Update best_model_acc69.87_metadata.json - Run 20251012_141246
Browse files
weights/David-fully_shared-weighted_sum/20251012_141246/best_model_acc69.87_metadata.json
ADDED
|
@@ -0,0 +1,251 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 2,
|
| 3 |
+
"optimizer_state_dict": {
|
| 4 |
+
"state": {
|
| 5 |
+
"0": {
|
| 6 |
+
"step": "tensor(3756.)",
|
| 7 |
+
"exp_avg": "tensor([[-5.1282e-04, -1.8069e-03, 3.8467e-03, ..., -4.5071e-04,\n -1.5242e-03, -6.7662e-04],\n [-4.9425e-04, -3.8872e-03, 8.1979e-04, ..., -1.5170e-04,\n 4.2225e-04, 1.5177e-04],\n [ 4.4098e-04, -1.3676e-04, 2.0125e-04, ..., -5.6593e-04,\n -4.3852e-04, -9.6662e-05],\n ...,\n [ 6.8255e-04, -2.4197e-03, 3.6309e-04, ..., -3.6951e-04,\n -7.6403e-04, -5.5311e-04],\n [ 3.4969e-04, -2.0538e-03, -2.5826e-03, ..., -1.5261e-04,\n 2.2250e-05, -3.8426e-04],\n [ 5.1188e-04, 1.7978e-03, 5.6984e-04, ..., -7.2885e-04,\n 2.3756e-04, 1.2626e-03]], device='cuda:0')",
|
| 8 |
+
"exp_avg_sq": "tensor([[4.3677e-06, 2.8684e-05, 1.2545e-05, ..., 2.7990e-06, 2.0325e-06,\n 2.4483e-06],\n [7.8124e-06, 4.7808e-05, 1.6527e-05, ..., 4.8475e-06, 3.7860e-06,\n 3.7397e-06],\n [4.2822e-06, 1.9946e-05, 1.0627e-05, ..., 2.9982e-06, 2.0986e-06,\n 2.7400e-06],\n ...,\n [1.0023e-05, 4.6593e-05, 2.5447e-05, ..., 8.6805e-06, 3.5114e-06,\n 9.9850e-06],\n [3.4912e-06, 2.6399e-05, 1.6046e-05, ..., 2.9484e-06, 1.8660e-06,\n 2.2302e-06],\n [1.0286e-05, 5.3918e-05, 2.3976e-05, ..., 5.4892e-06, 3.7292e-06,\n 3.9925e-06]], device='cuda:0')"
|
| 9 |
+
},
|
| 10 |
+
"1": {
|
| 11 |
+
"step": "tensor(3756.)",
|
| 12 |
+
"exp_avg": "tensor([-1.4255e-02, -1.7429e-02, -1.4369e-03, -1.5860e-02, 2.4850e-02,\n -3.1161e-02, 2.8110e-02, -5.1534e-02, 2.9153e-02, -2.2904e-03,\n 2.3722e-02, 6.5154e-03, 1.2955e-02, 2.3281e-02, -2.3921e-02,\n -1.7445e-02, 1.2257e-02, -2.5723e-02, 2.6587e-02, -5.1181e-02,\n 1.4078e-02, -2.3552e-03, 1.4307e-02, -7.9446e-03, -1.3228e-02,\n 1.5358e-02, 4.1472e-03, -1.3335e-02, -2.3042e-02, -1.5220e-02,\n 1.4166e-02, -1.0739e-02, 1.1633e-02, -2.7870e-03, -1.0891e-02,\n 1.1091e-02, -1.8680e-02, 4.7923e-03, -2.3064e-02, 1.6778e-02,\n 9.4276e-03, -2.1996e-02, -1.2473e-02, 2.8413e-02, 2.1145e-02,\n 1.8040e-02, 1.8892e-02, 3.5657e-02, 2.3672e-02, 1.3978e-02,\n 1.0297e-02, -2.8424e-02, -7.3923e-03, -1.3931e-02, -1.6343e-03,\n 9.6438e-03, -1.0937e-02, 2.3224e-02, -3.5859e-03, -5.3420e-02,\n -1.5661e-02, 2.6106e-02, 1.3404e-02, 2.5665e-02, 2.3201e-02,\n -1.3421e-02, -1.5187e-02, 1.5093e-02, 2.2707e-02, 1.8863e-02,\n -5.0320e-03, 2.4879e-02, -3.2277e-03, -1.6981e-03, -3.5395e-02,\n 1.0851e-02, -1.1235e-02, -1.4035e-02, -1.8554e-03, -1.4879e-02,\n 1.6353e-02, -1.7177e-02, -3.6586e-02, 1.4019e-02, 3.3692e-02,\n -3.8917e-02, 1.0824e-02, 1.0028e-02, -1.4208e-02, -1.5659e-03,\n 1.9637e-02, 3.4607e-02, -2.3749e-03, 1.7911e-02, 4.7467e-03,\n -5.5257e-03, -3.0690e-04, -2.0636e-02, -5.5741e-03, 3.4466e-02,\n -1.0625e-02, 6.0976e-03, -7.6574e-03, 4.3836e-02, 3.6158e-02,\n 6.5411e-03, 4.6887e-02, 2.8918e-03, -4.8728e-03, -7.9181e-02,\n 7.9945e-03, -2.5518e-02, -2.6316e-03, 8.5638e-03, -2.6370e-03,\n -2.9238e-03, -5.2411e-04, 2.4355e-02, 1.8629e-02, 8.4759e-03,\n 2.0407e-02, -5.6805e-02, -1.5662e-03, -1.0971e-03, 1.6802e-02,\n -1.6344e-02, 2.4147e-02, 4.1619e-03, 4.8996e-03, -3.1324e-02,\n 7.3548e-04, 3.6705e-03, -1.5056e-02, -4.5280e-02, -9.3660e-03,\n -1.6968e-02, -3.9965e-03, 6.1707e-04, -1.1390e-02, -9.7548e-03,\n -1.3074e-03, -8.2523e-04, 6.3772e-03, -2.1490e-02, -1.8934e-02,\n 1.9771e-02, -3.5844e-02, 4.5269e-03, 4.5064e-03, 1.5126e-02,\n 2.4978e-03, 7.4933e-03, 1.0131e-02, 1.8943e-02, -9.7679e-03,\n 6.0188e-03, 4.0893e-02, 9.8221e-03, 3.0524e-03, 2.7150e-02,\n 1.9952e-02, 5.2079e-03, -2.3833e-03, -7.0037e-03, 3.4771e-02,\n 1.0544e-02, 1.9827e-02, -1.7007e-02, -4.7398e-02, -5.2514e-02,\n 3.7329e-02, -4.7748e-02, -2.0122e-02, 2.2415e-02, -3.2224e-03,\n 2.7437e-02, -1.7444e-02, -2.1704e-02, -1.2059e-02, -2.1734e-03,\n -5.7376e-03, -1.8165e-02, -1.1038e-02, 3.0465e-02, -4.9724e-03,\n 1.6105e-02, -1.4725e-02, 6.9043e-03, 4.6574e-02, 1.2863e-02,\n -6.2336e-03, 5.5427e-02, -9.4873e-03, 2.8956e-02, -3.1223e-02,\n 2.0518e-02, -2.2657e-03, -3.2693e-02, 1.6381e-02, -9.9133e-03,\n 8.6844e-04, 2.3570e-02, 8.0049e-03, 3.5122e-02, -3.7207e-02,\n -9.0648e-03, 1.8155e-02, 1.1116e-02, 9.1401e-03, -3.4643e-03,\n -1.1809e-02, 1.3212e-02, -1.3844e-02, -5.5379e-03, -2.9721e-03,\n 4.9281e-03, 1.2826e-02, -9.3373e-03, 1.2189e-02, -3.1024e-02,\n 1.8964e-02, 1.0482e-03, -7.0562e-02, -1.3350e-02, -6.0344e-02,\n -1.5159e-02, -4.1496e-03, -1.2962e-03, 1.1805e-02, -1.3752e-02,\n -7.7286e-03, -2.9281e-04, -1.6601e-02, 1.1870e-03, 1.6660e-02,\n 1.5674e-02, -1.1843e-02, 1.4354e-02, -3.3468e-02, 2.3114e-02,\n 3.0233e-03, -3.1130e-02, -5.4174e-03, -6.7430e-03, 2.0263e-02,\n -1.3765e-02, 1.7033e-02, 1.6376e-02, -2.1871e-02, -1.2658e-02,\n 8.9020e-03, 2.6257e-02, -1.9029e-02, -2.3954e-02, -6.5403e-03,\n -2.6069e-03, 1.2913e-02, -1.5120e-02, 2.5506e-02, 3.2572e-02,\n -4.4677e-03, 2.8049e-02, -6.1418e-03, 2.4198e-02, -5.6070e-02,\n -7.1119e-03, 5.1588e-05, -1.3852e-03, -1.3575e-02, 3.2372e-03,\n -8.9531e-03, 1.5040e-02, 3.0446e-02, 6.1889e-02, -3.8240e-03,\n -2.5519e-02, -1.5132e-02, -6.5424e-03, 2.0394e-02, 2.2815e-02,\n -3.5695e-02, 4.7660e-03, 2.7675e-02, -4.7882e-03, 3.8173e-03,\n -2.5800e-02, 2.4386e-02, -6.4826e-03, 1.2559e-02, -1.4047e-03,\n 1.7719e-02, -4.4879e-02, 1.9942e-03, 9.1627e-03, -5.7775e-02,\n 8.5031e-03, 1.5875e-03, -3.1919e-03, 1.0848e-02, 7.9421e-02,\n 8.6447e-03, -1.3437e-02, -6.0144e-03, 1.4354e-02, -1.0426e-02,\n 2.7296e-02, 2.2907e-02, 7.4398e-03, 7.2754e-03, 1.8063e-02,\n 1.0656e-02, -1.1258e-02, 4.0771e-02, 1.1695e-02, -2.3708e-03,\n -4.0431e-03, -4.5782e-02, -1.1644e-02, 3.1177e-02, 7.2370e-03,\n -4.7481e-02, -2.8037e-03, -4.8167e-02, 1.1179e-02, 1.7628e-03,\n 2.3841e-02, 1.4055e-03, 8.2538e-03, 5.2495e-03, -3.3605e-02,\n -2.2332e-02, 2.5498e-02, 2.2015e-02, -3.0085e-02, 