Update best_model_acc70.47_metadata.json - Run 20251012_141246
Browse files
weights/David-fully_shared-weighted_sum/20251012_141246/best_model_acc70.47_metadata.json
ADDED
|
@@ -0,0 +1,258 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 3,
|
| 3 |
+
"optimizer_state_dict": {
|
| 4 |
+
"state": {
|
| 5 |
+
"0": {
|
| 6 |
+
"step": "tensor(5008.)",
|
| 7 |
+
"exp_avg": "tensor([[-3.0439e-04, -1.2638e-03, 9.0815e-04, ..., 5.5382e-04,\n 3.3799e-05, -3.1934e-04],\n [-2.5792e-04, -2.0910e-03, 7.0001e-04, ..., 2.9106e-04,\n 2.3716e-04, 3.8460e-04],\n [-2.7868e-04, 7.0527e-04, -1.5997e-04, ..., -4.7133e-04,\n 2.2337e-05, 1.8719e-04],\n ...,\n [ 2.4204e-04, -3.9945e-03, -4.4159e-04, ..., 1.1548e-04,\n 2.3947e-04, 2.1437e-04],\n [-1.4930e-03, -4.0168e-04, -2.4823e-04, ..., -7.3704e-04,\n -1.4586e-04, 1.1135e-03],\n [-1.1672e-03, 1.8088e-04, -7.8920e-04, ..., -2.4491e-03,\n -1.0256e-04, -9.7940e-04]], device='cuda:0')",
|
| 8 |
+
"exp_avg_sq": "tensor([[4.1701e-06, 2.7568e-05, 1.2674e-05, ..., 3.0292e-06, 2.1470e-06,\n 2.4044e-06],\n [7.9701e-06, 4.7258e-05, 1.6028e-05, ..., 4.4461e-06, 3.5715e-06,\n 3.5131e-06],\n [3.8127e-06, 2.0071e-05, 1.1649e-05, ..., 2.6304e-06, 2.1596e-06,\n 2.6966e-06],\n ...,\n [9.1427e-06, 4.4013e-05, 2.3539e-05, ..., 8.7201e-06, 3.4374e-06,\n 9.7062e-06],\n [3.5759e-06, 2.5879e-05, 1.5778e-05, ..., 3.1565e-06, 1.7900e-06,\n 2.1601e-06],\n [1.0052e-05, 5.0066e-05, 2.1692e-05, ..., 5.6329e-06, 3.5013e-06,\n 3.4152e-06]], device='cuda:0')"
|
| 9 |
+
},
|
| 10 |
+
"1": {
|
| 11 |
+
"step": "tensor(5008.)",
|
| 12 |
+
"exp_avg": "tensor([ 2.4243e-03, -8.5013e-04, -5.2422e-03, -3.8375e-04, 3.1773e-03,\n -1.9358e-02, -5.8851e-03, -2.2709e-02, -1.8964e-02, 3.8169e-03,\n -4.9320e-03, -1.1352e-02, 1.0686e-02, -1.4502e-02, 1.6617e-02,\n -5.4667e-02, 4.0225e-03, -2.4636e-02, 1.3224e-02, 1.1783e-02,\n -5.7657e-03, -5.7105e-03, 2.9567e-03, -2.4735e-02, -1.4029e-02,\n -9.7400e-03, 2.2377e-02, 2.3119e-02, -1.2840e-02, -4.3134e-03,\n -1.0725e-02, 4.5216e-02, 7.8719e-03, -2.2110e-02, -7.3033e-04,\n -3.6112e-02, -6.1424e-03, 3.9496e-03, -1.0987e-03, -2.1590e-03,\n 9.4075e-03, 1.6307e-02, 9.2102e-03, 2.8749e-03, 1.0137e-02,\n 5.4039e-03, 7.0307e-03, 1.3238e-02, -2.2105e-02, -1.1850e-02,\n 1.9330e-03, -1.3518e-02, -2.3559e-02, 3.4206e-02, 8.3824e-03,\n -1.1959e-02, -5.8444e-03, -2.6135e-02, -1.0795e-02, 1.1920e-02,\n 4.0418e-02, 2.6382e-02, -4.2998e-02, -4.5609e-03, 5.8468e-03,\n 1.3620e-02, 1.2469e-02, 1.8971e-03, 4.4493e-03, -1.2354e-03,\n 1.3073e-03, -1.2886e-03, 6.6212e-04, -2.1047e-02, -8.5861e-03,\n 2.1285e-03, -9.3415e-03, 1.0289e-02, 1.9614e-02, 2.0836e-02,\n -2.5651e-02, 7.1762e-03, -7.0389e-03, 2.7191e-03, -2.3954e-02,\n 2.3363e-03, -3.6812e-02, 1.8846e-02, 2.2274e-02, -3.7874e-03,\n 1.9122e-03, 1.0402e-03, 8.1689e-03, 2.0028e-02, 2.2185e-02,\n 1.9287e-03, 1.1999e-02, 1.3015e-02, -1.1467e-03, -5.1108e-03,\n -2.2392e-03, 1.1830e-03, 1.5531e-02, -1.1133e-02, -6.5686e-03,\n 2.3827e-02, 1.3286e-02, -3.6239e-03, 1.4576e-02, 6.1491e-03,\n 1.0413e-02, -3.3454e-04, 2.5617e-02, -1.3917e-02, 3.3622e-02,\n 2.2787e-03, -9.6399e-03, -2.1922e-02, 2.1205e-02, 9.6859e-03,\n 4.0384e-03, 9.6089e-03, 1.1627e-02, 1.4227e-02, -9.3298e-02,\n -1.4962e-02, -9.8815e-03, -1.9284e-03, -1.1341e-02, -2.8572e-02,\n 2.6365e-02, -8.4033e-03, -1.1267e-02, -4.1134e-03, 1.3227e-02,\n 1.5651e-02, -1.9749e-02, 1.7928e-02, 1.7892e-02, -1.4762e-02,\n 8.2630e-03, 5.2767e-03, -6.6801e-03, -6.4645e-03, 1.1751e-02,\n 1.4294e-02, 2.4533e-02, -1.0547e-02, -5.4050e-04, -1.2102e-02,\n -4.1204e-03, -1.1758e-02, 5.3652e-03, 1.5218e-02, 2.2805e-02,\n 6.7015e-03, 2.6498e-02, 1.6052e-02, 4.8282e-03, 5.3960e-03,\n 1.4863e-02, -1.6281e-02, -1.2958e-02, 2.2301e-02, 2.5019e-02,\n -2.1805e-02, 6.8261e-03, 8.9266e-03, -3.1486e-03, -3.0654e-02,\n 1.2002e-02, 8.8481e-04, 9.5850e-03, 1.8192e-04, -1.4625e-02,\n -1.4924e-02, 3.7741e-02, -5.5797e-03, 2.4373e-02, -5.1456e-02,\n 9.3330e-03, 8.7766e-03, 8.0937e-03, -6.8327e-03, 2.9262e-02,\n 2.6497e-02, -1.7965e-02, -5.6506e-02, 2.5354e-02, -8.5739e-04,\n 1.3334e-02, -1.6831e-02, 6.1886e-03, -2.1549e-02, -2.5818e-02,\n 2.2249e-02, -2.1240e-03, 1.4452e-02, 1.0309e-02, 2.9214e-03,\n -1.8740e-02, 9.4850e-03, -3.1860e-02, 3.8937e-02, -1.9748e-02,\n -1.3102e-02, -1.1588e-02, 6.0613e-03, 3.8787e-03, 1.1830e-02,\n 2.5494e-02, 6.0811e-03, 2.2445e-02, 1.2228e-02, -1.3482e-02,\n -1.0421e-02, -6.1065e-02, -8.2119e-03, -8.9526e-03, 9.9038e-03,\n 2.8217e-02, -9.6497e-03, -1.1520e-02, -5.9738e-03, -2.5734e-02,\n 9.5100e-03, 2.2608e-02, -4.7071e-02, 5.7535e-03, 1.6333e-02,\n -9.3328e-03, -1.4112e-02, 1.7139e-02, -5.0020e-03, 9.8147e-03,\n 1.0538e-02, -2.1752e-03, -4.2427e-04, -4.7802e-03, -1.7064e-02,\n 1.2316e-02, 1.6193e-02, -3.5094e-02, 6.0564e-03, -6.8591e-03,\n 1.1384e-02, -1.5022e-02, -5.7512e-04, -4.8773e-03, 6.8374e-03,\n 1.4703e-03, -3.0438e-03, 1.8145e-02, -8.6099e-03, -9.0361e-03,\n -3.4152e-02, -6.9523e-03, -7.6154e-03, 6.9522e-03, -2.2281e-03,\n 7.5506e-03, 2.6929e-02, -1.4048e-04, -1.2103e-02, -3.2524e-04,\n -1.9230e-02, -8.0186e-03, 1.5169e-02, -7.3955e-03, -4.4947e-03,\n -8.7165e-03, 9.9267e-03, 1.5210e-02, 7.5870e-03, 2.5001e-02,\n -1.1291e-02, 3.1086e-02, 3.9886e-05, 2.7340e-02, 3.1986e-02,\n 5.5108e-03, -1.0107e-03, -1.1321e-02, -1.3377e-02, 5.0625e-03,\n 1.5022e-03, 9.0310e-03, -8.0518e-03, 9.0508e-03, -9.0905e-03,\n 1.5710e-02, 1.4866e-02, 2.2734e-03, 9.5856e-03, -2.5116e-03,\n 5.0736e-03, -2.9051e-02, 1.9267e-02, 2.9432e-02, -1.7091e-02,\n 8.9209e-03, 4.4394e-02, 2.6959e-02, -8.9313e-03, 6.5262e-03,\n -8.0171e-03, 2.7571e-03, -4.7748e-03, 1.3865e-02, -6.5752e-03,\n -9.8687e-02, -2.2751e-02, -1.0735e-02, 1.2652e-02, 1.7228e-02,\n -1.8310e-02, 1.3612e-02, 