| { | |
| "epoch": 4, | |
| "optimizer_state_dict": { | |
| "state": { | |
| "0": { | |
| "step": "tensor(6260.)", | |
| "exp_avg": "tensor([[-5.8938e-05, -1.0730e-04, 8.0182e-05, ..., -1.9117e-05,\n -1.8629e-04, 2.5552e-04],\n [ 3.2350e-04, 1.0969e-04, -6.3806e-04, ..., 1.6798e-04,\n -4.1147e-05, -2.7883e-04],\n [ 4.8186e-06, -4.3921e-04, -7.6474e-05, ..., -7.4579e-05,\n -6.5048e-05, -1.3001e-04],\n ...,\n [ 2.7341e-04, -1.3374e-04, 2.2510e-04, ..., 2.2098e-04,\n -4.7319e-05, -2.0046e-04],\n [-8.0803e-06, 7.6843e-05, -5.8125e-05, ..., 4.5216e-05,\n -3.3094e-05, 4.5730e-05],\n [-9.5892e-05, -1.7754e-04, 1.8862e-05, ..., -1.1035e-04,\n -5.7614e-05, -3.2277e-05]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[5.2952e-07, 8.9589e-07, 4.7349e-07, ..., 3.7375e-07, 3.3526e-07,\n 2.5569e-07],\n [3.1020e-07, 3.1433e-07, 3.8827e-07, ..., 1.8587e-07, 2.0136e-07,\n 1.9000e-07],\n [3.9126e-07, 4.0905e-07, 3.1031e-07, ..., 2.2411e-07, 2.3018e-07,\n 1.7189e-07],\n ...,\n [5.5864e-07, 3.5190e-07, 2.3949e-07, ..., 2.9737e-07, 2.4339e-07,\n 2.7539e-07],\n [2.6327e-07, 2.4835e-07, 1.6827e-07, ..., 1.9028e-07, 1.5703e-07,\n 1.1832e-07],\n [5.6091e-07, 6.0792e-07, 2.7790e-07, ..., 3.2537e-07, 3.2751e-07,\n 2.7770e-07]], device='cuda:0')" | |
| }, | |
| "1": { | |
| "step": "tensor(6260.)", | |
| "exp_avg": "tensor([-0.0068, 0.0158, -0.0071, ..., 0.0028, 0.0084, 0.0024],\n device='cuda:0')", | |
| "exp_avg_sq": "tensor([0.0008, 0.0005, 0.0005, ..., 0.0006, 0.0004, 0.0007], device='cuda:0')" | |
| }, | |
| "2": { | |
| "step": "tensor(6260.)", | |
| "exp_avg": "tensor([[-9.4157e-05, 6.8343e-05, 8.5860e-06, ..., -2.0608e-04,\n -2.9644e-05, 3.2769e-05],\n [-2.3410e-04, 3.2868e-05, -3.0236e-05, ..., 5.1995e-06,\n -4.8005e-05, -4.3380e-05],\n [ 1.7437e-05, 4.0101e-07, 4.1894e-05, ..., 2.4205e-06,\n -3.4673e-07, -1.8191e-06],\n ...,\n [-3.3203e-06, -7.0877e-05, 2.5506e-06, ..., -5.3619e-06,\n 2.0255e-06, 7.6671e-07],\n [ 4.6797e-06, 3.2860e-05, -2.6168e-05, ..., 1.9320e-05,\n -3.5511e-05, -3.0347e-06],\n [-5.4684e-05, 1.4414e-08, -4.5865e-08, ..., -7.0358e-05,\n 2.6544e-08, 2.2381e-06]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[3.9153e-07, 4.8874e-07, 2.1018e-08, ..., 3.1086e-07, 6.0362e-08,\n 1.1412e-07],\n [4.6134e-07, 2.6121e-07, 1.2464e-07, ..., 4.4007e-07, 5.4885e-08,\n 1.6479e-07],\n [1.3269e-09, 8.7719e-10, 5.4821e-09, ..., 7.8799e-10, 6.7528e-11,\n 4.6056e-09],\n ...,\n [1.8251e-09, 2.0045e-08, 2.6827e-09, ..., 3.6975e-08, 1.9177e-09,\n 3.7644e-09],\n [1.1493e-08, 8.4289e-09, 1.9244e-08, ..., 3.2403e-08, 7.5397e-08,\n 1.8727e-08],\n [7.9751e-09, 3.7165e-09, 2.8975e-10, ..., 1.8166e-07, 1.2682e-09,\n 9.6544e-09]], device='cuda:0')" | |
| }, | |
| "3": { | |
| "step": "tensor(6260.)", | |