1.9280e-02,\n -1.5545e-02, -1.0492e-02, 5.9703e-03, -3.6561e-03, -7.0754e-03,\n -5.5848e-02, -3.1192e-02, -3.2975e-03, 3.3136e-02, 6.1243e-03,\n 2.0121e-02, -1.4502e-02, 7.9186e-03, -8.6687e-03, -4.0311e-03,\n -1.8577e-02, 3.6946e-02, -8.8441e-03, 9.3628e-03, -2.8559e-02,\n -5.8829e-03, -8.9988e-03, 3.0501e-03, -2.7434e-02, -1.5944e-02,\n 3.0584e-02, -3.4951e-02, -4.0043e-02, -1.7494e-02, 2.6877e-02,\n 1.8801e-02, -6.5205e-03, -1.7808e-02, -2.8063e-02, -3.4563e-03,\n -1.0193e-02, -1.3519e-03, 1.4158e-02, -1.3357e-02, 1.4662e-02,\n 3.1759e-02, 2.3299e-02, 2.2801e-02, 1.6106e-02, -1.1098e-02,\n 1.5057e-02, 3.6226e-03, -1.7602e-02, -1.5314e-02, -3.9627e-02,\n 6.8139e-02, 8.0705e-04, -3.4062e-03, 6.7188e-03, -3.4910e-02,\n 1.1566e-02, 1.2729e-02, -6.5061e-03, -4.1888e-02, -7.0573e-03,\n -5.4524e-04, 2.1974e-02, 2.4095e-02, 2.9447e-02, -2.2365e-02,\n -1.1549e-02, -4.6290e-03, 1.1450e-02, 1.3905e-02, 2.2551e-02,\n 4.6688e-02, -4.0011e-03, -1.9564e-02, -9.2004e-03, -3.4919e-02,\n 1.6655e-02, 4.8478e-03, -1.5637e-03, 1.0506e-03, -6.1508e-03,\n -1.1711e-02, 2.1589e-02, 3.4279e-02, -1.4546e-02, -1.5230e-02,\n 4.2678e-02, 3.0476e-02, 2.3720e-03, -2.6752e-02, -4.0102e-03,\n -3.6202e-02, 2.3622e-03, -7.7354e-03, 6.2482e-03, 8.2080e-03,\n 5.0929e-03, -1.5668e-02, -4.6139e-02, 9.2474e-03, -9.2137e-03,\n -8.3459e-03, -3.0361e-02, -1.0694e-02, -1.6979e-03, -6.9064e-03,\n 2.8304e-02, 1.6010e-02, -1.1776e-02, 1.0324e-02, 2.5138e-03,\n -4.1980e-03, 2.1287e-02, 1.7605e-02, 1.9882e-02, 1.2172e-04,\n -2.7392e-02, 1.1579e-03, 3.1266e-03, -7.8805e-03, 2.3212e-02,\n 8.8162e-03, -1.0174e-02, -1.7105e-02, 4.4913e-02, -8.5784e-03,\n 6.5764e-03, -9.3736e-03, 1.2461e-02, 7.6065e-03, 4.7172e-03,\n -3.8011e-03, -2.9922e-03, -6.7235e-03, 2.5287e-02, -2.8385e-03,\n 1.9191e-02, 5.9655e-03, -1.5364e-02, -5.3431e-04, 4.4472e-02,\n -1.3154e-02, 2.5025e-02, -4.1851e-06, -4.6314e-03, 1.6846e-02,\n 2.7155e-03, -9.5421e-03, 4.9021e-03, 4.3820e-03, -2.0072e-02,\n 6.7053e-04, 7.4226e-04, -1.7608e-02, -3.9608e-02, -1.1947e-02,\n -2.2722e-02, -4.7142e-02, -3.2867e-03, 3.3100e-02, -4.3785e-02,\n 6.7138e-03, 4.0571e-03, -5.2462e-02, 3.5313e-02, 4.8415e-03,\n 1.1184e-02, 2.2603e-02, 1.1727e-02, -1.2257e-03, -1.4775e-02,\n 1.1809e-02, -2.2602e-03, -8.8450e-04, -1.1740e-02, -3.6093e-03,\n 1.4647e-02, 3.8674e-03], device='cuda:0')",
|
| 13 |
+
"exp_avg_sq": "tensor([0.0034, 0.0061, 0.0031, 0.0034, 0.0044, 0.0060, 0.0024, 0.0055, 0.0060,\n 0.0051, 0.0031, 0.0040, 0.0061, 0.0030, 0.0075, 0.0076, 0.0047, 0.0040,\n 0.0046, 0.0040, 0.0043, 0.0087, 0.0045, 0.0031, 0.0025, 0.0070, 0.0040,\n 0.0056, 0.0055, 0.0050, 0.0038, 0.0093, 0.0040, 0.0084, 0.0040, 0.0074,\n 0.0051, 0.0045, 0.0037, 0.0045, 0.0041, 0.0046, 0.0031, 0.0042, 0.0084,\n 0.0039, 0.0032, 0.0042, 0.0056, 0.0051, 0.0058, 0.0038, 0.0042, 0.0048,\n 0.0039, 0.0038, 0.0053, 0.0023, 0.0028, 0.0040, 0.0064, 0.0033, 0.0053,\n 0.0053, 0.0048, 0.0070, 0.0044, 0.0042, 0.0052, 0.0039, 0.0055, 0.0031,\n 0.0043, 0.0093, 0.0045, 0.0026, 0.0069, 0.0037, 0.0052, 0.0065, 0.0055,\n 0.0053, 0.0038, 0.0065, 0.0047, 0.0057, 0.0042, 0.0071, 0.0075, 0.0039,\n 0.0101, 0.0056, 0.0045, 0.0038, 0.0061, 0.0024, 0.0038, 0.0062, 0.0032,\n 0.0048, 0.0060, 0.0056, 0.0057, 0.0031, 0.0033, 0.0062, 0.0063, 0.0036,\n 0.0052, 0.0063, 0.0051, 0.0056, 0.0074, 0.0036, 0.0050, 0.0028, 0.0025,\n 0.0034, 0.0071, 0.0039, 0.0057, 0.0100, 0.0056, 0.0041, 0.0083, 0.0053,\n 0.0035, 0.0025, 0.0068, 0.0066, 0.0074, 0.0051, 0.0047, 0.0040, 0.0029,\n 0.0015, 0.0034, 0.0051, 0.0019, 0.0066, 0.0051, 0.0049, 0.0051, 0.0041,\n 0.0045, 0.0035, 0.0050, 0.0032, 0.0026, 0.0047, 0.0045, 0.0025, 0.0044,\n 0.0041, 0.0038, 0.0025, 0.0045, 0.0036, 0.0054, 0.0032, 0.0038, 0.0023,\n 0.0030, 0.0035, 0.0047, 0.0038, 0.0027, 0.0034, 0.0066, 0.0063, 0.0065,\n 0.0069, 0.0059, 0.0039, 0.0054, 0.0027, 0.0061, 0.0044, 0.0061, 0.0042,\n 0.0047, 0.0035, 0.0027, 0.0056, 0.0047, 0.0061, 0.0055, 0.0081, 0.0058,\n 0.0055, 0.0040, 0.0046, 0.0049, 0.0040, 0.0047, 0.0081, 0.0025, 0.0043,\n 0.0027, 0.0045, 0.0027, 0.0032, 0.0039, 0.0052, 0.0060, 0.0058, 0.0105,\n 0.0041, 0.0023, 0.0050, 0.0033, 0.0025, 0.0067, 0.0028, 0.0033, 0.0051,\n 0.0053, 0.0043, 0.0024, 0.0056, 0.0039, 0.0048, 0.0063, 0.0055, 0.0054,\n 0.0028, 0.0020, 0.0050, 0.0074, 0.0043, 0.0040, 0.0038, 0.0051, 0.0067,\n 0.0049, 0.0027, 0.0030, 0.0037, 0.0058, 0.0043, 0.0015, 0.0043, 0.0048,\n 0.0048, 0.0049, 0.0041, 0.0040, 0.0035, 0.0042, 0.0053, 0.0020, 0.0032,\n 0.0065, 0.0062, 0.0051, 0.0077, 0.0030, 0.0040, 0.0031, 0.0069, 0.0036,\n 0.0053, 0.0056, 0.0052, 0.0036, 0.0027, 0.0036, 0.0030, 0.0038, 0.0078,\n 0.0053, 0.0036, 0.0036, 0.0074, 0.0060, 0.0025, 0.0040, 0.0044, 0.0063,\n 0.0044, 0.0040, 0.0073, 0.0073, 0.0022, 0.0018, 0.0043, 0.0044, 0.0057,\n 0.0035, 0.0033, 0.0029, 0.0021, 0.0045, 0.0049, 0.0042, 0.0056, 0.0049,\n 0.0020, 0.0040, 0.0084, 0.0044, 0.0025, 0.0035, 0.0045, 0.0055, 0.0069,\n 0.0044, 0.0041, 0.0064, 0.0048, 0.0063, 0.0039, 0.0067, 0.0040, 0.0055,\n 0.0037, 0.0049, 0.0060, 0.0042, 0.0050, 0.0056, 0.0043, 0.0040, 0.0050,\n 0.0038, 0.0067, 0.0048, 0.0031, 0.0039, 0.0052, 0.0042, 0.0053, 0.0040,\n 0.0044, 0.0054, 0.0054, 0.0034, 0.0021, 0.0056, 0.0047, 0.0074, 0.0088,\n 0.0020, 0.0045, 0.0054, 0.0067, 0.0036, 0.0037, 0.0052, 0.0027, 0.0027,\n 0.0064, 0.0041, 0.0033, 0.0044, 0.0043, 0.0028, 0.0038, 0.0050, 0.0030,\n 0.0034, 0.0088, 0.0036, 0.0110, 0.0056, 0.0055, 0.0017, 0.0038, 0.0036,\n 0.0031, 0.0060, 0.0051, 0.0055, 0.0034, 0.0034, 0.0069, 0.0051, 0.0053,\n 0.0077, 0.0060, 0.0040, 0.0019, 0.0027, 0.0029, 0.0045, 0.0090, 0.0056,\n 0.0035, 0.0032, 0.0076, 0.0031, 0.0037, 0.0024, 0.0051, 0.0064, 0.0041,\n 0.0040, 0.0045, 0.0074, 0.0102, 0.0036, 0.0041, 0.0036, 0.0042, 0.0038,\n 0.0089, 0.0040, 0.0023, 0.0041, 0.0047, 0.0039, 0.0039, 0.0029, 0.0061,\n 0.0059, 0.0066, 0.0051, 0.0106, 0.0040, 0.0085, 0.0066, 0.0043, 0.0092,\n 0.0028, 0.0057, 0.0027, 0.0028, 0.0044, 0.0041, 0.0032, 0.0038, 0.0098,\n 0.0077, 0.0039, 0.0040, 0.0038, 0.0065, 0.0081, 0.0037, 0.0041, 0.0064,\n 0.0039, 0.0024, 0.0043, 0.0069, 0.0044, 0.0027, 0.0032, 0.0071, 0.0032,\n 0.0043, 0.0053, 0.0037, 0.0046, 0.0043, 0.0057, 0.0050, 0.0029, 0.0043,\n 0.0038, 0.0026, 0.0033, 0.0032, 0.0051, 0.0066, 0.0041, 0.0041, 0.0082,\n 0.0044, 0.0091, 0.0039, 0.0054, 0.0035, 0.0036, 0.0066, 0.0020, 0.0066,\n 0.0033, 0.0036, 0.0046, 0.0057, 0.0055, 0.0043, 0.0077, 0.0065, 0.0032,\n 0.0037, 0.0083, 0.0060, 0.0045, 0.0030, 0.0054, 0.0033, 0.0045, 0.0040,\n 0.0048, 0.0033, 0.0074, 0.0037, 0.0048, 0.0057, 0.0025, 0.0031, 0.0040,\n 0.0027, 0.0030, 0.0043, 0.0007, 0.0058, 0.0081, 0.0033, 0.0056],\n device='cuda:0')"