2.0725e-02, 1.0180e-03, -9.1683e-03,\n -1.6622e-02, 7.0989e-03, 3.1653e-02, -4.7998e-02, 2.7526e-03,\n 4.2457e-02, 2.0658e-03, 2.4072e-03, 3.5495e-03, -4.5000e-03,\n 1.7232e-02, -3.1475e-03, -5.7724e-02, 2.6666e-02, 3.1201e-02,\n 4.6177e-03, -5.3078e-03, 5.8918e-03, -3.9324e-02, 3.8592e-02,\n 7.8600e-03, 8.3587e-03, 1.4978e-02, -2.0945e-02, 2.6487e-02,\n -3.7187e-02, -5.3889e-03, 1.4120e-02, 2.1807e-02, -1.3182e-02,\n 1.0572e-02, -1.4188e-02, 1.2114e-03, 2.2210e-02, -1.1374e-02,\n 6.0740e-03, 5.7350e-03, 9.5004e-03, 5.3072e-03, 1.7179e-03,\n 6.1309e-03, 7.8299e-02, 1.1475e-02, -4.7023e-02, 1.6888e-02,\n -3.9229e-03, 6.2326e-03, -1.3998e-02, 7.5118e-03, -9.0922e-03,\n -1.4320e-02, 2.3053e-02, 9.8167e-03, -1.3675e-03, 3.0142e-03,\n -1.0070e-02, -2.5025e-03, 1.5466e-03, 4.5001e-03, -3.5074e-02,\n -5.7579e-03, -1.0238e-02, -1.0083e-02, 1.5585e-03, 1.0556e-02,\n -8.9796e-03, 2.1423e-02, -2.0190e-03, -9.6638e-03, 1.6917e-02,\n 2.0799e-02, 1.0056e-02, -2.8389e-03, -2.2730e-02, 3.5561e-03,\n 1.7113e-02, 2.5512e-02, 1.8370e-02, -5.2637e-02, 2.1709e-02,\n -2.4716e-03, 4.3185e-03, 2.0641e-02, -6.2556e-03, -1.4137e-02,\n 2.0672e-02, 1.2666e-03, 3.8481e-03, 1.2629e-02, -1.1552e-03,\n -4.6952e-03, -1.8167e-02, 7.6389e-03, 4.3964e-02, 1.9429e-02,\n -1.3789e-02, 5.0313e-03, -4.3294e-02, -3.2675e-03, 3.6612e-03,\n -2.3821e-03, 1.6892e-03, 2.9409e-02, -7.0049e-03, -8.8465e-03,\n 1.9304e-02, 2.8943e-02, 1.7985e-02, -1.1636e-02, 1.9671e-02,\n 1.8276e-02, -1.0311e-02, -2.0016e-02, 9.7860e-03, -1.1220e-03,\n -1.4127e-02, 7.8587e-03, 1.1355e-02, -9.7547e-03, -1.3760e-03,\n -2.0792e-02, 1.0052e-02, 2.5971e-02, 2.8259e-02, 3.3219e-03,\n -1.6401e-02, -2.0638e-03, -2.3205e-02, 1.3103e-02, 1.1499e-02,\n 3.5879e-03, -1.1835e-02, -4.8976e-03, 2.8567e-02, -5.9754e-03,\n 6.9775e-03, 5.3113e-03, -2.6783e-02, -9.8149e-03, 4.2149e-04,\n -5.7762e-03, -1.9819e-02, 7.6192e-04, -6.7934e-03, -5.4587e-02,\n -1.4696e-02, 2.6597e-02, -9.5552e-03, -6.8685e-03, -4.2968e-02,\n 9.8085e-03, 2.3174e-02, 5.9678e-03, -5.7416e-03, -2.3846e-02,\n -6.7874e-03, -2.1284e-02, -5.9179e-03, -1.3977e-02, -2.7261e-02,\n -8.4449e-03, -1.3114e-02, -3.1455e-03, 5.4051e-03, -8.5884e-02,\n 1.6486e-02, -2.1378e-02, 4.9907e-03, 1.5578e-02, -3.6323e-02,\n 1.1104e-03, -3.2952e-03, -5.2101e-04, -1.3572e-03, 3.1171e-05,\n 8.5228e-03, -1.6927e-02, -1.2524e-02, 2.1571e-02, 1.6116e-02,\n -1.3650e-03, 1.8693e-03, -4.8000e-03, -1.3955e-02, -1.1120e-02,\n -1.2644e-02, -1.3413e-02, -2.6868e-02, -9.3948e-03, 3.1542e-02,\n -8.4935e-03, -1.1167e-02], device='cuda:0')",
|
| 13 |
+
"exp_avg_sq": "tensor([0.0035, 0.0052, 0.0029, 0.0027, 0.0039, 0.0056, 0.0021, 0.0046, 0.0053,\n 0.0046, 0.0028, 0.0036, 0.0063, 0.0025, 0.0064, 0.0071, 0.0045, 0.0037,\n 0.0038, 0.0041, 0.0038, 0.0086, 0.0048, 0.0030, 0.0022, 0.0068, 0.0038,\n 0.0052, 0.0056, 0.0049, 0.0038, 0.0099, 0.0032, 0.0091, 0.0038, 0.0070,\n 0.0042, 0.0040, 0.0035, 0.0049, 0.0041, 0.0043, 0.0032, 0.0036, 0.0074,\n 0.0035, 0.0028, 0.0037, 0.0055, 0.0049, 0.0051, 0.0037, 0.0038, 0.0042,\n 0.0036, 0.0035, 0.0043, 0.0022, 0.0027, 0.0038, 0.0058, 0.0030, 0.0063,\n 0.0048, 0.0053, 0.0065, 0.0036, 0.0037, 0.0048, 0.0033, 0.0052, 0.0029,\n 0.0038, 0.0092, 0.0042, 0.0023, 0.0068, 0.0036, 0.0046, 0.0062, 0.0052,\n 0.0058, 0.0030, 0.0066, 0.0046, 0.0052, 0.0038, 0.0069, 0.0067, 0.0035,\n 0.0091, 0.0050, 0.0040, 0.0035, 0.0057, 0.0021, 0.0035, 0.0052, 0.0026,\n 0.0039, 0.0052, 0.0051, 0.0048, 0.0033, 0.0032, 0.0060, 0.0061, 0.0031,\n 0.0048, 0.0055, 0.0049, 0.0051, 0.0064, 0.0033, 0.0045, 0.0026, 0.0024,\n 0.0034, 0.0070, 0.0044, 0.0049, 0.0094, 0.0057, 0.0044, 0.0083, 0.0050,\n 0.0031, 0.0022, 0.0059, 0.0061, 0.0072, 0.0045, 0.0046, 0.0037, 0.0023,\n 0.0014, 0.0031, 0.0048, 0.0018, 0.0064, 0.0048, 0.0040, 0.0055, 0.0037,\n 0.0051, 0.0032, 0.0043, 0.0029, 0.0026, 0.0044, 0.0045, 0.0025, 0.0041,\n 0.0037, 0.0035, 0.0024, 0.0041, 0.0036, 0.0051, 0.0026, 0.0035, 0.0024,\n 0.0030, 0.0033, 0.0040, 0.0034, 0.0026, 0.0031, 0.0059, 0.0070, 0.0057,\n 0.0059, 0.0055, 0.0041, 0.0049, 0.0024, 0.0053, 0.0039, 0.0052, 0.0041,\n 0.0046, 0.0029, 0.0025, 0.0046, 0.0046, 0.0057, 0.0056, 0.0068, 0.0055,\n 0.0051, 0.0038, 0.0036, 0.0046, 0.0041, 0.0043, 0.0078, 0.0024, 0.0036,\n 0.0025, 0.0038, 0.0023, 0.0028, 0.0040, 0.0051, 0.0059, 0.0052, 0.0106,\n 0.0041, 0.0021, 0.0042, 0.0032, 0.0022, 0.0063, 0.0027, 0.0034, 0.0051,\n 0.0048, 0.0043, 0.0025, 0.0051, 0.0035, 0.0048, 0.0060, 0.0052, 0.0053,\n 0.0025, 0.0018, 0.0043, 0.0066, 0.0036, 0.0036, 0.0036, 0.0044, 0.0064,\n 0.0047, 0.0024, 0.0030, 0.0032, 0.0049, 0.0039, 0.0013, 0.0038, 0.0046,\n 0.0047, 0.0048, 0.0034, 0.0036, 0.0035, 0.0037, 0.0058, 0.0016, 0.0031,\n 0.0062, 0.0056, 0.0044, 0.0080, 0.0029, 0.0046, 0.0026, 0.0067, 0.0032,\n 0.0057, 0.0050, 0.0048, 0.0034, 0.0024, 0.0032, 0.0025, 0.0035, 0.0076,\n 0.0046, 0.0035, 0.0041, 0.0068, 0.0052, 0.0021, 0.0036, 0.0039, 0.0060,\n 0.0040, 0.0031, 0.0073, 0.0063, 0.0020, 0.0016, 0.0039, 0.0040, 0.0053,\n 0.0029, 0.0030, 0.0027, 0.0021, 0.0039, 0.0045, 0.0040, 0.0051, 0.0039,\n 0.0018, 0.0044, 0.0084, 0.0040, 0.0024, 0.0033, 0.0039, 0.0051, 0.0064,\n 0.0045, 0.0037, 0.0067, 0.0044, 0.0070, 0.0038, 0.0061, 0.0038, 0.0052,\n 0.0036, 0.0042, 0.0054, 0.0039, 0.0047, 0.0045, 0.0042, 0.0035, 0.0042,\n 0.0035, 0.0065, 0.0045, 0.0031, 0.0038, 0.0046, 0.0037, 0.0049, 0.0036,\n 0.0042, 0.0056, 0.0049, 0.0029, 0.0019, 0.0050, 0.0046, 0.0065, 0.0087,\n 0.0019, 0.0040, 0.0047, 0.0060, 0.0031, 0.0030, 0.0048, 0.0027, 0.0023,\n 0.0057, 0.0036, 0.0029, 0.0037, 0.0035, 0.0025, 0.0036, 0.0047, 0.0030,\n 0.0032, 0.0092, 0.0029, 0.0119, 0.0049, 0.0046, 0.0015, 0.0034, 0.0032,\n 0.0030, 0.0052, 0.0046, 0.0047, 0.0031, 0.0031, 0.0062, 