| "exp_avg": "tensor([-9.8131e-03, -1.5546e-02, -1.6927e-03, 6.9462e-03, -6.7657e-03,\n 1.7916e-02, 5.6335e-03, -5.6690e-03, 1.5700e-02, -3.1653e-04,\n 5.8696e-03, 7.6766e-03, 2.3932e-03, -6.8757e-03, -2.3732e-03,\n -4.2066e-03, -3.3856e-03, 1.0301e-02, -1.8936e-02, -1.0456e-03,\n 2.5144e-03, 6.4290e-03, 3.0678e-03, -2.7091e-02, 6.0622e-03,\n 1.0306e-02, 1.1041e-02, 1.1133e-02, 3.4427e-03, 2.5804e-03,\n -2.0072e-02, -3.0521e-03, -1.6293e-02, 6.9154e-03, -3.9660e-03,\n -7.5013e-03, 3.8696e-03, -3.8856e-03, -5.0542e-03, 6.7389e-03,\n -6.2442e-05, -4.6184e-03, 1.0267e-02, 5.6052e-45, 1.6012e-02,\n -5.5841e-03, -8.5813e-03, -1.4302e-03, -1.2584e-02, 1.2870e-02,\n 1.3603e-02, 2.1711e-02, -1.9012e-02, 2.3844e-02, 1.3757e-02,\n 3.8144e-02, 1.9998e-02, 2.0947e-02, 1.7590e-02, 1.4222e-02,\n 7.2729e-03, -8.6973e-03, -3.3344e-02, 1.8164e-02, -1.6577e-02,\n -8.4603e-03, 2.8164e-03, 1.5063e-03, -7.1452e-04, 3.3825e-03,\n -1.0001e-03, 1.3101e-03, -8.2617e-03, 8.8901e-03, 1.1129e-02,\n 1.3797e-12, 9.3750e-03, -4.2928e-03, 1.1533e-02, 6.8859e-03,\n -7.8826e-03, 5.0383e-04, -3.6035e-02, -1.4855e-02, -8.3759e-03,\n -3.9742e-03, 2.4363e-03, 2.2695e-02, 2.4134e-02, -6.2095e-03,\n -6.3404e-03, 9.0265e-03, 9.5040e-03, -3.0083e-03, -1.4026e-02,\n 1.6580e-02, -3.7338e-03, 3.6342e-02, -1.4849e-03, 1.1814e-02,\n 6.2800e-03, -2.5046e-02, 3.4333e-03, 5.6052e-45, -1.7817e-02,\n 1.7998e-02, -1.1494e-02, -6.4829e-03, 2.6398e-03, 7.7073e-03,\n 8.5199e-35, 1.2626e-02, -8.6576e-04, -4.2440e-03, 2.9013e-02,\n 1.2098e-02, 1.4869e-02, -4.8896e-03, -7.9741e-03, -2.2373e-03,\n -1.5504e-03, -1.6189e-02, -1.1165e-02, -1.7134e-03, -1.9570e-03,\n 1.8197e-02, 8.6582e-03, -9.3705e-03, -1.3550e-02, 1.1731e-02,\n -4.5059e-03, 7.4930e-03, -9.3355e-03, 1.6815e-03, 2.9815e-03,\n 1.7771e-02, -1.2413e-02, 1.0264e-02, 4.5959e-03, -8.5827e-03,\n 1.4805e-02, 4.1681e-15, 5.0720e-03, 1.8000e-02, -7.8543e-07,\n -6.8710e-03, -1.5761e-04, -3.6113e-03, 9.6119e-03, 1.2898e-02,\n 1.1325e-02, -4.2546e-03, -1.8522e-02, 8.4032e-03, 2.1966e-03,\n -9.7267e-03, 5.6052e-45, 1.3146e-02, 5.8783e-03, 6.1296e-04,\n 5.6381e-03, -8.4520e-03, -4.9639e-03, -9.5616e-03, 1.6231e-03,\n -5.4163e-03, -2.9891e-02, 4.5783e-03, 2.4962e-02, 1.5597e-02,\n -1.4449e-02, -7.7031e-03, 4.7943e-04, -8.0656e-03, 4.0757e-03,\n 1.9952e-02, 1.5394e-02, -4.4500e-03, -3.4570e-03, -7.8593e-03,\n -5.8641e-03, -1.0263e-02, 2.8176e-03, 6.9769e-03, 2.6267e-03,\n -7.4955e-03, 2.0979e-03, -9.8697e-03, -1.5688e-03, -1.6755e-02,\n 4.8768e-03, -1.8059e-03, -3.7789e-03, 9.7125e-03, 9.8311e-03,\n -8.6778e-03, -4.3296e-03, -2.5171e-03, 1.7599e-02, -2.1603e-03,\n 3.8856e-04, 3.3987e-03, 1.5440e-03, 6.9254e-03, 2.2880e-02,\n 5.6141e-03, 1.2790e-02, 1.1084e-02, -1.4145e-02, 1.1108e-02,\n 4.2465e-04, -1.5820e-04, -3.3379e-03, -2.6639e-02, -1.1472e-03,\n -1.5200e-03, 1.1512e-02, 2.7139e-04, 1.6471e-02, -1.1445e-02,\n 3.0807e-03, 5.6052e-45, 1.2529e-02, -4.9045e-03, -4.2989e-03,\n -7.4669e-03, -2.0794e-03, -5.0865e-03, -2.4981e-03, 9.2352e-03,\n 9.7226e-03, -1.1576e-02, -8.7490e-03, 8.1402e-03, -6.9570e-03,\n 8.9664e-03, 1.4908e-02, -4.9196e-03, -8.2108e-03, 1.4688e-02,\n -4.4439e-03, -1.8461e-03, 3.8516e-03, 4.5827e-03, 2.1046e-02,\n -4.4964e-03, 3.5546e-03, -1.0179e-02, -7.5028e-03, 2.3459e-02,\n -5.9884e-03, -8.6088e-03, -4.4504e-03, 1.5183e-04, -1.5113e-02,\n 1.2801e-02, -8.7578e-03, 3.3260e-04, -5.5654e-03, -1.3499e-02,\n 5.6052e-45, -6.1953e-04, -2.5939e-03, 3.8372e-03, -5.1841e-03,\n 3.7719e-03, -1.5203e-02, -9.2531e-03, 1.2689e-02, -1.3146e-02,\n 5.7281e-03, 1.0027e-02, -8.1972e-03, -2.0711e-02, -4.6079e-03,\n -1.6652e-02, -3.5226e-03, 1.2232e-02, 6.0470e-03, 2.3759e-03,\n 6.5004e-03, 1.3729e-02, 2.2350e-03, 3.2432e-03, -5.0167e-03,\n -1.1818e-02, -3.0393e-02, 1.6854e-02, -1.8319e-03, -1.0238e-02,\n 1.2304e-02, -2.0425e-02, -4.6732e-03, 1.0474e-02, -2.4947e-02,\n 8.7277e-03, -1.0153e-03, -7.3858e-03, 5.6052e-45, -3.8278e-03,\n -9.3566e-04, -3.4597e-03, -1.8282e-03, -1.6436e-02, 1.2313e-02,\n -1.6268e-02, 2.5583e-03, 4.7070e-03, -1.1531e-02, 1.3627e-02,\n 5.0420e-03, -1.4651e-02, -1.2063e-02, 1.5515e-02, 1.2089e-03,\n 1.7735e-02, 1.3702e-02, -4.7914e-02, 5.7573e-03, -1.2561e-02,\n -9.3551e-03, -1.7075e-02, 6.9045e-03, -5.6243e-03, 3.6487e-02,\n 5.6052e-45, 5.6052e-45, -1.9819e-02, 3.5777e-03, -1.0864e-02,\n -4.0384e-03, -9.9724e-05, 1.6338e-02, 2.6230e-02, 1.7074e-03,\n 7.3622e-03, 6.1570e-03, 1.5659e-03, 3.1976e-03, 1.7096e-02,\n -4.2093e-03, 9.0569e-03, -1.4553e-02, 2.0056e-02, -7.6002e-03,\n 2.3456e-03, 2.6388e-02, -4.2826e-04, -1.2056e-03, 1.6909e-06,\n -3.3161e-03, 2.7300e-03, -9.4239e-03, -1.6052e-03, 5.6052e-45,\n 2.5109e-03, -1.0912e-03, -9.7342e-03, 1.9687e-03, 3.6726e-03,\n 4.9191e-02, 6.1480e-03, -2.0142e-02, 1.7908e-02, -1.0032e-02,\n -2.4246e-03, -1.3837e-03, 1.8261e-02, 4.2247e-03, 6.6992e-03,\n -1.7524e-02, 3.1055e-03, -4.1805e-03, 1.8453e-03, 2.8205e-03,\n -1.6907e-03, 8.1097e-03, -4.1872e-03, -1.2711e-02, -8.5391e-03,\n 1.4167e-02, 3.6070e-03, -2.0300e-02, 2.5453e-02, 1.2457e-02,\n 8.1677e-03, 1.2991e-02, -1.2082e-02, -4.6532e-03, -3.2704e-03,\n -2.4762e-02, 6.1068e-04, 5.3949e-03, 5.0209e-03, 8.8213e-03,\n 5.8303e-03, 1.7247e-02, -9.2788e-04, -3.0314e-03, -2.0920e-03,\n 1.5846e-02, -1.0852e-02, -1.0871e-02, 1.4993e-02, -4.4597e-04,\n 7.0044e-04, 7.3132e-05, -4.3138e-03, -1.8233e-02, 4.5378e-04,\n -1.0286e-02, 2.1716e-02, -4.9681e-03, -1.9790e-02, -5.0931e-03,\n 1.5898e-02, 4.7825e-03, -1.4644e-02, -5.5929e-02, 6.1009e-03,\n -4.8571e-03, 4.1758e-03, 1.3805e-02, -1.0051e-02, 3.9756e-03,\n 1.2315e-03, 5.0062e-03, 1.0427e-02, -2.9244e-02, -1.1986e-02,\n -9.6503e-03, -2.2894e-02, -1.9892e-02, 1.8624e-26, 9.5737e-03,\n -2.0759e-02, -2.3902e-05, 5.7899e-03, 1.2006e-02, 1.1891e-02,\n 1.0911e-03, 1.2771e-02, 2.3805e-03, 