|
| 14 |
+
},
|
| 15 |
+
"2": {
|
| 16 |
+
"step": "tensor(3756.)",
|
| 17 |
+
"exp_avg": "tensor([-3.8232e-03, -1.6397e-03, -1.0210e-03, -4.0746e-03, 4.6383e-03,\n -3.3967e-03, 1.2117e-02, -4.0425e-03, 3.0530e-03, -2.4434e-03,\n 6.9064e-03, 8.6173e-04, 1.6080e-03, 4.2205e-03, -2.1715e-03,\n -1.3820e-03, 3.0267e-03, -3.5267e-03, 2.1569e-03, -7.0411e-03,\n 3.9578e-04, 6.1193e-04, 8.3370e-04, -3.9721e-03, -3.4225e-03,\n 8.9434e-04, 1.1877e-03, -1.3331e-03, -4.1333e-03, -2.6295e-03,\n 3.0341e-03, -1.0307e-03, 1.6278e-03, -6.5423e-04, -2.0573e-03,\n 3.9550e-04, -2.3171e-03, 4.8506e-04, -5.8707e-03, 3.9251e-03,\n 9.7879e-04, -3.9052e-03, -3.3605e-03, 6.2907e-03, 2.0978e-03,\n 3.5550e-03, 4.3128e-03, 6.0364e-03, 3.0349e-03, 7.0363e-04,\n 1.5791e-03, -5.8049e-03, -7.0300e-04, -1.7464e-03, -3.1062e-04,\n 2.0174e-03, -2.0791e-03, 7.0522e-03, -5.2365e-04, -1.1335e-02,\n -2.0603e-03, 5.9836e-03, 1.9626e-03, 3.4149e-03, 2.3977e-03,\n -1.1504e-03, -1.1616e-03, 1.9426e-03, 3.1548e-03, 2.4246e-03,\n -3.6964e-04, 5.8146e-03, -6.6673e-04, -7.0421e-04, -4.1908e-03,\n 3.3199e-03, -1.3366e-03, -3.5823e-03, -7.9915e-04, -2.4573e-03,\n 3.6869e-04, -7.5964e-05, -9.0390e-03, 1.5305e-03, 5.4214e-03,\n -4.3001e-03, 3.1054e-03, -1.4007e-04, -1.6551e-03, -4.7637e-04,\n 1.3845e-03, 6.4701e-03, -4.1187e-04, 4.0654e-03, 8.5460e-04,\n -1.1094e-03, 1.0472e-03, -1.8597e-03, -1.0058e-03, 3.4279e-03,\n -1.5466e-03, 2.3631e-03, -1.1406e-03, 1.0256e-02, 6.3160e-03,\n 9.5589e-05, 4.9202e-03, 1.0530e-03, -2.1668e-03, -7.8717e-03,\n 1.6505e-03, -3.4160e-03, 5.3260e-04, 1.3272e-03, -2.5545e-04,\n -2.3050e-04, -3.7615e-04, 4.2416e-03, 2.3084e-03, 4.8959e-04,\n 1.9260e-03, -5.9499e-03, -5.2166e-04, -1.8965e-03, 2.3404e-03,\n -3.3071e-03, 4.3783e-03, 1.9403e-03, 6.4966e-04, -3.1669e-03,\n -1.8722e-03, 2.9091e-04, -1.8168e-03, -7.9114e-03, -3.5881e-03,\n -1.2222e-02, -1.4945e-03, 7.8308e-04, -4.2892e-03, -4.3289e-04,\n -1.5073e-03, 1.0961e-04, 3.3182e-04, -4.3809e-03, -3.2402e-03,\n 3.9983e-03, -3.5059e-03, 6.6061e-04, 2.1192e-03, 9.4257e-04,\n 1.3534e-04, 4.7296e-03, 3.9229e-03, 3.8738e-03, -1.5724e-03,\n 1.7034e-03, 7.3767e-03, 1.9455e-03, -9.3838e-06, 3.8434e-03,\n 2.8693e-03, 2.2100e-03, -6.8345e-04, -8.1221e-04, 6.3223e-03,\n 1.1825e-03, 6.0730e-03, -3.0812e-03, -4.1030e-03, -3.6914e-03,\n 3.7369e-03, -3.2966e-03, -1.4117e-03, 4.9811e-03, -6.0739e-04,\n 5.9174e-03, -1.1100e-03, -8.1785e-04, -1.2234e-03, 9.8462e-05,\n -2.1136e-04, -4.1965e-03, -4.1048e-03, 4.0671e-03, -1.3424e-03,\n 1.4690e-03, -2.4006e-03, 5.7845e-04, 6.6211e-03, 2.6349e-03,\n -8.6628e-04, 6.5579e-03, -9.5245e-04, 5.8284e-03, -5.6592e-03,\n 2.2054e-03, -2.4150e-03, -4.3273e-03, 6.0820e-03, -8.2204e-04,\n -5.1147e-04, 4.8256e-03, 3.1654e-03, 3.3415e-03, -3.0285e-03,\n 4.3116e-04, 2.6375e-04, 1.3278e-03, 1.8873e-03, 2.1694e-03,\n -2.3544e-03, 4.4367e-03, -1.3492e-03, -2.5118e-03, -2.3766e-04,\n 1.3999e-03, 7.2722e-04, -1.4761e-03, 3.5014e-03, -4.7280e-03,\n 4.4040e-03, -5.6488e-05, -5.2856e-03, -2.8972e-04, -8.5121e-03,\n -3.2751e-03, -2.4823e-03, -4.8345e-04, 9.0014e-04, -1.8586e-03,\n -4.7867e-04, -5.1450e-04, -1.2842e-03, -1.0990e-04, 2.1474e-03,\n 4.4235e-03, -3.2873e-03, 2.9311e-03, -3.6021e-03, 2.9713e-03,\n 2.8872e-03, -4.5157e-03, -1.2427e-03, -2.4789e-03, 3.4719e-03,\n -1.4907e-03, 2.9819e-03, 2.8654e-03, -2.8922e-03, -2.4596e-03,\n 3.9484e-03, 6.6545e-03, -1.5605e-03, -2.5654e-03, -9.0281e-04,\n 1.3405e-04, 2.1900e-03, -2.4502e-03, 6.1020e-03, 3.8476e-03,\n 7.1517e-04, 2.8352e-03, -1.6999e-03, 4.6878e-03, -1.3612e-02,\n -1.5944e-03, 6.6080e-04, 1.7011e-04, -3.7364e-03, 1.2944e-04,\n -7.4860e-04, 2.4901e-03, 6.2165e-03, 4.9753e-03, 7.3049e-04,\n -5.5854e-03, -2.8718e-03, -5.4057e-04, 1.5831e-03, 4.2426e-03,\n -5.9121e-03, 2.7685e-04, 2.9045e-03, -1.3611e-03, 2.6355e-03,\n -2.4626e-03, 1.8971e-03, -2.3983e-03, 3.5778e-03, 6.4848e-04,\n 3.3054e-03, -1.9309e-02, 2.3411e-04, 1.8142e-03, -6.4247e-03,\n 1.4360e-03, 2.4054e-04, -2.4325e-03, 2.3963e-03, 7.4911e-03,\n 1.7857e-04, -2.9299e-03, -1.7652e-03, 1.8811e-03, -2.0270e-03,\n 2.3461e-03, 3.0829e-03, 2.0107e-03, 5.0216e-04, 2.9952e-03,\n 1.4499e-03, -3.3243e-04, 4.9163e-03, 2.2471e-03, 2.5017e-05,\n -9.0892e-04, -6.7616e-03, -4.9244e-04, 5.4089e-03, 1.0053e-03,\n -6.6529e-03, -8.5690e-04, -7.3705e-03, 8.9773e-04, 6.7272e-04,\n 2.2827e-03, 2.3043e-04, 2.3791e-03, 1.3387e-03, -7.7609e-03,\n -2.6471e-03, 3.0461e-03, 2.5404e-03, -5.0199e-03, 2.9787e-03,\n -9.4906e-04, -1.6315e-03, 2.2848e-03, 1.0783e-03, -8.9133e-04,\n -5.6765e-03, -2.3406e-03, -1.3567e-03, 4.3125e-03, 7.8894e-04,\n 1.3243e-03, -2.4020e-03, 1.1041e-03, -2.4089e-03, -1.8501e-03,\n -4.6368e-03, 4.3222e-03, -1.9785e-03, 1.9397e-03, -4.2973e-03,\n -1.3585e-03, -2.5821e-03, 1.1199e-03, -1.8066e-03, -2.9276e-03,\n 7.6289e-03, -3.5350e-03, -7.4869e-03, -1.2727e-03, 3.5138e-03,\n 2.1153e-03, -2.4189e-03, -2.5223e-03, -5.0938e-03, 3.1868e-04,\n -5.2456e-04, -1.7126e-04, 1.4906e-03, -2.2434e-03, 4.0231e-03,\n 3.1576e-03, 2.3327e-03, 2.9064e-03, 1.0345e-03, -1.0266e-03,\n 1.0089e-03, 8.6231e-04, -3.4490e-03, -2.6455e-03, -6.1605e-03,\n 5.7907e-03, -7.5402e-05, -1.1617e-03, 1.2122e-03, -4.9077e-03,\n 1.5875e-03, 1.0089e-03, -8.4651e-04, -3.7033e-03, -1.2708e-03,\n -8.9101e-05, 1.7581e-03, 