0.0047, 0.0048,\n 0.0068, 0.0048, 0.0034, 0.0017, 0.0025, 0.0024, 0.0042, 0.0089, 0.0052,\n 0.0031, 0.0030, 0.0074, 0.0029, 0.0037, 0.0022, 0.0041, 0.0054, 0.0035,\n 0.0037, 0.0040, 0.0071, 0.0105, 0.0029, 0.0034, 0.0031, 0.0033, 0.0035,\n 0.0077, 0.0038, 0.0022, 0.0040, 0.0043, 0.0035, 0.0033, 0.0024, 0.0055,\n 0.0051, 0.0062, 0.0044, 0.0098, 0.0036, 0.0077, 0.0059, 0.0039, 0.0078,\n 0.0029, 0.0053, 0.0025, 0.0026, 0.0040, 0.0033, 0.0031, 0.0036, 0.0094,\n 0.0069, 0.0036, 0.0042, 0.0031, 0.0069, 0.0078, 0.0034, 0.0039, 0.0064,\n 0.0038, 0.0023, 0.0039, 0.0074, 0.0038, 0.0025, 0.0030, 0.0060, 0.0030,\n 0.0038, 0.0050, 0.0033, 0.0041, 0.0041, 0.0060, 0.0050, 0.0025, 0.0040,\n 0.0038, 0.0024, 0.0032, 0.0033, 0.0045, 0.0066, 0.0030, 0.0034, 0.0084,\n 0.0043, 0.0087, 0.0036, 0.0047, 0.0032, 0.0033, 0.0059, 0.0020, 0.0064,\n 0.0027, 0.0034, 0.0042, 0.0052, 0.0050, 0.0040, 0.0079, 0.0070, 0.0032,\n 0.0034, 0.0078, 0.0053, 0.0044, 0.0029, 0.0050, 0.0033, 0.0037, 0.0038,\n 0.0057, 0.0033, 0.0068, 0.0031, 0.0044, 0.0051, 0.0023, 0.0027, 0.0035,\n 0.0024, 0.0030, 0.0043, 0.0006, 0.0050, 0.0075, 0.0032, 0.0048],\n device='cuda:0')"
|
| 14 |
+
},
|
| 15 |
+
"2": {
|
| 16 |
+
"step": "tensor(5008.)",
|
| 17 |
+
"exp_avg": "tensor([ 1.0760e-03, -1.9924e-04, -1.0585e-03, -3.6501e-04, 1.8805e-03,\n -4.1317e-03, -1.2585e-03, -1.5408e-03, -3.3328e-03, 7.6370e-05,\n -1.4217e-03, -2.2055e-03, 1.3632e-03, -3.4025e-03, 1.9553e-03,\n -7.3807e-03, 4.4939e-04, -3.9161e-03, 1.8200e-03, 2.3310e-03,\n -1.6538e-03, -9.6346e-04, 6.5816e-04, -6.3538e-03, -4.2029e-03,\n -1.2815e-03, 4.0625e-03, 2.7927e-03, -9.4310e-04, -2.3484e-04,\n -5.3390e-04, 3.1583e-03, 1.5606e-03, -3.4860e-03, 6.0764e-04,\n -3.9333e-03, 6.4129e-04, 1.9194e-03, 3.3717e-04, -6.0615e-04,\n 2.1533e-03, 3.0271e-03, 1.6567e-03, 2.7906e-06, 5.6205e-04,\n 2.2313e-03, 2.1865e-03, 2.9951e-03, -3.8193e-03, -1.8873e-03,\n -4.5699e-04, -4.0301e-03, -5.7637e-03, 4.4634e-03, 2.0250e-03,\n -2.0362e-03, -3.2582e-04, -6.9376e-03, -2.5112e-03, 3.4172e-03,\n 4.6652e-03, 7.3805e-03, -5.0588e-03, -3.1703e-04, 1.7010e-03,\n 2.0559e-03, 3.0957e-03, 6.8300e-04, 1.3795e-03, -3.1506e-04,\n 9.8390e-04, 9.9259e-04, -4.8994e-04, -2.4305e-03, -6.5025e-04,\n 2.0417e-03, -7.5291e-04, 3.9514e-03, 2.5827e-03, 2.5108e-03,\n -2.8077e-03, 2.7788e-03, -1.6651e-04, 4.9973e-04, -3.6839e-03,\n -5.5308e-04, -8.0804e-03, 1.3720e-03, 1.2752e-03, -1.8158e-03,\n 7.4181e-04, 2.1030e-03, 9.1909e-04, 4.6448e-03, 2.0664e-03,\n 1.5627e-03, 2.3060e-03, 2.1078e-03, -8.7513e-05, -5.0009e-04,\n -7.1153e-04, 1.3170e-03, 1.2506e-03, -1.8092e-03, -1.4188e-03,\n 2.4026e-03, 1.8485e-03, -7.5610e-04, 2.8564e-03, -5.1572e-04,\n 2.3000e-03, 9.8853e-04, 3.8448e-03, -1.9461e-03, 4.8592e-03,\n 1.5923e-03, -2.9991e-03, -5.1885e-03, 3.2936e-03, 8.4376e-04,\n -3.5436e-05, 1.8326e-03, 1.6482e-03, 3.3899e-03, -7.4999e-03,\n -1.3389e-03, -8.8365e-04, -1.5079e-03, -9.8218e-04, -4.7880e-03,\n 2.6630e-03, -1.6142e-03, -1.4999e-03, -1.6338e-03, 4.5918e-03,\n 1.3893e-02, -2.7300e-03, 3.1242e-03, 1.2378e-02, -3.8089e-04,\n 7.2884e-04, 1.0066e-03, 1.0328e-04, -2.2640e-03, 2.4184e-03,\n 3.2678e-03, 3.9608e-03, -2.0471e-03, 4.0945e-04, -1.8978e-03,\n -8.9517e-04, -4.1060e-03, 1.6608e-03, 5.7546e-03, 3.9770e-03,\n 3.2127e-03, 3.6615e-03, 6.8869e-03, -2.9030e-04, 1.7846e-03,\n 2.1285e-03, -3.8009e-03, -2.2956e-03, 5.8408e-03, 3.7337e-03,\n -4.8652e-03, 3.2431e-03, 2.7328e-03, 1.8817e-04, -3.8184e-03,\n 2.1042e-03, -2.9157e-04, -1.9590e-04, -1.8022e-03, -1.6456e-03,\n -4.2975e-03, 5.3972e-03, -2.6994e-03, 3.0899e-03, -7.5949e-03,\n 2.6107e-03, 2.8993e-03, 2.9146e-03, -1.8592e-03, 4.2277e-03,\n 2.7645e-03, -2.5342e-03, -5.9323e-03, 4.3078e-03, 5.0717e-04,\n 3.0284e-03, -1.7403e-03, 4.4655e-04, -4.9284e-03, -3.6535e-03,\n 2.1565e-03, 7.5630e-04, 6.9602e-04, 3.9993e-03, 3.9049e-04,\n -3.1912e-03, 3.2173e-03, -2.6310e-03, 3.8522e-03, -2.0296e-03,\n -1.6826e-03, -1.7039e-03, 4.8088e-04, 2.2580e-03, 1.5511e-03,\n 7.5667e-03, 1.2902e-03, 2.0092e-03, 3.1987e-03, -4.0342e-03,\n -1.9889e-03, -7.1501e-03, -1.6472e-03, -3.5188e-03, 1.1383e-03,\n 7.2219e-03, -1.1193e-03, -1.6463e-03, -2.6309e-04, -3.6223e-03,\n 4.5149e-03, 1.4193e-02, -7.5970e-03, 1.1809e-04, 3.2096e-03,\n 4.6998e-04, -2.4627e-03, 2.7943e-03, -8.1360e-04, 2.5079e-03,\n 4.3966e-03, 4.3459e-04, 1.3021e-04, -7.4115e-04, -1.6682e-03,\n 1.2128e-02, 4.1821e-03, -5.7264e-03, 9.6836e-04, -8.7097e-04,\n 4.1537e-03, -5.3405e-03, 5.6894e-04, -1.5949e-03, 1.0667e-03,\n 3.8671e-03, -2.0235e-03, 1.5057e-03, -8.3241e-04, -2.2067e-03,\n -4.5485e-03, -2.0256e-03, -2.0713e-03, 3.4733e-03, 5.2437e-04,\n 1.5098e-03, 4.0789e-03, 3.5602e-04, -1.3296e-03, -3.3853e-04,\n -6.2656e-03, -3.3723e-03, 4.4595e-03, -2.0321e-04, -8.2743e-04,\n -1.7756e-03, 2.8006e-03, 2.8960e-03, 1.5568e-04, 3.0431e-03,\n -3.2921e-03, 7.7648e-03, 1.1132e-03, 4.5998e-03, 8.8927e-03,\n 7.2997e-04, -3.3341e-04, -1.3475e-03, -8.6352e-03, 5.3741e-03,\n 3.4103e-04, -2.4627e-04, -1.5709e-04, 2.4570e-03, -1.6053e-03,\n 6.0014e-03, 7.5074e-03, 5.7500e-04, 1.5468e-03, 1.0660e-03,\n 1.8716e-03, -3.7410e-03, 1.3488e-02, 5.8733e-03, -2.4013e-03,\n 2.1345e-03, 1.1800e-02, 5.7752e-03, -3.5541e-04, 2.6520e-03,\n -3.5412e-04, 9.0048e-04, 7.0304e-04, 2.9790e-03, -1.2613e-03,\n -1.0982e-02, -5.9923e-03, -1.0814e-03, 3.1954e-03, 2.6095e-03,\n -2.8608e-03, 3.6323e-03, 3.2709e-03, 6.4025e-04, -9.8489e-04,\n -2.5235e-03, -4.8096e-04, 5.5454e-03, -9.7755e-03, -2.4052e-04,\n 5.3678e-03, 1.4298e-03, 1.3890e-04, 5.8200e-05, 2.7262e-04,\n 3.1338e-03, 7.3380e-05, -6.6152e-03, 4.8547e-03, 7.2005e-03,\n 1.3932e-03, -5.6825e-04, 2.8054e-03, -4.8849e-03, 5.3676e-03,\n 3.9600e-04, 1.3255e-03, 1.0298e-02, -2.7812e-03, 