], + "bp_grad_norms_F": [ + 0.0023859951179474592, + 0.002494273241609335, + 0.0025624327827244997, + 0.0023940331302583218, + 0.0017616016557440162 + ], + "gamma_dfa": NaN, + "acc_eval": 0.611328125, + "loss_eval": 1.1454447507858276, + "epoch": 65 + }, + { + "hidden_norms": [ + 267.8343811035156, + 247.6757049560547, + 231.40367126464844, + 225.93670654296875, + 220.4804229736328 + ], + "bp_grad_norms_per_sample_med": [ + 4.405072468216531e-05, + 4.688911576522514e-05, + 4.961229205946438e-05, + 4.743603858514689e-05, + 3.724359339685179e-05 + ], + "bp_grad_norms_F": [ + 0.002465154742822051, + 0.0025785581674426794, + 0.002648484194651246, + 0.0024633395951241255, + 0.0017954027280211449 + ], + "gamma_dfa": NaN, + "acc_eval": 0.615234375, + "loss_eval": 1.165484070777893, + "epoch": 66 + }, + { + "hidden_norms": [ + 267.8901062011719, + 248.67857360839844, + 231.63746643066406, + 225.49383544921875, + 220.6399383544922 + ], + "bp_grad_norms_per_sample_med": [ + 4.755932604894042e-05, + 4.9960097385337576e-05, + 5.191187665332109e-05, + 4.949339563609101e-05, + 3.946756987716071e-05 + ], + "bp_grad_norms_F": [ + 0.0024303828831762075, + 0.0025356977712363005, + 0.002610789379104972, + 0.00243638688698411, + 0.0017875435296446085 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6103515625, + "loss_eval": 1.1472151279449463, + "epoch": 67 + }, + { + "hidden_norms": [ + 267.3486022949219, + 247.69305419921875, + 229.9873504638672, + 223.6610870361328, + 218.91578674316406 + ], + "bp_grad_norms_per_sample_med": [ + 4.743155659525655e-05, + 5.06583419337403e-05, + 5.422766844276339e-05, + 5.0596820074133575e-05, + 3.959470632253215e-05 + ], + "bp_grad_norms_F": [ + 0.0025001303292810917, + 0.0026160534471273422, + 0.002682509133592248, + 0.002501503797248006, + 0.0018251334549859166 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6201171875, + "loss_eval": 1.157745599746704, + "epoch": 68 + }, + { + "hidden_norms": [ + 266.2055969238281, + 246.88201904296875, + 229.0934295654297, + 223.09255981445312, + 217.90614318847656 + ], + "bp_grad_norms_per_sample_med": [ + 4.6090037358226255e-05, + 4.8821068048710003e-05, + 5.145218892721459e-05, + 4.892798824585043e-05, + 3.874724279739894e-05 + ], + "bp_grad_norms_F": [ + 0.002517222659662366, + 0.00262960116378963, + 0.002704967511817813, + 0.002524228999391198, + 0.0018283555982634425 + ], + "gamma_dfa": NaN, + "acc_eval": 0.615234375, + "loss_eval": 1.1513168811798096, + "epoch": 69 + }, + { + "hidden_norms": [ + 263.8387451171875, + 244.32431030273438, + 227.78903198242188, + 222.13900756835938, + 216.27389526367188 + ], + "bp_grad_norms_per_sample_med": [ + 4.680602069129236e-05, + 5.040669202571735e-05, + 5.318928378983401e-05, + 4.9446778575656936e-05, + 3.913935870514251e-05 + ], + "bp_grad_norms_F": [ + 0.0025120435748249292, + 0.0026304670609533787, + 0.0027123456820845604, + 0.002537800231948495, + 0.001847997889854014 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6240234375, + "loss_eval": 1.1497204303741455, + "epoch": 70 + }, + { + "hidden_norms": [ + 263.52716064453125, + 243.21226501464844, + 226.427001953125, + 220.67845153808594, + 214.70176696777344 + ], + "bp_grad_norms_per_sample_med": [ + 4.6976576413726434e-05, + 5.0085352995665744e-05, + 5.2343326387926936e-05, + 4.9784572183853015e-05, + 4.0186921978602186e-05 + ], + "bp_grad_norms_F": [ + 0.0026124012656509876, + 0.0027241536881774664, + 0.002797869499772787, + 0.002603790955618024, + 0.0018792233895510435 + ], + "gamma_dfa": NaN, + "acc_eval": 0.609375, + "loss_eval": 1.153285026550293, + "epoch": 71 + }, + { + "hidden_norms": [ + 264.0121765136719, + 243.2773895263672, + 226.7365264892578, + 220.29107666015625, + 214.1591339111328 + ], + "bp_grad_norms_per_sample_med": [ + 4.6254041080828756e-05, + 4.949727008352056e-05, + 5.206605055718683e-05, + 5.012763722334057e-05, + 3.8815273001091555e-05 + ], + "bp_grad_norms_F": [ + 0.0026143237482756376, + 0.0027275518514215946, + 0.0028062777128070593, + 0.002607470378279686, + 0.0018867084290832281 + ], + "gamma_dfa": NaN, + "acc_eval": 0.615234375, + "loss_eval": 1.1514866352081299, + "epoch": 72 + }, + { + "hidden_norms": [ + 263.4013671875, + 243.0095672607422, + 225.4296112060547, + 219.55067443847656, + 213.40794372558594 + ], + "bp_grad_norms_per_sample_med": [ + 4.8506815801374614e-05, + 5.151949153514579e-05, + 5.5085791245801374e-05, + 5.175785190658644e-05, + 4.0316641388926655e-05 + ], + "bp_grad_norms_F": [ + 0.002606831956654787, + 0.002728143008425832, + 0.0028109808918088675, + 0.0026184367015957832, + 0.0018970600795000792 + ], + "gamma_dfa": NaN, + "acc_eval": 0.615234375, + "loss_eval": 1.1346663236618042, + "epoch": 73 + }, + { + "hidden_norms": [ + 262.27703857421875, + 242.7241668701172, + 224.5229949951172, + 218.4025115966797, + 212.19247436523438 + ], + "bp_grad_norms_per_sample_med": [ + 4.704251477960497e-05, + 5.0965834816452116e-05, + 5.3327348723541945e-05, + 5.0449500122340396e-05, + 4.0479797462467104e-05 + ], + "bp_grad_norms_F": [ + 0.0026277368888258934, + 0.0027542279567569494, + 0.00283243996091187, + 0.00264621595852077, + 0.0019058829639106989 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6240234375, + "loss_eval": 1.137591004371643, + "epoch": 74 + }, + { + "hidden_norms": [ + 262.1747131347656, + 241.88607788085938, + 224.1046142578125, + 217.9982147216797, + 211.7519989013672 + ], + "bp_grad_norms_per_sample_med": [ + 4.847371383220889e-05, + 5.121564754517749e-05, + 5.3943567763781175e-05, + 5.087414683657698e-05, + 4.037184771732427e-05 + ], + "bp_grad_norms_F": [ + 0.0026811074931174517, + 0.002809705911204219, + 0.002887698821723461, + 0.002692745067179203, + 0.0019356919219717383 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6142578125, + "loss_eval": 1.1538643836975098, + "epoch": 75 + }, + { + "hidden_norms": [ + 261.7838439941406, + 239.75706481933594, + 222.1133575439453, + 216.12124633789062, + 210.44593811035156 + ], + "bp_grad_norms_per_sample_med": [ + 4.8369467549491674e-05, + 5.07068180013448e-05, + 5.281560879666358e-05, + 5.086012970423326e-05, + 4.062180232722312e-05 + ], + "bp_grad_norms_F": [ + 0.0027064597234129906, + 0.0028339733835309744, + 0.0029109998140484095, + 0.002717787167057395, + 0.0019469966646283865 + ], + "gamma_dfa": NaN, + "acc_eval": 0.615234375, + "loss_eval": 1.1476709842681885, + "epoch": 76 + }, + { + "hidden_norms": [ + 260.708251953125, + 239.65792846679688, + 222.502197265625, + 216.31419372558594, + 210.15716552734375 + ], + "bp_grad_norms_per_sample_med": [ + 4.919906132272445e-05, + 5.164166941540316e-05, + 5.366719051380642e-05, + 5.108323966851458e-05, + 4.000464832643047e-05 + ], + "bp_grad_norms_F": [ + 0.002715736860409379, + 0.0028464714996516705, + 0.0029330456163734198, + 0.0027443754952400923, + 0.001961463363841176 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6181640625, + "loss_eval": 1.1511735916137695, + "epoch": 77 + }, + { + "hidden_norms": [ + 260.4088439941406, + 239.9808349609375, + 222.2389373779297, + 215.33692932128906, + 208.58534240722656 + ], + "bp_grad_norms_per_sample_med": [ + 4.8405916459159926e-05, + 5.163977766642347e-05, + 5.34093014721293e-05, + 5.167788185644895e-05, + 4.020798587589525e-05 + ], + "bp_grad_norms_F": [ + 0.0027501434087753296, + 0.0028795108664780855, + 0.0029617082327604294, + 0.002769613405689597, + 0.001984042814001441 + ], + "gamma_dfa": NaN, + "acc_eval": 0.615234375, + "loss_eval": 1.1523363590240479, + "epoch": 78 + }, + { + "hidden_norms": [ + 259.8567199707031, + 238.8525848388672, + 221.2687225341797, + 213.9132537841797, + 207.3290557861328 + ], + "bp_grad_norms_per_sample_med": [ + 4.8219208110822365e-05, + 5.1672555855475366e-05, + 5.364803655538708e-05, + 5.1243885536678135e-05, + 4.079756035935134e-05 + ], + "bp_grad_norms_F": [ + 0.0028046013321727514, + 0.002939376747235656, + 0.0030214993748813868, + 0.002813557395711541, + 0.0019953204318881035 + ], + "gamma_dfa": NaN, + "acc_eval": 0.611328125, + "loss_eval": 1.1514203548431396, + "epoch": 79 + }, + { + "hidden_norms": [ + 259.52423095703125, + 238.27757263183594, + 219.9165802001953, + 213.4214324951172, + 206.20606994628906 + ], + "bp_grad_norms_per_sample_med": [ + 4.8479021643288434e-05, + 5.219021113589406e-05, + 5.5094871640903875e-05, + 5.1899030950153247e-05, + 4.124108454561792e-05 + ], + "bp_grad_norms_F": [ + 0.0028202880639582872, + 0.002952422248199582, + 0.0030410154722630978, + 0.002832787809893489, + 0.0020088190212845802 + ], + "gamma_dfa": NaN, + "acc_eval": 0.625, + "loss_eval": 1.1545820236206055, + "epoch": 80 + }, + { + "hidden_norms": [ + 258.1958923339844, + 237.08778381347656, + 219.2870635986328, + 212.87725830078125, + 205.84408569335938 + ], + "bp_grad_norms_per_sample_med": [ + 5.005837374483235e-05, + 5.344217424863018e-05, + 5.582463927567005e-05, + 5.345495083020069e-05, + 4.113194518140517e-05 + ], + "bp_grad_norms_F": [ + 0.0028330644126981497, + 0.002965509658679366, + 0.0030466399621218443, + 0.00284536718390882, + 0.0020232615061104298 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6220703125, + "loss_eval": 1.1552448272705078, + "epoch": 81 + }, + { + "hidden_norms": [ + 258.44268798828125, + 236.90562438964844, + 218.65948486328125, + 212.63186645507812, + 204.90692138671875 + ], + "bp_grad_norms_per_sample_med": [ + 4.896440077573061e-05, + 5.1820486987708136e-05, + 5.4638629080727696e-05, + 5.3578904044115916e-05, + 4.117973367101513e-05 + ], + "bp_grad_norms_F": [ + 0.002839416265487671, + 0.0029778846073895693, + 0.003062452422454953, + 0.002849552081897855, + 0.0020236214622855186 + ], + "gamma_dfa": NaN, + "acc_eval": 0.62109375, + "loss_eval": 1.1515182256698608, + "epoch": 82 + }, + { + "hidden_norms": [ + 258.24615478515625, + 236.87181091308594, + 219.06484985351562, + 213.2528076171875, + 205.03671264648438 + ], + "bp_grad_norms_per_sample_med": [ + 4.8622885515214875e-05, + 5.181535379961133e-05, + 5.509376205736771e-05, + 5.2298331866040826e-05, + 4.074195385328494e-05 + ], + "bp_grad_norms_F": [ + 0.002838765038177371, + 0.00297414674423635, + 0.003058358561247587, + 0.0028491863049566746, + 0.0020313323475420475 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6142578125, + "loss_eval": 1.1609394550323486, + "epoch": 83 + }, + { + "hidden_norms": [ + 257.9989929199219, + 236.4859161376953, + 218.73072814941406, + 212.0786895751953, + 204.2919464111328 + ], + "bp_grad_norms_per_sample_med": [ + 5.0208276661578566e-05, + 5.2276191127020866e-05, + 5.509322727448307e-05, + 5.315374437486753e-05, + 4.122377504245378e-05 + ], + "bp_grad_norms_F": [ + 0.0028634185437113047, + 0.0029970910400152206, + 0.0030890253838151693, + 0.002881886437535286, + 0.0020408176351338625 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6171875, + "loss_eval": 1.1592046022415161, + "epoch": 84 + }, + { + "hidden_norms": [ + 257.24053955078125, + 235.7548065185547, + 217.84329223632812, + 211.033935546875, + 203.8216094970703 + ], + "bp_grad_norms_per_sample_med": [ + 5.0660164561122656e-05, + 5.414208135334775e-05, + 5.5454143875977024e-05, + 5.3000938351033255e-05, + 4.292939411243424e-05 + ], + "bp_grad_norms_F": [ + 0.0028948886319994926, + 0.003031315514817834, + 0.0031204961705952883, + 0.002907233312726021, + 0.0020496752113103867 + ], + "gamma_dfa": NaN, + "acc_eval": 0.619140625, + "loss_eval": 1.1583250761032104, + "epoch": 85 + }, + { + "hidden_norms": [ + 256.78289794921875, + 235.2476348876953, + 217.21603393554688, + 210.80035400390625, + 203.15478515625 + ], + "bp_grad_norms_per_sample_med": [ + 4.9510348617332056e-05, + 5.1982155127916485e-05, + 5.440499808173627e-05, + 5.1704166253330186e-05, + 4.0308059396920726e-05 + ], + "bp_grad_norms_F": [ + 0.0028964560478925705, + 0.0030334230978041887, + 0.0031210912857204676, + 0.002908239373937249, + 0.0020543786231428385 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6181640625, + "loss_eval": 1.1597179174423218, + "epoch": 86 + }, + { + "hidden_norms": [ + 256.531982421875, + 235.45941162109375, + 217.1429443359375, + 210.1690673828125, + 202.77899169921875 + ], + "bp_grad_norms_per_sample_med": [ + 4.943124076817185e-05, + 5.149367279955186e-05, + 5.385246913647279e-05, + 5.262702325126156e-05, + 4.104706385987811e-05 + ], + "bp_grad_norms_F": [ + 0.0029305708594620228, + 0.0030687344260513783, + 0.003160761669278145, + 0.002940374892205, + 0.0020649591460824013 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6220703125, + "loss_eval": 1.161116600036621, + "epoch": 87 + }, + { + "hidden_norms": [ + 256.3059387207031, + 235.24874877929688, + 217.37142944335938, + 210.53717041015625, + 203.14259338378906 + ], + "bp_grad_norms_per_sample_med": [ + 4.934536627843045e-05, + 5.2089759265072644e-05, + 5.424721530289389e-05, + 5.292960850056261e-05, + 4.041651845909655e-05 + ], + "bp_grad_norms_F": [ + 0.002913407515734434, + 0.0030489542987197638, + 0.003135726321488619, + 0.002922008978202939, + 0.002062067622318864 + ], + "gamma_dfa": NaN, + "acc_eval": 0.61328125, + "loss_eval": 1.155902624130249, + "epoch": 88 + }, + { + "hidden_norms": [ + 256.7462463378906, + 235.3153839111328, + 216.87977600097656, + 210.572509765625, + 202.77845764160156 + ], + "bp_grad_norms_per_sample_med": [ + 4.990033630747348e-05, + 5.337029870133847e-05, + 5.522817809833214e-05, + 5.3465773817151785e-05, + 4.164372876402922e-05 + ], + "bp_grad_norms_F": [ + 0.0029192205984145403, + 0.0030557813588529825, + 0.003147577866911888, + 0.0029307191725820303, + 0.0020654413383454084 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6162109375, + "loss_eval": 1.1561503410339355, + "epoch": 89 + }, + { + "hidden_norms": [ + 256.68145751953125, + 234.77310180664062, + 216.7720947265625, + 209.664794921875, + 202.22833251953125 + ], + "bp_grad_norms_per_sample_med": [ + 4.9965801736107096e-05, + 5.341583164408803e-05, + 5.593279274762608e-05, + 5.420796514954418e-05, + 4.210533734294586e-05 + ], + "bp_grad_norms_F": [ + 0.002943370258435607, + 0.00308181531727314, + 0.003169203409925103, + 0.002951863221824169, + 0.0020759364124387503 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6142578125, + "loss_eval": 1.1583542823791504, + "epoch": 90 + }, + { + "hidden_norms": [ + 256.4252014160156, + 234.56448364257812, + 216.34176635742188, + 209.5242156982422, + 201.7559356689453 + ], + "bp_grad_norms_per_sample_med": [ + 4.9815931561170146e-05, + 5.2233976020943373e-05, + 5.5747830629115924e-05, + 5.41718618478626e-05, + 4.169955354882404e-05 + ], + "bp_grad_norms_F": [ + 0.0029554881621152163, + 0.0030950764194130898, + 0.0031847492791712284, + 0.0029650654178112745, + 0.002081435639411211 + ], + "gamma_dfa": NaN, + "acc_eval": 0.619140625, + "loss_eval": 1.159005045890808, + "epoch": 91 + }, + { + "hidden_norms": [ + 256.15167236328125, + 234.47537231445312, + 216.2799072265625, + 209.95578002929688, + 201.94126892089844 + ], + "bp_grad_norms_per_sample_med": [ + 5.007637446396984e-05, + 5.260824036668055e-05, + 5.540058555197902e-05, + 5.322153810993768e-05, + 4.1122035327134654e-05 + ], + "bp_grad_norms_F": [ + 0.002944375155493617, + 0.00308293872512877, + 0.003174267942085862, + 0.0029555640649050474, + 0.0020777545869350433 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6259765625, + "loss_eval": 1.1595215797424316, + "epoch": 92 + }, + { + "hidden_norms": [ + 256.1369934082031, + 234.51614379882812, + 216.30044555664062, + 209.4623565673828, + 201.70799255371094 + ], + "bp_grad_norms_per_sample_med": [ + 5.048588718636893e-05, + 5.3128016588743776e-05, + 5.624579353025183e-05, + 5.411298116086982e-05, + 4.109516885364428e-05 + ], + "bp_grad_norms_F": [ + 0.0029427227564156055, + 0.0030841128900647163, + 0.003178349928930402, + 0.0029583934228867292, + 0.002080487785860896 + ], + "gamma_dfa": NaN, + "acc_eval": 0.62109375, + "loss_eval": 1.1565905809402466, + "epoch": 93 + }, + { + "hidden_norms": [ + 255.9738311767578, + 234.28948974609375, + 216.3280029296875, + 209.589111328125, + 201.76205444335938 + ], + "bp_grad_norms_per_sample_med": [ + 5.070818224339746e-05, + 5.330924614099786e-05, + 5.596653136308305e-05, + 5.407314529293217e-05, + 4.100686783203855e-05 + ], + "bp_grad_norms_F": [ + 0.0029472638852894306, + 0.003088710131123662, + 0.0031828396022319794, + 0.0029631764627993107, + 0.0020829145796597004 + ], + "gamma_dfa": NaN, + "acc_eval": 0.619140625, + "loss_eval": 1.1584088802337646, + "epoch": 94 + }, + { + "hidden_norms": [ + 255.89332580566406, + 234.19253540039062, + 216.33103942871094, + 209.45057678222656, + 201.54258728027344 + ], + "bp_grad_norms_per_sample_med": [ + 5.041498661739752e-05, + 5.279783727019094e-05, + 5.5226330005098134e-05, + 5.407658318290487e-05, + 4.1154507925966755e-05 + ], + "bp_grad_norms_F": [ + 0.002949801040813327, + 0.00309200631454587, + 0.0031869453378021717, + 0.0029658437706530094, + 0.002085147425532341 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6201171875, + "loss_eval": 1.158942699432373, + "epoch": 95 + }, + { + "hidden_norms": [ + 255.63841247558594, + 234.11630249023438, + 216.2378692626953, + 209.49703979492188, + 201.46917724609375 + ], + "bp_grad_norms_per_sample_med": [ + 5.020374737796374e-05, + 5.243354826234281e-05, + 5.587147097685374e-05, + 5.334949673851952e-05, + 4.121083111385815e-05 + ], + "bp_grad_norms_F": [ + 0.002950438065454364, + 0.003093136241659522, + 0.0031878354493528605, + 0.0029664800968021154, + 0.0020843464881181717 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6201171875, + "loss_eval": 1.1582318544387817, + "epoch": 96 + }, + { + "hidden_norms": [ + 255.8027801513672, + 234.07070922851562, + 216.1463623046875, + 209.4476318359375, + 201.27279663085938 + ], + "bp_grad_norms_per_sample_med": [ + 5.0307549827266484e-05, + 5.278814569464885e-05, + 5.604944453807548e-05, + 5.306081584421918e-05, + 4.1182753193425015e-05 + ], + "bp_grad_norms_F": [ + 0.002952505135908723, + 0.003095670836046338, + 0.0031897351145744324, + 0.00296790711581707, + 0.002085329731926322 + ], + "gamma_dfa": NaN, + "acc_eval": 0.619140625, + "loss_eval": 1.15840482711792, + "epoch": 97 + }, + { + "hidden_norms": [ + 255.60986328125, + 234.11065673828125, + 216.12860107421875, + 209.42921447753906, + 201.28346252441406 + ], + "bp_grad_norms_per_sample_med": [ + 5.0218295655213296e-05, + 5.269264511298388e-05, + 5.5988577514654025e-05, + 5.322946162777953e-05, + 4.136392453801818e-05 + ], + "bp_grad_norms_F": [ + 0.002954497467726469, + 0.003097717184573412, + 0.0031918887980282307, + 0.002969518303871155, + 0.00208606431260705 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6181640625, + "loss_eval": 1.1585593223571777, + "epoch": 98 + }, + { + "hidden_norms": [ + 255.57244873046875, + 234.11569213867188, + 216.13626098632812, + 209.40383911132812, + 201.28627014160156 + ], + "bp_grad_norms_per_sample_med": [ + 5.005559796700254e-05, + 5.2759678510483354e-05, + 5.5868193157948554e-05, + 5.3356721764430404e-05, + 4.125368650420569e-05 + ], + "bp_grad_norms_F": [ + 0.0029547216836363077, + 0.00309770368039608, + 0.003191797062754631, + 0.0029695071280002594, + 0.0020862380042672157 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6181640625, + "loss_eval": 1.1585021018981934, + "epoch": 99 + }, + { + "hidden_norms": [ + 255.5677947998047, + 234.11875915527344, + 216.1315460205078, + 209.4070281982422, + 201.2687530517578 + ], + "bp_grad_norms_per_sample_med": [ + 5.006368883186951e-05, + 5.279530887492001e-05, + 5.585233884630725e-05, + 5.334027810022235e-05, + 4.123226608498953e-05 + ], + "bp_grad_norms_F": [ + 0.002954971743747592, + 0.0030978918075561523, + 0.003191987983882427, + 0.002969692926853895, + 0.002086336025968194 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6181640625, + "loss_eval": 1.1584858894348145, + "epoch": 100 + } + ], + "dfa_log": [ + { + "hidden_norms": [ + 8.690464973449707, + 8.719978332519531, + 8.73093318939209, + 8.75937271118164, + 8.790542602539062 + ], + "bp_grad_norms_per_sample_med": [ + 0.0009977436857298017, + 0.000992251792922616, + 0.0009848480112850666, + 0.0009761073160916567, + 0.0009735578205436468 + ], + "bp_grad_norms_F": [ + 0.035340383648872375, + 0.03499162197113037, + 0.03454425558447838, + 0.034113530069589615, + 0.033874060958623886 + ], + "gamma_dfa": 0.006830460682976991, + "acc_eval": 0.095703125, + "loss_eval": 2.486203908920288, + "epoch": 0 + }, + { + "hidden_norms": [ + 861.2440185546875, + 2373.370849609375, + 4675.99951171875, + 6090.40869140625, + 6962.984375 + ], + "bp_grad_norms_per_sample_med": [ + 2.380779051236459e-06, + 1.3197138741816161e-06, + 1.3154849511920474e-06, + 1.3285966815601569e-06, + 1.3258621720524388e-06 + ], + "bp_grad_norms_F": [ + 0.0001391920231981203, + 6.988063250901178e-05, + 6.874280370539054e-05, + 6.720778765156865e-05, + 6.707101420033723e-05 + ], + "gamma_dfa": 0.10659917898010463, + "acc_eval": 0.3037109375, + "loss_eval": 1.9665180444717407, + "epoch": 1 + }, + { + "hidden_norms": [ + 1679.144775390625, + 8850.173828125, + 31676.23046875, + 40712.2109375, + 45375.55859375 + ], + "bp_grad_norms_per_sample_med": [ + 1.1123192962259054e-06, + 2.7329409135745664e-07, + 2.7436919936008053e-07, + 2.75225175982996e-07, + 2.744452558545163e-07 + ], + "bp_grad_norms_F": [ + 7.363831537077203e-05, + 2.2035641450202093e-05, + 1.9731569409486838e-05, + 1.9022963897441514e-05, + 1.7921549442689866e-05 + ], + "gamma_dfa": 0.10788296448299661, + "acc_eval": 0.29296875, + "loss_eval": 1.972828984260559, + "epoch": 2 + }, + { + "hidden_norms": [ + 2530.16015625, + 21436.486328125, + 88828.296875, + 114079.484375, + 124374.78125 + ], + "bp_grad_norms_per_sample_med": [ + 8.110604312605574e-07, + 1.1556701196013819e-07, + 1.1451407289087001e-07, + 1.1523399479074214e-07, + 1.1543902900257308e-07 + ], + "bp_grad_norms_F": [ + 0.00011594755778787658, + 1.8132053810404614e-05, + 1.1369732419552747e-05, + 8.215727575588971e-06, + 7.219489361887099e-06 + ], + "gamma_dfa": 0.1084698709892109, + "acc_eval": 0.318359375, + "loss_eval": 1.9430747032165527, + "epoch": 3 + }, + { + "hidden_norms": [ + 3363.359375, + 41233.19921875, + 182355.03125, + 240225.46875, + 257539.390625 + ], + "bp_grad_norms_per_sample_med": [ + 6.176944111757621e-07, + 6.474806468759198e-08, + 6.307719502274267e-08, + 6.379124783961743e-08, + 6.374663996666641e-08 + ], + "bp_grad_norms_F": [ + 3.5978406231151894e-05, + 4.323610028222902e-06, + 3.5177260997443227e-06, + 3.4877305097325006e-06, + 3.479522774796351e-06 + ], + "gamma_dfa": 0.10987455540453084, + "acc_eval": 0.3173828125, + "loss_eval": 1.959622859954834, + "epoch": 4 + }, + { + "hidden_norms": [ + 4170.119140625, + 70935.8515625, + 335571.09375, + 441693.5625, + 469995.8125 + ], + "bp_grad_norms_per_sample_med": [ + 4.856212285631045e-07, + 4.0158354863706336e-08, + 3.8916198263905244e-08, + 3.9119736783277403e-08, + 3.916203894505088e-08 + ], + "bp_grad_norms_F": [ + 2.611691706988495e-05, + 2.208936621173052e-06, + 1.6750394706832594e-06, + 1.6634863868603134e-06, + 1.659758027017233e-06 + ], + "gamma_dfa": 0.1083616423420608, + "acc_eval": 0.3017578125, + "loss_eval": 1.932759165763855, + "epoch": 5 + }, + { + "hidden_norms": [ + 5013.92236328125, + 106898.6171875, + 530492.6875, + 719779.5, + 763058.3125 + ], + "bp_grad_norms_per_sample_med": [ + 4.242272950705228e-07, + 2.814423716301917e-08, + 2.683668043346188e-08, + 2.6872848835068908e-08, + 2.6873932412740942e-08 + ], + "bp_grad_norms_F": [ + 2.2560918296221644e-05, + 1.2798993793694535e-06, + 1.0339434766137856e-06, + 1.0326252777304035e-06, + 1.030524344969308e-06 + ], + "gamma_dfa": 0.11125951120629907, + "acc_eval": 0.3095703125, + "loss_eval": 1.9531997442245483, + "epoch": 6 + }, + { + "hidden_norms": [ + 5811.36279296875, + 156264.234375, + 809496.6875, + 1104418.875, + 1163013.625 + ], + "bp_grad_norms_per_sample_med": [ + 3.5694648659045924e-07, + 2.0201103012595922e-08, + 1.89678583950581e-08, + 1.8967025283700423e-08, + 1.8932304612917505e-08 + ], + "bp_grad_norms_F": [ + 1.9050115952268243e-05, + 9.341621307612513e-07, + 6.890153372296481e-07, + 6.891143584653037e-07, + 6.886222649882257e-07 + ], + "gamma_dfa": 0.11049338441807777, + "acc_eval": 0.3076171875, + "loss_eval": 1.9362893104553223, + "epoch": 7 + }, + { + "hidden_norms": [ + 6660.841796875, + 217521.65625, + 1158874.5, + 1615614.625, + 1694436.0 + ], + "bp_grad_norms_per_sample_med": [ + 3.212987280676316e-07, + 1.5324568281016582e-08, + 1.4045938634410504e-08, + 1.4053354924215e-08, + 1.4027999206689401e-08 + ], + "bp_grad_norms_F": [ + 1.7352898794342764e-05, + 6.539408445860317e-07, + 5.044145154897706e-07, + 5.047700142313261e-07, + 5.048477760283276e-07 + ], + "gamma_dfa": 0.11245601065456867, + "acc_eval": 0.2978515625, + "loss_eval": 1.9441646337509155, + "epoch": 8 + }, + { + "hidden_norms": [ + 7480.24169921875, + 295557.40625, + 1585967.0, + 2246827.75, + 2356427.0 + ], + "bp_grad_norms_per_sample_med": [ + 2.889860297727864e-07, + 1.193609300287335e-08, + 1.1067995409064224e-08, + 1.103960478587851e-08, + 1.1055576010221557e-08 + ], + "bp_grad_norms_F": [ + 1.5521236491622403e-05, + 4.5597263920171827e-07, + 3.8742672359148855e-07, + 3.8764648024880444e-07, + 3.877208314406744e-07 + ], + "gamma_dfa": 0.11317481694277376, + "acc_eval": 0.30078125, + "loss_eval": 1.9356626272201538, + "epoch": 9 + }, + { + "hidden_norms": [ + 8273.8408203125, + 388312.4375, + 2104540.5, + 3043837.75, + 3180222.75 + ], + "bp_grad_norms_per_sample_med": [ + 2.619119925384439e-07, + 9.19370091168048e-09, + 8.51864179196582e-09, + 8.482276214749618e-09, + 8.476813917468462e-09 + ], + "bp_grad_norms_F": [ + 1.4210399967851117e-05, + 3.5443522961031704e-07, + 3.0404962103602884e-07, + 3.0431687036980293e-07, + 3.0445681886703824e-07 + ], + "gamma_dfa": 0.11495429277420044, + "acc_eval": 0.30859375, + "loss_eval": 1.936495304107666, + "epoch": 10 + }, + { + "hidden_norms": [ + 9086.7265625, + 501434.3125, + 2767834.0, + 4046107.75, + 4214970.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.471596189934644e-07, + 7.524303846651037e-09, + 7.025830139895106e-09, + 7.019247405537499e-09, + 7.011437208603866e-09 + ], + "bp_grad_norms_F": [ + 1.3406845027930103e-05, + 4.39807990915142e-07, + 2.471441860052437e-07, + 2.474318421263888e-07, + 2.475488827258232e-07 + ], + "gamma_dfa": 0.11651878873817623, + "acc_eval": 0.30859375, + "loss_eval": 1.9542189836502075, + "epoch": 11 + }, + { + "hidden_norms": [ + 9854.2626953125, + 612623.625, + 3485830.25, + 5139234.5, + 5346583.0 + ], + "bp_grad_norms_per_sample_med": [ + 2.3279393701614026e-07, + 6.378383154981293e-09, + 5.839252636974379e-09, + 5.8299334249056756e-09, + 5.825462778830115e-09 + ], + "bp_grad_norms_F": [ + 1.2396733836794738e-05, + 2.516441668376501e-07, + 2.0780194631697668e-07, + 2.0795941679807584e-07, + 2.080748373600727e-07 + ], + "gamma_dfa": 0.11470737145282328, + "acc_eval": 0.3046875, + "loss_eval": 1.9374217987060547, + "epoch": 12 + }, + { + "hidden_norms": [ + 10653.666015625, + 745430.125, + 4325399.0, + 6434218.5, + 6688000.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.1187355514484807e-07, + 5.294193083926757e-09, + 4.98529528769609e-09, + 4.991081326011226e-09, + 4.989251234377434e-09 + ], + "bp_grad_norms_F": [ + 1.1210328011657111e-05, + 2.000140426616781e-07, + 1.7647043648594263e-07, + 1.7654315342952032e-07, + 1.7665543339262513e-07 + ], + "gamma_dfa": 0.11383607180323452, + "acc_eval": 0.291015625, + "loss_eval": 1.9233126640319824, + "epoch": 13 + }, + { + "hidden_norms": [ + 11404.5263671875, + 893037.0625, + 5298411.0, + 7870617.0, + 8170729.5 + ], + "bp_grad_norms_per_sample_med": [ + 1.9957101926593168e-07, + 4.642851436642559e-09, + 4.288450483613815e-09, + 4.285100718703916e-09, + 4.289933297485504e-09 + ], + "bp_grad_norms_F": [ + 1.0866175216506235e-05, + 1.758858729772328e-07, + 1.538420519864303e-07, + 1.5386244456294662e-07, + 1.5397839092656795e-07 + ], + "gamma_dfa": 0.1139617893495597, + "acc_eval": 0.30859375, + "loss_eval": 1.932279109954834, + "epoch": 14 + }, + { + "hidden_norms": [ + 12171.2412109375, + 1062066.875, + 6388639.0, + 9528923.0, + 9891447.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.9334000000981177e-07, + 4.044033108385747e-09, + 3.810601167941741e-09, + 3.801854830953744e-09, + 3.8046659156520946e-09 + ], + "bp_grad_norms_F": [ + 1.0322088201064616e-05, + 1.529211743900305e-07, + 1.350435638869385e-07, + 1.3504063645086717e-07, + 1.3514510044387862e-07 + ], + "gamma_dfa": 0.11320225725648925, + "acc_eval": 0.3193359375, + "loss_eval": 1.921454906463623, + "epoch": 15 + }, + { + "hidden_norms": [ + 12924.974609375, + 1273988.625, + 7574387.5, + 11500420.0, + 11920183.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.8497563303299103e-07, + 3.6160896499382034e-09, + 3.388503477808058e-09, + 3.385580482628825e-09, + 3.3854716807724117e-09 + ], + "bp_grad_norms_F": [ + 9.76746559899766e-06, + 1.3485876593222201e-07, + 1.1896084117779537e-07, + 1.1894425000491538e-07, + 1.1903343022368063e-07 + ], + "gamma_dfa": 0.11276883230311796, + "acc_eval": 0.306640625, + "loss_eval": 1.9356281757354736, + "epoch": 16 + }, + { + "hidden_norms": [ + 13632.109375, + 1465148.75, + 8873670.0, + 13535396.0, + 14019399.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.7523743167657813e-07, + 3.18161341716916e-09, + 3.0166942277531916e-09, + 3.0173783471809656e-09, + 3.0215525637089513e-09 + ], + "bp_grad_norms_F": [ + 9.427177246834617e-06, + 1.1933295240851294e-07, + 1.0663712401992598e-07, + 1.0661205607220836e-07, + 1.0669979388922002e-07 + ], + "gamma_dfa": 0.11264261469477788, + "acc_eval": 0.3203125, + "loss_eval": 1.926081657409668, + "epoch": 17 + }, + { + "hidden_norms": [ + 14394.1845703125, + 1699121.5, + 10379340.0, + 15814831.0, + 16377349.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.6784264289526618e-07, + 2.8936604223872564e-09, + 2.7336535257660444e-09, + 2.7314879247342105e-09, + 2.733708148738856e-09 + ], + "bp_grad_norms_F": [ + 9.04797616385622e-06, + 1.087723191517398e-07, + 9.662341682314946e-08, + 9.658959498892727e-08, + 9.665905764677518e-08 + ], + "gamma_dfa": 0.11253939801827073, + "acc_eval": 0.3154296875, + "loss_eval": 1.922888994216919, + "epoch": 18 + }, + { + "hidden_norms": [ + 15138.509765625, + 1969752.0, + 12071423.0, + 18393228.0, + 19044056.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5837277089758572e-07, + 2.609559235366987e-09, + 2.4664128517315476e-09, + 2.4639881246457662e-09, + 2.4657431652030937e-09 + ], + "bp_grad_norms_F": [ + 8.502225682605058e-06, + 9.664965716638108e-08, + 8.704243725787819e-08, + 8.700892806245974e-08, + 8.707356613513184e-08 + ], + "gamma_dfa": 0.11155363742727786, + "acc_eval": 0.3125, + "loss_eval": 1.9243314266204834, + "epoch": 19 + }, + { + "hidden_norms": [ + 15811.982421875, + 2268261.5, + 13867645.0, + 21375086.0, + 22149420.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5073908343765652e-07, + 2.376487229227564e-09, + 2.2317030445861974e-09, + 2.2317414583028494e-09, + 2.2311954506193388e-09 + ], + "bp_grad_norms_F": [ + 8.195236659958027e-06, + 8.728681422098816e-08, + 7.846901439734211e-08, + 7.84395552955175e-08, + 7.849664029890846e-08 + ], + "gamma_dfa": 0.11315422086045146, + "acc_eval": 0.3095703125, + "loss_eval": 1.924652338027954, + "epoch": 20 + }, + { + "hidden_norms": [ + 16487.935546875, + 2556303.5, + 15702076.0, + 24321856.0, + 25185988.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.443007136003871e-07, + 2.174122659681643e-09, + 2.0514809850880056e-09, + 2.052057856971601e-09, + 2.05182382195801e-09 + ], + "bp_grad_norms_F": [ + 8.002187314559706e-06, + 8.025331510452816e-08, + 7.266314838716426e-08, + 7.2638087544874e-08, + 7.268938162496852e-08 + ], + "gamma_dfa": 0.11267373809823766, + "acc_eval": 0.310546875, + "loss_eval": 1.9348869323730469, + "epoch": 21 + }, + { + "hidden_norms": [ + 17153.216796875, + 2908610.25, + 17831918.0, + 27718528.0, + 28682196.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.43127067531168e-07, + 2.0134633960111614e-09, + 1.908325941712974e-09, + 1.9070569567958273e-09, + 1.9050578892176873e-09 + ], + "bp_grad_norms_F": [ + 7.550354439445073e-06, + 7.309487415341209e-08, + 6.660556550741603e-08, + 6.657593587533484e-08, + 6.66217090383725e-08 + ], + "gamma_dfa": 0.11266892295680009, + "acc_eval": 0.310546875, + "loss_eval": 1.9210036993026733, + "epoch": 22 + }, + { + "hidden_norms": [ + 17796.001953125, + 3236156.25, + 20246900.0, + 31494576.0, + 32571052.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.3518946673229948e-07, + 1.8474217711172969e-09, + 1.7507182370479768e-09, + 1.7496273319039801e-09, + 1.7511733174657707e-09 + ], + "bp_grad_norms_F": [ + 7.284599632839672e-06, + 6.709063882226474e-08, + 6.151606157800416e-08, + 6.149030440383285e-08, + 6.153235432293513e-08 + ], + "gamma_dfa": 0.11186322406865656, + "acc_eval": 0.3115234375, + "loss_eval": 1.9283114671707153, + "epoch": 23 + }, + { + "hidden_norms": [ + 18460.63671875, + 3604280.25, + 22626418.0, + 35429216.0, + 36622356.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.3137012899733236e-07, + 1.7567419741126855e-09, + 1.6593408869169934e-09, + 1.6595901319860218e-09, + 1.6608053821087765e-09 + ], + "bp_grad_norms_F": [ + 7.130098310881294e-06, + 6.264485108431472e-08, + 5.731577701340029e-08, + 5.729528851361465e-08, + 5.733248187311801e-08 + ], + "gamma_dfa": 0.11299802124267444, + "acc_eval": 0.2978515625, + "loss_eval": 1.942992925643921, + "epoch": 24 + }, + { + "hidden_norms": [ + 19115.904296875, + 4038094.25, + 25405536.0, + 39835936.0, + 41154148.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.2640579427625198e-07, + 1.5953075527264104e-09, + 1.5003157605164574e-09, + 1.4991035079958692e-09, + 1.4979777418488993e-09 + ], + "bp_grad_norms_F": [ + 6.799941729696002e-06, + 5.7522207441706996e-08, + 5.297441418861126e-08, + 5.2952817242157835e-08, + 5.298943506204523e-08 + ], + "gamma_dfa": 0.11197824770351872, + "acc_eval": 0.298828125, + "loss_eval": 1.9319748878479004, + "epoch": 25 + }, + { + "hidden_norms": [ + 19780.572265625, + 4490110.5, + 28153634.0, + 44291160.0, + 45731404.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.2269519800156559e-07, + 1.4798114955638653e-09, + 1.4102843337582271e-09, + 1.4097069067631196e-09, + 1.4109671209183716e-09 + ], + "bp_grad_norms_F": [ + 6.577545264008222e-06, + 5.327013141709358e-08, + 4.94958669605694e-08, + 4.9476955865657146e-08, + 4.950964083150211e-08 + ], + "gamma_dfa": 0.11252723945653997, + "acc_eval": 0.3115234375, + "loss_eval": 1.9247620105743408, + "epoch": 26 + }, + { + "hidden_norms": [ + 20381.6796875, + 4929967.0, + 30880866.0, + 48905308.0, + 50483824.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.1911821928833888e-07, + 1.377626124288156e-09, + 1.3317531522005766e-09, + 1.3313244950907688e-09, + 1.3306112878197496e-09 + ], + "bp_grad_norms_F": [ + 6.326347829599399e-06, + 4.955709798082353e-08, + 4.649340468176888e-08, + 4.6477893533847237e-08, + 4.650862805988254e-08 + ], + "gamma_dfa": 0.11152978462632746, + "acc_eval": 0.2978515625, + "loss_eval": 1.9170689582824707, + "epoch": 27 + }, + { + "hidden_norms": [ + 21001.544921875, + 5444194.5, + 33907076.0, + 53832084.0, + 55554724.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.1559290413742929e-07, + 1.3061883796794405e-09, + 1.2378150726632953e-09, + 1.2379912650573033e-09, + 1.2384089309591673e-09 + ], + "bp_grad_norms_F": [ + 6.23212235950632e-06, + 4.7021419646853246e-08, + 4.3998962695468435e-08, + 4.3984215380987735e-08, + 4.4011432720481025e-08 + ], + "gamma_dfa": 0.11285935762862209, + "acc_eval": 0.3193359375, + "loss_eval": 1.9198743104934692, + "epoch": 28 + }, + { + "hidden_norms": [ + 21623.64453125, + 5895143.0, + 37025488.0, + 58791348.0, + 60687928.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.1377478870144841e-07, + 1.2547849426169932e-09, + 1.1888731110687445e-09, + 1.1877619998656996e-09, + 1.1896320595283782e-09 + ], + "bp_grad_norms_F": [ + 6.024939921189798e-06, + 4.443810297516393e-08, + 4.161294953064498e-08, + 4.159920763413538e-08, + 4.162497546644772e-08 + ], + "gamma_dfa": 0.11188641694025137, + "acc_eval": 0.3115234375, + "loss_eval": 1.917163372039795, + "epoch": 29 + }, + { + "hidden_norms": [ + 22195.29296875, + 6369258.5, + 40117404.0, + 64235468.0, + 66270544.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.1021134582733794e-07, + 1.153794726427293e-09, + 1.1123471033158694e-09, + 1.1119479781385166e-09, + 1.1122585075185043e-09 + ], + "bp_grad_norms_F": [ + 5.96408472119947e-06, + 4.2022573865097e-08, + 3.954471239353552e-08, + 3.953241289877951e-08, + 3.955623029128219e-08 + ], + "gamma_dfa": 0.11316103948047385, + "acc_eval": 0.3125, + "loss_eval": 1.9332935810089111, + "epoch": 30 + }, + { + "hidden_norms": [ + 22786.513671875, + 6899202.5, + 43496572.0, + 69758712.0, + 71960648.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.0609301170916297e-07, + 1.1096534802135238e-09, + 1.05872854927469e-09, + 1.0590027743617725e-09, + 1.0579482845329835e-09 + ], + "bp_grad_norms_F": [ + 5.711301128030755e-06, + 3.962547268088201e-08, + 3.742779952631281e-08, + 3.741677900848117e-08, + 3.743992138538488e-08 + ], + "gamma_dfa": 0.11225346029095817, + "acc_eval": 0.31640625, + "loss_eval": 1.9247596263885498, + "epoch": 31 + }, + { + "hidden_norms": [ + 23367.078125, + 7480530.0, + 47061616.0, + 75610000.0, + 77981320.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.054115728038596e-07, + 1.0618838031106748e-09, + 1.0191184562913236e-09, + 1.01949926278877e-09, + 1.0192180432966325e-09 + ], + "bp_grad_norms_F": [ + 5.768360097135883e-06, + 3.826109207238915e-08, + 3.593911657162607e-08, + 3.592855790657268e-08, + 3.594948694285449e-08 + ], + "gamma_dfa": 0.11492819089471595, + "acc_eval": 0.314453125, + "loss_eval": 1.939026117324829, + "epoch": 32 + }, + { + "hidden_norms": [ + 23912.85546875, + 8053234.0, + 51027352.0, + 81931568.0, + 84515616.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.0266481353937706e-07, + 9.919576271499864e-10, + 9.572211912001194e-10, + 9.572932446744176e-10, + 9.577664217275128e-10 + ], + "bp_grad_norms_F": [ + 5.472421435115393e-06, + 3.587240016145188e-08, + 3.39712649122248e-08, + 3.396091940999213e-08, + 3.398050552050336e-08 + ], + "gamma_dfa": 0.11253271601162851, + "acc_eval": 0.3125, + "loss_eval": 1.9187253713607788, + "epoch": 33 + }, + { + "hidden_norms": [ + 24486.8359375, + 8682399.0, + 55030492.0, + 88236184.0, + 91025880.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.0101911129822838e-07, + 9.702203485062455e-10, + 9.30652499508966e-10, + 9.30728105696943e-10, + 9.303261494508774e-10 + ], + "bp_grad_norms_F": [ + 5.286908617563313e-06, + 3.4072087373715476e-08, + 3.2386065385026086e-08, + 3.2376764380614986e-08, + 3.2396076932172946e-08 + ], + "gamma_dfa": 0.11125053715659305, + "acc_eval": 0.3125, + "loss_eval": 1.9214520454406738, + "epoch": 34 + }, + { + "hidden_norms": [ + 25051.91796875, + 9255517.0, + 58776788.0, + 94748472.0, + 97725744.0 + ], + "bp_grad_norms_per_sample_med": [ + 9.793318156425812e-08, + 9.150468716079274e-10, + 8.747003676923271e-10, + 8.746902091516517e-10, + 8.743750723461119e-10 + ], + "bp_grad_norms_F": [ + 5.294927632348845e-06, + 3.279593840943562e-08, + 3.115108881956985e-08, + 3.114205782139834e-08, + 3.115993862934374e-08 + ], + "gamma_dfa": 0.11356210591475246, + "acc_eval": 0.314453125, + "loss_eval": 1.9334189891815186, + "epoch": 35 + }, + { + "hidden_norms": [ + 25539.400390625, + 9852175.0, + 62733436.0, + 101270040.0, + 104412528.0 + ], + "bp_grad_norms_per_sample_med": [ + 9.590952032567657e-08, + 8.831190223546059e-10, + 8.506708115696426e-10, + 8.503630022360653e-10, + 8.506205184666271e-10 + ], + "bp_grad_norms_F": [ + 5.156022325536469e-06, + 3.1347209272780674e-08, + 2.992178593785866e-08, + 2.9913682197957314e-08, + 2.9930919964726854e-08 + ], + "gamma_dfa": 0.11300312746607233, + "acc_eval": 0.3291015625, + "loss_eval": 1.9240946769714355, + "epoch": 36 + }, + { + "hidden_norms": [ + 26058.23046875, + 10503947.0, + 66954420.0, + 108182184.0, + 111535672.0 + ], + "bp_grad_norms_per_sample_med": [ + 9.312070403666439e-08, + 8.298656206662258e-10, + 8.076765367626138e-10, + 8.076487811869981e-10, + 8.083505531608637e-10 + ], + "bp_grad_norms_F": [ + 5.084062195237493e-06, + 3.008365112577849e-08, + 2.8773875726528786e-08, + 2.8766294235538226e-08, + 2.878227967073599e-08 + ], + "gamma_dfa": 0.11286510003083094, + "acc_eval": 0.322265625, + "loss_eval": 1.9284429550170898, + "epoch": 37 + }, + { + "hidden_norms": [ + 26547.951171875, + 11128081.0, + 71119440.0, + 115305784.0, + 118851064.0 + ], + "bp_grad_norms_per_sample_med": [ + 9.278523549482998e-08, + 8.024952369289906e-10, + 7.701798643289237e-10, + 7.699446635811569e-10, + 7.706733029522184e-10 + ], + "bp_grad_norms_F": [ + 4.915013960271608e-06, + 2.8781727223758935e-08, + 2.7522533585511155e-08, + 2.751522920618754e-08, + 2.7530596469205193e-08 + ], + "gamma_dfa": 0.11193350053144968, + "acc_eval": 0.3046875, + "loss_eval": 1.9189305305480957, + "epoch": 38 + }, + { + "hidden_norms": [ + 27052.158203125, + 11815623.0, + 75267568.0, + 122747224.0, + 126538816.0 + ], + "bp_grad_norms_per_sample_med": [ + 9.042137349979384e-08, + 7.699976212194315e-10, + 7.406790181185841e-10, + 7.406811275423308e-10, + 7.407266466863405e-10 + ], + "bp_grad_norms_F": [ + 4.783120857609902e-06, + 2.7648876965713498e-08, + 2.6492134708178128e-08, + 2.6485189152936073e-08, + 2.6500073246893407e-08 + ], + "gamma_dfa": 0.1119252087228233, + "acc_eval": 0.3076171875, + "loss_eval": 1.9200191497802734, + "epoch": 39 + }, + { + "hidden_norms": [ + 27573.439453125, + 12446571.0, + 79631776.0, + 129906592.0, + 133912576.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.999781897500725e-08, + 7.431448234562765e-10, + 7.212022645752825e-10, + 7.211930497241781e-10, + 7.215293917894883e-10 + ], + "bp_grad_norms_F": [ + 4.7411135710717645e-06, + 2.6746526771148638e-08, + 2.567137080689008e-08, + 2.566465262532347e-08, + 2.5678824400188205e-08 + ], + "gamma_dfa": 0.1126710368462227, + "acc_eval": 0.3291015625, + "loss_eval": 1.9156131744384766, + "epoch": 40 + }, + { + "hidden_norms": [ + 28020.1953125, + 13139046.0, + 84357848.0, + 137652880.0, + 141892912.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.93968845616655e-08, + 7.234106647047156e-10, + 7.015109604324721e-10, + 7.013014613477253e-10, + 7.019884118442121e-10 + ], + "bp_grad_norms_F": [ + 4.706852905655978e-06, + 2.5948086346261334e-08, + 2.480862271170281e-08, + 2.4802661258149783e-08, + 2.4816182886411298e-08 + ], + "gamma_dfa": 0.11299434299689892, + "acc_eval": 0.3134765625, + "loss_eval": 1.917637825012207, + "epoch": 41 + }, + { + "hidden_norms": [ + 28489.8828125, + 13728548.0, + 88388480.0, + 144432224.0, + 148906448.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.832166287220389e-08, + 7.007013858029154e-10, + 6.849794620400473e-10, + 6.849559253119253e-10, + 6.857461820608535e-10 + ], + "bp_grad_norms_F": [ + 4.678155619330937e-06, + 2.5201579489930737e-08, + 2.418770428391781e-08, + 2.4181909807907687e-08, + 2.419472266979028e-08 + ], + "gamma_dfa": 0.11410953500489995, + "acc_eval": 0.314453125, + "loss_eval": 1.9333336353302002, + "epoch": 42 + }, + { + "hidden_norms": [ + 28918.5546875, + 14414132.0, + 92681240.0, + 151898304.0, + 156571040.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.454645694655483e-08, + 6.792187368098723e-10, + 6.553820264265653e-10, + 6.554308207284976e-10, + 6.556304388283252e-10 + ], + "bp_grad_norms_F": [ + 4.511462066147942e-06, + 2.4275013998931172e-08, + 2.338800797474505e-08, + 2.338238225263467e-08, + 2.3394919779207157e-08 + ], + "gamma_dfa": 0.11213390928969602, + "acc_eval": 0.3037109375, + "loss_eval": 1.9155462980270386, + "epoch": 43 + }, + { + "hidden_norms": [ + 29332.80078125, + 15055747.0, + 96975336.0, + 159279984.0, + 164189344.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.274933804841567e-08, + 6.447524736330479e-10, + 6.278008113369538e-10, + 6.277455777414787e-10, + 6.275664987676066e-10 + ], + "bp_grad_norms_F": [ + 4.426053692441201e-06, + 2.3487833900048827e-08, + 2.272233956546188e-08, + 2.2717028258512073e-08, + 2.272915367029782e-08 + ], + "gamma_dfa": 0.11223121372131573, + "acc_eval": 0.3173828125, + "loss_eval": 1.9162096977233887, + "epoch": 44 + }, + { + "hidden_norms": [ + 29736.349609375, + 15823423.0, + 101458080.0, + 166481584.0, + 171638352.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.434898290943238e-08, + 6.55440757224568e-10, + 6.384029416217629e-10, + 6.384751061183636e-10, + 6.382625539202991e-10 + ], + "bp_grad_norms_F": [ + 4.4675143726635724e-06, + 2.308258295613541e-08, + 2.2235047580920764e-08, + 2.2229942331364327e-08, + 2.2241332331418562e-08 + ], + "gamma_dfa": 0.11349940555010107, + "acc_eval": 0.3134765625, + "loss_eval": 1.935542106628418, + "epoch": 45 + }, + { + "hidden_norms": [ + 30168.33984375, + 16512403.0, + 106240736.0, + 174293312.0, + 179686352.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.255891259523196e-08, + 6.180955192114368e-10, + 6.070984825967685e-10, + 6.070359770404821e-10, + 6.068190394614703e-10 + ], + "bp_grad_norms_F": [ + 4.380563041195273e-06, + 2.227099038520919e-08, + 2.156298606337259e-08, + 2.1558207663474604e-08, + 2.1569489305761635e-08 + ], + "gamma_dfa": 0.11322060551538016, + "acc_eval": 0.3212890625, + "loss_eval": 1.9260894060134888, + "epoch": 46 + }, + { + "hidden_norms": [ + 30560.10546875, + 17240642.0, + 111300408.0, + 182109792.0, + 187762288.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.14508851476603e-08, + 6.030238530740917e-10, + 5.890909982042558e-10, + 5.890405385677866e-10, + 5.88881721164114e-10 + ], + "bp_grad_norms_F": [ + 4.2523201955191325e-06, + 2.1578937747790405e-08, + 2.0878760054188206e-08, + 2.0874194817110947e-08, + 2.08853556671329e-08 + ], + "gamma_dfa": 0.11227845895700739, + "acc_eval": 0.3193359375, + "loss_eval": 1.9136857986450195, + "epoch": 47 + }, + { + "hidden_norms": [ + 30927.453125, + 17981622.0, + 115698520.0, + 189446720.0, + 195329056.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.999796736157805e-08, + 5.827110460820961e-10, + 5.688347570753649e-10, + 5.688115534141502e-10, + 5.69402469619007e-10 + ], + "bp_grad_norms_F": [ + 4.254153282090556e-06, + 2.1111715042820833e-08, + 2.044158087244341e-08, + 2.043698899001356e-08, + 2.0447874504725405e-08 + ], + "gamma_dfa": 0.11285097184736514, + "acc_eval": 0.3251953125, + "loss_eval": 1.9130818843841553, + "epoch": 48 + }, + { + "hidden_norms": [ + 31297.09765625, + 18680606.0, + 120312440.0, + 197032272.0, + 203114064.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.911415877970285e-08, + 5.74776504169705e-10, + 5.532440061628563e-10, + 5.531197722064007e-10, + 5.533422053893844e-10 + ], + "bp_grad_norms_F": [ + 4.17833052779315e-06, + 2.0621358842731752e-08, + 1.9960678443453617e-08, + 1.9956562624656726e-08, + 1.9966993392017685e-08 + ], + "gamma_dfa": 0.11243592749451636, + "acc_eval": 0.3125, + "loss_eval": 1.9191365242004395, + "epoch": 49 + }, + { + "hidden_norms": [ + 31637.244140625, + 19360764.0, + 124754384.0, + 204064768.0, + 210382992.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.003268447964729e-08, + 5.749308251701279e-10, + 5.599746222273438e-10, + 5.599968821989876e-10, + 5.599694041791281e-10 + ], + "bp_grad_norms_F": [ + 4.128593900531996e-06, + 2.0178502424528233e-08, + 1.9575201903876405e-08, + 1.9571228193626666e-08, + 1.9581117172151608e-08 + ], + "gamma_dfa": 0.11233749791426817, + "acc_eval": 0.330078125, + "loss_eval": 1.9152377843856812, + "epoch": 50 + }, + { + "hidden_norms": [ + 31964.60546875, + 20067444.0, + 128969536.0, + 211772368.0, + 218247696.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.774531951554309e-08, + 5.427598370744136e-10, + 5.298457228519737e-10, + 5.297982053065198e-10, + 5.30141264221129e-10 + ], + "bp_grad_norms_F": [ + 4.058078957314137e-06, + 1.9621586133666824e-08, + 1.90722975190738e-08, + 1.9068517431719556e-08, + 1.907839575210346e-08 + ], + "gamma_dfa": 0.11183859535958618, + "acc_eval": 0.328125, + "loss_eval": 1.9089019298553467, + "epoch": 51 + }, + { + "hidden_norms": [ + 32292.658203125, + 20829408.0, + 133566216.0, + 219264704.0, + 225938912.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.692310788343093e-08, + 5.312774109533791e-10, + 5.173871886476888e-10, + 5.171740258269608e-10, + 5.173406147918058e-10 + ], + "bp_grad_norms_F": [ + 4.047169568366371e-06, + 1.926170511978853e-08, + 1.867934962263007e-08, + 1.8675748947316606e-08, + 1.868521692927061e-08 + ], + "gamma_dfa": 0.11216733865148854, + "acc_eval": 0.330078125, + "loss_eval": 1.917893886566162, + "epoch": 52 + }, + { + "hidden_norms": [ + 32582.02734375, + 21555154.0, + 138235216.0, + 226622112.0, + 233550816.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.588336359276582e-08, + 5.294207294781472e-10, + 5.160025184913763e-10, + 5.156744475875996e-10, + 5.168100947194887e-10 + ], + "bp_grad_norms_F": [ + 3.991732683061855e-06, + 1.8877956975416055e-08, + 1.8321758332717764e-08, + 1.831830331866513e-08, + 1.832757234865312e-08 + ], + "gamma_dfa": 0.11230919507215731, + "acc_eval": 0.314453125, + "loss_eval": 1.913461685180664, + "epoch": 53 + }, + { + "hidden_norms": [ + 32871.28125, + 22246662.0, + 142493184.0, + 233885040.0, + 241005264.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.552973357860537e-08, + 5.124695112712629e-10, + 4.966513311721599e-10, + 4.966223543512172e-10, + 4.969377687125132e-10 + ], + "bp_grad_norms_F": [ + 4.012842964584706e-06, + 1.8570215587487837e-08, + 1.8001765411668202e-08, + 1.799840099181438e-08, + 1.800739823920594e-08 + ], + "gamma_dfa": 0.11296637258055853, + "acc_eval": 0.33203125, + "loss_eval": 1.9206554889678955, + "epoch": 54 + }, + { + "hidden_norms": [ + 33169.98828125, + 22902584.0, + 146822496.0, + 241191920.0, + 248487904.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.486269026912851e-08, + 4.992010138593628e-10, + 4.874242676145002e-10, + 4.874663450671335e-10, + 4.884102011715186e-10 + ], + "bp_grad_norms_F": [ + 3.938752797694178e-06, + 1.8166964821375586e-08, + 1.767327617585579e-08, + 1.7670013008341812e-08, + 1.7678777552987412e-08 + ], + "gamma_dfa": 0.11258088904469332, + "acc_eval": 0.3134765625, + "loss_eval": 1.913904070854187, + "epoch": 55 + }, + { + "hidden_norms": [ + 33456.9140625, + 23628216.0, + 151346624.0, + 248325424.0, + 255824304.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.482947239623172e-08, + 5.02439700955648e-10, + 4.890933769097217e-10, + 4.891700933207233e-10, + 4.892242166931737e-10 + ], + "bp_grad_norms_F": [ + 3.895439022016944e-06, + 1.781435265968412e-08, + 1.7326955870089478e-08, + 1.7324007117736073e-08, + 1.7332443036366385e-08 + ], + "gamma_dfa": 0.11223017568408977, + "acc_eval": 0.314453125, + "loss_eval": 1.9154590368270874, + "epoch": 56 + }, + { + "hidden_norms": [ + 33716.85546875, + 24280678.0, + 155270816.0, + 255050304.0, + 262737184.