4.7974e-03, 3.9791e-03, -2.7438e-03,\n -2.8367e-03, -4.2353e-04, 2.4457e-03, 1.6684e-03, 3.1654e-03,\n 4.3703e-03, -1.2334e-03, -4.6973e-03, -1.6569e-03, -4.5868e-03,\n 2.0037e-03, 1.4512e-03, -3.8849e-04, -9.9309e-04, -8.6930e-04,\n -1.2446e-03, 1.7638e-03, 4.2534e-03, -1.5525e-03, -7.6845e-05,\n 5.8191e-03, 4.4860e-03, 2.0676e-04, -5.6273e-03, -8.2557e-04,\n -1.1065e-02, -9.2922e-04, -2.6523e-03, 3.7929e-05, 2.7240e-03,\n 8.5019e-04, -1.8443e-03, -7.6718e-03, 2.0663e-03, -3.2680e-03,\n -1.5347e-03, -2.5689e-03, -9.4767e-04, -1.5669e-03, -1.3933e-03,\n 3.5156e-03, 1.1046e-03, -4.3386e-03, 1.6627e-03, -1.0207e-03,\n -5.8478e-04, 3.9060e-03, 2.6958e-03, 2.0970e-03, 4.9914e-04,\n -3.6578e-03, 7.5467e-04, 8.3560e-04, -7.5659e-04, 5.3464e-03,\n 5.1763e-04, -1.8288e-03, -3.5557e-03, 5.8412e-03, -2.5337e-03,\n 7.1687e-04, -2.6906e-03, 4.1935e-03, -1.2517e-03, 6.4318e-04,\n -6.5397e-04, -5.4968e-04, 7.6766e-05, 6.2681e-03, -1.3416e-03,\n 3.4218e-03, 2.0072e-03, -2.7761e-03, 7.4963e-04, 4.5388e-03,\n -5.4899e-03, 2.7788e-03, -5.0380e-04, -9.7595e-04, 2.7079e-03,\n 6.0335e-05, 8.9102e-05, 1.0114e-03, 9.4225e-04, -1.2840e-03,\n -6.4696e-04, 1.0087e-04, -1.9680e-03, -3.5246e-03, -2.3386e-03,\n -5.0812e-03, -6.0286e-03, -7.2772e-04, 5.3849e-03, -5.7066e-03,\n 1.5333e-03, 1.3602e-03, -3.7028e-03, 8.2098e-03, -5.1497e-04,\n 9.9816e-04, 5.5319e-03, 2.4035e-03, 1.0995e-03, -2.7137e-03,\n 2.4697e-03, -1.5286e-03, 6.5226e-02, -1.3668e-03, 1.9327e-04,\n 4.5763e-03, -3.3860e-04], device='cuda:0')",
|
| 18 |
+
"exp_avg_sq": "tensor([9.6168e-05, 5.5701e-05, 1.0642e-04, 8.6665e-05, 9.3219e-05, 6.2647e-05,\n 1.8632e-04, 4.2113e-05, 3.9415e-05, 5.6621e-05, 1.3859e-04, 5.1311e-05,\n 5.2565e-05, 9.4228e-05, 4.6351e-05, 6.6350e-05, 6.8699e-05, 5.1146e-05,\n 6.2194e-05, 7.6297e-05, 3.2476e-05, 5.5471e-05, 4.6977e-05, 1.0289e-04,\n 1.5221e-04, 6.0421e-05, 7.4801e-05, 6.8150e-05, 6.8895e-05, 6.7756e-05,\n 7.1341e-05, 7.4975e-05, 9.6087e-05, 6.5403e-05, 9.3067e-05, 3.8217e-05,\n 5.7230e-05, 6.4178e-05, 8.7975e-05, 9.8471e-05, 1.2092e-04, 9.8692e-05,\n 1.1432e-04, 8.3053e-05, 4.7126e-05, 1.0149e-04, 1.0264e-04, 8.8348e-05,\n 5.1339e-05, 5.8539e-05, 6.1255e-05, 1.0782e-04, 1.0446e-04, 4.9303e-05,\n 9.5678e-05, 1.2607e-04, 4.2222e-05, 1.3150e-04, 1.6972e-04, 1.1589e-04,\n 6.5476e-05, 1.1373e-04, 7.9460e-05, 6.7900e-05, 9.2633e-05, 9.3173e-05,\n 6.0757e-05, 6.2745e-05, 4.9153e-05, 5.6286e-05, 5.8082e-05, 9.3240e-05,\n 5.2230e-05, 6.1410e-05, 6.4350e-05, 1.6291e-04, 5.1348e-05, 1.6187e-04,\n 4.8463e-05, 3.9851e-05, 6.8544e-05, 9.1249e-05, 1.2848e-04, 3.7420e-05,\n 1.3498e-04, 6.6051e-05, 9.0856e-05, 5.4639e-05, 3.6494e-05, 9.1842e-05,\n 6.0154e-05, 9.5500e-05, 3.7674e-05, 7.5844e-05, 6.5925e-05, 1.6916e-04,\n 6.4942e-05, 5.8595e-05, 1.1093e-04, 3.6879e-05, 4.4575e-05, 5.6019e-05,\n 5.3446e-05, 9.3076e-05, 6.6885e-05, 4.9247e-05, 9.1384e-05, 5.9920e-05,\n 8.4508e-05, 5.4257e-05, 8.6140e-05, 3.9920e-05, 6.4953e-05, 4.2788e-05,\n 6.0224e-05, 1.7370e-04, 1.2947e-04, 6.1560e-05, 9.7266e-05, 6.8281e-05,\n 5.4598e-05, 5.5647e-05, 7.5656e-05, 1.1696e-04, 7.0792e-05, 6.3557e-05,\n 5.1532e-05, 1.2344e-04, 5.7105e-05, 7.0826e-05, 5.4756e-05, 6.6369e-05,\n 6.0646e-05, 1.1344e-04, 1.2205e-04, 4.5047e-04, 5.3759e-05, 5.1978e-05,\n 2.7290e-04, 1.0257e-04, 6.5447e-05, 4.0020e-05, 3.7471e-05, 7.6116e-05,\n 1.0189e-04, 8.6103e-05, 7.1779e-05, 1.1996e-04, 2.3965e-04, 1.0062e-04,\n 5.6080e-05, 3.7266e-04, 9.8890e-05, 1.1597e-04, 5.1336e-05, 2.2619e-04,\n 5.7077e-05, 1.4663e-04, 6.6801e-05, 7.1277e-05, 5.2605e-05, 1.7049e-04,\n 8.7258e-05, 8.2399e-05, 6.7100e-05, 5.9220e-05, 1.8117e-04, 8.6778e-05,\n 5.0730e-05, 5.4697e-05, 5.7292e-05, 4.4153e-05, 5.0992e-05, 8.9692e-05,\n 6.1623e-05, 9.5654e-05, 5.4566e-05, 4.3633e-05, 6.2905e-05, 5.3515e-05,\n 1.1241e-04, 1.2368e-04, 1.8261e-04, 5.9995e-05, 8.5189e-05, 4.4971e-05,\n 6.8166e-05, 5.2004e-05, 8.7948e-05, 6.7528e-05, 7.4595e-05, 5.1676e-05,\n 5.7712e-05, 1.1430e-04, 7.4959e-05, 4.0628e-05, 1.8902e-04, 4.6834e-05,\n 2.2320e-04, 8.4341e-05, 9.7246e-05, 9.7494e-05, 9.6704e-05, 5.4077e-05,\n 5.9017e-05, 6.2355e-05, 6.8672e-05, 4.2398e-05, 1.6997e-04, 4.8124e-05,\n 1.3716e-04, 1.7790e-04, 6.3981e-05, 1.0710e-04, 7.9157e-05, 8.5675e-05,\n 4.5850e-05, 1.3366e-04, 1.1911e-04, 4.5676e-05, 1.2008e-04, 1.1266e-04,\n 4.2735e-05, 3.9232e-05, 8.5256e-05, 2.3796e-04, 2.7869e-04, 6.1953e-05,\n 5.5546e-05, 9.1156e-05, 5.8087e-05, 6.3714e-05, 5.1727e-05, 5.0837e-05,\n 4.8912e-05, 1.5690e-04, 1.4618e-04, 9.5436e-05, 4.4259e-05, 7.7949e-05,\n 4.8602e-04, 8.1730e-05, 8.9570e-05, 9.3653e-05, 1.0242e-04, 1.0018e-04,\n 1.0793e-04, 1.0172e-04, 7.7579e-05, 6.5240e-05, 3.1128e-04, 1.3673e-04,\n 4.8604e-05, 5.9652e-05, 1.1858e-04, 7.0504e-05, 1.0986e-04, 1.2541e-04,\n 1.1251e-04, 7.5516e-05, 7.3188e-05, 8.0533e-05, 4.9523e-05, 7.1976e-05,\n 1.1281e-04, 1.2888e-04, 1.2894e-04, 1.5643e-04, 1.4068e-04, 4.7572e-05,\n 7.1576e-05, 9.8587e-05, 1.1414e-04, 3.7332e-05, 5.7950e-05, 1.0004e-04,\n 1.1630e-04, 5.4720e-05, 4.3919e-05, 1.6694e-04, 7.4773e-05, 6.4344e-05,\n 5.5466e-05, 3.1312e-04, 2.3163e-04, 4.8671e-05, 5.8282e-05, 5.2649e-05,\n 1.1673e-04, 1.4505e-04, 1.2041e-04, 2.4477e-04, 6.1627e-05, 7.1042e-05,\n 5.4568e-05, 8.0500e-05, 7.5674e-05, 2.9698e-04, 6.7669e-05, 7.1135e-05,\n 6.3367e-05, 9.2269e-05, 9.4848e-05, 7.3340e-05, 1.1173e-04, 4.3297e-05,\n 7.6723e-05, 5.5055e-05, 8.1377e-05, 6.6960e-05, 9.0751e-05, 9.2065e-05,\n 7.0950e-05, 1.4190e-04, 4.8784e-05, 7.8965e-05, 7.5281e-05, 7.0025e-05,\n 9.9225e-05, 3.8240e-05, 7.5202e-05, 7.0902e-05, 6.0698e-05, 8.2984e-05,\n 6.8298e-05, 5.9530e-05, 1.5637e-04, 1.4435e-04, 1.0828e-04, 1.4800e-04,\n 5.7593e-05, 7.3886e-05, 4.4522e-05, 6.8591e-05, 7.7450e-05, 3.1030e-05,\n 7.3315e-05, 1.8411e-04, 5.2605e-05, 6.1554e-05, 6.5826e-05, 5.9970e-05,\n 2.7556e-04, 5.7184e-05, 3.8165e-05, 4.7162e-05, 6.6876e-05, 8.0989e-05,\n 6.3877e-05, 1.8840e-04, 1.3101e-04, 6.6667e-05, 6.7065e-05, 7.4637e-05,\n 