3.2363e-03, 2.6912e-03,\n -9.7566e-04, -2.8401e-02, 4.6579e-03, 1.1159e-02, 4.3149e-03,\n 1.3615e-02, 3.4267e-03, 8.7986e-03, 1.9150e-03, 1.2428e-02,\n 8.0887e-04, -9.9660e-03, -6.3126e-03, 1.1521e-02, -8.8777e-03,\n 3.2697e-03, 2.3531e-02, 1.5991e-03, 9.0705e-04, 1.1142e-02,\n -8.1575e-03, -2.1980e-02, 1.0534e-02, 1.5299e-02, -2.1538e-02,\n -5.8132e-04, 1.3116e-42, -6.9623e-03, 1.9060e-02, -5.1298e-03,\n -4.4228e-03, -5.5474e-03, 2.1613e-04, -4.2964e-03, 5.2568e-03,\n 6.2892e-03, 6.5574e-03, -1.1350e-02, 2.7090e-03, 4.9994e-03,\n -1.2527e-02, -8.9306e-03, 5.6052e-45, -2.3949e-03, 5.5749e-04,\n -3.7454e-03, -8.7451e-04, 1.4319e-02, 7.5584e-03, -8.6649e-03,\n -7.5471e-03, 1.5498e-02, -5.1384e-03, 2.6283e-03, 5.6052e-45,\n -1.4437e-03, -9.8281e-04, 5.6052e-45, -1.6984e-02, 1.3427e-02,\n 1.8984e-03, 8.9322e-03, 6.9099e-03, 1.0841e-02, 4.3314e-03,\n 6.9923e-03, 4.4178e-03], device='cuda:0')", | |
| "exp_avg_sq": "tensor([2.6125e-03, 2.1868e-03, 5.0719e-05, 2.0078e-03, 1.0132e-03, 2.0049e-03,\n 2.1303e-03, 1.9892e-03, 2.2660e-03, 2.1666e-03, 9.2473e-04, 2.6591e-03,\n 8.8785e-04, 1.9365e-03, 2.2825e-03, 6.0448e-04, 1.4993e-03, 2.6318e-03,\n 1.4362e-03, 1.0068e-03, 1.6717e-03, 2.3392e-03, 1.4478e-03, 2.3933e-03,\n 2.3633e-03, 6.0168e-04, 2.3430e-03, 2.0100e-03, 2.3132e-03, 2.2524e-03,\n 1.8728e-03, 2.1110e-03, 1.9248e-03, 1.7096e-03, 2.0939e-03, 1.9215e-03,\n 1.1159e-03, 2.3576e-03, 2.0096e-03, 2.3771e-03, 1.7342e-03, 2.0992e-03,\n 8.7721e-04, 6.9075e-11, 2.1280e-03, 8.3135e-04, 2.6536e-03, 1.9391e-03,\n 2.5293e-03, 2.6126e-03, 9.4995e-04, 2.5013e-03, 1.1286e-03, 2.2905e-03,\n 2.5150e-03, 1.7555e-03, 2.4683e-03, 2.9266e-03, 2.3219e-03, 1.7237e-03,\n 2.2353e-03, 2.2496e-03, 2.2942e-03, 2.4539e-03, 2.7359e-03, 2.2534e-03,\n 2.0172e-03, 2.1988e-03, 2.6511e-03, 8.7214e-04, 2.2709e-03, 4.7531e-04,\n 1.7137e-03, 2.2471e-03, 2.1729e-03, 6.2032e-10, 2.0394e-03, 2.5949e-03,\n 2.3184e-03, 2.0583e-03, 2.0734e-03, 2.1209e-03, 1.8971e-03, 2.0631e-03,\n 2.6023e-03, 1.7904e-03, 2.1458e-03, 1.0752e-03, 2.1445e-03, 2.5338e-04,\n 1.4822e-03, 1.8954e-03, 5.7452e-04, 1.6543e-04, 2.4514e-03, 2.5245e-03,\n 7.3677e-04, 2.4545e-03, 2.1723e-03, 1.7622e-03, 1.6343e-03, 1.1413e-03,\n 1.3320e-03, 4.0818e-08, 2.1628e-03, 2.1441e-03, 2.4104e-03, 1.5729e-03,\n 2.0905e-03, 1.0835e-03, 3.7553e-08, 2.1000e-03, 1.3023e-03, 1.4745e-03,\n 2.8444e-03, 2.4835e-03, 2.5512e-03, 2.2178e-03, 1.9541e-03, 2.6199e-04,\n 4.5669e-05, 2.5731e-03, 1.9573e-03, 2.5787e-03, 2.3326e-03, 1.4135e-03,\n 2.4080e-03, 2.1238e-03, 2.5445e-03, 2.2499e-03, 2.2720e-03, 2.7928e-03,\n 8.8953e-04, 2.4567e-03, 2.3663e-03, 2.1039e-03, 2.3241e-03, 1.6384e-03,\n 2.0717e-03, 2.3973e-03, 2.2532e-03, 2.9173e-07, 2.8567e-03, 1.8422e-03,\n 2.0875e-03, 2.3655e-03, 2.0334e-03, 1.9806e-03, 2.2821e-03, 2.0834e-03,\n 1.8857e-03, 5.0023e-04, 2.5251e-03, 