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.482827157900829e-08, + 4.958616850458952e-10, + 4.826584132366918e-10, + 4.826221089437865e-10, + 4.828137889489881e-10 + ], + "bp_grad_norms_F": [ + 3.898983777617104e-06, + 1.7600514823357116e-08, + 1.7106302152569697e-08, + 1.7103344518432095e-08, + 1.7111482009113388e-08 + ], + "gamma_dfa": 0.11287979638109391, + "acc_eval": 0.30859375, + "loss_eval": 1.9285999536514282, + "epoch": 57 + }, + { + "hidden_norms": [ + 33981.390625, + 24925848.0, + 159148752.0, + 261662432.0, + 269540768.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.396101153744894e-08, + 4.846232304345222e-10, + 4.728849534174628e-10, + 4.72891226177552e-10, + 4.725210223099907e-10 + ], + "bp_grad_norms_F": [ + 3.847520019917283e-06, + 1.7274359720431676e-08, + 1.6806101399424733e-08, + 1.6803195279635474e-08, + 1.681126526875687e-08 + ], + "gamma_dfa": 0.11251267153238587, + "acc_eval": 0.330078125, + "loss_eval": 1.913767695426941, + "epoch": 58 + }, + { + "hidden_norms": [ + 34242.87890625, + 25494510.0, + 162965072.0, + 268190960.0, + 276245600.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.276207725226413e-08, + 4.64974586611433e-10, + 4.573621481540613e-10, + 4.5720988106623395e-10, + 4.578690204759539e-10 + ], + "bp_grad_norms_F": [ + 3.846054823952727e-06, + 1.7068730429059542e-08, + 1.6637597965996065e-08, + 1.663470783341836e-08, + 1.6642776046182917e-08 + ], + "gamma_dfa": 0.11268835317605408, + "acc_eval": 0.318359375, + "loss_eval": 1.9175846576690674, + "epoch": 59 + }, + { + "hidden_norms": [ + 34498.37890625, + 26013872.0, + 166717360.0, + 274736192.0, + 282985952.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.275441760157264e-08, + 4.638447681504232e-10, + 4.554343846496778e-10, + 4.5541673210358624e-10, + 4.558452226799403e-10 + ], + "bp_grad_norms_F": [ + 3.7814804727531737e-06, + 1.6729790885960938e-08, + 1.629800472358056e-08, + 1.629528156854576e-08, + 1.630315793477166e-08 + ], + "gamma_dfa": 0.11192971542368468, + "acc_eval": 0.3125, + "loss_eval": 1.9148613214492798, + "epoch": 60 + }, + { + "hidden_norms": [ + 34741.75390625, + 26574266.0, + 170266464.0, + 280835296.0, + 289264896.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.206668328763044e-08, + 4.543688481017938e-10, + 4.4642781138470866e-10, + 4.464762171085823e-10, + 4.4638817642272954e-10 + ], + "bp_grad_norms_F": [ + 3.800365902861813e-06, + 1.6550599113429598e-08, + 1.6135599523181554e-08, + 1.6132924329781417e-08, + 1.6140655034746487e-08 + ], + "gamma_dfa": 0.11263992198291817, + "acc_eval": 0.328125, + "loss_eval": 1.9202332496643066, + "epoch": 61 + }, + { + "hidden_norms": [ + 34981.7578125, + 27132752.0, + 173824032.0, + 286969856.0, + 295549056.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.146741154429037e-08, + 4.5107212409689623e-10, + 4.4140177624107935e-10, + 4.4141457156143815e-10, + 4.4168110835407504e-10 + ], + "bp_grad_norms_F": [ + 3.7655997857655166e-06, + 1.6349241960256222e-08, + 1.5922701379622595e-08, + 1.5920058160645567e-08, + 1.592757570278991e-08 + ], + "gamma_dfa": 0.11310465578208095, + "acc_eval": 0.314453125, + "loss_eval": 1.9224812984466553, + "epoch": 62 + }, + { + "hidden_norms": [ + 35201.25, + 27683016.0, + 177040880.0, + 292723872.0, + 301473504.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.122528700165276e-08, + 4.486005733550513e-10, + 4.4048292791032395e-10, + 4.4043477198663084e-10, + 4.4109013663806707e-10 + ], + "bp_grad_norms_F": [ + 3.748537892533932e-06, + 1.6115603074240425e-08, + 1.5725172275438126e-08, + 1.572264451965566e-08, + 1.5729948898979274e-08 + ], + "gamma_dfa": 0.11326105792613816, + "acc_eval": 0.3095703125, + "loss_eval": 1.9233062267303467, + "epoch": 63 + }, + { + "hidden_norms": [ + 35399.75390625, + 28156860.0, + 180111792.0, + 298457728.0, + 307298976.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.082014974457707e-08, + 4.4495629403229486e-10, + 4.324987312731565e-10, + 4.3260212079232474e-10, + 4.3259393289751813e-10 + ], + "bp_grad_norms_F": [ + 3.685790488816565e-06, + 1.5895359695150546e-08, + 1.5473323955461638e-08, + 1.547100403342938e-08, + 1.5478214265840506e-08 + ], + "gamma_dfa": 0.11222807004196511, + "acc_eval": 0.3125, + "loss_eval": 1.9198863506317139, + "epoch": 64 + }, + { + "hidden_norms": [ + 35584.6640625, + 28666806.0, + 183522992.0, + 303940064.0, + 312955008.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.149574088316513e-08, + 4.4883607941414994e-10, + 4.353337967888393e-10, + 4.3539158389727106e-10, + 4.35594643688475e-10 + ], + "bp_grad_norms_F": [ + 3.6930814530933276e-06, + 1.5769751726679715e-08, + 1.5353194271483517e-08, + 1.535078197889561e-08, + 1.5357914051605803e-08 + ], + "gamma_dfa": 0.11321029500413715, + "acc_eval": 0.3095703125, + "loss_eval": 1.9162211418151855, + "epoch": 65 + }, + { + "hidden_norms": [ + 35768.86328125, + 29145662.0, + 186551392.0, + 308909184.0, + 318089856.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.985555245364594e-08, + 4.355726057614362e-10, + 4.2523776166980554e-10, + 4.251851093428627e-10, + 4.2568140679044575e-10 + ], + "bp_grad_norms_F": [ + 3.7230390717013506e-06, + 1.5615523096812467e-08, + 1.523139303571952e-08, + 1.522908021911462e-08, + 1.5236116368555486e-08 + ], + "gamma_dfa": 0.11393742701784504, + "acc_eval": 0.31640625, + "loss_eval": 1.9276196956634521, + "epoch": 66 + }, + { + "hidden_norms": [ + 35944.30859375, + 29545796.0, + 189503824.0, + 313822880.0, + 323180096.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.027381343505112e-08, + 4.2508935260698877e-10, + 4.156285593470699e-10, + 4.156731903126598e-10, + 4.1562600583411324e-10 + ], + "bp_grad_norms_F": [ + 3.67820075553027e-06, + 1.5450812185235918e-08, + 1.5081118576176777e-08, + 1.5078812864999236e-08, + 1.5085870330722173e-08 + ], + "gamma_dfa": 0.11295431066082529, + "acc_eval": 0.3251953125, + "loss_eval": 1.9145984649658203, + "epoch": 67 + }, + { + "hidden_norms": [ + 36107.91015625, + 29963964.0, + 192509232.0, + 318614048.0, + 328116416.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.995779955332182e-08, + 4.257695862541766e-10, + 4.1527120631101866e-10, + 4.1522696392348735e-10, + 4.1541275974665837e-10 + ], + "bp_grad_norms_F": [ + 3.6363949220685754e-06, + 1.5271682585193957e-08, + 1.4899570466297973e-08, + 1.4897324263074552e-08, + 1.4904214751254585e-08 + ], + "gamma_dfa": 0.11261070579712396, + "acc_eval": 0.322265625, + "loss_eval": 1.9148796796798706, + "epoch": 68 + }, + { + "hidden_norms": [ + 36263.38671875, + 30397186.0, + 195132224.0, + 323205344.0, + 332809856.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.94296602432587e-08, + 4.221426264106043e-10, + 4.1666703420872864e-10, + 4.166992584320184e-10, + 4.166444134146019e-10 + ], + "bp_grad_norms_F": [ + 3.677051836348255e-06, + 1.5217779036902357e-08, + 1.486149248108859e-08, + 1.4859250718757266e-08, + 1.4866162523219373e-08 + ], + "gamma_dfa": 0.11349719034842565, + "acc_eval": 0.3134765625, + "loss_eval": 1.9238262176513672, + "epoch": 69 + }, + { + "hidden_norms": [ + 36417.9921875, + 30770712.0, + 197719328.0, + 327539232.0, + 337245152.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.906716976118332e-08, + 4.1953429619212557e-10, + 4.080901727654407e-10, + 4.080701054842706e-10, + 4.0843109450072745e-10 + ], + "bp_grad_norms_F": [ + 3.6080327845411375e-06, + 1.500828084033401e-08, + 1.4656276192681617e-08, + 1.465407439837918e-08, + 1.466091692492455e-08 + ], + "gamma_dfa": 0.11264980123269197, + "acc_eval": 0.3115234375, + "loss_eval": 1.9172749519348145, + "epoch": 70 + }, + { + "hidden_norms": [ + 36554.734375, + 31127202.0, + 200166992.0, + 331766240.0, + 341638624.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.895474768953136e-08, + 4.153002663986882e-10, + 4.053070101761591e-10, + 4.0528172484677327e-10, + 4.0563402636806245e-10 + ], + "bp_grad_norms_F": [ + 3.5790299079963006e-06, + 1.4871366360580396e-08, + 1.4521869928785236e-08, + 1.4519704549798007e-08, + 1.4526481351140319e-08 + ], + "gamma_dfa": 0.1125820265888251, + "acc_eval": 0.3291015625, + "loss_eval": 1.90960693359375, + "epoch": 71 + }, + { + "hidden_norms": [ + 36685.06640625, + 31478428.0, + 202703456.0, + 335890528.0, + 345860736.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.840604527269534e-08, + 4.11831152513642e-10, + 4.0155134772845713e-10, + 4.014005239305618e-10, + 4.018080868029017e-10 + ], + "bp_grad_norms_F": [ + 3.5732464311877266e-06, + 1.4777253198872131e-08, + 1.4430981742918902e-08, + 1.4428859884674239e-08, + 1.443555053270984e-08 + ], + "gamma_dfa": 0.11229235199152754, + "acc_eval": 0.3095703125, + "loss_eval": 1.912062644958496, + "epoch": 72 + }, + { + "hidden_norms": [ + 36808.640625, + 31756406.0, + 205006400.0, + 339625888.0, + 349722048.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.845510824859957e-08, + 4.056126268192628e-10, + 3.9775802096464474e-10, + 3.9773195847914167e-10, + 3.978435358931165e-10 + ], + "bp_grad_norms_F": [ + 3.601402113417862e-06, + 1.4699119255112691e-08, + 1.434870711136682e-08, + 1.4346603904868971e-08, + 1.4353199517813664e-08 + ], + "gamma_dfa": 0.11323098125649267, + "acc_eval": 0.3271484375, + "loss_eval": 1.9194645881652832, + "epoch": 73 + }, + { + "hidden_norms": [ + 36922.67578125, + 32105574.0, + 207090144.0, + 343260640.0, + 353436544.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.8447327805643e-08, + 4.050982882475296e-10, + 3.977037865698918e-10, + 3.9765896131527256e-10, + 3.979234441953139e-10 + ], + "bp_grad_norms_F": [ + 3.603466893764562e-06, + 1.4652233204515142e-08, + 1.4306199780378392e-08, + 1.4304094797523703e-08, + 1.4310627349800598e-08 + ], + "gamma_dfa": 0.11329636031587142, + "acc_eval": 0.3056640625, + "loss_eval": 1.922640323638916, + "epoch": 74 + }, + { + "hidden_norms": [ + 37024.05859375, + 32409028.0, + 208845456.0, + 346325312.0, + 356587104.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.819700359983472e-08, + 4.0326031403026263e-10, + 3.9583156197231517e-10, + 3.9586384170675615e-10, + 3.9581374289276994e-10 + ], + "bp_grad_norms_F": [ + 3.5432237837085268e-06, + 1.4486117194678627e-08, + 1.4156262828635136e-08, + 1.4154186267489877e-08, + 1.4160684180808403e-08 + ], + "gamma_dfa": 0.11267639175457589, + "acc_eval": 0.31640625, + "loss_eval": 1.9154051542282104, + "epoch": 75 + }, + { + "hidden_norms": [ + 37125.02734375, + 32663892.0, + 210781344.0, + 349533728.0, + 359873568.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.772787486397647e-08, + 3.9591485645473767e-10, + 3.875866017022389e-10, + 3.8754283115949306e-10, + 3.8766378995802597e-10 + ], + "bp_grad_norms_F": [ + 3.5503553590388037e-06, + 1.4420066918319208e-08, + 1.40893732236691e-08, + 1.4087291333453322e-08, + 1.4093810563053921e-08 + ], + "gamma_dfa": 0.11297615164312447, + "acc_eval": 0.32421875, + "loss_eval": 1.913941502571106, + "epoch": 76 + }, + { + "hidden_norms": [ + 37216.296875, + 32930840.0, + 212456432.0, + 352475008.0, + 362884992.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.74248425980295e-08, + 3.9426073517034865e-10, + 3.872507869928654e-10, + 3.8720734951702696e-10, + 3.8765768373139053e-10 + ], + "bp_grad_norms_F": [ + 3.55011752617429e-06, + 1.4357588895563822e-08, + 1.402936344874206e-08, + 1.402732330291201e-08, + 1.4033815887160017e-08 + ], + "gamma_dfa": 0.11311322844358074, + "acc_eval": 0.3203125, + "loss_eval": 1.914380431175232, + "epoch": 77 + }, + { + "hidden_norms": [ + 37298.37890625, + 33147848.0, + 214118832.0, + 355022880.0, + 365529568.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.774458682912154e-08, + 3.987353780487979e-10, + 3.904769285689724e-10, + 3.904657153164237e-10, + 3.907260348601227e-10 + ], + "bp_grad_norms_F": [ + 3.509487669361988e-06, + 1.4249152968659473e-08, + 1.3931584774695693e-08, + 1.3929625453101835e-08, + 1.3936028331329453e-08 + ], + "gamma_dfa": 0.11238172389857937, + "acc_eval": 0.314453125, + "loss_eval": 1.9106721878051758, + "epoch": 78 + }, + { + "hidden_norms": [ + 37372.10546875, + 33388374.0, + 215761392.0, + 357596800.0, + 368190304.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.751073300392818e-08, + 3.9374201121766816e-10, + 3.8655822987010424e-10, + 3.866116038420131e-10, + 3.8653891198947576e-10 + ], + "bp_grad_norms_F": [ + 3.537302973199985e-06, + 1.4222394817409167e-08, + 1.3910858243093571e-08, + 1.3908881157931319e-08, + 1.3915280483445258e-08 + ], + "gamma_dfa": 0.11306001050525083, + "acc_eval": 0.322265625, + "loss_eval": 1.9138463735580444, + "epoch": 79 + }, + { + "hidden_norms": [ + 37442.84375, + 33566284.0, + 216908128.0, + 359755968.0, + 370379136.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.704155453007843e-08, + 3.923593672183756e-10, + 3.850708640840139e-10, + 3.8509884170423447e-10, + 3.849042473635933e-10 + ], + "bp_grad_norms_F": [ + 3.509628641040763e-06, + 1.4144676541150147e-08, + 1.3834353218555862e-08, + 1.3832397449675682e-08, + 1.3838723056380786e-08 + ], + "gamma_dfa": 0.11266809623703011, + "acc_eval": 0.3251953125, + "loss_eval": 1.9121947288513184, + "epoch": 80 + }, + { + "hidden_norms": [ + 37512.11328125, + 33717608.0, + 218115872.0, + 361736960.0, + 372400416.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.701832688804643e-08, + 3.917752788851203e-10, + 3.835655681960759e-10, + 3.835243511662867e-10, + 3.835320394607322e-10 + ], + "bp_grad_norms_F": [ + 3.5144942103215726e-06, + 1.4111047441645042e-08, + 1.3798888254257236e-08, + 1.3796943143518092e-08, + 1.3803241216692186e-08 + ], + "gamma_dfa": 0.11281233225781762, + "acc_eval": 0.3193359375, + "loss_eval": 1.9152591228485107, + "epoch": 81 + }, + { + "hidden_norms": [ + 37568.84375, + 33899664.0, + 219258560.0, + 363709504.0, + 374424128.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.695909604559347e-08, + 3.8868031015937277e-10, + 3.823757421805851e-10, + 3.8241659838789133e-10, + 3.8222841558521736e-10 + ], + "bp_grad_norms_F": [ + 3.516150854920852e-06, + 1.40706539752955e-08, + 1.3756586092483758e-08, + 1.375462410635464e-08, + 1.3760891093284044e-08 + ], + "gamma_dfa": 0.11307295318874822, + "acc_eval": 0.326171875, + "loss_eval": 1.9149377346038818, + "epoch": 82 + }, + { + "hidden_norms": [ + 37613.421875, + 34060644.0, + 220322080.0, + 365325472.0, + 376109056.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.71154296583154e-08, + 3.8974332094987574e-10, + 3.834943751446218e-10, + 3.8340422503502225e-10, + 3.8375369548759863e-10 + ], + "bp_grad_norms_F": [ + 3.4982940633199178e-06, + 1.4013506799415154e-08, + 1.370163893454901e-08, + 1.3699707146486162e-08, + 1.3705942158992457e-08 + ], + "gamma_dfa": 0.1128659905607492, + "acc_eval": 0.3212890625, + "loss_eval": 1.912517786026001, + "epoch": 83 + }, + { + "hidden_norms": [ + 37660.46875, + 34177292.0, + 221168784.0, + 366778016.0, + 377600544.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.68943869186478e-08, + 3.894674860394076e-10, + 3.8076572450584933e-10, + 3.8073258434856427e-10, + 3.8100084198688933e-10 + ], + "bp_grad_norms_F": [ + 3.4944750950671732e-06, + 1.397998961039093e-08, + 1.3671985321650482e-08, + 1.3670069520799188e-08, + 1.3676272558882374e-08 + ], + "gamma_dfa": 0.11273636969053769, + "acc_eval": 0.318359375, + "loss_eval": 1.9171113967895508, + "epoch": 84 + }, + { + "hidden_norms": [ + 37701.8671875, + 34327872.0, + 222020640.0, + 368143552.0, + 379001216.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.639802307972786e-08, + 3.87883669628053e-10, + 3.8044753458699176e-10, + 3.804264403495239e-10, + 3.8037359373355173e-10 + ], + "bp_grad_norms_F": [ + 3.4919537483801832e-06, + 1.3943878052202763e-08, + 1.3639727569625393e-08, + 1.3637805551525162e-08, + 1.364400148418099e-08 + ], + "gamma_dfa": 0.11286781356830033, + "acc_eval": 0.328125, + "loss_eval": 1.9123945236206055, + "epoch": 85 + }, + { + "hidden_norms": [ + 37739.8125, + 34429904.0, + 222737072.0, + 369295712.0, + 380185984.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.741821323430486e-08, + 3.8846115213431176e-10, + 3.8069486452130263e-10, + 3.8060576912357647e-10, + 3.809500215279371e-10 + ], + "bp_grad_norms_F": [ + 3.500515958876349e-06, + 1.3945418153582523e-08, + 1.364027113481825e-08, + 1.3638369544821671e-08, + 1.364453350305439e-08 + ], + "gamma_dfa": 0.11293645015939546, + "acc_eval": 0.3154296875, + "loss_eval": 1.917877197265625, + "epoch": 86 + }, + { + "hidden_norms": [ + 37773.1484375, + 34523684.0, + 223309680.0, + 370332384.0, + 381249280.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.685363729275196e-08, + 3.8528755186284513e-10, + 3.7927344598287505e-10, + 3.7934516639026583e-10, + 3.7926364826468273e-10 + ], + "bp_grad_norms_F": [ + 3.4931999834952876e-06, + 1.3911479967987361e-08, + 1.360801249461474e-08, + 1.3606110904618163e-08, + 1.3612289961884017e-08 + ], + "gamma_dfa": 0.11282484548610228, + "acc_eval": 0.322265625, + "loss_eval": 1.9136494398117065, + "epoch": 87 + }, + { + "hidden_norms": [ + 37800.29296875, + 34600436.0, + 223845264.0, + 371228288.0, + 382171744.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.65767956320451e-08, + 3.853606878045923e-10, + 3.7911099259879677e-10, + 3.789283054000947e-10, + 3.791919833684432e-10 + ], + "bp_grad_norms_F": [ + 3.4772997423715424e-06, + 1.3876966242776234e-08, + 1.3575736979021258e-08, + 1.3573847823522556e-08, + 1.3580010893576855e-08 + ], + "gamma_dfa": 0.11266115616490424, + "acc_eval": 0.318359375, + "loss_eval": 1.912358045578003, + "epoch": 88 + }, + { + "hidden_norms": [ + 37822.953125, + 34659448.0, + 224278352.0, + 371994688.0, + 382956800.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.695878340678973e-08, + 3.8400196911148043e-10, + 3.7754030457470833e-10, + 3.775957047036371e-10, + 3.7752789783240814e-10 + ], + "bp_grad_norms_F": [ + 3.488719130473328e-06, + 1.387363113281026e-08, + 1.3571511026100325e-08, + 1.3569623646958462e-08, + 1.3575775170693305e-08 + ], + "gamma_dfa": 0.11286654848299804, + "acc_eval": 0.318359375, + "loss_eval": 1.9152348041534424, + "epoch": 89 + }, + { + "hidden_norms": [ + 37843.71484375, + 34715836.0, + 224633712.0, + 372602496.0, + 383583840.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.682009967562408e-08, + 3.831650552399424e-10, + 3.7603772873318064e-10, + 3.7600736413345714e-10, + 3.763372669052245e-10 + ], + "bp_grad_norms_F": [ + 3.4864976896642474e-06, + 1.3852580416084948e-08, + 1.355203504971314e-08, + 1.3550148558749697e-08, + 1.3556297417949281e-08 + ], + "gamma_dfa": 0.11291119743145828, + "acc_eval": 0.326171875, + "loss_eval": 1.914948582649231, + "epoch": 90 + }, + { + "hidden_norms": [ + 37859.484375, + 34761876.0, + 224938096.0, + 373110784.0, + 384106720.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.677319674963655e-08, + 3.8410433167435087e-10, + 3.768680645332978e-10, + 3.768474976517666e-10, + 3.767945122579164e-10 + ], + "bp_grad_norms_F": [ + 3.480063014649204e-06, + 1.3833711953736838e-08, + 1.3534608989118624e-08, + 1.3532734044474637e-08, + 1.3538858922856889e-08 + ], + "gamma_dfa": 0.1127637956833496, + "acc_eval": 0.322265625, + "loss_eval": 1.9153952598571777, + "epoch": 91 + }, + { + "hidden_norms": [ + 37872.0625, + 34797108.0, + 225202544.0, + 373570400.0, + 384579200.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.67909034746117e-08, + 3.821318816932262e-10, + 3.753052590926842e-10, + 3.7535435870594824e-10, + 3.7532754681990355e-10 + ], + "bp_grad_norms_F": [ + 3.4843556022678968e-06, + 1.3827068379157481e-08, + 1.352985190550271e-08, + 1.3527971631788205e-08, + 1.3534107168311493e-08 + ], + "gamma_dfa": 0.1129197701375233, + "acc_eval": 0.318359375, + "loss_eval": 1.9156945943832397, + "epoch": 92 + }, + { + "hidden_norms": [ + 37881.54296875, + 34826116.0, + 225403072.0, + 373889952.0, + 384906848.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.663099583192889e-08, + 3.844494445015556e-10, + 3.765425748980533e-10, + 3.7651176620911997e-10, + 3.7661743168548867e-10 + ], + "bp_grad_norms_F": [ + 3.480702616798226e-06, + 1.382177039488397e-08, + 1.3523890451949683e-08, + 1.3522016395484115e-08, + 1.3528141273866368e-08 + ], + "gamma_dfa": 0.11285195982236473, + "acc_eval": 0.322265625, + "loss_eval": 1.9148646593093872, + "epoch": 93 + }, + { + "hidden_norms": [ + 37889.18359375, + 34847784.0, + 225551344.0, + 374143840.0, + 385168448.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.623226767032975e-08, + 3.819731198007048e-10, + 3.7512581929632915e-10, + 3.752123889366743e-10, + 3.7511962980296687e-10 + ], + "bp_grad_norms_F": [ + 3.4782924558385275e-06, + 1.3813751031932497e-08, + 1.3516644692401769e-08, + 1.3514763530508844e-08, + 1.3520896402496874e-08 + ], + "gamma_dfa": 0.11281965267244232, + "acc_eval": 0.318359375, + "loss_eval": 1.9142093658447266, + "epoch": 94 + }, + { + "hidden_norms": [ + 37895.046875, + 34865808.0, + 225670928.0, + 374345440.0, + 385375392.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.64825137164371e-08, + 3.829765116147854e-10, + 3.7528671836817296e-10, + 3.7535685670775365e-10, + 3.752768373832538e-10 + ], + "bp_grad_norms_F": [ + 3.4792733458743896e-06, + 1.3809607679604596e-08, + 1.3512363672418815e-08, + 1.3510485175061149e-08, + 1.351661271797866e-08 + ], + "gamma_dfa": 0.11288163481003721, + "acc_eval": 0.32421875, + "loss_eval": 1.9146695137023926, + "epoch": 95 + }, + { + "hidden_norms": [ + 37898.87109375, + 34878456.0, + 225750400.0, + 374473248.0, + 385506496.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.652722817079848e-08, + 3.8302755411834255e-10, + 3.759152988891401e-10, + 3.7595143664859165e-10, + 3.7588568368995823e-10 + ], + "bp_grad_norms_F": [ + 3.480078930806485e-06, + 1.3808159060602065e-08, + 1.3511034957502943e-08, + 1.3509157348323697e-08, + 1.351528045034911e-08 + ], + "gamma_dfa": 0.1128929610213163, + "acc_eval": 0.3203125, + "loss_eval": 1.9147298336029053, + "epoch": 96 + }, + { + "hidden_norms": [ + 37900.7109375, + 34885860.0, + 225787696.0, + 374537472.0, + 385572896.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.630104820715133e-08, + 3.8308795025088216e-10, + 3.7579228617801164e-10, + 3.7582850720419003e-10, + 3.7576297629016153e-10 + ], + "bp_grad_norms_F": [ + 3.4770955608109944e-06, + 1.3803015619373582e-08, + 1.3505847995531894e-08, + 1.3503973050887907e-08, + 1.3510093488378061e-08 + ], + "gamma_dfa": 0.11285188281362935, + "acc_eval": 0.3212890625, + "loss_eval": 1.9143996238708496, + "epoch": 97 + }, + { + "hidden_norms": [ + 37902.46875, + 34890740.0, + 225820192.0, + 374593792.0, + 385630976.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.624973991620209e-08, + 3.8267702895389277e-10, + 3.7545791475857015e-10, + 3.7555664134103495e-10, + 3.75430270205257e-10 + ], + "bp_grad_norms_F": [ + 3.476953224890167e-06, + 1.3801756182374447e-08, + 1.350452816240022e-08, + 1.3502653217756233e-08, + 1.3508774543424806e-08 + ], + "gamma_dfa": 0.11284657929081732, + "acc_eval": 0.3203125, + "loss_eval": 1.9143402576446533, + "epoch": 98 + }, + { + "hidden_norms": [ + 37903.22265625, + 34893256.0, + 225836736.0, + 374620576.0, + 385658624.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.627380599866228e-08, + 3.8272779390169376e-10, + 3.7550038078926207e-10, + 3.7559910737172686e-10, + 3.7547268072479767e-10 + ], + "bp_grad_norms_F": [ + 3.477006657703896e-06, + 1.3801344067587706e-08, + 1.3504034335198867e-08, + 1.350215939055488e-08, + 1.3508279828045033e-08 + ], + "gamma_dfa": 0.11284785682073561, + "acc_eval": 0.3203125, + "loss_eval": 1.9143691062927246, + "epoch": 99 + }, + { + "hidden_norms": [ + 37903.25, + 34893376.0, + 225837280.0, + 374621408.0, + 385659456.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.626758874972438e-08, + 3.826984007471168e-10, + 3.755481758904722e-10, + 3.756469302285126e-10, + 3.755204758260078e-10 + ], + "bp_grad_norms_F": [ + 3.476960500847781e-06, + 1.3801311204986177e-08, + 1.3503997919883659e-08, + 1.3502122975239672e-08, + 1.3508243412729826e-08 + ], + "gamma_dfa": 0.11284651984533411, + "acc_eval": 0.3203125, + "loss_eval": 1.9143500328063965, + "epoch": 100 + } + ] +}
\ No newline at end of file diff --git a/results/snapshot_evolution_v2/snapshot_evolution_s42.json b/results/snapshot_evolution_v2/snapshot_evolution_s42.json new file mode 100644 index 0000000..fb4a73e --- /dev/null +++ b/results/snapshot_evolution_v2/snapshot_evolution_s42.json @@ -0,0 +1,5473 @@ +{ + "config": { + "output_dir": "results/snapshot_evolution_v2", + "epochs": 100, + "lr": 0.001, + "wd": 0.01, + "seed": 42, + "depth": 4, + "d_hidden": 256, + "log_every": 1 + }, + "depth": 4, + "d_hidden": 256, + "num_classes": 10, + "bp_log": [ + { + "hidden_norms": [ + 8.718070030212402, + 8.780964851379395, + 8.812002182006836, + 8.828720092773438, + 8.893179893493652 + ], + "bp_grad_norms_per_sample_med": [ + 0.0010068593546748161, + 0.000997236929833889, + 0.0009934091940522194, + 0.0009870644425973296, + 0.0009829498594626784 + ], + "bp_grad_norms_F": [ + 0.03544403985142708, + 0.03507080301642418, + 0.03477022796869278, + 0.03443896397948265, + 0.03407429903745651 + ], + "gamma_dfa": NaN, + "acc_eval": 0.115234375, + "loss_eval": 2.383650302886963, + "epoch": 0 + }, + { + "hidden_norms": [ + 77.06888580322266, + 79.12034606933594, + 82.7389144897461, + 87.31297302246094, + 93.80765533447266 + ], + "bp_grad_norms_per_sample_med": [ + 5.941955532762222e-05, + 5.619435978587717e-05, + 5.499917097040452e-05, + 5.5899512517498806e-05, + 5.7389137509744614e-05 + ], + "bp_grad_norms_F": [ + 0.0026216788683086634, + 0.0023027872666716576, + 0.002111479640007019, + 0.002080164849758148, + 0.0021061261650174856 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3798828125, + "loss_eval": 1.6751981973648071, + "epoch": 1 + }, + { + "hidden_norms": [ + 98.86981201171875, + 98.9476089477539, + 102.55016326904297, + 106.3559341430664, + 112.6939697265625 + ], + "bp_grad_norms_per_sample_med": [ + 4.658113539335318e-05, + 4.402571721584536e-05, + 4.2862800910370424e-05, + 4.381770850159228e-05, + 4.3870313675142825e-05 + ], + "bp_grad_norms_F": [ + 0.00196655560284853, + 0.0018041934818029404, + 0.0016866555670276284, + 0.0016513779992237687, + 0.0016559252981096506 + ], + "gamma_dfa": NaN, + "acc_eval": 0.4541015625, + "loss_eval": 1.5228025913238525, + "epoch": 2 + }, + { + "hidden_norms": [ + 110.40647888183594, + 112.07202911376953, + 114.70049285888672, + 117.17343139648438, + 124.54205322265625 + ], + "bp_grad_norms_per_sample_med": [ + 4.4490061554824933e-05, + 4.3751915654866025e-05, + 4.270448334864341e-05, + 4.182140401098877e-05, + 4.151134271523915e-05 + ], + "bp_grad_norms_F": [ + 0.0019047901732847095, + 0.0017693191766738892, + 0.0016821600729599595, + 0.0016251102788373828, + 0.0015800207620486617 + ], + "gamma_dfa": NaN, + "acc_eval": 0.4697265625, + "loss_eval": 1.4416429996490479, + "epoch": 3 + }, + { + "hidden_norms": [ + 125.09136199951172, + 126.65565490722656, + 128.4573974609375, + 130.39990234375, + 138.1091766357422 + ], + "bp_grad_norms_per_sample_med": [ + 4.1048842831514776e-05, + 4.0168823034036905e-05, + 3.916600326192565e-05, + 3.953508348786272e-05, + 3.938759255106561e-05 + ], + "bp_grad_norms_F": [ + 0.0017511667683720589, + 0.001639657886698842, + 0.0015659942291676998, + 0.0015107291983440518, + 0.0014574137749150395 + ], + "gamma_dfa": NaN, + "acc_eval": 0.50390625, + "loss_eval": 1.3718887567520142, + "epoch": 4 + }, + { + "hidden_norms": [ + 139.76705932617188, + 139.05606079101562, + 139.5422821044922, + 141.0078582763672, + 148.9493865966797 + ], + "bp_grad_norms_per_sample_med": [ + 3.957379158237018e-05, + 4.0101109334500507e-05, + 3.9158370782388374e-05, + 3.927717261831276e-05, + 3.796629243879579e-05 + ], + "bp_grad_norms_F": [ + 0.0017042789841070771, + 0.001630594371818006, + 0.0015700262738391757, + 0.0015037605771794915, + 0.0014294973807409406 + ], + "gamma_dfa": NaN, + "acc_eval": 0.498046875, + "loss_eval": 1.367867112159729, + "epoch": 5 + }, + { + "hidden_norms": [ + 147.73207092285156, + 148.01638793945312, + 148.1798553466797, + 149.07754516601562, + 156.61810302734375 + ], + "bp_grad_norms_per_sample_med": [ + 3.987981835962273e-05, + 3.9670325350016356e-05, + 3.970034231315367e-05, + 3.8273319660220295e-05, + 3.6929654015693814e-05 + ], + "bp_grad_norms_F": [ + 0.0017687880899757147, + 0.001693942816928029, + 0.0016310750506818295, + 0.0015193652361631393, + 0.0014129421906545758 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5234375, + "loss_eval": 1.3461639881134033, + "epoch": 6 + }, + { + "hidden_norms": [ + 159.30731201171875, + 158.4180450439453, + 158.36903381347656, + 158.61819458007812, + 165.9392547607422 + ], + "bp_grad_norms_per_sample_med": [ + 3.744649075088091e-05, + 3.805281448876485e-05, + 3.727989678736776e-05, + 3.5814937291434035e-05, + 3.422912777750753e-05 + ], + "bp_grad_norms_F": [ + 0.0017357978504151106, + 0.0016663926653563976, + 0.0016097063198685646, + 0.0014966910239309072, + 0.001379394088871777 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5283203125, + "loss_eval": 1.3227663040161133, + "epoch": 7 + }, + { + "hidden_norms": [ + 171.02230834960938, + 169.76348876953125, + 168.44949340820312, + 168.26429748535156, + 176.89520263671875 + ], + "bp_grad_norms_per_sample_med": [ + 3.716369246831164e-05, + 3.8349648093571886e-05, + 3.801397542702034e-05, + 3.6082456063013524e-05, + 3.359258334967308e-05 + ], + "bp_grad_norms_F": [ + 0.0016445739893242717, + 0.001604317338205874, + 0.0015634173760190606, + 0.0014586447505280375, + 0.0013399318559095263 + ], + "gamma_dfa": NaN, + "acc_eval": 0.541015625, + "loss_eval": 1.2933616638183594, + "epoch": 8 + }, + { + "hidden_norms": [ + 177.41796875, + 177.29603576660156, + 175.89486694335938, + 175.0777587890625, + 181.38400268554688 + ], + "bp_grad_norms_per_sample_med": [ + 3.602267679525539e-05, + 3.6796918720938265e-05, + 3.648500933195464e-05, + 3.566688246792182e-05, + 3.4070406400132924e-05 + ], + "bp_grad_norms_F": [ + 0.0015903833555057645, + 0.0015696624759584665, + 0.0015270623844116926, + 0.0014402325032278895, + 0.0013178731314837933 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5478515625, + "loss_eval": 1.2806801795959473, + "epoch": 9 + }, + { + "hidden_norms": [ + 187.4056854248047, + 186.72698974609375, + 186.21583557128906, + 185.0518035888672, + 193.2371063232422 + ], + "bp_grad_norms_per_sample_med": [ + 3.3276784961344674e-05, + 3.436602492001839e-05, + 3.432494486332871e-05, + 3.371347702341154e-05, + 3.1834206311032176e-05 + ], + "bp_grad_norms_F": [ + 0.0014970082556828856, + 0.0014613966923207045, + 0.001435543643310666, + 0.001374539453536272, + 0.0012561712646856904 + ], + "gamma_dfa": NaN, + "acc_eval": 0.56640625, + "loss_eval": 1.2740942239761353, + "epoch": 10 + }, + { + "hidden_norms": [ + 194.4965057373047, + 191.25531005859375, + 189.80703735351562, + 188.18850708007812, + 195.7776336669922 + ], + "bp_grad_norms_per_sample_med": [ + 3.298965748399496e-05, + 3.40789083566051e-05, + 3.381206624908373e-05, + 3.3352327591273934e-05, + 3.1442876206710935e-05 + ], + "bp_grad_norms_F": [ + 0.0015152172418311238, + 0.001501554623246193, + 0.0014814225723966956, + 0.001403055852279067, + 0.001272345893085003 + ], + "gamma_dfa": NaN, + "acc_eval": 0.560546875, + "loss_eval": 1.265188455581665, + "epoch": 11 + }, + { + "hidden_norms": [ + 200.5504913330078, + 197.28366088867188, + 195.27011108398438, + 192.64669799804688, + 200.8507843017578 + ], + "bp_grad_norms_per_sample_med": [ + 3.453825775068253e-05, + 3.5414250305620953e-05, + 3.575523442123085e-05, + 3.455601472523995e-05, + 3.2858857593964785e-05 + ], + "bp_grad_norms_F": [ + 0.001518857548944652, + 0.001502488274127245, + 0.0014838415663689375, + 0.001404650043696165, + 0.0012748484732583165 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5458984375, + "loss_eval": 1.2733395099639893, + "epoch": 12 + }, + { + "hidden_norms": [ + 207.21469116210938, + 202.24969482421875, + 198.8241424560547, + 197.5497283935547, + 207.32147216796875 + ], + "bp_grad_norms_per_sample_med": [ + 3.2123080018209293e-05, + 3.345572258695029e-05, + 3.3888838515849784e-05, + 3.344774086144753e-05, + 3.095310239586979e-05 + ], + "bp_grad_norms_F": [ + 0.0014620382571592927, + 0.0014708929229527712, + 0.0014657732099294662, + 0.0013834136771038175, + 0.0012381445849314332 + ], + "gamma_dfa": NaN, + "acc_eval": 0.56640625, + "loss_eval": 1.2357412576675415, + "epoch": 13 + }, + { + "hidden_norms": [ + 214.6268310546875, + 208.3655242919922, + 205.5032501220703, + 202.8442840576172, + 211.64646911621094 + ], + "bp_grad_norms_per_sample_med": [ + 3.207657573511824e-05, + 3.362155257491395e-05, + 3.38435493176803e-05, + 3.330525942146778e-05, + 3.142434434266761e-05 + ], + "bp_grad_norms_F": [ + 0.0014900992391631007, + 0.0015044353203848004, + 0.001489310641773045, + 0.0014063691487535834, + 0.0012448005145415664 + ], + "gamma_dfa": NaN, + "acc_eval": 0.56640625, + "loss_eval": 1.2451550960540771, + "epoch": 14 + }, + { + "hidden_norms": [ + 220.40109252929688, + 213.53439331054688, + 209.3384552001953, + 206.62062072753906, + 213.82391357421875 + ], + "bp_grad_norms_per_sample_med": [ + 3.227095294278115e-05, + 3.415472019696608e-05, + 3.4918764868052676e-05, + 3.409514465602115e-05, + 3.082074908888899e-05 + ], + "bp_grad_norms_F": [ + 0.0015189005061984062, + 0.001544533297419548, + 0.0015320393722504377, + 0.0014468894805759192, + 0.0012609416153281927 + ], + "gamma_dfa": NaN, + "acc_eval": 0.564453125, + "loss_eval": 1.248854398727417, + "epoch": 15 + }, + { + "hidden_norms": [ + 226.32350158691406, + 219.70506286621094, + 213.84132385253906, + 210.67050170898438, + 219.11776733398438 + ], + "bp_grad_norms_per_sample_med": [ + 3.360298796906136e-05, + 3.4720324038062245e-05, + 3.473780452623032e-05, + 3.412307341932319e-05, + 3.0327399144880474e-05 + ], + "bp_grad_norms_F": [ + 0.0014919604873284698, + 0.0015073713148012757, + 0.001499581034295261, + 0.001411686884239316, + 0.0012323390692472458 + ], + "gamma_dfa": NaN, + "acc_eval": 0.57421875, + "loss_eval": 1.2207037210464478, + "epoch": 16 + }, + { + "hidden_norms": [ + 230.78538513183594, + 223.432373046875, + 216.73037719726562, + 212.86383056640625, + 220.8701171875 + ], + "bp_grad_norms_per_sample_med": [ + 3.253064278396778e-05, + 3.3939610148081556e-05, + 3.490943709039129e-05, + 3.4084565413650125e-05, + 3.0307599445222877e-05 + ], + "bp_grad_norms_F": [ + 0.0015068502398207784, + 0.0015421892749145627, + 0.0015326113207265735, + 0.0014369196724146605, + 0.0012492147507146 + ], + "gamma_dfa": NaN, + "acc_eval": 0.560546875, + "loss_eval": 1.216906189918518, + "epoch": 17 + }, + { + "hidden_norms": [ + 234.5084991455078, + 225.73875427246094, + 219.08822631835938, + 214.42320251464844, + 222.92369079589844 + ], + "bp_grad_norms_per_sample_med": [ + 3.451363227213733e-05, + 3.6047003959538415e-05, + 3.654924512375146e-05, + 3.617065158323385e-05, + 3.275999551988207e-05 + ], + "bp_grad_norms_F": [ + 0.001566705177538097, + 0.0015968933003023267, + 0.0015981856267899275, + 0.0014931216137483716, + 0.0012800253462046385 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5556640625, + "loss_eval": 1.2483153343200684, + "epoch": 18 + }, + { + "hidden_norms": [ + 238.0281219482422, + 229.677734375, + 223.0265350341797, + 217.872314453125, + 225.45947265625 + ], + "bp_grad_norms_per_sample_med": [ + 3.4922632039524615e-05, + 3.6500507121672854e-05, + 3.762893902603537e-05, + 3.615105015342124e-05, + 3.2680742151569575e-05 + ], + "bp_grad_norms_F": [ + 0.0015319561352953315, + 0.001577642629854381, + 0.001574998372234404, + 0.0014754259027540684, + 0.0012637749314308167 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5576171875, + "loss_eval": 1.2469550371170044, + "epoch": 19 + }, + { + "hidden_norms": [ + 241.45298767089844, + 233.28753662109375, + 226.9265899658203, + 222.43453979492188, + 229.83859252929688 + ], + "bp_grad_norms_per_sample_med": [ + 3.368386751390062e-05, + 3.570445187506266e-05, + 3.622082658694126e-05, + 3.5024619137402624e-05, + 3.180970816174522e-05 + ], + "bp_grad_norms_F": [ + 0.0015120706520974636, + 0.0015433274675160646, + 0.0015402185963466763, + 0.0014465245185419917, + 0.001252179266884923 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5634765625, + "loss_eval": 1.2113828659057617, + "epoch": 20 + }, + { + "hidden_norms": [ + 246.53602600097656, + 236.25527954101562, + 231.07940673828125, + 227.7357940673828, + 235.99017333984375 + ], + "bp_grad_norms_per_sample_med": [ + 3.2015348551794887e-05, + 3.388321783859283e-05, + 3.521092003211379e-05, + 3.46749002346769e-05, + 3.079533780692145e-05 + ], + "bp_grad_norms_F": [ + 0.0014757646713405848, + 0.001517578144557774, + 0.0015156606677919626, + 0.00144051609095186, + 0.0012491638772189617 + ], + "gamma_dfa": NaN, + "acc_eval": 0.576171875, + "loss_eval": 1.2042850255966187, + "epoch": 21 + }, + { + "hidden_norms": [ + 250.74209594726562, + 241.11993408203125, + 232.76829528808594, + 228.410400390625, + 237.100341796875 + ], + "bp_grad_norms_per_sample_med": [ + 3.173810910084285e-05, + 3.328004459035583e-05, + 3.405071402085014e-05, + 3.34872274834197e-05, + 3.07901827909518e-05 + ], + "bp_grad_norms_F": [ + 0.0014801323413848877, + 0.0015176001470535994, + 0.0015198889886960387, + 0.0014302636263892055, + 0.001232756651006639 + ], + "gamma_dfa": NaN, + "acc_eval": 0.587890625, + "loss_eval": 1.1819555759429932, + "epoch": 22 + }, + { + "hidden_norms": [ + 251.96310424804688, + 242.98260498046875, + 235.83056640625, + 229.86978149414062, + 238.271240234375 + ], + "bp_grad_norms_per_sample_med": [ + 3.078249574173242e-05, + 3.301461765659042e-05, + 3.3559550502104685e-05, + 3.2682331948308274e-05, + 2.9418402846204117e-05 + ], + "bp_grad_norms_F": [ + 0.00149905972648412, + 0.0015357647789642215, + 0.0015457755653187633, + 0.0014610268408432603, + 0.0012446421897038817 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5732421875, + "loss_eval": 1.1770297288894653, + "epoch": 23 + }, + { + "hidden_norms": [ + 255.878662109375, + 244.79376220703125, + 236.5006561279297, + 233.29434204101562, + 239.50732421875 + ], + "bp_grad_norms_per_sample_med": [ + 3.191530049662106e-05, + 3.370656486367807e-05, + 3.539594035828486e-05, + 3.486185596557334e-05, + 3.131559424218722e-05 + ], + "bp_grad_norms_F": [ + 0.001507170731201768, + 0.0015490618534386158, + 0.0015714912442490458, + 0.0014840371441096067, + 0.0012607484823092818 + ], + "gamma_dfa": NaN, + "acc_eval": 0.572265625, + "loss_eval": 1.2011631727218628, + "epoch": 24 + }, + { + "hidden_norms": [ + 256.66448974609375, + 244.18772888183594, + 236.46058654785156, + 231.88583374023438, + 238.06246948242188 + ], + "bp_grad_norms_per_sample_med": [ + 3.2962809200398624e-05, + 3.441200897214003e-05, + 3.570063927327283e-05, + 3.488562651909888e-05, + 3.0770879675401375e-05 + ], + "bp_grad_norms_F": [ + 0.0015825299778953195, + 0.0016350955702364445, + 0.0016404123743996024, + 0.0015316286589950323, + 0.0012829666957259178 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5703125, + "loss_eval": 1.192138433456421, + "epoch": 25 + }, + { + "hidden_norms": [ + 259.5942687988281, + 247.43173217773438, + 239.85533142089844, + 233.21835327148438, + 241.53346252441406 + ], + "bp_grad_norms_per_sample_med": [ + 3.1691903132013977e-05, + 3.34988217218779e-05, + 3.45351618307177e-05, + 3.441058288444765e-05, + 3.0752617021789774e-05 + ], + "bp_grad_norms_F": [ + 0.0015797498635947704, + 0.0016322932206094265, + 0.001641901326365769, + 0.0015325341373682022, + 0.0012732355389744043 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5791015625, + "loss_eval": 1.1775751113891602, + "epoch": 26 + }, + { + "hidden_norms": [ + 260.6902770996094, + 249.26528930664062, + 240.86239624023438, + 235.91477966308594, + 242.1553497314453 + ], + "bp_grad_norms_per_sample_med": [ + 3.32598174281884e-05, + 3.471180752967484e-05, + 3.609888517530635e-05, + 3.496996214380488e-05, + 3.091244798270054e-05 + ], + "bp_grad_norms_F": [ + 0.0015408779727295041, + 0.001589508610777557, + 0.0016157848294824362, + 0.0015165223740041256, + 0.0012689991854131222 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5859375, + "loss_eval": 1.1941479444503784, + "epoch": 27 + }, + { + "hidden_norms": [ + 265.0726623535156, + 253.6020050048828, + 244.876953125, + 239.7056427001953, + 248.2157440185547 + ], + "bp_grad_norms_per_sample_med": [ + 3.255613773944788e-05, + 3.457435013842769e-05, + 3.5964570997748524e-05, + 3.529110108502209e-05, + 3.089939127676189e-05 + ], + "bp_grad_norms_F": [ + 0.001513357157818973, + 0.0015594394644722342, + 0.0015772230690345168, + 0.001489063142798841, + 0.0012491336092352867 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5810546875, + "loss_eval": 1.1790356636047363, + "epoch": 28 + }, + { + "hidden_norms": [ + 266.9394836425781, + 253.8067169189453, + 243.13002014160156, + 236.81365966796875, + 245.42434692382812 + ], + "bp_grad_norms_per_sample_med": [ + 3.378500332473777e-05, + 3.5516346542863175e-05, + 3.742165063158609e-05, + 3.6865043512079865e-05, + 3.170729542034678e-05 + ], + "bp_grad_norms_F": [ + 0.0015557212755084038, + 0.0016067115357145667, + 0.001632526982575655, + 0.0015416526002809405, + 0.0012708577560260892 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5888671875, + "loss_eval": 1.1785690784454346, + "epoch": 29 + }, + { + "hidden_norms": [ + 271.3736572265625, + 257.6708984375, + 246.3377685546875, + 240.7875213623047, + 248.55162048339844 + ], + "bp_grad_norms_per_sample_med": [ + 3.2452040613861755e-05, + 3.456772174104117e-05, + 3.550515975803137e-05, + 3.530189496814273e-05, + 3.098960587522015e-05 + ], + "bp_grad_norms_F": [ + 0.0015420741401612759, + 0.0016120158834382892, + 0.0016289768973365426, + 0.0015373170608654618, + 0.0012695658951997757 + ], + "gamma_dfa": NaN, + "acc_eval": 0.57421875, + "loss_eval": 1.1686367988586426, + "epoch": 30 + }, + { + "hidden_norms": [ + 270.6155090332031, + 256.5137634277344, + 246.4404296875, + 238.71493530273438, + 245.14651489257812 + ], + "bp_grad_norms_per_sample_med": [ + 3.4225693525513634e-05, + 3.6208981327945367e-05, + 3.771808042074554e-05, + 3.671547892736271e-05, + 3.175914389430545e-05 + ], + "bp_grad_norms_F": [ + 0.0016278985422104597, + 0.00169221474789083, + 0.0017131771892309189, + 0.0016123101813718677, + 0.001308745937421918 + ], + "gamma_dfa": NaN, + "acc_eval": 0.572265625, + "loss_eval": 1.1706045866012573, + "epoch": 31 + }, + { + "hidden_norms": [ + 271.2703552246094, + 258.5479431152344, + 247.48568725585938, + 241.0449676513672, + 248.61419677734375 + ], + "bp_grad_norms_per_sample_med": [ + 3.370125705259852e-05, + 3.563391510397196e-05, + 3.776583253056742e-05, + 3.622193253249861e-05, + 3.1177536584436893e-05 + ], + "bp_grad_norms_F": [ + 0.001631051884032786, + 0.0016799280419945717, + 0.0016939701745286584, + 0.001596588990651071, + 0.0012989162933081388 + ], + "gamma_dfa": NaN, + "acc_eval": 0.576171875, + "loss_eval": 1.1757416725158691, + "epoch": 32 + }, + { + "hidden_norms": [ + 271.5987854003906, + 257.3157043457031, + 247.04425048828125, + 239.22933959960938, + 247.08534240722656 + ], + "bp_grad_norms_per_sample_med": [ + 3.276485949754715e-05, + 3.4356282412773e-05, + 3.570897752069868e-05, + 3.5391843994148076e-05, + 3.004215977853164e-05 + ], + "bp_grad_norms_F": [ + 0.0015858211554586887, + 0.0016445693327113986, + 0.0016775003168731928, + 0.0015801271656528115, + 0.0012829708866775036 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5947265625, + "loss_eval": 1.1316245794296265, + "epoch": 33 + }, + { + "hidden_norms": [ + 276.85565185546875, + 262.14959716796875, + 248.80125427246094, + 239.6945037841797, + 246.55982971191406 + ], + "bp_grad_norms_per_sample_med": [ + 3.424682654440403e-05, + 3.617212496465072e-05, + 3.731884135049768e-05, + 3.583814759622328e-05, + 3.1317249522544444e-05 + ], + "bp_grad_norms_F": [ + 0.001694253645837307, + 0.0017545269802212715, + 0.0017868074355646968, + 0.0016729168128222227, + 0.0013285611057654023 + ], + "gamma_dfa": NaN, + "acc_eval": 0.599609375, + "loss_eval": 1.1637687683105469, + "epoch": 34 + }, + { + "hidden_norms": [ + 274.4824523925781, + 260.23223876953125, + 249.41390991210938, + 242.22657775878906, + 248.33303833007812 + ], + "bp_grad_norms_per_sample_med": [ + 3.4118878829758614e-05, + 3.6190944229019806e-05, + 3.7162455555517226e-05, + 3.6398294469108805e-05, + 3.1276995287043974e-05 + ], + "bp_grad_norms_F": [ + 0.0016947221010923386, + 0.0017584856832399964, + 0.0017846780829131603, + 0.0016663862625136971, + 0.0013457784662023187 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5830078125, + "loss_eval": 1.1625440120697021, + "epoch": 35 + }, + { + "hidden_norms": [ + 279.3048095703125, + 263.830810546875, + 250.07839965820312, + 242.9677276611328, + 248.742431640625 + ], + "bp_grad_norms_per_sample_med": [ + 3.378376641194336e-05, + 3.623673183028586e-05, + 3.8036723708501086e-05, + 3.8026879337849095e-05, + 3.20856343023479e-05 + ], + "bp_grad_norms_F": [ + 0.0016582749085500836, + 0.001725532696582377, + 0.0017548021860420704, + 0.0016622358234599233, + 0.0013314742827787995 + ], + "gamma_dfa": NaN, + "acc_eval": 0.58203125, + "loss_eval": 1.1663323640823364, + "epoch": 36 + }, + { + "hidden_norms": [ + 277.88153076171875, + 263.3263244628906, + 254.0471954345703, + 246.00294494628906, + 251.6592559814453 + ], + "bp_grad_norms_per_sample_med": [ + 3.402382935746573e-05, + 3.644815296866e-05, + 3.812061549979262e-05, + 3.8299189327517524e-05, + 3.244871550123207e-05 + ], + "bp_grad_norms_F": [ + 0.0016575837507843971, + 0.001726161572150886, + 0.0017651193775236607, + 0.001658888068050146, + 0.001325216144323349 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5849609375, + "loss_eval": 1.1578618288040161, + "epoch": 37 + }, + { + "hidden_norms": [ + 278.96038818359375, + 264.09832763671875, + 252.11532592773438, + 243.7300567626953, + 250.29537963867188 + ], + "bp_grad_norms_per_sample_med": [ + 3.398891203687526e-05, + 3.60828016710002e-05, + 3.8209349440876395e-05, + 3.6874149373034015e-05, + 3.130899858660996e-05 + ], + "bp_grad_norms_F": [ + 0.0016938851913437247, + 0.0017578925471752882, + 0.0017999019473791122, + 0.0016839306335896254, + 0.0013463495997712016 + ], + "gamma_dfa": NaN, + "acc_eval": 0.59375, + "loss_eval": 1.1660709381103516, + "epoch": 38 + }, + { + "hidden_norms": [ + 279.11102294921875, + 262.8247375488281, + 250.32635498046875, + 241.73477172851562, + 246.8911590576172 + ], + "bp_grad_norms_per_sample_med": [ + 3.476023630355485e-05, + 3.673156606964767e-05, + 3.8200654671527445e-05, + 3.745179856196046e-05, + 3.111823389190249e-05 + ], + "bp_grad_norms_F": [ + 0.0017367384862154722, + 0.001807119813747704, + 0.0018404393922537565, + 0.0017188042402267456, + 0.0013523433590307832 + ], + "gamma_dfa": NaN, + "acc_eval": 0.58984375, + "loss_eval": 1.1548914909362793, + "epoch": 39 + }, + { + "hidden_norms": [ + 278.79083251953125, + 263.49237060546875, + 252.0679931640625, + 243.65948486328125, + 249.34564208984375 + ], + "bp_grad_norms_per_sample_med": [ + 3.3709933632053435e-05, + 3.595885937102139e-05, + 3.795271186390892e-05, + 3.70899579138495e-05, + 3.1362116715172306e-05 + ], + "bp_grad_norms_F": [ + 0.0017184949247166514, + 0.0017862631939351559, + 0.0018348938319832087, + 0.0017180118011310697, + 0.0013606568099930882 + ], + "gamma_dfa": NaN, + "acc_eval": 0.591796875, + "loss_eval": 1.159794569015503, + "epoch": 40 + }, + { + "hidden_norms": [ + 281.3595886230469, + 263.4467468261719, + 250.73536682128906, + 243.140380859375, + 247.22425842285156 + ], + "bp_grad_norms_per_sample_med": [ + 3.411353100091219e-05, + 3.656598346424289e-05, + 3.8585232687182724e-05, + 3.752345946850255e-05, + 3.1474613933824e-05 + ], + "bp_grad_norms_F": [ + 0.0017623946769163013, + 0.0018276257906109095, + 0.0018694576574489474, + 0.0017436364432796836, + 0.0013814391568303108 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5888671875, + "loss_eval": 1.153226375579834, + "epoch": 41 + }, + { + "hidden_norms": [ + 279.79888916015625, + 263.5989685058594, + 252.2397918701172, + 243.49703979492188, + 248.86973571777344 + ], + "bp_grad_norms_per_sample_med": [ + 3.528413435560651e-05, + 3.805026790359989e-05, + 4.0152310248231515e-05, + 3.962377377320081e-05, + 3.250985173508525e-05 + ], + "bp_grad_norms_F": [ + 0.0017320383340120316, + 0.0018105976050719619, + 0.001847731531597674, + 0.0017405269900336862, + 0.0013762396993115544 + ], + "gamma_dfa": NaN, + "acc_eval": 0.58984375, + "loss_eval": 1.1387929916381836, + "epoch": 42 + }, + { + "hidden_norms": [ + 280.80633544921875, + 263.2684326171875, + 251.36338806152344, + 242.25201416015625, + 245.46954345703125 + ], + "bp_grad_norms_per_sample_med": [ + 3.463789835222997e-05, + 3.694587940117344e-05, + 3.909278166247532e-05, + 3.8173999200807884e-05, + 3.152083809254691e-05 + ], + "bp_grad_norms_F": [ + 0.001763346022926271, + 0.0018382675480097532, + 0.0018745084526017308, + 0.0017674838891252875, + 0.0013920797500759363 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5986328125, + "loss_eval": 1.1115086078643799, + "epoch": 43 + }, + { + "hidden_norms": [ + 278.4877624511719, + 261.4671630859375, + 247.66079711914062, + 240.21690368652344, + 244.1111602783203 + ], + "bp_grad_norms_per_sample_med": [ + 3.559369361028075e-05, + 3.895946065313183e-05, + 4.048463233630173e-05, + 3.892362292390317e-05, + 3.229418143746443e-05 + ], + "bp_grad_norms_F": [ + 0.0018469772767275572, + 0.0019197544315829873, + 0.001963041489943862, + 0.0018302135867998004, + 0.001429378753527999 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6025390625, + "loss_eval": 1.16768217086792, + "epoch": 44 + }, + { + "hidden_norms": [ + 280.8171691894531, + 263.16705322265625, + 251.45835876464844, + 243.88563537597656, + 246.92945861816406 + ], + "bp_grad_norms_per_sample_med": [ + 3.3486583561170846e-05, + 3.533214839990251e-05, + 3.789112452068366e-05, + 3.651758743217215e-05, + 3.081141403526999e-05 + ], + "bp_grad_norms_F": [ + 0.0017694556154310703, + 0.0018461478175595403, + 0.0018877091351896524, + 0.001781664090231061, + 0.0014001834206283092 + ], + "gamma_dfa": NaN, + "acc_eval": 0.611328125, + "loss_eval": 1.1236202716827393, + "epoch": 45 + }, + { + "hidden_norms": [ + 280.0203552246094, + 261.3996887207031, + 248.5751953125, + 240.40728759765625, + 243.84178161621094 + ], + "bp_grad_norms_per_sample_med": [ + 3.66695094271563e-05, + 3.912827742169611e-05, + 4.0782215364743024e-05, + 3.97856165363919e-05, + 3.258495053160004e-05 + ], + "bp_grad_norms_F": [ + 0.0018954386468976736, + 0.0019765326287597418, + 0.0020179273560643196, + 0.0018787897424772382, + 0.0014408754650503397 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5947265625, + "loss_eval": 1.142503261566162, + "epoch": 46 + }, + { + "hidden_norms": [ + 278.37445068359375, + 260.7147521972656, + 247.63418579101562, + 239.64564514160156, + 241.75396728515625 + ], + "bp_grad_norms_per_sample_med": [ + 3.824803570751101e-05, + 4.062152947881259e-05, + 4.1070070437854156e-05, + 