6.8482e-05, 1.1247e-04, 1.5769e-04, 7.9301e-05, 3.6952e-05, 1.2051e-04,\n 1.4890e-04, 7.9362e-05, 9.0029e-05, 7.5405e-05, 6.2290e-05, 4.1517e-05,\n 5.6088e-04, 3.3891e-05, 7.4827e-05, 1.5084e-04, 5.9689e-05, 4.3360e-05,\n 4.5068e-05, 1.0804e-04, 1.2517e-04, 6.3465e-05, 4.1845e-05, 8.2388e-05,\n 5.6875e-05, 6.8548e-05, 3.3477e-05, 2.5251e-04, 1.4175e-04, 1.0302e-04,\n 5.7852e-05, 4.6229e-05, 4.1553e-05, 1.5325e-04, 1.1906e-04, 5.8060e-05,\n 1.0056e-04, 6.9376e-05, 9.1686e-05, 7.5711e-05, 6.5340e-05, 7.2854e-05,\n 4.7828e-05, 6.8399e-05, 6.5467e-05, 1.0981e-04, 1.2876e-04, 6.0495e-05,\n 1.3455e-04, 4.7653e-05, 1.1251e-04, 6.1181e-05, 6.2493e-05, 1.3060e-04,\n 6.7152e-05, 6.6508e-05, 7.2772e-05, 6.8023e-05, 9.4625e-05, 5.2255e-05,\n 6.0429e-05, 4.0293e-05, 8.8497e-05, 6.8487e-05, 7.6317e-05, 6.4180e-05,\n 4.2293e-05, 6.3329e-05, 6.3575e-05, 1.5501e-04, 6.8306e-05, 1.3812e-04,\n 8.9009e-05, 8.6875e-05, 3.7820e-05, 1.3975e-04, 8.0261e-05, 5.1738e-05,\n 8.7930e-05, 7.5628e-05, 6.9927e-05, 5.0949e-05, 5.4870e-05, 8.1954e-05,\n 1.0945e-04, 5.8937e-05, 8.2851e-05, 7.3097e-05, 1.7476e-04, 7.1253e-05,\n 5.2914e-05, 7.2343e-05, 1.1247e-04, 7.0059e-05, 3.9815e-05, 1.2611e-04,\n 5.2057e-05, 5.0983e-05, 6.4289e-05, 9.9487e-05, 9.8102e-05, 3.7717e-05,\n 9.0624e-05, 1.2070e-04, 8.1908e-05, 1.0498e-04, 1.3605e-04, 1.0545e-04,\n 1.0549e-04, 4.5952e-05, 6.8166e-05, 6.1712e-05, 6.0346e-05, 5.9200e-05,\n 1.4954e-04, 7.8733e-05, 8.2456e-05, 5.4241e-05, 9.4397e-05, 7.1237e-05,\n 6.2395e-05, 2.5891e-04, 5.6632e-05, 6.9524e-05, 9.8059e-05, 9.0600e-05,\n 5.1372e-05, 4.6658e-05, 1.2291e-04, 1.3925e-04, 5.8913e-05, 1.3619e-04,\n 1.0577e-04, 7.9560e-05, 4.6509e-05, 7.7260e-05, 7.3028e-05, 4.9912e-05,\n 1.0725e-04, 9.0816e-05, 6.3932e-05, 9.6702e-05, 1.4026e-04, 4.2530e-05,\n 1.2383e-04, 3.7171e-05, 4.5485e-05, 2.0025e-04, 8.8744e-05, 9.7183e-05,\n 9.4149e-05, 1.3035e-04, 1.1539e-04, 4.7811e-02, 8.3708e-05, 6.0849e-05,\n 1.3335e-04, 6.8517e-05], device='cuda:0')"
|
| 19 |
+
},
|
| 20 |
+
"3": {
|
| 21 |
+
"step": "tensor(3756.)",
|
| 22 |
+
"exp_avg": "tensor([-2.5951e-03, -1.7995e-03, -3.2956e-04, -3.0269e-03, 4.5551e-03,\n -3.7965e-03, 6.4864e-03, -6.9376e-03, 3.7037e-03, -1.2758e-03,\n 5.1614e-03, 1.0910e-03, 1.7433e-03, 3.8663e-03, -3.3189e-03,\n -2.3177e-03, 1.8113e-03, -3.3134e-03, 3.6348e-03, -9.1819e-03,\n 1.9023e-03, -5.6421e-04, 2.1344e-03, -1.3443e-03, -2.9883e-03,\n 1.9473e-03, 7.6099e-04, -1.6108e-03, -3.7588e-03, -2.7968e-03,\n 2.3559e-03, -1.3525e-03, 1.8600e-03, -8.2280e-05, -1.9714e-03,\n 1.1321e-03, -2.5444e-03, 1.0359e-03, -4.2058e-03, 2.7422e-03,\n 1.4761e-03, -3.4474e-03, -2.8026e-03, 4.8455e-03, 2.9099e-03,\n 3.0711e-03, 3.2642e-03, 6.3806e-03, 3.3717e-03, 1.5873e-03,\n 1.7912e-03, -4.2398e-03, -1.1134e-03, -1.7221e-03, -4.9186e-04,\n 1.9012e-03, -1.3283e-03, 5.0805e-03, -6.5430e-05, -9.1005e-03,\n -2.4788e-03, 5.0533e-03, 2.2336e-03, 3.9655e-03, 3.2666e-03,\n -2.3668e-03, -2.2364e-03, 2.5287e-03, 3.7433e-03, 3.1272e-03,\n -6.7291e-04, 4.6988e-03, -1.5266e-04, -7.1187e-04, -5.1465e-03,\n 1.9544e-03, -1.2065e-03, -2.8387e-03, -2.0042e-04, -1.8590e-03,\n 1.6663e-03, -9.7326e-04, -6.7153e-03, 1.6727e-03, 5.8425e-03,\n -5.7890e-03, 1.9194e-03, 8.4617e-04, -1.5502e-03, -7.5285e-04,\n 2.5784e-03, 5.8919e-03, -3.2042e-04, 3.9240e-03, 7.7772e-04,\n -9.5631e-04, -1.7553e-04, -3.3069e-03, -5.6288e-04, 3.5285e-03,\n -1.4563e-03, 8.9457e-04, -1.3342e-03, 8.2776e-03, 5.5026e-03,\n 5.5508e-04, 6.3487e-03, 2.4866e-04, -1.3034e-03, -1.0184e-02,\n 1.4704e-03, -3.1561e-03, -1.3311e-03, 1.3403e-03, -2.5799e-05,\n 1.4565e-05, -7.8099e-05, 4.0390e-03, 2.7194e-03, 8.0401e-04,\n 2.9376e-03, -5.9013e-03, -7.1538e-05, -3.2182e-04, 3.0648e-03,\n -2.8122e-03, 3.8148e-03, 1.0052e-03, 7.4655e-04, -4.7815e-03,\n 9.9959e-05, 7.7151e-04, -2.3564e-03, -6.5890e-03, -1.5862e-03,\n -4.1487e-03, -7.1233e-04, 4.5944e-04, -2.3402e-03, -1.3785e-03,\n -2.0697e-04, -3.0170e-04, 9.0770e-04, -3.7640e-03, -2.3538e-03,\n 2.9719e-03, -5.0791e-03, 8.4040e-04, 1.5313e-03, 2.7439e-03,\n 3.3597e-04, 2.2289e-03, 1.5728e-03, 3.6779e-03, -1.7607e-03,\n 1.3729e-03, 5.7663e-03, 1.5431e-03, 2.6398e-04, 3.8553e-03,\n 3.1762e-03, 1.1432e-03, -5.0920e-04, -8.6882e-04, 5.2535e-03,\n 1.4474e-03, 4.2717e-03, -2.6841e-03, -6.0207e-03, -5.9701e-03,\n 5.6275e-03, -5.6339e-03, -2.4287e-03, 3.5272e-03, -9.5934e-04,\n 4.8066e-03, -2.4598e-03, -3.1320e-03, -2.2000e-03, 4.3587e-05,\n -6.1056e-04, -3.4441e-03, -2.0015e-03, 4.5283e-03, -8.6107e-04,\n 1.8610e-03, -2.3314e-03, 6.0485e-04, 7.5936e-03, 2.3138e-03,\n -1.3617e-03, 7.5821e-03, -1.2004e-03, 4.8771e-03, -4.8066e-03,\n 2.6738e-03, -7.0053e-04, -4.8842e-03, 3.3484e-03, -1.1274e-03,\n 2.1622e-04, 4.2635e-03, 1.5749e-03, 4.6333e-03, -4.9248e-03,\n -1.2856e-03, 2.2352e-03, 1.9683e-03, 1.6985e-03, -8.3800e-05,\n -2.0866e-03, 2.6607e-03, -2.0954e-03, -9.0745e-04, -2.3684e-04,\n 1.0712e-03, 1.8231e-03, -2.2269e-03, 2.6031e-03, -3.6388e-03,\n 3.9342e-03, 3.7590e-04, -7.7941e-03, -1.5236e-03, -9.2455e-03,\n -3.0485e-03, -1.3462e-03, 1.0387e-04, 1.1846e-03, -2.0704e-03,\n -8.8671e-04, 2.8862e-04, -2.1250e-03, 8.7214e-04, 2.7775e-03,\n 3.2478e-03, -3.0666e-03, 2.8043e-03, -4.6288e-03, 3.8551e-03,\n 1.1485e-03, -5.1464e-03, -6.7856e-04, -1.3368e-03, 3.5811e-03,\n -2.4646e-03, 2.6418e-03, 2.9466e-03, -2.7348e-03, -2.2950e-03,\n 2.6233e-03, 5.1924e-03, -2.2731e-03, -3.6679e-03, -1.6690e-03,\n -5.5802e-04, 2.1393e-03, -2.8971e-03, 5.2283e-03, 5.0128e-03,\n -5.7256e-04, 3.8357e-03, -1.2963e-03, 5.3628e-03, -9.5067e-03,\n -1.1406e-03, -1.7073e-04, -2.1333e-04, -2.9957e-03, 4.5758e-04,\n -1.3380e-03, 2.5009e-03, 5.4122e-03, 6.8415e-03, 2.3793e-04,\n -4.5983e-03, -2.9453e-03, -5.7655e-04, 2.4966e-03, 4.1889e-03,\n -6.2390e-03, 6.7919e-04, 3.1521e-03, -7.6242e-04, 1.0228e-03,\n -3.7283e-03, 2.9883e-03, -9.7195e-04, 2.6678e-03, -3.0145e-05,\n 3.2596e-03, -1.0855e-02, 1.6982e-04, 1.6946e-03, -8.2407e-03,\n 1.5404e-03, 3.8151e-04, -6.7733e-04, 