1.8801e-03, 2.3159e-03, 2.2441e-03,\n 3.4629e-10, 2.2182e-03, 1.2212e-03, 7.7967e-04, 2.2927e-03, 2.3659e-03,\n 2.1787e-03, 1.9373e-03, 2.2254e-03, 2.1515e-03, 2.1631e-03, 8.5829e-04,\n 2.1926e-03, 1.9583e-03, 2.4266e-03, 1.8932e-03, 1.6585e-03, 2.3456e-03,\n 2.1349e-03, 5.4703e-04, 2.2006e-03, 2.3905e-03, 2.2344e-03, 2.0369e-03,\n 2.0564e-03, 2.1006e-03, 2.8016e-04, 1.0178e-03, 2.2378e-03, 2.4372e-03,\n 1.9077e-03, 2.1609e-03, 1.4411e-03, 1.3787e-03, 2.4927e-03, 1.0359e-03,\n 2.4392e-03, 2.4551e-03, 5.7372e-04, 6.0724e-04, 4.6947e-04, 1.1671e-03,\n 1.6242e-03, 2.4646e-03, 1.0543e-03, 2.0590e-03, 1.7073e-03, 1.4345e-03,\n 2.3075e-03, 2.0428e-03, 9.0425e-04, 1.3145e-03, 1.2631e-03, 2.6683e-03,\n 2.6233e-05, 2.3555e-03, 2.2694e-03, 2.3043e-03, 2.3207e-03, 2.9618e-03,\n 2.6839e-03, 2.3266e-03, 1.7628e-03, 2.4710e-03, 2.3672e-03, 8.6989e-09,\n 1.5029e-03, 2.1329e-03, 1.4662e-03, 2.1289e-03, 5.8336e-04, 1.7406e-03,\n 2.6728e-03, 1.9156e-03, 7.4026e-04, 2.3287e-03, 2.5068e-03, 2.3045e-03,\n 2.2402e-03, 1.2569e-03, 2.4143e-03, 1.5179e-03, 2.2210e-03, 1.6664e-03,\n 2.0580e-03, 1.9902e-03, 2.6756e-03, 1.1533e-03, 2.3843e-03, 1.1576e-03,\n 2.1888e-03, 2.2781e-03, 2.7884e-03, 2.1953e-03, 2.3637e-03, 1.3471e-03,\n 2.4828e-03, 9.4887e-04, 2.3243e-03, 2.0525e-03, 2.3128e-03, 3.8433e-04,\n 2.3175e-03, 1.8396e-03, 4.2507e-09, 7.1352e-04, 3.4186e-05, 8.1641e-04,\n 1.7586e-03, 1.9858e-03, 2.3176e-03, 2.4726e-03, 2.4416e-03, 1.5165e-03,\n 2.3544e-03, 1.0457e-03, 2.5987e-03, 2.3891e-03, 2.0503e-03, 1.9405e-03,\n 1.8802e-03, 1.4475e-03, 1.2394e-03, 7.8397e-04, 2.5658e-03, 2.7546e-03,\n 2.0984e-03, 1.1726e-03, 1.1996e-03, 2.6076e-03, 1.3913e-03, 2.7250e-03,\n 1.9333e-03, 1.4413e-03, 1.5256e-03, 2.5462e-03, 2.1945e-03, 2.1621e-03,\n 1.3739e-03, 2.0717e-03, 2.2232e-03, 8.4511e-04, 1.2907e-07, 1.3169e-03,\n 1.5375e-03, 2.2075e-03, 2.0347e-03, 2.2193e-03, 2.4335e-03, 1.8398e-03,\n 2.1413e-03, 2.4325e-03, 2.4049e-03, 9.5552e-04, 2.1720e-03, 2.2246e-03,\n 2.3411e-03, 2.0426e-03, 2.3665e-03, 2.1655e-03, 8.5161e-04, 2.5800e-03,\n 6.6619e-04, 2.5082e-03, 2.1251e-03, 2.3911e-03, 2.3207e-03, 2.4030e-03,\n 2.0053e-03, 1.9740e-08, 8.6932e-08, 1.9819e-03, 2.0162e-03, 2.4714e-03,\n 2.2444e-03, 4.4311e-04, 2.0232e-03, 2.4803e-03, 4.1376e-04, 2.1139e-03,\n 9.1067e-04, 1.8365e-04, 5.6037e-04, 2.8509e-03, 2.3353e-03, 2.2614e-03,\n 2.5119e-03, 2.1252e-03, 1.7994e-03, 2.3652e-03, 2.2485e-03, 2.3763e-03,\n 1.9616e-03, 1.9514e-07, 1.0533e-03, 1.9545e-03, 2.1863e-03, 2.9157e-03,\n 1.7896e-08, 9.2829e-04, 2.3482e-03, 2.3013e-03, 2.0623e-03, 2.5113e-03,\n 2.6312e-03, 2.4231e-03, 2.4741e-03, 2.6204e-03, 2.2984e-03, 1.6990e-03,\n 2.3055e-03, 2.2526e-03, 2.2751e-03, 1.7021e-03, 1.2115e-03, 1.8342e-03,\n 2.7926e-04, 1.5652e-04, 5.9656e-04, 5.4603e-04, 1.6113e-03, 6.1666e-04,\n 2.3287e-03, 8.2289e-04, 2.4363e-03, 