4.01996003347449e-05, + 3.288290827185847e-05 + ], + "bp_grad_norms_F": [ + 0.001872657099738717, + 0.0019627343863248825, + 0.002008425537496805, + 0.0018816223600879312, + 0.001456336583942175 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6064453125, + "loss_eval": 1.137413740158081, + "epoch": 47 + }, + { + "hidden_norms": [ + 281.14813232421875, + 263.6340637207031, + 250.13082885742188, + 242.0965118408203, + 244.81060791015625 + ], + "bp_grad_norms_per_sample_med": [ + 3.531872062012553e-05, + 3.7883237382629886e-05, + 4.0227863792097196e-05, + 3.96844552597031e-05, + 3.296942304586992e-05 + ], + "bp_grad_norms_F": [ + 0.0018434273079037666, + 0.0019304269226267934, + 0.0019746439065784216, + 0.0018603875068947673, + 0.0014465745771303773 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6025390625, + "loss_eval": 1.1396205425262451, + "epoch": 48 + }, + { + "hidden_norms": [ + 281.31280517578125, + 262.5829772949219, + 247.9232940673828, + 239.8461456298828, + 242.20977783203125 + ], + "bp_grad_norms_per_sample_med": [ + 3.71285859728232e-05, + 3.963968629250303e-05, + 4.135111521463841e-05, + 4.0518349123885855e-05, + 3.323471173644066e-05 + ], + "bp_grad_norms_F": [ + 0.00189371092710644, + 0.0019808171782642603, + 0.002019244944676757, + 0.001904282602481544, + 0.0014800610952079296 + ], + "gamma_dfa": NaN, + "acc_eval": 0.59375, + "loss_eval": 1.1411675214767456, + "epoch": 49 + }, + { + "hidden_norms": [ + 278.64208984375, + 260.1700744628906, + 246.7251434326172, + 239.43145751953125, + 240.350830078125 + ], + "bp_grad_norms_per_sample_med": [ + 3.6054647353012115e-05, + 3.858766285702586e-05, + 4.05439204769209e-05, + 4.046479079988785e-05, + 3.3791853638831526e-05 + ], + "bp_grad_norms_F": [ + 0.0018870895728468895, + 0.00198533502407372, + 0.0020502021070569754, + 0.001929490827023983, + 0.0014933926286175847 + ], + "gamma_dfa": NaN, + "acc_eval": 0.59375, + "loss_eval": 1.1310430765151978, + "epoch": 50 + }, + { + "hidden_norms": [ + 281.7696533203125, + 261.5150451660156, + 247.67295837402344, + 239.7654266357422, + 240.27789306640625 + ], + "bp_grad_norms_per_sample_med": [ + 3.671262675197795e-05, + 3.92102992918808e-05, + 4.1406125092180446e-05, + 4.029196497867815e-05, + 3.32514064211864e-05 + ], + "bp_grad_norms_F": [ + 0.001880357856862247, + 0.001981948269531131, + 0.0020440546795725822, + 0.0019350070506334305, + 0.0014877711655572057 + ], + "gamma_dfa": NaN, + "acc_eval": 0.60546875, + "loss_eval": 1.1188440322875977, + "epoch": 51 + }, + { + "hidden_norms": [ + 279.01318359375, + 259.6091003417969, + 244.6603240966797, + 235.1455535888672, + 237.438232421875 + ], + "bp_grad_norms_per_sample_med": [ + 4.0185608668252826e-05, + 4.309232099330984e-05, + 4.5118234993424267e-05, + 4.371534669189714e-05, + 3.5620210837805644e-05 + ], + "bp_grad_norms_F": [ + 0.0020104716531932354, + 0.002107330597937107, + 0.002167000202462077, + 0.0020260754972696304, + 0.0015411453787237406 + ], + "gamma_dfa": NaN, + "acc_eval": 0.603515625, + "loss_eval": 1.1525764465332031, + "epoch": 52 + }, + { + "hidden_norms": [ + 278.2245178222656, + 257.14617919921875, + 243.9732666015625, + 234.4013214111328, + 236.68521118164062 + ], + "bp_grad_norms_per_sample_med": [ + 3.85418206860777e-05, + 4.142443503951654e-05, + 4.363177140476182e-05, + 4.2996281990781426e-05, + 3.424972965149209e-05 + ], + "bp_grad_norms_F": [ + 0.0020445864647626877, + 0.002142493613064289, + 0.0022005243226885796, + 0.002054274780675769, + 0.0015541499014943838 + ], + "gamma_dfa": NaN, + "acc_eval": 0.59765625, + "loss_eval": 1.162191390991211, + "epoch": 53 + }, + { + "hidden_norms": [ + 278.7825622558594, + 258.4292297363281, + 243.01690673828125, + 234.15765380859375, + 235.1446533203125 + ], + "bp_grad_norms_per_sample_med": [ + 3.855194881907664e-05, + 4.1293307731393725e-05, + 4.296215047361329e-05, + 4.1900369978975505e-05, + 3.372962964931503e-05 + ], + "bp_grad_norms_F": [ + 0.002041360829025507, + 0.0021457262337207794, + 0.0021996963769197464, + 0.0020641626324504614, + 0.001561368815600872 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6171875, + "loss_eval": 1.130479335784912, + "epoch": 54 + }, + { + "hidden_norms": [ + 275.3794860839844, + 256.83026123046875, + 242.16006469726562, + 231.49655151367188, + 232.30279541015625 + ], + "bp_grad_norms_per_sample_med": [ + 4.011294731753878e-05, + 4.240096313878894e-05, + 4.396942676976323e-05, + 4.347893627709709e-05, + 3.485888009890914e-05 + ], + "bp_grad_norms_F": [ + 0.0021268154960125685, + 0.002230642130598426, + 0.0022892621345818043, + 0.002145433332771063, + 0.0016037407331168652 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6005859375, + "loss_eval": 1.1599653959274292, + "epoch": 55 + }, + { + "hidden_norms": [ + 277.5028076171875, + 257.1715087890625, + 241.6505584716797, + 234.09323120117188, + 233.63377380371094 + ], + "bp_grad_norms_per_sample_med": [ + 3.876453047269024e-05, + 4.156330396654084e-05, + 4.3727923184633255e-05, + 4.2131912778131664e-05, + 3.510245369398035e-05 + ], + "bp_grad_norms_F": [ + 0.002074806485325098, + 0.002173666376620531, + 0.002245684852823615, + 0.0021073734387755394, + 0.0015908328350633383 + ], + "gamma_dfa": NaN, + "acc_eval": 0.611328125, + "loss_eval": 1.144984245300293, + "epoch": 56 + }, + { + "hidden_norms": [ + 276.0826416015625, + 255.17193603515625, + 239.51519775390625, + 231.28643798828125, + 230.85292053222656 + ], + "bp_grad_norms_per_sample_med": [ + 3.960869798902422e-05, + 4.2907246097456664e-05, + 4.45116929768119e-05, + 4.3612952140392736e-05, + 3.500159073155373e-05 + ], + "bp_grad_norms_F": [ + 0.0021277335472404957, + 0.0022348251659423113, + 0.0022981096990406513, + 0.002155488822609186, + 0.001618923619389534 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6142578125, + "loss_eval": 1.1534931659698486, + "epoch": 57 + }, + { + "hidden_norms": [ + 275.6867370605469, + 254.65744018554688, + 238.1124267578125, + 229.36215209960938, + 227.90867614746094 + ], + "bp_grad_norms_per_sample_med": [ + 3.966662916354835e-05, + 4.33680324931629e-05, + 4.5270633563632146e-05, + 4.528860517893918e-05, + 3.5298704460728914e-05 + ], + "bp_grad_norms_F": [ + 0.002194908680394292, + 0.002305977512151003, + 0.0023776311427354813, + 0.002223761286586523, + 0.001651395345106721 + ], + "gamma_dfa": NaN, + "acc_eval": 0.611328125, + "loss_eval": 1.1494864225387573, + "epoch": 58 + }, + { + "hidden_norms": [ + 274.8414306640625, + 253.18215942382812, + 238.03150939941406, + 228.1765594482422, + 229.46058654785156 + ], + "bp_grad_norms_per_sample_med": [ + 4.132632238906808e-05, + 4.4773249101126567e-05, + 4.6669130824739113e-05, + 4.591346441884525e-05, + 3.6523833841783926e-05 + ], + "bp_grad_norms_F": [ + 0.0022293017245829105, + 0.002341218525543809, + 0.002405093051493168, + 0.0022504546213895082, + 0.0016715668607503176 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5927734375, + "loss_eval": 1.1698435544967651, + "epoch": 59 + }, + { + "hidden_norms": [ + 273.4884338378906, + 251.2228240966797, + 236.2550048828125, + 226.7325439453125, + 225.68386840820312 + ], + "bp_grad_norms_per_sample_med": [ + 4.192883352516219e-05, + 4.3471976823639125e-05, + 4.7195244405884296e-05, + 4.490713035920635e-05, + 3.646116965683177e-05 + ], + "bp_grad_norms_F": [ + 0.0022312228102236986, + 0.002349371323361993, + 0.0024164437782019377, + 0.002271530916914344, + 0.0016918154433369637 + ], + "gamma_dfa": NaN, + "acc_eval": 0.609375, + "loss_eval": 1.1560128927230835, + "epoch": 60 + }, + { + "hidden_norms": [ + 273.4084777832031, + 250.5165557861328, + 236.19557189941406, + 227.0643768310547, + 226.594970703125 + ], + "bp_grad_norms_per_sample_med": [ + 4.2211744585074484e-05, + 4.488111881073564e-05, + 4.760283627547324e-05, + 4.589638047036715e-05, + 3.658623973024078e-05 + ], + "bp_grad_norms_F": [ + 0.0022313676308840513, + 0.002337696962058544, + 0.0024093035608530045, + 0.002264693146571517, + 0.0016880000475794077 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6044921875, + "loss_eval": 1.1498074531555176, + "epoch": 61 + }, + { + "hidden_norms": [ + 273.1318664550781, + 249.35223388671875, + 234.15737915039062, + 223.48074340820312, + 223.0059356689453 + ], + "bp_grad_norms_per_sample_med": [ + 4.4729218643624336e-05, + 4.768320650327951e-05, + 4.822494520340115e-05, + 4.791786341229454e-05, + 3.7625926779583097e-05 + ], + "bp_grad_norms_F": [ + 0.0023362624924629927, + 0.0024679312482476234, + 0.0025367215275764465, + 0.0023599599953740835, + 0.0017239096341654658 + ], + "gamma_dfa": NaN, + "acc_eval": 0.603515625, + "loss_eval": 1.1377537250518799, + "epoch": 62 + }, + { + "hidden_norms": [ + 270.9543762207031, + 249.6943817138672, + 234.0796356201172, + 223.80093383789062, + 222.75062561035156 + ], + "bp_grad_norms_per_sample_med": [ + 4.4321575842332095e-05, + 4.7786033974261954e-05, + 4.9388974730391055e-05, + 4.749266372527927e-05, + 3.7502803024835885e-05 + ], + "bp_grad_norms_F": [ + 0.0023458057548850775, + 0.0024719720240682364, + 0.0025427823420614004, + 0.0023789280094206333, + 0.0017488099401816726 + ], + "gamma_dfa": NaN, + "acc_eval": 0.609375, + "loss_eval": 1.1472076177597046, + "epoch": 63 + }, + { + "hidden_norms": [ + 270.6864318847656, + 249.71315002441406, + 233.9838409423828, + 223.17515563964844, + 222.17759704589844 + ], + "bp_grad_norms_per_sample_med": [ + 4.36245281889569e-05, + 4.725098915514536e-05, + 4.936556069878861e-05, + 4.7286572225857526e-05, + 3.7781614082632586e-05 + ], + "bp_grad_norms_F": [ + 0.0023510511964559555, + 0.0024691869039088488, + 0.0025394195690751076, + 0.0023858139757066965, + 0.0017493697814643383 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5986328125, + "loss_eval": 1.1437795162200928, + "epoch": 64 + }, + { + "hidden_norms": [ + 269.1813659667969, + 247.57632446289062, + 231.57135009765625, + 220.31541442871094, + 220.33135986328125 + ], + "bp_grad_norms_per_sample_med": [ + 4.363871266832575e-05, + 4.776224886882119e-05, + 4.8712514399085194e-05, + 4.777491994900629e-05, + 3.733497578650713e-05 + ], + "bp_grad_norms_F": [ + 0.0023870510049164295, + 0.0025214161723852158, + 0.0025991688016802073, + 0.0024342697579413652, + 0.0017686353530734777 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6123046875, + "loss_eval": 1.138253927230835, + "epoch": 65 + }, + { + "hidden_norms": [ + 268.92877197265625, + 245.84384155273438, + 230.63307189941406, + 220.8726806640625, + 219.60350036621094 + ], + "bp_grad_norms_per_sample_med": [ + 4.505042306846008e-05, + 4.780786548508331e-05, + 5.067480378784239e-05, + 4.878537220065482e-05, + 3.842872683890164e-05 + ], + "bp_grad_norms_F": [ + 0.00237255753017962, + 0.002493572887033224, + 0.0025717453099787235, + 0.0024270617868751287, + 0.0017787005053833127 + ], + "gamma_dfa": NaN, + "acc_eval": 0.611328125, + "loss_eval": 1.1350346803665161, + "epoch": 66 + }, + { + "hidden_norms": [ + 268.9022521972656, + 244.85621643066406, + 229.2279510498047, + 218.72445678710938, + 217.325439453125 + ], + "bp_grad_norms_per_sample_med": [ + 4.515323234954849e-05, + 4.8419675295008346e-05, + 5.0062186346622184e-05, + 4.777919821208343e-05, + 3.962183836847544e-05 + ], + "bp_grad_norms_F": [ + 0.0024519655853509903, + 0.002586368238553405, + 0.00266969739459455, + 0.0024983694311231375, + 0.001813769806176424 + ], + "gamma_dfa": NaN, + "acc_eval": 0.58984375, + "loss_eval": 1.143377661705017, + "epoch": 67 + }, + { + "hidden_norms": [ + 267.7603454589844, + 245.58932495117188, + 229.04205322265625, + 219.0813446044922, + 217.887939453125 + ], + "bp_grad_norms_per_sample_med": [ + 4.398225064505823e-05, + 4.6881206799298525e-05, + 4.986919157090597e-05, + 4.9659283831715584e-05, + 3.928233854821883e-05 + ], + "bp_grad_norms_F": [ + 0.0024525129701942205, + 0.0025901535991579294, + 0.002675000112503767, + 0.0025050004478543997, + 0.0018165758810937405 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6083984375, + "loss_eval": 1.1445319652557373, + "epoch": 68 + }, + { + "hidden_norms": [ + 267.32110595703125, + 244.51095581054688, + 228.1964569091797, + 217.7543182373047, + 215.99093627929688 + ], + "bp_grad_norms_per_sample_med": [ + 4.567872019833885e-05, + 4.9028451030608267e-05, + 5.0534836191218346e-05, + 4.8578112910036e-05, + 3.865628605126403e-05 + ], + "bp_grad_norms_F": [ + 0.002467149868607521, + 0.0026051453314721584, + 0.0026927595026791096, + 0.002530105412006378, + 0.0018421142594888806 + ], + "gamma_dfa": NaN, + "acc_eval": 0.615234375, + "loss_eval": 1.143081545829773, + "epoch": 69 + }, + { + "hidden_norms": [ + 266.6060791015625, + 242.973388671875, + 227.23739624023438, + 216.60504150390625, + 213.26922607421875 + ], + "bp_grad_norms_per_sample_med": [ + 4.524671021499671e-05, + 4.8752917791716754e-05, + 5.0053116865456104e-05, + 4.897990220342763e-05, + 3.949753590859473e-05 + ], + "bp_grad_norms_F": [ + 0.002536450745537877, + 0.002682002494111657, + 0.002782325493171811, + 0.002600395353510976, + 0.0018774428172037005 + ], + "gamma_dfa": NaN, + "acc_eval": 0.607421875, + "loss_eval": 1.1569209098815918, + "epoch": 70 + }, + { + "hidden_norms": [ + 265.7056579589844, + 243.6046142578125, + 226.88746643066406, + 217.55291748046875, + 213.99151611328125 + ], + "bp_grad_norms_per_sample_med": [ + 4.667519897338934e-05, + 4.9728132580639794e-05, + 5.27437987329904e-05, + 5.151727600605227e-05, + 4.0550570702180266e-05 + ], + "bp_grad_norms_F": [ + 0.002509700832888484, + 0.0026521242689341307, + 0.0027410376351326704, + 0.002574845217168331, + 0.001871992484666407 + ], + "gamma_dfa": NaN, + "acc_eval": 0.611328125, + "loss_eval": 1.1381304264068604, + "epoch": 71 + }, + { + "hidden_norms": [ + 264.63116455078125, + 241.50344848632812, + 225.20465087890625, + 214.67364501953125, + 212.77584838867188 + ], + "bp_grad_norms_per_sample_med": [ + 4.526918928604573e-05, + 5.0124705012422055e-05, + 5.2680206863442436e-05, + 5.1697836170205846e-05, + 3.9593822293682024e-05 + ], + "bp_grad_norms_F": [ + 0.0025687876623123884, + 0.0027084490284323692, + 0.0027945584151893854, + 0.0026290949899703264, + 0.0018978636944666505 + ], + "gamma_dfa": NaN, + "acc_eval": 0.609375, + "loss_eval": 1.149982213973999, + "epoch": 72 + }, + { + "hidden_norms": [ + 264.168701171875, + 241.78074645996094, + 224.8336944580078, + 214.45762634277344, + 212.4181671142578 + ], + "bp_grad_norms_per_sample_med": [ + 4.745498517877422e-05, + 5.042303018854e-05, + 5.3622185077983886e-05, + 5.276402589515783e-05, + 4.0586572140455246e-05 + ], + "bp_grad_norms_F": [ + 0.0026231552474200726, + 0.002779489615932107, + 0.002866284688934684, + 0.00268669705837965, + 0.0019228557357564569 + ], + "gamma_dfa": NaN, + "acc_eval": 0.611328125, + "loss_eval": 1.158836007118225, + "epoch": 73 + }, + { + "hidden_norms": [ + 263.89593505859375, + 240.3160858154297, + 223.6101531982422, + 212.4967803955078, + 210.63546752929688 + ], + "bp_grad_norms_per_sample_med": [ + 4.678544428315945e-05, + 5.029747626394965e-05, + 5.3069161367602646e-05, + 5.2157745813019574e-05, + 4.0450311644235626e-05 + ], + "bp_grad_norms_F": [ + 0.0026451845187693834, + 0.002798875328153372, + 0.0028935885056853294, + 0.0027081011794507504, + 0.001933218096382916 + ], + "gamma_dfa": NaN, + "acc_eval": 0.607421875, + "loss_eval": 1.1447510719299316, + "epoch": 74 + }, + { + "hidden_norms": [ + 262.4220275878906, + 239.43899536132812, + 223.32151794433594, + 212.5982666015625, + 209.80105590820312 + ], + "bp_grad_norms_per_sample_med": [ + 4.713214002549648e-05, + 5.087011959403753e-05, + 5.283053906168789e-05, + 5.280825644149445e-05, + 4.054068267578259e-05 + ], + "bp_grad_norms_F": [ + 0.002688886132091284, + 0.002841175301000476, + 0.002931408118456602, + 0.0027343749534338713, + 0.001953375292941928 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6201171875, + "loss_eval": 1.1561517715454102, + "epoch": 75 + }, + { + "hidden_norms": [ + 262.65814208984375, + 240.0749969482422, + 223.0806121826172, + 212.08688354492188, + 209.22561645507812 + ], + "bp_grad_norms_per_sample_med": [ + 4.8153047828236595e-05, + 5.145415343577042e-05, + 5.4130876378621906e-05, + 5.3563955589197576e-05, + 4.1630475607234985e-05 + ], + "bp_grad_norms_F": [ + 0.0026582488790154457, + 0.0028067713137716055, + 0.00290639977902174, + 0.0027312941383570433, + 0.0019485733937472105 + ], + "gamma_dfa": NaN, + "acc_eval": 0.609375, + "loss_eval": 1.1465811729431152, + "epoch": 76 + }, + { + "hidden_norms": [ + 262.0201721191406, + 239.08547973632812, + 222.36146545410156, + 211.62014770507812, + 208.1699676513672 + ], + "bp_grad_norms_per_sample_med": [ + 4.763290053233504e-05, + 5.171096199774183e-05, + 5.4121221182867885e-05, + 5.251873881206848e-05, + 4.096185875823721e-05 + ], + "bp_grad_norms_F": [ + 0.0027110630180686712, + 0.002856222679838538, + 0.002954155672341585, + 0.0027671835850924253, + 0.001963438233360648 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6142578125, + "loss_eval": 1.140420913696289, + "epoch": 77 + }, + { + "hidden_norms": [ + 261.04132080078125, + 238.44912719726562, + 221.8912353515625, + 210.62893676757812, + 207.68223571777344 + ], + "bp_grad_norms_per_sample_med": [ + 4.7598918172298e-05, + 5.09894234710373e-05, + 5.384815813158639e-05, + 5.2337087254272774e-05, + 4.065291432198137e-05 + ], + "bp_grad_norms_F": [ + 0.00274560390971601, + 0.0029035566840320826, + 0.0030046424362808466, + 0.0028060651384294033, + 0.001984622096642852 + ], + "gamma_dfa": NaN, + "acc_eval": 0.615234375, + "loss_eval": 1.1556674242019653, + "epoch": 78 + }, + { + "hidden_norms": [ + 260.81304931640625, + 237.52015686035156, + 221.12814331054688, + 210.04925537109375, + 206.80441284179688 + ], + "bp_grad_norms_per_sample_med": [ + 4.8430883907712996e-05, + 5.440305176307447e-05, + 5.596410119323991e-05, + 5.4586391343036667e-05, + 4.195213477942161e-05 + ], + "bp_grad_norms_F": [ + 0.002767855068668723, + 0.002925564767792821, + 0.003015928901731968, + 0.002823467366397381, + 0.0020000736694782972 + ], + "gamma_dfa": NaN, + "acc_eval": 0.599609375, + "loss_eval": 1.155595302581787, + "epoch": 79 + }, + { + "hidden_norms": [ + 260.0649108886719, + 236.74310302734375, + 220.48731994628906, + 209.0772247314453, + 205.55821228027344 + ], + "bp_grad_norms_per_sample_med": [ + 4.8615416744723916e-05, + 5.309982589096762e-05, + 5.548154513235204e-05, + 5.3888015827396885e-05, + 4.1955223423428833e-05 + ], + "bp_grad_norms_F": [ + 0.0028222037944942713, + 0.0029849831480532885, + 0.003080391325056553, + 0.0028690374456346035, + 0.002019484294578433 + ], + "gamma_dfa": NaN, + "acc_eval": 0.609375, + "loss_eval": 1.151352882385254, + "epoch": 80 + }, + { + "hidden_norms": [ + 259.8204650878906, + 236.13253784179688, + 219.08865356445312, + 207.6793975830078, + 205.0387725830078 + ], + "bp_grad_norms_per_sample_med": [ + 4.734982212539762e-05, + 5.1937022362835705e-05, + 5.472155680763535e-05, + 5.394391337176785e-05, + 4.0832986996974796e-05 + ], + "bp_grad_norms_F": [ + 0.00282403570599854, + 0.0029848285485059023, + 0.0030928945634514093, + 0.002879998181015253, + 0.0020237534772604704 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6103515625, + "loss_eval": 1.153958797454834, + "epoch": 81 + }, + { + "hidden_norms": [ + 259.70703125, + 235.5811767578125, + 218.87939453125, + 207.87005615234375, + 204.21360778808594 + ], + "bp_grad_norms_per_sample_med": [ + 4.972740134689957e-05, + 5.442327164928429e-05, + 5.538179539144039e-05, + 5.4411018936662003e-05, + 4.185390207567252e-05 + ], + "bp_grad_norms_F": [ + 0.002822998445481062, + 0.00298913661390543, + 0.0030840388499200344, + 0.0028817548882216215, + 0.0020279802847653627 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6201171875, + "loss_eval": 1.153503656387329, + "epoch": 82 + }, + { + "hidden_norms": [ + 258.22747802734375, + 235.3745880126953, + 218.20066833496094, + 206.82144165039062, + 203.41139221191406 + ], + "bp_grad_norms_per_sample_med": [ + 5.077639434603043e-05, + 5.5536078434670344e-05, + 5.664999844157137e-05, + 5.608425635728054e-05, + 4.293184611015022e-05 + ], + "bp_grad_norms_F": [ + 0.002843277994543314, + 0.0030130224768072367, + 0.0031098793260753155, + 0.00290561281144619, + 0.0020425335969775915 + ], + "gamma_dfa": NaN, + "acc_eval": 0.603515625, + "loss_eval": 1.1597235202789307, + "epoch": 83 + }, + { + "hidden_norms": [ + 258.78253173828125, + 234.73965454101562, + 217.91439819335938, + 207.1282958984375, + 203.4480743408203 + ], + "bp_grad_norms_per_sample_med": [ + 4.941036968375556e-05, + 5.3143365221330896e-05, + 5.529402187676169e-05, + 5.3718889830634e-05, + 4.207601887173951e-05 + ], + "bp_grad_norms_F": [ + 0.0028217420913279057, + 0.0029905554838478565, + 0.0030917164403945208, + 0.002896753139793873, + 0.00204362440854311 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6103515625, + "loss_eval": 1.1545956134796143, + "epoch": 84 + }, + { + "hidden_norms": [ + 258.10736083984375, + 234.35299682617188, + 217.81224060058594, + 206.38864135742188, + 202.70126342773438 + ], + "bp_grad_norms_per_sample_med": [ + 4.899106716038659e-05, + 5.2254512411309406e-05, + 5.481128755491227e-05, + 5.298088217386976e-05, + 4.121303936699405e-05 + ], + "bp_grad_norms_F": [ + 0.002849399344995618, + 0.0030259499326348305, + 0.0031236386857926846, + 0.002921136561781168, + 0.0020506957080215216 + ], + "gamma_dfa": NaN, + "acc_eval": 0.615234375, + "loss_eval": 1.150564193725586, + "epoch": 85 + }, + { + "hidden_norms": [ + 258.50750732421875, + 234.4120330810547, + 217.49217224121094, + 205.7490692138672, + 202.62380981445312 + ], + "bp_grad_norms_per_sample_med": [ + 4.976892887498252e-05, + 5.383255484048277e-05, + 5.497025267686695e-05, + 5.391201557358727e-05, + 4.2558931454550475e-05 + ], + "bp_grad_norms_F": [ + 0.0028686418663710356, + 0.003045660676434636, + 0.0031442255713045597, + 0.002935740165412426, + 0.0020568587351590395 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6103515625, + "loss_eval": 1.152902603149414, + "epoch": 86 + }, + { + "hidden_norms": [ + 258.063720703125, + 233.87294006347656, + 217.34030151367188, + 205.4007110595703, + 202.4425506591797 + ], + "bp_grad_norms_per_sample_med": [ + 5.047345257480629e-05, + 5.492625859915279e-05, + 5.640966628561728e-05, + 5.587706982623786e-05, + 4.286773400963284e-05 + ], + "bp_grad_norms_F": [ + 0.002875820267945528, + 0.003056521760299802, + 0.0031533341389149427, + 0.002944495528936386, + 0.0020668748766183853 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6083984375, + "loss_eval": 1.152883529663086, + "epoch": 87 + }, + { + "hidden_norms": [ + 258.1524353027344, + 234.54095458984375, + 216.9737548828125, + 205.64048767089844, + 202.26617431640625 + ], + "bp_grad_norms_per_sample_med": [ + 4.9309845053358003e-05, + 5.373455860535614e-05, + 5.6388864322798327e-05, + 5.448140291264281e-05, + 4.2345352994743735e-05 + ], + "bp_grad_norms_F": [ + 0.0028876049909740686, + 0.0030658405739814043, + 0.0031632501631975174, + 0.002954283496364951, + 0.0020715948194265366 + ], + "gamma_dfa": NaN, + "acc_eval": 0.611328125, + "loss_eval": 1.155066967010498, + "epoch": 88 + }, + { + "hidden_norms": [ + 257.7385559082031, + 233.87037658691406, + 216.7277069091797, + 205.37454223632812, + 202.0509033203125 + ], + "bp_grad_norms_per_sample_med": [ + 5.0028131227009e-05, + 5.394254549173638e-05, + 5.677127046510577e-05, + 5.581674486165866e-05, + 4.2508807382546365e-05 + ], + "bp_grad_norms_F": [ + 0.0028917219024151564, + 0.0030712694860994816, + 0.0031723883002996445, + 0.0029576809611171484, + 0.0020730902906507254 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6083984375, + "loss_eval": 1.1531357765197754, + "epoch": 89 + }, + { + "hidden_norms": [ + 257.52423095703125, + 233.51254272460938, + 216.73731994628906, + 205.52561950683594, + 202.19493103027344 + ], + "bp_grad_norms_per_sample_med": [ + 4.9573794967727736e-05, + 5.4153944802237675e-05, + 5.742486973758787e-05, + 5.468217204906978e-05, + 4.2126855987589806e-05 + ], + "bp_grad_norms_F": [ + 0.0029005431570112705, + 0.0030793119221925735, + 0.003179334569722414, + 0.0029661250300705433, + 0.002075627911835909 + ], + "gamma_dfa": NaN, + "acc_eval": 0.603515625, + "loss_eval": 1.155139684677124, + "epoch": 90 + }, + { + "hidden_norms": [ + 257.5526428222656, + 233.24456787109375, + 216.63601684570312, + 205.40318298339844, + 201.8704376220703 + ], + "bp_grad_norms_per_sample_med": [ + 4.996290954295546e-05, + 5.331254214979708e-05, + 5.576194598688744e-05, + 5.50017248315271e-05, + 4.1135557694360614e-05 + ], + "bp_grad_norms_F": [ + 0.002902554115280509, + 0.003083745948970318, + 0.0031855429988354445, + 0.0029742431361228228, + 0.002079661935567856 + ], + "gamma_dfa": NaN, + "acc_eval": 0.611328125, + "loss_eval": 1.15346360206604, + "epoch": 91 + }, + { + "hidden_norms": [ + 257.3648681640625, + 233.42010498046875, + 216.6300506591797, + 204.97442626953125, + 201.4558868408203 + ], + "bp_grad_norms_per_sample_med": [ + 5.06279211549554e-05, + 5.411348683992401e-05, + 5.6813423725543544e-05, + 5.523188519873656e-05, + 4.197460293653421e-05 + ], + "bp_grad_norms_F": [ + 0.0029165328014642, + 0.003098647342994809, + 0.00319958315230906, + 0.002985388273373246, + 0.002088340697810054 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6015625, + "loss_eval": 1.155455470085144, + "epoch": 92 + }, + { + "hidden_norms": [ + 257.30535888671875, + 233.4147491455078, + 216.60826110839844, + 205.0714874267578, + 201.24066162109375 + ], + "bp_grad_norms_per_sample_med": [ + 5.053050699643791e-05, + 5.4108801123220474e-05, + 5.702309499611147e-05, + 5.574637179961428e-05, + 4.271106809028424e-05 + ], + "bp_grad_norms_F": [ + 0.002917014993727207, + 0.0030978922732174397, + 0.0031987964175641537, + 0.0029851715080440044, + 0.0020893686451017857 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6103515625, + "loss_eval": 1.1536216735839844, + "epoch": 93 + }, + { + "hidden_norms": [ + 257.18878173828125, + 232.8798370361328, + 216.17039489746094, + 204.66229248046875, + 201.07859802246094 + ], + "bp_grad_norms_per_sample_med": [ + 5.010717359255068e-05, + 5.397907079895958e-05, + 5.614722249447368e-05, + 5.553716255235486e-05, + 4.292877929401584e-05 + ], + "bp_grad_norms_F": [ + 0.0029321948532015085, + 0.003115386702120304, + 0.0032163059804588556, + 0.002997474977746606, + 0.002096242969855666 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6123046875, + "loss_eval": 1.1570696830749512, + "epoch": 94 + }, + { + "hidden_norms": [ + 257.0701904296875, + 232.91481018066406, + 216.06378173828125, + 204.648681640625, + 201.00148010253906 + ], + "bp_grad_norms_per_sample_med": [ + 4.9896454584086314e-05, + 5.386451448430307e-05, + 5.5900127335917205e-05, + 5.5407243053196e-05, + 4.247200922691263e-05 + ], + "bp_grad_norms_F": [ + 0.0029298008885234594, + 0.0031122476793825626, + 0.0032105366699397564, + 0.0029950770549476147, + 0.0020956522785127163 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6083984375, + "loss_eval": 1.1554739475250244, + "epoch": 95 + }, + { + "hidden_norms": [ + 257.0519104003906, + 232.83355712890625, + 216.1780242919922, + 204.67999267578125, + 201.0236358642578 + ], + "bp_grad_norms_per_sample_med": [ + 5.054255234426819e-05, + 5.421326932264492e-05, + 5.641246752929874e-05, + 5.574019087362103e-05, + 4.288824857212603e-05 + ], + "bp_grad_norms_F": [ + 0.0029319566674530506, + 0.0031146046239882708, + 0.003213704563677311, + 0.002997281262651086, + 0.0020963428542017937 + ], + "gamma_dfa": NaN, + "acc_eval": 0.607421875, + "loss_eval": 1.1545898914337158, + "epoch": 96 + }, + { + "hidden_norms": [ + 257.1473388671875, + 232.85020446777344, + 216.2519989013672, + 204.68023681640625, + 200.93080139160156 + ], + "bp_grad_norms_per_sample_med": [ + 5.0549009756650776e-05, + 5.3789182857144624e-05, + 5.673874693457037e-05, + 5.580837023444474e-05, + 4.2498151742620394e-05 + ], + "bp_grad_norms_F": [ + 0.0029327664524316788, + 0.0031159960199147463, + 0.003215631702914834, + 0.002998515497893095, + 0.0020966045558452606 + ], + "gamma_dfa": NaN, + "acc_eval": 0.609375, + "loss_eval": 1.1546047925949097, + "epoch": 97 + }, + { + "hidden_norms": [ + 257.09906005859375, + 232.74813842773438, + 216.1462860107422, + 204.6047821044922, + 200.89031982421875 + ], + "bp_grad_norms_per_sample_med": [ + 5.086465898784809e-05, + 5.3742358431918547e-05, + 5.673644773196429e-05, + 5.5752621847204864e-05, + 4.2624127672752365e-05 + ], + "bp_grad_norms_F": [ + 0.0029341024346649647, + 0.003117120126262307, + 0.0032170764170587063, + 0.0029998423997312784, + 0.002097061835229397 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6083984375, + "loss_eval": 1.1545934677124023, + "epoch": 98 + }, + { + "hidden_norms": [ + 257.09326171875, + 232.7879638671875, + 216.1444091796875, + 204.5867919921875, + 200.89596557617188 + ], + "bp_grad_norms_per_sample_med": [ + 5.078025787952356e-05, + 5.40847031516023e-05, + 5.667618097504601e-05, + 5.571893780143e-05, + 4.271125726518221e-05 + ], + "bp_grad_norms_F": [ + 0.0029357182793319225, + 0.0031186225824058056, + 0.003218533471226692, + 0.0030010156333446503, + 0.002097515854984522 + ], + "gamma_dfa": NaN, + "acc_eval": 0.609375, + "loss_eval": 1.154747486114502, + "epoch": 99 + }, + { + "hidden_norms": [ + 257.0869140625, + 232.78091430664062, + 216.12852478027344, + 204.58143615722656, + 200.87417602539062 + ], + "bp_grad_norms_per_sample_med": [ + 5.08715384057723e-05, + 5.4204192565521225e-05, + 5.669457823387347e-05, + 5.573080852627754e-05, + 4.275495302863419e-05 + ], + "bp_grad_norms_F": [ + 0.002935809548944235, + 0.003118706401437521, + 0.0032186671160161495, + 0.003001126926392317, + 0.0020975489169359207 + ], + "gamma_dfa": NaN, + "acc_eval": 0.609375, + "loss_eval": 1.1547397375106812, + "epoch": 100 + } + ], + "dfa_log": [ + { + "hidden_norms": [ + 8.718070030212402, + 8.780964851379395, + 8.812002182006836, + 8.828720092773438, + 8.893179893493652 + ], + "bp_grad_norms_per_sample_med": [ + 0.0010068593546748161, + 0.000997236929833889, + 0.0009934091940522194, + 0.0009870644425973296, + 0.0009829498594626784 + ], + "bp_grad_norms_F": [ + 0.03544403985142708, + 0.03507080301642418, + 0.03477022796869278, + 0.03443896397948265, + 0.03407429903745651 + ], + "gamma_dfa": 0.007406219025142491, + "acc_eval": 0.115234375, + "loss_eval": 2.383650302886963, + "epoch": 0 + }, + { + "hidden_norms": [ + 802.1658935546875, + 3058.3173828125, + 4910.8408203125, + 6491.31494140625, + 7304.1533203125 + ], + "bp_grad_norms_per_sample_med": [ + 2.526000798752648e-06, + 1.3625027577290894e-06, + 1.3747999219049234e-06, + 1.378283172925876e-06, + 1.3869492931917193e-06 + ], + "bp_grad_norms_F": [ + 0.00013306058826856315, + 7.476000610040501e-05, + 7.328101492021233e-05, + 7.293069211300462e-05, + 7.309897773666307e-05 + ], + "gamma_dfa": 0.08989996102172881, + "acc_eval": 0.310546875, + "loss_eval": 1.9283097982406616, + "epoch": 1 + }, + { + "hidden_norms": [ + 1580.1480712890625, + 13320.0517578125, + 22793.435546875, + 32440.041015625, + 37758.640625 + ], + "bp_grad_norms_per_sample_med": [ + 1.1427439403632889e-06, + 3.0901847480890865e-07, + 3.1107475706448895e-07, + 3.1377521736430936e-07, + 3.1722706239634135e-07 + ], + "bp_grad_norms_F": [ + 6.523138290503994e-05, + 2.3476462956750765e-05, + 2.0436000340851024e-05, + 1.96047385543352e-05, + 1.9654715288197622e-05 + ], + "gamma_dfa": 0.0858152944711037, + "acc_eval": 0.2880859375, + "loss_eval": 1.9359606504440308, + "epoch": 2 + }, + { + "hidden_norms": [ + 2343.090576171875, + 32438.32421875, + 58119.84375, + 87185.09375, + 104586.15625 + ], + "bp_grad_norms_per_sample_med": [ + 8.229510513046989e-07, + 1.3056541092737461e-07, + 1.2754330214193033e-07, + 1.2950138739142858e-07, + 1.2936459370394005e-07 + ], + "bp_grad_norms_F": [ + 0.0001579313538968563, + 2.960947858809959e-05, + 1.2371615412121173e-05, + 8.468440682918299e-06, + 7.067524620651966e-06 + ], + "gamma_dfa": 0.08629915304481983, + "acc_eval": 0.3017578125, + "loss_eval": 1.9288904666900635, + "epoch": 3 + }, + { + "hidden_norms": [ + 3129.4423828125, + 62780.5, + 117608.9375, + 182640.09375, + 223504.390625 + ], + "bp_grad_norms_per_sample_med": [ + 6.484323762379063e-07, + 7.041825256237644e-08, + 6.729855783760286e-08, + 6.816127040565334e-08, + 6.818439857170233e-08 + ], + "bp_grad_norms_F": [ + 2.9993034331710078e-05, + 2.8730657959386008e-06, + 2.3409718323819106e-06, + 2.367017486903933e-06, + 2.3597553990839515e-06 + ], + "gamma_dfa": 0.08862219587899745, + "acc_eval": 0.3076171875, + "loss_eval": 1.926429033279419, + "epoch": 4 + }, + { + "hidden_norms": [ + 3942.00390625, + 109015.7890625, + 203168.390625, + 329067.625, + 404167.84375 + ], + "bp_grad_norms_per_sample_med": [ + 5.148021386958135e-07, + 4.3746137379230277e-08, + 4.211498705330996e-08, + 4.271550579915129e-08, + 4.2616289164243426e-08 + ], + "bp_grad_norms_F": [ + 2.426102219033055e-05, + 1.769960135789006e-06, + 1.4341048881760798e-06, + 1.45279318530811e-06, + 1.44924592859752e-06 + ], + "gamma_dfa": 0.0876110177487135, + "acc_eval": 0.328125, + "loss_eval": 1.9219281673431396, + "epoch": 5 + }, + { + "hidden_norms": [ + 4737.880859375, + 173361.203125, + 331121.75, + 547707.8125, + 677363.375 + ], + "bp_grad_norms_per_sample_med": [ + 4.4596529846785415e-07, + 2.929639286719521e-08, + 2.8080462399771022e-08, + 2.8515311001342525e-08, + 2.847208158129888e-08 + ], + "bp_grad_norms_F": [ + 2.1308444047463126e-05, + 1.213497625940363e-06, + 9.622256129659945e-07, + 9.734020522955689e-07, + 9.715020041767275e-07 + ], + "gamma_dfa": 0.08417161786928773, + "acc_eval": 0.2919921875, + "loss_eval": 1.942756175994873, + "epoch": 6 + }, + { + "hidden_norms": [ + 5535.86865234375, + 258337.40625, + 501117.59375, + 823776.9375, + 1026549.875 + ], + "bp_grad_norms_per_sample_med": [ + 3.856409307445574e-07, + 2.2078769035260848e-08, + 2.0866384176088104e-08, + 2.09813144635973e-08, + 2.1003458527957264e-08 + ], + "bp_grad_norms_F": [ + 1.847265775722917e-05, + 8.849666528476519e-07, + 7.187628057181428e-07, + 7.262269718921743e-07, + 7.249743703141576e-07 + ], + "gamma_dfa": 0.08587896963581443, + "acc_eval": 0.2919921875, + "loss_eval": 1.9307396411895752, + "epoch": 7 + }, + { + "hidden_norms": [ + 6304.064453125, + 367445.5625, + 717525.125, + 1200281.0, + 1512872.125 + ], + "bp_grad_norms_per_sample_med": [ + 3.410295903449878e-07, + 1.670872862291617e-08, + 1.5770844186135946e-08, + 1.5891666649281433e-08, + 1.588542275499094e-08 + ], + "bp_grad_norms_F": [ + 1.6556636182940565e-05, + 6.716044254062581e-07, + 5.55544488634041e-07, + 5.606318040918268e-07, + 5.59783927656099e-07 + ], + "gamma_dfa": 0.08515941491350532, + "acc_eval": 0.306640625, + "loss_eval": 1.9279191493988037, + "epoch": 8 + }, + { + "hidden_norms": [ + 7086.48388671875, + 505478.71875, + 997925.8125, + 1708849.375, + 2150479.25 + ], + "bp_grad_norms_per_sample_med": [ + 3.013988703060022e-07, + 1.3297809964285534e-08, + 1.2481795153007624e-08, + 1.262175253202713e-08, + 1.2595491760691857e-08 + ], + "bp_grad_norms_F": [ + 1.4646101590187754e-05, + 5.195778953748231e-07, + 4.4164838186588895e-07, + 4.451969459751126e-07, + 4.4467057591646153e-07 + ], + "gamma_dfa": 0.08757842611521482, + "acc_eval": 0.318359375, + "loss_eval": 1.9144790172576904, + "epoch": 9 + }, + { + "hidden_norms": [ + 7891.7685546875, + 675713.625, + 1320934.25, + 2319375.25, + 2933561.75 + ], + "bp_grad_norms_per_sample_med": [ + 2.7613364750322944e-07, + 1.1096022767276281e-08, + 1.0442049891423721e-08, + 1.0500828651061056e-08, + 1.0485931234427426e-08 + ], + "bp_grad_norms_F": [ + 1.3369051885092631e-05, + 4.295688142974541e-07, + 3.6954631355001766e-07, + 3.72132660686475e-07, + 3.718281220699282e-07 + ], + "gamma_dfa": 0.08759273961186409, + "acc_eval": 0.3037109375, + "loss_eval": 1.928476095199585, + "epoch": 10 + }, + { + "hidden_norms": [ + 8605.7783203125, + 881309.9375, + 1719981.25, + 3048018.25, + 3848896.25 + ], + "bp_grad_norms_per_sample_med": [ + 2.546977952988527e-07, + 9.457612470953336e-09, + 8.959561981214392e-09, + 8.996098976865596e-09, + 8.995014511015142e-09 + ], + "bp_grad_norms_F": [ + 1.2278720532776788e-05, + 3.6486676435742993e-07, + 3.18511354180373e-07, + 3.204301890491479e-07, + 3.2022188634073245e-07 + ], + "gamma_dfa": 0.08920952118933201, + "acc_eval": 0.3017578125, + "loss_eval": 1.9193304777145386, + "epoch": 11 + }, + { + "hidden_norms": [ + 9378.4287109375, + 1120098.75, + 2197340.0, + 3930402.25, + 4945595.0 + ], + "bp_grad_norms_per_sample_med": [ + 2.424853278171213e-07, + 8.285154784459792e-09, + 7.84627474104127e-09, + 7.900293752527432e-09, + 7.906582943917329e-09 + ], + "bp_grad_norms_F": [ + 1.1596725016715936e-05, + 3.161075028401683e-07, + 2.796709850372281e-07, + 2.81156104620095e-07, + 2.810372734529665e-07 + ], + "gamma_dfa": 0.09209982817992568, + "acc_eval": 0.287109375, + "loss_eval": 1.927558422088623, + "epoch": 12 + }, + { + "hidden_norms": [ + 10124.396484375, + 1412441.875, + 2752346.75, + 5010036.0, + 6325289.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.236940588318248e-07, + 7.092171738776187e-09, + 6.815563668993718e-09, + 6.8297678623707725e-09, + 6.830232379684276e-09 + ], + "bp_grad_norms_F": [ + 1.0770681910798885e-05, + 2.7517026524037647e-07, + 2.434409793750092e-07, + 2.4456548430862313e-07, + 2.445065661049739e-07 + ], + "gamma_dfa": 0.0919318727683276, + "acc_eval": 0.2900390625, + "loss_eval": 1.9280297756195068, + "epoch": 13 + }, + { + "hidden_norms": [ + 10873.9169921875, + 1726814.125, + 3365165.0, + 6225292.5, + 7941070.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.0750607632180618e-07, + 6.175024491739123e-09, + 5.938003866390318e-09, + 5.979505779407646e-09, + 5.979241990416995e-09 + ], + "bp_grad_norms_F": [ + 9.84854341368191e-06, + 2.3993459308258025e-07, + 2.1497004354387172e-07, + 2.1585482556929492e-07, + 2.1582005160780682e-07 + ], + "gamma_dfa": 0.09486197168007493, + "acc_eval": 0.2978515625, + "loss_eval": 1.9190003871917725, + "epoch": 14 + }, + { + "hidden_norms": [ + 11534.0087890625, + 2084976.25, + 4091674.75, + 7667605.5, + 9821730.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.9337822720899567e-07, + 5.5730837722478554e-09, + 5.387873702744628e-09, + 5.409793057964407e-09, + 5.415522696949893e-09 + ], + "bp_grad_norms_F": [ + 9.290296475228388e-06, + 2.1226249202754843e-07, + 1.9121620198347955e-07, + 1.9190700584204023e-07, + 1.9189823774468096e-07 + ], + "gamma_dfa": 0.09578572702594101, + "acc_eval": 0.28515625, + "loss_eval": 1.9317920207977295, + "epoch": 15 + }, + { + "hidden_norms": [ + 12231.0380859375, + 2494592.5, + 4969235.5, + 9387796.0, + 12017026.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.7950755193396617e-07, + 4.997119606997558e-09, + 4.8298063326512874e-09, + 4.8657629037052175e-09, + 4.859395552614387e-09 + ], + "bp_grad_norms_F": [ + 8.774502930464223e-06, + 1.8752842834146577e-07, + 1.7071489821773866e-07, + 1.7126839679804107e-07, + 1.7126781415299774e-07 + ], + "gamma_dfa": 0.09745451644994318, + "acc_eval": 0.2978515625, + "loss_eval": 1.927154541015625, + "epoch": 16 + }, + { + "hidden_norms": [ + 12895.0341796875, + 2937750.0, + 5861416.5, + 11095317.0, + 14303387.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.7269709928768862e-07, + 4.551929055907067e-09, + 4.4097947515808755e-09, + 4.41813474694186e-09, + 4.416958798714177e-09 + ], + "bp_grad_norms_F": [ + 8.354931196663529e-06, + 1.7010511044190935e-07, + 1.556661288759642e-07, + 1.5612842219070444e-07, + 1.5612160098044114e-07 + ], + "gamma_dfa": 0.09986255329567939, + "acc_eval": 0.3115234375, + "loss_eval": 1.9158110618591309, + "epoch": 17 + }, + { + "hidden_norms": [ + 13624.58984375, + 3430302.75, + 6892795.0, + 13179784.0, + 16997646.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.642013387481711e-07, + 4.094950600830316e-09, + 4.012986831725129e-09, + 4.022568500516854e-09, + 4.023625876925507e-09 + ], + "bp_grad_norms_F": [ + 7.884121259849053e-06, + 1.54415545239317e-07, + 1.407240830531009e-07, + 1.411058434541701e-07, + 1.4110656820776057e-07 + ], + "gamma_dfa": 0.09882167540490627, + "acc_eval": 0.296875, + "loss_eval": 1.9182151556015015, + "epoch": 18 + }, + { + "hidden_norms": [ + 14323.6708984375, + 3985548.0, + 8076905.5, + 15465362.0, + 20015594.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.551940442823252e-07, + 3.79724474086629e-09, + 3.6876826037257615e-09, + 3.700401984829682e-09, + 3.6983835993709135e-09 + ], + "bp_grad_norms_F": [ + 7.607863608427579e-06, + 1.393282218487002e-07, + 1.283839452526081e-07, + 1.2869591614617093e-07, + 1.2869809040694236e-07 + ], + "gamma_dfa": 0.1002046266803518, + "acc_eval": 0.294921875, + "loss_eval": 1.9388974905014038, + "epoch": 19 + }, + { + "hidden_norms": [ + 14955.4462890625, + 4541631.5, + 9327444.0, + 17893654.0, + 23108114.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.4713883444983367e-07, + 3.464018405452407e-09, + 3.3744134153579353e-09, + 3.379040602879968e-09, + 3.381670721225305e-09 + ], + "bp_grad_norms_F": [ + 7.235442353703547e-06, + 1.2747574373861426e-07, + 1.1753004969250469e-07, + 1.1780696951291247e-07, + 1.1780796427274254e-07 + ], + "gamma_dfa": 0.10065551439765841, + "acc_eval": 0.30859375, + "loss_eval": 1.9343281984329224, + "epoch": 20 + }, + { + "hidden_norms": [ + 15583.8564453125, + 5145675.5, + 10628445.0, + 20547210.0, + 26561584.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.4314694851691456e-07, + 3.170894435911009e-09, + 3.083378885548882e-09, + 3.0910656256821767e-09, + 3.0932543193529227e-09 + ], + "bp_grad_norms_F": [ + 6.938716524018673e-06, + 1.1743370009753562e-07, + 1.0794892091325892e-07, + 1.0819184126376058e-07, + 1.0819199047773509e-07 + ], + "gamma_dfa": 0.10184943513013422, + "acc_eval": 0.3154296875, + "loss_eval": 1.9154099225997925, + "epoch": 21 + }, + { + "hidden_norms": [ + 16234.9208984375, + 5839592.5, + 12062968.0, + 23496628.0, + 30428386.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.379862197836701e-07, + 2.91886981251821e-09, + 2.8514124394973805e-09, + 2.8584146161136914e-09, + 2.860225389866855e-09 + ], + "bp_grad_norms_F": [ + 6.72459282213822e-06, + 1.0762735769276333e-07, + 9.920048427147776e-08, + 9.94118352082296e-08, + 9.941232548271728e-08 + ], + "gamma_dfa": 0.1024534439202398, + "acc_eval": 0.3095703125, + "loss_eval": 1.9225246906280518, + "epoch": 22 + }, + { + "hidden_norms": [ + 16833.765625, + 6566770.0, + 13630788.0, + 26729236.0, + 34665300.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.331996628550769e-07, + 2.688527622751735e-09, + 2.6381341555747895e-09, + 2.6441848710589966e-09, + 2.644968466469777e-09 + ], + "bp_grad_norms_F": [ + 6.5681074374879245e-06, + 9.902939268613409e-08, + 9.131563416531208e-08, + 9.149899682370233e-08, + 9.149922419737777e-08 + ], + "gamma_dfa": 0.10199526121141389, + "acc_eval": 0.30859375, + "loss_eval": 1.9233064651489258, + "epoch": 23 + }, + { + "hidden_norms": [ + 17456.634765625, + 7337101.0, + 15401232.0, + 30171576.0, + 39161244.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.2747005939672817e-07, + 2.4809998500074926e-09, + 2.4282977850731413e-09, + 2.432346990488554e-09, + 2.431272960734532e-09 + ], + "bp_grad_norms_F": [ + 6.5036538217100315e-06, + 9.14754565428666e-08, + 8.484188640522916e-08, + 8.500099824004792e-08, + 8.500632731056612e-08 + ], + "gamma_dfa": 0.10456769005395472, + "acc_eval": 0.30078125, + "loss_eval": 1.9299731254577637, + "epoch": 24 + }, + { + "hidden_norms": [ + 18049.205078125, + 8139419.5, + 17132936.0, + 33801196.0, + 43890588.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.2153788020441425e-07, + 2.2737314253618024e-09, + 2.2332342641817604e-09, + 2.232388940370811e-09, + 2.237900753598865e-09 + ], + "bp_grad_norms_F": [ + 6.099615802668268e-06, + 8.523231542767462e-08, + 7.842250226985925e-08, + 7.855853567662052e-08, + 7.856372263859157e-08 + ], + "gamma_dfa": 0.10287670505931601, + "acc_eval": 0.3232421875, + "loss_eval": 1.923604965209961, + "epoch": 25 + }, + { + "hidden_norms": [ + 18644.9375, + 9035569.0, + 19136636.0, + 37813052.0, + 49194300.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.212478508705317e-07, + 2.0938955014315752e-09, + 2.0560451119422396e-09, + 2.058360148993188e-09, + 2.061209425363586e-09 + ], + "bp_grad_norms_F": [ + 5.810702987218974e-06, + 7.886949049407121e-08, + 7.294480752761956e-08, + 7.307144755941408e-08, + 7.307325233796291e-08 + ], + "gamma_dfa": 0.10249267728067935, + "acc_eval": 0.3037109375, + "loss_eval": 1.9200568199157715, + "epoch": 26 + }, + { + "hidden_norms": [ + 19280.703125, + 9952065.0, + 21071430.0, + 41963900.0, + 54690120.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.1414142875310063e-07, + 1.9668882078605066e-09, + 1.928355919389446e-09, + 1.9296169107008154e-09, + 1.929810311551705e-09 + ], + "bp_grad_norms_F": [ + 5.566818344959756e-06, + 7.357547104902551e-08, + 6.781976935599232e-08, + 6.792832607516175e-08, + 6.79302942785398e-08 + ], + "gamma_dfa": 0.10335587273584679, + "acc_eval": 0.310546875, + "loss_eval": 1.92685067653656, + "epoch": 27 + }, + { + "hidden_norms": [ + 19887.205078125, + 10899353.0, + 23080260.0, + 46142640.0, + 60105148.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.1361055385350483e-07, + 1.8462845696731733e-09, + 1.815088412904231e-09, + 1.8113950339682106e-09, + 1.8156832704008252e-09 + ], + "bp_grad_norms_F": [ + 5.398809662437998e-06, + 6.89453401037099e-08, + 6.4051612014282e-08, + 6.414857267600382e-08, + 6.415204012455433e-08 + ], + "gamma_dfa": 0.10455695656128228, + "acc_eval": 0.302734375, + "loss_eval": 1.923227310180664, + "epoch": 28 + }, + { + "hidden_norms": [ + 20438.587890625, + 11953376.0, + 25259200.0, + 50259240.0, + 65732368.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.1073134231764925e-07, + 1.7497674420496878e-09, + 1.7245866956727696e-09, + 1.7265968654811559e-09, + 1.727246901062074e-09 + ], + "bp_grad_norms_F": [ + 5.281461653794395e-06, + 6.489771209317041e-08, + 6.026093046784808e-08, + 6.034837696233808e-08, + 6.035197941400838e-08 + ], + "gamma_dfa": 0.10511547370697372, + "acc_eval": 0.326171875, + "loss_eval": 1.9300421476364136, + "epoch": 29 + }, + { + "hidden_norms": [ + 20991.0234375, + 13017462.0, + 27475572.0, + 54971628.0, + 71927184.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.0625635837868685e-07, + 1.6529324575742521e-09, + 1.6255721213553898e-09, + 1.6247864165208625e-09, + 1.6244732226056158e-09 + ], + "bp_grad_norms_F": [ + 5.098012934467988e-06, + 6.077872427567854e-08, + 5.674080227890954e-08, + 5.681971870785674e-08, + 5.6821534144546604e-08 + ], + "gamma_dfa": 0.10590779440826736, + "acc_eval": 0.3017578125, + "loss_eval": 1.9211739301681519, + "epoch": 30 + }, + { + "hidden_norms": [ + 21617.34765625, + 14125094.0, + 29881326.0, + 59926964.0, + 78189832.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.0433515029717455e-07, + 1.5471924852406005e-09, + 1.5234034034250499e-09, + 1.5246930384904545e-09, + 1.5239296491387222e-09 + ], + "bp_grad_norms_F": [ + 4.977115622750716e-06, + 5.780140810429657e-08, + 5.362210941939338e-08, + 5.3694908075385683e-08, + 5.369727773540944e-08 + ], + "gamma_dfa": 0.10481705865822732, + "acc_eval": 0.3212890625, + "loss_eval": 1.9209015369415283, + "epoch": 31 + }, + { + "hidden_norms": [ + 22126.13671875, + 15300531.0, + 32579362.0, + 65167536.0, + 85083400.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.0149238249823611e-07, + 1.4551810867402537e-09, + 1.4291762218121562e-09, + 1.432274854273885e-09, + 1.4318950469771607e-09 + ], + "bp_grad_norms_F": [ + 4.9215527724300046e-06, + 5.434217342781267e-08, + 5.088266163966182e-08, + 5.094950239481477e-08, + 5.0953229191463834e-08 + ], + "gamma_dfa": 0.10599250381346792, + "acc_eval": 0.30859375, + "loss_eval": 1.9330241680145264, + "epoch": 32 + }, + { + "hidden_norms": [ + 22662.396484375, + 16481282.0, + 35263600.0, + 70793304.0, + 92166136.0 + ], + "bp_grad_norms_per_sample_med": [ + 9.882934648430819e-08, + 1.3810197430075277e-09, + 1.3668104426045602e-09, + 1.3675354182396404e-09, + 1.366150526038723e-09 + ], + "bp_grad_norms_F": [ + 4.6740801735722926e-06, + 5.137474090588512e-08, + 4.796589792022132e-08, + 4.802501862855024e-08, + 4.8027377630432966e-08 + ], + "gamma_dfa": 0.10529429838061333, + "acc_eval": 0.3115234375, + "loss_eval": 1.9205418825149536, + "epoch": 33 + }, + { + "hidden_norms": [ + 23192.3125, + 17779744.0, + 38108720.0, + 76522688.0, + 99643120.0 + ], + "bp_grad_norms_per_sample_med": [ + 9.530750588737646e-08, + 1.3089170858293642e-09, + 1.2901538726239892e-09, + 1.2906937740808644e-09, + 1.2914372904404559e-09 + ], + "bp_grad_norms_F": [ + 4.669429472414777e-06, + 4.88593805414439e-08, + 4.5740907950175824e-08, + 4.579441181817856e-08, + 4.5798362435789386e-08 + ], + "gamma_dfa": 0.1058788642694708, + "acc_eval": 0.30078125, + "loss_eval": 1.9393997192382812, + "epoch": 34 + }, + { + "hidden_norms": [ + 23683.251953125, + 18992256.0, + 40936536.0, + 82030200.0, + 106929248.0 + ], + "bp_grad_norms_per_sample_med": [ + 9.458356942104729e-08, + 1.2407530558533608e-09, + 1.2231146095942336e-09, + 1.2233244417458877e-09, + 1.2247586278490985e-09 + ], + "bp_grad_norms_F": [ + 4.483116299525136e-06, + 4.651869289773458e-08, + 4.358627236911161e-08, + 4.363666050721804e-08, + 4.3638891611408326e-08 + ], + "gamma_dfa": 0.10620562738040462, + "acc_eval": 0.3251953125, + "loss_eval": 1.9221802949905396, + "epoch": 35 + }, + { + "hidden_norms": [ + 24223.7265625, + 20421224.0, + 43883308.0, + 87849264.0, + 114523200.0 + ], + "bp_grad_norms_per_sample_med": [ + 9.247051480087976e-08, + 1.182353992490448e-09, + 1.1679496259020539e-09, + 1.169668362166476e-09, + 1.1697178781133744e-09 + ], + "bp_grad_norms_F": [ + 4.388592969917227e-06, + 4.4431864409943955e-08, + 4.157495325785021e-08, + 4.162143696362364e-08, + 4.16235543809762e-08 + ], + "gamma_dfa": 0.10484841075958684, + "acc_eval": 0.2998046875, + "loss_eval": 1.9306102991104126, + "epoch": 36 + }, + { + "hidden_norms": [ + 24732.05078125, + 21765008.0, + 46887340.0, + 93818392.0, + 122372552.0 + ], + "bp_grad_norms_per_sample_med": [ + 9.012441637423763e-08, + 1.1452657711075176e-09, + 1.1280866241136778e-09, + 1.1283430856323662e-09, + 1.1287816237270931e-09 + ], + "bp_grad_norms_F": [ + 4.297942268749466e-06, + 4.29258228962226e-08, + 3.9989565436826524e-08, + 4.003141995667647e-08, + 4.003394948881578e-08 + ], + "gamma_dfa": 0.10523941312567331, + "acc_eval": 0.302734375, + "loss_eval": 1.9360214471817017, + "epoch": 37 + }, + { + "hidden_norms": [ + 25179.26953125, + 23094218.0, + 50076040.0, + 100053064.0, + 130605352.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.80345041309738e-08, + 1.0831726626747695e-09, + 1.0690409668612233e-09, + 1.0700328401114234e-09, + 1.0692404739387484e-09 + ], + "bp_grad_norms_F": [ + 4.235343567415839e-06, + 4.072955306355652e-08, + 3.817867622046833e-08, + 3.8218924913735464e-08, + 3.822109917450689e-08 + ], + "gamma_dfa": 0.10568258634884842, + "acc_eval": 0.3193359375, + "loss_eval": 1.927825689315796, + "epoch": 38 + }, + { + "hidden_norms": [ + 25586.23828125, + 24470240.0, + 53013036.0, + 105981368.0, + 138543408.