2.2548e-03,\n -3.2902e-03, -3.1466e-04, 2.3151e-03, 3.5020e-03, -2.1404e-03,\n 3.5722e-03, -1.7201e-03, 7.1400e-05, 5.2289e-03, -1.9076e-03,\n 1.6478e-03, 2.9100e-03, 2.6463e-03, -9.3678e-04, 2.3682e-04,\n 4.0577e-03, 8.3947e-03, 3.1041e-03, -5.8881e-03, 1.9997e-03,\n 6.3157e-06, 5.7734e-03, -1.0855e-03, 2.0493e-03, -1.5225e-03,\n -1.0818e-03, 2.3146e-03, 2.2726e-03, -3.3404e-04, 1.3841e-03,\n -1.0460e-04, -3.2271e-04, 2.9169e-04, 1.1957e-03, -3.9265e-03,\n -1.5158e-03, -4.8737e-03, -2.6840e-03, 2.5119e-04, 2.2675e-03,\n -4.3496e-04, 2.1929e-03, -7.5741e-04, -1.2767e-03, 2.2268e-03,\n 4.2693e-03, 1.7566e-03, 2.2998e-05, -2.8282e-03, -1.1140e-05,\n 2.2572e-03, 3.0200e-03, 2.7229e-03, -3.4812e-03, 3.3490e-03,\n 2.2433e-04, 7.5598e-04, 5.4190e-03, -3.6793e-04, -3.9798e-03,\n 1.5115e-03, -1.5924e-03, 1.2935e-03, 2.3063e-03, -4.3544e-04,\n 4.7752e-05, -3.5182e-03, 3.0538e-03, 6.2237e-03, 2.4674e-03,\n -2.9227e-03, 3.5460e-05, -3.5901e-03, -4.5757e-04, 7.3636e-04,\n -1.1571e-04, -4.0782e-04, 3.3397e-03, -1.9265e-03, -4.8521e-04,\n 5.5345e-03, 7.0311e-03, 2.7982e-03, -2.0407e-03, 6.9817e-03,\n 3.1650e-03, -5.3249e-04, -2.8690e-03, 1.2999e-03, 2.1724e-04,\n -2.3800e-03, 1.1447e-03, 2.2322e-03, -3.2777e-03, 1.1536e-04,\n -3.0402e-03, 3.2118e-03, 1.0838e-02, 5.2197e-03, -2.9021e-04,\n -2.3780e-03, 3.2910e-04, -4.3813e-03, 4.2833e-04, 3.2216e-03,\n 3.4763e-04, -2.1738e-03, -7.9203e-04, 5.7498e-03, -2.3314e-03,\n -4.7377e-04, 1.8323e-03, -5.4208e-03, -2.0634e-03, 3.3921e-05,\n 9.5779e-05, -4.8986e-03, 6.5574e-05, -1.9130e-03, -8.1843e-03,\n -2.7711e-03, 5.8904e-03, -8.1747e-04, -3.0500e-03, -4.9923e-03,\n 8.3620e-04, 3.2250e-03, -3.1839e-04, -1.6571e-03, -4.4163e-03,\n -2.2036e-03, -2.2436e-03, -9.6584e-04, -1.8378e-03, -4.0269e-03,\n -1.3929e-03, -1.4195e-03, -7.8772e-04, 4.4017e-04, -1.2614e-02,\n 6.1262e-03, -5.2261e-03, 9.8007e-04, 1.2340e-03, -7.0803e-03,\n 2.4849e-04, -2.3804e-04, 1.9051e-04, -2.0024e-04, 8.7644e-04,\n 9.2664e-04, -3.0737e-03, -1.1227e-03, 5.7350e-03, 2.8291e-03,\n 1.0170e-03, 1.2765e-03, -2.6955e-03, -1.9165e-03, -2.1941e-03,\n -1.7692e-03, -3.2617e-03, 4.2621e-02, -1.2467e-03, 4.3511e-03,\n -2.7755e-03, -9.8040e-05], device='cuda:0')",
|
| 18 |
+
"exp_avg_sq": "tensor([1.4124e-04, 6.5849e-05, 1.3965e-04, 1.0277e-04, 1.3422e-04, 7.9421e-05,\n 2.4171e-04, 4.9901e-05, 4.9325e-05, 7.2099e-05, 2.0670e-04, 6.2177e-05,\n 7.4313e-05, 1.1605e-04, 5.3366e-05, 8.2839e-05, 9.7343e-05, 6.8816e-05,\n 7.5626e-05, 1.0735e-04, 3.9988e-05, 7.6535e-05, 6.4383e-05, 1.6600e-04,\n 2.0832e-04, 8.3585e-05, 1.0410e-04, 9.1864e-05, 9.8470e-05, 9.9075e-05,\n 1.0143e-04, 1.0264e-04, 1.1376e-04, 9.2314e-05, 1.3401e-04, 5.3227e-05,\n 7.3963e-05, 8.5542e-05, 1.2348e-04, 1.5794e-04, 1.6455e-04, 1.2863e-04,\n 1.7065e-04, 9.7924e-05, 5.7260e-05, 1.3875e-04, 1.3454e-04, 1.1495e-04,\n 6.8206e-05, 8.4227e-05, 7.8703e-05, 1.5470e-04, 1.4525e-04, 6.2726e-05,\n 1.3343e-04, 1.6864e-04, 4.6367e-05, 1.8937e-04, 2.4776e-04, 1.6169e-04,\n 8.8166e-05, 1.4949e-04, 1.3876e-04, 8.8524e-05, 1.5268e-04, 1.2840e-04,\n 7.5672e-05, 8.0947e-05, 6.8772e-05, 7.5767e-05, 8.0129e-05, 1.3752e-04,\n 6.5823e-05, 7.7167e-05, 8.4598e-05, 2.1437e-04, 6.9690e-05, 2.4753e-04,\n 6.0546e-05, 5.1684e-05, 9.3439e-05, 1.4443e-04, 1.5326e-04, 4.6653e-05,\n 1.8293e-04, 9.2830e-05, 1.2840e-04, 6.9392e-05, 4.4580e-05, 1.2539e-04,\n 6.9563e-05, 1.2346e-04, 4.9493e-05, 1.0766e-04, 9.5459e-05, 2.3067e-04,\n 8.9814e-05, 7.0676e-05, 1.4116e-04, 4.1317e-05, 4.9896e-05, 8.1196e-05,\n 6.1510e-05, 1.3802e-04, 9.4718e-05, 6.2764e-05, 1.1793e-04, 7.8695e-05,\n 1.0893e-04, 6.8735e-05, 1.2580e-04, 4.7587e-05, 8.1663e-05, 5.7276e-05,\n 7.6966e-05, 2.4486e-04, 2.0611e-04, 9.0651e-05, 1.3050e-04, 1.1275e-04,\n 6.5071e-05, 6.6924e-05, 1.1506e-04, 1.8121e-04, 9.3479e-05, 8.7633e-05,\n 6.3612e-05, 1.5823e-04, 6.6918e-05, 9.7128e-05, 7.2684e-05, 9.4971e-05,\n 8.4115e-05, 1.4676e-04, 1.4543e-04, 6.3450e-04, 7.7229e-05, 7.0468e-05,\n 4.0119e-04, 1.4765e-04, 9.1711e-05, 4.9098e-05, 5.2469e-05, 9.8531e-05,\n 1.6600e-04, 1.1819e-04, 9.0771e-05, 1.5894e-04, 3.9479e-04, 1.3051e-04,\n 7.9641e-05, 5.5951e-04, 1.3235e-04, 1.6343e-04, 6.8394e-05, 3.5854e-04,\n 7.5326e-05, 2.1827e-04, 9.1672e-05, 8.2189e-05, 7.2272e-05, 2.6673e-04,\n 1.3571e-04, 1.1789e-04, 8.4320e-05, 8.3168e-05, 2.6849e-04, 1.1491e-04,\n 6.1910e-05, 8.2218e-05, 6.7806e-05, 5.5781e-05, 6.9184e-05, 1.4116e-04,\n 7.7205e-05, 1.3015e-04, 7.1157e-05, 5.7818e-05, 7.8478e-05, 7.2927e-05,\n 1.5907e-04, 1.5749e-04, 2.6441e-04, 7.1291e-05, 1.2097e-04, 5.3638e-05,\n 9.1276e-05, 6.0862e-05, 1.2157e-04, 8.8698e-05, 1.0560e-04, 5.7009e-05,\n 7.6289e-05, 1.7204e-04, 1.0579e-04, 5.0155e-05, 2.9593e-04, 5.7364e-05,\n 3.1310e-04, 1.0228e-04, 1.2704e-04, 1.2638e-04, 1.4434e-04, 7.3119e-05,\n 7.9660e-05, 7.8461e-05, 9.0183e-05, 5.8831e-05, 2.3788e-04, 5.7610e-05,\n 2.1797e-04, 2.5406e-04, 8.2861e-05, 1.5013e-04, 1.1775e-04, 1.2561e-04,\n 5.9568e-05, 1.8942e-04, 2.0882e-04, 5.1065e-05, 1.6692e-04, 1.6414e-04,\n 5.4666e-05, 5.0593e-05, 1.2256e-04, 3.1038e-04, 3.8497e-04, 7.8244e-05,\n 6.6529e-05, 1.1106e-04, 8.2780e-05, 8.7677e-05, 6.6490e-05, 6.6594e-05,\n 7.1993e-05, 2.1312e-04, 2.1847e-04, 1.2077e-04, 5.2287e-05, 1.0062e-04,\n 7.4146e-04, 1.0988e-04, 1.1709e-04, 1.4612e-04, 1.5341e-04, 1.3174e-04,\n 1.4755e-04, 1.4976e-04, 9.6952e-05, 9.7009e-05, 4.1025e-04, 2.1222e-04,\n 6.4224e-05, 7.6106e-05, 1.4429e-04, 9.7140e-05, 1.6106e-04, 2.0249e-04,\n 1.5364e-04, 1.0738e-04, 1.0043e-04, 1.2144e-04, 5.9498e-05, 1.1413e-04,\n 1.4499e-04, 1.6976e-04, 1.8218e-04, 2.0654e-04, 1.9299e-04, 5.6444e-05,\n 8.9298e-05, 1.4499e-04, 1.8828e-04, 4.7120e-05, 7.3163e-05, 1.3453e-04,\n 1.5986e-04, 7.3800e-05, 5.9458e-05, 