2.2765e-03, 1.6607e-03, 1.9663e-03,\n 1.2385e-03, 2.2049e-03, 2.6350e-03, 1.9659e-03, 2.3572e-03, 1.0832e-03,\n 2.6044e-03, 2.3147e-03, 2.3602e-03, 2.2912e-03, 2.4396e-03, 1.8313e-03,\n 2.4977e-03, 8.2840e-05, 7.6493e-04, 2.4836e-03, 2.0153e-03, 2.8734e-03,\n 2.2055e-03, 1.6475e-03, 1.2223e-03, 2.4504e-03, 1.9433e-03, 1.1137e-03,\n 1.6721e-03, 1.5459e-06, 7.0614e-04, 2.3199e-03, 2.2382e-03, 2.0027e-03,\n 1.0439e-03, 2.4866e-03, 2.3220e-03, 2.2651e-03, 1.8960e-03, 8.8446e-04,\n 2.3357e-03, 1.8609e-03, 1.9484e-03, 2.1082e-03, 2.3924e-03, 8.0313e-04,\n 2.1121e-03, 2.2498e-03, 1.8111e-03, 2.5661e-03, 1.4449e-03, 1.3437e-03,\n 2.2832e-03, 3.5471e-08, 9.0418e-04, 2.0917e-03, 2.1935e-03, 1.9116e-03,\n 2.4084e-03, 2.6080e-03, 1.1367e-03, 2.2981e-03, 2.2655e-03, 1.0147e-03,\n 2.1324e-03, 3.0307e-03, 2.3596e-03, 1.2397e-03, 2.1829e-03, 1.7435e-03,\n 2.4314e-03, 8.0006e-04, 2.0079e-03, 2.0783e-03, 1.0996e-03, 2.2827e-03,\n 2.4053e-03, 1.8660e-03, 1.9977e-03, 2.0427e-03, 2.4042e-03, 2.2336e-03,\n 3.1926e-04, 1.2732e-03, 1.2815e-03, 2.4449e-03, 1.9246e-03, 2.9635e-03,\n 1.9751e-03, 2.3030e-03, 2.7361e-03, 1.6384e-07, 2.3080e-03, 2.2656e-03,\n 5.7760e-04, 2.3242e-03, 1.8621e-03, 1.1781e-03, 2.3175e-03, 1.6376e-03,\n 2.2961e-03, 6.5516e-05, 1.0642e-03, 1.8956e-03, 2.1472e-03, 8.8240e-04,\n 2.2490e-03, 1.4340e-07, 2.1697e-03, 2.1769e-03, 1.9057e-03, 2.0080e-03,\n 2.1360e-03, 2.1710e-03, 2.0935e-03, 1.2766e-03, 2.0026e-03, 2.3095e-03,\n 8.7796e-04, 1.3307e-08, 8.2713e-04, 2.3146e-03, 2.1223e-09, 2.1006e-03,\n 2.2825e-03, 1.1784e-04, 2.0655e-03, 1.9814e-03, 1.5181e-03, 4.9282e-04,\n 9.8451e-04, 6.5205e-04], device='cuda:0')" | |
| }, | |
| "4": { | |
| "step": "tensor(6260.)", | |
| "exp_avg": "tensor([[-1.3696e-04, 1.3360e-04, 4.3022e-05, ..., 1.8117e-05,\n -3.7524e-06, -6.2092e-06],\n [ 4.7460e-04, 2.2737e-04, 5.2812e-06, ..., 1.1790e-05,\n -2.8303e-05, -2.3222e-05],\n [-1.6619e-04, -1.6797e-04, -1.7393e-05, ..., 2.6658e-05,\n 8.7483e-05, -6.1632e-06],\n ...,\n [-2.7077e-04, -1.1349e-04, 4.8902e-05, ..., -3.5663e-05,\n 8.0127e-05, 5.6280e-07],\n [-4.9770e-04, -5.2166e-05, -2.3442e-05, ..., 2.2118e-05,\n -3.0998e-06, -1.5163e-05],\n [ 1.0683e-04, 6.2915e-05, -3.0317e-05, ..., -3.2692e-05,\n -4.7489e-05, 6.5322e-06]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[2.8699e-07, 3.3533e-07, 9.0435e-09, ..., 1.0669e-07, 4.4584e-08,\n 5.4326e-09],\n [5.5001e-07, 6.9598e-07, 1.1643e-08, ..., 1.1631e-08, 7.9185e-08,\n 2.3460e-08],\n [4.4530e-07, 6.0069e-07, 4.5371e-09, ..., 3.6378e-08, 4.2821e-08,\n 1.8219e-08],\n ...,\n [4.2281e-07, 7.1814e-07, 7.5493e-09, ..., 2.3532e-08, 5.9713e-08,\n 1.6166e-08],\n [4.2245e-07, 6.9378e-07, 9.3833e-09, ..., 5.0033e-08, 6.3601e-08,\n 1.3951e-08],\n [5.2584e-07, 8.0056e-07, 1.7899e-08, ..., 2.5917e-08, 7.1535e-08,\n 9.6266e-09]], device='cuda:0')" | |