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.783595717432036e-08, + 1.0381262516290235e-09, + 1.0265919225815878e-09, + 1.0273364381419015e-09, + 1.0274008310773297e-09 + ], + "bp_grad_norms_F": [ + 4.177269602223532e-06, + 3.924894187434802e-08, + 3.684685978555535e-08, + 3.688498040332888e-08, + 3.688741045948518e-08 + ], + "gamma_dfa": 0.10561428684741259, + "acc_eval": 0.302734375, + "loss_eval": 1.9321579933166504, + "epoch": 39 + }, + { + "hidden_norms": [ + 26049.974609375, + 25812240.0, + 56223324.0, + 112200904.0, + 146978224.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.555475972116255e-08, + 9.982903392824483e-10, + 9.87055215340149e-10, + 9.8631947054173e-10, + 9.876408579856388e-10 + ], + "bp_grad_norms_F": [ + 4.088252808287507e-06, + 3.767732792425704e-08, + 3.5331073178213046e-08, + 3.536514370239274e-08, + 3.5367978767908426e-08 + ], + "gamma_dfa": 0.10579964506905526, + "acc_eval": 0.318359375, + "loss_eval": 1.930686116218567, + "epoch": 40 + }, + { + "hidden_norms": [ + 26467.328125, + 27200558.0, + 59437036.0, + 118688848.0, + 155514592.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.46039256430231e-08, + 9.647596055373242e-10, + 9.512606258255119e-10, + 9.511659238015113e-10, + 9.522146404705722e-10 + ], + "bp_grad_norms_F": [ + 4.015588729089359e-06, + 3.636354861669133e-08, + 3.413835614196614e-08, + 3.417060767674229e-08, + 3.417307326003538e-08 + ], + "gamma_dfa": 0.10565257369307801, + "acc_eval": 0.314453125, + "loss_eval": 1.9317667484283447, + "epoch": 41 + }, + { + "hidden_norms": [ + 26919.3671875, + 28678974.0, + 62716300.0, + 125549296.0, + 163974208.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.232915860162393e-08, + 9.29719468079071e-10, + 9.197844152986079e-10, + 9.194760508535182e-10, + 9.196678418810222e-10 + ], + "bp_grad_norms_F": [ + 3.951602138840826e-06, + 3.508949220076829e-08, + 3.2971975372220186e-08, + 3.3001178678659926e-08, + 3.300409190387654e-08 + ], + "gamma_dfa": 0.10566475696396083, + "acc_eval": 0.3056640625, + "loss_eval": 1.9284584522247314, + "epoch": 42 + }, + { + "hidden_norms": [ + 27317.033203125, + 30078810.0, + 66046968.0, + 132107504.0, + 172659440.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.130339068657122e-08, + 9.03509489447174e-10, + 8.925334360476711e-10, + 8.917908633776506e-10, + 8.925368222278962e-10 + ], + "bp_grad_norms_F": [ + 3.9174719859147444e-06, + 3.398368519924588e-08, + 3.1965900149089066e-08, + 3.199273024279137e-08, + 3.199540898890518e-08 + ], + "gamma_dfa": 0.10646540904417634, + "acc_eval": 0.3115234375, + "loss_eval": 1.9241154193878174, + "epoch": 43 + }, + { + "hidden_norms": [ + 27715.529296875, + 31383778.0, + 69356640.0, + 138681856.0, + 181338928.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.031172171740764e-08, + 8.735605017129444e-10, + 8.597746958827202e-10, + 8.612066615398817e-10, + 8.621173774869817e-10 + ], + "bp_grad_norms_F": [ + 3.837071290035965e-06, + 3.286310601424702e-08, + 3.101721191001161e-08, + 3.104238288642591e-08, + 3.104513979224066e-08 + ], + "gamma_dfa": 0.10568622383289039, + "acc_eval": 0.2939453125, + "loss_eval": 1.938409447669983, + "epoch": 44 + }, + { + "hidden_norms": [ + 28149.3203125, + 32840858.0, + 72848112.0, + 145667440.0, + 190265088.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.894184506085367e-08, + 8.401777051858517e-10, + 8.299890774665641e-10, + 8.304122944835512e-10, + 8.303523979513727e-10 + ], + "bp_grad_norms_F": [ + 3.798121497311513e-06, + 3.184389640864538e-08, + 3.004597459721481e-08, + 3.0069060130699654e-08, + 3.007191295978373e-08 + ], + "gamma_dfa": 0.10558789351489395, + "acc_eval": 0.3037109375, + "loss_eval": 1.9300565719604492, + "epoch": 45 + }, + { + "hidden_norms": [ + 28522.55078125, + 34358040.0, + 76134080.0, + 152412848.0, + 198973936.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.830858095303483e-08, + 8.109956595170331e-10, + 8.036356025087343e-10, + 8.033995135825478e-10, + 8.031720843959533e-10 + ], + "bp_grad_norms_F": [ + 3.7768336369481403e-06, + 3.096561229654071e-08, + 2.9296373327269976e-08, + 2.9317961391939207e-08, + 2.9320871064442144e-08 + ], + "gamma_dfa": 0.10623026502435096, + "acc_eval": 0.3115234375, + "loss_eval": 1.930029273033142, + "epoch": 46 + }, + { + "hidden_norms": [ + 28924.912109375, + 35808620.0, + 79629336.0, + 159269824.0, + 208087264.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.754207587140627e-08, + 7.940589297206202e-10, + 7.852188899093449e-10, + 7.844204730211857e-10, + 7.861447048895798e-10 + ], + "bp_grad_norms_F": [ + 3.695072564369184e-06, + 3.0030403053160626e-08, + 2.8479737679276695e-08, + 2.849986913133762e-08, + 2.8502654458861798e-08 + ], + "gamma_dfa": 0.1059467513114214, + "acc_eval": 0.30078125, + "loss_eval": 1.9284684658050537, + "epoch": 47 + }, + { + "hidden_norms": [ + 29289.236328125, + 37312204.0, + 83049256.0, + 165977216.0, + 216710016.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.562951509498816e-08, + 7.782112176890621e-10, + 7.675478586044449e-10, + 7.666810519779688e-10, + 7.681388303204528e-10 + ], + "bp_grad_norms_F": [ + 3.6269300380809e-06, + 2.9297559933638695e-08, + 2.782604902051844e-08, + 2.7844757610751003e-08, + 2.7847393724300673e-08 + ], + "gamma_dfa": 0.10581977141555399, + "acc_eval": 0.3037109375, + "loss_eval": 1.9282777309417725, + "epoch": 48 + }, + { + "hidden_norms": [ + 29638.296875, + 38754036.0, + 86627536.0, + 172893504.0, + 225686464.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.50456763398688e-08, + 7.526539391733422e-10, + 7.443530236628249e-10, + 7.439299176681402e-10, + 7.444729277494844e-10 + ], + "bp_grad_norms_F": [ + 3.6355218071548734e-06, + 2.845291824371543e-08, + 2.7127033064289208e-08, + 2.714476821097378e-08, + 2.7147471826083347e-08 + ], + "gamma_dfa": 0.1066790189652238, + "acc_eval": 0.3056640625, + "loss_eval": 1.925389051437378, + "epoch": 49 + }, + { + "hidden_norms": [ + 29950.615234375, + 40150236.0, + 90007640.0, + 179977856.0, + 235206976.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.382724476201474e-08, + 7.370593024802474e-10, + 7.280710478951846e-10, + 7.283330605289962e-10, + 7.279294389483937e-10 + ], + "bp_grad_norms_F": [ + 3.5918935736845015e-06, + 2.7770946431360244e-08, + 2.6450262424759785e-08, + 2.6465910352158062e-08, + 2.6468752523101102e-08 + ], + "gamma_dfa": 0.10654840966162737, + "acc_eval": 0.3076171875, + "loss_eval": 1.929931402206421, + "epoch": 50 + }, + { + "hidden_norms": [ + 30266.310546875, + 41604512.0, + 93401064.0, + 187075808.0, + 244203616.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.343555097349963e-08, + 7.144400626657443e-10, + 7.057154305378788e-10, + 7.05836944447924e-10, + 7.073042707084198e-10 + ], + "bp_grad_norms_F": [ + 3.5134323752572527e-06, + 2.7063506990998576e-08, + 2.580901536930469e-08, + 2.582418012764265e-08, + 2.582707381293403e-08 + ], + "gamma_dfa": 0.10640880587743595, + "acc_eval": 0.314453125, + "loss_eval": 1.9268465042114258, + "epoch": 51 + }, + { + "hidden_norms": [ + 30588.291015625, + 43073320.0, + 96898208.0, + 193783280.0, + 253267328.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.257300893570573e-08, + 6.925075513031231e-10, + 6.871804236752155e-10, + 6.855109813130866e-10, + 6.860176871015256e-10 + ], + "bp_grad_norms_F": [ + 3.4990612221008632e-06, + 2.652852693074692e-08, + 2.5297810068991566e-08, + 2.5311907236869047e-08, + 2.531497678148753e-08 + ], + "gamma_dfa": 0.10629434209840838, + "acc_eval": 0.3125, + "loss_eval": 1.932523488998413, + "epoch": 52 + }, + { + "hidden_norms": [ + 30898.34765625, + 44412940.0, + 100019336.0, + 199966816.0, + 261667760.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.140450719589353e-08, + 6.861456958162648e-10, + 6.789319106914604e-10, + 6.784582895491553e-10, + 6.785109696316738e-10 + ], + "bp_grad_norms_F": [ + 3.470987167020212e-06, + 2.5969489669819268e-08, + 2.478421379237261e-08, + 2.479695204726795e-08, + 2.4800137055080995e-08 + ], + "gamma_dfa": 0.1070190458704019, + "acc_eval": 0.310546875, + "loss_eval": 1.932159662246704, + "epoch": 53 + }, + { + "hidden_norms": [ + 31188.375, + 45820600.0, + 103429992.0, + 206547120.0, + 270181600.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.079076169702603e-08, + 6.688025688816879e-10, + 6.633480986728557e-10, + 6.629659043966285e-10, + 6.630911930649575e-10 + ], + "bp_grad_norms_F": [ + 3.4183981370006222e-06, + 2.540217280966317e-08, + 2.4278467236626966e-08, + 2.4290830680229192e-08, + 2.429379897250783e-08 + ], + "gamma_dfa": 0.10550818023330066, + "acc_eval": 0.2998046875, + "loss_eval": 1.9357173442840576, + "epoch": 54 + }, + { + "hidden_norms": [ + 31478.2109375, + 47178892.0, + 107044784.0, + 213196560.0, + 278895232.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.016877390242371e-08, + 6.609012226377331e-10, + 6.520708972779232e-10, + 6.519841333485488e-10, + 6.530669338644657e-10 + ], + "bp_grad_norms_F": [ + 3.405668167033582e-06, + 2.4965938649756936e-08, + 2.3893584000234114e-08, + 2.390493669679472e-08, + 2.390810749375305e-08 + ], + "gamma_dfa": 0.10673638083972037, + "acc_eval": 0.294921875, + "loss_eval": 1.9359304904937744, + "epoch": 55 + }, + { + "hidden_norms": [ + 31753.5859375, + 48563264.0, + 110454832.0, + 219550528.0, + 287320896.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.007908919831607e-08, + 6.483197867446222e-10, + 6.408613084651904e-10, + 6.410805775125539e-10, + 6.409432429244077e-10 + ], + "bp_grad_norms_F": [ + 3.369002797626308e-06, + 2.447991498399915e-08, + 2.3494733270013057e-08, + 2.350535943662635e-08, + 2.350850003551841e-08 + ], + "gamma_dfa": 0.10726616549072787, + "acc_eval": 0.3076171875, + "loss_eval": 1.9281866550445557, + "epoch": 56 + }, + { + "hidden_norms": [ + 32002.76953125, + 49866268.0, + 113724368.0, + 225759120.0, + 295231328.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.914736161434121e-08, + 6.332004365283694e-10, + 6.232644400583354e-10, + 6.233273341926804e-10, + 6.240686856173738e-10 + ], + "bp_grad_norms_F": [ + 3.3845158213807736e-06, + 2.415495181651295e-08, + 2.3192731291032942e-08, + 2.3202973764568924e-08, + 2.3206132127029377e-08 + ], + "gamma_dfa": 0.10715722179156728, + "acc_eval": 0.3095703125, + "loss_eval": 1.9343407154083252, + "epoch": 57 + }, + { + "hidden_norms": [ + 32277.291015625, + 51206064.0, + 116617608.0, + 231756352.0, + 303064864.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.926612172719615e-08, + 6.240719607752965e-10, + 6.182420131395361e-10, + 6.17405682135086e-10, + 6.178215161689593e-10 + ], + "bp_grad_norms_F": [ + 3.317889877507696e-06, + 2.3704412654979024e-08, + 2.2745311412109004e-08, + 2.2754765183208292e-08, + 2.2757769002623718e-08 + ], + "gamma_dfa": 0.10657632350921631, + "acc_eval": 0.302734375, + "loss_eval": 1.9305057525634766, + "epoch": 58 + }, + { + "hidden_norms": [ + 32521.890625, + 52538716.0, + 119605040.0, + 237773904.0, + 310920672.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.805787933217289e-08, + 6.190028489783117e-10, + 6.10553496649402e-10, + 6.107545580391616e-10, + 6.10922701316241e-10 + ], + "bp_grad_norms_F": [ + 3.2823897981870687e-06, + 2.3313557306892108e-08, + 2.2385792775025948e-08, + 2.239448981811165e-08, + 2.239757002087117e-08 + ], + "gamma_dfa": 0.10671919275773689, + "acc_eval": 0.302734375, + "loss_eval": 1.9269468784332275, + "epoch": 59 + }, + { + "hidden_norms": [ + 32717.80859375, + 53698120.0, + 122676120.0, + 243745136.0, + 318772768.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.75130422678194e-08, + 6.033377686343044e-10, + 5.948920800413759e-10, + 5.949861159315617e-10, + 5.951407144877408e-10 + ], + "bp_grad_norms_F": [ + 3.2588357044005534e-06, + 2.2926835541170476e-08, + 2.205681148836902e-08, + 2.206508220581327e-08, + 2.2068283200837868e-08 + ], + "gamma_dfa": 0.10711025857017376, + "acc_eval": 0.314453125, + "loss_eval": 1.9273767471313477, + "epoch": 60 + }, + { + "hidden_norms": [ + 32950.74609375, + 54885932.0, + 125689616.0, + 249312240.0, + 326217856.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.673791119737871e-08, + 5.923626589243725e-10, + 5.863842189590684e-10, + 5.866296337586618e-10, + 5.870876007563197e-10 + ], + "bp_grad_norms_F": [ + 3.2307948458765168e-06, + 2.2597017590442192e-08, + 2.1738133071380616e-08, + 2.1745805156569986e-08, + 2.1749071876797643e-08 + ], + "gamma_dfa": 0.10673619594308548, + "acc_eval": 0.31640625, + "loss_eval": 1.9271256923675537, + "epoch": 61 + }, + { + "hidden_norms": [ + 33148.1328125, + 55976648.0, + 128534072.0, + 254655232.0, + 333162784.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.663199769718631e-08, + 5.852057172184288e-10, + 5.787633705622852e-10, + 5.792484825128952e-10, + 5.793157620281875e-10 + ], + "bp_grad_norms_F": [ + 3.197097157681128e-06, + 2.2320827852695402e-08, + 2.1484760637235922e-08, + 2.149219824332249e-08, + 2.1495376145708178e-08 + ], + "gamma_dfa": 0.10659754439257085, + "acc_eval": 0.310546875, + "loss_eval": 1.9244905710220337, + "epoch": 62 + }, + { + "hidden_norms": [ + 33369.18359375, + 57164716.0, + 131352720.0, + 259876736.0, + 339964160.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.583201184184873e-08, + 5.737598174349046e-10, + 5.67273561458137e-10, + 5.66656666034504e-10, + 5.665773961105458e-10 + ], + "bp_grad_norms_F": [ + 3.2037539767770795e-06, + 2.2096731555620863e-08, + 2.127841902677119e-08, + 2.1285371687440602e-08, + 2.1288625973170383e-08 + ], + "gamma_dfa": 0.10679806087864563, + "acc_eval": 0.3154296875, + "loss_eval": 1.9308257102966309, + "epoch": 63 + }, + { + "hidden_norms": [ + 33540.88671875, + 58301528.0, + 133722112.0, + 264798368.0, + 346396256.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.606099844930213e-08, + 5.686319748399171e-10, + 5.629098853709991e-10, + 5.619399945366865e-10, + 5.626447086015673e-10 + ], + "bp_grad_norms_F": [ + 3.1530219075648347e-06, + 2.1766210167584177e-08, + 2.0992644067518995e-08, + 2.0999330274662498e-08, + 2.1002534822400776e-08 + ], + "gamma_dfa": 0.10647483140928671, + "acc_eval": 0.3115234375, + "loss_eval": 1.925765037536621, + "epoch": 64 + }, + { + "hidden_norms": [ + 33728.984375, + 59420104.0, + 136277376.0, + 270265568.0, + 353408672.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.57889458466343e-08, + 5.627094901150542e-10, + 5.575621631059846e-10, + 5.567652450189087e-10, + 5.571588745922895e-10 + ], + "bp_grad_norms_F": [ + 3.138873807984055e-06, + 2.1483266721133987e-08, + 2.0736742101234995e-08, + 2.0742897177683517e-08, + 2.0746110607205992e-08 + ], + "gamma_dfa": 0.10677585859230021, + "acc_eval": 0.3154296875, + "loss_eval": 1.9289934635162354, + "epoch": 65 + }, + { + "hidden_norms": [ + 33919.515625, + 60492864.0, + 138846208.0, + 275456896.0, + 359821632.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.451340084367985e-08, + 5.607461717183071e-10, + 5.556192728128906e-10, + 5.553609239150603e-10, + 5.557289628477236e-10 + ], + "bp_grad_norms_F": [ + 3.125772991552367e-06, + 2.1269308092541905e-08, + 2.0518179155715188e-08, + 2.052397540808215e-08, + 2.0527094690692138e-08 + ], + "gamma_dfa": 0.10671760967670707, + "acc_eval": 0.3125, + "loss_eval": 1.9296305179595947, + "epoch": 66 + }, + { + "hidden_norms": [ + 34098.5, + 61508056.0, + 141295472.0, + 280204000.0, + 365860768.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.474934366451635e-08, + 5.519530943409734e-10, + 5.459506735583375e-10, + 5.446967876743258e-10, + 5.452004958605983e-10 + ], + "bp_grad_norms_F": [ + 3.1189638320938684e-06, + 2.1083655710185667e-08, + 2.0349753881987453e-08, + 2.0355370722313637e-08, + 2.0358520202989894e-08 + ], + "gamma_dfa": 0.10652847628807649, + "acc_eval": 0.3046875, + "loss_eval": 1.9298161268234253, + "epoch": 67 + }, + { + "hidden_norms": [ + 34267.96875, + 62529272.0, + 143621744.0, + 284772064.0, + 371754336.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.392743046035321e-08, + 5.466314068058864e-10, + 5.412825743178473e-10, + 5.413390846698007e-10, + 5.416116999334974e-10 + ], + "bp_grad_norms_F": [ + 3.0790877190156607e-06, + 2.0871810946232472e-08, + 2.0159163227617682e-08, + 2.0164401703937074e-08, + 2.0167590264463797e-08 + ], + "gamma_dfa": 0.10652959482831648, + "acc_eval": 0.298828125, + "loss_eval": 1.926388144493103, + "epoch": 68 + }, + { + "hidden_norms": [ + 34420.42578125, + 63473892.0, + 145906128.0, + 289224256.0, + 377581856.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.427427479138714e-08, + 5.377295275721394e-10, + 5.307561612433176e-10, + 5.305922923248829e-10, + 5.308011252758149e-10 + ], + "bp_grad_norms_F": [ + 3.096157115578535e-06, + 2.067219107004803e-08, + 2.0005694878477698e-08, + 2.001087118230771e-08, + 2.001400645212925e-08 + ], + "gamma_dfa": 0.10710431921324925, + "acc_eval": 0.310546875, + "loss_eval": 1.9277684688568115, + "epoch": 69 + }, + { + "hidden_norms": [ + 34573.87890625, + 64328388.0, + 148036320.0, + 293257856.0, + 382751104.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.389419127117435e-08, + 5.373848033229933e-10, + 5.314943485323909e-10, + 5.314927387090052e-10, + 5.31539035009132e-10 + ], + "bp_grad_norms_F": [ + 3.072862682529376e-06, + 2.049388392322271e-08, + 1.98319511923728e-08, + 1.9836669196138246e-08, + 1.9840014076066836e-08 + ], + "gamma_dfa": 0.10700461147644091, + "acc_eval": 0.3076171875, + "loss_eval": 1.9307092428207397, + "epoch": 70 + }, + { + "hidden_norms": [ + 34690.74609375, + 65142160.0, + 150006800.0, + 297046144.0, + 387685792.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.348407310952098e-08, + 5.322000062868426e-10, + 5.271332259582096e-10, + 5.264539360005926e-10, + 5.263872671079639e-10 + ], + "bp_grad_norms_F": [ + 3.055365141335642e-06, + 2.0346906381973895e-08, + 1.9688986441224188e-08, + 1.969357477094036e-08, + 1.969680063496071e-08 + ], + "gamma_dfa": 0.10688555391971022, + "acc_eval": 0.314453125, + "loss_eval": 1.9277830123901367, + "epoch": 71 + }, + { + "hidden_norms": [ + 34818.79296875, + 65891160.0, + 151813680.0, + 300644832.0, + 392342848.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.302482802311715e-08, + 5.310150652526602e-10, + 5.265936020570905e-10, + 5.26124921407245e-10, + 5.26482579754628e-10 + ], + "bp_grad_norms_F": [ + 3.027237426067586e-06, + 2.0201985861945104e-08, + 1.9550000729395833e-08, + 1.955433681644081e-08, + 1.9557578667672715e-08 + ], + "gamma_dfa": 0.10652808679151349, + "acc_eval": 0.2998046875, + "loss_eval": 1.9273805618286133, + "epoch": 72 + }, + { + "hidden_norms": [ + 34951.515625, + 66590476.0, + 153730848.0, + 304118720.0, + 396925408.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.403244157127119e-08, + 5.26472032635894e-10, + 5.205495479110311e-10, + 5.201504782448296e-10, + 5.20502474454787e-10 + ], + "bp_grad_norms_F": [ + 3.0631524623458972e-06, + 2.0121730059941e-08, + 1.9494045488954725e-08, + 1.9498189729461046e-08, + 1.950143335704979e-08 + ], + "gamma_dfa": 0.1070192107144976, + "acc_eval": 0.3017578125, + "loss_eval": 1.934456467628479, + "epoch": 73 + }, + { + "hidden_norms": [ + 35056.31640625, + 67311408.0, + 155368816.0, + 307494752.0, + 401124448.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.303329769252741e-08, + 5.209726539057158e-10, + 5.160313842900166e-10, + 5.156270965755994e-10, + 5.161130411934778e-10 + ], + "bp_grad_norms_F": [ + 3.0127189347695094e-06, + 1.993531562050066e-08, + 1.93110221147208e-08, + 1.9314887467203334e-08, + 1.9318180832783582e-08 + ], + "gamma_dfa": 0.10658633662387729, + "acc_eval": 0.3076171875, + "loss_eval": 1.9260145425796509, + "epoch": 74 + }, + { + "hidden_norms": [ + 35158.68359375, + 67933192.0, + 157091536.0, + 310654048.0, + 405321344.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.339039515523837e-08, + 5.170803230036825e-10, + 5.107546607874269e-10, + 5.104218714357955e-10, + 5.108970468903351e-10 + ], + "bp_grad_norms_F": [ + 3.0164017061906634e-06, + 1.9819051289005074e-08, + 1.921591774589615e-08, + 1.921967651696832e-08, + 1.9223040936822144e-08 + ], + "gamma_dfa": 0.10703902837121859, + "acc_eval": 0.3095703125, + "loss_eval": 1.927987813949585, + "epoch": 75 + }, + { + "hidden_norms": [ + 35270.68359375, + 68510120.0, + 158496816.0, + 313641408.0, + 409130592.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.246813200050383e-08, + 5.133981018090594e-10, + 5.070178166199923e-10, + 5.070903141835004e-10, + 5.071925657240683e-10 + ], + "bp_grad_norms_F": [ + 3.004009158757981e-06, + 1.970923690919335e-08, + 1.910464320076244e-08, + 1.9108172821802327e-08, + 1.9111526583515115e-08 + ], + "gamma_dfa": 0.1070477613247931, + "acc_eval": 0.3193359375, + "loss_eval": 1.9260926246643066, + "epoch": 76 + }, + { + "hidden_norms": [ + 35364.3125, + 69078320.0, + 159830400.0, + 316294976.0, + 412518496.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.217867820623724e-08, + 5.12098419225282e-10, + 5.052269158589695e-10, + 5.058938268298618e-10, + 5.053178431246863e-10 + ], + "bp_grad_norms_F": [ + 2.9913805974501884e-06, + 1.9599063705300068e-08, + 1.900307999846973e-08, + 1.9006522578024487e-08, + 1.9009846141671005e-08 + ], + "gamma_dfa": 0.10688473540358245, + "acc_eval": 0.3076171875, + "loss_eval": 1.926918864250183, + "epoch": 77 + }, + { + "hidden_norms": [ + 35441.375, + 69602704.0, + 161117808.0, + 318818240.0, + 415766208.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.249013040360296e-08, + 5.102671063461628e-10, + 5.039960671027188e-10, + 5.038492956188634e-10, + 5.034740402365401e-10 + ], + "bp_grad_norms_F": [ + 2.9863124382245587e-06, + 1.9512324200832154e-08, + 1.8926254341522508e-08, + 1.8929529943534362e-08, + 1.8932860612608238e-08 + ], + "gamma_dfa": 0.10701000291737728, + "acc_eval": 0.306640625, + "loss_eval": 1.9271972179412842, + "epoch": 78 + }, + { + "hidden_norms": [ + 35514.78125, + 70081152.0, + 162190624.0, + 321009312.0, + 418699392.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.149930698029493e-08, + 5.099357047733122e-10, + 5.055039165036135e-10, + 5.048302331722709e-10, + 5.054044960317583e-10 + ], + "bp_grad_norms_F": [ + 2.969128445329261e-06, + 1.9427163877594467e-08, + 1.883810085701043e-08, + 1.884115441441736e-08, + 1.884450639977331e-08 + ], + "gamma_dfa": 0.10680000087450026, + "acc_eval": 0.3017578125, + "loss_eval": 1.9285144805908203, + "epoch": 79 + }, + { + "hidden_norms": [ + 35582.9765625, + 70499392.0, + 163284976.0, + 323095264.0, + 421363552.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.187113399391819e-08, + 5.05410657769545e-10, + 4.990756696798826e-10, + 4.979930912085706e-10, + 4.985904467069702e-10 + ], + "bp_grad_norms_F": [ + 2.9715449727518717e-06, + 1.9354965630213883e-08, + 1.877567434860339e-08, + 1.8778711918798763e-08, + 1.8782076338652587e-08 + ], + "gamma_dfa": 0.10694103027344681, + "acc_eval": 0.310546875, + "loss_eval": 1.92802095413208, + "epoch": 80 + }, + { + "hidden_norms": [ + 35639.56640625, + 70899536.0, + 164253712.0, + 324923200.0, + 423745824.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.238182237439105e-08, + 5.018311322046998e-10, + 4.970572842211141e-10, + 4.953790155859394e-10, + 4.961450694729308e-10 + ], + "bp_grad_norms_F": [ + 2.971921048811055e-06, + 1.9296171771543413e-08, + 1.872837351868384e-08, + 1.8731368456315067e-08, + 1.873476662694884e-08 + ], + "gamma_dfa": 0.10704115682892734, + "acc_eval": 0.3076171875, + "loss_eval": 1.926428198814392, + "epoch": 81 + }, + { + "hidden_norms": [ + 35689.78515625, + 71256024.0, + 165127008.0, + 326618208.0, + 425949792.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.238117578050151e-08, + 4.972192657604069e-10, + 4.926972718699574e-10, + 4.921834606541609e-10, + 4.923969565417963e-10 + ], + "bp_grad_norms_F": [ + 2.9655700473085744e-06, + 1.9235280035445612e-08, + 1.8671270751724478e-08, + 1.8674217727721043e-08, + 1.8677596358429582e-08 + ], + "gamma_dfa": 0.10696214074414456, + "acc_eval": 0.3056640625, + "loss_eval": 1.924986481666565, + "epoch": 82 + }, + { + "hidden_norms": [ + 35742.58984375, + 71595640.0, + 165912336.0, + 328130016.0, + 427838752.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.213700487478491e-08, + 4.983659041002397e-10, + 4.936055453264032e-10, + 4.941617670617404e-10, + 4.939196274200697e-10 + ], + "bp_grad_norms_F": [ + 2.951217993540922e-06, + 1.919361736213432e-08, + 1.863161003257119e-08, + 1.8634468190725784e-08, + 1.8637845045077484e-08 + ], + "gamma_dfa": 0.10690056857129093, + "acc_eval": 0.3046875, + "loss_eval": 1.9250625371932983, + "epoch": 83 + }, + { + "hidden_norms": [ + 35784.6640625, + 71885704.0, + 166691376.0, + 329546848.0, + 429668704.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.175997668833588e-08, + 4.971995593017198e-10, + 4.919967211414189e-10, + 4.91977791838849e-10, + 4.926699603835516e-10 + ], + "bp_grad_norms_F": [ + 2.9550594717875356e-06, + 1.9144454910247077e-08, + 1.8583381944381472e-08, + 1.858614595562358e-08, + 1.8589524586332118e-08 + ], + "gamma_dfa": 0.10689649073174223, + "acc_eval": 0.3076171875, + "loss_eval": 1.926330327987671, + "epoch": 84 + }, + { + "hidden_norms": [ + 35822.33984375, + 72122744.0, + 167344016.0, + 330816224.0, + 431313312.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.139300978702522e-08, + 4.980582613001161e-10, + 4.935602482269985e-10, + 4.934832542602408e-10, + 4.934415098745148e-10 + ], + "bp_grad_norms_F": [ + 2.9577784061984858e-06, + 1.9112555094125128e-08, + 1.8551357783280764e-08, + 1.8553992120473595e-08, + 1.8557404501962083e-08 + ], + "gamma_dfa": 0.10689723303948995, + "acc_eval": 0.3056640625, + "loss_eval": 1.9307332038879395, + "epoch": 85 + }, + { + "hidden_norms": [ + 35859.390625, + 72358304.0, + 167947920.0, + 331984992.0, + 432823232.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.135296359843778e-08, + 4.947317000514317e-10, + 4.903184525062443e-10, + 4.896648087004962e-10, + 4.905649775288623e-10 + ], + "bp_grad_norms_F": [ + 2.9511427328543505e-06, + 1.906450464161935e-08, + 1.8504644927475056e-08, + 1.8507259724742653e-08, + 1.8510663224446944e-08 + ], + "gamma_dfa": 0.10684622721601045, + "acc_eval": 0.314453125, + "loss_eval": 1.9271869659423828, + "epoch": 86 + }, + { + "hidden_norms": [ + 35888.20703125, + 72553136.0, + 168401376.0, + 332911328.0, + 434057600.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.163116239576993e-08, + 4.935953312745767e-10, + 4.904400219274407e-10, + 4.890517990574494e-10, + 4.891231308867816e-10 + ], + "bp_grad_norms_F": [ + 2.9438597266562283e-06, + 1.902795432329185e-08, + 1.8475304841558682e-08, + 1.8477877006262133e-08, + 1.848129471682114e-08 + ], + "gamma_dfa": 0.10682923735294025, + "acc_eval": 0.30859375, + "loss_eval": 1.9263050556182861, + "epoch": 87 + }, + { + "hidden_norms": [ + 35914.35546875, + 72735848.0, + 168823680.0, + 333752736.0, + 435137376.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.153870657499283e-08, + 4.920966967247864e-10, + 4.894144534084433e-10, + 4.876969939004994e-10, + 4.884483928435657e-10 + ], + "bp_grad_norms_F": [ + 2.9408777209027903e-06, + 1.8999060102942167e-08, + 1.8450597494279464e-08, + 1.8453114591920894e-08, + 1.84565323024799e-08 + ], + "gamma_dfa": 0.10694600266288035, + "acc_eval": 0.3115234375, + "loss_eval": 1.9252078533172607, + "epoch": 88 + }, + { + "hidden_norms": [ + 35936.5546875, + 72887360.0, + 169178624.0, + 334467840.0, + 436061952.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.168602340039797e-08, + 4.908231043820876e-10, + 4.854965318656923e-10, + 4.846267276370497e-10, + 4.847209300606892e-10 + ], + "bp_grad_norms_F": [ + 2.945510459539946e-06, + 1.8982424521141184e-08, + 1.8438125692910035e-08, + 1.844062857969675e-08, + 1.8444072935608347e-08 + ], + "gamma_dfa": 0.10697454003093299, + "acc_eval": 0.3134765625, + "loss_eval": 1.9264906644821167, + "epoch": 89 + }, + { + "hidden_norms": [ + 35955.3671875, + 73005808.0, + 169498240.0, + 335085920.0, + 436862656.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.152810527737529e-08, + 4.922047214250824e-10, + 4.872572900715966e-10, + 4.873856318532432e-10, + 4.87353157829773e-10 + ], + "bp_grad_norms_F": [ + 2.9327068205020623e-06, + 1.8955052638602865e-08, + 1.8409036073308016e-08, + 1.8411496327530585e-08, + 1.841491403808959e-08 + ], + "gamma_dfa": 0.10689280136284651, + "acc_eval": 0.3095703125, + "loss_eval": 1.9255037307739258, + "epoch": 90 + }, + { + "hidden_norms": [ + 35973.34375, + 73111368.0, + 169770208.0, + 335584768.0, + 437507744.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.107820382794671e-08, + 4.92519802719471e-10, + 4.876242187812352e-10, + 4.871342773604681e-10, + 4.873002001914983e-10 + ], + "bp_grad_norms_F": [ + 2.9368579816946294e-06, + 1.8944499302619988e-08, + 1.839602603581625e-08, + 1.8398441881117833e-08, + 1.8401848933535803e-08 + ], + "gamma_dfa": 0.10685028225998394, + "acc_eval": 0.306640625, + "loss_eval": 1.9265596866607666, + "epoch": 91 + }, + { + "hidden_norms": [ + 35985.046875, + 73193336.0, + 169984976.0, + 335981920.0, + 438037376.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.100844984757714e-08, + 4.92902163529152e-10, + 4.862484304091197e-10, + 4.871591463562197e-10, + 4.871424374996991e-10 + ], + "bp_grad_norms_F": [ + 2.930574055426405e-06, + 1.8926412437281215e-08, + 1.8379486377284593e-08, + 1.8381888011731462e-08, + 1.8385312827717826e-08 + ], + "gamma_dfa": 0.106783474504482, + "acc_eval": 0.30859375, + "loss_eval": 1.9269328117370605, + "epoch": 92 + }, + { + "hidden_norms": [ + 35994.07421875, + 73256784.0, + 170148048.0, + 336300032.0, + 438460544.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.113810968599864e-08, + 4.918689899824358e-10, + 4.860293834063611e-10, + 4.857191315821296e-10, + 4.859731506101639e-10 + ], + "bp_grad_norms_F": [ + 2.9313273444131482e-06, + 1.8918141719836967e-08, + 1.8372302790226058e-08, + 1.837469731924557e-08, + 1.8378123911588773e-08 + ], + "gamma_dfa": 0.10685011067107553, + "acc_eval": 0.3095703125, + "loss_eval": 1.9262909889221191, + "epoch": 93 + }, + { + "hidden_norms": [ + 36002.55078125, + 73308792.0, + 170289472.0, + 336572864.0, + 438812896.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.121508988599089e-08, + 4.917221074762779e-10, + 4.853899504553283e-10, + 4.847242052186118e-10, + 4.849394219519354e-10 + ], + "bp_grad_norms_F": [ + 2.933722271336592e-06, + 1.8911242349872737e-08, + 1.8366764109600808e-08, + 1.836915508590664e-08, + 1.8372583454606684e-08 + ], + "gamma_dfa": 0.10693995938345324, + "acc_eval": 0.3076171875, + "loss_eval": 1.9261877536773682, + "epoch": 94 + }, + { + "hidden_norms": [ + 36008.01953125, + 73348264.0, + 170386624.0, + 336761248.0, + 439049248.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.123732987362018e-08, + 4.916126949972011e-10, + 4.860848945575924e-10, + 4.87389184566922e-10, + 4.868180303319036e-10 + ], + "bp_grad_norms_F": [ + 2.9306668238859857e-06, + 1.8901717524499873e-08, + 1.835805107930355e-08, + 1.8360427844754668e-08, + 1.836385976616839e-08 + ], + "gamma_dfa": 0.10689331469620811, + "acc_eval": 0.30859375, + "loss_eval": 1.9260027408599854, + "epoch": 95 + }, + { + "hidden_norms": [ + 36011.9296875, + 73372720.0, + 170447072.0, + 336878464.0, + 439192800.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.115104156378948e-08, + 4.911633877391353e-10, + 4.862875657707377e-10, + 4.868677128122556e-10, + 4.867456437906981e-10 + ], + "bp_grad_norms_F": [ + 2.9297382297954755e-06, + 1.889710254943111e-08, + 1.8353631503487122e-08, + 1.835600293986772e-08, + 1.8359434861281443e-08 + ], + "gamma_dfa": 0.1068731502891751, + "acc_eval": 0.306640625, + "loss_eval": 1.9261021614074707, + "epoch": 96 + }, + { + "hidden_norms": [ + 36014.109375, + 73385416.0, + 170481184.0, + 336945120.0, + 439280128.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.12362214269524e-08, + 4.916780316222003e-10, + 4.858832225451692e-10, + 4.867402036978774e-10, + 4.862423241824843e-10 + ], + "bp_grad_norms_F": [ + 2.930293021563557e-06, + 1.8895260467388653e-08, + 1.8352444897118403e-08, + 1.835481810985584e-08, + 1.8358248254912723e-08 + ], + "gamma_dfa": 0.10688256371940952, + "acc_eval": 0.3095703125, + "loss_eval": 1.926184892654419, + "epoch": 97 + }, + { + "hidden_norms": [ + 36015.9375, + 73396752.0, + 170510784.0, + 337001056.0, + 439352832.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.120947659837839e-08, + 4.916590468084792e-10, + 4.857179658479538e-10, + 4.86417739420375e-10, + 4.85919415815772e-10 + ], + "bp_grad_norms_F": [ + 2.930209348050994e-06, + 1.889360845552801e-08, + 1.835070939648631e-08, + 1.8353080832866908e-08, + 1.835651275428063e-08 + ], + "gamma_dfa": 0.10688149025372695, + "acc_eval": 0.306640625, + "loss_eval": 1.9262256622314453, + "epoch": 98 + }, + { + "hidden_norms": [ + 36016.734375, + 73401888.0, + 170523312.0, + 337026592.0, + 439384864.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.119820028516187e-08, + 4.917350415745148e-10, + 4.857160784688119e-10, + 4.860361002556601e-10, + 4.855379431845108e-10 + ], + "bp_grad_norms_F": [ + 2.9303794235602254e-06, + 1.8893027586841526e-08, + 1.835026353091962e-08, + 1.835263319094338e-08, + 1.83560651123571e-08 + ], + "gamma_dfa": 0.10688879482768243, + "acc_eval": 0.3076171875, + "loss_eval": 1.926223874092102, + "epoch": 99 + }, + { + "hidden_norms": [ + 36016.7734375, + 73401992.0, + 170523904.0, + 337027392.0, + 439386048.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.120198747794348e-08, + 4.91758134213427e-10, + 4.857037549932386e-10, + 4.860785107752008e-10, + 4.855802981929003e-10 + ], + "bp_grad_norms_F": [ + 2.9303721476026112e-06, + 1.8893056008550957e-08, + 1.8350315045267962e-08, + 1.8352682928934883e-08, + 1.8356114850348604e-08 + ], + "gamma_dfa": 0.1068887785077095, + "acc_eval": 0.3076171875, + "loss_eval": 1.926220417022705, + "epoch": 100 + } + ] +}
\ No newline at end of file diff --git a/results/snapshot_evolution_v2/snapshot_evolution_s456.json b/results/snapshot_evolution_v2/snapshot_evolution_s456.json new file mode 100644 index 0000000..f78065f --- /dev/null +++ b/results/snapshot_evolution_v2/snapshot_evolution_s456.json @@ -0,0 +1,5473 @@ +{ + "config": { + "output_dir": "results/snapshot_evolution_v2", + "epochs": 100, + "lr": 0.001, + "wd": 0.01, + "seed": 456, + "depth": 4, + "d_hidden": 256, + "log_every": 1 + }, + "depth": 4, + "d_hidden": 256, + "num_classes": 10, + "bp_log": [ + { + "hidden_norms": [ + 8.696551322937012, + 8.737784385681152, + 8.813130378723145, + 8.844642639160156, + 8.889528274536133 + ], + "bp_grad_norms_per_sample_med": [ + 0.0010079912608489394, + 0.0010035138111561537, + 0.0009970925748348236, + 0.0009936393471434712, + 0.000986678060144186 + ], + "bp_grad_norms_F": [ + 0.035574451088905334, + 0.035288579761981964, + 0.035048045217990875, + 0.0346725769340992, + 0.03439188003540039 + ], + "gamma_dfa": NaN, + "acc_eval": 0.0927734375, + "loss_eval": 2.4863264560699463, + "epoch": 0 + }, + { + "hidden_norms": [ + 79.9596176147461, + 82.38219451904297, + 86.97193145751953, + 92.46615600585938, + 99.17337799072266 + ], + "bp_grad_norms_per_sample_med": [ + 5.704927025362849e-05, + 5.414035331341438e-05, + 5.3104751714272425e-05, + 5.354730456019752e-05, + 5.5780339607736096e-05 + ], + "bp_grad_norms_F": [ + 0.00240120361559093, + 0.0021707164123654366, + 0.002052225871011615, + 0.0020026240963488817, + 0.002049861242994666 + ], + "gamma_dfa": NaN, + "acc_eval": 0.400390625, + "loss_eval": 1.6942739486694336, + "epoch": 1 + }, + { + "hidden_norms": [ + 95.21885681152344, + 97.9500732421875, + 101.84932708740234, + 108.15538787841797, + 115.16011810302734 + ], + "bp_grad_norms_per_sample_med": [ + 4.467168764676899e-05, + 4.3202508095419034e-05, + 4.3004638428101316e-05, + 4.286228431737982e-05, + 4.4292191887507215e-05 + ], + "bp_grad_norms_F": [ + 0.0018949288642033935, + 0.0017146588070318103, + 0.0016399234300479293, + 0.001598119386471808, + 0.0016228299355134368 + ], + "gamma_dfa": NaN, + "acc_eval": 0.4599609375, + "loss_eval": 1.5051867961883545, + "epoch": 2 + }, + { + "hidden_norms": [ + 112.65824890136719, + 111.87681579589844, + 114.76032257080078, + 118.74363708496094, + 125.05441284179688 + ], + "bp_grad_norms_per_sample_med": [ + 4.4033156882505864e-05, + 4.2794465116458014e-05, + 4.2031828343169764e-05, + 4.134616028750315e-05, + 4.226301462040283e-05 + ], + "bp_grad_norms_F": [ + 0.0018527565989643335, + 0.0017638156423345208, + 0.0016669610049575567, + 0.0015881577273830771, + 0.0015705813420936465 + ], + "gamma_dfa": NaN, + "acc_eval": 0.4931640625, + "loss_eval": 1.4331163167953491, + "epoch": 3 + }, + { + "hidden_norms": [ + 126.20967102050781, + 125.90274810791016, + 128.2918243408203, + 131.685302734375, + 136.71499633789062 + ], + "bp_grad_norms_per_sample_med": [ + 4.358301157481037e-05, + 4.267259282642044e-05, + 4.196130612399429e-05, + 4.08500200137496e-05, + 4.0281276596942917e-05 + ], + "bp_grad_norms_F": [ + 0.0018887611804530025, + 0.0017927174922078848, + 0.0016837789444252849, + 0.0015876276884227991, + 0.0015263468958437443 + ], + "gamma_dfa": NaN, + "acc_eval": 0.490234375, + "loss_eval": 1.4055290222167969, + "epoch": 4 + }, + { + "hidden_norms": [ + 137.673583984375, + 136.8435516357422, + 138.64651489257812, + 141.41473388671875, + 147.87835693359375 + ], + "bp_grad_norms_per_sample_med": [ + 4.085261025466025e-05, + 4.06657018174883e-05, + 3.941181785194203e-05, + 3.858502532239072e-05, + 3.7998870539013296e-05 + ], + "bp_grad_norms_F": [ + 0.001770049100741744, + 0.0016922445502132177, + 0.0016029543476179242, + 0.0015241446672007442, + 0.0014664788031950593 + ], + "gamma_dfa": NaN, + "acc_eval": 0.50390625, + "loss_eval": 1.3763911724090576, + "epoch": 5 + }, + { + "hidden_norms": [ + 150.45315551757812, + 148.60609436035156, + 150.3750457763672, + 152.9460906982422, + 158.2902069091797 + ], + "bp_grad_norms_per_sample_med": [ + 3.8344696804415435e-05, + 3.8422465877374634e-05, + 3.834897142951377e-05, + 3.7070902180857956e-05, + 3.6289642594056204e-05 + ], + "bp_grad_norms_F": [ + 0.0017990845954045653, + 0.001743242726661265, + 0.0016358466818928719, + 0.0015305348206311464, + 0.0014392968732863665 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5166015625, + "loss_eval": 1.373401165008545, + "epoch": 6 + }, + { + "hidden_norms": [ + 162.06126403808594, + 159.53811645507812, + 158.62655639648438, + 160.54307556152344, + 165.60415649414062 + ], + "bp_grad_norms_per_sample_med": [ + 3.797870522248559e-05, + 3.766914960579015e-05, + 3.7873480323469266e-05, + 3.735603968380019e-05, + 3.6521956644719467e-05 + ], + "bp_grad_norms_F": [ + 0.001707478309981525, + 0.0016470234841108322, + 0.0015834521036595106, + 0.001506388420239091, + 0.001406461582519114 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5283203125, + "loss_eval": 1.3223559856414795, + "epoch": 7 + }, + { + "hidden_norms": [ + 169.8936309814453, + 168.42221069335938, + 169.54115295410156, + 171.452880859375, + 176.4298553466797 + ], + "bp_grad_norms_per_sample_med": [ + 3.760562321986072e-05, + 3.8139285607030615e-05, + 3.7332967622205615e-05, + 3.670771184260957e-05, + 3.4867567592300475e-05 + ], + "bp_grad_norms_F": [ + 0.001674071652814746, + 0.0016396756982430816, + 0.001556537114083767, + 0.001464220928028226, + 0.0013669237960129976 + ], + "gamma_dfa": NaN, + "acc_eval": 0.529296875, + "loss_eval": 1.3310301303863525, + "epoch": 8 + }, + { + "hidden_norms": [ + 177.821044921875, + 176.60687255859375, + 177.76409912109375, + 179.53733825683594, + 184.176513671875 + ], + "bp_grad_norms_per_sample_med": [ + 3.643108357209712e-05, + 3.605252277338877e-05, + 3.6002034903503954e-05, + 3.529640525812283e-05, + 3.386645403224975e-05 + ], + "bp_grad_norms_F": [ + 0.0016080558998510242, + 0.0015692447777837515, + 0.0014910849276930094, + 0.0014210963854566216, + 0.0013221093686297536 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5390625, + "loss_eval": 1.3072502613067627, + "epoch": 9 + }, + { + "hidden_norms": [ + 188.1353759765625, + 184.39834594726562, + 184.91822814941406, + 185.82460021972656, + 191.45692443847656 + ], + "bp_grad_norms_per_sample_med": [ + 3.544713035807945e-05, + 3.5891316656488925e-05, + 3.613864828366786e-05, + 3.507022120174952e-05, + 3.361281778779812e-05 + ], + "bp_grad_norms_F": [ + 0.0015651460271328688, + 0.0015296414494514465, + 0.0014787326799705625, + 0.0014131872449070215, + 0.0013114992761984468 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5537109375, + "loss_eval": 1.2935423851013184, + "epoch": 10 + }, + { + "hidden_norms": [ + 193.53599548339844, + 189.33700561523438, + 190.08908081054688, + 192.31288146972656, + 198.0936737060547 + ], + "bp_grad_norms_per_sample_med": [ + 3.3899334084708244e-05, + 3.468272552709095e-05, + 3.463856774033047e-05, + 3.383589137229137e-05, + 3.240162550355308e-05 + ], + "bp_grad_norms_F": [ + 0.001495138043537736, + 0.0014781414065510035, + 0.0014301009941846132, + 0.0013711795909330249, + 0.0012839946430176497 + ], + "gamma_dfa": NaN, + "acc_eval": 0.544921875, + "loss_eval": 1.2766207456588745, + "epoch": 11 + }, + { + "hidden_norms": [ + 199.85023498535156, + 194.9699249267578, + 193.4657440185547, + 194.60928344726562, + 199.97662353515625 + ], + "bp_grad_norms_per_sample_med": [ + 3.289094092906453e-05, + 3.539410317898728e-05, + 3.491900861263275e-05, + 3.4397991839796305e-05, + 3.252939859521575e-05 + ], + "bp_grad_norms_F": [ + 0.0015656249597668648, + 0.0015614606672897935, + 0.0015110056847333908, + 0.0014306820230558515, + 0.00130200176499784 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5439453125, + "loss_eval": 1.2812860012054443, + "epoch": 12 + }, + { + "hidden_norms": [ + 206.15440368652344, + 199.86099243164062, + 199.35470581054688, + 201.66831970214844, + 206.71054077148438 + ], + "bp_grad_norms_per_sample_med": [ + 3.430957076488994e-05, + 3.5125638532917947e-05, + 3.523785198922269e-05, + 3.404217568458989e-05, + 3.245573316235095e-05 + ], + "bp_grad_norms_F": [ + 0.0015102275647222996, + 0.0015037795528769493, + 0.001468595233745873, + 0.0013989906292408705, + 0.0012804438592866063 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5517578125, + "loss_eval": 1.2559274435043335, + "epoch": 13 + }, + { + "hidden_norms": [ + 209.39404296875, + 203.8543243408203, + 201.63771057128906, + 202.7413787841797, + 209.64585876464844 + ], + "bp_grad_norms_per_sample_med": [ + 3.3599189919186756e-05, + 3.511353497742675e-05, + 3.4795291867339984e-05, + 3.415898390812799e-05, + 3.225719046895392e-05 + ], + "bp_grad_norms_F": [ + 0.0015785910654813051, + 0.0015854343073442578, + 0.0015357385855168104, + 0.00144979078322649, + 0.0013052559224888682 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5478515625, + "loss_eval": 1.2778449058532715, + "epoch": 14 + }, + { + "hidden_norms": [ + 214.74293518066406, + 208.2928924560547, + 206.78591918945312, + 207.79452514648438, + 213.0249481201172 + ], + "bp_grad_norms_per_sample_med": [ + 3.347836172906682e-05, + 3.457088314462453e-05, + 3.509481393848546e-05, + 3.4477459848858416e-05, + 3.169552292092703e-05 + ], + "bp_grad_norms_F": [ + 0.0015018024714663625, + 0.0015100837917998433, + 0.0014814585447311401, + 0.0014098759274929762, + 0.0012745971325784922 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5693359375, + "loss_eval": 1.2316113710403442, + "epoch": 15 + }, + { + "hidden_norms": [ + 223.09906005859375, + 214.6353302001953, + 211.16726684570312, + 210.60520935058594, + 214.28952026367188 + ], + "bp_grad_norms_per_sample_med": [ + 3.5336081054992974e-05, + 3.693878170452081e-05, + 3.7104789953446016e-05, + 3.635902612586506e-05, + 3.3355980121996254e-05 + ], + "bp_grad_norms_F": [ + 0.0015461220173165202, + 0.0015768769662827253, + 0.0015498694265261292, + 0.001453379518352449, + 0.0013068238040432334 + ], + "gamma_dfa": NaN, + "acc_eval": 0.55078125, + "loss_eval": 1.2516858577728271, + "epoch": 16 + }, + { + "hidden_norms": [ + 225.36517333984375, + 217.36669921875, + 214.27880859375, + 215.62054443359375, + 219.76197814941406 + ], + "bp_grad_norms_per_sample_med": [ + 3.474476397968829e-05, + 3.609717168728821e-05, + 3.612042564782314e-05, + 3.547673986759037e-05, + 3.216006007278338e-05 + ], + "bp_grad_norms_F": [ + 0.001510953065007925, + 0.0015347315929830074, + 0.0015160846523940563, + 0.001440546941012144, + 0.0012782461708411574 + ], + "gamma_dfa": NaN, + "acc_eval": 0.583984375, + "loss_eval": 1.2262325286865234, + "epoch": 17 + }, + { + "hidden_norms": [ + 230.7099151611328, + 222.3380889892578, + 218.67819213867188, + 217.3147735595703, + 221.999755859375 + ], + "bp_grad_norms_per_sample_med": [ + 3.285402272013016e-05, + 3.4367294574622065e-05, + 3.490627204882912e-05, + 3.413260492379777e-05, + 3.064925112994388e-05 + ], + "bp_grad_norms_F": [ + 0.0015693235909566283, + 0.0015950833912938833, + 0.0015582602936774492, + 0.001467122812755406, + 0.0012723723193630576 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5751953125, + "loss_eval": 1.2127145528793335, + "epoch": 18 + }, + { + "hidden_norms": [ + 235.28932189941406, + 226.01202392578125, + 222.56756591796875, + 222.25152587890625, + 226.0247344970703 + ], + "bp_grad_norms_per_sample_med": [ + 3.336059671710245e-05, + 3.488900983938947e-05, + 3.518831363180652e-05, + 3.5047873097937554e-05, + 3.1955056329024956e-05 + ], + "bp_grad_norms_F": [ + 0.0015264152316376567, + 0.0015387749299407005, + 0.0015209508128464222, + 0.001451890799216926, + 0.00127940543461591 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5673828125, + "loss_eval": 1.2340281009674072, + "epoch": 19 + }, + { + "hidden_norms": [ + 239.98486328125, + 230.95614624023438, + 226.7176513671875, + 225.3818359375, + 231.0238037109375 + ], + "bp_grad_norms_per_sample_med": [ + 3.4207379940198734e-05, + 3.5167759051546454e-05, + 3.5256896808277816e-05, + 3.462574022705667e-05, + 3.117474989267066e-05 + ], + "bp_grad_norms_F": [ + 0.0015223841182887554, + 0.0015598630998283625, + 0.0015445069875568151, + 0.0014689662493765354, + 0.0012761331163346767 + ], + "gamma_dfa": NaN, + "acc_eval": 0.56640625, + "loss_eval": 1.2441283464431763, + "epoch": 20 + }, + { + "hidden_norms": [ + 245.0663299560547, + 233.2012939453125, + 227.60426330566406, + 226.96006774902344, + 229.30145263671875 + ], + "bp_grad_norms_per_sample_med": [ + 3.25386063195765e-05, + 3.39963698934298e-05, + 3.503257903503254e-05, + 3.4624255931703374e-05, + 3.013665264006704e-05 + ], + "bp_grad_norms_F": [ + 0.001554132904857397, + 0.0015968671068549156, + 0.001575724920257926, + 0.0014765377854928374, + 0.0012655528262257576 + ], + "gamma_dfa": NaN, + "acc_eval": 0.595703125, + "loss_eval": 1.1905839443206787, + "epoch": 21 + }, + { + "hidden_norms": [ + 246.93258666992188, + 235.25485229492188, + 232.71095275878906, + 231.27796936035156, + 235.4611358642578 + ], + "bp_grad_norms_per_sample_med": [ + 3.248557914048433e-05, + 3.448772622505203e-05, + 3.525316424202174e-05, + 3.4218086511828005e-05, + 2.997798401338514e-05 + ], + "bp_grad_norms_F": [ + 0.001511804643087089, + 0.0015418389812111855, + 0.001536028808914125, + 0.0014609270729124546, + 0.001262559788301587 + ], + "gamma_dfa": NaN, + "acc_eval": 0.576171875, + "loss_eval": 1.2032543420791626, + "epoch": 22 + }, + { + "hidden_norms": [ + 251.2838592529297, + 237.0211639404297, + 234.6432342529297, + 233.30433654785156, + 238.11598205566406 + ], + "bp_grad_norms_per_sample_med": [ + 3.3030770282493904e-05, + 3.484229455352761e-05, + 3.59142868546769e-05, + 3.533583367243409e-05, + 3.1214847695082426e-05 + ], + "bp_grad_norms_F": [ + 0.0015120860189199448, + 0.00155613012611866, + 0.0015533699188381433, + 0.001477457000873983, + 0.0012758751399815083 + ], + "gamma_dfa": NaN, + "acc_eval": 0.568359375, + "loss_eval": 1.220862627029419, + "epoch": 23 + }, + { + "hidden_norms": [ + 253.3640594482422, + 242.6001739501953, + 237.11773681640625, + 237.31105041503906, + 241.1314697265625 + ], + "bp_grad_norms_per_sample_med": [ + 3.295885835541412e-05, + 3.492692485451698e-05, + 3.562758502084762e-05, + 3.515293064992875e-05, + 3.0970571970101446e-05 + ], + "bp_grad_norms_F": [ + 0.001516172313131392, + 0.0015499184373766184, + 0.0015383078716695309, + 0.0014658188447356224, + 0.0012659834465011954 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5888671875, + "loss_eval": 1.1925604343414307, + "epoch": 24 + }, + { + "hidden_norms": [ + 258.6325378417969, + 245.05575561523438, + 239.04722595214844, + 237.04385375976562, + 240.51011657714844 + ], + "bp_grad_norms_per_sample_med": [ + 3.258220021962188e-05, + 3.459660729276948e-05, + 3.553628994268365e-05, + 3.461805317783728e-05, + 3.001722870976664e-05 + ], + "bp_grad_norms_F": [ + 0.0015354488277807832, + 0.001572331297211349, + 0.0015721773961558938, + 0.0014915125211700797, + 0.0012740870006382465 + ], + "gamma_dfa": NaN, + "acc_eval": 0.58203125, + "loss_eval": 1.1843492984771729, + "epoch": 25 + }, + { + "hidden_norms": [ + 262.35577392578125, + 247.456298828125, + 241.52687072753906, + 239.8603057861328, + 242.48818969726562 + ], + "bp_grad_norms_per_sample_med": [ + 3.335328074172139e-05, + 3.581988130463287e-05, + 3.6700876080431044e-05, + 3.52635579474736e-05, + 3.091046164627187e-05 + ], + "bp_grad_norms_F": [ + 0.0015506797935813665, + 0.0016048819525167346, + 0.0016078166663646698, + 0.0015189298428595066, + 0.001287299208343029 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6025390625, + "loss_eval": 1.2021183967590332, + "epoch": 26 + }, + { + "hidden_norms": [ + 262.69769287109375, + 248.58924865722656, + 243.61415100097656, + 242.7408905029297, + 243.58761596679688 + ], + "bp_grad_norms_per_sample_med": [ + 3.363215728313662e-05, + 3.542406557244249e-05, + 3.617940456024371e-05, + 3.522956103552133e-05, + 3.093081613769755e-05 + ], + "bp_grad_norms_F": [ + 0.0015515872510150075, + 0.0015918170101940632, + 0.0015870564384385943, + 0.0015067857457324862, + 0.001285396981984377 + ], + "gamma_dfa": NaN, + "acc_eval": 0.576171875, + "loss_eval": 1.1753299236297607, + "epoch": 27 + }, + { + "hidden_norms": [ + 264.57330322265625, + 252.62904357910156, + 246.43833923339844, + 245.4523468017578, + 247.43896484375 + ], + "bp_grad_norms_per_sample_med": [ + 3.3458851248724386e-05, + 3.510742317303084e-05, + 3.596700116759166e-05, + 3.530932735884562e-05, + 3.0871200578985736e-05 + ], + "bp_grad_norms_F": [ + 0.0015561054460704327, + 0.0015992044936865568, + 0.0016016490990296006, + 0.0015203922521322966, + 0.001274992828257382 + ], + "gamma_dfa": NaN, + "acc_eval": 0.587890625, + "loss_eval": 1.1872658729553223, + "epoch": 28 + }, + { + "hidden_norms": [ + 268.63336181640625, + 253.74609375, + 247.2230224609375, + 244.07066345214844, + 248.91635131835938 + ], + "bp_grad_norms_per_sample_med": [ + 3.388441837159917e-05, + 3.55305674020201e-05, + 3.711709723575041e-05, + 3.565939550753683e-05, + 3.06438232655637e-05 + ], + "bp_grad_norms_F": [ + 0.0015880814753472805, + 0.001632494037039578, + 