2.2744e-04, 8.0506e-05, 9.1567e-05,\n 6.8672e-05, 4.5386e-04, 3.3359e-04, 6.0604e-05, 7.7098e-05, 7.0753e-05,\n 1.4903e-04, 1.9131e-04, 1.8039e-04, 3.5739e-04, 7.9989e-05, 9.1450e-05,\n 7.7840e-05, 1.0502e-04, 8.6503e-05, 4.1384e-04, 1.1977e-04, 9.4591e-05,\n 8.5816e-05, 1.2587e-04, 1.2913e-04, 9.2100e-05, 1.5364e-04, 5.5147e-05,\n 1.1434e-04, 7.2206e-05, 1.1912e-04, 8.9865e-05, 1.3083e-04, 1.3692e-04,\n 8.6461e-05, 1.8539e-04, 6.4203e-05, 1.2036e-04, 9.1853e-05, 9.1266e-05,\n 1.2826e-04, 4.9336e-05, 9.5147e-05, 1.0606e-04, 8.5607e-05, 1.0879e-04,\n 9.0387e-05, 8.0070e-05, 2.2812e-04, 2.2321e-04, 1.5276e-04, 1.8655e-04,\n 7.0936e-05, 9.4679e-05, 5.7680e-05, 9.9155e-05, 1.1065e-04, 3.7480e-05,\n 9.7159e-05, 2.4936e-04, 6.5508e-05, 8.1866e-05, 7.2815e-05, 7.9745e-05,\n 4.0143e-04, 7.8228e-05, 4.8093e-05, 5.8736e-05, 8.2138e-05, 9.8299e-05,\n 8.2835e-05, 2.8774e-04, 1.7804e-04, 7.9997e-05, 8.1695e-05, 9.6544e-05,\n 8.4501e-05, 1.4656e-04, 2.0974e-04, 1.1101e-04, 4.6248e-05, 1.9609e-04,\n 1.9456e-04, 1.1693e-04, 1.1460e-04, 9.6882e-05, 7.6281e-05, 5.0896e-05,\n 8.2161e-04, 4.7396e-05, 1.0336e-04, 2.0838e-04, 7.1173e-05, 5.5470e-05,\n 5.4022e-05, 1.4769e-04, 1.7074e-04, 7.5676e-05, 5.5009e-05, 1.1351e-04,\n 7.5352e-05, 8.1224e-05, 4.2310e-05, 3.6187e-04, 1.8514e-04, 1.2618e-04,\n 7.5935e-05, 5.8082e-05, 5.3598e-05, 2.0704e-04, 1.5869e-04, 8.6110e-05,\n 1.3430e-04, 9.7489e-05, 1.4032e-04, 8.9337e-05, 7.3535e-05, 9.0282e-05,\n 6.6644e-05, 8.7402e-05, 8.4674e-05, 1.4181e-04, 1.6253e-04, 7.6692e-05,\n 1.8118e-04, 5.2446e-05, 1.5685e-04, 7.4415e-05, 8.9577e-05, 1.9657e-04,\n 9.2098e-05, 9.5134e-05, 9.3537e-05, 9.5785e-05, 1.2997e-04, 6.7884e-05,\n 6.8621e-05, 5.3502e-05, 1.2079e-04, 8.0327e-05, 1.0744e-04, 8.6784e-05,\n 5.2469e-05, 8.5863e-05, 7.0892e-05, 2.5145e-04, 8.9643e-05, 1.8638e-04,\n 1.3377e-04, 1.1449e-04, 4.6369e-05, 2.1019e-04, 1.1543e-04, 6.2819e-05,\n 1.2121e-04, 1.0798e-04, 9.6368e-05, 6.6104e-05, 7.1405e-05, 1.0906e-04,\n 1.4803e-04, 8.4382e-05, 1.2217e-04, 9.8784e-05, 2.5779e-04, 9.7755e-05,\n 7.7773e-05, 9.3283e-05, 1.4556e-04, 9.9902e-05, 4.3868e-05, 1.7682e-04,\n 6.8879e-05, 6.5671e-05, 8.5997e-05, 1.3068e-04, 1.4210e-04, 4.7437e-05,\n 1.3316e-04, 1.5329e-04, 1.0883e-04, 1.5712e-04, 1.9201e-04, 1.5604e-04,\n 1.5539e-04, 6.1075e-05, 9.7723e-05, 7.0199e-05, 7.3318e-05, 7.6915e-05,\n 2.1255e-04, 1.0116e-04, 1.0937e-04, 6.7676e-05, 1.3457e-04, 1.0270e-04,\n 8.1325e-05, 3.9137e-04, 7.9392e-05, 8.2452e-05, 1.4229e-04, 1.2877e-04,\n 7.6856e-05, 6.4541e-05, 1.6234e-04, 1.9188e-04, 9.7638e-05, 2.1437e-04,\n 1.4912e-04, 1.0542e-04, 5.7781e-05, 1.0458e-04, 1.0367e-04, 6.1220e-05,\n 1.7313e-04, 1.0941e-04, 8.4274e-05, 1.4432e-04, 2.1253e-04, 5.2812e-05,\n 1.7730e-04, 4.8591e-05, 6.2405e-05, 2.9676e-04, 1.1649e-04, 1.2318e-04,\n 1.2446e-04, 2.0783e-04, 1.8429e-04, 1.2496e-01, 1.0711e-04, 8.2360e-05,\n 1.9484e-04, 8.3154e-05], device='cuda:0')"
|
| 19 |
+
},
|
| 20 |
+
"3": {
|
| 21 |
+
"step": "tensor(5008.)",
|
| 22 |
+
"exp_avg": "tensor([ 9.3995e-04, 5.3493e-04, -9.2469e-05, 7.9293e-04, 1.5494e-03,\n -2.9398e-03, -8.9516e-04, -3.0337e-03, -2.5389e-03, 7.1951e-04,\n -1.1581e-03, -1.9381e-03, 2.1704e-03, -2.9857e-03, 3.0286e-03,\n -7.4560e-03, 4.4365e-04, -3.8586e-03, 2.3106e-03, 2.4502e-03,\n -1.3609e-03, -7.1620e-04, 1.2495e-03, -5.3785e-03, -3.2744e-03,\n -1.3431e-03, 4.0722e-03, 4.3499e-03, -2.0809e-03, -9.9803e-04,\n -1.4977e-03, 6.4351e-03, 2.4206e-03, -3.7451e-03, 7.3609e-04,\n -5.3168e-03, -3.8251e-04, 7.1577e-04, 8.3345e-04, -2.6863e-04,\n 2.1463e-03, 2.9690e-03, 1.9017e-03, 8.1933e-04, 1.4243e-03,\n 1.6495e-03, 1.7887e-03, 2.9858e-03, -3.6765e-03, -2.7388e-03,\n 5.8965e-04, -2.3046e-03, -4.7142e-03, 6.4551e-03, 1.3523e-03,\n -1.8898e-03, -6.9407e-04, -5.3256e-03, -2.3611e-03, 2.6244e-03,\n 6.9907e-03, 5.9308e-03, -6.4750e-03, -6.0156e-04, 1.3847e-03,\n 3.1007e-03, 2.3071e-03, 6.8298e-04, 8.5911e-04, 1.9065e-04,\n 6.2515e-04, 3.7935e-04, -2.1990e-05, -3.0945e-03, -1.0614e-03,\n 1.0077e-03, -8.3223e-04, 2.5223e-03, 3.3869e-03, 3.0351e-03,\n -3.2081e-03, 2.1576e-03, -4.1565e-04, 3.8700e-04, -4.2826e-03,\n 4.6964e-04, -8.3721e-03, 2.5122e-03, 2.9182e-03, -9.7626e-04,\n 8.0074e-04, 6.9284e-04, 1.3955e-03, 4.8382e-03, 3.9510e-03,\n 1.2017e-03, 2.4939e-03, 3.1056e-03, -1.5853e-04, -8.4449e-04,\n 3.1369e-04, 3.8497e-04, 2.9391e-03, -1.9534e-03, -5.0195e-04,\n 3.2240e-03, 2.4815e-03, -9.4458e-04, 3.4231e-03, 1.0539e-03,\n 2.1498e-03, -6.3291e-04, 4.9783e-03, -1.9917e-03, 5.1772e-03,\n 1.3922e-03, -1.7816e-03, -4.0558e-03, 4.1370e-03, 2.5139e-03,\n 3.4585e-04, 1.7201e-03, 2.2200e-03, 3.2070e-03, -1.4173e-02,\n -2.2935e-03, -1.3570e-03, -7.3760e-04, -1.3799e-03, -4.6109e-03,\n 3.3627e-03, -1.2456e-03, -1.2575e-03, -2.9359e-04, 3.0646e-03,\n 6.4006e-03, -2.5721e-03, 3.0454e-03, 6.0609e-03, -1.2892e-03,\n 1.9711e-03, 1.1068e-03, -8.5692e-04, -5.2004e-04, 2.6332e-03,\n 2.6749e-03, 4.5995e-03, -2.3088e-03, -2.9037e-06, -1.7287e-03,\n -6.6841e-04, -3.0889e-03, 1.3581e-03, 3.8288e-03, 4.1425e-03,\n 2.2256e-03, 4.6231e-03, 4.0351e-03, 7.3718e-04, 1.2772e-03,\n 2.5559e-03, -2.7679e-03, -1.8396e-03, 5.1660e-03, 4.4771e-03,\n -3.5873e-03, 2.2983e-03, 1.7264e-03, -4.5878e-04, -5.3942e-03,\n 1.6000e-03, 2.3746e-04, 1.1601e-03, 3.2040e-04, -2.2283e-03,\n -3.0040e-03, 5.5235e-03, -9.1621e-04, 3.7846e-03, -9.0241e-03,\n 2.3917e-03, 2.5384e-03, 3.0034e-03, -1.3303e-03, 5.5763e-03,\n 3.5962e-03, -2.3103e-03, -7.5684e-03, 4.4205e-03, 9.7167e-05,\n 2.9433e-03, -2.7392e-03, 1.1306e-03, -4.5036e-03, -4.0120e-03,\n 3.1670e-03, 3.8540e-04, 2.1480e-03, 2.7352e-03, 4.9127e-04,\n -3.3695e-03, 2.7876e-03, -4.2993e-03, 6.4551e-03, -2.6179e-03,\n -1.8947e-03, -1.7365e-03, 1.1199e-03, 