2.2451e-03, 1.0537e-02,\n 1.2316e-03, -2.3546e-03, -1.0068e-03, 2.5431e-03, -2.0887e-03,\n 3.7325e-03, 3.1804e-03, 1.7812e-03, 4.7140e-04, 3.2608e-03,\n 1.7480e-03, -1.8733e-03, 5.5655e-03, 1.7794e-03, -3.3741e-04,\n -3.8731e-04, -7.5570e-03, -1.8605e-03, 4.6805e-03, 6.9399e-04,\n -7.5959e-03, -6.1124e-04, -7.4626e-03, 2.0356e-03, 2.6502e-04,\n 3.6421e-03, -2.0233e-05, 1.9142e-03, 1.2463e-03, -6.1384e-03,\n -3.3679e-03, 3.2188e-03, 3.1764e-03, -4.7223e-03, 2.9905e-03,\n -1.6764e-03, -1.5089e-03, 1.7080e-03, 2.8376e-05, -1.1281e-03,\n -5.8314e-03, -3.6901e-03, -9.1803e-04, 5.2064e-03, 3.3064e-04,\n 2.2967e-03, -2.2131e-03, 1.1978e-03, -1.6021e-03, -1.1057e-03,\n -4.1175e-03, 5.1464e-03, -9.9526e-04, 1.6474e-03, -3.9670e-03,\n -1.3401e-03, -2.1783e-03, 6.2561e-04, -3.3924e-03, -3.1284e-03,\n 5.3930e-03, -5.3472e-03, -7.5375e-03, -2.0810e-03, 3.8033e-03,\n 2.5937e-03, -2.1525e-03, -2.6872e-03, -4.7715e-03, -7.7848e-05,\n -1.4770e-03, -8.4897e-05, 2.1579e-03, -2.7624e-03, 3.1366e-03,\n 4.6243e-03, 3.1186e-03, 3.0918e-03, 2.0418e-03, -2.0133e-03,\n 1.8991e-03, 7.7032e-04, -3.0663e-03, -2.7248e-03, -5.6349e-03,\n 7.1838e-03, 1.5665e-04, -1.0350e-03, 1.0265e-03, -5.1220e-03,\n 2.0531e-03, 1.9616e-03, -6.8571e-04, -6.3147e-03, -7.6521e-04,\n -3.4404e-04, 3.0887e-03, 3.9217e-03, 4.3005e-03, -3.3023e-03,\n -2.3543e-03, -6.3546e-04, 2.2603e-03, 1.6140e-03, 4.4361e-03,\n 5.9111e-03, -3.5827e-04, -4.3015e-03, -1.2677e-03, -5.3198e-03,\n 2.5576e-03, 1.0168e-03, -3.0957e-04, -4.3073e-04, -1.3525e-03,\n -1.2944e-03, 2.5900e-03, 4.5832e-03, -2.4337e-03, -2.1446e-03,\n 5.0370e-03, 4.9755e-03, 1.8830e-04, -4.8597e-03, -7.4052e-04,\n -6.4605e-03, 4.2243e-04, -1.5008e-03, 3.9887e-04, 2.0528e-03,\n 3.9953e-04, -1.8109e-03, -6.9804e-03, 1.9128e-03, -1.3338e-03,\n -1.5692e-03, -3.3139e-03, -1.9828e-03, -3.4454e-04, -9.0147e-04,\n 4.4259e-03, 2.3460e-03, -2.5229e-03, 2.0350e-03, 1.7167e-04,\n -6.2756e-04, 3.9813e-03, 2.6670e-03, 2.4128e-03, 2.8180e-04,\n -4.2629e-03, 8.0089e-04, 3.9442e-04, -3.0638e-04, 4.0134e-03,\n 1.2239e-03, -1.5313e-03, -3.3815e-03, 7.2087e-03, -2.0010e-03,\n 1.0306e-03, -1.6078e-03, 3.1942e-03, 7.4135e-04, 9.0146e-04,\n -3.6422e-04, -4.7467e-04, -2.6454e-04, 4.7286e-03, -6.1334e-04,\n 3.1836e-03, 9.8611e-04, -2.6580e-03, 4.2775e-04, 6.2486e-03,\n -3.3021e-03, 3.5554e-03, 1.4777e-04, -8.2929e-04, 3.7326e-03,\n 5.3244e-05, -1.2886e-03, 8.8888e-04, 2.1578e-04, -2.2871e-03,\n 1.2848e-04, 8.5944e-05, -2.8067e-03, -4.7514e-03, -2.0380e-03,\n -3.9176e-03, -5.6416e-03, -6.8947e-04, 4.9518e-03, -7.0921e-03,\n 1.4025e-03, 5.6785e-04, -6.2369e-03, 6.5355e-03, 6.9820e-04,\n 1.2637e-03, 5.0896e-03, 2.4483e-03, -5.8608e-04, -2.8507e-03,\n 2.4420e-03, -5.6670e-04, 3.8474e-03, -1.7227e-03, -3.6214e-04,\n 3.0643e-03, 7.0291e-04], device='cuda:0')",
|
| 23 |
+
"exp_avg_sq": "tensor([6.4778e-05, 6.8930e-05, 7.2817e-05, 7.2361e-05, 9.1340e-05, 7.6225e-05,\n 7.8936e-05, 7.0199e-05, 6.6033e-05, 6.3915e-05, 9.0267e-05, 5.8384e-05,\n 8.2390e-05, 6.5914e-05, 8.3250e-05, 9.3731e-05, 6.9873e-05, 5.0148e-05,\n 6.5907e-05, 7.1988e-05, 4.9623e-05, 9.3907e-05, 5.4909e-05, 7.2679e-05,\n 7.6596e-05, 9.1007e-05, 6.9332e-05, 8.8766e-05, 8.2449e-05, 8.5124e-05,\n 6.5848e-05, 1.3090e-04, 8.5626e-05, 1.1024e-04, 8.6421e-05, 7.2192e-05,\n 7.3108e-05, 7.1878e-05, 7.0837e-05, 9.9984e-05, 9.3764e-05, 1.0015e-04,\n 7.7231e-05, 7.6841e-05, 8.4232e-05, 8.0646e-05, 7.1116e-05, 7.9160e-05,\n 7.3270e-05, 8.5335e-05, 8.2843e-05, 8.1515e-05, 8.2305e-05, 7.1665e-05,\n 7.1226e-05, 8.6916e-05, 6.1390e-05, 6.6272e-05, 9.0172e-05, 8.1987e-05,\n 8.6548e-05, 7.6128e-05, 9.7000e-05, 8.2111e-05, 9.3557e-05, 1.2045e-04,\n 6.8723e-05, 6.5061e-05, 6.7247e-05, 6.7742e-05, 8.5528e-05, 7.5615e-05,\n 6.6086e-05, 1.0420e-04, 6.6522e-05, 8.3809e-05, 7.8889e-05, 8.4902e-05,\n 7.4441e-05, 6.1987e-05, 7.2534e-05, 8.9324e-05, 8.7819e-05, 7.0028e-05,\n 1.0770e-04, 7.7608e-05, 9.0224e-05, 8.1195e-05, 6.9452e-05, 8.7250e-05,\n 1.0675e-04, 9.0452e-05, 5.0930e-05, 7.4194e-05, 8.9640e-05, 7.3545e-05,\n 6.7374e-05, 9.3075e-05, 7.5209e-05, 5.2197e-05, 6.9284e-05, 7.2810e-05,\n 6.9370e-05, 6.8823e-05, 5.5769e-05, 6.8998e-05, 9.4245e-05, 6.3593e-05,\n 8.8530e-05, 8.9705e-05, 9.1641e-05, 5.1810e-05, 1.0775e-04, 5.1666e-05,\n 6.7871e-05, 9.1675e-05, 7.2388e-05, 6.1352e-05, 1.1882e-04, 7.0367e-05,\n 7.1932e-05, 8.2974e-05, 8.8327e-05, 7.6756e-05, 1.1051e-04, 8.3251e-05,\n 5.2568e-05, 6.6439e-05, 8.4868e-05, 1.0214e-04, 7.4891e-05, 9.0290e-05,\n 7.0467e-05, 8.1827e-05, 7.0707e-05, 8.7756e-05, 5.8077e-05, 6.3466e-05,\n 8.0926e-05, 1.2537e-04, 8.2467e-05, 5.8116e-05, 6.2327e-05, 7.4773e-05,\n 8.4296e-05, 6.9050e-05, 7.6877e-05, 7.6802e-05, 1.0276e-04, 8.9515e-05,\n 6.6591e-05, 1.1156e-04, 8.4894e-05, 1.0241e-04, 5.5762e-05, 9.8653e-05,\n 5.9072e-05, 9.4799e-05, 8.3639e-05, 5.7579e-05, 6.1630e-05, 7.6750e-05,\n 6.7997e-05, 6.7918e-05, 7.5523e-05, 6.0712e-05, 8.3961e-05, 6.0657e-05,\n 6.9193e-05, 8.0243e-05, 8.4894e-05, 7.8415e-05, 6.8718e-05, 7.6892e-05,\n 7.8360e-05, 6.5164e-05, 7.3896e-05, 5.3240e-05, 9.4785e-05, 5.9362e-05,\n 1.0349e-04, 8.7455e-05, 8.3351e-05, 7.6283e-05, 8.5986e-05, 6.3941e-05,\n 7.4136e-05, 8.2980e-05, 9.0801e-05, 8.5564e-05, 7.6928e-05, 6.5935e-05,\n 6.6861e-05, 7.9157e-05, 7.5741e-05, 7.5991e-05, 8.5949e-05, 5.9480e-05,\n 8.1899e-05, 7.6347e-05, 6.2768e-05, 6.9582e-05, 7.2373e-05, 6.6750e-05,\n 8.8281e-05, 8.5161e-05, 1.1991e-04, 5.4127e-05, 7.2896e-05, 6.8532e-05,\n 8.8434e-05, 8.0866e-05, 9.4338e-05, 6.0796e-05, 5.8480e-05, 7.8620e-05,\n 6.9785e-05, 1.0083e-04, 6.5408e-05, 5.6484e-05, 8.5452e-05, 9.4465e-05,\n 6.1388e-05, 5.3154e-05, 9.4413e-05, 9.3035e-05, 8.8240e-05, 7.8404e-05,\n 8.9316e-05, 8.3281e-05, 6.0305e-05, 5.9281e-05, 7.3400e-05, 7.9748e-05,\n 7.1312e-05, 7.5872e-05, 9.2766e-05, 7.7908e-05, 7.1008e-05, 7.8454e-05,\n 1.0022e-04, 8.0303e-05, 8.5269e-05, 8.7098e-05, 1.0236e-04, 9.3070e-05,\n 8.4632e-05, 6.9639e-05, 7.6597e-05, 