0.0016352345701307058, + 0.001542495796456933, + 0.001282386132515967 + ], + "gamma_dfa": NaN, + "acc_eval": 0.59375, + "loss_eval": 1.1748178005218506, + "epoch": 29 + }, + { + "hidden_norms": [ + 269.5155944824219, + 254.65687561035156, + 249.08070373535156, + 246.4084930419922, + 248.03065490722656 + ], + "bp_grad_norms_per_sample_med": [ + 3.340150215080939e-05, + 3.542837657732889e-05, + 3.656153057818301e-05, + 3.62997452612035e-05, + 3.134588405373506e-05 + ], + "bp_grad_norms_F": [ + 0.0015988025115802884, + 0.0016428054077550769, + 0.0016343789175152779, + 0.00154898373875767, + 0.001290567102842033 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5888671875, + "loss_eval": 1.184614658355713, + "epoch": 30 + }, + { + "hidden_norms": [ + 273.0482482910156, + 255.88157653808594, + 249.6538543701172, + 246.09181213378906, + 250.09861755371094 + ], + "bp_grad_norms_per_sample_med": [ + 3.485401612124406e-05, + 3.613552325987257e-05, + 3.727355215232819e-05, + 3.6410303437151015e-05, + 3.111722253379412e-05 + ], + "bp_grad_norms_F": [ + 0.0016218513483181596, + 0.001677986583672464, + 0.0016929754056036472, + 0.001593436230905354, + 0.0013143697287887335 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5673828125, + "loss_eval": 1.2012369632720947, + "epoch": 31 + }, + { + "hidden_norms": [ + 274.05010986328125, + 255.77792358398438, + 247.96664428710938, + 246.15379333496094, + 248.44847106933594 + ], + "bp_grad_norms_per_sample_med": [ + 3.431778895901516e-05, + 3.609380655689165e-05, + 3.7377911212388426e-05, + 3.6012417695019394e-05, + 3.0985673220129684e-05 + ], + "bp_grad_norms_F": [ + 0.0016116296174004674, + 0.0016683804569765925, + 0.0016786488704383373, + 0.0015687323175370693, + 0.001289981184527278 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5986328125, + "loss_eval": 1.1654620170593262, + "epoch": 32 + }, + { + "hidden_norms": [ + 274.6156921386719, + 257.166259765625, + 251.9473876953125, + 251.01661682128906, + 252.84222412109375 + ], + "bp_grad_norms_per_sample_med": [ + 3.185174864483997e-05, + 3.433173696976155e-05, + 3.572784407879226e-05, + 3.4680619137361646e-05, + 2.997389128722716e-05 + ], + "bp_grad_norms_F": [ + 0.0015743848634883761, + 0.001610185718163848, + 0.0016222007106989622, + 0.0015360275283455849, + 0.0012806184822693467 + ], + "gamma_dfa": NaN, + "acc_eval": 0.58984375, + "loss_eval": 1.1679120063781738, + "epoch": 33 + }, + { + "hidden_norms": [ + 274.3385009765625, + 256.1404724121094, + 249.90264892578125, + 247.3904266357422, + 249.5623016357422 + ], + "bp_grad_norms_per_sample_med": [ + 3.419954009586945e-05, + 3.6830995668424293e-05, + 3.8464357203338295e-05, + 3.64230691047851e-05, + 3.0958552088122815e-05 + ], + "bp_grad_norms_F": [ + 0.0016346886986866593, + 0.0016823107143864036, + 0.0016956342151388526, + 0.0015958002768456936, + 0.0013100137002766132 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5927734375, + "loss_eval": 1.1599082946777344, + "epoch": 34 + }, + { + "hidden_norms": [ + 276.65985107421875, + 258.50543212890625, + 251.1997833251953, + 247.7036895751953, + 248.07545471191406 + ], + "bp_grad_norms_per_sample_med": [ + 3.384043520782143e-05, + 3.567537714843638e-05, + 3.686825220938772e-05, + 3.5754670534515753e-05, + 3.0301211154437624e-05 + ], + "bp_grad_norms_F": [ + 0.0016451155534014106, + 0.0016956959152594209, + 0.001703266054391861, + 0.0016015227884054184, + 0.0013142916141077876 + ], + "gamma_dfa": NaN, + "acc_eval": 0.615234375, + "loss_eval": 1.148215889930725, + "epoch": 35 + }, + { + "hidden_norms": [ + 279.6298522949219, + 261.1481018066406, + 253.67742919921875, + 250.86221313476562, + 250.6768035888672 + ], + "bp_grad_norms_per_sample_med": [ + 3.341220144648105e-05, + 3.499272861517966e-05, + 3.6701327189803123e-05, + 3.602671858971007e-05, + 3.081390605075285e-05 + ], + "bp_grad_norms_F": [ + 0.0016287019243463874, + 0.0016769150970503688, + 0.0016951658762991428, + 0.0015969424275681376, + 0.0013187138829380274 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6015625, + "loss_eval": 1.1451219320297241, + "epoch": 36 + }, + { + "hidden_norms": [ + 277.2866516113281, + 258.4331359863281, + 251.4995880126953, + 249.0448455810547, + 248.5166015625 + ], + "bp_grad_norms_per_sample_med": [ + 3.2723561162129045e-05, + 3.4428499930072576e-05, + 3.571901106624864e-05, + 3.4436179703334346e-05, + 2.9252587410155684e-05 + ], + "bp_grad_norms_F": [ + 0.0016638770466670394, + 0.0017149352934211493, + 0.0017208203207701445, + 0.0016291715437546372, + 0.0013402353506535292 + ], + "gamma_dfa": NaN, + "acc_eval": 0.599609375, + "loss_eval": 1.145239233970642, + "epoch": 37 + }, + { + "hidden_norms": [ + 279.7109680175781, + 259.3131103515625, + 250.73587036132812, + 247.214111328125, + 249.20944213867188 + ], + "bp_grad_norms_per_sample_med": [ + 3.449835639912635e-05, + 3.682902752188966e-05, + 3.780073166126385e-05, + 3.652105442597531e-05, + 3.046718666155357e-05 + ], + "bp_grad_norms_F": [ + 0.001753124757669866, + 0.0018144651548936963, + 0.0018306451383978128, + 0.0017095789080485702, + 0.001379602705128491 + ], + "gamma_dfa": NaN, + "acc_eval": 0.5908203125, + "loss_eval": 1.1677792072296143, + "epoch": 38 + }, + { + "hidden_norms": [ + 279.9946594238281, + 260.2458190917969, + 252.0008087158203, + 250.56716918945312, + 250.73135375976562 + ], + "bp_grad_norms_per_sample_med": [ + 3.489958180580288e-05, + 3.730110984179191e-05, + 3.791403287323192e-05, + 3.650010330602527e-05, + 3.076100256294012e-05 + ], + "bp_grad_norms_F": [ + 0.0017145961755886674, + 0.0017580300336703658, + 0.001775139826349914, + 0.0016669962788000703, + 0.0013597821816802025 + ], + "gamma_dfa": NaN, + "acc_eval": 0.615234375, + "loss_eval": 1.1510668992996216, + "epoch": 39 + }, + { + "hidden_norms": [ + 280.5132751464844, + 260.75970458984375, + 252.2814178466797, + 249.84548950195312, + 248.41783142089844 + ], + "bp_grad_norms_per_sample_med": [ + 3.552919588400982e-05, + 3.777973688556813e-05, + 3.952483166358434e-05, + 3.783141073654406e-05, + 3.1490348192164674e-05 + ], + "bp_grad_norms_F": [ + 0.0017431442392989993, + 0.0017984459409490228, + 0.0018199888290837407, + 0.0017078405944630504, + 0.0013792469399049878 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6103515625, + "loss_eval": 1.1412110328674316, + "epoch": 40 + }, + { + "hidden_norms": [ + 279.3873596191406, + 259.55096435546875, + 251.9462432861328, + 247.29443359375, + 247.82852172851562 + ], + "bp_grad_norms_per_sample_med": [ + 3.456972990534268e-05, + 3.654780448414385e-05, + 3.848340566037223e-05, + 3.736113285413012e-05, + 3.1597948691342026e-05 + ], + "bp_grad_norms_F": [ + 0.0017544032307341695, + 0.0018107325304299593, + 0.0018319039372727275, + 0.0017261328175663948, + 0.0013868761016055942 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6083984375, + "loss_eval": 1.1418483257293701, + "epoch": 41 + }, + { + "hidden_norms": [ + 282.4360656738281, + 261.22149658203125, + 251.49090576171875, + 247.5189208984375, + 247.03427124023438 + ], + "bp_grad_norms_per_sample_med": [ + 3.5179280530428514e-05, + 3.728548108483665e-05, + 3.779769394895993e-05, + 3.6327692214399576e-05, + 3.026284321094863e-05 + ], + "bp_grad_norms_F": [ + 0.0017881252570077777, + 0.0018441356951370835, + 0.0018691252917051315, + 0.0017487212317064404, + 0.0013921407517045736 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6142578125, + "loss_eval": 1.1468144655227661, + "epoch": 42 + }, + { + "hidden_norms": [ + 283.1267395019531, + 261.29425048828125, + 252.38563537597656, + 248.00192260742188, + 246.58505249023438 + ], + "bp_grad_norms_per_sample_med": [ + 3.485176421236247e-05, + 3.699021544889547e-05, + 3.812833165284246e-05, + 3.7145182432141155e-05, + 3.1198313081404194e-05 + ], + "bp_grad_norms_F": [ + 0.0017692983383312821, + 0.0018438888946548104, + 0.001876220339909196, + 0.0017641744343563914, + 0.0014044356066733599 + ], + "gamma_dfa": NaN, + "acc_eval": 0.62109375, + "loss_eval": 1.1329892873764038, + "epoch": 43 + }, + { + "hidden_norms": [ + 281.29327392578125, + 259.43359375, + 249.52906799316406, + 246.0892333984375, + 243.162109375 + ], + "bp_grad_norms_per_sample_med": [ + 3.586204911698587e-05, + 3.787377499975264e-05, + 3.9215286960825324e-05, + 3.7436715501826257e-05, + 3.160976120852865e-05 + ], + "bp_grad_norms_F": [ + 0.0018376735970377922, + 0.0019067999674007297, + 0.0019337635021656752, + 0.0017983472207561135, + 0.001438757055439055 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6181640625, + "loss_eval": 1.144508957862854, + "epoch": 44 + }, + { + "hidden_norms": [ + 282.5230407714844, + 259.55322265625, + 248.72308349609375, + 245.0394287109375, + 243.0651092529297 + ], + "bp_grad_norms_per_sample_med": [ + 3.6309906136011705e-05, + 3.868114436045289e-05, + 4.0484872442903e-05, + 3.8203681469894946e-05, + 3.144023139611818e-05 + ], + "bp_grad_norms_F": [ + 0.0018340122187510133, + 0.0019134157337248325, + 0.0019425065256655216, + 0.0018252682639285922, + 0.001443033921532333 + ], + "gamma_dfa": NaN, + "acc_eval": 0.61328125, + "loss_eval": 1.1290326118469238, + "epoch": 45 + }, + { + "hidden_norms": [ + 280.8598327636719, + 260.14111328125, + 250.84779357910156, + 245.9561767578125, + 243.0910186767578 + ], + "bp_grad_norms_per_sample_med": [ + 3.4626806154847145e-05, + 3.693445250974037e-05, + 3.8311845855787396e-05, + 3.7721722037531435e-05, + 3.145459049846977e-05 + ], + "bp_grad_norms_F": [ + 0.0018591763218864799, + 0.0019209657330065966, + 0.0019388735527172685, + 0.001830682042054832, + 0.001447839429602027 + ], + "gamma_dfa": NaN, + "acc_eval": 0.611328125, + "loss_eval": 1.1423537731170654, + "epoch": 46 + }, + { + "hidden_norms": [ + 280.17266845703125, + 259.26568603515625, + 248.45452880859375, + 243.45750427246094, + 241.2735137939453 + ], + "bp_grad_norms_per_sample_med": [ + 3.609486884670332e-05, + 3.887504499289207e-05, + 4.01619981857948e-05, + 3.9294092857744545e-05, + 3.1412804673891515e-05 + ], + "bp_grad_norms_F": [ + 0.0018975000130012631, + 0.0019804267212748528, + 0.002001592656597495, + 0.0018729616422206163, + 0.0014571875799447298 + ], + "gamma_dfa": NaN, + "acc_eval": 0.61328125, + "loss_eval": 1.1366385221481323, + "epoch": 47 + }, + { + "hidden_norms": [ + 281.79833984375, + 259.25860595703125, + 249.18551635742188, + 244.24977111816406, + 240.67100524902344 + ], + "bp_grad_norms_per_sample_med": [ + 3.47710374626331e-05, + 3.7153975426917896e-05, + 3.885528349201195e-05, + 3.792768620769493e-05, + 3.113777347607538e-05 + ], + "bp_grad_norms_F": [ + 0.0019318967824801803, + 0.0019981141667813063, + 0.0020311782136559486, + 0.0018996316939592361, + 0.0014788025291636586 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6162109375, + "loss_eval": 1.138331651687622, + "epoch": 48 + }, + { + "hidden_norms": [ + 282.3840637207031, + 258.1807861328125, + 248.13995361328125, + 243.45587158203125, + 239.51858520507812 + ], + "bp_grad_norms_per_sample_med": [ + 3.68369001080282e-05, + 3.935660788556561e-05, + 4.189974788459949e-05, + 3.9929436752572656e-05, + 3.188087430316955e-05 + ], + "bp_grad_norms_F": [ + 0.0019372758688405156, + 0.0020036858040839434, + 0.0020287453662604094, + 0.001896715839393437, + 0.0014835885958746076 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6103515625, + "loss_eval": 1.1377842426300049, + "epoch": 49 + }, + { + "hidden_norms": [ + 280.9283142089844, + 257.81524658203125, + 248.47984313964844, + 241.96157836914062, + 238.4756317138672 + ], + "bp_grad_norms_per_sample_med": [ + 3.7310775951482356e-05, + 3.98769298044499e-05, + 4.1753111872822046e-05, + 4.021048516733572e-05, + 3.21856678056065e-05 + ], + "bp_grad_norms_F": [ + 0.0019433718407526612, + 0.0020235073752701283, + 0.002055324148386717, + 0.0019370814552530646, + 0.0015049076173454523 + ], + "gamma_dfa": NaN, + "acc_eval": 0.615234375, + "loss_eval": 1.133098840713501, + "epoch": 50 + }, + { + "hidden_norms": [ + 279.66400146484375, + 256.7322082519531, + 246.65223693847656, + 241.45947265625, + 236.65582275390625 + ], + "bp_grad_norms_per_sample_med": [ + 3.620017014327459e-05, + 3.800772174145095e-05, + 4.072795491083525e-05, + 3.888807259500027e-05, + 3.182498039677739e-05 + ], + "bp_grad_norms_F": [ + 0.0019729172345250845, + 0.002036863239482045, + 0.0020802877843379974, + 0.0019662927370518446, + 0.0015235996106639504 + ], + "gamma_dfa": NaN, + "acc_eval": 0.619140625, + "loss_eval": 1.1315571069717407, + "epoch": 51 + }, + { + "hidden_norms": [ + 280.00152587890625, + 258.0238037109375, + 247.4223175048828, + 240.8101348876953, + 237.3229217529297 + ], + "bp_grad_norms_per_sample_med": [ + 3.6401659599505365e-05, + 3.855170143651776e-05, + 4.065703978994861e-05, + 3.844475213554688e-05, + 3.1645795388612896e-05 + ], + "bp_grad_norms_F": [ + 0.0019983802922070026, + 0.0020701857283711433, + 0.002107742242515087, + 0.001981276087462902, + 0.0015306670684367418 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6171875, + "loss_eval": 1.137194037437439, + "epoch": 52 + }, + { + "hidden_norms": [ + 278.19708251953125, + 255.23036193847656, + 245.5857696533203, + 239.48196411132812, + 235.09983825683594 + ], + "bp_grad_norms_per_sample_med": [ + 3.97727417293936e-05, + 4.185501529718749e-05, + 4.392741175252013e-05, + 4.302372326492332e-05, + 3.48737376043573e-05 + ], + "bp_grad_norms_F": [ + 0.002044696593657136, + 0.00211329385638237, + 0.002144050318747759, + 0.0020148823969066143, + 0.0015573638956993818 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6103515625, + "loss_eval": 1.1354646682739258, + "epoch": 53 + }, + { + "hidden_norms": [ + 276.7043762207031, + 255.62408447265625, + 243.71484375, + 238.2984161376953, + 233.8153076171875 + ], + "bp_grad_norms_per_sample_med": [ + 3.715257116709836e-05, + 3.887641287292354e-05, + 4.099323268746957e-05, + 3.948296944145113e-05, + 3.221148654120043e-05 + ], + "bp_grad_norms_F": [ + 0.0020365805830806494, + 0.00209893099963665, + 0.0021342532709240913, + 0.0020123852882534266, + 0.0015702869277447462 + ], + "gamma_dfa": NaN, + "acc_eval": 0.623046875, + "loss_eval": 1.1295127868652344, + "epoch": 54 + }, + { + "hidden_norms": [ + 279.308349609375, + 255.30587768554688, + 243.6786651611328, + 237.71823120117188, + 232.89991760253906 + ], + "bp_grad_norms_per_sample_med": [ + 3.9407084841514006e-05, + 4.1900668293237686e-05, + 4.364689812064171e-05, + 4.154618363827467e-05, + 3.264051701989956e-05 + ], + "bp_grad_norms_F": [ + 0.0021011638455092907, + 0.002175872679799795, + 0.002217113971710205, + 0.002071501687169075, + 0.0015880379360169172 + ], + "gamma_dfa": NaN, + "acc_eval": 0.626953125, + "loss_eval": 1.1201554536819458, + "epoch": 55 + }, + { + "hidden_norms": [ + 276.9933166503906, + 253.379638671875, + 244.23390197753906, + 238.01930236816406, + 233.3076171875 + ], + "bp_grad_norms_per_sample_med": [ + 3.799827391048893e-05, + 4.017200990347192e-05, + 4.2665858927648515e-05, + 4.083341627847403e-05, + 3.210103386663832e-05 + ], + "bp_grad_norms_F": [ + 0.002064710482954979, + 0.0021362933330237865, + 0.0021728368010371923, + 0.0020478416699916124, + 0.0015873984666541219 + ], + "gamma_dfa": NaN, + "acc_eval": 0.630859375, + "loss_eval": 1.1222728490829468, + "epoch": 56 + }, + { + "hidden_norms": [ + 276.41168212890625, + 251.91033935546875, + 241.81382751464844, + 234.8279266357422, + 230.53956604003906 + ], + "bp_grad_norms_per_sample_med": [ + 3.770063631236553e-05, + 3.94962407881394e-05, + 4.150882887188345e-05, + 4.040042040287517e-05, + 3.2475280022481456e-05 + ], + "bp_grad_norms_F": [ + 0.002123177284374833, + 0.0021965675987303257, + 0.00223328173160553, + 0.0020905195269733667, + 0.0016039537731558084 + ], + "gamma_dfa": NaN, + "acc_eval": 0.630859375, + "loss_eval": 1.1091526746749878, + "epoch": 57 + }, + { + "hidden_norms": [ + 274.760986328125, + 250.0972442626953, + 239.511474609375, + 233.29798889160156, + 228.16070556640625 + ], + "bp_grad_norms_per_sample_med": [ + 3.728809315362014e-05, + 4.0525967051507905e-05, + 4.239326517563313e-05, + 4.087138222530484e-05, + 3.216422919649631e-05 + ], + "bp_grad_norms_F": [ + 0.0021635335870087147, + 0.0022461507469415665, + 0.0022873859852552414, + 0.002144098049029708, + 0.001636566361412406 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6416015625, + "loss_eval": 1.1180500984191895, + "epoch": 58 + }, + { + "hidden_norms": [ + 273.91064453125, + 249.49452209472656, + 238.54373168945312, + 232.80734252929688, + 225.6833953857422 + ], + "bp_grad_norms_per_sample_med": [ + 3.966117583331652e-05, + 4.1986208088928834e-05, + 4.316144259064458e-05, + 4.332280877861194e-05, + 3.44519030477386e-05 + ], + "bp_grad_norms_F": [ + 0.0021825931034982204, + 0.0022617392241954803, + 0.0023092019837349653, + 0.00216560042463243, + 0.0016612708568572998 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6279296875, + "loss_eval": 1.131789207458496, + "epoch": 59 + }, + { + "hidden_norms": [ + 274.90008544921875, + 250.0712127685547, + 238.57467651367188, + 231.15719604492188, + 225.55587768554688 + ], + "bp_grad_norms_per_sample_med": [ + 4.130819797865115e-05, + 4.363877815194428e-05, + 4.489632920012809e-05, + 4.4512398744700477e-05, + 3.470887531875633e-05 + ], + "bp_grad_norms_F": [ + 0.002240521367639303, + 0.002310203853994608, + 0.0023576009552925825, + 0.002204722724854946, + 0.0016799044096842408 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6240234375, + "loss_eval": 1.1323366165161133, + "epoch": 60 + }, + { + "hidden_norms": [ + 275.4504699707031, + 249.5341796875, + 237.540283203125, + 230.39208984375, + 224.43321228027344 + ], + "bp_grad_norms_per_sample_med": [ + 4.056218676851131e-05, + 4.288182390155271e-05, + 4.5818003854947165e-05, + 4.356648423708975e-05, + 3.492152245598845e-05 + ], + "bp_grad_norms_F": [ + 0.002249655080959201, + 0.0023316002916544676, + 0.002382595557719469, + 0.002231382764875889, + 0.00168902100995183 + ], + "gamma_dfa": NaN, + "acc_eval": 0.62890625, + "loss_eval": 1.1275749206542969, + "epoch": 61 + }, + { + "hidden_norms": [ + 273.6263122558594, + 248.43479919433594, + 234.8235321044922, + 228.0020751953125, + 222.4278106689453 + ], + "bp_grad_norms_per_sample_med": [ + 4.0597744373371825e-05, + 4.323143730289303e-05, + 4.647616151487455e-05, + 4.4688425987260416e-05, + 3.521826511132531e-05 + ], + "bp_grad_norms_F": [ + 0.002307126997038722, + 0.0023877634666860104, + 0.0024417925160378218, + 0.0022813305258750916, + 0.001714664394967258 + ], + "gamma_dfa": NaN, + "acc_eval": 0.626953125, + "loss_eval": 1.1252907514572144, + "epoch": 62 + }, + { + "hidden_norms": [ + 271.7508239746094, + 246.79209899902344, + 233.84280395507812, + 226.96934509277344, + 220.57888793945312 + ], + "bp_grad_norms_per_sample_med": [ + 4.177554728812538e-05, + 4.393904600874521e-05, + 4.635817458620295e-05, + 4.428510874276981e-05, + 3.396430111024529e-05 + ], + "bp_grad_norms_F": [ + 0.002355988370254636, + 0.0024527544155716896, + 0.0025022663176059723, + 0.0023387460969388485, + 0.0017370774876326323 + ], + "gamma_dfa": NaN, + "acc_eval": 0.62109375, + "loss_eval": 1.1249933242797852, + "epoch": 63 + }, + { + "hidden_norms": [ + 271.7035217285156, + 246.70675659179688, + 233.96661376953125, + 226.53175354003906, + 220.52359008789062 + ], + "bp_grad_norms_per_sample_med": [ + 4.2499312257859856e-05, + 4.503092714003287e-05, + 4.670799171435647e-05, + 4.5866876462241635e-05, + 3.720118911587633e-05 + ], + "bp_grad_norms_F": [ + 0.002408105880022049, + 0.0024944571778178215, + 0.002545328810811043, + 0.002381332218647003, + 0.0017641197191551328 + ], + "gamma_dfa": NaN, + "acc_eval": 0.625, + "loss_eval": 1.1418037414550781, + "epoch": 64 + }, + { + "hidden_norms": [ + 271.0970153808594, + 246.78770446777344, + 234.2436065673828, + 226.7012176513672, + 219.61680603027344 + ], + "bp_grad_norms_per_sample_med": [ + 4.040581188746728e-05, + 4.3224725231993943e-05, + 4.4494401663541794e-05, + 4.3532105337362736e-05, + 3.4612559829838574e-05 + ], + "bp_grad_norms_F": [ + 0.0024068003986030817, + 0.002485552802681923, + 0.002538430504500866, + 0.002368941670283675, + 0.0017652952810749412 + ], + "gamma_dfa": NaN, + "acc_eval": 0.626953125, + "loss_eval": 1.1279454231262207, + "epoch": 65 + }, + { + "hidden_norms": [ + 270.66473388671875, + 244.8525390625, + 232.8356475830078, + 224.2222137451172, + 217.24554443359375 + ], + "bp_grad_norms_per_sample_med": [ + 4.1549992602085695e-05, + 4.493525193538517e-05, + 4.7448877012357116e-05, + 4.479651397559792e-05, + 3.5401590139372274e-05 + ], + "bp_grad_norms_F": [ + 0.002463088370859623, + 0.002551029436290264, + 0.002594881458207965, + 0.0024160456378012896, + 0.001803302438929677 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6376953125, + "loss_eval": 1.1330325603485107, + "epoch": 66 + }, + { + "hidden_norms": [ + 269.8645935058594, + 243.50497436523438, + 231.37692260742188, + 222.27967834472656, + 215.22059631347656 + ], + "bp_grad_norms_per_sample_med": [ + 4.1519095248077065e-05, + 4.487609476200305e-05, + 4.708006963483058e-05, + 4.4878328480990604e-05, + 3.578457108233124e-05 + ], + "bp_grad_norms_F": [ + 0.0024854058865457773, + 0.0025721373967826366, + 0.0026270432863384485, + 0.0024401266127824783, + 0.0018134783022105694 + ], + "gamma_dfa": NaN, + "acc_eval": 0.63671875, + "loss_eval": 1.123876690864563, + "epoch": 67 + }, + { + "hidden_norms": [ + 269.28570556640625, + 242.04359436035156, + 229.53497314453125, + 221.24327087402344, + 215.1306610107422 + ], + "bp_grad_norms_per_sample_med": [ + 4.279898348613642e-05, + 4.6176737669156864e-05, + 4.813459236174822e-05, + 4.609950337908231e-05, + 3.566763552953489e-05 + ], + "bp_grad_norms_F": [ + 0.0024860126432031393, + 0.002575797727331519, + 0.0026342689525336027, + 0.0024608916137367487, + 0.0018216874450445175 + ], + "gamma_dfa": NaN, + "acc_eval": 0.62890625, + "loss_eval": 1.122079849243164, + "epoch": 68 + }, + { + "hidden_norms": [ + 268.3481140136719, + 242.3759765625, + 229.19241333007812, + 220.78233337402344, + 213.58587646484375 + ], + "bp_grad_norms_per_sample_med": [ + 4.163750054431148e-05, + 4.50872685178183e-05, + 4.8797548515722156e-05, + 4.597716542775743e-05, + 3.505409767967649e-05 + ], + "bp_grad_norms_F": [ + 0.0025033215060830116, + 0.002596172271296382, + 0.002654082840308547, + 0.002480126451700926, + 0.0018305372213944793 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6328125, + "loss_eval": 1.1238645315170288, + "epoch": 69 + }, + { + "hidden_norms": [ + 267.92425537109375, + 239.90985107421875, + 228.08212280273438, + 219.12835693359375, + 212.38052368164062 + ], + "bp_grad_norms_per_sample_med": [ + 4.42373457190115e-05, + 4.6515069698216394e-05, + 4.823393464903347e-05, + 4.762738535646349e-05, + 3.6444154829951e-05 + ], + "bp_grad_norms_F": [ + 0.0025684935972094536, + 0.002675230847671628, + 0.0027295781765133142, + 0.0025394319090992212, + 0.0018643162911757827 + ], + "gamma_dfa": NaN, + "acc_eval": 0.626953125, + "loss_eval": 1.135793685913086, + "epoch": 70 + }, + { + "hidden_norms": [ + 266.2279052734375, + 239.37680053710938, + 226.3264923095703, + 218.04910278320312, + 210.68836975097656 + ], + "bp_grad_norms_per_sample_med": [ + 4.4824198994319886e-05, + 4.7638000978622586e-05, + 4.962517414242029e-05, + 4.7396664740517735e-05, + 3.5957502404926345e-05 + ], + "bp_grad_norms_F": [ + 0.002600227715447545, + 0.002714222064241767, + 0.002763595897704363, + 0.0025755097158253193, + 0.0018772372277453542 + ], + "gamma_dfa": NaN, + "acc_eval": 0.623046875, + "loss_eval": 1.123576283454895, + "epoch": 71 + }, + { + "hidden_norms": [ + 266.4837951660156, + 239.67581176757812, + 227.51805114746094, + 218.58811950683594, + 211.26596069335938 + ], + "bp_grad_norms_per_sample_med": [ + 4.245663512847386e-05, + 4.560768138617277e-05, + 4.797001020051539e-05, + 4.539322981145233e-05, + 3.4807548217941076e-05 + ], + "bp_grad_norms_F": [ + 0.0025984421372413635, + 0.002703956561163068, + 0.002762222196906805, + 0.002577459439635277, + 0.0018865292659029365 + ], + "gamma_dfa": NaN, + "acc_eval": 0.61328125, + "loss_eval": 1.1362444162368774, + "epoch": 72 + }, + { + "hidden_norms": [ + 264.8100891113281, + 237.96121215820312, + 226.14822387695312, + 217.3454132080078, + 208.98077392578125 + ], + "bp_grad_norms_per_sample_med": [ + 4.217765308567323e-05, + 4.603620618581772e-05, + 4.735886977869086e-05, + 4.5346474507823586e-05, + 3.594472218537703e-05 + ], + "bp_grad_norms_F": [ + 0.0026375274173915386, + 0.0027397233061492443, + 0.0028012648690491915, + 0.0026011927984654903, + 0.0018980697495862842 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6318359375, + "loss_eval": 1.1286330223083496, + "epoch": 73 + }, + { + "hidden_norms": [ + 264.8172302246094, + 238.0478057861328, + 225.0303497314453, + 217.1344757080078, + 209.569580078125 + ], + "bp_grad_norms_per_sample_med": [ + 4.242224895278923e-05, + 4.6696859499206766e-05, + 4.831598562304862e-05, + 4.5893459173385054e-05, + 3.54531584889628e-05 + ], + "bp_grad_norms_F": [ + 0.0026500027161091566, + 0.002760941395536065, + 0.0028223067056387663, + 0.0026202781591564417, + 0.001915424014441669 + ], + "gamma_dfa": NaN, + "acc_eval": 0.626953125, + "loss_eval": 1.1304678916931152, + "epoch": 74 + }, + { + "hidden_norms": [ + 264.7013854980469, + 237.4438018798828, + 225.43527221679688, + 216.18502807617188, + 208.08302307128906 + ], + "bp_grad_norms_per_sample_med": [ + 4.43761091446504e-05, + 4.801220347872004e-05, + 4.9817157560028136e-05, + 4.829527097172104e-05, + 3.73583534383215e-05 + ], + "bp_grad_norms_F": [ + 0.002686889609321952, + 0.0028029615059494972, + 0.002866453491151333, + 0.002669648267328739, + 0.0019349679350852966 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6240234375, + "loss_eval": 1.1350867748260498, + "epoch": 75 + }, + { + "hidden_norms": [ + 264.3004455566406, + 236.1824188232422, + 223.4772186279297, + 214.64682006835938, + 207.52272033691406 + ], + "bp_grad_norms_per_sample_med": [ + 4.3003376049455255e-05, + 4.6196739276638255e-05, + 4.8772617446957156e-05, + 4.7264005843317136e-05, + 3.6179160815663636e-05 + ], + "bp_grad_norms_F": [ + 0.0027191757690161467, + 0.0028310383204370737, + 0.0028950762934982777, + 0.002689448418095708, + 0.0019433586858212948 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6259765625, + "loss_eval": 1.1378443241119385, + "epoch": 76 + }, + { + "hidden_norms": [ + 262.62811279296875, + 235.80126953125, + 222.7261199951172, + 213.42430114746094, + 205.99493408203125 + ], + "bp_grad_norms_per_sample_med": [ + 4.4399519538274035e-05, + 4.737091148854233e-05, + 4.900431667920202e-05, + 4.768445069203153e-05, + 3.608842962421477e-05 + ], + "bp_grad_norms_F": [ + 0.0027450472116470337, + 0.0028562990482896566, + 0.0029378861654549837, + 0.0027161173056811094, + 0.0019581648521125317 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6220703125, + "loss_eval": 1.137312889099121, + "epoch": 77 + }, + { + "hidden_norms": [ + 262.612060546875, + 235.62489318847656, + 222.1566925048828, + 212.93875122070312, + 205.24142456054688 + ], + "bp_grad_norms_per_sample_med": [ + 4.443951911525801e-05, + 4.7095887566683814e-05, + 4.906025424133986e-05, + 4.8352503654314205e-05, + 3.6001758417114615e-05 + ], + "bp_grad_norms_F": [ + 0.002745443256571889, + 0.002859732834622264, + 0.002942001214250922, + 0.002725319005548954, + 0.001969744684174657 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6279296875, + "loss_eval": 1.126712441444397, + "epoch": 78 + }, + { + "hidden_norms": [ + 262.6611633300781, + 235.52174377441406, + 221.89093017578125, + 213.1799774169922, + 204.9930877685547 + ], + "bp_grad_norms_per_sample_med": [ + 4.5766752009512857e-05, + 4.834473293158226e-05, + 5.181043525226414e-05, + 4.826926306122914e-05, + 3.573719732230529e-05 + ], + "bp_grad_norms_F": [ + 0.0027218428440392017, + 0.002838012296706438, + 0.002919211983680725, + 0.0027209871914237738, + 0.001970200566574931 + ], + "gamma_dfa": NaN, + "acc_eval": 0.62109375, + "loss_eval": 1.1352139711380005, + "epoch": 79 + }, + { + "hidden_norms": [ + 261.58856201171875, + 234.9381103515625, + 221.2636260986328, + 211.9308624267578, + 204.54647827148438 + ], + "bp_grad_norms_per_sample_med": [ + 4.5460306864697486e-05, + 4.857210296904668e-05, + 4.9856924306368455e-05, + 4.823243580176495e-05, + 3.731804099516012e-05 + ], + "bp_grad_norms_F": [ + 0.002767088357359171, + 0.0028805267065763474, + 0.0029621378052979708, + 0.00275496207177639, + 0.001986011164262891 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6259765625, + "loss_eval": 1.134087324142456, + "epoch": 80 + }, + { + "hidden_norms": [ + 261.3812561035156, + 234.79383850097656, + 220.678955078125, + 211.66114807128906, + 203.9405517578125 + ], + "bp_grad_norms_per_sample_med": [ + 4.64644399471581e-05, + 5.052189590060152e-05, + 5.2927574870409444e-05, + 5.0570299208629876e-05, + 3.7252495530992746e-05 + ], + "bp_grad_norms_F": [ + 0.0027735778130590916, + 0.002885127207264304, + 0.0029686433263123035, + 0.002763227093964815, + 0.0019980284851044416 + ], + "gamma_dfa": NaN, + "acc_eval": 0.62890625, + "loss_eval": 1.135728359222412, + "epoch": 81 + }, + { + "hidden_norms": [ + 261.2560729980469, + 233.66014099121094, + 220.25682067871094, + 211.13511657714844, + 203.18130493164062 + ], + "bp_grad_norms_per_sample_med": [ + 4.6268629375845194e-05, + 4.9701800890034065e-05, + 5.169105133973062e-05, + 5.069339385954663e-05, + 3.80978781322483e-05 + ], + "bp_grad_norms_F": [ + 0.0028223153203725815, + 0.0029419492930173874, + 0.0030234854202717543, + 0.002805375959724188, + 0.002015760401263833 + ], + "gamma_dfa": NaN, + "acc_eval": 0.619140625, + "loss_eval": 1.1317980289459229, + "epoch": 82 + }, + { + "hidden_norms": [ + 260.62255859375, + 233.12937927246094, + 219.89907836914062, + 210.4191436767578, + 202.494873046875 + ], + "bp_grad_norms_per_sample_med": [ + 4.5126867917133495e-05, + 4.8572586820228025e-05, + 5.1520797569537535e-05, + 4.971070302417502e-05, + 3.838812335743569e-05 + ], + "bp_grad_norms_F": [ + 0.0028271761257201433, + 0.002944816369563341, + 0.003029909450560808, + 0.0028170389123260975, + 0.002025141380727291 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6259765625, + "loss_eval": 1.1392892599105835, + "epoch": 83 + }, + { + "hidden_norms": [ + 260.0465393066406, + 232.27760314941406, + 218.845458984375, + 209.7770233154297, + 201.92831420898438 + ], + "bp_grad_norms_per_sample_med": [ + 4.730001455754973e-05, + 5.118208355270326e-05, + 5.1580365834524855e-05, + 4.936498589813709e-05, + 3.706002826220356e-05 + ], + "bp_grad_norms_F": [ + 0.0028656981885433197, + 0.002991823712363839, + 0.0030798574443906546, + 0.0028517318423837423, + 0.0020406947005540133 + ], + "gamma_dfa": NaN, + "acc_eval": 0.625, + "loss_eval": 1.1393200159072876, + "epoch": 84 + }, + { + "hidden_norms": [ + 259.99285888671875, + 232.2432403564453, + 218.6685333251953, + 209.5802001953125, + 201.7836456298828 + ], + "bp_grad_norms_per_sample_med": [ + 4.4883607188239694e-05, + 4.9325379222864285e-05, + 5.1563201850512996e-05, + 4.9446705816080794e-05, + 3.641278817667626e-05 + ], + "bp_grad_norms_F": [ + 0.002843427937477827, + 0.0029685739427804947, + 0.0030620063189417124, + 0.0028350763022899628, + 0.0020346841774880886 + ], + "gamma_dfa": NaN, + "acc_eval": 0.625, + "loss_eval": 1.136763572692871, + "epoch": 85 + }, + { + "hidden_norms": [ + 259.7002868652344, + 232.1937713623047, + 218.6339569091797, + 209.42572021484375, + 201.48532104492188 + ], + "bp_grad_norms_per_sample_med": [ + 4.557083229883574e-05, + 5.006462743040174e-05, + 5.1964529120596126e-05, + 5.040188261773437e-05, + 3.70733359886799e-05 + ], + "bp_grad_norms_F": [ + 0.002866365248337388, + 0.0029951538890600204, + 0.0030862949788570404, + 0.0028592217713594437, + 0.002047886373475194 + ], + "gamma_dfa": NaN, + "acc_eval": 0.630859375, + "loss_eval": 1.139750361442566, + "epoch": 86 + }, + { + "hidden_norms": [ + 259.22808837890625, + 232.3558807373047, + 218.42848205566406, + 209.1579132080078, + 201.5212860107422 + ], + "bp_grad_norms_per_sample_med": [ + 4.5765504182782024e-05, + 4.875436570728198e-05, + 5.104453521198593e-05, + 5.005718412576243e-05, + 3.675639891298488e-05 + ], + "bp_grad_norms_F": [ + 0.002878013998270035, + 0.003002135781571269, + 0.0030916042160242796, + 0.002861205255612731, + 0.0020479748491197824 + ], + "gamma_dfa": NaN, + "acc_eval": 0.62890625, + "loss_eval": 1.1344283819198608, + "epoch": 87 + }, + { + "hidden_norms": [ + 259.1551513671875, + 232.05300903320312, + 218.15708923339844, + 208.9161834716797, + 201.6800537109375 + ], + "bp_grad_norms_per_sample_med": [ + 4.611797339748591e-05, + 4.964259278494865e-05, + 5.252738264971413e-05, + 5.0600574468262494e-05, + 3.820879283011891e-05 + ], + "bp_grad_norms_F": [ + 0.0028855716809630394, + 0.003010405693203211, + 0.0031016215216368437, + 0.002874146681278944, + 0.0020545709412544966 + ], + "gamma_dfa": NaN, + "acc_eval": 0.626953125, + "loss_eval": 1.135157823562622, + "epoch": 88 + }, + { + "hidden_norms": [ + 259.21893310546875, + 231.72747802734375, + 217.7665252685547, + 208.59022521972656, + 201.1155548095703 + ], + "bp_grad_norms_per_sample_med": [ + 4.6835561079205945e-05, + 5.060776311438531e-05, + 5.325842721504159e-05, + 5.128757766215131e-05, + 3.774667493416928e-05 + ], + "bp_grad_norms_F": [ + 0.002893448108807206, + 0.0030188425444066525, + 0.0031124993693083525, + 0.002883171197026968, + 0.002058332087472081 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6298828125, + "loss_eval": 1.1357543468475342, + "epoch": 89 + }, + { + "hidden_norms": [ + 258.88385009765625, + 230.9880828857422, + 217.33352661132812, + 208.3406524658203, + 200.54083251953125 + ], + "bp_grad_norms_per_sample_med": [ + 4.6940938773332164e-05, + 5.0942355301231146e-05, + 5.368283382267691e-05, + 5.184212204767391e-05, + 3.746367656276561e-05 + ], + "bp_grad_norms_F": [ + 0.0029224599711596966, + 0.0030530549120157957, + 0.003145798807963729, + 0.0029114321805536747, + 0.0020718672312796116 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6279296875, + "loss_eval": 1.1398890018463135, + "epoch": 90 + }, + { + "hidden_norms": [ + 258.7764587402344, + 231.19357299804688, + 217.22779846191406, + 207.97528076171875, + 199.982421875 + ], + "bp_grad_norms_per_sample_med": [ + 4.70306295028422e-05, + 5.185308327781968e-05, + 5.281997437123209e-05, + 5.1426530262688175e-05, + 3.7254685594234616e-05 + ], + "bp_grad_norms_F": [ + 0.002924826927483082, + 0.0030514320824295282, + 0.0031429482623934746, + 0.0029119658283889294, + 0.0020754521246999502 + ], + "gamma_dfa": NaN, + "acc_eval": 0.62890625, + "loss_eval": 1.138663411140442, + "epoch": 91 + }, + { + "hidden_norms": [ + 258.7837829589844, + 230.58901977539062, + 217.062255859375, + 207.7244873046875, + 199.89395141601562 + ], + "bp_grad_norms_per_sample_med": [ + 4.688404078478925e-05, + 5.141457950230688e-05, + 5.3491079597733915e-05, + 5.087396493763663e-05, + 3.739634121302515e-05 + ], + "bp_grad_norms_F": [ + 0.0029230716172605753, + 0.0030518025159835815, + 0.0031419326551258564, + 0.002910938812419772, + 0.00207503535784781 + ], + "gamma_dfa": NaN, + "acc_eval": 0.626953125, + "loss_eval": 1.1365739107131958, + "epoch": 92 + }, + { + "hidden_norms": [ + 258.7561950683594, + 230.4254913330078, + 217.07546997070312, + 207.6160125732422, + 199.7816162109375 + ], + "bp_grad_norms_per_sample_med": [ + 4.680828351411037e-05, + 5.096250242786482e-05, + 5.211709503782913e-05, + 5.09246347064618e-05, + 3.779645703616552e-05 + ], + "bp_grad_norms_F": [ + 0.002931967144832015, + 0.0030605625361204147, + 0.00315046776086092, + 0.0029180829878896475, + 0.002077121753245592 + ], + "gamma_dfa": NaN, + "acc_eval": 0.62890625, + "loss_eval": 1.1363377571105957, + "epoch": 93 + }, + { + "hidden_norms": [ + 258.76409912109375, + 230.5386505126953, + 216.76475524902344, + 207.16859436035156, + 199.2827911376953 + ], + "bp_grad_norms_per_sample_med": [ + 4.749056097352877e-05, + 5.2020419389009476e-05, + 5.3769603255204856e-05, + 5.178397987037897e-05, + 3.789216498262249e-05 + ], + "bp_grad_norms_F": [ + 0.0029493479523807764, + 0.0030784173868596554, + 0.0031682881526649, + 0.0029316784348338842, + 0.002084067091345787 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6298828125, + "loss_eval": 1.1382873058319092, + "epoch": 94 + }, + { + "hidden_norms": [ + 258.7884826660156, + 230.40945434570312, + 216.70376586914062, + 207.1806182861328, + 199.41917419433594 + ], + "bp_grad_norms_per_sample_med": [ + 4.741501834359951e-05, + 5.169968062546104e-05, + 5.352263542590663e-05, + 5.053802306065336e-05, + 3.758866660064086e-05 + ], + "bp_grad_norms_F": [ + 0.002947731176391244, + 0.003076745430007577, + 0.0031680618412792683, + 0.0029321794863790274, + 0.0020845714025199413 + ], + "gamma_dfa": NaN, + "acc_eval": 0.630859375, + "loss_eval": 1.1382077932357788, + "epoch": 95 + }, + { + "hidden_norms": [ + 258.766845703125, + 230.38992309570312, + 216.7628173828125, + 207.26092529296875, + 199.36366271972656 + ], + "bp_grad_norms_per_sample_med": [ + 4.718741183751263e-05, + 5.1928480388596654e-05, + 5.3666040912503377e-05, + 5.089692058390938e-05, + 3.752821430680342e-05 + ], + "bp_grad_norms_F": [ + 0.002943811472505331, + 0.003073457395657897, + 0.0031646883580833673, + 0.0029298998415470123, + 0.002084212377667427 + ], + "gamma_dfa": NaN, + "acc_eval": 0.62890625, + "loss_eval": 1.138296365737915, + "epoch": 96 + }, + { + "hidden_norms": [ + 258.8481140136719, + 230.38389587402344, + 216.7748260498047, + 207.24560546875, + 199.39401245117188 + ], + "bp_grad_norms_per_sample_med": [ + 4.728109706775285e-05, + 5.156427505426109e-05, + 5.3663239668821916e-05, + 5.139272616361268e-05, + 3.7748181057395414e-05 + ], + "bp_grad_norms_F": [ + 0.002947240835055709, + 0.003076959168538451, + 0.0031680231913924217, + 0.0029323517810553312, + 0.0020851469598710537 + ], + "gamma_dfa": NaN, + "acc_eval": 0.62890625, + "loss_eval": 1.1388041973114014, + "epoch": 97 + }, + { + "hidden_norms": [ + 258.77734375, + 230.3692626953125, + 216.7212677001953, + 207.1890869140625, + 199.34091186523438 + ], + "bp_grad_norms_per_sample_med": [ + 4.7204313887050375e-05, + 5.154019163455814e-05, + 5.373918611439876e-05, + 5.1062332204310223e-05, + 3.7904177588643506e-05 + ], + "bp_grad_norms_F": [ + 0.0029484080150723457, + 0.0030780842062085867, + 0.0031695417128503323, + 0.0029333126731216908, + 0.002085541607812047 + ], + "gamma_dfa": NaN, + "acc_eval": 0.62890625, + "loss_eval": 1.138832688331604, + "epoch": 98 + }, + { + "hidden_norms": [ + 258.7897644042969, + 230.38552856445312, + 216.73997497558594, + 207.1963653564453, + 199.38656616210938 + ], + "bp_grad_norms_per_sample_med": [ + 4.726059705717489e-05, + 5.1730185077758506e-05, + 5.3854040743317455e-05, + 5.0960083171958104e-05, + 3.793121504713781e-05 + ], + "bp_grad_norms_F": [ + 0.00294809159822762, + 0.003077775239944458, + 0.003169153118506074, + 0.0029330949764698744, + 0.0020853858441114426 + ], + "gamma_dfa": NaN, + "acc_eval": 0.62890625, + "loss_eval": 1.1388249397277832, + "epoch": 99 + }, + { + "hidden_norms": [ + 258.7891845703125, + 230.38888549804688, + 216.73963928222656, + 207.19769287109375, + 199.4077606201172 + ], + "bp_grad_norms_per_sample_med": [ + 4.725577309727669e-05, + 5.174486432224512e-05, + 5.38085114385467e-05, + 5.098402107250877e-05, + 3.7933263229206204e-05 + ], + "bp_grad_norms_F": [ + 0.002948042470961809, + 0.0030777468346059322, + 0.003169137053191662, + 0.002933081705123186, + 0.0020853858441114426 + ], + "gamma_dfa": NaN, + "acc_eval": 0.6279296875, + "loss_eval": 1.1387826204299927, + "epoch": 100 + } + ], + "dfa_log": [ + { + "hidden_norms": [ + 8.696551322937012, + 8.737784385681152, + 8.813130378723145, + 8.844642639160156, + 8.889528274536133 + ], + "bp_grad_norms_per_sample_med": [ + 0.0010079912608489394, + 0.0010035138111561537, + 0.0009970925748348236, + 0.0009936393471434712, + 0.000986678060144186 + ], + "bp_grad_norms_F": [ + 0.035574451088905334, + 0.035288579761981964, + 0.035048045217990875, + 0.0346725769340992, + 0.03439188003540039 + ], + "gamma_dfa": 0.0033523982419865206, + "acc_eval": 0.0927734375, + "loss_eval": 2.4863264560699463, + "epoch": 0 + }, + { + "hidden_norms": [ + 900.3377685546875, + 3463.92626953125, + 6660.5166015625, + 10478.5126953125, + 13466.513671875 + ], + "bp_grad_norms_per_sample_med": [ + 2.017142833210528e-06, + 7.972002435963077e-07, + 7.783806381667091e-07, + 7.758338256280695e-07, + 7.810436954969191e-07 + ], + "bp_grad_norms_F": [ + 0.00013047464017290622, + 7.321321754716337e-05, + 7.131113670766354e-05, + 7.113779429346323e-05, + 7.09498199285008e-05 + ], + "gamma_dfa": 0.1028625147882849, + "acc_eval": 0.2861328125, + "loss_eval": 1.970402717590332, + "epoch": 1 + }, + { + "hidden_norms": [ + 1784.805419921875, + 14085.2138671875, + 36523.9375, + 61422.50390625, + 73870.3359375 + ], + "bp_grad_norms_per_sample_med": [ + 9.087140711017128e-07, + 1.772240381114898e-07, + 1.705584367073243e-07, + 1.7156033038645546e-07, + 1.728751612972701e-07 + ], + "bp_grad_norms_F": [ + 4.7445286327274516e-05, + 1.1973520486208145e-05, + 9.698592293716501e-06, + 9.335580216429662e-06, + 9.154121471510734e-06 + ], + "gamma_dfa": 0.09588160619023256, + "acc_eval": 0.298828125, + "loss_eval": 1.9628738164901733, + "epoch": 2 + }, + { + "hidden_norms": [ + 2670.88134765625, + 35761.4921875, + 99927.7890625, + 170223.953125, + 201448.390625 + ], + "bp_grad_norms_per_sample_med": [ + 5.815078338855528e-07, + 7.772520405069372e-08, + 7.527528111950232e-08, + 7.557918735301428e-08, + 7.57173737042649e-08 + ], + "bp_grad_norms_F": [ + 2.8944486984983087e-05, + 3.2005805223889183e-06, + 2.6765355869429186e-06, + 2.639568720042007e-06, + 2.6436800908413716e-06 + ], + "gamma_dfa": 0.09442083747126162, + "acc_eval": 0.2900390625, + "loss_eval": 1.9821033477783203, + "epoch": 3 + }, + { + "hidden_norms": [ + 3506.001220703125, + 70233.9375, + 205054.078125, + 348792.375, + 416698.5625 + ], + "bp_grad_norms_per_sample_med": [ + 4.5228594558466284e-07, + 4.338783554658221e-08, + 4.114394869247917e-08, + 4.122260932604149e-08, + 4.124275321260029e-08 + ], + "bp_grad_norms_F": [ + 2.319143823115155e-05, + 1.7165241388283903e-06, + 1.3772670399703202e-06, + 1.3777447520624264e-06, + 1.3796359326079255e-06 + ], + "gamma_dfa": 0.09172606549691409, + "acc_eval": 0.298828125, + "loss_eval": 1.9661527872085571, + "epoch": 4 + }, + { + "hidden_norms": [ + 4351.04052734375, + 120982.0390625, + 361547.03125, + 633667.3125, + 764198.75 + ], + "bp_grad_norms_per_sample_med": [ + 3.632259790720127e-07, + 2.6118947005215887e-08, + 2.4913996199416033e-08, + 2.494128636953974e-08, + 2.4942012899487054e-08 + ], + "bp_grad_norms_F": [ + 1.894752676889766e-05, + 1.1090510270150844e-06, + 8.424218549407669e-07, + 8.403117703892349e-07, + 8.407336054006009e-07 + ], + "gamma_dfa": 0.08992433547973633, + "acc_eval": 0.2900390625, + "loss_eval": 1.980475664138794, + "epoch": 5 + }, + { + "hidden_norms": [ + 5174.9833984375, + 189817.0625, + 576763.3125, + 1012734.125, + 1235217.5 + ], + "bp_grad_norms_per_sample_med": [ + 3.340334444601467e-07, + 1.826493090106851e-08, + 1.7263264595612782e-08, + 1.730784759956805e-08, + 1.7305337607353977e-08 + ], + "bp_grad_norms_F": [ + 1.7139096598839387e-05, + 7.582453918075771e-07, + 5.794162802885694e-07, + 5.812879635413992e-07, + 5.813373604723893e-07 + ], + "gamma_dfa": 0.08752730302512646, + "acc_eval": 0.3056640625, + "loss_eval": 1.9696743488311768, + "epoch": 6 + }, + { + "hidden_norms": [ + 5963.3935546875, + 276686.75, + 861517.375, + 1555059.75, + 1913322.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.834741223978199e-07, + 1.3120247110975924e-08, + 1.2326365705916942e-08, + 1.2353129186237766e-08, + 1.2349524958210623e-08 + ], + "bp_grad_norms_F": [ + 1.4978587387304287e-05, + 5.169860628484457e-07, + 4.2094282548532647e-07, + 4.222391680741566e-07, + 4.222146685606276e-07 + ], + "gamma_dfa": 0.08884745300747454, + "acc_eval": 0.2919921875, + "loss_eval": 1.9592503309249878, + "epoch": 7 + }, + { + "hidden_norms": [ + 6751.1630859375, + 390990.65625, + 1268157.625, + 2303462.75, + 2850455.0 + ], + "bp_grad_norms_per_sample_med": [ + 2.5272595394199016e-07, + 9.644952392307005e-09, + 9.213153795428752e-09, + 9.189856875480018e-09, + 9.217835383878992e-09 + ], + "bp_grad_norms_F": [ + 1.3214407772466075e-05, + 3.7820277043465467e-07, + 3.14011003865744e-07, + 3.1468064776163374e-07, + 3.146610936255456e-07 + ], + "gamma_dfa": 0.08774545323103666, + "acc_eval": 0.3115234375, + "loss_eval": 1.9636309146881104, + "epoch": 8 + }, + { + "hidden_norms": [ + 7557.10302734375, + 541741.25, + 1767146.375, + 3237576.25, + 4005094.0 + ], + "bp_grad_norms_per_sample_med": [ + 2.3122942138797953e-07, + 7.519741274109037e-09, + 7.2226606917524805e-09, + 7.218183384338772e-09, + 7.221534037427091e-09 + ], + "bp_grad_norms_F": [ + 1.196970697492361e-05, + 2.9184533900661336e-07, + 2.4720966962377133e-07, + 2.4752529270699597e-07, + 2.4754368155299744e-07 + ], + "gamma_dfa": 0.08941872720606625, + "acc_eval": 0.2900390625, + "loss_eval": 1.9587393999099731, + "epoch": 9 + }, + { + "hidden_norms": [ + 8353.015625, + 720288.5625, + 2367034.25, + 4437438.5, + 5500873.0 + ], + "bp_grad_norms_per_sample_med": [ + 2.1009967099416826e-07, + 6.096115612308495e-09, + 5.8281148795913396e-09, + 5.816529036195561e-09, + 5.823292070772368e-09 + ], + "bp_grad_norms_F": [ + 1.0853039384528529e-05, + 2.2833219759377243e-07, + 1.979581298883204e-07, + 1.9814572738141578e-07, + 1.9820384977720096e-07 + ], + "gamma_dfa": 0.09132041316479445, + "acc_eval": 0.2978515625, + "loss_eval": 1.9605882167816162, + "epoch": 10 + }, + { + "hidden_norms": [ + 9134.4443359375, + 924688.8125, + 3108971.25, + 5809310.0, + 7180875.5 + ], + "bp_grad_norms_per_sample_med": [ + 1.9402843065563502e-07, + 4.977510403847418e-09, + 4.8141264308299014e-09, + 4.791541829973767e-09, + 4.794175723077387e-09 + ], + "bp_grad_norms_F": [ + 1.0063903573609423e-05, + 1.9364638603747153e-07, + 1.6622837506474752e-07, + 1.6636218447274587e-07, + 1.6641583044929575e-07 + ], + "gamma_dfa": 0.08994872949551791, + "acc_eval": 0.2822265625, + "loss_eval": 1.9717106819152832, + "epoch": 11 + }, + { + "hidden_norms": [ + 9934.0107421875, + 1171186.5, + 4059807.75, + 7477705.0, + 9248714.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.802942222184356e-07, + 4.213630333538276e-09, + 4.032348677185382e-09, + 4.031016409555832e-09, + 4.033940292913485e-09 + ], + "bp_grad_norms_F": [ + 9.375374247611035e-06, + 1.60658984782458e-07, + 1.398711617639492e-07, + 1.399746452079853e-07, + 1.4004145043600147e-07 + ], + "gamma_dfa": 0.09168496390338987, + "acc_eval": 0.2998046875, + "loss_eval": 1.9513648748397827, + "epoch": 12 + }, + { + "hidden_norms": [ + 10690.2216796875, + 1474348.375, + 5210694.5, + 9539616.0, + 11795722.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.7011585384807404e-07, + 3.5948770626959003e-09, + 3.444458496204561e-09, + 3.439692752849055e-09, + 3.4398179860062328e-09 + ], + "bp_grad_norms_F": [ + 8.913790225051343e-06, + 1.346561759874021e-07, + 1.1880944583708697e-07, + 1.1887690476442003e-07, + 1.1894194784645151e-07 + ], + "gamma_dfa": 0.0931803259300068, + "acc_eval": 0.3134765625, + "loss_eval": 1.959463357925415, + "epoch": 13 + }, + { + "hidden_norms": [ + 11428.3076171875, + 1789460.25, + 6431273.0, + 11810251.0, + 14610525.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5823535193248972e-07, + 3.1298097447063356e-09, + 3.012203153573978e-09, + 3.010908633527265e-09, + 3.010648619294898e-09 + ], + "bp_grad_norms_F": [ + 8.233506378019229e-06, + 1.1668681310084139e-07, + 1.0298786889961775e-07, + 1.0303519104581937e-07, + 1.0309597797686365e-07 + ], + "gamma_dfa": 0.09336384548805654, + "acc_eval": 0.2861328125, + "loss_eval": 1.9515889883041382, + "epoch": 14 + }, + { + "hidden_norms": [ + 12175.4384765625, + 2177908.0, + 7886216.5, + 14475684.0, + 17927020.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5137742082060868e-07, + 2.705691448667835e-09, + 2.5981530260565933e-09, + 2.5995214869567462e-09, + 2.6000293029682098e-09 + ], + "bp_grad_norms_F": [ + 7.800716048222966e-06, + 1.0053086896277819e-07, + 8.984200405848242e-08, + 8.987952071493055e-08, + 8.993622913067156e-08 + ], + "gamma_dfa": 0.09357930033002049, + "acc_eval": 0.2978515625, + "loss_eval": 1.9580912590026855, + "epoch": 15 + }, + { + "hidden_norms": [ + 12918.6787109375, + 2622644.5, + 9573671.0, + 17460572.0, + 21682046.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.432233602827182e-07, + 2.359643591631766e-09, + 2.2714441438864696e-09, + 2.2688442236074025e-09, + 2.2694950363444377e-09 + ], + "bp_grad_norms_F": [ + 7.43454029361601e-06, + 8.846858889910436e-08, + 7.942762891843813e-08, + 7.945371294226788e-08, + 7.950914238108453e-08 + ], + "gamma_dfa": 0.09332277777139097, + "acc_eval": 0.30859375, + "loss_eval": 1.9656270742416382, + "epoch": 16 + }, + { + "hidden_norms": [ + 13676.931640625, + 3144179.5, + 11421599.0, + 20832254.0, + 25907438.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.3402345189206244e-07, + 2.1064472388587774e-09, + 2.0263852817947736e-09, + 2.0199610872850826e-09, + 2.0222397090208233e-09 + ], + "bp_grad_norms_F": [ + 6.909830062795663e-06, + 7.773559929091789e-08, + 7.076827301943922e-08, + 7.077783692466255e-08, + 7.082799413637986e-08 + ], + "gamma_dfa": 0.0945073509356007, + "acc_eval": 0.296875, + "loss_eval": 1.9510960578918457, + "epoch": 17 + }, + { + "hidden_norms": [ + 14410.64453125, + 3691622.0, + 13302875.0, + 24659928.0, + 30700242.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.2569752527724631e-07, + 1.8540845525549798e-09, + 1.8018908587436044e-09, + 1.8003264434796051e-09, + 1.802896276714705e-09 + ], + "bp_grad_norms_F": [ + 6.58303315503872e-06, + 6.894391191281102e-08, + 6.3350469758916e-08, + 6.336826174901944e-08, + 6.340740554833246e-08 + ], + "gamma_dfa": 0.09388192358892411, + "acc_eval": 0.30859375, + "loss_eval": 1.9456725120544434, + "epoch": 18 + }, + { + "hidden_norms": [ + 15102.634765625, + 4270070.0, + 15687016.0, + 28858048.0, + 35882240.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.2314617947595252e-07, + 1.6925943979728686e-09, + 1.6383979728473719e-09, + 1.6367472932543592e-09, + 1.6394127166918793e-09 + ], + "bp_grad_norms_F": [ + 6.343140285025584e-06, + 6.240136940505181e-08, + 5.757489418556361e-08, + 5.7587019597349354e-08, + 5.7624333749117795e-08 + ], + "gamma_dfa": 0.09518493304494768, + "acc_eval": 0.302734375, + "loss_eval": 1.9497833251953125, + "epoch": 19 + }, + { + "hidden_norms": [ + 15809.6728515625, + 4913181.0, + 18109688.0, + 33281892.0, + 41436228.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.1704661773137559e-07, + 1.533212445892218e-09, + 1.483607681151966e-09, + 1.4837684414459318e-09, + 1.4838255069093975e-09 + ], + "bp_grad_norms_F": [ + 6.079233116906835e-06, + 5.7217945936827164e-08, + 5.270470992968512e-08, + 5.2718213794378244e-08, + 5.2752689327917324e-08 + ], + "gamma_dfa": 0.09407007653499022, + "acc_eval": 0.2958984375, + "loss_eval": 1.961036205291748, + "epoch": 20 + }, + { + "hidden_norms": [ + 16522.783203125, + 5620503.0, + 20819020.0, + 38391960.0, + 47723272.