| }, | |
| "5": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([[ 1.9494e-05, 7.0051e-06, -3.2682e-06, ..., -1.9544e-06,\n 1.8390e-07, 1.5588e-07],\n [ 1.4529e-07, -6.8810e-06, 5.9431e-06, ..., 6.0483e-06,\n -1.5329e-06, 1.5239e-06],\n [-1.7346e-05, 6.8581e-07, -2.6905e-05, ..., -6.2609e-07,\n -1.1453e-06, -8.2278e-07],\n ...,\n [-3.9712e-06, 6.1368e-06, -8.0675e-06, ..., -1.8386e-06,\n 8.8590e-07, 7.6704e-07],\n [ 1.3317e-06, 1.2702e-05, -7.2322e-06, ..., 3.0252e-06,\n 2.3881e-06, 6.0639e-06],\n [ 9.4251e-06, -1.0249e-05, -4.7948e-06, ..., -4.5530e-06,\n -5.4284e-06, 3.2429e-07]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[3.8087e-09, 1.9501e-09, 3.1005e-10, ..., 3.2051e-10, 5.2854e-10,\n 5.4474e-10],\n [1.4943e-10, 9.8423e-10, 2.1185e-10, ..., 2.8002e-09, 2.0762e-10,\n 4.5338e-10],\n [1.7719e-09, 2.5870e-10, 2.9427e-10, ..., 2.9233e-10, 2.2890e-10,\n 5.2653e-10],\n ...,\n [1.0323e-09, 5.6697e-10, 4.8303e-10, ..., 7.5165e-10, 5.1092e-10,\n 3.7146e-10],\n [3.3707e-10, 2.7873e-09, 4.0367e-10, ..., 1.4169e-09, 4.8054e-10,\n 1.0919e-09],\n [2.9401e-09, 1.2387e-10, 1.5390e-10, ..., 2.7543e-10, 8.7368e-10,\n 2.8697e-10]], device='cuda:0')" | |
| }, | |
| "6": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([ 1.6006e-03, -7.8564e-04, -4.9745e-04, ..., -3.3313e-05,\n 5.3103e-04, -1.1981e-03], device='cuda:0')", | |
| "exp_avg_sq": "tensor([7.9636e-06, 6.6279e-06, 8.9004e-06, ..., 7.5030e-06, 8.0549e-06,\n 8.1558e-06], device='cuda:0')" | |
| }, | |
| "7": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([[-7.5676e-06, 3.7859e-06, -1.4781e-06, ..., -3.9294e-06,\n -2.4208e-05, 4.9028e-06],\n [-6.9484e-06, -2.9819e-06, -2.7441e-06, ..., -5.9271e-06,\n -1.5014e-05, -3.2692e-06],\n [ 5.0979e-06, -6.1285e-06, -4.2667e-06, ..., 5.1584e-06,\n 1.8542e-05, 5.9064e-06],\n ...,\n [-2.3607e-06, 5.2226e-06, 1.0742e-05, ..., -1.1877e-05,\n -6.4642e-06, 4.6107e-06],\n [-1.0722e-05, -3.5663e-06, 3.7635e-06, ..., -9.9505e-06,\n -1.8677e-05, -7.8065e-06],\n [-7.0040e-06, -3.7658e-06, 6.7026e-06, ..., -6.2246e-07,\n -1.4885e-05, -2.6397e-06]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[3.6245e-10, 2.2329e-10, 4.5823e-10, ..., 2.7173e-10, 7.3399e-10,\n 4.7935e-10],\n [6.8720e-10, 5.9961e-10, 6.6910e-10, ..., 7.4165e-10, 1.1265e-09,\n 7.8219e-10],\n [7.5174e-10, 5.1944e-10, 8.3053e-10, ..., 7.2094e-10, 9.9498e-10,\n 6.9835e-10],\n ...,\n [8.2339e-10, 8.3513e-10, 1.0441e-09, ..., 9.6042e-10, 1.3126e-09,\n 9.9008e-10],\n [7.6150e-10, 7.4384e-10, 7.3682e-10, ..., 8.3939e-10, 9.1103e-10,\n 6.6738e-10],\n [8.6501e-10, 4.9953e-10, 1.0287e-09, ..., 8.0046e-10, 1.0384e-09,\n 8.0412e-10]], device='cuda:0')" | |
| }, | |
| "32": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([2.9611e-14], device='cuda:0')", | |
| "exp_avg_sq": "tensor([5.2951e-05], device='cuda:0')" | |