1.6042e-03, 2.1639e-03,\n 5.9779e-03, 1.8356e-03, 3.9251e-03, 2.2803e-03, -1.9082e-03,\n -1.3564e-03, -9.9513e-03, -1.2085e-03, -1.8434e-03, 1.6422e-03,\n 6.3195e-03, -8.1061e-04, -1.2943e-03, -5.8647e-04, -4.5078e-03,\n 2.8416e-03, 7.5427e-03, -7.0448e-03, 6.8060e-04, 3.2456e-03,\n -7.6301e-04, -2.5210e-03, 3.0675e-03, -5.3197e-04, 2.8423e-03,\n 2.6317e-03, -2.0153e-04, 4.2221e-04, -8.5223e-04, -2.4661e-03,\n 5.8843e-03, 3.3626e-03, -6.6645e-03, 1.5920e-03, -1.4581e-03,\n 3.0774e-03, -3.8252e-03, 1.0260e-03, -6.4725e-04, 1.2208e-03,\n 1.9212e-03, -5.0870e-04, 3.2386e-03, -1.1543e-03, -1.5287e-03,\n -4.8898e-03, -1.0672e-03, -1.7036e-03, 2.3553e-03, 2.3559e-04,\n 2.0695e-03, 5.1938e-03, 2.1975e-04, -1.9068e-03, 4.8429e-04,\n -4.1005e-03, -1.5331e-03, 4.5410e-03, -9.5963e-04, -4.4084e-04,\n -1.9203e-03, 2.0167e-03, 2.7930e-03, 1.1424e-03, 4.1286e-03,\n -1.9620e-03, 6.7313e-03, 6.9431e-04, 4.1377e-03, 7.4892e-03,\n 1.0774e-03, -5.8148e-04, -2.0324e-03, -3.4396e-03, 2.5459e-03,\n 2.0739e-04, 1.4052e-03, -8.9298e-04, 2.1601e-03, -1.0888e-03,\n 4.7125e-03, 4.9038e-03, 4.7550e-04, 1.7088e-03, -5.0850e-04,\n 1.1038e-03, -4.6990e-03, 6.1433e-03, 5.9827e-03, -2.3882e-03,\n 1.8379e-03, 1.0338e-02, 5.5797e-03, -5.3412e-04, 1.7067e-03,\n -9.3760e-04, 4.9994e-04, -5.3858e-04, 2.3430e-03, -7.2119e-04,\n -1.6663e-02, -4.6842e-03, -1.3807e-03, 2.8567e-03, 2.7801e-03,\n -2.8160e-03, 3.2939e-03, 3.3099e-03, 6.5704e-04, -1.6222e-03,\n -2.3838e-03, 1.5990e-03, 5.8883e-03, -7.8628e-03, 5.4428e-04,\n 7.0941e-03, 1.2604e-03, 1.2193e-03, 8.2649e-04, -6.5924e-04,\n 3.4451e-03, -5.7147e-04, -1.0282e-02, 4.9728e-03, 7.1955e-03,\n 4.4466e-04, -5.4105e-04, 2.1839e-03, -5.2782e-03, 5.9574e-03,\n 1.2064e-03, 8.7723e-04, 5.3173e-03, -3.8961e-03, 3.6711e-03,\n -4.5890e-03, -6.0709e-04, 2.9456e-03, 3.7873e-03, -2.1091e-03,\n 2.5349e-03, -1.7635e-03, 4.2707e-04, 4.0094e-03, -1.7286e-03,\n 1.8440e-03, 1.6442e-03, 2.1929e-03, 1.0052e-03, -1.6898e-04,\n 1.9221e-03, 1.2520e-02, 3.2543e-03, -6.1463e-03, 3.1407e-03,\n -6.4178e-04, 2.9133e-03, -2.3615e-03, 2.2220e-03, -1.1252e-03,\n -2.0678e-03, 3.2585e-03, 1.7110e-03, 2.0133e-04, 9.9620e-04,\n -8.5631e-04, -6.4947e-04, 1.5038e-04, 1.0388e-03, -5.9933e-03,\n -6.9472e-04, -2.0862e-03, -1.7936e-03, 8.3626e-04, 1.9117e-03,\n -1.1752e-03, 3.9943e-03, 8.6494e-05, -1.1850e-03, 2.4166e-03,\n 4.8612e-03, 1.8702e-03, -6.2417e-05, -3.4200e-03, 4.3606e-04,\n 3.7036e-03, 3.9358e-03, 3.3533e-03, -8.0649e-03, 3.0782e-03,\n 1.9876e-04, 3.6608e-04, 4.9251e-03, -1.0570e-03, -3.1083e-03,\n 3.0692e-03, 1.3297e-04, 1.0655e-03, 2.4066e-03, -1.4128e-04,\n -1.4716e-04, -3.0262e-03, 2.1086e-03, 7.7068e-03, 2.7993e-03,\n -2.0927e-03, 8.5632e-04, -6.7327e-03, -1.2880e-03, 9.0940e-04,\n -2.8788e-04, 2.6240e-04, 4.5101e-03, -1.5800e-03, -6.9475e-04,\n 4.6650e-03, 6.2453e-03, 2.9589e-03, -1.6889e-03, 4.3495e-03,\n 4.0939e-03, -1.4954e-03, -3.7679e-03, 1.8829e-03, 5.0169e-04,\n -2.0377e-03, 1.4705e-03, 1.8068e-03, -7.9973e-04, 2.3428e-04,\n -4.0738e-03, 2.1010e-03, 7.5543e-03, 5.6507e-03, 6.1676e-04,\n -2.6442e-03, 7.3531e-05, -3.9252e-03, 1.2137e-03, 2.1417e-03,\n 1.1458e-04, -1.6244e-03, -6.8217e-04, 5.8181e-03, -6.4444e-04,\n 5.8866e-04, 1.0482e-03, -5.0150e-03, -1.5274e-03, -5.1017e-05,\n -4.3485e-04, -3.4411e-03, 4.5546e-04, -7.5076e-04, -8.7652e-03,\n -2.5562e-03, 5.1971e-03, -1.1175e-03, -1.3917e-03, -5.5715e-03,\n 1.9473e-03, 3.7744e-03, 1.5753e-03, -1.6271e-03, -4.1889e-03,\n -1.2088e-03, -3.4298e-03, -1.2195e-03, -2.3678e-03, -4.4478e-03,\n -1.4865e-03, -2.2742e-03, -3.1901e-04, 1.6735e-03, -1.1667e-02,\n 4.1815e-03, -4.2072e-03, 1.5710e-03, 2.2331e-03, -5.8231e-03,\n 9.7620e-04, -2.3506e-04, 2.1908e-05, 2.1538e-05, 4.6389e-04,\n 1.5950e-03, -3.0704e-03, -1.1162e-03, 5.8593e-03, 2.9577e-03,\n 1.3388e-04, 8.1901e-04, -1.1075e-03, -2.0862e-03, -2.0075e-03,\n -1.8437e-03, -2.4605e-03, 2.5083e-03, -1.0154e-03, 5.0384e-03,\n -1.8101e-03, -1.3158e-03], device='cuda:0')",
|
| 23 |
+
"exp_avg_sq": "tensor([9.9146e-05, 8.4103e-05, 1.0039e-04, 8.6257e-05, 1.2663e-04, 1.0446e-04,\n 1.0169e-04, 8.8746e-05, 8.3938e-05, 8.5618e-05, 1.2869e-04, 7.5735e-05,\n 1.2443e-04, 8.0566e-05, 1.0633e-04, 1.2942e-04, 1.0224e-04, 7.0880e-05,\n 7.8441e-05, 1.0664e-04, 6.5302e-05, 1.3418e-04, 8.5792e-05, 1.1253e-04,\n 1.0326e-04, 1.3577e-04, 9.8272e-05, 1.2573e-04, 1.2265e-04, 1.3149e-04,\n 9.7335e-05, 1.9996e-04, 9.8888e-05, 1.7563e-04, 1.2321e-04, 1.0392e-04,\n 9.6964e-05, 9.5777e-05, 9.7161e-05, 1.6627e-04, 1.3486e-04, 1.3383e-04,\n 1.2315e-04, 9.3076e-05, 1.1186e-04, 1.0889e-04, 9.3798e-05, 1.0045e-04,\n 1.0619e-04, 1.2866e-04, 1.1147e-04, 1.1945e-04, 1.1440e-04, 9.5678e-05,\n 9.4787e-05, 1.1836e-04, 7.1087e-05, 9.5196e-05, 1.2986e-04, 1.1271e-04,\n 1.2095e-04, 1.0130e-04, 1.7030e-04, 1.0608e-04, 1.5571e-04, 1.7216e-04,\n 8.5789e-05, 8.2551e-05, 9.8587e-05, 8.5040e-05, 1.2816e-04, 1.0746e-04,\n 8.6522e-05, 1.4832e-04, 8.9216e-05, 1.0852e-04, 1.1654e-04, 1.1711e-04,\n 9.6246e-05, 8.8854e-05, 1.0200e-04, 1.4503e-04, 1.0571e-04, 1.0267e-04,\n 1.5522e-04, 1.0849e-04, 1.3093e-04, 1.1375e-04, 9.4083e-05, 1.2114e-04,\n 1.4474e-04, 1.2389e-04, 6.8669e-05, 1.0688e-04, 1.3426e-04, 9.8493e-05,\n 9.4314e-05, 1.2040e-04, 9.3013e-05, 6.0467e-05, 8.5556e-05, 1.0312e-04,\n 8.6734e-05, 1.0447e-04, 7.9891e-05, 9.6553e-05, 1.2847e-04, 8.3968e-05,\n 1.2248e-04, 1.1084e-04, 1.3911e-04, 6.8670e-05, 1.4057e-04, 7.3868e-05,\n 8.9940e-05, 1.3079e-04, 1.0924e-04, 9.2835e-05, 1.6816e-04, 1.2380e-04,\n 9.1662e-05, 1.0939e-04, 1.3800e-04, 1.2032e-04, 1.6712e-04, 1.1630e-04,\n 6.9001e-05, 8.3550e-05, 1.0654e-04, 1.4342e-04, 1.0946e-04, 1.2329e-04,\n 1.0052e-04, 1.0421e-04, 7.8709e-05, 1.2022e-04, 7.9804e-05, 9.3144e-05,\n 