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.1146808986950418e-07, + 1.4118024527220996e-09, + 1.3678460586419305e-09, + 1.3669514409286876e-09, + 1.3668450815629285e-09 + ], + "bp_grad_norms_F": [ + 5.778397280664649e-06, + 5.162353389209784e-08, + 4.8153967924235985e-08, + 4.816252285877454e-08, + 4.8193449231348495e-08 + ], + "gamma_dfa": 0.09494083630852401, + "acc_eval": 0.2958984375, + "loss_eval": 1.9576481580734253, + "epoch": 21 + }, + { + "hidden_norms": [ + 17230.27734375, + 6437730.5, + 23789486.0, + 43960328.0, + 54569096.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.0952022932997352e-07, + 1.2858689668604484e-09, + 1.246260317167014e-09, + 1.2462959553261044e-09, + 1.2484926426026277e-09 + ], + "bp_grad_norms_F": [ + 5.637425147142494e-06, + 4.732414282671016e-08, + 4.430494371376881e-08, + 4.431280586913999e-08, + 4.434128086927558e-08 + ], + "gamma_dfa": 0.09481688600499183, + "acc_eval": 0.2900390625, + "loss_eval": 1.9625442028045654, + "epoch": 22 + }, + { + "hidden_norms": [ + 17916.12890625, + 7217473.5, + 26969020.0, + 49888804.0, + 62010184.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.043977349013403e-07, + 1.1912204556097095e-09, + 1.159051632448893e-09, + 1.1590449711107453e-09, + 1.159014662022173e-09 + ], + "bp_grad_norms_F": [ + 5.385433269111672e-06, + 4.3334623001101136e-08, + 4.083162963297582e-08, + 4.0836408032873805e-08, + 4.0863898931320364e-08 + ], + "gamma_dfa": 0.0946647283853963, + "acc_eval": 0.2939453125, + "loss_eval": 1.9656811952590942, + "epoch": 23 + }, + { + "hidden_norms": [ + 18617.46875, + 8095365.0, + 30241864.0, + 56059896.0, + 69702984.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.001103129283365e-07, + 1.1034658742303805e-09, + 1.0674757744411068e-09, + 1.0666105776380164e-09, + 1.0677670969627684e-09 + ], + "bp_grad_norms_F": [ + 5.150916877028067e-06, + 3.999087283546032e-08, + 3.779740609388682e-08, + 3.780221291549424e-08, + 3.782682256314729e-08 + ], + "gamma_dfa": 0.09509630355751142, + "acc_eval": 0.296875, + "loss_eval": 1.9476948976516724, + "epoch": 24 + }, + { + "hidden_norms": [ + 19248.7890625, + 9047595.0, + 33634812.0, + 62495088.0, + 77796536.0 + ], + "bp_grad_norms_per_sample_med": [ + 9.749370377676314e-08, + 1.0280801765460978e-09, + 9.980430926148642e-10, + 9.978470272287154e-10, + 9.990493987643845e-10 + ], + "bp_grad_norms_F": [ + 5.017087005398935e-06, + 3.7442834610601494e-08, + 3.536156611971819e-08, + 3.536548476290591e-08, + 3.538822213045023e-08 + ], + "gamma_dfa": 0.09479212731821463, + "acc_eval": 0.2861328125, + "loss_eval": 1.9604804515838623, + "epoch": 25 + }, + { + "hidden_norms": [ + 19902.470703125, + 10030725.0, + 37252656.0, + 69351704.0, + 86270768.0 + ], + "bp_grad_norms_per_sample_med": [ + 9.411947843318558e-08, + 9.563448921667828e-10, + 9.361866837309663e-10, + 9.359075736625755e-10, + 9.361766917237446e-10 + ], + "bp_grad_norms_F": [ + 4.873440502706217e-06, + 3.4947159832654506e-08, + 3.313299146157078e-08, + 3.3137865784738096e-08, + 3.315915364510147e-08 + ], + "gamma_dfa": 0.09510701123508625, + "acc_eval": 0.2978515625, + "loss_eval": 1.9498941898345947, + "epoch": 26 + }, + { + "hidden_norms": [ + 20513.66796875, + 11086393.0, + 41223968.0, + 76828584.0, + 95694480.0 + ], + "bp_grad_norms_per_sample_med": [ + 9.249640697817085e-08, + 9.003223166992314e-10, + 8.743661350507637e-10, + 8.736535939135592e-10, + 8.752278901624777e-10 + ], + "bp_grad_norms_F": [ + 4.75014485346037e-06, + 3.290759664764664e-08, + 3.1125043875590563e-08, + 3.1127818544973707e-08, + 3.11488683735206e-08 + ], + "gamma_dfa": 0.09552363055991009, + "acc_eval": 0.2841796875, + "loss_eval": 1.9527215957641602, + "epoch": 27 + }, + { + "hidden_norms": [ + 21161.345703125, + 12263719.0, + 45533684.0, + 84877296.0, + 105753224.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.900443049242313e-08, + 8.457264888406257e-10, + 8.261510919815862e-10, + 8.261567541190118e-10, + 8.276057061884501e-10 + ], + "bp_grad_norms_F": [ + 4.624974280886818e-06, + 3.0792985938887796e-08, + 2.9227694042788244e-08, + 2.9232776199705768e-08, + 2.9251353339532216e-08 + ], + "gamma_dfa": 0.09527035459177569, + "acc_eval": 0.2890625, + "loss_eval": 1.950648307800293, + "epoch": 28 + }, + { + "hidden_norms": [ + 21787.568359375, + 13358760.0, + 49723396.0, + 92840264.0, + 115901424.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.79872459336184e-08, + 8.034463094830357e-10, + 7.846168159630906e-10, + 7.837334670135476e-10, + 7.841179927581265e-10 + ], + "bp_grad_norms_F": [ + 4.530191290541552e-06, + 2.9184068495169413e-08, + 2.7723606521590227e-08, + 2.7726789753046432e-08, + 2.7744931685447227e-08 + ], + "gamma_dfa": 0.09598725114483386, + "acc_eval": 0.2919921875, + "loss_eval": 1.9504928588867188, + "epoch": 29 + }, + { + "hidden_norms": [ + 22378.34765625, + 14611837.0, + 54335560.0, + 101350864.0, + 126710488.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.497670478391228e-08, + 7.560595483013799e-10, + 7.353461728420996e-10, + 7.342743080229752e-10, + 7.348425756781296e-10 + ], + "bp_grad_norms_F": [ + 4.383283339848276e-06, + 2.7485034692631416e-08, + 2.6182895851434296e-08, + 2.618475392068831e-08, + 2.620201833281044e-08 + ], + "gamma_dfa": 0.09509110247017816, + "acc_eval": 0.2958984375, + "loss_eval": 1.9564056396484375, + "epoch": 30 + }, + { + "hidden_norms": [ + 22972.798828125, + 15931348.0, + 59291124.0, + 110237776.0, + 138020368.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.252197858382715e-08, + 7.193445838993284e-10, + 6.977300404109599e-10, + 6.966639487515636e-10, + 6.977723399081981e-10 + ], + "bp_grad_norms_F": [ + 4.255563453625655e-06, + 2.5971447215056287e-08, + 2.4833044065530885e-08, + 2.4835205891804435e-08, + 2.4851219748711628e-08 + ], + "gamma_dfa": 0.0951943755208049, + "acc_eval": 0.30859375, + "loss_eval": 1.9508299827575684, + "epoch": 31 + }, + { + "hidden_norms": [ + 23525.689453125, + 17283022.0, + 64234984.0, + 119317480.0, + 149347744.0 + ], + "bp_grad_norms_per_sample_med": [ + 8.036072784989301e-08, + 6.884249836858203e-10, + 6.666004415123439e-10, + 6.656397100179845e-10, + 6.674217845059616e-10 + ], + "bp_grad_norms_F": [ + 4.212505245959619e-06, + 2.4864668546342727e-08, + 2.375563212808629e-08, + 2.3757841915994504e-08, + 2.3773273127858374e-08 + ], + "gamma_dfa": 0.09534151147818193, + "acc_eval": 0.294921875, + "loss_eval": 1.95639967918396, + "epoch": 32 + }, + { + "hidden_norms": [ + 24102.95703125, + 18811968.0, + 69193392.0, + 128823560.0, + 161700704.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.921565980950618e-08, + 6.526117424243694e-10, + 6.390621365426341e-10, + 6.381262185328751e-10, + 6.384171524764781e-10 + ], + "bp_grad_norms_F": [ + 4.09630820286111e-06, + 2.3687046990517047e-08, + 2.2686869272092736e-08, + 2.2687553169475905e-08, + 2.2702355551018627e-08 + ], + "gamma_dfa": 0.0947199320216896, + "acc_eval": 0.2880859375, + "loss_eval": 1.960526943206787, + "epoch": 33 + }, + { + "hidden_norms": [ + 24687.955078125, + 20313542.0, + 74389296.0, + 138258576.0, + 173791248.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.937118340350935e-08, + 6.360312831965587e-10, + 6.205357339084117e-10, + 6.196035906569364e-10, + 6.196675395031548e-10 + ], + "bp_grad_norms_F": [ + 4.070574505021796e-06, + 2.2864428572688666e-08, + 2.1894738466698982e-08, + 2.1895061763643753e-08, + 2.1909569269951135e-08 + ], + "gamma_dfa": 0.09556430295924656, + "acc_eval": 0.2822265625, + "loss_eval": 1.97263765335083, + "epoch": 34 + }, + { + "hidden_norms": [ + 25226.55078125, + 21851366.0, + 80000824.0, + 148574480.0, + 186595952.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.592127104771862e-08, + 6.044152955908544e-10, + 5.892630272619215e-10, + 5.896260146798227e-10, + 5.900769872724254e-10 + ], + "bp_grad_norms_F": [ + 3.966818894696189e-06, + 2.1824051898988728e-08, + 2.093374718015184e-08, + 2.093443285389185e-08, + 2.0947767964685227e-08 + ], + "gamma_dfa": 0.09587649442255497, + "acc_eval": 0.298828125, + "loss_eval": 1.9495991468429565, + "epoch": 35 + }, + { + "hidden_norms": [ + 25759.50390625, + 23391128.0, + 85561480.0, + 158667136.0, + 198961760.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.493858333873504e-08, + 5.860726903783586e-10, + 5.731417562770957e-10, + 5.733850616529423e-10, + 5.73569580719635e-10 + ], + "bp_grad_norms_F": [ + 3.896147973136976e-06, + 2.099926454945944e-08, + 2.0218029916918567e-08, + 2.0218159590967844e-08, + 2.0230830344303286e-08 + ], + "gamma_dfa": 0.09563440945930779, + "acc_eval": 0.294921875, + "loss_eval": 1.954901099205017, + "epoch": 36 + }, + { + "hidden_norms": [ + 26267.0859375, + 25111884.0, + 91593392.0, + 169479968.0, + 212491408.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.387680511783401e-08, + 5.612941222921108e-10, + 5.475105369079358e-10, + 5.468157038279742e-10, + 5.479569020749864e-10 + ], + "bp_grad_norms_F": [ + 3.838411430479027e-06, + 2.0261547106770195e-08, + 1.9493551661753372e-08, + 1.949278782831243e-08, + 1.9505332460312275e-08 + ], + "gamma_dfa": 0.09528586096712388, + "acc_eval": 0.302734375, + "loss_eval": 1.9556396007537842, + "epoch": 37 + }, + { + "hidden_norms": [ + 26761.587890625, + 26706742.0, + 97554944.0, + 180407616.0, + 226265728.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.260722156843258e-08, + 5.407673198121188e-10, + 5.32608401826451e-10, + 5.31698407524317e-10, + 5.328552044048251e-10 + ], + "bp_grad_norms_F": [ + 3.7372303722804645e-06, + 1.9519287519642603e-08, + 1.8807236656925852e-08, + 1.8806394663783976e-08, + 1.8818212765836506e-08 + ], + "gamma_dfa": 0.09491665422683582, + "acc_eval": 0.2890625, + "loss_eval": 1.963973879814148, + "epoch": 38 + }, + { + "hidden_norms": [ + 27302.328125, + 28336562.0, + 103781320.0, + 191401696.0, + 240130064.0 + ], + "bp_grad_norms_per_sample_med": [ + 7.187927053564636e-08, + 5.305865746763061e-10, + 5.205576525391109e-10, + 5.20661624925367e-10, + 5.217564713611011e-10 + ], + "bp_grad_norms_F": [ + 3.734877054739627e-06, + 1.88775892695503e-08, + 1.8220953634795478e-08, + 1.8220212893993448e-08, + 1.8231482101782603e-08 + ], + "gamma_dfa": 0.0957885464449646, + "acc_eval": 0.302734375, + "loss_eval": 1.9604723453521729, + "epoch": 39 + }, + { + "hidden_norms": [ + 27801.44140625, + 30051226.0, + 109945288.0, + 203002704.0, + 254881776.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.894305215610075e-08, + 5.044222262107212e-10, + 4.932855790507062e-10, + 4.934126440758746e-10, + 4.940316489232544e-10 + ], + "bp_grad_norms_F": [ + 3.610647127061384e-06, + 1.8150354108570355e-08, + 1.753623024569606e-08, + 1.7534704355171016e-08, + 1.7545806585417267e-08 + ], + "gamma_dfa": 0.09517717303242534, + "acc_eval": 0.2958984375, + "loss_eval": 1.9589688777923584, + "epoch": 40 + }, + { + "hidden_norms": [ + 28266.703125, + 31897594.0, + 116263888.0, + 215162448.0, + 270215296.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.807566421684896e-08, + 4.835814526593651e-10, + 4.775287942848649e-10, + 4.773109685274335e-10, + 4.784843632421598e-10 + ], + "bp_grad_norms_F": [ + 3.590413371057366e-06, + 1.7596850199197434e-08, + 1.705259222717359e-08, + 1.7050926004458233e-08, + 1.7061644541627174e-08 + ], + "gamma_dfa": 0.0961545761092566, + "acc_eval": 0.30859375, + "loss_eval": 1.9595096111297607, + "epoch": 41 + }, + { + "hidden_norms": [ + 28755.66015625, + 33582328.0, + 122605680.0, + 226378480.0, + 284438688.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.689399612014313e-08, + 4.746543713629592e-10, + 4.678599174745557e-10, + 4.68083238835959e-10, + 4.683119447790318e-10 + ], + "bp_grad_norms_F": [ + 3.5121436212648405e-06, + 1.7124380136124273e-08, + 1.6555107507087996e-08, + 1.655343240258844e-08, + 1.656368908697914e-08 + ], + "gamma_dfa": 0.09569378412561491, + "acc_eval": 0.287109375, + "loss_eval": 1.9612011909484863, + "epoch": 42 + }, + { + "hidden_norms": [ + 29227.5703125, + 35443612.0, + 129229904.0, + 238040368.0, + 299000064.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.582912703834154e-08, + 4.575871348500016e-10, + 4.493441174702184e-10, + 4.48520082185766e-10, + 4.485415650012925e-10 + ], + "bp_grad_norms_F": [ + 3.4321058137720684e-06, + 1.6602150765265833e-08, + 1.6083028242519504e-08, + 1.6080972997656318e-08, + 1.609118882583971e-08 + ], + "gamma_dfa": 0.09606127908773487, + "acc_eval": 0.306640625, + "loss_eval": 1.9551353454589844, + "epoch": 43 + }, + { + "hidden_norms": [ + 29665.798828125, + 37243032.0, + 135839760.0, + 250377328.0, + 314433920.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.588776813032382e-08, + 4.4936304677278827e-10, + 4.4115311403913893e-10, + 4.406963405312325e-10, + 4.4055534220710513e-10 + ], + "bp_grad_norms_F": [ + 3.3984799756581197e-06, + 1.6157997606569552e-08, + 1.564128204734061e-08, + 1.5638891071034777e-08, + 1.564865925729464e-08 + ], + "gamma_dfa": 0.09547593681054423, + "acc_eval": 0.2900390625, + "loss_eval": 1.9503705501556396, + "epoch": 44 + }, + { + "hidden_norms": [ + 30096.123046875, + 39014544.0, + 141817168.0, + 261217760.0, + 328197248.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.549055342475185e-08, + 4.3991607578952596e-10, + 4.348142956800416e-10, + 4.3428130536149467e-10, + 4.349899329625373e-10 + ], + "bp_grad_norms_F": [ + 3.382166596566094e-06, + 1.5776649320287106e-08, + 1.5299850275596327e-08, + 1.5297549893489304e-08, + 1.5307026757227504e-08 + ], + "gamma_dfa": 0.09610002994304523, + "acc_eval": 0.2998046875, + "loss_eval": 1.952854037284851, + "epoch": 45 + }, + { + "hidden_norms": [ + 30554.974609375, + 40873780.0, + 148087312.0, + 272436128.0, + 342258336.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.514299855098216e-08, + 4.291983157767021e-10, + 4.228755678958862e-10, + 4.2251957488304015e-10, + 4.227345695717588e-10 + ], + "bp_grad_norms_F": [ + 3.3793355669331504e-06, + 1.5485536408732514e-08, + 1.4993654318118388e-08, + 1.499121538017789e-08, + 1.5000590991576246e-08 + ], + "gamma_dfa": 0.09557013235462364, + "acc_eval": 0.3017578125, + "loss_eval": 1.9633105993270874, + "epoch": 46 + }, + { + "hidden_norms": [ + 30981.4765625, + 42548124.0, + 153904848.0, + 283882112.0, + 356627616.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.343542224840348e-08, + 4.1842326825580756e-10, + 4.139384113255318e-10, + 4.141494092113618e-10, + 4.1477912771092917e-10 + ], + "bp_grad_norms_F": [ + 3.3046233056666097e-06, + 1.508391989091251e-08, + 1.4620022525946297e-08, + 1.4617588028897899e-08, + 1.462664656060042e-08 + ], + "gamma_dfa": 0.09567815247282851, + "acc_eval": 0.296875, + "loss_eval": 1.958325982093811, + "epoch": 47 + }, + { + "hidden_norms": [ + 31383.2578125, + 44333160.0, + 160058736.0, + 295685184.0, + 371288096.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.18988522660402e-08, + 4.057836289206307e-10, + 4.0004452528386025e-10, + 3.9987613220660023e-10, + 4.004964138104583e-10 + ], + "bp_grad_norms_F": [ + 3.2255043151963037e-06, + 1.4663311453944061e-08, + 1.4252069746589768e-08, + 1.424969386931707e-08, + 1.4258660918642363e-08 + ], + "gamma_dfa": 0.09609431066201068, + "acc_eval": 0.302734375, + "loss_eval": 1.9508600234985352, + "epoch": 48 + }, + { + "hidden_norms": [ + 31750.03125, + 46204072.0, + 166757840.0, + 307620864.0, + 385856832.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.133605978675405e-08, + 3.9670436380312424e-10, + 3.904292722456404e-10, + 3.9002356899686674e-10, + 3.9039230181892037e-10 + ], + "bp_grad_norms_F": [ + 3.2063508115243167e-06, + 1.4354414545891814e-08, + 1.3938280751801813e-08, + 1.3935605558401676e-08, + 1.394456727865645e-08 + ], + "gamma_dfa": 0.09565898558139452, + "acc_eval": 0.310546875, + "loss_eval": 1.9467211961746216, + "epoch": 49 + }, + { + "hidden_norms": [ + 32133.1484375, + 47907484.0, + 173083168.0, + 319475424.0, + 400677984.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.030914079246941e-08, + 3.8645733835274143e-10, + 3.81215115030642e-10, + 3.812981042017327e-10, + 3.811861937208505e-10 + ], + "bp_grad_norms_F": [ + 3.177025973855052e-06, + 1.4072985443647212e-08, + 1.368274382684831e-08, + 1.3679918531295243e-08, + 1.368857649453048e-08 + ], + "gamma_dfa": 0.09606298332801089, + "acc_eval": 0.30078125, + "loss_eval": 1.9524433612823486, + "epoch": 50 + }, + { + "hidden_norms": [ + 32483.716796875, + 49794588.0, + 179046160.0, + 330962240.0, + 414954912.0 + ], + "bp_grad_norms_per_sample_med": [ + 6.051003254015086e-08, + 3.794887737385011e-10, + 3.751928767670165e-10, + 3.752734789586043e-10, + 3.7574471312140645e-10 + ], + "bp_grad_norms_F": [ + 3.154751766487607e-06, + 1.3766489281863414e-08, + 1.339402011524271e-08, + 1.3391274755747418e-08, + 1.3399712450734569e-08 + ], + "gamma_dfa": 0.09573800596626825, + "acc_eval": 0.3046875, + "loss_eval": 1.952356219291687, + "epoch": 51 + }, + { + "hidden_norms": [ + 32825.296875, + 51662352.0, + 185889568.0, + 343223360.0, + 430225696.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.9168662858155585e-08, + 3.7636679883767954e-10, + 3.7184264001233203e-10, + 3.7154498921943e-10, + 3.7157116272723556e-10 + ], + "bp_grad_norms_F": [ + 3.0909043289284455e-06, + 1.346105005239906e-08, + 1.3102917861829155e-08, + 1.3100156515122308e-08, + 1.3108366836434016e-08 + ], + "gamma_dfa": 0.09539097778178984, + "acc_eval": 0.291015625, + "loss_eval": 1.9585516452789307, + "epoch": 52 + }, + { + "hidden_norms": [ + 33148.9453125, + 53487464.0, + 192413680.0, + 354830112.0, + 444651488.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.949418024897568e-08, + 3.692673389288359e-10, + 3.6510769407804844e-10, + 3.6480282683548637e-10, + 3.65057900575394e-10 + ], + "bp_grad_norms_F": [ + 3.0807416351308348e-06, + 1.3205962545725924e-08, + 1.2856014031115137e-08, + 1.2853148767533185e-08, + 1.2861222309368259e-08 + ], + "gamma_dfa": 0.09566472223013989, + "acc_eval": 0.2880859375, + "loss_eval": 1.9538320302963257, + "epoch": 53 + }, + { + "hidden_norms": [ + 33465.359375, + 55258308.0, + 198680688.0, + 365875680.0, + 458634560.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.9228732141036744e-08, + 3.640579226971141e-10, + 3.6045816331764513e-10, + 3.6029040861862427e-10, + 3.6025293859154317e-10 + ], + "bp_grad_norms_F": [ + 3.058711854464491e-06, + 1.294814744312589e-08, + 1.2619584488504643e-08, + 1.2616848898971966e-08, + 1.2624772338654111e-08 + ], + "gamma_dfa": 0.09607254836737411, + "acc_eval": 0.2998046875, + "loss_eval": 1.9489495754241943, + "epoch": 54 + }, + { + "hidden_norms": [ + 33763.53125, + 56920656.0, + 204554208.0, + 376681216.0, + 472155680.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.756867338391203e-08, + 3.4894195866108646e-10, + 3.4671995829960167e-10, + 3.464649955819965e-10, + 3.4666045034548176e-10 + ], + "bp_grad_norms_F": [ + 3.092292217843351e-06, + 1.2733584853208413e-08, + 1.2427238793577544e-08, + 1.2424394846277664e-08, + 1.2432122886707475e-08 + ], + "gamma_dfa": 0.09629741777825984, + "acc_eval": 0.3017578125, + "loss_eval": 1.9477816820144653, + "epoch": 55 + }, + { + "hidden_norms": [ + 34061.03125, + 58650568.0, + 210577584.0, + 387585472.0, + 485747168.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.70959208801014e-08, + 3.4928168690662176e-10, + 3.46003392603933e-10, + 3.4562114281655454e-10, + 3.4606426058125805e-10 + ], + "bp_grad_norms_F": [ + 3.178826091243536e-06, + 1.2512856528701377e-08, + 1.2208544397651622e-08, + 1.2205639166040783e-08, + 1.2213387634574246e-08 + ], + "gamma_dfa": 0.09542209599749185, + "acc_eval": 0.3046875, + "loss_eval": 1.9549853801727295, + "epoch": 56 + }, + { + "hidden_norms": [ + 34344.76171875, + 60294200.0, + 216821792.0, + 398493824.0, + 499371840.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.7367511629990986e-08, + 3.416208427253764e-10, + 3.378898272288211e-10, + 3.377436663676292e-10, + 3.386138314187548e-10 + ], + "bp_grad_norms_F": [ + 3.2603361432848033e-06, + 1.2295201301526504e-08, + 1.2022674411582557e-08, + 1.2019833128817936e-08, + 1.2027326690144946e-08 + ], + "gamma_dfa": 0.09621504079404986, + "acc_eval": 0.294921875, + "loss_eval": 1.9597327709197998, + "epoch": 57 + }, + { + "hidden_norms": [ + 34621.87109375, + 62004936.0, + 222707536.0, + 409299680.0, + 512662208.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.6534943837505125e-08, + 3.356194766546139e-10, + 3.3068575655548216e-10, + 3.3087257933495096e-10, + 3.3104252672444545e-10 + ], + "bp_grad_norms_F": [ + 3.325955731270369e-06, + 1.2073825494951507e-08, + 1.1812358202689666e-08, + 1.1809524025352403e-08, + 1.1816909228912209e-08 + ], + "gamma_dfa": 0.09621665377926547, + "acc_eval": 0.3017578125, + "loss_eval": 1.9411261081695557, + "epoch": 58 + }, + { + "hidden_norms": [ + 34892.69921875, + 63619864.0, + 228332912.0, + 419524224.0, + 525474048.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.6131870707076814e-08, + 3.3116545616884707e-10, + 3.281262206389357e-10, + 3.2806185545908306e-10, + 3.281940552657403e-10 + ], + "bp_grad_norms_F": [ + 3.306513235656894e-06, + 1.1909767394513437e-08, + 1.165356344756674e-08, + 1.1650684861308491e-08, + 1.1658000786951561e-08 + ], + "gamma_dfa": 0.09604647590094828, + "acc_eval": 0.30078125, + "loss_eval": 1.9494699239730835, + "epoch": 59 + }, + { + "hidden_norms": [ + 35134.12890625, + 65285940.0, + 233831760.0, + 429654048.0, + 537931904.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.542536740676951e-08, + 3.2549649087165733e-10, + 3.224807365587168e-10, + 3.2198288479889925e-10, + 3.2189514942437825e-10 + ], + "bp_grad_norms_F": [ + 3.2077998639579164e-06, + 1.1713971126425804e-08, + 1.1479991179896842e-08, + 1.147721562233528e-08, + 1.1484392992144876e-08 + ], + "gamma_dfa": 0.09610223810159368, + "acc_eval": 0.298828125, + "loss_eval": 1.947687029838562, + "epoch": 60 + }, + { + "hidden_norms": [ + 35383.89453125, + 66896316.0, + 239557328.0, + 439032672.0, + 549329344.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.55890657949476e-08, + 3.2204855449080583e-10, + 3.1836161484832814e-10, + 3.185524344306856e-10, + 3.1883792828146795e-10 + ], + "bp_grad_norms_F": [ + 3.19811397275771e-06, + 1.1587792059231106e-08, + 1.1364262419988336e-08, + 1.1361324325775968e-08, + 1.1368459951199839e-08 + ], + "gamma_dfa": 0.09620415412791772, + "acc_eval": 0.3046875, + "loss_eval": 1.9550621509552002, + "epoch": 61 + }, + { + "hidden_norms": [ + 35603.75390625, + 68393976.0, + 244906672.0, + 448738336.0, + 561222784.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.4847117070266904e-08, + 3.208481536010055e-10, + 3.1700905789300293e-10, + 3.1679742162893376e-10, + 3.169834394967097e-10 + ], + "bp_grad_norms_F": [ + 3.1413667329616146e-06, + 1.1423910706298557e-08, + 1.1206890526693769e-08, + 1.1204000394116065e-08, + 1.1211046313519546e-08 + ], + "gamma_dfa": 0.09554143912100699, + "acc_eval": 0.302734375, + "loss_eval": 1.9572508335113525, + "epoch": 62 + }, + { + "hidden_norms": [ + 35840.75, + 69909448.0, + 249732624.0, + 457790144.0, + 572499584.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.482150200464275e-08, + 3.132541726014182e-10, + 3.1085201079861235e-10, + 3.1094149477439714e-10, + 3.1132280087220465e-10 + ], + "bp_grad_norms_F": [ + 3.1111580938159022e-06, + 1.1274166489272375e-08, + 1.1068144623038734e-08, + 1.1065306892987792e-08, + 1.1072265770906142e-08 + ], + "gamma_dfa": 0.09598036286843126, + "acc_eval": 0.2939453125, + "loss_eval": 1.9514048099517822, + "epoch": 63 + }, + { + "hidden_norms": [ + 36044.14453125, + 71301840.0, + 254043168.0, + 466339968.0, + 583323328.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.41697069422753e-08, + 3.088812261520246e-10, + 3.0549318630335165e-10, + 3.051613683968668e-10, + 3.050981411956144e-10 + ], + "bp_grad_norms_F": [ + 3.063583790208213e-06, + 1.1131353616633533e-08, + 1.0936378025405702e-08, + 1.0933466576545925e-08, + 1.0940373940115933e-08 + ], + "gamma_dfa": 0.09560732997124433, + "acc_eval": 0.3076171875, + "loss_eval": 1.9496949911117554, + "epoch": 64 + }, + { + "hidden_norms": [ + 36240.3046875, + 72753064.0, + 259111488.0, + 475471840.0, + 594591872.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.367900612895937e-08, + 3.0714322751812517e-10, + 3.043856555695612e-10, + 3.04040070897571e-10, + 3.043043039774318e-10 + ], + "bp_grad_norms_F": [ + 3.096502268817858e-06, + 1.1002981636920595e-08, + 1.0808172135057248e-08, + 1.0805252692591694e-08, + 1.0812099660029162e-08 + ], + "gamma_dfa": 0.09588889799488243, + "acc_eval": 0.2978515625, + "loss_eval": 1.951111078262329, + "epoch": 65 + }, + { + "hidden_norms": [ + 36427.96484375, + 74044192.0, + 263715872.0, + 483414560.0, + 604480512.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.3715726977543454e-08, + 3.022930794571721e-10, + 2.992986969374556e-10, + 2.9908994725325044e-10, + 2.9924074329557016e-10 + ], + "bp_grad_norms_F": [ + 3.0333421818795614e-06, + 1.0890616408687492e-08, + 1.0706762587631147e-08, + 1.0703858244198727e-08, + 1.071066169089363e-08 + ], + "gamma_dfa": 0.09630945282333414, + "acc_eval": 0.30078125, + "loss_eval": 1.949332594871521, + "epoch": 66 + }, + { + "hidden_norms": [ + 36613.015625, + 75303768.0, + 267802576.0, + 490809920.0, + 613719616.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.348211473688025e-08, + 3.012197435925401e-10, + 2.9878169383046327e-10, + 2.9841951132425493e-10, + 2.984068825373498e-10 + ], + "bp_grad_norms_F": [ + 3.008998646691907e-06, + 1.0781116444036343e-08, + 1.0603616651394532e-08, + 1.0600682998074262e-08, + 1.0607434930420823e-08 + ], + "gamma_dfa": 0.0960872639734589, + "acc_eval": 0.310546875, + "loss_eval": 1.9464266300201416, + "epoch": 67 + }, + { + "hidden_norms": [ + 36784.2265625, + 76478576.0, + 271792224.0, + 498011968.0, + 622710336.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.403291680750044e-08, + 3.0129157502223336e-10, + 2.977850743768329e-10, + 2.975975854635493e-10, + 2.9797145306709183e-10 + ], + "bp_grad_norms_F": [ + 2.9997727324371226e-06, + 1.0701310948491027e-08, + 1.052768094922385e-08, + 1.0524773941256171e-08, + 1.0531463701113353e-08 + ], + "gamma_dfa": 0.09624087948759552, + "acc_eval": 0.2939453125, + "loss_eval": 1.9560136795043945, + "epoch": 68 + }, + { + "hidden_norms": [ + 36953.8671875, + 77622176.0, + 275419712.0, + 504592704.0, + 630923392.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.3204310290766443e-08, + 2.9739929963135125e-10, + 2.9424951364376284e-10, + 2.9405439194718497e-10, + 2.94715835069681e-10 + ], + "bp_grad_norms_F": [ + 2.95851918963308e-06, + 1.0595774924127e-08, + 1.0427880781094245e-08, + 1.042505370918434e-08, + 1.0431726593651547e-08 + ], + "gamma_dfa": 0.09637677710998105, + "acc_eval": 0.2978515625, + "loss_eval": 1.950129747390747, + "epoch": 69 + }, + { + "hidden_norms": [ + 37103.0546875, + 78713944.0, + 279082240.0, + 511376384.0, + 639270976.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.3287564583115454e-08, + 2.9558042125010786e-10, + 2.94046592630437e-10, + 2.936023091315576e-10, + 2.9353278141464045e-10 + ], + "bp_grad_norms_F": [ + 2.9398986498563318e-06, + 1.0504810354916572e-08, + 1.034081797968156e-08, + 1.0337980249630618e-08, + 1.0344593626143705e-08 + ], + "gamma_dfa": 0.09638382289995207, + "acc_eval": 0.29296875, + "loss_eval": 1.9502758979797363, + "epoch": 70 + }, + { + "hidden_norms": [ + 37248.1015625, + 79737480.0, + 282511072.0, + 517720224.0, + 647238976.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.260229585246634e-08, + 2.899096518405031e-10, + 2.8717414557455356e-10, + 2.8704966181791747e-10, + 2.871551885164081e-10 + ], + "bp_grad_norms_F": [ + 2.9307011573109776e-06, + 1.0423523377767197e-08, + 1.0264206373733487e-08, + 1.0261346439222052e-08, + 1.0267889649639983e-08 + ], + "gamma_dfa": 0.0963258501906239, + "acc_eval": 0.3017578125, + "loss_eval": 1.949232816696167, + "epoch": 71 + }, + { + "hidden_norms": [ + 37381.0390625, + 80657776.0, + 285833920.0, + 523738208.0, + 654706368.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.2669339112298985e-08, + 2.929272380214343e-10, + 2.8994415202099333e-10, + 2.898675188767186e-10, + 2.900694684448979e-10 + ], + "bp_grad_norms_F": [ + 2.912831860157894e-06, + 1.0341300260563457e-08, + 1.018425699328418e-08, + 1.0181377518847512e-08, + 1.0187909182945987e-08 + ], + "gamma_dfa": 0.09613002402329585, + "acc_eval": 0.2958984375, + "loss_eval": 1.9482614994049072, + "epoch": 72 + }, + { + "hidden_norms": [ + 37505.28515625, + 81612512.0, + 289044000.0, + 529475296.0, + 661837376.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.200942965188915e-08, + 2.875680804592662e-10, + 2.8434027354862224e-10, + 2.8415880759524725e-10, + 2.840882251664567e-10 + ], + "bp_grad_norms_F": [ + 2.9117632038833108e-06, + 1.0266585803719863e-08, + 1.0111568471415922e-08, + 1.0108765380323348e-08, + 1.0115227766505086e-08 + ], + "gamma_dfa": 0.09620579873444512, + "acc_eval": 0.2958984375, + "loss_eval": 1.949747085571289, + "epoch": 73 + }, + { + "hidden_norms": [ + 37622.05078125, + 82508960.0, + 292035424.0, + 534928544.0, + 668760064.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.211632014834322e-08, + 2.8521121575586506e-10, + 2.8250474182200946e-10, + 2.822122813217476e-10, + 2.8265498275281686e-10 + ], + "bp_grad_norms_F": [ + 2.9012180675636046e-06, + 1.0211213208322079e-08, + 1.0059093114023199e-08, + 1.0056256272150677e-08, + 1.0062689348444565e-08 + ], + "gamma_dfa": 0.09616703895881074, + "acc_eval": 0.2998046875, + "loss_eval": 1.9510722160339355, + "epoch": 74 + }, + { + "hidden_norms": [ + 37739.125, + 83378024.0, + 294890528.0, + 540384832.0, + 675452224.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.20015284166675e-08, + 2.812508559379978e-10, + 2.7935267987722057e-10, + 2.7908611532900807e-10, + 2.792987507937994e-10 + ], + "bp_grad_norms_F": [ + 2.858193283827859e-06, + 1.0136964156970407e-08, + 9.987989102455685e-09, + 9.985158477832101e-09, + 9.991581784163373e-09 + ], + "gamma_dfa": 0.09625962336212979, + "acc_eval": 0.306640625, + "loss_eval": 1.949669599533081, + "epoch": 75 + }, + { + "hidden_norms": [ + 37841.23828125, + 84099792.0, + 297324992.0, + 545087552.0, + 681202944.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.180555717743118e-08, + 2.81841883165157e-10, + 2.790032094246442e-10, + 2.7877883335136744e-10, + 2.790592201762365e-10 + ], + "bp_grad_norms_F": [ + 2.8341921733954223e-06, + 1.0072185752107998e-08, + 9.925220645357058e-09, + 9.922425547870262e-09, + 9.928815103421584e-09 + ], + "gamma_dfa": 0.09619013674819143, + "acc_eval": 0.2958984375, + "loss_eval": 1.947105884552002, + "epoch": 76 + }, + { + "hidden_norms": [ + 37933.76953125, + 84846568.0, + 299677280.0, + 549384448.0, + 686714688.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.136923775239666e-08, + 2.810240096184913e-10, + 2.7924595968897847e-10, + 2.7894245246962157e-10, + 2.789228015220857e-10 + ], + "bp_grad_norms_F": [ + 2.822413307512761e-06, + 1.0021487639733095e-08, + 9.876163886701761e-09, + 9.873332373899757e-09, + 9.879697948633748e-09 + ], + "gamma_dfa": 0.09583658421615837, + "acc_eval": 0.306640625, + "loss_eval": 1.9495177268981934, + "epoch": 77 + }, + { + "hidden_norms": [ + 38020.0, + 85562224.0, + 301864096.0, + 553597760.0, + 691941760.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.160018545780076e-08, + 2.8198060553208393e-10, + 2.797063691772905e-10, + 2.7946911451692813e-10, + 2.796869680299352e-10 + ], + "bp_grad_norms_F": [ + 2.7969890652457252e-06, + 9.96650406648314e-09, + 9.82430670148915e-09, + 9.821477853222405e-09, + 9.827816782603804e-09 + ], + "gamma_dfa": 0.09604255547674256, + "acc_eval": 0.298828125, + "loss_eval": 1.9492969512939453, + "epoch": 78 + }, + { + "hidden_norms": [ + 38095.6171875, + 86194680.0, + 304047168.0, + 557639872.0, + 696924224.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.120950063997043e-08, + 2.782411523405415e-10, + 2.750368821580196e-10, + 2.7475391406461824e-10, + 2.748750949077561e-10 + ], + "bp_grad_norms_F": [ + 2.7781038625107612e-06, + 9.913026843832995e-09, + 9.771905951083681e-09, + 9.769114406310564e-09, + 9.775444453907767e-09 + ], + "gamma_dfa": 0.09590470486728009, + "acc_eval": 0.3037109375, + "loss_eval": 1.9471888542175293, + "epoch": 79 + }, + { + "hidden_norms": [ + 38161.02734375, + 86761784.0, + 305935040.0, + 560989248.0, + 701039488.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.1234330555871566e-08, + 2.762826911695271e-10, + 2.737765292248895e-10, + 2.7344454478495095e-10, + 2.735327242486818e-10 + ], + "bp_grad_norms_F": [ + 2.7851888262375724e-06, + 9.883654783493512e-09, + 9.742406881230181e-09, + 9.73962688277652e-09, + 9.74590719238222e-09 + ], + "gamma_dfa": 0.0961023918025603, + "acc_eval": 0.306640625, + "loss_eval": 1.9466722011566162, + "epoch": 80 + }, + { + "hidden_norms": [ + 38225.30859375, + 87283448.0, + 307635232.0, + 564114432.0, + 704943360.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.112997669698416e-08, + 2.7494201360056536e-10, + 2.7230390164945106e-10, + 2.724628855865774e-10, + 2.7264715485308955e-10 + ], + "bp_grad_norms_F": [ + 2.775808070509811e-06, + 9.851329529908526e-09, + 9.711465409623088e-09, + 9.708682746634167e-09, + 9.714950621741991e-09 + ], + "gamma_dfa": 0.09609865310994792, + "acc_eval": 0.30078125, + "loss_eval": 1.9487848281860352, + "epoch": 81 + }, + { + "hidden_norms": [ + 38286.74609375, + 87760048.0, + 309161312.0, + 566976960.0, + 708521920.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.100326561091606e-08, + 2.750530914141791e-10, + 2.7221699894219853e-10, + 2.720162428637707e-10, + 2.72086408958927e-10 + ], + "bp_grad_norms_F": [ + 2.76197988569038e-06, + 9.813816426174071e-09, + 9.67496749382235e-09, + 9.672206147115503e-09, + 9.678448265049155e-09 + ], + "gamma_dfa": 0.09604053818475222, + "acc_eval": 0.3017578125, + "loss_eval": 1.946324110031128, + "epoch": 82 + }, + { + "hidden_norms": [ + 38345.609375, + 88198008.0, + 310597792.0, + 569628928.0, + 711849600.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.150431903189201e-08, + 2.739733162560043e-10, + 2.706747326275405e-10, + 2.703801071923806e-10, + 2.703643420254309e-10 + ], + "bp_grad_norms_F": [ + 2.754284651018679e-06, + 9.79015535307326e-09, + 9.652834087603424e-09, + 9.650069188182897e-09, + 9.656329069684944e-09 + ], + "gamma_dfa": 0.09618356187274912, + "acc_eval": 0.3017578125, + "loss_eval": 1.952025294303894, + "epoch": 83 + }, + { + "hidden_norms": [ + 38393.00390625, + 88631992.0, + 311868064.0, + 571956672.0, + 714769600.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.135545677603659e-08, + 2.741361304625656e-10, + 2.7016161530113436e-10, + 2.7007868164119486e-10, + 2.702876256144293e-10 + ], + "bp_grad_norms_F": [ + 2.7439270979812136e-06, + 9.758214680744004e-09, + 9.621899721423688e-09, + 9.619145480144198e-09, + 9.625373387223135e-09 + ], + "gamma_dfa": 0.09624775896372739, + "acc_eval": 0.302734375, + "loss_eval": 1.9458943605422974, + "epoch": 84 + }, + { + "hidden_norms": [ + 38432.8515625, + 89003520.0, + 313081504.0, + 574038272.0, + 717387392.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.1023032909824906e-08, + 2.709044100157598e-10, + 2.6825267007701825e-10, + 2.6804419794856926e-10, + 2.6862953528272726e-10 + ], + "bp_grad_norms_F": [ + 2.7511816824699054e-06, + 9.751657259471358e-09, + 9.615223284242802e-09, + 9.612453943930177e-09, + 9.618667640154399e-09 + ], + "gamma_dfa": 0.09613665863071219, + "acc_eval": 0.3095703125, + "loss_eval": 1.9508092403411865, + "epoch": 85 + }, + { + "hidden_norms": [ + 38471.12890625, + 89306096.0, + 314023616.0, + 575863488.0, + 719706496.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.104102740460803e-08, + 2.722161940305057e-10, + 2.698590240157728e-10, + 2.697040091259595e-10, + 2.698894441266475e-10 + ], + "bp_grad_norms_F": [ + 2.731306722125737e-06, + 9.717195048608573e-09, + 9.582218574166745e-09, + 9.579466997422514e-09, + 9.585680693646736e-09 + ], + "gamma_dfa": 0.09615167022275273, + "acc_eval": 0.296875, + "loss_eval": 1.9497017860412598, + "epoch": 86 + }, + { + "hidden_norms": [ + 38506.49609375, + 89593104.0, + 314968672.0, + 577511168.0, + 721761728.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.099763100702148e-08, + 2.7151769721456276e-10, + 2.698610224172171e-10, + 2.6943663966605413e-10, + 2.6953861365086595e-10 + ], + "bp_grad_norms_F": [ + 2.73057457889081e-06, + 9.698882585951196e-09, + 9.564125491579034e-09, + 9.561379243905321e-09, + 9.567577841096409e-09 + ], + "gamma_dfa": 0.09604650201799814, + "acc_eval": 0.2998046875, + "loss_eval": 1.948227882385254, + "epoch": 87 + }, + { + "hidden_norms": [ + 38536.46875, + 89827632.0, + 315715328.0, + 578913664.0, + 723505408.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.074277709127273e-08, + 2.7221166987168033e-10, + 2.7026783588901537e-10, + 2.700663026544703e-10, + 2.7004307123768e-10 + ], + "bp_grad_norms_F": [ + 2.727413175307447e-06, + 9.684548274435656e-09, + 9.549998125635284e-09, + 9.547248325247892e-09, + 9.553437152476363e-09 + ], + "gamma_dfa": 0.09602321147394832, + "acc_eval": 0.2978515625, + "loss_eval": 1.9482835531234741, + "epoch": 88 + }, + { + "hidden_norms": [ + 38560.86328125, + 90036944.0, + 316345728.0, + 580143808.0, + 725043392.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.10075821580358e-08, + 2.70407807256845e-10, + 2.6787719265009e-10, + 2.677959520802631e-10, + 2.679190480581184e-10 + ], + "bp_grad_norms_F": [ + 2.7208343453821726e-06, + 9.672045386821537e-09, + 9.538242196072133e-09, + 9.535500389290519e-09, + 9.541698986481606e-09 + ], + "gamma_dfa": 0.09616725159867201, + "acc_eval": 0.3017578125, + "loss_eval": 1.9490463733673096, + "epoch": 89 + }, + { + "hidden_norms": [ + 38581.62890625, + 90197056.0, + 316901888.0, + 581085248.0, + 726214080.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.062888064344406e-08, + 2.7050772732906125e-10, + 2.680498045748436e-10, + 2.6770730077174676e-10, + 2.6768456895531756e-10 + ], + "bp_grad_norms_F": [ + 2.7148839762958232e-06, + 9.659022026653474e-09, + 9.525311206459719e-09, + 9.522568511499685e-09, + 9.528759115084995e-09 + ], + "gamma_dfa": 0.09604255046724575, + "acc_eval": 0.2998046875, + "loss_eval": 1.9472681283950806, + "epoch": 90 + }, + { + "hidden_norms": [ + 38598.60546875, + 90331488.0, + 317399072.0, + 581959360.0, + 727306816.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.070089059699967e-08, + 2.706254109696715e-10, + 2.6912555517455417e-10, + 2.687201572371123e-10, + 2.686441624710767e-10 + ], + "bp_grad_norms_F": [ + 2.7157270778843667e-06, + 9.650602983413137e-09, + 9.516874399650987e-09, + 9.514130816512534e-09, + 9.520310761956807e-09 + ], + "gamma_dfa": 0.09604466614109697, + "acc_eval": 0.296875, + "loss_eval": 1.9475460052490234, + "epoch": 91 + }, + { + "hidden_norms": [ + 38612.046875, + 90438432.0, + 317761152.0, + 582634240.0, + 728132928.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.071476039120171e-08, + 2.693741063541921e-10, + 2.670912935265335e-10, + 2.670463572496118e-10, + 2.6704632949403617e-10 + ], + "bp_grad_norms_F": [ + 2.7161215712112607e-06, + 9.646365484172748e-09, + 9.512908683007026e-09, + 9.51017486983119e-09, + 9.516351262561784e-09 + ], + "gamma_dfa": 0.09614986698579742, + "acc_eval": 0.3017578125, + "loss_eval": 1.947176218032837, + "epoch": 92 + }, + { + "hidden_norms": [ + 38622.60546875, + 90514344.0, + 318023872.0, + 583138880.0, + 728758976.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.0669719087181875e-08, + 2.7007937553058525e-10, + 2.6770055616687216e-10, + 2.676309729388038e-10, + 2.678066102212995e-10 + ], + "bp_grad_norms_F": [ + 2.715124537644442e-06, + 9.639773423941733e-09, + 9.506337939058085e-09, + 9.503595244098051e-09, + 9.509768084114967e-09 + ], + "gamma_dfa": 0.09608547511743382, + "acc_eval": 0.30078125, + "loss_eval": 1.9474897384643555, + "epoch": 93 + }, + { + "hidden_norms": [ + 38631.16796875, + 90589928.0, + 318254208.0, + 583549632.0, + 729273152.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.0695071251993795e-08, + 2.704435009270867e-10, + 2.6853075318911124e-10, + 2.682237765228024e-10, + 2.6814858666845964e-10 + ], + "bp_grad_norms_F": [ + 2.7134594802191714e-06, + 9.634878672670766e-09, + 9.501549769197482e-09, + 9.498809738772707e-09, + 9.504978137897524e-09 + ], + "gamma_dfa": 0.0960953844114556, + "acc_eval": 0.2978515625, + "loss_eval": 1.9472770690917969, + "epoch": 94 + }, + { + "hidden_norms": [ + 38638.31640625, + 90643280.0, + 318415136.0, + 583858560.0, + 729656640.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.078240761235975e-08, + 2.6967122979115743e-10, + 2.669718335290838e-10, + 2.6691132637424175e-10, + 2.6697202781811313e-10 + ], + "bp_grad_norms_F": [ + 2.711297611313057e-06, + 9.631804687160184e-09, + 9.498727138179675e-09, + 9.495988884111739e-09, + 9.502159947771815e-09 + ], + "gamma_dfa": 0.09613717825050117, + "acc_eval": 0.2998046875, + "loss_eval": 1.94759202003479, + "epoch": 95 + }, + { + "hidden_norms": [ + 38642.26171875, + 90677816.0, + 318520992.0, + 584046976.0, + 729889984.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.069002995128358e-08, + 2.6985705336990407e-10, + 2.6735619274020905e-10, + 2.671637633344659e-10, + 2.6708837919109385e-10 + ], + "bp_grad_norms_F": [ + 2.7110704650112893e-06, + 9.629748554118578e-09, + 9.496692321420142e-09, + 9.493953179173786e-09, + 9.500122466477023e-09 + ], + "gamma_dfa": 0.09612438916155952, + "acc_eval": 0.2998046875, + "loss_eval": 1.9474642276763916, + "epoch": 96 + }, + { + "hidden_norms": [ + 38644.16015625, + 90695256.0, + 318576288.0, + 584137856.0, + 730003328.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.070784681038276e-08, + 2.6936791686082984e-10, + 2.671484977678773e-10, + 2.668835430430505e-10, + 2.671484977678773e-10 + ], + "bp_grad_norms_F": [ + 2.710563421715051e-06, + 9.628578823139833e-09, + 9.495578545681838e-09, + 9.492838515257063e-09, + 9.499009578917139e-09 + ], + "gamma_dfa": 0.09612188487153617, + "acc_eval": 0.298828125, + "loss_eval": 1.9476267099380493, + "epoch": 97 + }, + { + "hidden_norms": [ + 38646.0625, + 90710400.0, + 318627968.0, + 584229760.0, + 730119232.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.075278863841959e-08, + 2.692224221334527e-10, + 2.671664001141494e-10, + 2.669355292361786e-10, + 2.6700280875147087e-10 + ], + "bp_grad_norms_F": [ + 2.7103426418761956e-06, + 9.62763913037179e-09, + 9.494667274623225e-09, + 9.49192813237687e-09, + 9.498098307858527e-09 + ], + "gamma_dfa": 0.09612667236069683, + "acc_eval": 0.2998046875, + "loss_eval": 1.9476723670959473, + "epoch": 98 + }, + { + "hidden_norms": [ + 38646.953125, + 90717408.0, + 318651872.0, + 584275776.0, + 730176960.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.076975639894954e-08, + 2.691815659261465e-10, + 2.671429466527542e-10, + 2.669321430559535e-10, + 2.669318377446217e-10 + ], + "bp_grad_norms_F": [ + 2.7102296371595003e-06, + 9.62716750763093e-09, + 9.494214303629178e-09, + 9.491475161382823e-09, + 9.4976462250429e-09 + ], + "gamma_dfa": 0.09612801273397054, + "acc_eval": 0.2998046875, + "loss_eval": 1.9476864337921143, + "epoch": 99 + }, + { + "hidden_norms": [ + 38646.98046875, + 90717592.0, + 318652416.0, + 584276736.0, + 730178112.0 + ], + "bp_grad_norms_per_sample_med": [ + 5.076372033840926e-08, + 2.692036038531853e-10, + 2.6715371581609304e-10, + 2.669119369969053e-10, + 2.6695023969125486e-10 + ], + "bp_grad_norms_F": [ + 2.710211447265465e-06, + 9.627171060344608e-09, + 9.494216968164437e-09, + 9.491476937739662e-09, + 9.497648001399739e-09 + ], + "gamma_dfa": 0.09612751441454748, + "acc_eval": 0.2998046875, + "loss_eval": 1.947685956954956, + "epoch": 100 + } + ] +}