8.4146e-05, 9.5226e-05, 9.2729e-05,\n 6.5700e-05, 8.4154e-05, 1.0356e-04, 9.0999e-05, 6.9529e-05, 9.2237e-05,\n 7.8636e-05, 9.1761e-05, 7.5640e-05, 8.7124e-05, 6.4360e-05, 9.8520e-05,\n 7.7250e-05, 6.5979e-05, 8.6946e-05, 8.6644e-05, 9.4409e-05, 7.6740e-05,\n 8.5480e-05, 6.9180e-05, 7.6496e-05, 6.3631e-05, 7.7779e-05, 6.0929e-05,\n 8.4170e-05, 6.1701e-05, 6.6012e-05, 1.1791e-04, 7.5256e-05, 1.0533e-04,\n 8.1057e-05, 9.3197e-05, 7.3821e-05, 6.5775e-05, 6.5703e-05, 7.8570e-05,\n 8.4494e-05, 9.7106e-05, 7.8018e-05, 7.9242e-05, 7.4246e-05, 8.0500e-05,\n 6.2479e-05, 9.7404e-05, 8.5880e-05, 9.1780e-05, 7.3332e-05, 1.0472e-04,\n 6.8994e-05, 6.2115e-05, 7.3042e-05, 7.7083e-05, 1.1029e-04, 7.1237e-05,\n 6.4797e-05, 6.2751e-05, 1.0064e-04, 7.9509e-05, 1.0880e-04, 8.2692e-05,\n 1.0031e-04, 9.9418e-05, 7.1278e-05, 7.0703e-05, 7.9786e-05, 9.5753e-05,\n 7.7248e-05, 5.7183e-05, 9.0929e-05, 6.6065e-05, 6.6642e-05, 8.9777e-05,\n 6.7077e-05, 8.8939e-05, 1.0215e-04, 8.8618e-05, 8.7282e-05, 1.1736e-04,\n 5.8351e-05, 7.8745e-05, 5.5490e-05, 7.0701e-05, 8.7285e-05, 5.0178e-05,\n 6.2019e-05, 6.9398e-05, 7.3027e-05, 6.9083e-05, 8.5803e-05, 8.3323e-05,\n 8.6115e-05, 7.2935e-05, 6.0361e-05, 6.7880e-05, 5.6083e-05, 7.2937e-05,\n 7.5409e-05, 8.2865e-05, 7.6347e-05, 9.4705e-05, 6.6539e-05, 6.0763e-05,\n 6.8314e-05, 1.0283e-04, 8.2259e-05, 6.7123e-05, 5.1264e-05, 7.6025e-05,\n 8.7850e-05, 1.2742e-04, 7.8741e-05, 1.1834e-04, 7.7785e-05, 7.5760e-05,\n 1.0822e-04, 4.7977e-05, 6.7523e-05, 8.6116e-05, 6.0029e-05, 6.4109e-05,\n 6.5107e-05, 7.5348e-05, 8.6149e-05, 9.2829e-05, 5.7826e-05, 8.8211e-05,\n 1.1026e-04, 1.0041e-04, 4.8314e-05, 7.8447e-05, 7.2270e-05, 7.3820e-05,\n 6.3359e-05, 8.6933e-05, 7.3481e-05, 9.9203e-05, 7.1461e-05, 1.0200e-04,\n 6.7338e-05, 7.0756e-05, 5.7504e-05, 7.5816e-05, 8.6807e-05, 7.9380e-05,\n 5.6600e-05, 7.0744e-05, 9.7026e-05, 1.4749e-04, 8.8113e-05, 6.5137e-05,\n 1.0072e-04, 4.8268e-05, 8.4052e-05, 1.0184e-04, 6.2267e-05, 6.5002e-05,\n 6.4862e-05, 8.7360e-05, 6.2456e-05, 6.7248e-05, 7.0133e-05, 8.3633e-05,\n 8.0540e-05, 6.5275e-05, 7.5192e-05, 1.1047e-04, 7.7879e-05, 1.1341e-04,\n 6.4001e-05, 6.4285e-05, 1.0541e-04, 8.4485e-05, 8.5989e-05, 7.1867e-05,\n 6.6135e-05, 7.0508e-05, 5.2325e-05, 8.0112e-05, 7.3580e-05, 9.9173e-05,\n 1.1967e-04, 7.0044e-05, 6.3798e-05, 6.0664e-05, 7.1368e-05, 1.0872e-04,\n 8.1429e-05, 5.8918e-05, 1.1485e-04, 6.7334e-05, 8.0067e-05, 7.6994e-05,\n 9.1628e-05, 7.3013e-05, 6.5114e-05, 5.7044e-05, 7.6302e-05, 7.5283e-05,\n 6.2769e-05, 5.5535e-05, 6.3032e-05, 9.2329e-05, 8.4781e-05, 4.9118e-05,\n 9.7670e-05, 7.2756e-05, 7.2531e-05, 8.2850e-05, 7.0791e-05, 7.7695e-05,\n 8.0708e-05, 6.3668e-05, 9.6037e-05, 6.6317e-05, 6.1993e-05, 8.0809e-05,\n 1.0779e-04, 1.0491e-04, 7.0323e-05, 7.0868e-05, 7.4078e-05, 6.6519e-05,\n 9.2758e-05, 8.8959e-05, 8.7197e-05, 6.1222e-05, 8.2696e-05, 9.4417e-05,\n 8.0743e-05, 7.5376e-05, 8.7228e-05, 1.4586e-04, 7.3078e-05, 8.1471e-05,\n 7.2303e-05, 1.1140e-04, 6.5338e-05, 7.4517e-05, 6.2221e-05, 5.8278e-05,\n 7.5590e-05, 8.2765e-05, 6.7946e-05, 8.0095e-05, 7.8928e-05, 7.5859e-05,\n 9.1486e-05, 5.7707e-05, 6.9164e-05, 8.2736e-05, 6.6431e-05, 7.8592e-05,\n 6.5607e-05, 8.1008e-05, 9.1264e-05, 1.4560e-04, 1.0181e-04, 1.0794e-04,\n 8.5865e-05, 7.1658e-05], device='cuda:0')"
|
| 24 |
+
},
|
| 25 |
+
"4": {
|
| 26 |
+
"step": "tensor(3756.)",
|
| 27 |
+
"exp_avg": "tensor([[-1.8166e-04, -1.4551e-04, 1.5379e-04, ..., -2.4146e-05,\n -2.4896e-05, 3.0513e-06],\n [ 3.7192e-04, 1.0308e-04, 1.4716e-04, ..., -2.4657e-04,\n -3.6889e-04, -4.3552e-04],\n [-5.5728e-04, -1.1507e-04, -1.1001e-04, ..., 3.8197e-04,\n -7.1066e-04, 1.0456e-04],\n ...,\n [-5.3134e-04, -3.2760e-04, -4.7402e-05, ..., 5.1233e-04,\n -5.6859e-04, -4.5190e-05],\n [ 1.3909e-04, -3.0858e-05, 5.0687e-05, ..., 2.1157e-04,\n 4.5776e-04, 2.9802e-04],\n [ 5.9601e-04, 7.6984e-07, 1.3280e-04, ..., -2.3257e-04,\n -1.1097e-04, 5.1589e-04]], device='cuda:0')",
|
| 28 |
+
"exp_avg_sq": "tensor([[8.2731e-07, 5.9595e-07, 6.9423e-07, ..., 5.7473e-07, 7.4810e-07,\n 5.7568e-07],\n [1.5119e-06, 1.1203e-06, 1.1401e-06, ..., 1.6122e-06, 1.5266e-06,\n 1.7077e-06],\n [1.2480e-06, 1.2007e-06, 1.0906e-06, ..., 1.6883e-06, 1.4146e-06,\n 1.2574e-06],\n ...,\n [1.3845e-06, 1.6778e-06, 1.2627e-06, ..., 1.3317e-06, 1.7706e-06,\n 1.1668e-06],\n [1.1621e-06, 1.3618e-06, 1.2718e-06, ..., 1.5193e-06, 1.6794e-06,\n 1.3274e-06],\n [1.5637e-06, 1.5391e-06, 1.2052e-06, ..., 1.6961e-06, 1.6528e-06,\n 1.4305e-06]], device='cuda:0')"
|
| 29 |
+
},
|
| 30 |
+
"5": {
|
| 31 |
+
"step": "tensor(3756.)",
|
| 32 |
+
"exp_avg": "tensor([[ 3.1499e-04, -6.4592e-05, 1.8783e-05, ..., -9.7207e-06,\n 4.0348e-05, 4.0990e-05],\n [ 2.5795e-04, 1.0899e-04, 9.6960e-05, ..., -1.1355e-04,\n -2.3406e-04, -3.6127e-04],\n [-3.5330e-05, -6.3925e-05, -1.1464e-04, ..., 3.0496e-04,\n -3.1659e-04, 4.5647e-05],\n ...,\n [ 1.5368e-04, -2.9553e-04, -1.1519e-04, ..., -7.2532e-05,\n 1.9635e-04, -2.1949e-04],\n [-6.2868e-06, 5.1236e-05, 4.6614e-05, ..., 3.0361e-05,\n 1.4917e-04, -3.8046e-05],\n [-2.1048e-04, 4.0262e-05, -2.8385e-05, ..., 6.5582e-05,\n 4.7009e-05, 1.5673e-04]], device='cuda:0')",
|
| 33 |
+
"exp_avg_sq": "tensor([[1.8257e-07, 1.5002e-07, 1.8231e-07, ..., 1.1647e-07, 2.2682e-07,\n 1.5474e-07],\n [3.6776e-07, 2.9056e-07, 3.5886e-07, ..., 3.6522e-07, 3.8800e-07,\n 4.4745e-07],\n [3.5064e-07, 3.5725e-07, 3.1704e-07, ..., 5.2026e-07, 4.2944e-07,\n 3.6776e-07],\n ...,\n [3.5290e-07, 5.2947e-07, 4.6637e-07, ..., 4.4746e-07, 5.5009e-07,\n 3.3635e-07],\n [3.2450e-07, 3.7283e-07, 3.3587e-07, ..., 4.3587e-07, 4.8314e-07,\n 3.6586e-07],\n [3.1819e-07, 3.9803e-07, 3.7453e-07, ..., 4.2185e-07, 5.0269e-07,\n 6.1920e-07]], device='cuda:0')"
|
| 34 |
+
},
|
| 35 |
+
"6": {
|
| 36 |
+
"step": "tensor(3756.)",
|
| 37 |
+
"exp_avg": "tensor([ 0.0015, -0.0015], device='cuda:0')",
|
| 38 |
+
"exp_avg_sq": "tensor([1.2167e-05, 1.2167e-05], device='cuda:0')"
|
| 39 |
+
}
|
| 40 |
+
},