1.1542e-04, 1.8559e-04, 1.2271e-04, 7.0139e-05, 9.8245e-05, 9.8939e-05,\n 1.4620e-04, 9.1417e-05, 9.8479e-05, 1.0217e-04, 1.6400e-04, 1.2003e-04,\n 9.9455e-05, 1.6394e-04, 1.1391e-04, 1.4349e-04, 7.7520e-05, 1.5171e-04,\n 8.2146e-05, 1.4261e-04, 1.1454e-04, 6.6722e-05, 8.2628e-05, 1.1801e-04,\n 1.0282e-04, 9.7658e-05, 9.7879e-05, 8.1329e-05, 1.2180e-04, 7.9855e-05,\n 8.9305e-05, 1.3202e-04, 1.0882e-04, 1.0390e-04, 9.6872e-05, 1.2120e-04,\n 1.0729e-04, 8.2154e-05, 1.0029e-04, 7.2729e-05, 1.2386e-04, 8.7855e-05,\n 1.5337e-04, 1.1196e-04, 1.1970e-04, 9.3028e-05, 1.2774e-04, 8.6016e-05,\n 1.0545e-04, 1.0335e-04, 1.3223e-04, 1.1775e-04, 1.0830e-04, 7.4462e-05,\n 9.4030e-05, 1.1631e-04, 1.0680e-04, 1.0743e-04, 1.2958e-04, 7.3172e-05,\n 1.1407e-04, 9.3777e-05, 8.2345e-05, 8.9150e-05, 1.0752e-04, 9.7627e-05,\n 1.2778e-04, 1.1534e-04, 1.7801e-04, 7.6890e-05, 1.0104e-04, 8.7324e-05,\n 1.3805e-04, 1.1200e-04, 1.3268e-04, 8.7606e-05, 8.7231e-05, 1.2034e-04,\n 9.9251e-05, 1.4628e-04, 1.0757e-04, 7.3136e-05, 1.1665e-04, 1.3646e-04,\n 8.8382e-05, 7.7234e-05, 1.4046e-04, 1.2008e-04, 1.1798e-04, 9.8332e-05,\n 1.1220e-04, 1.0423e-04, 8.3265e-05, 8.5247e-05, 9.5551e-05, 1.1254e-04,\n 1.0556e-04, 1.0157e-04, 1.3882e-04, 9.9902e-05, 8.9171e-05, 1.0250e-04,\n 1.4240e-04, 1.0860e-04, 1.1644e-04, 1.3189e-04, 1.5158e-04, 1.2011e-04,\n 1.1477e-04, 1.0233e-04, 9.7488e-05, 1.3467e-04, 1.1787e-04, 1.4406e-04,\n 9.6114e-05, 1.1415e-04, 1.2539e-04, 1.3570e-04, 1.0486e-04, 1.5311e-04,\n 1.0146e-04, 1.3224e-04, 1.0222e-04, 1.3572e-04, 8.2841e-05, 1.4813e-04,\n 1.0354e-04, 8.6952e-05, 1.2163e-04, 1.1411e-04, 1.2834e-04, 1.0708e-04,\n 1.1285e-04, 9.9967e-05, 1.3011e-04, 8.3935e-05, 1.0017e-04, 7.5994e-05,\n 1.1165e-04, 8.5254e-05, 1.0013e-04, 1.5794e-04, 8.0569e-05, 1.5641e-04,\n 1.0878e-04, 1.2697e-04, 9.8664e-05, 8.5662e-05, 9.1790e-05, 1.1022e-04,\n 1.0499e-04, 1.2761e-04, 1.1276e-04, 1.1332e-04, 9.3035e-05, 1.0710e-04,\n 8.9329e-05, 1.2958e-04, 1.0110e-04, 1.2248e-04, 1.3040e-04, 1.5238e-04,\n 9.1117e-05, 8.8666e-05, 1.0269e-04, 9.9711e-05, 1.5406e-04, 9.9667e-05,\n 9.9111e-05, 8.5138e-05, 1.5805e-04, 1.0620e-04, 1.7860e-04, 1.2117e-04,\n 1.3142e-04, 1.3165e-04, 9.9516e-05, 1.0433e-04, 9.8781e-05, 1.2666e-04,\n 1.0320e-04, 7.8912e-05, 1.1459e-04, 9.4621e-05, 9.1533e-05, 1.0636e-04,\n 8.8211e-05, 1.2896e-04, 1.4077e-04, 1.3468e-04, 1.3012e-04, 1.5409e-04,\n 7.8091e-05, 1.0577e-04, 7.7208e-05, 1.0288e-04, 1.3335e-04, 6.7926e-05,\n 8.1656e-05, 8.9355e-05, 9.3965e-05, 9.7924e-05, 1.0672e-04, 1.2122e-04,\n 1.2222e-04, 9.7585e-05, 8.0788e-05, 9.2451e-05, 7.2778e-05, 8.8919e-05,\n 9.8765e-05, 1.2398e-04, 1.0120e-04, 1.2315e-04, 8.7318e-05, 7.6194e-05,\n 8.4452e-05, 1.2695e-04, 1.0856e-04, 9.5518e-05, 7.0758e-05, 1.1910e-04,\n 1.1759e-04, 2.0036e-04, 9.6404e-05, 1.8972e-04, 9.9041e-05, 9.3855e-05,\n 1.5090e-04, 6.8762e-05, 9.2345e-05, 1.2132e-04, 7.6584e-05, 8.9249e-05,\n 8.3033e-05, 1.0611e-04, 1.1960e-04, 1.2451e-04, 7.5428e-05, 1.1792e-04,\n 1.5009e-04, 1.1943e-04, 6.1696e-05, 1.0772e-04, 9.4514e-05, 9.1039e-05,\n 8.5732e-05, 1.2514e-04, 1.0165e-04, 1.3277e-04, 9.3500e-05, 1.5441e-04,\n 9.1974e-05, 1.0249e-04, 8.3534e-05, 8.7453e-05, 1.0736e-04, 1.0186e-04,\n 7.8395e-05, 9.0419e-05, 1.4172e-04, 2.1217e-04, 1.0785e-04, 8.1546e-05,\n 1.3602e-04, 5.4616e-05, 1.2020e-04, 1.3123e-04, 9.1332e-05, 9.7466e-05,\n 9.3985e-05, 1.2369e-04, 8.6052e-05, 8.9958e-05, 9.3982e-05, 1.1359e-04,\n 9.5315e-05, 9.3998e-05, 9.6766e-05, 1.5004e-04, 1.0979e-04, 1.6002e-04,\n 8.4518e-05, 8.5139e-05, 1.2923e-04, 1.3394e-04, 1.2078e-04, 9.4645e-05,\n 9.6646e-05, 9.4626e-05, 6.5946e-05, 1.1798e-04, 1.0583e-04, 1.3395e-04,\n 1.6550e-04, 9.8046e-05, 9.5393e-05, 7.6339e-05, 1.0859e-04, 1.5327e-04,\n 1.1449e-04, 8.3778e-05, 1.7394e-04, 9.5075e-05, 1.1703e-04, 1.0878e-04,\n 1.4211e-04, 9.3261e-05, 8.5334e-05, 8.1499e-05, 9.2768e-05, 1.0473e-04,\n 8.2421e-05, 7.5808e-05, 8.5633e-05, 1.2249e-04, 1.2047e-04, 7.3874e-05,\n 1.5171e-04, 9.1562e-05, 9.9637e-05, 1.2266e-04, 9.4611e-05, 1.1457e-04,\n 1.1890e-04, 8.5027e-05, 1.4557e-04, 7.2169e-05, 7.4639e-05, 1.2335e-04,\n 1.5602e-04, 1.5124e-04, 9.7914e-05, 8.9994e-05, 1.0205e-04, 9.6434e-05,\n 1.2734e-04, 1.3004e-04, 1.2516e-04, 7.3189e-05, 1.1713e-04, 1.3054e-04,\n 1.1357e-04, 1.0794e-04, 1.1883e-04, 2.1336e-04, 1.1809e-04, 1.2658e-04,\n 1.0059e-04, 1.4989e-04, 8.3989e-05, 1.0307e-04, 8.7079e-05, 7.9842e-05,\n 1.1982e-04, 1.0120e-04, 9.4825e-05, 1.3270e-04, 1.1248e-04, 1.0099e-04,\n 1.1845e-04, 8.0143e-05, 9.5401e-05, 1.1579e-04, 8.4454e-05, 1.0021e-04,\n 8.7122e-05, 1.2666e-04, 1.3999e-04, 3.5266e-04, 1.3194e-04, 1.5433e-04,\n 1.2398e-04, 9.2838e-05], device='cuda:0')"
|
| 24 |
+
},
|
| 25 |
+
"4": {
|
| 26 |
+
"step": "tensor(5008.)",
|
| 27 |
+
"exp_avg": "tensor([[ 3.3870e-04, 5.7780e-05, 5.8557e-05, ..., 3.0503e-04,\n -2.9286e-04, -6.3647e-05],\n [ 4.4750e-04, -1.4423e-04, 8.3188e-05, ..., -3.1997e-04,\n 5.2137e-05, 4.7515e-04],\n [-1.4607e-04, 1.3980e-04, -7.5376e-05, ..., 1.3429e-03,\n -2.6009e-04, 3.2687e-04],\n ...,\n [ 5.8070e-04, -4.6753e-04, 1.2029e-04, ..., 2.2311e-04,\n 3.6787e-06, -6.7717e-05],\n [-9.4568e-05, -8.4748e-05, 8.9639e-05, ..., -7.9685e-04,\n -1.2736e-04, -2.0490e-04],\n [-1.5651e-05, -1.7649e-04, 2.6340e-04, ..., 3.4075e-04,\n 8.1045e-04, 1.4185e-04]], device='cuda:0')",
|
| 28 |
+
"exp_avg_sq": "tensor([[5.3774e-07, 3.5043e-07, 3.9298e-07, ..., 4.3656e-07, 5.0919e-07,\n 3.9534e-07],\n [9.2913e-07, 6.8370e-07, 6.0700e-07, ..., 1.1959e-06, 1.0253e-06,\n 1.3121e-06],\n [7.8140e-07, 8.2446e-07, 6.3426e-07, ..., 1.3773e-06, 1.0084e-06,\n 9.3031e-07],\n ...,\n [8.7309e-07, 1.2266e-06, 7.0933e-07, ..., 1.0514e-06, 1.1772e-06,\n 8.3726e-07],\n [6.8128e-07, 9.2824e-07, 6.9016e-07, ..., 1.2652e-06, 1.1831e-06,\n 1.0257e-06],\n [1.0231e-06, 1.0769e-06, 6.8512e-07, ..., 1.1817e-06, 1.2245e-06,\n 1.0696e-06]], device='cuda:0')"