\ No newline at end of file diff --git a/results/snapshot_evolution_v2/snapshot_fa_canonical_s42.json b/results/snapshot_evolution_v2/snapshot_fa_canonical_s42.json new file mode 100644 index 0000000..d29ad2d --- /dev/null +++ b/results/snapshot_evolution_v2/snapshot_fa_canonical_s42.json @@ -0,0 +1,2734 @@ +{ + "fa_log": [ + { + "hidden_norms": [ + 8.718070030212402, + 8.780964851379395, + 8.812002182006836, + 8.828720092773438, + 8.893179893493652 + ], + "bp_grad_norms_per_sample_med": [ + 0.0010068593546748161, + 0.000997236929833889, + 0.0009934091940522194, + 0.0009870644425973296, + 0.0009829498594626784 + ], + "bp_grad_norms_F": [ + 0.03544403985142708, + 0.03507080301642418, + 0.03477022796869278, + 0.03443896397948265, + 0.03407429903745651 + ], + "gamma_dfa": NaN, + "acc_eval": 0.115234375, + "loss_eval": 2.383650302886963, + "epoch": 0 + }, + { + "hidden_norms": [ + 634.95556640625, + 4686.67578125, + 8548.919921875, + 9006.0517578125, + 8484.302734375 + ], + "bp_grad_norms_per_sample_med": [ + 2.975244342451333e-06, + 1.1454501418484142e-06, + 1.1543062328200904e-06, + 1.164633545158722e-06, + 1.1064564660046017e-06 + ], + "bp_grad_norms_F": [ + 0.0001923188247019425, + 9.00624945643358e-05, + 6.289214798016474e-05, + 5.813155075884424e-05, + 5.113427323522046e-05 + ], + "gamma_dfa": NaN, + "acc_eval": 0.27734375, + "loss_eval": 1.9464800357818604, + "epoch": 1 + }, + { + "hidden_norms": [ + 838.77099609375, + 8185.21484375, + 20076.259765625, + 21714.076171875, + 19607.720703125 + ], + "bp_grad_norms_per_sample_med": [ + 2.675848463695729e-06, + 5.853078732798167e-07, + 5.769501285612932e-07, + 5.794665298708424e-07, + 5.573179464590794e-07 + ], + "bp_grad_norms_F": [ + 0.00015455791435670108, + 2.4559509256505407e-05, + 2.031458825513255e-05, + 2.0197476260364056e-05, + 1.898946902656462e-05 + ], + "gamma_dfa": NaN, + "acc_eval": 0.2958984375, + "loss_eval": 1.932114601135254, + "epoch": 2 + }, + { + "hidden_norms": [ + 966.05859375, + 10760.646484375, + 31344.212890625, + 35515.1015625, + 30314.041015625 + ], + "bp_grad_norms_per_sample_med": [ + 2.748855649770121e-06, + 4.3298211949149845e-07, + 4.17521391682385e-07, + 4.17111380102142e-07, + 3.9905967241793405e-07 + ], + "bp_grad_norms_F": [ + 0.00015430455096065998, + 1.7030582966981456e-05, + 1.4290850231191143e-05, + 1.4243430996430106e-05, + 1.3498008229362313e-05 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3203125, + "loss_eval": 1.9043192863464355, + "epoch": 3 + }, + { + "hidden_norms": [ + 1071.293212890625, + 12608.09765625, + 40531.1171875, + 47819.078125, + 38897.390625 + ], + "bp_grad_norms_per_sample_med": [ + 2.733014980549342e-06, + 3.670443504688592e-07, + 3.5215543903177604e-07, + 3.5238662121628295e-07, + 3.3813631716839154e-07 + ], + "bp_grad_norms_F": [ + 0.0001624817814445123, + 1.4486134205071721e-05, + 1.1960014489886817e-05, + 1.1927315426873975e-05, + 1.1350254681019578e-05 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3115234375, + "loss_eval": 1.898539662361145, + "epoch": 4 + }, + { + "hidden_norms": [ + 1170.5045166015625, + 14295.359375, + 49885.03515625, + 61276.62890625, + 47640.921875 + ], + "bp_grad_norms_per_sample_med": [ + 2.690757355594542e-06, + 3.267070951551432e-07, + 3.0663491656923725e-07, + 3.0629081493316335e-07, + 2.9683877755815047e-07 + ], + "bp_grad_norms_F": [ + 0.0001624024735065177, + 1.2482374586397782e-05, + 1.0314163773728069e-05, + 1.0298784218321089e-05, + 9.855078133114148e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.322265625, + "loss_eval": 1.877720832824707, + "epoch": 5 + }, + { + "hidden_norms": [ + 1288.9276123046875, + 15613.23046875, + 57501.26171875, + 73376.796875, + 54583.2109375 + ], + "bp_grad_norms_per_sample_med": [ + 2.7759115255321376e-06, + 3.0544967444257054e-07, + 2.801086793624563e-07, + 2.7996006224384473e-07, + 2.7101373234472703e-07 + ], + "bp_grad_norms_F": [ + 0.00017124316946137697, + 1.1784563866967801e-05, + 9.484625479672104e-06, + 9.465281436860096e-06, + 9.092000254895538e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.32421875, + "loss_eval": 1.8702466487884521, + "epoch": 6 + }, + { + "hidden_norms": [ + 1437.197021484375, + 16777.841796875, + 65706.7109375, + 87232.4140625, + 61590.078125 + ], + "bp_grad_norms_per_sample_med": [ + 2.883781235141214e-06, + 2.9475407359313976e-07, + 2.6204861569567584e-07, + 2.6302603828298743e-07, + 2.5517954327369807e-07 + ], + "bp_grad_norms_F": [ + 0.00017113517969846725, + 1.1177045962540433e-05, + 8.825853001326323e-06, + 8.807965969026554e-06, + 8.485528269375209e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.33984375, + "loss_eval": 1.8411049842834473, + "epoch": 7 + }, + { + "hidden_norms": [ + 1575.2144775390625, + 17884.404296875, + 73215.0, + 101692.0703125, + 67757.890625 + ], + "bp_grad_norms_per_sample_med": [ + 2.8855845357611543e-06, + 2.785824051443342e-07, + 2.509512739834463e-07, + 2.5104517931140435e-07, + 2.451099874178908e-07 + ], + "bp_grad_norms_F": [ + 0.0001778283913154155, + 1.0928491974482313e-05, + 8.524083568772767e-06, + 8.510252882842906e-06, + 8.22034417069517e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.349609375, + "loss_eval": 1.8426380157470703, + "epoch": 8 + }, + { + "hidden_norms": [ + 1746.853515625, + 18578.95703125, + 80693.078125, + 117047.6171875, + 73706.9765625 + ], + "bp_grad_norms_per_sample_med": [ + 2.964599843835458e-06, + 2.738325974860345e-07, + 2.414699338260107e-07, + 2.418815370219818e-07, + 2.3542932581221976e-07 + ], + "bp_grad_norms_F": [ + 0.0001854781003203243, + 1.0656134691089392e-05, + 8.172302841558121e-06, + 8.16663850855548e-06, + 7.904830454208422e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3671875, + "loss_eval": 1.8274496793746948, + "epoch": 9 + }, + { + "hidden_norms": [ + 1938.662109375, + 19766.384765625, + 88785.7734375, + 134051.8125, + 79702.1796875 + ], + "bp_grad_norms_per_sample_med": [ + 2.93986431643134e-06, + 2.6766733185468183e-07, + 2.3016899319827644e-07, + 2.3030897011722118e-07, + 2.2565886581560335e-07 + ], + "bp_grad_norms_F": [ + 0.0001912089210236445, + 1.0689387636375614e-05, + 7.921232281660195e-06, + 7.924772944534197e-06, + 7.683997864660341e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3505859375, + "loss_eval": 1.8229554891586304, + "epoch": 10 + }, + { + "hidden_norms": [ + 2088.147705078125, + 20979.9453125, + 96985.96875, + 150667.6875, + 85126.3359375 + ], + "bp_grad_norms_per_sample_med": [ + 2.9922123303549597e-06, + 2.649965722412162e-07, + 2.2647282094112597e-07, + 2.272310553053103e-07, + 2.2135814958801348e-07 + ], + "bp_grad_norms_F": [ + 0.00019602400425355881, + 1.0553159881965257e-05, + 7.715923857176676e-06, + 7.728580385446548e-06, + 7.498896593460813e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.375, + "loss_eval": 1.8172534704208374, + "epoch": 11 + }, + { + "hidden_norms": [ + 2276.7451171875, + 22148.2578125, + 105343.15625, + 167593.515625, + 90718.5 + ], + "bp_grad_norms_per_sample_med": [ + 3.033006578334607e-06, + 2.6026333443951444e-07, + 2.1839655062194652e-07, + 2.189913459460513e-07, + 2.1411695172446343e-07 + ], + "bp_grad_norms_F": [ + 0.00019336632976774126, + 1.0404637578176335e-05, + 7.483776698791189e-06, + 7.508507223974448e-06, + 7.2862567321863025e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.349609375, + "loss_eval": 1.8100277185440063, + "epoch": 12 + }, + { + "hidden_norms": [ + 2470.05517578125, + 22943.138671875, + 111998.0078125, + 181164.46875, + 93911.8515625 + ], + "bp_grad_norms_per_sample_med": [ + 3.0791732115176274e-06, + 2.678765156360896e-07, + 2.2002203081683547e-07, + 2.2065630389533908e-07, + 2.1573649178208143e-07 + ], + "bp_grad_norms_F": [ + 0.00020370102720335126, + 1.0737571756180841e-05, + 7.488396931876196e-06, + 7.527175512223039e-06, + 7.304198788915528e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3447265625, + "loss_eval": 1.8219082355499268, + "epoch": 13 + }, + { + "hidden_norms": [ + 2655.71875, + 23620.771484375, + 118446.4453125, + 194098.046875, + 97110.40625 + ], + "bp_grad_norms_per_sample_med": [ + 3.03382375932415e-06, + 2.674543395642104e-07, + 2.1672505567948974e-07, + 2.1837567487636989e-07, + 2.127215310565589e-07 + ], + "bp_grad_norms_F": [ + 0.00019606151909101754, + 1.0727502740337513e-05, + 7.367311354755657e-06, + 7.4189733823004644e-06, + 7.2012962846201845e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3447265625, + "loss_eval": 1.800294041633606, + "epoch": 14 + }, + { + "hidden_norms": [ + 2864.51904296875, + 23962.1328125, + 123932.375, + 204264.46875, + 99147.40625 + ], + "bp_grad_norms_per_sample_med": [ + 3.0961696211306844e-06, + 2.757700485744863e-07, + 2.1832956065281905e-07, + 2.1977299979880627e-07, + 2.1560039442647394e-07 + ], + "bp_grad_norms_F": [ + 0.00020294770365580916, + 1.0936742910416797e-05, + 7.422260750900023e-06, + 7.487843049602816e-06, + 7.265869498951361e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.349609375, + "loss_eval": 1.8283555507659912, + "epoch": 15 + }, + { + "hidden_norms": [ + 3058.654296875, + 24515.28515625, + 129439.953125, + 215162.921875, + 101313.0703125 + ], + "bp_grad_norms_per_sample_med": [ + 3.0670128126075724e-06, + 2.704452413127001e-07, + 2.1300424180026312e-07, + 2.1552395423896087e-07, + 2.103476077763844e-07 + ], + "bp_grad_norms_F": [ + 0.00020699974265880883, + 1.1274227290414274e-05, + 7.348373856075341e-06, + 7.4233907980669755e-06, + 7.201060725492425e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3603515625, + "loss_eval": 1.8033101558685303, + "epoch": 16 + }, + { + "hidden_norms": [ + 3205.80859375, + 24759.69921875, + 132512.875, + 221758.34375, + 102296.3515625 + ], + "bp_grad_norms_per_sample_med": [ + 3.0818489449302433e-06, + 2.7902248689315456e-07, + 2.1470867750394973e-07, + 2.1584412479569437e-07, + 2.1120550286468642e-07 + ], + "bp_grad_norms_F": [ + 0.00019346778572071344, + 1.1271905350440647e-05, + 7.35169760446297e-06, + 7.441570687660715e-06, + 7.2173174885392655e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3623046875, + "loss_eval": 1.789717197418213, + "epoch": 17 + }, + { + "hidden_norms": [ + 3365.503173828125, + 24903.955078125, + 137239.453125, + 228336.28125, + 103774.4140625 + ], + "bp_grad_norms_per_sample_med": [ + 3.0317510208988097e-06, + 2.810125181440526e-07, + 2.1507479175397748e-07, + 2.1818523521233146e-07, + 2.1249555004487775e-07 + ], + "bp_grad_norms_F": [ + 0.00019401832832954824, + 1.144540510722436e-05, + 7.354777153523173e-06, + 7.453731996065471e-06, + 7.234774784592446e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3515625, + "loss_eval": 1.7909660339355469, + "epoch": 18 + }, + { + "hidden_norms": [ + 3517.615478515625, + 25088.482421875, + 140183.171875, + 234581.5, + 105112.640625 + ], + "bp_grad_norms_per_sample_med": [ + 3.0766200325160753e-06, + 2.8522768502625695e-07, + 2.1497518787327863e-07, + 2.1691003837531753e-07, + 2.1271530670219363e-07 + ], + "bp_grad_norms_F": [ + 0.00020355181186459959, + 1.1879135854542255e-05, + 7.4169261097267736e-06, + 7.52632467992953e-06, + 7.302520316443406e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.34765625, + "loss_eval": 1.8049894571304321, + "epoch": 19 + }, + { + "hidden_norms": [ + 3643.212890625, + 25572.47265625, + 143095.625, + 240103.59375, + 106815.9609375 + ], + "bp_grad_norms_per_sample_med": [ + 2.9240111416584114e-06, + 2.7889984721696237e-07, + 2.1295824126355e-07, + 2.145163620070889e-07, + 2.1029546815043432e-07 + ], + "bp_grad_norms_F": [ + 0.00019266517483629286, + 1.1654319678200409e-05, + 7.349848146986915e-06, + 7.464911050192313e-06, + 7.246624591061845e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3681640625, + "loss_eval": 1.7967158555984497, + "epoch": 20 + }, + { + "hidden_norms": [ + 3727.509521484375, + 26266.642578125, + 147179.796875, + 243896.734375, + 107961.3359375 + ], + "bp_grad_norms_per_sample_med": [ + 3.034621840924956e-06, + 2.95065575528497e-07, + 2.1733650612532074e-07, + 2.1865858457204013e-07, + 2.140783124104928e-07 + ], + "bp_grad_norms_F": [ + 0.00019002222688868642, + 1.179500304715475e-05, + 7.40604900784092e-06, + 7.5316506809031125e-06, + 7.309085049200803e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3603515625, + "loss_eval": 1.7879303693771362, + "epoch": 21 + }, + { + "hidden_norms": [ + 3814.343994140625, + 26307.63671875, + 149997.296875, + 248837.015625, + 109946.75 + ], + "bp_grad_norms_per_sample_med": [ + 3.0296112072392134e-06, + 2.891725046083593e-07, + 2.132598950765896e-07, + 2.164179448982395e-07, + 2.1202281175192184e-07 + ], + "bp_grad_norms_F": [ + 0.00019085934036411345, + 1.1828436981886625e-05, + 7.313527476071613e-06, + 7.4397207754373085e-06, + 7.220100542326691e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3798828125, + "loss_eval": 1.769087791442871, + "epoch": 22 + }, + { + "hidden_norms": [ + 3903.16357421875, + 26762.390625, + 151661.765625, + 252546.703125, + 112163.4375 + ], + "bp_grad_norms_per_sample_med": [ + 3.10251039081777e-06, + 2.9397492085081467e-07, + 2.119260216204566e-07, + 2.151843716546864e-07, + 2.0996778005155647e-07 + ], + "bp_grad_norms_F": [ + 0.0001914280146593228, + 1.1861766324727796e-05, + 7.28844088371261e-06, + 7.418064342346042e-06, + 7.190795258793514e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.349609375, + "loss_eval": 1.7941722869873047, + "epoch": 23 + }, + { + "hidden_norms": [ + 3989.95361328125, + 27301.33984375, + 155319.25, + 259232.390625, + 115495.6796875 + ], + "bp_grad_norms_per_sample_med": [ + 2.962329745059833e-06, + 2.8580453204085643e-07, + 2.0789140364740888e-07, + 2.0986689719393326e-07, + 2.051628200661071e-07 + ], + "bp_grad_norms_F": [ + 0.00018467944755684584, + 1.1658386029012036e-05, + 7.1250747168960515e-06, + 7.254411684698425e-06, + 7.026835191936698e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3642578125, + "loss_eval": 1.772944450378418, + "epoch": 24 + }, + { + "hidden_norms": [ + 4044.4833984375, + 27958.3046875, + 158277.765625, + 265767.4375, + 119247.171875 + ], + "bp_grad_norms_per_sample_med": [ + 2.939599653473124e-06, + 2.9012642244197195e-07, + 2.0495993169333815e-07, + 2.0806970724152052e-07, + 2.0302766756685742e-07 + ], + "bp_grad_norms_F": [ + 0.00017772591672837734, + 1.151475134975044e-05, + 7.005810857663164e-06, + 7.14335192242288e-06, + 6.918045983184129e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.359375, + "loss_eval": 1.7681654691696167, + "epoch": 25 + }, + { + "hidden_norms": [ + 4102.31298828125, + 28819.13671875, + 162886.59375, + 274888.375, + 124251.71875 + ], + "bp_grad_norms_per_sample_med": [ + 3.00775354844518e-06, + 2.8842808319495816e-07, + 1.9920004490359133e-07, + 2.0228824837431603e-07, + 1.9777876048010512e-07 + ], + "bp_grad_norms_F": [ + 0.00017743176431395113, + 1.1497872947074939e-05, + 6.855066203570459e-06, + 6.992981980147306e-06, + 6.764100362488534e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.361328125, + "loss_eval": 1.7643167972564697, + "epoch": 26 + }, + { + "hidden_norms": [ + 4156.68994140625, + 29263.09765625, + 167567.5625, + 286674.25, + 130533.7578125 + ], + "bp_grad_norms_per_sample_med": [ + 2.9521122542064404e-06, + 2.7994269657938275e-07, + 1.9472663836950233e-07, + 1.9846331156259112e-07, + 1.9457104372122558e-07 + ], + "bp_grad_norms_F": [ + 0.0001732901146169752, + 1.1156066648254637e-05, + 6.697610388073372e-06, + 6.834617579443147e-06, + 6.61404874335858e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.357421875, + "loss_eval": 1.7718162536621094, + "epoch": 27 + }, + { + "hidden_norms": [ + 4236.7529296875, + 30445.798828125, + 173905.171875, + 301198.6875, + 138844.546875 + ], + "bp_grad_norms_per_sample_med": [ + 2.825209094226011e-06, + 2.710314390697022e-07, + 1.868730663545648e-07, + 1.8901035048202175e-07, + 1.8439881444010098e-07 + ], + "bp_grad_norms_F": [ + 0.00016928529657889158, + 1.0838041816896293e-05, + 6.40136795482249e-06, + 6.533743089676136e-06, + 6.317317456705496e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3740234375, + "loss_eval": 1.7575714588165283, + "epoch": 28 + }, + { + "hidden_norms": [ + 4307.15966796875, + 30990.375, + 179990.59375, + 314925.53125, + 146306.109375 + ], + "bp_grad_norms_per_sample_med": [ + 2.882675744331209e-06, + 2.69783697603998e-07, + 1.8120582012670639e-07, + 1.8295163783932367e-07, + 1.7808179109124467e-07 + ], + "bp_grad_norms_F": [ + 0.0001705507020233199, + 1.0706517059588805e-05, + 6.220788691280177e-06, + 6.352388027153211e-06, + 6.134147952252533e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.38671875, + "loss_eval": 1.765124797821045, + "epoch": 29 + }, + { + "hidden_norms": [ + 4360.767578125, + 32488.16015625, + 189370.28125, + 333218.25, + 156773.703125 + ], + "bp_grad_norms_per_sample_med": [ + 2.8181013931316556e-06, + 2.653597448443179e-07, + 1.7396521911905438e-07, + 1.755686014348612e-07, + 1.7160699883334019e-07 + ], + "bp_grad_norms_F": [ + 0.00016797029820736498, + 1.0374783414590638e-05, + 5.9415992836875375e-06, + 6.068700713512953e-06, + 5.864551440026844e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3564453125, + "loss_eval": 1.755420207977295, + "epoch": 30 + }, + { + "hidden_norms": [ + 4420.99462890625, + 33248.23046875, + 197272.421875, + 350132.625, + 166295.0625 + ], + "bp_grad_norms_per_sample_med": [ + 2.8225279038451845e-06, + 2.5861882591016183e-07, + 1.676798291327941e-07, + 1.704754311049328e-07, + 1.6635479482829396e-07 + ], + "bp_grad_norms_F": [ + 0.00016677004168741405, + 1.0174546332564205e-05, + 5.748868716182187e-06, + 5.872501787962392e-06, + 5.676161890733056e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.365234375, + "loss_eval": 1.749579906463623, + "epoch": 31 + }, + { + "hidden_norms": [ + 4467.9111328125, + 34465.12109375, + 206817.0625, + 370568.6875, + 177992.796875 + ], + "bp_grad_norms_per_sample_med": [ + 2.797579327307176e-06, + 2.4461206749037956e-07, + 1.6084129583759932e-07, + 1.623946417339539e-07, + 1.5842591949422058e-07 + ], + "bp_grad_norms_F": [ + 0.0001667520118644461, + 9.997165761888027e-06, + 5.521564617083641e-06, + 5.637548838421935e-06, + 5.447109742817702e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3818359375, + "loss_eval": 1.7478370666503906, + "epoch": 32 + }, + { + "hidden_norms": [ + 4500.455078125, + 35232.3984375, + 215903.1875, + 387229.5, + 187304.4375 + ], + "bp_grad_norms_per_sample_med": [ + 2.744678795352229e-06, + 2.4473743565067707e-07, + 1.5567927391657577e-07, + 1.5809170861302846e-07, + 1.5448634371750813e-07 + ], + "bp_grad_norms_F": [ + 0.00016586907440796494, + 9.867669177765492e-06, + 5.356784186005825e-06, + 5.468943072628463e-06, + 5.285164206725312e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.380859375, + "loss_eval": 1.748937964439392, + "epoch": 33 + }, + { + "hidden_norms": [ + 4559.64990234375, + 36405.82421875, + 224989.71875, + 406526.59375, + 197279.796875 + ], + "bp_grad_norms_per_sample_med": [ + 2.676637450349517e-06, + 2.350530081685065e-07, + 1.5119638874239172e-07, + 1.5273755593625538e-07, + 1.4942507675641536e-07 + ], + "bp_grad_norms_F": [ + 0.00016609806334599853, + 9.700875125417951e-06, + 5.207733465795172e-06, + 5.317946943250718e-06, + 5.140104349266039e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.38671875, + "loss_eval": 1.747586965560913, + "epoch": 34 + }, + { + "hidden_norms": [ + 4600.470703125, + 37569.390625, + 234249.046875, + 424273.0625, + 207188.484375 + ], + "bp_grad_norms_per_sample_med": [ + 2.6595262170303613e-06, + 2.3298444773445226e-07, + 1.4830226291451254e-07, + 1.5017273824469157e-07, + 1.4676733428586886e-07 + ], + "bp_grad_norms_F": [ + 0.0001615065266378224, + 9.520201274426654e-06, + 5.091071670904057e-06, + 5.198567123443354e-06, + 5.027129645895911e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.375, + "loss_eval": 1.7455894947052002, + "epoch": 35 + }, + { + "hidden_norms": [ + 4655.232421875, + 38777.5, + 243977.203125, + 443307.78125, + 217682.03125 + ], + "bp_grad_norms_per_sample_med": [ + 2.7372755084797973e-06, + 2.3094322898487007e-07, + 1.4448018248458538e-07, + 1.465641616960056e-07, + 1.4282748850291682e-07 + ], + "bp_grad_norms_F": [ + 0.00016169788432307541, + 9.384648365085013e-06, + 4.931174771627411e-06, + 5.035632966610137e-06, + 4.871027158515062e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3779296875, + "loss_eval": 1.739393949508667, + "epoch": 36 + }, + { + "hidden_norms": [ + 4674.50634765625, + 39366.46875, + 253330.671875, + 458747.40625, + 225757.234375 + ], + "bp_grad_norms_per_sample_med": [ + 2.720141083045746e-06, + 2.3003921967301721e-07, + 1.4151289917663235e-07, + 1.4314728957742773e-07, + 1.4079574839342968e-07 + ], + "bp_grad_norms_F": [ + 0.00016218192467931658, + 9.336513357993681e-06, + 4.852768142882269e-06, + 4.95553604196175e-06, + 4.79518030260806e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3818359375, + "loss_eval": 1.7406929731369019, + "epoch": 37 + }, + { + "hidden_norms": [ + 4730.9697265625, + 40562.7578125, + 261285.078125, + 476049.71875, + 235002.046875 + ], + "bp_grad_norms_per_sample_med": [ + 2.727818809944438e-06, + 2.3385531733310927e-07, + 1.4104001877512928e-07, + 1.4261416936278692e-07, + 1.39172499302731e-07 + ], + "bp_grad_norms_F": [ + 0.00016533929738216102, + 9.380160008731764e-06, + 4.782661108038155e-06, + 4.884728696197271e-06, + 4.725173766928492e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3759765625, + "loss_eval": 1.7494721412658691, + "epoch": 38 + }, + { + "hidden_norms": [ + 4769.4833984375, + 41614.734375, + 272033.25, + 494194.5, + 244904.296875 + ], + "bp_grad_norms_per_sample_med": [ + 2.728792651396361e-06, + 2.2869069482567284e-07, + 1.3784755026335915e-07, + 1.3816679711453617e-07, + 1.3538770815557655e-07 + ], + "bp_grad_norms_F": [ + 0.0001621924457140267, + 9.226439942722209e-06, + 4.664412244892446e-06, + 4.759738658322021e-06, + 4.6096511141513474e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.380859375, + "loss_eval": 1.744396686553955, + "epoch": 39 + }, + { + "hidden_norms": [ + 4807.61669921875, + 41796.7890625, + 279173.46875, + 509024.46875, + 251937.796875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7552744086278835e-06, + 2.2436680069404247e-07, + 1.334142893938406e-07, + 1.3510819485418324e-07, + 1.3202044613080943e-07 + ], + "bp_grad_norms_F": [ + 0.00016327289631590247, + 9.199145097227301e-06, + 4.6049044613027945e-06, + 4.701471880252939e-06, + 4.55197186965961e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.37890625, + "loss_eval": 1.738051176071167, + "epoch": 40 + }, + { + "hidden_norms": [ + 4846.89013671875, + 43419.1953125, + 288358.6875, + 526415.1875, + 261093.328125 + ], + "bp_grad_norms_per_sample_med": [ + 2.7159530873177573e-06, + 2.2739833127616293e-07, + 1.3493598771674442e-07, + 1.3701293255508062e-07, + 1.3374133800425625e-07 + ], + "bp_grad_norms_F": [ + 0.00016205597785301507, + 9.20344336918788e-06, + 4.573969818011392e-06, + 4.670639100368135e-06, + 4.519301455729874e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3759765625, + "loss_eval": 1.769773244857788, + "epoch": 41 + }, + { + "hidden_norms": [ + 4880.62109375, + 44368.0, + 297320.0, + 543147.8125, + 270529.90625 + ], + "bp_grad_norms_per_sample_med": [ + 2.704309054024634e-06, + 2.2085772855007235e-07, + 1.31169457517899e-07, + 1.3294274481268076e-07, + 1.296902354397389e-07 + ], + "bp_grad_norms_F": [ + 0.000160810406669043, + 9.048065294336993e-06, + 4.454658665054012e-06, + 4.546407581074163e-06, + 4.4059761421522126e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.388671875, + "loss_eval": 1.7465946674346924, + "epoch": 42 + }, + { + "hidden_norms": [ + 4899.1767578125, + 45389.57421875, + 307315.8125, + 560231.5625, + 278935.6875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7406351819081465e-06, + 2.2317071568522806e-07, + 1.2826720308112272e-07, + 1.3064222059711028e-07, + 1.270270928444006e-07 + ], + "bp_grad_norms_F": [ + 0.00016026092635001987, + 8.999614692584146e-06, + 4.385735792311607e-06, + 4.476342382986331e-06, + 4.342015927250031e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.38671875, + "loss_eval": 1.73671293258667, + "epoch": 43 + }, + { + "hidden_norms": [ + 4931.53857421875, + 45391.578125, + 314715.1875, + 576663.1875, + 287751.21875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7501505428517703e-06, + 2.1621856660658523e-07, + 1.2502815138759615e-07, + 1.2642922797567735e-07, + 1.2348743894108338e-07 + ], + "bp_grad_norms_F": [ + 0.00015757110668346286, + 8.882019756129012e-06, + 4.295293820177903e-06, + 4.383605300972704e-06, + 4.2555047912173904e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.376953125, + "loss_eval": 1.7194021940231323, + "epoch": 44 + }, + { + "hidden_norms": [ + 4952.10205078125, + 46544.41015625, + 321747.78125, + 590124.0, + 293778.09375 + ], + "bp_grad_norms_per_sample_med": [ + 2.7445450996310683e-06, + 2.20402526451835e-07, + 1.2519362257989997e-07, + 1.2609659449935862e-07, + 1.243117537796934e-07 + ], + "bp_grad_norms_F": [ + 0.00015927865752018988, + 8.919399988371879e-06, + 4.280226676200982e-06, + 4.3682503019226715e-06, + 4.240421731083188e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.376953125, + "loss_eval": 1.7289605140686035, + "epoch": 45 + }, + { + "hidden_norms": [ + 4985.6728515625, + 47313.03515625, + 330026.15625, + 605693.8125, + 301341.84375 + ], + "bp_grad_norms_per_sample_med": [ + 2.751742158579873e-06, + 2.1928519799985224e-07, + 1.2418719563811464e-07, + 1.259736563952174e-07, + 1.2327782883403415e-07 + ], + "bp_grad_norms_F": [ + 0.0001606208534212783, + 8.962500942288898e-06, + 4.251486188877607e-06, + 4.337850896263262e-06, + 4.2124443098146e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.390625, + "loss_eval": 1.7371184825897217, + "epoch": 46 + }, + { + "hidden_norms": [ + 4993.990234375, + 47521.25390625, + 335434.3125, + 618341.0625, + 307157.4375 + ], + "bp_grad_norms_per_sample_med": [ + 2.7009368750441354e-06, + 2.1855973386664118e-07, + 1.2373016033961903e-07, + 1.255795467614007e-07, + 1.2245830305346317e-07 + ], + "bp_grad_norms_F": [ + 0.00015943347534630448, + 8.89301736606285e-06, + 4.223482847010018e-06, + 4.306960363464896e-06, + 4.1826651795418e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.384765625, + "loss_eval": 1.737732172012329, + "epoch": 47 + }, + { + "hidden_norms": [ + 5012.49951171875, + 48445.1953125, + 344918.46875, + 633703.8125, + 314618.65625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7400287763157394e-06, + 2.1858940613128652e-07, + 1.2219133793678338e-07, + 1.242535176970705e-07, + 1.2144892025389709e-07 + ], + "bp_grad_norms_F": [ + 0.0001590315659996122, + 8.84339078766061e-06, + 4.180086762062274e-06, + 4.265210009180009e-06, + 4.143982550885994e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.375, + "loss_eval": 1.7298643589019775, + "epoch": 48 + }, + { + "hidden_norms": [ + 5054.08984375, + 49201.015625, + 353929.5625, + 650649.9375, + 323234.34375 + ], + "bp_grad_norms_per_sample_med": [ + 2.710038643272128e-06, + 2.1170153274852055e-07, + 1.1893747853264358e-07, + 1.2112403169339814e-07, + 1.1869307314782418e-07 + ], + "bp_grad_norms_F": [ + 0.00015847652684897184, + 8.75549994816538e-06, + 4.11285736845457e-06, + 4.190015260974178e-06, + 4.073814125149511e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7279736995697021, + "epoch": 49 + }, + { + "hidden_norms": [ + 5079.68505859375, + 49818.85546875, + 363584.96875, + 665326.4375, + 330221.40625 + ], + "bp_grad_norms_per_sample_med": [ + 2.728055960687925e-06, + 2.1385960735642584e-07, + 1.1923917497824732e-07, + 1.2005260430214548e-07, + 1.1790963583280245e-07 + ], + "bp_grad_norms_F": [ + 0.00015702334349043667, + 8.663697371957824e-06, + 4.082327905052807e-06, + 4.160873686487321e-06, + 4.0494828681403305e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7221198081970215, + "epoch": 50 + }, + { + "hidden_norms": [ + 5103.2626953125, + 50236.62109375, + 366800.8125, + 675614.9375, + 334505.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.6967995836457703e-06, + 2.110563741553051e-07, + 1.1874865890604269e-07, + 1.1944737821067974e-07, + 1.17826154166778e-07 + ], + "bp_grad_norms_F": [ + 0.00015892359078861773, + 8.714905561646447e-06, + 4.066732344654156e-06, + 4.143567366554635e-06, + 4.033038294437574e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3955078125, + "loss_eval": 1.7186274528503418, + "epoch": 51 + }, + { + "hidden_norms": [ + 5129.00830078125, + 51124.4453125, + 375390.84375, + 689126.5, + 340674.15625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7418275294621708e-06, + 2.1331975119665003e-07, + 1.1826992363239697e-07, + 1.1942577771151264e-07, + 1.1719654224862097e-07 + ], + "bp_grad_norms_F": [ + 0.00015799023094587028, + 8.683533451403491e-06, + 4.048787559440825e-06, + 4.127006377530051e-06, + 4.019248081021942e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7216696739196777, + "epoch": 52 + }, + { + "hidden_norms": [ + 5147.63671875, + 51383.28515625, + 382888.34375, + 701867.8125, + 346661.03125 + ], + "bp_grad_norms_per_sample_med": [ + 2.725489366639522e-06, + 2.1360351354360319e-07, + 1.1696960910967391e-07, + 1.1808428013182493e-07, + 1.1615074413384718e-07 + ], + "bp_grad_norms_F": [ + 0.0001584803540026769, + 8.698161764186807e-06, + 4.018610525235999e-06, + 4.092659764864948e-06, + 3.9867554733064026e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.39453125, + "loss_eval": 1.7209246158599854, + "epoch": 53 + }, + { + "hidden_norms": [ + 5179.68603515625, + 52093.16796875, + 388203.75, + 712999.25, + 350465.40625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7409823815105483e-06, + 2.16206004211017e-07, + 1.1817400746849671e-07, + 1.1920442943846865e-07, + 1.1731118831903586e-07 + ], + "bp_grad_norms_F": [ + 0.00015836946840863675, + 8.735608389542904e-06, + 4.036091922898777e-06, + 4.111263024242362e-06, + 4.005241407867288e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3837890625, + "loss_eval": 1.7180767059326172, + "epoch": 54 + }, + { + "hidden_norms": [ + 5197.8408203125, + 53041.44140625, + 398597.125, + 728487.4375, + 358157.9375 + ], + "bp_grad_norms_per_sample_med": [ + 2.763840029729181e-06, + 2.1291030805059563e-07, + 1.1598489635389342e-07, + 1.1754219286785883e-07, + 1.1593917292884726e-07 + ], + "bp_grad_norms_F": [ + 0.0001594589266460389, + 8.711985174159054e-06, + 4.000611625087913e-06, + 4.07353172704461e-06, + 3.971201294916682e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.38671875, + "loss_eval": 1.7179795503616333, + "epoch": 55 + }, + { + "hidden_norms": [ + 5217.0146484375, + 53347.57421875, + 403986.625, + 738288.125, + 361666.125 + ], + "bp_grad_norms_per_sample_med": [ + 2.7599005534284515e-06, + 2.135690380100641e-07, + 1.1644724651205252e-07, + 1.166295646726212e-07, + 1.1492483054098557e-07 + ], + "bp_grad_norms_F": [ + 0.0001592675835127011, + 8.73087174113607e-06, + 4.000759417976951e-06, + 4.0730465116212144e-06, + 3.972524154960411e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.390625, + "loss_eval": 1.719031810760498, + "epoch": 56 + }, + { + "hidden_norms": [ + 5245.46875, + 53932.046875, + 410963.8125, + 750011.5, + 367171.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.727937499003019e-06, + 2.1207910094744875e-07, + 1.1586875103830607e-07, + 1.1747369654813156e-07, + 1.1563617619003708e-07 + ], + "bp_grad_norms_F": [ + 0.0001595055073266849, + 8.686207365826704e-06, + 3.98215252062073e-06, + 4.0539648580306675e-06, + 3.954477506340481e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3837890625, + "loss_eval": 1.7162381410598755, + "epoch": 57 + }, + { + "hidden_norms": [ + 5273.81396484375, + 54359.40234375, + 415035.71875, + 757034.5, + 368386.46875 + ], + "bp_grad_norms_per_sample_med": [ + 2.8185475002828753e-06, + 2.183954137535693e-07, + 1.180805142553254e-07, + 1.196149383986267e-07, + 1.1741713734636505e-07 + ], + "bp_grad_norms_F": [ + 0.00016158133803401142, + 8.819352842692751e-06, + 4.027346676593879e-06, + 4.101018930668943e-06, + 3.999525233666645e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3828125, + "loss_eval": 1.7294056415557861, + "epoch": 58 + }, + { + "hidden_norms": [ + 5279.26513671875, + 54860.85546875, + 420308.09375, + 767223.875, + 372461.6875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7814132863568375e-06, + 2.1827422358455806e-07, + 1.1745117944883532e-07, + 1.1896155172053113e-07, + 1.1699857083158349e-07 + ], + "bp_grad_norms_F": [ + 0.0001601481344550848, + 8.751929271966219e-06, + 4.01089027946e-06, + 4.083126441400964e-06, + 3.984826435043942e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3740234375, + "loss_eval": 1.7233260869979858, + "epoch": 59 + }, + { + "hidden_norms": [ + 5281.876953125, + 55172.76171875, + 426239.0, + 777712.75, + 377365.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.729702600845485e-06, + 2.1274833272855176e-07, + 1.14989177291136e-07, + 1.1596039684036441e-07, + 1.1409560585207146e-07 + ], + "bp_grad_norms_F": [ + 0.00016002384654711932, + 8.679173333803192e-06, + 3.97576059185667e-06, + 4.045130026497645e-06, + 3.949881829612423e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.392578125, + "loss_eval": 1.7190814018249512, + "epoch": 60 + }, + { + "hidden_norms": [ + 5293.4951171875, + 55554.5703125, + 430637.53125, + 784743.25, + 379242.96875 + ], + "bp_grad_norms_per_sample_med": [ + 2.768365220617852e-06, + 2.1479348788489006e-07, + 1.1614937989179452e-07, + 1.1718410775074517e-07, + 1.1549084888429206e-07 + ], + "bp_grad_norms_F": [ + 0.00016024228534661233, + 8.705451364221517e-06, + 3.991565336036729e-06, + 4.061004801769741e-06, + 3.96601899410598e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3837890625, + "loss_eval": 1.7091575860977173, + "epoch": 61 + }, + { + "hidden_norms": [ + 5305.26953125, + 56114.8203125, + 436835.59375, + 794036.0, + 382300.59375 + ], + "bp_grad_norms_per_sample_med": [ + 2.793117346300278e-06, + 2.1690949836283835e-07, + 1.1630865515144251e-07, + 1.1756346651736749e-07, + 1.1506058683607989e-07 + ], + "bp_grad_norms_F": [ + 0.00016134958423208445, + 8.74852321430808e-06, + 3.9943656702234875e-06, + 4.064597760589095e-06, + 3.969457793573383e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3798828125, + "loss_eval": 1.7154879570007324, + "epoch": 62 + }, + { + "hidden_norms": [ + 5302.8837890625, + 56558.90625, + 441340.78125, + 802203.0625, + 385590.65625 + ], + "bp_grad_norms_per_sample_med": [ + 2.8357644623611122e-06, + 2.1830236107689416e-07, + 1.174595283259805e-07, + 1.1849279246689548e-07, + 1.1699435020773308e-07 + ], + "bp_grad_norms_F": [ + 0.00016155940829776227, + 8.755636372370645e-06, + 4.0028130570135545e-06, + 4.07254128731438e-06, + 3.978685981564922e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.388671875, + "loss_eval": 1.7206263542175293, + "epoch": 63 + }, + { + "hidden_norms": [ + 5312.6650390625, + 56690.29296875, + 446444.65625, + 811108.375, + 389126.15625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7693422453012317e-06, + 2.0976162318220304e-07, + 1.1542074673798197e-07, + 1.161635623248003e-07, + 1.1368263841404769e-07 + ], + "bp_grad_norms_F": [ + 0.00016044928634073585, + 8.682382031111047e-06, + 3.968462351622293e-06, + 4.03698095396976e-06, + 3.944758191209985e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3935546875, + "loss_eval": 1.708216667175293, + "epoch": 64 + }, + { + "hidden_norms": [ + 5315.3154296875, + 57172.39453125, + 452320.78125, + 819345.4375, + 392431.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.7417515866545727e-06, + 2.1362264135404985e-07, + 1.1589729354000156e-07, + 1.165334211350455e-07, + 1.1489947127074629e-07 + ], + "bp_grad_norms_F": [ + 0.00016101017536129802, + 8.694333700987045e-06, + 3.9719238884572405e-06, + 4.039454324811231e-06, + 3.947835011786083e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3916015625, + "loss_eval": 1.7132065296173096, + "epoch": 65 + }, + { + "hidden_norms": [ + 5323.32421875, + 57353.53515625, + 455451.03125, + 825786.625, + 393529.90625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7779890388046624e-06, + 2.1516311221603246e-07, + 1.1656365472845209e-07, + 1.1796725374324524e-07, + 1.1570150348916286e-07 + ], + "bp_grad_norms_F": [ + 0.00016156666970346123, + 8.733955837669782e-06, + 3.986176579928724e-06, + 4.05381206292077e-06, + 3.9624515011382755e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3994140625, + "loss_eval": 1.715019702911377, + "epoch": 66 + }, + { + "hidden_norms": [ + 5342.94140625, + 57836.46875, + 461806.90625, + 835003.3125, + 397621.53125 + ], + "bp_grad_norms_per_sample_med": [ + 2.7782559754996328e-06, + 2.1394832572241285e-07, + 1.1512971553884199e-07, + 1.1638152841442206e-07, + 1.1455212955979732e-07 + ], + "bp_grad_norms_F": [ + 0.00016023094940464944, + 8.671009709360078e-06, + 3.967585598729784e-06, + 4.035073743580142e-06, + 3.945795015170006e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7059491872787476, + "epoch": 67 + }, + { + "hidden_norms": [ + 5353.94189453125, + 58288.83984375, + 465855.71875, + 840827.8125, + 399727.625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7943997338297777e-06, + 2.1217401524609159e-07, + 1.1551408363175142e-07, + 1.1611316352855283e-07, + 1.1411684397444333e-07 + ], + "bp_grad_norms_F": [ + 0.00015934662951622158, + 8.68945244292263e-06, + 3.9676901906204876e-06, + 4.034785888507031e-06, + 3.945680873584934e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.39453125, + "loss_eval": 1.703755259513855, + "epoch": 68 + }, + { + "hidden_norms": [ + 5350.173828125, + 58626.13671875, + 469388.5, + 847294.5, + 401525.625 + ], + "bp_grad_norms_per_sample_med": [ + 2.773014784906991e-06, + 2.1550022211158648e-07, + 1.1623347972999909e-07, + 1.1767290430952926e-07, + 1.1548346634526752e-07 + ], + "bp_grad_norms_F": [ + 0.00016030117694754153, + 8.702772902324796e-06, + 3.981078407377936e-06, + 4.047733909828821e-06, + 3.958340130338911e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3916015625, + "loss_eval": 1.7165981531143188, + "epoch": 69 + }, + { + "hidden_norms": [ + 5357.58251953125, + 58903.7734375, + 474182.15625, + 855194.375, + 404878.59375 + ], + "bp_grad_norms_per_sample_med": [ + 2.75364641311171e-06, + 2.1325861609966523e-07, + 1.1510206121556621e-07, + 1.1569208169248668e-07, + 1.1389116849613856e-07 + ], + "bp_grad_norms_F": [ + 0.00015995267312973738, + 8.656907994009089e-06, + 3.959937203035224e-06, + 4.026182978122961e-06, + 3.939108864869922e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3955078125, + "loss_eval": 1.7067822217941284, + "epoch": 70 + }, + { + "hidden_norms": [ + 5361.0703125, + 59035.59375, + 477104.5, + 859434.875, + 405244.875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7923097150051035e-06, + 2.1804021343996283e-07, + 1.1658453757945608e-07, + 1.1743448169454496e-07, + 1.1549956013823248e-07 + ], + "bp_grad_norms_F": [ + 0.0001615561341168359, + 8.730263289180584e-06, + 3.989315700891893e-06, + 4.055812496517319e-06, + 3.967216798628215e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.392578125, + "loss_eval": 1.713646411895752, + "epoch": 71 + }, + { + "hidden_norms": [ + 5369.7529296875, + 59398.04296875, + 480994.0625, + 865578.875, + 407758.6875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7749495075113373e-06, + 2.139253751920478e-07, + 1.1538476485384308e-07, + 1.1588154080754975e-07, + 1.1468705451989081e-07 + ], + "bp_grad_norms_F": [ + 0.00016089060227386653, + 8.69345421961043e-06, + 3.971756996179465e-06, + 4.03762214773451e-06, + 3.950478458136786e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3896484375, + "loss_eval": 1.7055977582931519, + "epoch": 72 + }, + { + "hidden_norms": [ + 5376.67578125, + 59606.38671875, + 484413.75, + 871224.125, + 409605.40625 + ], + "bp_grad_norms_per_sample_med": [ + 2.767387286439771e-06, + 2.163243522090852e-07, + 1.1596397797575264e-07, + 1.1663173182796527e-07, + 1.1466888594213742e-07 + ], + "bp_grad_norms_F": [ + 0.0001625583681743592, + 8.740000339457765e-06, + 3.985041985288262e-06, + 4.0507725316274446e-06, + 3.964435109082842e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3896484375, + "loss_eval": 1.717661738395691, + "epoch": 73 + }, + { + "hidden_norms": [ + 5388.92626953125, + 60038.8828125, + 487315.0, + 875672.0625, + 411209.46875 + ], + "bp_grad_norms_per_sample_med": [ + 2.811367039612378e-06, + 2.1376251879701158e-07, + 1.150508310843179e-07, + 1.1646184105984503e-07, + 1.1464238269809357e-07 + ], + "bp_grad_norms_F": [ + 0.0001601450494490564, + 8.67134895088384e-06, + 3.971680598624516e-06, + 4.03764806833351e-06, + 3.951014605263481e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.38671875, + "loss_eval": 1.7057151794433594, + "epoch": 74 + }, + { + "hidden_norms": [ + 5394.13818359375, + 60238.7265625, + 490408.5625, + 880235.0625, + 412592.1875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7481701181386597e-06, + 2.1436494535009842e-07, + 1.1508839747875754e-07, + 1.1683697209718957e-07, + 1.148769968040142e-07 + ], + "bp_grad_norms_F": [ + 0.00016096947365440428, + 8.67969538376201e-06, + 3.976166681241011e-06, + 4.041281044919742e-06, + 3.95562074118061e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3974609375, + "loss_eval": 1.711783528327942, + "epoch": 75 + }, + { + "hidden_norms": [ + 5398.15234375, + 60386.79296875, + 492903.0625, + 883714.5, + 413466.03125 + ], + "bp_grad_norms_per_sample_med": [ + 2.7468756798043614e-06, + 2.142960227047297e-07, + 1.155566522470508e-07, + 1.1617837003541354e-07, + 1.1447124137475839e-07 + ], + "bp_grad_norms_F": [ + 0.00016124852118082345, + 8.686849469086155e-06, + 3.9795741031412035e-06, + 4.044950401294045e-06, + 3.958904471801361e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3955078125, + "loss_eval": 1.7077192068099976, + "epoch": 76 + }, + { + "hidden_norms": [ + 5399.38623046875, + 60673.9140625, + 495671.34375, + 887343.3125, + 414498.4375 + ], + "bp_grad_norms_per_sample_med": [ + 2.782978071991238e-06, + 2.1490254198397452e-07, + 1.1598574900517633e-07, + 1.167711403127214e-07, + 1.1515145814655625e-07 + ], + "bp_grad_norms_F": [ + 0.00016076312749646604, + 8.698333658685442e-06, + 3.980411747761536e-06, + 4.046264621138107e-06, + 3.960503818234429e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3896484375, + "loss_eval": 1.7069586515426636, + "epoch": 77 + }, + { + "hidden_norms": [ + 5401.99755859375, + 60780.171875, + 498870.40625, + 892185.3125, + 416521.0625 + ], + "bp_grad_norms_per_sample_med": [ + 2.731095946728601e-06, + 2.1313489639851468e-07, + 1.1489591145164013e-07, + 1.1583144754467867e-07, + 1.1378151754115606e-07 + ], + "bp_grad_norms_F": [ + 0.00016048294492065907, + 8.671476280142087e-06, + 3.968764303863281e-06, + 4.033439381601056e-06, + 3.948562607547501e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.388671875, + "loss_eval": 1.7046585083007812, + "epoch": 78 + }, + { + "hidden_norms": [ + 5396.2353515625, + 60906.8125, + 499556.1875, + 893389.6875, + 416091.625 + ], + "bp_grad_norms_per_sample_med": [ + 2.8349218155199196e-06, + 2.165559180866694e-07, + 1.1532758037446911e-07, + 1.1579567171793315e-07, + 1.1431712465537203e-07 + ], + "bp_grad_norms_F": [ + 0.00016091841098386794, + 8.715818694327027e-06, + 3.983707301813411e-06, + 4.049661583849229e-06, + 3.963775725424057e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.39453125, + "loss_eval": 1.7074486017227173, + "epoch": 79 + }, + { + "hidden_norms": [ + 5399.12646484375, + 60870.21484375, + 502091.5625, + 896890.0625, + 417516.5625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7684259293891955e-06, + 2.148178452898719e-07, + 1.1549953882195041e-07, + 1.1610666916794798e-07, + 1.1443260916621512e-07 + ], + "bp_grad_norms_F": [ + 0.00016129721188917756, + 8.707292181497905e-06, + 3.980841484008124e-06, + 4.045855348522309e-06, + 3.9603528421139345e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.390625, + "loss_eval": 1.706810474395752, + "epoch": 80 + }, + { + "hidden_norms": [ + 5394.28564453125, + 61078.3671875, + 504476.65625, + 900378.125, + 418670.8125 + ], + "bp_grad_norms_per_sample_med": [ + 2.764037617453141e-06, + 2.1426544094538258e-07, + 1.156425071258127e-07, + 1.1587813730784546e-07, + 1.145299108884501e-07 + ], + "bp_grad_norms_F": [ + 0.00016113216406665742, + 8.712745511729736e-06, + 3.979288521804847e-06, + 4.0438499127049e-06, + 3.9589867810718715e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3896484375, + "loss_eval": 1.7053430080413818, + "epoch": 81 + }, + { + "hidden_norms": [ + 5396.7236328125, + 61213.38671875, + 505729.4375, + 902413.125, + 419251.0625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7847922865475994e-06, + 2.1385373827342846e-07, + 1.1590607584821555e-07, + 1.167506624710768e-07, + 1.1480997130774995e-07 + ], + "bp_grad_norms_F": [ + 0.0001612855412531644, + 8.721244739717804e-06, + 3.984357590525178e-06, + 4.048785740451422e-06, + 3.963945800933288e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.390625, + "loss_eval": 1.708754062652588, + "epoch": 82 + }, + { + "hidden_norms": [ + 5401.74169921875, + 61391.25, + 507562.5, + 904882.1875, + 420061.125 + ], + "bp_grad_norms_per_sample_med": [ + 2.7846790544572286e-06, + 2.140303649866837e-07, + 1.1582537240428792e-07, + 1.1669516197798657e-07, + 1.1512172903849205e-07 + ], + "bp_grad_norms_F": [ + 0.00016096761100925505, + 8.707896085979883e-06, + 3.9824740269978065e-06, + 4.0467593862558715e-06, + 3.962396021961467e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7056825160980225, + "epoch": 83 + }, + { + "hidden_norms": [ + 5407.85205078125, + 61435.578125, + 508690.6875, + 906681.875, + 420568.9375 + ], + "bp_grad_norms_per_sample_med": [ + 2.766970737866359e-06, + 2.144422381888944e-07, + 1.157119697836606e-07, + 1.1730713112001467e-07, + 1.1475398054017205e-07 + ], + "bp_grad_norms_F": [ + 0.00016140809748321772, + 8.720307960174978e-06, + 3.985259354521986e-06, + 4.050050392834237e-06, + 3.965534688177286e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.390625, + "loss_eval": 1.7068425416946411, + "epoch": 84 + }, + { + "hidden_norms": [ + 5412.962890625, + 61492.68359375, + 510179.25, + 908609.25, + 421252.96875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7769999633164844e-06, + 2.1622201984428102e-07, + 1.1603597727116721e-07, + 1.1669756361243344e-07, + 1.1481289163839392e-07 + ], + "bp_grad_norms_F": [ + 0.00016168547153938562, + 8.720241567061748e-06, + 3.984777777077397e-06, + 4.049422386742663e-06, + 3.9649389691476244e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.38671875, + "loss_eval": 1.7068109512329102, + "epoch": 85 + }, + { + "hidden_norms": [ + 5414.86083984375, + 61660.16015625, + 511083.3125, + 910132.875, + 421357.09375 + ], + "bp_grad_norms_per_sample_med": [ + 2.7625901566352695e-06, + 2.1562310337230883e-07, + 1.1591524184950686e-07, + 1.168542453910959e-07, + 1.1494233120856734e-07 + ], + "bp_grad_norms_F": [ + 0.00016218819655478, + 8.743301805225201e-06, + 3.990852292190539e-06, + 4.055550562043209e-06, + 3.9708543226879556e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.384765625, + "loss_eval": 1.7097513675689697, + "epoch": 86 + }, + { + "hidden_norms": [ + 5414.11181640625, + 61679.67578125, + 512268.53125, + 911841.3125, + 422135.65625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7821745334222214e-06, + 2.1437830355353071e-07, + 1.1559337309563489e-07, + 1.1630524454631086e-07, + 1.143072410059176e-07 + ], + "bp_grad_norms_F": [ + 0.00016158135258592665, + 8.71670272317715e-06, + 3.983236638305243e-06, + 4.047654329042416e-06, + 3.9634205677430145e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3916015625, + "loss_eval": 1.7057619094848633, + "epoch": 87 + }, + { + "hidden_norms": [ + 5415.7666015625, + 61667.71875, + 513312.21875, + 913202.4375, + 422493.96875 + ], + "bp_grad_norms_per_sample_med": [ + 2.759925337159075e-06, + 2.1446679454584228e-07, + 1.1589692405777896e-07, + 1.1621393980476569e-07, + 1.1463799864941393e-07 + ], + "bp_grad_norms_F": [ + 0.000161813193699345, + 8.719586730876472e-06, + 3.9842125261202455e-06, + 4.0484619603375904e-06, + 3.964251391153084e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.390625, + "loss_eval": 1.7056890726089478, + "epoch": 88 + }, + { + "hidden_norms": [ + 5416.599609375, + 61688.796875, + 513735.6875, + 914162.125, + 422800.5625 + ], + "bp_grad_norms_per_sample_med": [ + 2.7784999474533834e-06, + 2.1510695091819798e-07, + 1.1558491763707934e-07, + 1.1671390609535592e-07, + 1.1512633335541977e-07 + ], + "bp_grad_norms_F": [ + 0.0001624061696929857, + 8.740545126784127e-06, + 3.989640845247777e-06, + 4.053715656482382e-06, + 3.969642875745194e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7097570896148682, + "epoch": 89 + }, + { + "hidden_norms": [ + 5416.76220703125, + 61798.24609375, + 514785.84375, + 915175.75, + 423199.09375 + ], + "bp_grad_norms_per_sample_med": [ + 2.777203462756006e-06, + 2.1360766311318002e-07, + 1.1558247337006833e-07, + 1.1610516992277553e-07, + 1.1466264027149009e-07 + ], + "bp_grad_norms_F": [ + 0.0001616168738109991, + 8.719413926883135e-06, + 3.9822580220061354e-06, + 4.046389676659601e-06, + 3.9626197576581035e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3916015625, + "loss_eval": 1.7034153938293457, + "epoch": 90 + }, + { + "hidden_norms": [ + 5416.92138671875, + 61928.12109375, + 515030.59375, + 915798.25, + 423164.4375 + ], + "bp_grad_norms_per_sample_med": [ + 2.810946853060159e-06, + 2.1559939966664388e-07, + 1.1640541686119832e-07, + 1.1669366983824148e-07, + 1.1499378160806373e-07 + ], + "bp_grad_norms_F": [ + 0.00016196168144233525, + 8.733930371818133e-06, + 3.986880528827896e-06, + 4.051325959153473e-06, + 3.9673618630331475e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3857421875, + "loss_eval": 1.7049140930175781, + "epoch": 91 + }, + { + "hidden_norms": [ + 5416.93212890625, + 61895.34765625, + 515480.125, + 916358.0, + 423425.4375 + ], + "bp_grad_norms_per_sample_med": [ + 2.8099400424252963e-06, + 2.1475410960647423e-07, + 1.1618925555012538e-07, + 1.1643356145896178e-07, + 1.1480153716547647e-07 + ], + "bp_grad_norms_F": [ + 0.00016178778605535626, + 8.72767577675404e-06, + 3.985679995821556e-06, + 4.050090865348466e-06, + 3.966247277276125e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7039738893508911, + "epoch": 92 + }, + { + "hidden_norms": [ + 5417.095703125, + 61845.4921875, + 515914.6875, + 916804.125, + 423511.625 + ], + "bp_grad_norms_per_sample_med": [ + 2.805000349326292e-06, + 2.15853873442029e-07, + 1.1573133917863743e-07, + 1.1623403395333298e-07, + 1.146071184621178e-07 + ], + "bp_grad_norms_F": [ + 0.00016200476966332644, + 8.730704394110944e-06, + 3.986714546044823e-06, + 4.050946245115483e-06, + 3.967099928559037e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3896484375, + "loss_eval": 1.7049956321716309, + "epoch": 93 + }, + { + "hidden_norms": [ + 5417.3349609375, + 61879.95703125, + 516017.53125, + 916914.5625, + 423434.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.7787407361756777e-06, + 2.167610944070475e-07, + 1.1583406234194626e-07, + 1.1677065714366108e-07, + 1.1465835569879346e-07 + ], + "bp_grad_norms_F": [ + 0.0001623090065550059, + 8.742457794141956e-06, + 3.991137873526895e-06, + 4.055534191138577e-06, + 3.971431851823581e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7075328826904297, + "epoch": 94 + }, + { + "hidden_norms": [ + 5417.50390625, + 61889.82421875, + 516312.4375, + 917429.0, + 423711.03125 + ], + "bp_grad_norms_per_sample_med": [ + 2.7708026664186036e-06, + 2.1539200645293022e-07, + 1.1580210923511913e-07, + 1.16502334890356e-07, + 1.1404062405517834e-07 + ], + "bp_grad_norms_F": [ + 0.00016205437714233994, + 8.73019962455146e-06, + 3.986566753155785e-06, + 4.050732059113216e-06, + 3.966889835282927e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.388671875, + "loss_eval": 1.7053842544555664, + "epoch": 95 + }, + { + "hidden_norms": [ + 5417.48828125, + 61898.47265625, + 516375.0, + 917465.6875, + 423663.46875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7983871859760256e-06, + 2.1556444096404448e-07, + 1.160928064791733e-07, + 1.1642453046079027e-07, + 1.1440795333328424e-07 + ], + "bp_grad_norms_F": [ + 0.00016201405378524214, + 8.732341484574135e-06, + 3.987765467172721e-06, + 4.052003077958943e-06, + 3.9681326597929e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7055914402008057, + "epoch": 96 + }, + { + "hidden_norms": [ + 5417.54931640625, + 61916.09375, + 516515.78125, + 917698.75, + 423749.71875 + ], + "bp_grad_norms_per_sample_med": [ + 2.7831354145746445e-06, + 2.1518340531656577e-07, + 1.1582258707676374e-07, + 1.1653285980628425e-07, + 1.1448688042037247e-07 + ], + "bp_grad_norms_F": [ + 0.00016204272105824202, + 8.731909474590793e-06, + 3.987126092397375e-06, + 4.051275027450174e-06, + 3.967459178966237e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3896484375, + "loss_eval": 1.7053790092468262, + "epoch": 97 + }, + { + "hidden_norms": [ + 5417.61865234375, + 61911.2734375, + 516535.4375, + 917715.875, + 423733.46875 + ], + "bp_grad_norms_per_sample_med": [ + 2.784623802654096e-06, + 2.1553401552409923e-07, + 1.157067259782707e-07, + 1.1653578724235558e-07, + 1.1444880243516309e-07 + ], + "bp_grad_norms_F": [ + 0.0001621013943804428, + 8.73380577104399e-06, + 3.987797754234634e-06, + 4.05199125452782e-06, + 3.968138116761111e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3876953125, + "loss_eval": 1.7057509422302246, + "epoch": 98 + }, + { + "hidden_norms": [ + 5417.64111328125, + 61917.359375, + 516576.09375, + 917756.125, + 423741.78125 + ], + "bp_grad_norms_per_sample_med": [ + 2.785410288197454e-06, + 2.1554964746428595e-07, + 1.1583354364574916e-07, + 1.1650187303757775e-07, + 1.145084596032575e-07 + ], + "bp_grad_norms_F": [ + 0.00016210223839152604, + 8.733919457881711e-06, + 3.987840045738267e-06, + 4.052033091284102e-06, + 3.968175860791234e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.388671875, + "loss_eval": 1.7057793140411377, + "epoch": 99 + }, + { + "hidden_norms": [ + 5417.64697265625, + 61917.6640625, + 516580.8125, + 917765.6875, + 423747.84375 + ], + "bp_grad_norms_per_sample_med": [ + 2.786067398119485e-06, + 2.1562047436418652e-07, + 1.1590604742650612e-07, + 1.1655521348075126e-07, + 1.1442210023915322e-07 + ], + "bp_grad_norms_F": [ + 0.00016209435125347227, + 8.73375483934069e-06, + 3.987807303929003e-06, + 4.0519980757380836e-06, + 3.968141299992567e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.388671875, + "loss_eval": 1.705754280090332, + "epoch": 100 + } + ], + "seed": 42, + "depth": 4, + "d_hidden": 256 +}