|
| 41 |
+
"param_groups": [
|
| 42 |
+
{
|
| 43 |
+
"lr": 0.0007940987335200904,
|
| 44 |
+
"name": "shared",
|
| 45 |
+
"betas": [
|
| 46 |
+
0.9,
|
| 47 |
+
0.999
|
| 48 |
+
],
|
| 49 |
+
"eps": 1e-08,
|
| 50 |
+
"weight_decay": 1e-05,
|
| 51 |
+
"amsgrad": false,
|
| 52 |
+
"maximize": false,
|
| 53 |
+
"foreach": null,
|
| 54 |
+
"capturable": false,
|
| 55 |
+
"differentiable": false,
|
| 56 |
+
"fused": null,
|
| 57 |
+
"decoupled_weight_decay": true,
|
| 58 |
+
"initial_lr": 0.001,
|
| 59 |
+
"params": [
|
| 60 |
+
0,
|
| 61 |
+
1,
|
| 62 |
+
2,
|
| 63 |
+
3
|
| 64 |
+
]
|
| 65 |
+
},
|
| 66 |
+
{
|
| 67 |
+
"lr": 0.0007940987335200904,
|
| 68 |
+
"name": "scale_256",
|
| 69 |
+
"betas": [
|
| 70 |
+
0.9,
|
| 71 |
+
0.999
|
| 72 |
+
],
|
| 73 |
+
"eps": 1e-08,
|
| 74 |
+
"weight_decay": 1e-05,
|
| 75 |
+
"amsgrad": false,
|
| 76 |
+
"maximize": false,
|
| 77 |
+
"foreach": null,
|
| 78 |
+
"capturable": false,
|
| 79 |
+
"differentiable": false,
|
| 80 |
+
"fused": null,
|
| 81 |
+
"decoupled_weight_decay": true,
|
| 82 |
+
"initial_lr": 0.001,
|
| 83 |
+
"params": [
|
| 84 |
+
4
|
| 85 |
+
]
|
| 86 |
+
},
|
| 87 |
+
{
|
| 88 |
+
"lr": 0.0007940987335200904,
|
| 89 |
+
"name": "scale_512",
|
| 90 |
+
"betas": [
|
| 91 |
+
0.9,
|
| 92 |
+
0.999
|
| 93 |
+
],
|
| 94 |
+
"eps": 1e-08,
|
| 95 |
+
"weight_decay": 1e-05,
|
| 96 |
+
"amsgrad": false,
|
| 97 |
+
"maximize": false,
|
| 98 |
+
"foreach": null,
|
| 99 |
+
"capturable": false,
|
| 100 |
+
"differentiable": false,
|
| 101 |
+
"fused": null,
|
| 102 |
+
"decoupled_weight_decay": true,
|
| 103 |
+
"initial_lr": 0.001,
|
| 104 |
+
"params": [
|
| 105 |
+
5
|
| 106 |
+
]
|
| 107 |
+
},
|
| 108 |
+
{
|
| 109 |
+
"lr": 0.00039715242044697206,
|
| 110 |
+
"name": "fusion",
|
| 111 |
+
"betas": [
|
| 112 |
+
0.9,
|
| 113 |
+
0.999
|
| 114 |
+
],
|
| 115 |
+
"eps": 1e-08,
|
| 116 |
+
"weight_decay": 1e-05,
|
| 117 |
+
"amsgrad": false,
|
| 118 |
+
"maximize": false,
|
| 119 |
+
"foreach": null,
|
| 120 |
+
"capturable": false,
|
| 121 |
+
"differentiable": false,
|
| 122 |
+
"fused": null,
|
| 123 |
+
"decoupled_weight_decay": true,
|
| 124 |
+
"initial_lr": 0.0005,
|
| 125 |
+
"params": [
|
| 126 |
+
6
|
| 127 |
+
]
|
| 128 |
+
}
|
| 129 |
+
]
|
| 130 |
+
},
|
| 131 |
+
"scheduler_state_dict": {
|
| 132 |
+
"T_0": 10,
|
| 133 |
+
"T_i": 10,
|
| 134 |
+
"T_mult": 2,
|
| 135 |
+
"eta_min": 1e-06,
|
| 136 |
+
"T_cur": 3,
|
| 137 |
+
"base_lrs": [
|
| 138 |
+
0.001,
|
| 139 |
+
0.001,
|
| 140 |
+
0.001,
|
| 141 |
+
0.0005
|
| 142 |
+
],
|
| 143 |
+
"last_epoch": 3,
|
| 144 |
+
"_step_count": 0,
|
| 145 |
+
"_is_initial": false,
|
| 146 |
+
"_get_lr_called_within_step": false,
|
| 147 |
+
"_last_lr": [
|
| 148 |
+
0.0007940987335200904,
|
| 149 |
+
0.0007940987335200904,
|
| 150 |
+
0.0007940987335200904,
|
| 151 |
+
0.00039715242044697206
|
| 152 |
+
]
|
| 153 |
+
},
|
| 154 |
+
"metrics": {
|
| 155 |
+
"best_val_acc": 69.866,
|
| 156 |
+
"best_epoch": 2,
|
| 157 |
+
"scale_accuracies": {
|
| 158 |
+
"256": 69.194,
|
| 159 |
+
"512": 69.844
|
| 160 |
+
},
|
| 161 |
+
"training_history": {
|
| 162 |
+
"epochs": [
|
| 163 |
+
1,
|
| 164 |
+
2,
|
| 165 |
+
3
|
| 166 |
+
],
|
| 167 |
+
"train_loss": [
|
| 168 |
+
5.60248446921571,
|
| 169 |
+
4.156974341351384,
|
| 170 |
+
3.7702821485531595
|
| 171 |
+
],
|
| 172 |
+
"train_acc": [
|
| 173 |
+
63.38018384800733,
|
| 174 |
+
69.48813074329888,
|
| 175 |
+
70.23393515443342
|
| 176 |
+
],
|
| 177 |
+
"val_acc": [
|
| 178 |
+
67.966,
|
| 179 |
+
69.586,
|
| 180 |
+
69.866
|
| 181 |
+
],
|
| 182 |
+
"scale_accs": {
|
| 183 |
+
"256": [
|
| 184 |
+
66.908,
|
| 185 |
+
68.868,
|
| 186 |
+
69.194
|
| 187 |
+
],
|
| 188 |
+
"512": [
|
| 189 |
+
67.774,
|
| 190 |
+
69.268,
|
| 191 |
+
69.844
|
| 192 |
+
]
|
| 193 |
+
},
|
| 194 |
+
"lr": [
|
| 195 |
+
0.0009755527298894294,
|
| 196 |
+
0.0009046039886902864,
|
| 197 |
+
0.0007940987335200904
|
| 198 |
+
]
|
| 199 |
+
}
|
| 200 |
+
},
|
| 201 |
+
"train_config": {
|
| 202 |
+
"name": "david_training",
|
| 203 |
+
"run_id": "20251012_141246",
|
| 204 |
+
"dataset_name": "AbstractPhil/imagenet-clip-features-orderly",
|
| 205 |
+
"model_variant": "clip_vit_laion_b32",
|
| 206 |
+
"num_classes": 1000,
|
| 207 |
+
"preset": "small_fast",
|
| 208 |
+
"custom_config_path": null,
|
| 209 |
+
"num_classes_override": null,
|
| 210 |
+
"use_belly_override": null,
|
| 211 |
+
"belly_expand_override": null,
|
| 212 |
+
"progressive_training_override": false,
|
| 213 |
+
"scale_warmup_epochs_override": null,
|
| 214 |
+
"num_epochs": 10,
|
| 215 |
+
"batch_size": 1024,
|
| 216 |
+
"learning_rate": 0.001,
|
| 217 |
+
"weight_decay": 1e-05,
|
| 218 |
+
"warmup_epochs": 3,
|
| 219 |
+
"use_rose_loss": true,
|
| 220 |
+
"rose_initial_weight": 0.1,
|
| 221 |
+
"rose_max_weight": 0.5,
|
| 222 |
+
"rose_weight_schedule": "adaptive",
|
| 223 |
+
"use_cayley_loss": false,
|
| 224 |
+
"cayley_weight": 0.001,
|
| 225 |
+
"scale_loss_balance": null,
|
| 226 |
+
"use_mixed_precision": true,
|
| 227 |
+
"gradient_clip": 10.0,
|
| 228 |
+
"scheduler_type": "cosine_restarts",
|
| 229 |
+
"min_lr": 1e-06,
|
| 230 |
+
"freeze_strategy": "never",
|
| 231 |
+
"freeze_threshold": 90.0,
|
| 232 |
+
"unfreeze_on_plateau": true,
|
| 233 |
+
"patience": 10,
|
| 234 |
+
"track_gradients": true,
|
| 235 |
+
"gradient_scale_threshold": 1e-05,
|
| 236 |
+
"gradient_scale_multiplier": 10.0,
|
| 237 |
+
"log_interval": 50,
|
| 238 |
+
"val_interval": 1,
|
| 239 |
+
"save_interval": 5,
|
| 240 |
+
"log_fusion_weights": true,
|
| 241 |
+
"log_loss_components": true,
|
| 242 |
+
"save_format": "safetensors",
|
| 243 |
+
"hf_repo": "AbstractPhil/gated-david",
|
| 244 |
+
"upload_to_hub": true,
|
| 245 |
+
"base_dir": "./david_training",
|
| 246 |
+
"num_workers": 10,
|
| 247 |
+
"pin_memory": true,
|
| 248 |
+
"prefetch_factor": 4,
|
| 249 |
+
"persistent_workers": true
|
| 250 |
+
}
|
| 251 |
+
}
|