|
| 29 |
+
},
|
| 30 |
+
"5": {
|
| 31 |
+
"step": "tensor(5008.)",
|
| 32 |
+
"exp_avg": "tensor([[-1.1409e-04, 8.5129e-07, -4.1707e-05, ..., 3.9551e-05,\n 9.9652e-05, -1.5443e-05],\n [ 3.0692e-04, -1.3729e-04, 3.1686e-05, ..., 1.4986e-05,\n -3.9030e-05, 3.8597e-04],\n [-6.4513e-05, 8.2364e-05, 4.6654e-05, ..., 8.1078e-04,\n 1.3609e-04, 1.9124e-04],\n ...,\n [-9.8166e-05, -1.3892e-04, 8.9868e-05, ..., -1.0231e-04,\n 5.6953e-04, 7.4592e-05],\n [ 3.3890e-04, 1.8838e-04, -1.4125e-04, ..., 8.5272e-05,\n 1.6422e-05, -1.0121e-04],\n [-2.3709e-04, -1.1365e-04, -8.5998e-06, ..., -1.8348e-06,\n -4.2189e-04, -2.5793e-05]], device='cuda:0')",
|
| 33 |
+
"exp_avg_sq": "tensor([[1.2034e-07, 9.5384e-08, 1.1035e-07, ..., 9.6055e-08, 1.6457e-07,\n 1.1007e-07],\n [2.4729e-07, 1.8708e-07, 2.2358e-07, ..., 3.0345e-07, 2.9015e-07,\n 3.8112e-07],\n [2.3220e-07, 2.6291e-07, 2.0407e-07, ..., 4.7933e-07, 3.2345e-07,\n 3.0767e-07],\n ...,\n [2.3517e-07, 4.2812e-07, 3.2028e-07, ..., 3.7980e-07, 4.2314e-07,\n 2.3483e-07],\n [2.3217e-07, 2.7479e-07, 2.0518e-07, ..., 3.6554e-07, 3.8842e-07,\n 3.0151e-07],\n [2.1573e-07, 3.0068e-07, 2.5469e-07, ..., 3.3730e-07, 3.8665e-07,\n 5.1182e-07]], device='cuda:0')"
|
| 34 |
+
},
|
| 35 |
+
"6": {
|
| 36 |
+
"step": "tensor(5008.)",
|
| 37 |
+
"exp_avg": "tensor([ 0.0014, -0.0014], device='cuda:0')",
|
| 38 |
+
"exp_avg_sq": "tensor([1.0264e-05, 1.0264e-05], device='cuda:0')"
|
| 39 |
+
}
|
| 40 |
+
},
|
| 41 |
+
"param_groups": [
|
| 42 |
+
{
|
| 43 |
+
"lr": 0.0006548539886902864,
|
| 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": 0.0006548539886902864,
|
| 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": 0.0006548539886902864,
|
| 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": 0.0003275997400965494,
|
| 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": 4,
|
| 137 |
+
"base_lrs": [
|
| 138 |
+
0.001,
|
| 139 |
+
0.001,
|
| 140 |
+
0.001,
|
| 141 |
+
0.0005
|
| 142 |
+
],
|
| 143 |
+
"last_epoch": 4,
|
| 144 |
+
"_step_count": 0,
|
| 145 |
+
"_is_initial": false,
|
| 146 |
+
"_get_lr_called_within_step": false,
|
| 147 |
+
"_last_lr": [
|
| 148 |
+
0.0006548539886902864,
|
| 149 |
+
0.0006548539886902864,
|
| 150 |
+
0.0006548539886902864,
|
| 151 |
+
0.0003275997400965494
|
| 152 |
+
]
|
| 153 |
+
},
|
| 154 |
+
"metrics": {
|
| 155 |
+
"best_val_acc": 70.47,
|
| 156 |
+
"best_epoch": 3,
|
| 157 |
+
"scale_accuracies": {
|
| 158 |
+
"256": 69.78,
|
| 159 |
+
"512": 70.366
|
| 160 |
+
},
|
| 161 |
+
"training_history": {
|
| 162 |
+
"epochs": [
|
| 163 |
+
1,
|
| 164 |
+
2,
|
| 165 |
+
3,
|
| 166 |
+
4
|
| 167 |
+
],
|
| 168 |
+
"train_loss": [
|
| 169 |
+
5.60248446921571,
|
| 170 |
+
4.156974341351384,
|
| 171 |
+
3.7702821485531595,
|
| 172 |
+
3.570641661223512
|
| 173 |
+
],
|
| 174 |
+
"train_acc": [
|
| 175 |
+
63.38018384800733,
|
| 176 |
+
69.48813074329888,
|
| 177 |
+
70.23393515443342,
|
| 178 |
+
70.76774534467404
|
| 179 |
+
],
|
| 180 |
+
"val_acc": [
|
| 181 |
+
67.966,
|
| 182 |
+
69.586,
|
| 183 |
+
69.866,
|
| 184 |
+
70.47
|
| 185 |
+
],
|
| 186 |
+
"scale_accs": {
|
| 187 |
+
"256": [
|
| 188 |
+
66.908,
|
| 189 |
+
68.868,
|
| 190 |
+
69.194,
|
| 191 |
+
69.78
|
| 192 |
+
],
|
| 193 |
+
"512": [
|
| 194 |
+
67.774,
|
| 195 |
+
69.268,
|
| 196 |
+
69.844,
|
| 197 |
+
70.366
|
| 198 |
+
]
|
| 199 |
+
},
|
| 200 |
+
"lr": [
|
| 201 |
+
0.0009755527298894294,
|
| 202 |
+
0.0009046039886902864,
|
| 203 |
+
0.0007940987335200904,
|
| 204 |
+
0.0006548539886902864
|
| 205 |
+
]
|
| 206 |
+
}
|
| 207 |
+
},
|
| 208 |
+
"train_config": {
|
| 209 |
+
"name": "david_training",
|
| 210 |
+
"run_id": "20251012_141246",
|
| 211 |
+
"dataset_name": "AbstractPhil/imagenet-clip-features-orderly",
|
| 212 |
+
"model_variant": "clip_vit_laion_b32",
|
| 213 |
+
"num_classes": 1000,
|
| 214 |
+
"preset": "small_fast",
|
| 215 |
+
"custom_config_path": null,
|
| 216 |
+
"num_classes_override": null,
|
| 217 |
+
"use_belly_override": null,
|
| 218 |
+
"belly_expand_override": null,
|
| 219 |
+
"progressive_training_override": false,
|
| 220 |
+
"scale_warmup_epochs_override": null,
|
| 221 |
+
"num_epochs": 10,
|
| 222 |
+
"batch_size": 1024,
|
| 223 |
+
"learning_rate": 0.001,
|
| 224 |
+
"weight_decay": 1e-05,
|
| 225 |
+
"warmup_epochs": 3,
|
| 226 |
+
"use_rose_loss": true,
|
| 227 |
+
"rose_initial_weight": 0.1,
|
| 228 |
+
"rose_max_weight": 0.5,
|
| 229 |
+
"rose_weight_schedule": "adaptive",
|
| 230 |
+
"use_cayley_loss": false,
|
| 231 |
+
"cayley_weight": 0.001,
|
| 232 |
+
"scale_loss_balance": null,
|
| 233 |
+
"use_mixed_precision": true,
|
| 234 |
+
"gradient_clip": 10.0,
|
| 235 |
+
"scheduler_type": "cosine_restarts",
|
| 236 |
+
"min_lr": 1e-06,
|
| 237 |
+
"freeze_strategy": "never",
|
| 238 |
+
"freeze_threshold": 90.0,
|
| 239 |
+
"unfreeze_on_plateau": true,
|
| 240 |
+
"patience": 10,
|
| 241 |
+
"track_gradients": true,
|
| 242 |
+
"gradient_scale_threshold": 1e-05,
|
| 243 |
+
"gradient_scale_multiplier": 10.0,
|
| 244 |
+
"log_interval": 50,
|
| 245 |
+
"val_interval": 1,
|
| 246 |
+
"save_interval": 5,
|
| 247 |
+
"log_fusion_weights": true,
|
| 248 |
+
"log_loss_components": true,
|
| 249 |
+
"save_format": "safetensors",
|
| 250 |
+
"hf_repo": "AbstractPhil/gated-david",
|
| 251 |
+
"upload_to_hub": true,
|
| 252 |
+
"base_dir": "./david_training",
|
| 253 |
+
"num_workers": 10,
|
| 254 |
+
"pin_memory": true,
|
| 255 |
+
"prefetch_factor": 4,
|
| 256 |
+
"persistent_workers": true
|
| 257 |
+
}
|
| 258 |
+
}
|