\ No newline at end of file diff --git a/results/snapshot_evolution_v2/snapshot_fa_s42.json b/results/snapshot_evolution_v2/snapshot_fa_s42.json new file mode 100644 index 0000000..277acec --- /dev/null +++ b/results/snapshot_evolution_v2/snapshot_fa_s42.json @@ -0,0 +1,2734 @@ +{ + "fa_log": [ + { + "hidden_norms": [ + 8.718070030212402, + 8.780964851379395, + 8.812002182006836, + 8.828720092773438, + 8.893179893493652 + ], + "bp_grad_norms_per_sample_med": [ + 0.0010068593546748161, + 0.000997236929833889, + 0.0009934091940522194, + 0.0009870644425973296, + 0.0009829498594626784 + ], + "bp_grad_norms_F": [ + 0.03544403985142708, + 0.03507080301642418, + 0.03477022796869278, + 0.03443896397948265, + 0.03407429903745651 + ], + "gamma_dfa": NaN, + "acc_eval": 0.115234375, + "loss_eval": 2.383650302886963, + "epoch": 0 + }, + { + "hidden_norms": [ + 1156.72998046875, + 7932.5859375, + 11656.220703125, + 13101.765625, + 12094.2197265625 + ], + "bp_grad_norms_per_sample_med": [ + 1.5287704400179791e-06, + 8.219874985115894e-07, + 8.303726986014226e-07, + 8.411634553340264e-07, + 8.074007382674608e-07 + ], + "bp_grad_norms_F": [ + 0.0005889489548280835, + 0.0001249020715476945, + 0.00010261647548759356, + 8.311456622323021e-05, + 6.362409476423636e-05 + ], + "gamma_dfa": NaN, + "acc_eval": 0.2431640625, + "loss_eval": 2.0429816246032715, + "epoch": 1 + }, + { + "hidden_norms": [ + 2267.652099609375, + 33227.19921875, + 62863.03515625, + 73248.0546875, + 64100.21875 + ], + "bp_grad_norms_per_sample_med": [ + 6.826750791333325e-07, + 1.582481274908787e-07, + 1.5909284911685972e-07, + 1.6235998145930353e-07, + 1.57003455569793e-07 + ], + "bp_grad_norms_F": [ + 9.194230369757861e-05, + 3.062757241423242e-05, + 2.5034905775100924e-05, + 1.2339382010395639e-05, + 1.1088715837104246e-05 + ], + "gamma_dfa": NaN, + "acc_eval": 0.234375, + "loss_eval": 2.053070068359375, + "epoch": 2 + }, + { + "hidden_norms": [ + 3317.195556640625, + 69574.59375, + 208893.703125, + 272473.75, + 236858.453125 + ], + "bp_grad_norms_per_sample_med": [ + 5.585509370575892e-07, + 5.853191709093153e-08, + 5.772180244889569e-08, + 5.9099026117337417e-08, + 5.743537911939711e-08 + ], + "bp_grad_norms_F": [ + 5.4304215154843405e-05, + 1.798659286578186e-05, + 6.792580734327203e-06, + 5.748081548517803e-06, + 5.4292308959702495e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.2529296875, + "loss_eval": 2.109177350997925, + "epoch": 3 + }, + { + "hidden_norms": [ + 4279.80322265625, + 131147.296875, + 524522.6875, + 669556.125, + 582199.5 + ], + "bp_grad_norms_per_sample_med": [ + 4.5417544924930553e-07, + 2.7819011094720736e-08, + 2.7284025705398562e-08, + 2.7924896173203706e-08, + 2.717020919362767e-08 + ], + "bp_grad_norms_F": [ + 2.7083517124992795e-05, + 3.1135298286244506e-06, + 2.188051212215214e-06, + 1.643483983571059e-06, + 1.5356297353719128e-06 + ], + "gamma_dfa": NaN, + "acc_eval": 0.2314453125, + "loss_eval": 2.1304922103881836, + "epoch": 4 + }, + { + "hidden_norms": [ + 5095.1298828125, + 222828.6875, + 891649.1875, + 1124229.75, + 978070.3125 + ], + "bp_grad_norms_per_sample_med": [ + 3.9661114215050475e-07, + 1.7936926610673254e-08, + 1.787710601774961e-08, + 1.8164961090860743e-08, + 1.7773361449258118e-08 + ], + "bp_grad_norms_F": [ + 2.5229386665159836e-05, + 1.7291312133238534e-06, + 8.046768016356509e-07, + 6.81868698393373e-07, + 6.581432785424113e-07 + ], + "gamma_dfa": NaN, + "acc_eval": 0.2275390625, + "loss_eval": 2.1189823150634766, + "epoch": 5 + }, + { + "hidden_norms": [ + 5762.8369140625, + 345252.125, + 1352057.25, + 1685390.25, + 1472510.375 + ], + "bp_grad_norms_per_sample_med": [ + 3.5811123666462663e-07, + 1.2679316263586315e-08, + 1.2558022177699968e-08, + 1.2735733356805667e-08, + 1.2553626582700872e-08 + ], + "bp_grad_norms_F": [ + 1.9819366571027786e-05, + 7.985001957422355e-07, + 4.666244706186262e-07, + 4.659337093926297e-07, + 4.531105162186577e-07 + ], + "gamma_dfa": NaN, + "acc_eval": 0.2314453125, + "loss_eval": 2.1147799491882324, + "epoch": 6 + }, + { + "hidden_norms": [ + 6342.46337890625, + 489436.21875, + 1846600.125, + 2298539.25, + 2010095.0 + ], + "bp_grad_norms_per_sample_med": [ + 3.2295056939801725e-07, + 9.47152134500584e-09, + 9.449903082270339e-09, + 9.579181003971371e-09, + 9.398083200551355e-09 + ], + "bp_grad_norms_F": [ + 1.761323437676765e-05, + 3.5219920846429886e-07, + 3.464733993041591e-07, + 3.515923765462503e-07, + 3.4331398524045653e-07 + ], + "gamma_dfa": NaN, + "acc_eval": 0.263671875, + "loss_eval": 2.0468926429748535, + "epoch": 7 + }, + { + "hidden_norms": [ + 6895.3232421875, + 642604.125, + 2340199.25, + 2912852.0, + 2551695.75 + ], + "bp_grad_norms_per_sample_med": [ + 3.2115181625158584e-07, + 8.011411978259275e-09, + 8.093356207439228e-09, + 8.185297772911326e-09, + 8.036058929405954e-09 + ], + "bp_grad_norms_F": [ + 1.6510130080860108e-05, + 2.8924068828928284e-07, + 2.88436524442659e-07, + 2.9188095140852965e-07, + 2.860562631212815e-07 + ], + "gamma_dfa": NaN, + "acc_eval": 0.2373046875, + "loss_eval": 2.0705394744873047, + "epoch": 8 + }, + { + "hidden_norms": [ + 7418.32666015625, + 802952.8125, + 2905706.5, + 3629847.25, + 3176127.0 + ], + "bp_grad_norms_per_sample_med": [ + 2.9990744110364176e-07, + 6.746414982217175e-09, + 6.801009533319302e-09, + 6.844500521907548e-09, + 6.780156436292373e-09 + ], + "bp_grad_norms_F": [ + 1.5715995687060058e-05, + 2.419061502223485e-07, + 2.416737459043361e-07, + 2.441595654545381e-07, + 2.39673738633428e-07 + ], + "gamma_dfa": NaN, + "acc_eval": 0.271484375, + "loss_eval": 2.03157901763916, + "epoch": 9 + }, + { + "hidden_norms": [ + 8026.80126953125, + 964342.4375, + 3455197.75, + 4322591.0, + 3777789.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.9810524893036927e-07, + 5.931947377746383e-09, + 5.9767870652649435e-09, + 6.012730313642578e-09, + 5.922320855944463e-09 + ], + "bp_grad_norms_F": [ + 1.546428757137619e-05, + 2.1311132059054216e-07, + 2.1318879817044945e-07, + 2.1529912430651166e-07, + 2.115393868962201e-07 + ], + "gamma_dfa": NaN, + "acc_eval": 0.2421875, + "loss_eval": 2.029039144515991, + "epoch": 10 + }, + { + "hidden_norms": [ + 8568.4677734375, + 1146051.75, + 4149878.75, + 5211595.5, + 4551772.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.77834743656058e-07, + 5.156825633179096e-09, + 5.185044393840599e-09, + 5.243859568793141e-09, + 5.148059312176656e-09 + ], + "bp_grad_norms_F": [ + 1.4489689419860952e-05, + 1.8374593935277517e-07, + 1.8364994502917398e-07, + 1.8539546431384224e-07, + 1.8235705567803961e-07 + ], + "gamma_dfa": NaN, + "acc_eval": 0.279296875, + "loss_eval": 2.015061855316162, + "epoch": 11 + }, + { + "hidden_norms": [ + 9099.6689453125, + 1306440.25, + 4712579.5, + 5930700.5, + 5167806.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.789404618397384e-07, + 4.740023484828271e-09, + 4.688871069191691e-09, + 4.722021884617789e-09, + 4.664129082954105e-09 + ], + "bp_grad_norms_F": [ + 1.440922096662689e-05, + 1.6886964715467911e-07, + 1.6845007166921278e-07, + 1.6997903173887607e-07, + 1.673002003599322e-07 + ], + "gamma_dfa": NaN, + "acc_eval": 0.279296875, + "loss_eval": 2.0080716609954834, + "epoch": 12 + }, + { + "hidden_norms": [ + 9619.3310546875, + 1468314.25, + 5278606.0, + 6647205.0, + 5776389.0 + ], + "bp_grad_norms_per_sample_med": [ + 2.678217470020172e-07, + 4.431488953571261e-09, + 4.461886859985498e-09, + 4.481298443437254e-09, + 4.434627776106481e-09 + ], + "bp_grad_norms_F": [ + 1.4280021787271835e-05, + 1.5716064183379785e-07, + 1.5650925888621714e-07, + 1.5793726504398364e-07, + 1.5547414022876183e-07 + ], + "gamma_dfa": NaN, + "acc_eval": 0.294921875, + "loss_eval": 2.003105401992798, + "epoch": 13 + }, + { + "hidden_norms": [ + 10185.365234375, + 1637280.875, + 5941891.0, + 7477497.0, + 6482495.0 + ], + "bp_grad_norms_per_sample_med": [ + 2.580181046596408e-07, + 4.108497098087582e-09, + 4.027958855346014e-09, + 4.0670666834330405e-09, + 4.01310229491969e-09 + ], + "bp_grad_norms_F": [ + 1.3607143955596257e-05, + 1.4447114438098652e-07, + 1.4335948606003512e-07, + 1.4459322983384482e-07, + 1.424295135166176e-07 + ], + "gamma_dfa": NaN, + "acc_eval": 0.2841796875, + "loss_eval": 1.9895731210708618, + "epoch": 14 + }, + { + "hidden_norms": [ + 10773.4443359375, + 1777185.125, + 6485962.5, + 8158632.5, + 7057802.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.536919225804013e-07, + 3.8119583045670424e-09, + 3.815232130222057e-09, + 3.836773565524254e-09, + 3.790971536687948e-09 + ], + "bp_grad_norms_F": [ + 1.3741724615101703e-05, + 1.3785883368200302e-07, + 1.3661090747518756e-07, + 1.3781669849777245e-07, + 1.3573732360327995e-07 + ], + "gamma_dfa": NaN, + "acc_eval": 0.263671875, + "loss_eval": 2.0124382972717285, + "epoch": 15 + }, + { + "hidden_norms": [ + 11403.58984375, + 1931677.0, + 7089585.5, + 8906384.0, + 7683122.0 + ], + "bp_grad_norms_per_sample_med": [ + 2.446777500608732e-07, + 3.674859083702131e-09, + 3.648833679648078e-09, + 3.6761997890266684e-09, + 3.624884392650074e-09 + ], + "bp_grad_norms_F": [ + 1.3561247214965988e-05, + 1.2985380237751087e-07, + 1.2829688955662277e-07, + 1.294151275033073e-07, + 1.2746555455578346e-07 + ], + "gamma_dfa": NaN, + "acc_eval": 0.26953125, + "loss_eval": 2.0030646324157715, + "epoch": 16 + }, + { + "hidden_norms": [ + 12115.1162109375, + 2091857.0, + 7735447.5, + 9727085.0, + 8371443.5 + ], + "bp_grad_norms_per_sample_med": [ + 2.378987460360804e-07, + 3.4695852857424825e-09, + 3.42812689346772e-09, + 3.4526042025362358e-09, + 3.40047856539627e-09 + ], + "bp_grad_norms_F": [ + 1.3279566701385193e-05, + 1.2257437731477694e-07, + 1.2078584177288576e-07, + 1.2181956776657898e-07, + 1.1999593141354126e-07 + ], + "gamma_dfa": NaN, + "acc_eval": 0.2939453125, + "loss_eval": 1.9935777187347412, + "epoch": 17 + }, + { + "hidden_norms": [ + 12755.7646484375, + 2248501.0, + 8338276.5, + 10490721.0, + 8991669.0 + ], + "bp_grad_norms_per_sample_med": [ + 2.3127104498144035e-07, + 3.35100103221464e-09, + 3.3095433060736923e-09, + 3.326423581029303e-09, + 3.2682878625678313e-09 + ], + "bp_grad_norms_F": [ + 1.3015700460528024e-05, + 1.1814677947086238e-07, + 1.1589928305966168e-07, + 1.1685516199122503e-07, + 1.1511789210771894e-07 + ], + "gamma_dfa": NaN, + "acc_eval": 0.2900390625, + "loss_eval": 1.9912320375442505, + "epoch": 18 + }, + { + "hidden_norms": [ + 13395.4189453125, + 2393670.0, + 8949124.0, + 11235038.0, + 9603954.0 + ], + "bp_grad_norms_per_sample_med": [ + 2.2520477216403378e-07, + 3.2867890631393948e-09, + 3.2246825210080488e-09, + 3.250592239822936e-09, + 3.201830800492189e-09 + ], + "bp_grad_norms_F": [ + 1.3049841072643176e-05, + 1.1301781910333375e-07, + 1.1090982354744483e-07, + 1.1187000836798688e-07, + 1.101881252907333e-07 + ], + "gamma_dfa": NaN, + "acc_eval": 0.2626953125, + "loss_eval": 2.0112197399139404, + "epoch": 19 + }, + { + "hidden_norms": [ + 14034.6982421875, + 2529761.5, + 9576616.0, + 12035853.0, + 10258897.0 + ], + "bp_grad_norms_per_sample_med": [ + 2.1533313088184514e-07, + 3.174721152632287e-09, + 3.1246687459685063e-09, + 3.149874805430386e-09, + 3.09314329705046e-09 + ], + "bp_grad_norms_F": [ + 1.26855793496361e-05, + 1.0948959072720754e-07, + 1.0680034279175743e-07, + 1.0768705749342189e-07, + 1.0610038714276016e-07 + ], + "gamma_dfa": NaN, + "acc_eval": 0.25, + "loss_eval": 2.0056447982788086, + "epoch": 20 + }, + { + "hidden_norms": [ + 14717.9462890625, + 2650771.25, + 10157696.0, + 12784758.0, + 10849881.0 + ], + "bp_grad_norms_per_sample_med": [ + 2.086567434389508e-07, + 2.999795301050767e-09, + 2.966421552841325e-09, + 2.9786708655166194e-09, + 2.947827537624903e-09 + ], + "bp_grad_norms_F": [ + 1.2321271242399234e-05, + 1.0535904237940485e-07, + 1.0280631101977633e-07, + 1.0365675251478024e-07, + 1.021143916091205e-07 + ], + "gamma_dfa": NaN, + "acc_eval": 0.294921875, + "loss_eval": 1.9869788885116577, + "epoch": 21 + }, + { + "hidden_norms": [ + 15327.8466796875, + 2777500.75, + 10776735.0, + 13573723.0, + 11474654.0 + ], + "bp_grad_norms_per_sample_med": [ + 2.0232576503076416e-07, + 2.9384910060770153e-09, + 2.867603710043909e-09, + 2.8935454032819052e-09, + 2.8441446975335793e-09 + ], + "bp_grad_norms_F": [ + 1.231426631420618e-05, + 1.0333050681765599e-07, + 1.0039466502576033e-07, + 1.0124550442469626e-07, + 9.972216474807283e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.2939453125, + "loss_eval": 1.9892809391021729, + "epoch": 22 + }, + { + "hidden_norms": [ + 15945.8837890625, + 2878910.75, + 11304857.0, + 14241028.0, + 11994238.0 + ], + "bp_grad_norms_per_sample_med": [ + 2.0171727044271393e-07, + 2.9479860774728195e-09, + 2.8580404709543927e-09, + 2.8716116151628057e-09, + 2.8270419338838337e-09 + ], + "bp_grad_norms_F": [ + 1.2025901924062055e-05, + 1.0096052704966496e-07, + 9.789676624905042e-08, + 9.872773176766714e-08, + 9.722123195388122e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.2802734375, + "loss_eval": 1.9858973026275635, + "epoch": 23 + }, + { + "hidden_norms": [ + 16595.853515625, + 3002464.5, + 11949546.0, + 15061732.0, + 12646630.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.9120048477816454e-07, + 2.8694004949869623e-09, + 2.772703400211185e-09, + 2.777924779095997e-09, + 2.7492945697815685e-09 + ], + "bp_grad_norms_F": [ + 1.2024025636492297e-05, + 9.894926478182242e-08, + 9.581360416177631e-08, + 9.663772715384766e-08, + 9.51538083882042e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.296875, + "loss_eval": 1.9969089031219482, + "epoch": 24 + }, + { + "hidden_norms": [ + 17227.1640625, + 3125553.25, + 12574712.0, + 15844785.0, + 13247150.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.8941717883080855e-07, + 2.8377846739147117e-09, + 2.7375035571708395e-09, + 2.7606503749666444e-09, + 2.7206534802814986e-09 + ], + "bp_grad_norms_F": [ + 1.153391167463269e-05, + 9.671435208247203e-08, + 9.34372863525823e-08, + 9.424089597587226e-08, + 9.280081059159784e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.294921875, + "loss_eval": 1.9766383171081543, + "epoch": 25 + }, + { + "hidden_norms": [ + 17808.20703125, + 3256055.25, + 13319775.0, + 16758460.0, + 13956453.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.8628608700055338e-07, + 2.7554525328099544e-09, + 2.625358597185823e-09, + 2.6348250248275917e-09, + 2.614828797931068e-09 + ], + "bp_grad_norms_F": [ + 1.0964491593767889e-05, + 9.419229485274627e-08, + 9.09606185928169e-08, + 9.175563775443152e-08, + 9.03681396380307e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.2919921875, + "loss_eval": 1.9645390510559082, + "epoch": 26 + }, + { + "hidden_norms": [ + 18401.28125, + 3393640.75, + 14088409.0, + 17726368.0, + 14707189.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.7694672749257734e-07, + 2.722393421805691e-09, + 2.634128470901942e-09, + 2.6627771099185793e-09, + 2.6215960513553682e-09 + ], + "bp_grad_norms_F": [ + 1.1054430615331512e-05, + 9.26544814205954e-08, + 8.924564554035896e-08, + 9.002578593708677e-08, + 8.865336553753878e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.2861328125, + "loss_eval": 1.974367618560791, + "epoch": 27 + }, + { + "hidden_norms": [ + 18936.51953125, + 3515726.25, + 14734875.0, + 18530786.0, + 15301454.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.7923295558830432e-07, + 2.6605673220103654e-09, + 2.530292642077825e-09, + 2.5396913461150916e-09, + 2.517351882502794e-09 + ], + "bp_grad_norms_F": [ + 1.1157896551594604e-05, + 9.233546904852119e-08, + 8.888883229474231e-08, + 8.968125086994405e-08, + 8.831575115664236e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.2841796875, + "loss_eval": 1.986099362373352, + "epoch": 28 + }, + { + "hidden_norms": [ + 19457.962890625, + 3664789.5, + 15342228.0, + 19308666.0, + 15894652.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.7488744674665213e-07, + 2.6636790551037848e-09, + 2.4992619085395518e-09, + 2.507181573463413e-09, + 2.486318262384657e-09 + ], + "bp_grad_norms_F": [ + 1.092042657546699e-05, + 9.157383118463258e-08, + 8.810303597783786e-08, + 8.888122238204232e-08, + 8.75455583582152e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.32421875, + "loss_eval": 1.9868172407150269, + "epoch": 29 + }, + { + "hidden_norms": [ + 19933.78515625, + 3775119.25, + 16017107.0, + 20154288.0, + 16541561.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.715164188453855e-07, + 2.5937807457410145e-09, + 2.4676627408126706e-09, + 2.4873916260048645e-09, + 2.4605657511500567e-09 + ], + "bp_grad_norms_F": [ + 1.0760413715615869e-05, + 9.029215419786851e-08, + 8.696353148707203e-08, + 8.772182269467521e-08, + 8.64171596504093e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.291015625, + "loss_eval": 1.9754526615142822, + "epoch": 30 + }, + { + "hidden_norms": [ + 20432.466796875, + 3910461.25, + 16712518.0, + 21021696.0, + 17196572.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.710603356741558e-07, + 2.6387678708772455e-09, + 2.524718212271182e-09, + 2.5378650292395832e-09, + 2.5052109275947032e-09 + ], + "bp_grad_norms_F": [ + 1.0587626093183644e-05, + 8.93876332952459e-08, + 8.59824282883892e-08, + 8.675493745613494e-08, + 8.546451368829366e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3076171875, + "loss_eval": 1.9657469987869263, + "epoch": 31 + }, + { + "hidden_norms": [ + 20875.439453125, + 4054334.0, + 17463474.0, + 21948092.0, + 17887228.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.729485461510194e-07, + 2.6252673368531987e-09, + 2.4720219204965588e-09, + 2.491059358789016e-09, + 2.4621127359125694e-09 + ], + "bp_grad_norms_F": [ + 1.0830392966454383e-05, + 8.912193294463577e-08, + 8.547249308321625e-08, + 8.626851411008829e-08, + 8.495522507701025e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.31640625, + "loss_eval": 1.9752991199493408, + "epoch": 32 + }, + { + "hidden_norms": [ + 21372.0234375, + 4181554.25, + 18179600.0, + 22866024.0, + 18562826.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.6644050049308134e-07, + 2.5652076018900516e-09, + 2.4415069965755265e-09, + 2.4562532008332028e-09, + 2.423905742787724e-09 + ], + "bp_grad_norms_F": [ + 1.0476498573552817e-05, + 8.780278903941507e-08, + 8.420673225373321e-08, + 8.501019266304866e-08, + 8.373319104748589e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.318359375, + "loss_eval": 1.9647243022918701, + "epoch": 33 + }, + { + "hidden_norms": [ + 21698.390625, + 4323451.5, + 18873394.0, + 23742746.0, + 19211736.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.6982900774564769e-07, + 2.6139097553112833e-09, + 2.4535891096633122e-09, + 2.465766257842006e-09, + 2.4349082750063644e-09 + ], + "bp_grad_norms_F": [ + 1.0547244528424926e-05, + 8.761546865798664e-08, + 8.380350635661671e-08, + 8.460440881208342e-08, + 8.332586531878405e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3115234375, + "loss_eval": 1.9688763618469238, + "epoch": 34 + }, + { + "hidden_norms": [ + 22000.6484375, + 4419910.5, + 19288578.0, + 24275390.0, + 19571482.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.6632812105399353e-07, + 2.5976403250638214e-09, + 2.4608304283191273e-09, + 2.4786184216196716e-09, + 2.4511619400868767e-09 + ], + "bp_grad_norms_F": [ + 1.0402384759800043e-05, + 8.75945431744185e-08, + 8.378519567031617e-08, + 8.458554390244899e-08, + 8.331597456390227e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3115234375, + "loss_eval": 1.962469458580017, + "epoch": 35 + }, + { + "hidden_norms": [ + 22332.140625, + 4505425.5, + 19876824.0, + 24999434.0, + 20098762.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.673159886195208e-07, + 2.5706452522200607e-09, + 2.4236175288905315e-09, + 2.4205843995872556e-09, + 2.4001298726261666e-09 + ], + "bp_grad_norms_F": [ + 1.0201652912655845e-05, + 8.668673245892933e-08, + 8.301810083821692e-08, + 8.384706973174616e-08, + 8.257533323785538e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3046875, + "loss_eval": 1.9482183456420898, + "epoch": 36 + }, + { + "hidden_norms": [ + 22704.90625, + 4601120.5, + 20416742.0, + 25650594.0, + 20558608.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.6260217705621471e-07, + 2.501829188261695e-09, + 2.37497355115579e-09, + 2.379827890308661e-09, + 2.364261231235787e-09 + ], + "bp_grad_norms_F": [ + 1.0307686352462042e-05, + 8.687224806180893e-08, + 8.317324073914278e-08, + 8.40014422465174e-08, + 8.273995177887628e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.306640625, + "loss_eval": 1.9718880653381348, + "epoch": 37 + }, + { + "hidden_norms": [ + 22953.462890625, + 4708118.5, + 21018216.0, + 26396430.0, + 21081344.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.6642815126033383e-07, + 2.5344910614677474e-09, + 2.3881991939589398e-09, + 2.393568676595237e-09, + 2.366601137282487e-09 + ], + "bp_grad_norms_F": [ + 1.0291260878148023e-05, + 8.597661604881068e-08, + 8.222435354809932e-08, + 8.30701907261755e-08, + 8.179645050176987e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3173828125, + "loss_eval": 1.9656648635864258, + "epoch": 38 + }, + { + "hidden_norms": [ + 23329.68359375, + 4813138.0, + 21547542.0, + 27050624.0, + 21536360.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.6755477361130033e-07, + 2.5133575221047977e-09, + 2.3339741250794077e-09, + 2.3630628565030065e-09, + 2.3331871989995534e-09 + ], + "bp_grad_norms_F": [ + 1.0207740160694811e-05, + 8.593145395252577e-08, + 8.217772773377874e-08, + 8.303662468733819e-08, + 8.176558452532845e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3076171875, + "loss_eval": 1.9581431150436401, + "epoch": 39 + }, + { + "hidden_norms": [ + 23598.361328125, + 4903029.0, + 22069792.0, + 27713646.0, + 21984984.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.6272531411232194e-07, + 2.4648321161890863e-09, + 2.360529327560812e-09, + 2.364856310776986e-09, + 2.3259458803437383e-09 + ], + "bp_grad_norms_F": [ + 1.0116659723280463e-05, + 8.537404028174933e-08, + 8.159113207284463e-08, + 8.246142613188567e-08, + 8.120937167177544e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3212890625, + "loss_eval": 1.9615185260772705, + "epoch": 40 + }, + { + "hidden_norms": [ + 23904.275390625, + 5004298.0, + 22611816.0, + 28369744.0, + 22444866.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.6563248550482967e-07, + 2.5165425299178423e-09, + 2.3475779098447447e-09, + 2.3561286255358027e-09, + 2.3397561665916555e-09 + ], + "bp_grad_norms_F": [ + 1.0093037417391315e-05, + 8.496893855181042e-08, + 8.129408968216012e-08, + 8.217033098389948e-08, + 8.091691228173659e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3056640625, + "loss_eval": 1.9596519470214844, + "epoch": 41 + }, + { + "hidden_norms": [ + 24124.95703125, + 5079679.0, + 23056108.0, + 28905466.0, + 22810520.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.650763863381144e-07, + 2.5150639348936465e-09, + 2.375127650111608e-09, + 2.3864086262648243e-09, + 2.3595001508169844e-09 + ], + "bp_grad_norms_F": [ + 1.0080430001835339e-05, + 8.473855928059493e-08, + 8.099364379177132e-08, + 8.18785963474511e-08, + 8.06336757364079e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3251953125, + "loss_eval": 1.9462497234344482, + "epoch": 42 + }, + { + "hidden_norms": [ + 24350.154296875, + 5170977.5, + 23526890.0, + 29466298.0, + 23191828.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.6226854882006592e-07, + 2.5123039204544284e-09, + 2.381558061870237e-09, + 2.391710607341224e-09, + 2.3734676446451886e-09 + ], + "bp_grad_norms_F": [ + 1.0192407899012323e-05, + 8.457134015316115e-08, + 8.096912296196024e-08, + 8.18817795789073e-08, + 8.06216249316094e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.318359375, + "loss_eval": 1.950229287147522, + "epoch": 43 + }, + { + "hidden_norms": [ + 24543.87890625, + 5230886.0, + 23931748.0, + 29937192.0, + 23499106.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.6538373870389478e-07, + 2.4975805867910594e-09, + 2.3630661871720804e-09, + 2.3725827968945623e-09, + 2.3566677498365607e-09 + ], + "bp_grad_norms_F": [ + 1.0082419976242818e-05, + 8.452050792584487e-08, + 8.1015215869229e-08, + 8.194035672204336e-08, + 8.06845363854336e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.30078125, + "loss_eval": 1.9453504085540771, + "epoch": 44 + }, + { + "hidden_norms": [ + 24711.091796875, + 5298309.0, + 24318082.0, + 30413608.0, + 23808788.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.6054062257353507e-07, + 2.4891384509118097e-09, + 2.358970574434238e-09, + 2.368707230360201e-09, + 2.3476884880579973e-09 + ], + "bp_grad_norms_F": [ + 1.0053097867057659e-05, + 8.462841805112475e-08, + 8.112137095395155e-08, + 8.206667700960679e-08, + 8.080490943029872e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3076171875, + "loss_eval": 1.9472556114196777, + "epoch": 45 + }, + { + "hidden_norms": [ + 24881.759765625, + 5378830.0, + 24836314.0, + 31031698.0, + 24231208.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.623594982902432e-07, + 2.4946655852176036e-09, + 2.3718913499948258e-09, + 2.3750907907071905e-09, + 2.3478305966051494e-09 + ], + "bp_grad_norms_F": [ + 1.012617394735571e-05, + 8.473565316080567e-08, + 8.116323613194254e-08, + 8.212452939915238e-08, + 8.085635272436775e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3125, + "loss_eval": 1.9553910493850708, + "epoch": 46 + }, + { + "hidden_norms": [ + 25032.86328125, + 5446721.0, + 25258438.0, + 31534450.0, + 24555236.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.632960788811033e-07, + 2.470913029739563e-09, + 2.316862923734675e-09, + 2.3169177687520914e-09, + 2.3029997908707855e-09 + ], + "bp_grad_norms_F": [ + 9.900480108626653e-06, + 8.452943944803337e-08, + 8.106712812150363e-08, + 8.2038923210348e-08, + 8.077390845073751e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3115234375, + "loss_eval": 1.9489891529083252, + "epoch": 47 + }, + { + "hidden_norms": [ + 25306.201171875, + 5468743.5, + 25525060.0, + 31867246.0, + 24722362.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5847486167785974e-07, + 2.4746471538605874e-09, + 2.3425053008452323e-09, + 2.3598758502885175e-09, + 2.3311046426499615e-09 + ], + "bp_grad_norms_F": [ + 9.7980091595673e-06, + 8.43976621922593e-08, + 8.096849057892541e-08, + 8.195242884312393e-08, + 8.069165602364592e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3134765625, + "loss_eval": 1.9313395023345947, + "epoch": 48 + }, + { + "hidden_norms": [ + 25419.810546875, + 5521521.5, + 25869234.0, + 32269302.0, + 24978000.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.6377224199004559e-07, + 2.4982107493798367e-09, + 2.3510355884326373e-09, + 2.359785478134313e-09, + 2.3393140757832498e-09 + ], + "bp_grad_norms_F": [ + 1.003210672934074e-05, + 8.464159151344575e-08, + 8.12287339613249e-08, + 8.224073155815859e-08, + 8.095251047279817e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3095703125, + "loss_eval": 1.94147527217865, + "epoch": 49 + }, + { + "hidden_norms": [ + 25543.001953125, + 5573999.0, + 26298990.0, + 32786014.0, + 25316304.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.593823384382631e-07, + 2.512666297249666e-09, + 2.392667397543846e-09, + 2.4054993552624637e-09, + 2.386617126148849e-09 + ], + "bp_grad_norms_F": [ + 1.0020607987826224e-05, + 8.429628906014841e-08, + 8.090298564411569e-08, + 8.193208600459911e-08, + 8.064517231787249e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.32421875, + "loss_eval": 1.9412651062011719, + "epoch": 50 + }, + { + "hidden_norms": [ + 25755.05859375, + 5633394.5, + 26712786.0, + 33270568.0, + 25632750.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.615333218296655e-07, + 2.483285133081381e-09, + 2.346244309947565e-09, + 2.3569415308344333e-09, + 2.342600557980745e-09 + ], + "bp_grad_norms_F": [ + 9.841775863606017e-06, + 8.406106388747503e-08, + 8.07588094176026e-08, + 8.180128219237304e-08, + 8.051125632846379e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3193359375, + "loss_eval": 1.9356749057769775, + "epoch": 51 + }, + { + "hidden_norms": [ + 25838.6875, + 5676516.5, + 26956640.0, + 33554564.0, + 25797852.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5953406773405732e-07, + 2.4927520048123597e-09, + 2.379676899977312e-09, + 2.3930151193951588e-09, + 2.37515118683973e-09 + ], + "bp_grad_norms_F": [ + 1.0000761903938837e-05, + 8.431718612200712e-08, + 8.108807492135384e-08, + 8.215462798943918e-08, + 8.085050495765245e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.322265625, + "loss_eval": 1.9513205289840698, + "epoch": 52 + }, + { + "hidden_norms": [ + 25925.244140625, + 5739893.0, + 27348932.0, + 34049508.0, + 26115016.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5941535025376652e-07, + 2.4419115618457e-09, + 2.335727167235291e-09, + 2.3481498967470316e-09, + 2.335213800108704e-09 + ], + "bp_grad_norms_F": [ + 9.962137482943945e-06, + 8.399173623274692e-08, + 8.08289613019042e-08, + 8.190350797576684e-08, + 8.060307266077871e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3173828125, + "loss_eval": 1.9443122148513794, + "epoch": 53 + }, + { + "hidden_norms": [ + 26078.8125, + 5808761.5, + 27799278.0, + 34615776.0, + 26499736.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5867621527831943e-07, + 2.4137036813698387e-09, + 2.31744068379669e-09, + 2.327602111051874e-09, + 2.313133462550354e-09 + ], + "bp_grad_norms_F": [ + 9.748204320203513e-06, + 8.347069524461403e-08, + 8.032785814293675e-08, + 8.14167648854891e-08, + 8.011366503524187e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.314453125, + "loss_eval": 1.9361423254013062, + "epoch": 54 + }, + { + "hidden_norms": [ + 26192.453125, + 5875144.0, + 28166196.0, + 35073324.0, + 26791154.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5812705100870517e-07, + 2.4364927853071094e-09, + 2.339390015038134e-09, + 2.3421220518571317e-09, + 2.328011117214146e-09 + ], + "bp_grad_norms_F": [ + 9.899279575620312e-06, + 8.359175751593284e-08, + 8.043829780035594e-08, + 8.154502495472116e-08, + 8.023434361348336e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.306640625, + "loss_eval": 1.940393090248108, + "epoch": 55 + }, + { + "hidden_norms": [ + 26281.6796875, + 5935011.0, + 28625862.0, + 35638648.0, + 27180628.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.6062698193763936e-07, + 2.4364579243041362e-09, + 2.3309267849214166e-09, + 2.3373505353418977e-09, + 2.3235424695400297e-09 + ], + "bp_grad_norms_F": [ + 9.917155693983659e-06, + 8.292303022017222e-08, + 7.976373694873473e-08, + 8.087503999831824e-08, + 7.955991776498195e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.318359375, + "loss_eval": 1.945765495300293, + "epoch": 56 + }, + { + "hidden_norms": [ + 26394.865234375, + 5988690.5, + 28944822.0, + 36062032.0, + 27426714.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.6303006589168945e-07, + 2.4247555074907723e-09, + 2.309517688203755e-09, + 2.3198174492478074e-09, + 2.3075219512946887e-09 + ], + "bp_grad_norms_F": [ + 1.0007075616158545e-05, + 8.301859821813196e-08, + 7.986342609456187e-08, + 8.099387827087412e-08, + 7.966451676111319e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3115234375, + "loss_eval": 1.9433667659759521, + "epoch": 57 + }, + { + "hidden_norms": [ + 26515.357421875, + 6051740.5, + 29293128.0, + 36499980.0, + 27690768.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5923862406452827e-07, + 2.3941215676615e-09, + 2.28924412759568e-09, + 2.2969097734915067e-09, + 2.280708066848547e-09 + ], + "bp_grad_norms_F": [ + 9.82554138317937e-06, + 8.282027863515395e-08, + 7.976368721074323e-08, + 8.090249536962801e-08, + 7.957222436516531e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.306640625, + "loss_eval": 1.944448709487915, + "epoch": 58 + }, + { + "hidden_norms": [ + 26621.947265625, + 6089471.5, + 29639332.0, + 36907412.0, + 27955134.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5679145803915162e-07, + 2.3988511177464034e-09, + 2.303158996852517e-09, + 2.3152970651807436e-09, + 2.2871571303539895e-09 + ], + "bp_grad_norms_F": [ + 9.731273166835308e-06, + 8.24191062065438e-08, + 7.936159107657659e-08, + 8.050036370832458e-08, + 7.917343225471996e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3125, + "loss_eval": 1.9350972175598145, + "epoch": 59 + }, + { + "hidden_norms": [ + 26731.64453125, + 6146909.5, + 30036422.0, + 37405476.0, + 28295768.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5779031059537374e-07, + 2.4025526013105036e-09, + 2.3019863792939077e-09, + 2.3154580475193143e-09, + 2.293559120403188e-09 + ], + "bp_grad_norms_F": [ + 9.80158438323997e-06, + 8.19560526110763e-08, + 7.889645559089331e-08, + 8.003749485396838e-08, + 7.871490481647925e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.328125, + "loss_eval": 1.941221833229065, + "epoch": 60 + }, + { + "hidden_norms": [ + 26842.380859375, + 6189510.0, + 30305302.0, + 37746040.0, + 28515236.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5621530735643319e-07, + 2.3939163984465495e-09, + 2.3025825690581314e-09, + 2.3148469807665606e-09, + 2.2962292067774115e-09 + ], + "bp_grad_norms_F": [ + 9.731598765938543e-06, + 8.167720011442725e-08, + 7.863504691840717e-08, + 7.977818938798009e-08, + 7.845925864558012e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3408203125, + "loss_eval": 1.9333854913711548, + "epoch": 61 + }, + { + "hidden_norms": [ + 26940.765625, + 6243737.5, + 30559802.0, + 38046748.0, + 28709922.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5672847553105385e-07, + 2.3640371882294176e-09, + 2.2657933307357325e-09, + 2.2837092217287136e-09, + 2.2615944672566e-09 + ], + "bp_grad_norms_F": [ + 9.668071470514406e-06, + 8.165712017671467e-08, + 7.869160612017367e-08, + 7.984421301898692e-08, + 7.851971872696595e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3203125, + "loss_eval": 1.9349033832550049, + "epoch": 62 + }, + { + "hidden_norms": [ + 27024.9140625, + 6285780.0, + 30783454.0, + 38328276.0, + 28883832.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5834399391678744e-07, + 2.3645019275875256e-09, + 2.2709001346044033e-09, + 2.2806179167389473e-09, + 2.2639952135250496e-09 + ], + "bp_grad_norms_F": [ + 9.799305189517327e-06, + 8.167254605950802e-08, + 7.873082097376027e-08, + 7.989201833424886e-08, + 7.856029071717785e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.31640625, + "loss_eval": 1.9409070014953613, + "epoch": 63 + }, + { + "hidden_norms": [ + 27131.298828125, + 6330564.0, + 31012118.0, + 38625700.0, + 29050708.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.588112752415327e-07, + 2.403508059245496e-09, + 2.2917150399592856e-09, + 2.3033945861783423e-09, + 2.2864310444958846e-09 + ], + "bp_grad_norms_F": [ + 9.54241204453865e-06, + 8.120213124129805e-08, + 7.823086889402475e-08, + 7.938096047155341e-08, + 7.805995494436502e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3388671875, + "loss_eval": 1.9211525917053223, + "epoch": 64 + }, + { + "hidden_norms": [ + 27208.25, + 6368866.0, + 31290472.0, + 38967492.0, + 29272872.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5740459957669373e-07, + 2.3837900542389434e-09, + 2.284773703564724e-09, + 2.298153001234482e-09, + 2.2881960770604337e-09 + ], + "bp_grad_norms_F": [ + 9.658318958827294e-06, + 8.104061777203242e-08, + 7.810449886846982e-08, + 7.925908818151584e-08, + 7.793548206791456e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.341796875, + "loss_eval": 1.9278589487075806, + "epoch": 65 + }, + { + "hidden_norms": [ + 27276.62890625, + 6418036.5, + 31545874.0, + 39283716.0, + 29477138.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5503043471198907e-07, + 2.357751327508595e-09, + 2.2688111389612686e-09, + 2.2810615618595875e-09, + 2.2629353946257424e-09 + ], + "bp_grad_norms_F": [ + 9.6196936283377e-06, + 8.082280089638516e-08, + 7.795324563630857e-08, + 7.91092062968346e-08, + 7.778949395742529e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.33984375, + "loss_eval": 1.9305698871612549, + "epoch": 66 + }, + { + "hidden_norms": [ + 27340.607421875, + 6465153.0, + 31796498.0, + 39598816.0, + 29679376.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5558319432784629e-07, + 2.3627866330144798e-09, + 2.2779722552712656e-09, + 2.2912376440586968e-09, + 2.2693087409209056e-09 + ], + "bp_grad_norms_F": [ + 9.68782387644751e-06, + 8.066198375900058e-08, + 7.781451927257876e-08, + 7.897661191691441e-08, + 7.765357423750174e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3173828125, + "loss_eval": 1.9328408241271973, + "epoch": 67 + }, + { + "hidden_norms": [ + 27396.21484375, + 6507270.5, + 32041844.0, + 39890444.0, + 29881876.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5313091239477217e-07, + 2.336748128328736e-09, + 2.2449873071650472e-09, + 2.2587733905510277e-09, + 2.237245722014336e-09 + ], + "bp_grad_norms_F": [ + 9.509177289146464e-06, + 8.020302288969106e-08, + 7.739499352510393e-08, + 7.854830386122558e-08, + 7.723401296289012e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.318359375, + "loss_eval": 1.9248214960098267, + "epoch": 68 + }, + { + "hidden_norms": [ + 27461.697265625, + 6534744.5, + 32295470.0, + 40216712.0, + 30089670.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5820302223801264e-07, + 2.3342023869332706e-09, + 2.2277899525136036e-09, + 2.237228180490547e-09, + 2.2202666372095337e-09 + ], + "bp_grad_norms_F": [ + 9.662662705522962e-06, + 8.017656227821135e-08, + 7.73198536307973e-08, + 7.847864225141166e-08, + 7.715301109101347e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.32421875, + "loss_eval": 1.9348523616790771, + "epoch": 69 + }, + { + "hidden_norms": [ + 27519.265625, + 6576646.0, + 32533798.0, + 40497036.0, + 30273994.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5546247311704064e-07, + 2.332068760324546e-09, + 2.2422437240265936e-09, + 2.259832765361125e-09, + 2.2382042885737974e-09 + ], + "bp_grad_norms_F": [ + 9.630303793528583e-06, + 7.993662620719988e-08, + 7.707911464649442e-08, + 7.823470582479786e-08, + 7.691809855714382e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3212890625, + "loss_eval": 1.9298650026321411, + "epoch": 70 + }, + { + "hidden_norms": [ + 27578.833984375, + 6611492.0, + 32783444.0, + 40806224.0, + 30475316.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5486573090583988e-07, + 2.310989177800593e-09, + 2.214788796806033e-09, + 2.2242498953772838e-09, + 2.209040284029129e-09 + ], + "bp_grad_norms_F": [ + 9.58504097070545e-06, + 7.973878268785484e-08, + 7.692549530702308e-08, + 7.808419866250915e-08, + 7.67675771840004e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.314453125, + "loss_eval": 1.9349405765533447, + "epoch": 71 + }, + { + "hidden_norms": [ + 27637.01953125, + 6652277.0, + 32940606.0, + 41021348.0, + 30603354.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5176895828972192e-07, + 2.309274771405967e-09, + 2.2196366966653613e-09, + 2.2291524182094236e-09, + 2.212444005778025e-09 + ], + "bp_grad_norms_F": [ + 9.506617061560974e-06, + 7.949277147645262e-08, + 7.665874335316403e-08, + 7.781166289078101e-08, + 7.6499844681166e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.32421875, + "loss_eval": 1.9272165298461914, + "epoch": 72 + }, + { + "hidden_norms": [ + 27665.341796875, + 6675607.5, + 33128100.0, + 41256448.0, + 30754874.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.570084151580886e-07, + 2.2943660304974856e-09, + 2.2026656054663363e-09, + 2.2121668941110784e-09, + 2.1966797270067673e-09 + ], + "bp_grad_norms_F": [ + 9.71559165918734e-06, + 7.9600532387758e-08, + 7.679192037812754e-08, + 7.795681966626944e-08, + 7.663498990950757e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3115234375, + "loss_eval": 1.941681146621704, + "epoch": 73 + }, + { + "hidden_norms": [ + 27715.794921875, + 6709802.5, + 33331388.0, + 41504024.0, + 30915288.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5387240637210198e-07, + 2.3100528157016242e-09, + 2.2238448860179005e-09, + 2.2321375858780357e-09, + 2.2150188350167355e-09 + ], + "bp_grad_norms_F": [ + 9.473335921938997e-06, + 7.898351128687864e-08, + 7.616839781121598e-08, + 7.73214452465254e-08, + 7.601196472251104e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.33203125, + "loss_eval": 1.9246938228607178, + "epoch": 74 + }, + { + "hidden_norms": [ + 27766.177734375, + 6726091.5, + 33455028.0, + 41663816.0, + 31007976.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5595279023727926e-07, + 2.298697898694968e-09, + 2.2091579676697393e-09, + 2.215150729512061e-09, + 2.1983310727335947e-09 + ], + "bp_grad_norms_F": [ + 9.587340173311532e-06, + 7.91313112813441e-08, + 7.627861009495973e-08, + 7.743443575236597e-08, + 7.611913588334573e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3232421875, + "loss_eval": 1.931347370147705, + "epoch": 75 + }, + { + "hidden_norms": [ + 27818.396484375, + 6749941.5, + 33562528.0, + 41806788.0, + 31089946.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.542799026310604e-07, + 2.3069177679246877e-09, + 2.2149184708553094e-09, + 2.2274431188407107e-09, + 2.2138819666395193e-09 + ], + "bp_grad_norms_F": [ + 9.57066185947042e-06, + 7.899650000808833e-08, + 7.612889874053508e-08, + 7.72842909668725e-08, + 7.59681810791335e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.328125, + "loss_eval": 1.9290614128112793, + "epoch": 76 + }, + { + "hidden_norms": [ + 27874.181640625, + 6776020.5, + 33686440.0, + 41956108.0, + 31183962.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5547591658560123e-07, + 2.2904955709890373e-09, + 2.199639581590418e-09, + 2.2111712461025945e-09, + 2.1966135577144996e-09 + ], + "bp_grad_norms_F": [ + 9.538461199554149e-06, + 7.882318442398173e-08, + 7.600280582664709e-08, + 7.715882333059199e-08, + 7.58423581714851e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3369140625, + "loss_eval": 1.9268429279327393, + "epoch": 77 + }, + { + "hidden_norms": [ + 27928.552734375, + 6790532.0, + 33797660.0, + 42095308.0, + 31270040.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5469180425498053e-07, + 2.3010513494625684e-09, + 2.2103030516973377e-09, + 2.2237711672090654e-09, + 2.206205884647261e-09 + ], + "bp_grad_norms_F": [ + 9.510661584499758e-06, + 7.866238149745186e-08, + 7.584375794067455e-08, + 7.699873805222524e-08, + 7.568271342961452e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3291015625, + "loss_eval": 1.9229360818862915, + "epoch": 78 + }, + { + "hidden_norms": [ + 27959.18359375, + 6814900.0, + 33951204.0, + 42283796.0, + 31402056.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5280090792657575e-07, + 2.2722823622700616e-09, + 2.182985348042621e-09, + 2.195454484876791e-09, + 2.1783830295163398e-09 + ], + "bp_grad_norms_F": [ + 9.445489922654815e-06, + 7.852113270701011e-08, + 7.573528648663341e-08, + 7.688608150147047e-08, + 7.557830628002193e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3232421875, + "loss_eval": 1.9262363910675049, + "epoch": 79 + }, + { + "hidden_norms": [ + 27969.6171875, + 6828402.0, + 34031784.0, + 42376960.0, + 31453510.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5404552300424257e-07, + 2.2810822120078456e-09, + 2.1932544669311937e-09, + 2.216228756068972e-09, + 2.193953241302893e-09 + ], + "bp_grad_norms_F": [ + 9.498375220573507e-06, + 7.844284510838406e-08, + 7.562837112118359e-08, + 7.677955693452532e-08, + 7.546795188773103e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.330078125, + "loss_eval": 1.925222396850586, + "epoch": 80 + }, + { + "hidden_norms": [ + 27976.40625, + 6839237.5, + 34096948.0, + 42451852.0, + 31495834.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5588855717396655e-07, + 2.290766465407046e-09, + 2.2075170580393433e-09, + 2.2167860880273338e-09, + 2.1972801356184846e-09 + ], + "bp_grad_norms_F": [ + 9.528901500743814e-06, + 7.844620597552421e-08, + 7.561048676052451e-08, + 7.67637402532273e-08, + 7.544804248027503e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3427734375, + "loss_eval": 1.9254777431488037, + "epoch": 81 + }, + { + "hidden_norms": [ + 28004.76953125, + 6848695.0, + 34154160.0, + 42523512.0, + 31536314.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.562642069075082e-07, + 2.289735068217169e-09, + 2.2051671599854217e-09, + 2.2135104860154797e-09, + 2.188419445658951e-09 + ], + "bp_grad_norms_F": [ + 9.512452379567549e-06, + 7.840634452804807e-08, + 7.557503778343744e-08, + 7.672917234913257e-08, + 7.541159163793054e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.341796875, + "loss_eval": 1.9253835678100586, + "epoch": 82 + }, + { + "hidden_norms": [ + 28036.138671875, + 6859844.5, + 34231564.0, + 42620400.0, + 31595422.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5447380974364933e-07, + 2.2824857559555767e-09, + 2.1902821778496673e-09, + 2.2037338620606306e-09, + 2.183933034416441e-09 + ], + "bp_grad_norms_F": [ + 9.456872248847503e-06, + 7.838458770947909e-08, + 7.556796077778927e-08, + 7.67221948194674e-08, + 7.540831603591869e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.333984375, + "loss_eval": 1.9241688251495361, + "epoch": 83 + }, + { + "hidden_norms": [ + 28063.341796875, + 6866642.0, + 34298612.0, + 42705280.0, + 31647722.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.550155843688117e-07, + 2.281024480410565e-09, + 2.1895063540000592e-09, + 2.20287676988562e-09, + 2.1853003850935693e-09 + ], + "bp_grad_norms_F": [ + 9.501630302111153e-06, + 7.833479287455702e-08, + 7.55063354063168e-08, + 7.666048418286664e-08, + 7.534579538059916e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.333984375, + "loss_eval": 1.9253199100494385, + "epoch": 84 + }, + { + "hidden_norms": [ + 28090.240234375, + 6875769.0, + 34361584.0, + 42782612.0, + 31697334.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.536338629648526e-07, + 2.267983800763318e-09, + 2.1772310621059887e-09, + 2.1934687399749464e-09, + 2.1765453883659802e-09 + ], + "bp_grad_norms_F": [ + 9.536663128528744e-06, + 7.836315774056857e-08, + 7.554304914947352e-08, + 7.669789425790441e-08, + 7.538390178751797e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3203125, + "loss_eval": 1.9292113780975342, + "epoch": 85 + }, + { + "hidden_norms": [ + 28114.236328125, + 6884728.0, + 34407252.0, + 42840440.0, + 31732750.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5464524949493352e-07, + 2.280014399502761e-09, + 2.1917956338768363e-09, + 2.2101085406234233e-09, + 2.1894293045221502e-09 + ], + "bp_grad_norms_F": [ + 9.505700290901586e-06, + 7.817885716576711e-08, + 7.534623591709533e-08, + 7.650118760693658e-08, + 7.518445244159011e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3310546875, + "loss_eval": 1.9261689186096191, + "epoch": 86 + }, + { + "hidden_norms": [ + 28132.669921875, + 6893751.5, + 34460884.0, + 42910656.0, + 31780534.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5502969574754388e-07, + 2.2780193287275097e-09, + 2.1857382570544814e-09, + 2.200839510635433e-09, + 2.1766897173591815e-09 + ], + "bp_grad_norms_F": [ + 9.460140972805675e-06, + 7.810618996018093e-08, + 7.528173284754303e-08, + 7.643409105639876e-08, + 7.512058886049999e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.333984375, + "loss_eval": 1.9239721298217773, + "epoch": 87 + }, + { + "hidden_norms": [ + 28148.599609375, + 6899411.0, + 34495884.0, + 42954048.0, + 31804918.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5502014605317527e-07, + 2.2734785165567928e-09, + 2.191139492069283e-09, + 2.201265170143074e-09, + 2.182727776300908e-09 + ], + "bp_grad_norms_F": [ + 9.466436495131347e-06, + 7.804617752071863e-08, + 7.521869349602639e-08, + 7.63711227591557e-08, + 7.50560502638109e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3369140625, + "loss_eval": 1.9230663776397705, + "epoch": 88 + }, + { + "hidden_norms": [ + 28162.9609375, + 6901370.5, + 34523684.0, + 42989220.0, + 31824814.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5575578515836241e-07, + 2.2783777087198587e-09, + 2.1914692283075965e-09, + 2.19572404702717e-09, + 2.1778470138400507e-09 + ], + "bp_grad_norms_F": [ + 9.514466910331976e-06, + 7.809557445170867e-08, + 7.525108713934969e-08, + 7.640503696393353e-08, + 7.508738519845792e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3447265625, + "loss_eval": 1.9263484477996826, + "epoch": 89 + }, + { + "hidden_norms": [ + 28175.54296875, + 6907143.0, + 34559816.0, + 43035484.0, + 31854504.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5354446247783926e-07, + 2.272713128803616e-09, + 2.1856987331148048e-09, + 2.2020374412790034e-09, + 2.1765065305601183e-09 + ], + "bp_grad_norms_F": [ + 9.444916940992698e-06, + 7.799643242378806e-08, + 7.51774322793608e-08, + 7.632959864167788e-08, + 7.50167856722328e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3349609375, + "loss_eval": 1.9226629734039307, + "epoch": 90 + }, + { + "hidden_norms": [ + 28185.98046875, + 6908239.0, + 34575728.0, + 43054472.0, + 31863434.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5382977380795637e-07, + 2.274502586274707e-09, + 2.1870079081054428e-09, + 2.1993444843104726e-09, + 2.181675506918168e-09 + ], + "bp_grad_norms_F": [ + 9.480510016146582e-06, + 7.80089948193563e-08, + 7.51953308508746e-08, + 7.635033938413471e-08, + 7.503523846708049e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3359375, + "loss_eval": 1.92463219165802, + "epoch": 91 + }, + { + "hidden_norms": [ + 28195.05078125, + 6911853.0, + 34606972.0, + 43094588.0, + 31890320.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5355996652033355e-07, + 2.2723862791451666e-09, + 2.1887525125663387e-09, + 2.1977624165003817e-09, + 2.1820119044946296e-09 + ], + "bp_grad_norms_F": [ + 9.450015568290837e-06, + 7.795882339678428e-08, + 7.514577760048269e-08, + 7.629873266523646e-08, + 7.498609022604796e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3388671875, + "loss_eval": 1.9231452941894531, + "epoch": 92 + }, + { + "hidden_norms": [ + 28202.009765625, + 6911585.0, + 34614424.0, + 43102544.0, + 31893396.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5335498915192147e-07, + 2.273481403136657e-09, + 2.18971196730422e-09, + 2.201623994224633e-09, + 2.179575853133997e-09 + ], + "bp_grad_norms_F": [ + 9.466421943216119e-06, + 7.796668910486915e-08, + 7.514941557928978e-08, + 7.630385567836129e-08, + 7.498929477378624e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3349609375, + "loss_eval": 1.9242045879364014, + "epoch": 93 + }, + { + "hidden_norms": [ + 28207.30078125, + 6912594.5, + 34624228.0, + 43114124.0, + 31900214.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.541789487191636e-07, + 2.26694063520938e-09, + 2.1774959613196643e-09, + 2.1911708003585773e-09, + 2.177005020698175e-09 + ], + "bp_grad_norms_F": [ + 9.48972228798084e-06, + 7.800251466960617e-08, + 7.518539746342867e-08, + 7.634076126805667e-08, + 7.502466559117238e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3349609375, + "loss_eval": 1.9264299869537354, + "epoch": 94 + }, + { + "hidden_norms": [ + 28210.744140625, + 6914176.5, + 34633372.0, + 43125584.0, + 31907212.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5396986441373883e-07, + 2.2685893164009485e-09, + 2.182473091139059e-09, + 2.2001969135487798e-09, + 2.1797674776280473e-09 + ], + "bp_grad_norms_F": [ + 9.470595614402555e-06, + 7.796354850597709e-08, + 7.514324096291602e-08, + 7.629726184177343e-08, + 7.498252330151445e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.3388671875, + "loss_eval": 1.9243674278259277, + "epoch": 95 + }, + { + "hidden_norms": [ + 28213.134765625, + 6914654.0, + 34637116.0, + 43129968.0, + 31909774.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5392139118830528e-07, + 2.2676351907335857e-09, + 2.181399949563456e-09, + 2.194956438827944e-09, + 2.1772521563434566e-09 + ], + "bp_grad_norms_F": [ + 9.468548341828864e-06, + 7.796590040243245e-08, + 7.514832134347671e-08, + 7.630251985801806e-08, + 7.498784526660529e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.33984375, + "loss_eval": 1.9245264530181885, + "epoch": 96 + }, + { + "hidden_norms": [ + 28214.392578125, + 6914571.0, + 34637992.0, + 43130936.0, + 31909692.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5425968058480066e-07, + 2.2704826907471443e-09, + 2.18207008018112e-09, + 2.1958874718563948e-09, + 2.177970914729599e-09 + ], + "bp_grad_norms_F": [ + 9.473776117374655e-06, + 7.796563039619286e-08, + 7.514674393860332e-08, + 7.63012906190852e-08, + 7.498606890976589e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.337890625, + "loss_eval": 1.9245213270187378, + "epoch": 97 + }, + { + "hidden_norms": [ + 28215.587890625, + 6915082.0, + 34641648.0, + 43135608.0, + 31912816.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5412730647312856e-07, + 2.2691615253478403e-09, + 2.181160807523952e-09, + 2.1947776929209795e-09, + 2.1773913783107446e-09 + ], + "bp_grad_norms_F": [ + 9.472671081312e-06, + 7.796593592956924e-08, + 7.514773869843339e-08, + 7.630221432464168e-08, + 7.498719867271575e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.337890625, + "loss_eval": 1.9246418476104736, + "epoch": 98 + }, + { + "hidden_norms": [ + 28216.130859375, + 6915161.5, + 34642452.0, + 43136600.0, + 31913348.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5420548038491688e-07, + 2.2695016976825855e-09, + 2.1805732774993203e-09, + 2.1951678252918327e-09, + 2.1776345171531375e-09 + ], + "bp_grad_norms_F": [ + 9.474650141783059e-06, + 7.796723622277568e-08, + 7.514872635283609e-08, + 7.63033014550274e-08, + 7.498815079998167e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.337890625, + "loss_eval": 1.9247386455535889, + "epoch": 99 + }, + { + "hidden_norms": [ + 28216.1875, + 6915194.0, + 34642580.0, + 43136768.0, + 31913448.0 + ], + "bp_grad_norms_per_sample_med": [ + 1.5419134058447526e-07, + 2.2693393830763853e-09, + 2.180579494748258e-09, + 2.194876502770171e-09, + 2.1777446512771803e-09 + ], + "bp_grad_norms_F": [ + 9.474156286159996e-06, + 7.796705148166438e-08, + 7.51486410877078e-08, + 7.630320908447175e-08, + 7.498806553485338e-08 + ], + "gamma_dfa": NaN, + "acc_eval": 0.337890625, + "loss_eval": 1.9247117042541504, + "epoch": 100 + } + ], + "seed": 42, + "depth": 4, + "d_hidden": 256 +}
\ No newline at end of file |