| }, | |
| "33": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([ 2.1372e-16, -5.8856e-17, -1.5486e-16], device='cuda:0')", | |
| "exp_avg_sq": "tensor([1.5392e-08, 2.5728e-08, 2.4030e-09], device='cuda:0')" | |
| }, | |
| "34": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([-5.6052e-45, -1.7753e-16, -5.4738e-17, -2.7549e-17, -1.1068e-17,\n 2.8655e-16, 3.4156e-16, 3.2667e-16, 1.3247e-15, 3.9524e-16],\n device='cuda:0')", | |
| "exp_avg_sq": "tensor([2.0316e-05, 2.4408e-07, 2.3482e-07, 1.8478e-07, 2.2201e-07, 2.0602e-07,\n 3.4539e-07, 3.0888e-07, 3.3358e-07, 2.1723e-07], device='cuda:0')" | |
| }, | |
| "36": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([[ 4.6302e-19, 2.3932e-19, 1.0632e-18, ..., 2.7546e-19,\n 3.7119e-19, 4.6394e-20],\n [ 3.1411e-19, -5.0559e-19, -2.1059e-19, ..., -1.1402e-19,\n -1.2513e-19, -5.9795e-20],\n [ 2.6630e-19, 5.9625e-21, 7.6489e-19, ..., -1.6719e-20,\n 1.2725e-19, 1.2194e-19],\n ...,\n [ 3.6405e-19, 1.2599e-18, 3.8478e-19, ..., 2.4145e-19,\n 2.8915e-19, 1.9033e-19],\n [-2.4506e-18, 5.9950e-18, -1.5505e-18, ..., -7.1369e-19,\n -1.8098e-19, -5.0975e-19],\n [ 3.3803e-19, -1.4350e-18, 1.4313e-18, ..., -1.4526e-19,\n 7.7680e-20, 3.2205e-19]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[2.7733e-12, 6.1119e-13, 1.6517e-12, ..., 1.8978e-12, 2.3934e-12,\n 3.2296e-12],\n [7.6647e-15, 1.3750e-14, 1.9424e-14, ..., 1.1156e-14, 9.5052e-15,\n 1.8450e-14],\n [5.6024e-14, 2.9573e-14, 7.2472e-14, ..., 3.1013e-14, 1.1412e-13,\n 6.0126e-14],\n ...,\n [6.5647e-14, 4.6996e-14, 4.3662e-14, ..., 7.4369e-14, 5.0229e-14,\n 1.0761e-13],\n [1.4270e-12, 3.6499e-13, 1.1333e-12, ..., 7.5783e-13, 1.4260e-12,\n 1.5523e-12],\n [9.8311e-13, 5.2801e-13, 1.1939e-12, ..., 4.4870e-13, 1.2574e-12,\n 1.5805e-12]], device='cuda:0')" | |
| }, | |
| "37": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([ 3.0117e-16, -2.1780e-17, 1.4797e-16, 1.1795e-16, -4.7011e-18,\n 5.8041e-16, 2.6381e-16, -1.5242e-17, -8.9683e-16, -8.9381e-16,\n -1.1290e-15, 4.8412e-16, -8.9369e-16, 3.9253e-16, 2.0250e-16,\n 5.0453e-16, -7.3660e-16, 7.2177e-16, 7.1654e-16, -6.0833e-17,\n -1.3682e-16, 1.3695e-16, 2.6776e-17, 8.8923e-16, 1.8318e-17,\n -1.1090e-15, 1.6176e-16, -1.3593e-15, -6.6484e-16, 1.4048e-17,\n -1.3862e-16, -1.3812e-15, 4.8187e-16, 1.1190e-15, -3.1536e-16,\n -9.4935e-17, 6.4979e-17, 2.5887e-16, -3.2760e-18, 4.5856e-16,\n 9.9609e-17, -3.3281e-16, 5.1000e-16, 6.9931e-16, -7.7239e-16,\n 2.6191e-16, 6.1723e-16, -9.6550e-16, 4.6408e-16, 4.8705e-16,\n 9.3761e-16, 6.0333e-17, -1.1722e-16, 2.2743e-16, 2.0536e-16,\n 1.6242e-16, 2.9971e-16, -2.3475e-16, 2.2389e-17, 9.7998e-18,\n 1.5446e-16, 7.3147e-17, -1.4516e-15, -1.6356e-15, -4.8146e-16,\n 1.8005e-16, 3.5526e-18, 2.1948e-16, -2.3230e-16, 2.7964e-16,\n -1.1760e-16, 9.5736e-16, -5.7058e-18, -6.6686e-17, 8.6366e-17,\n 1.1431e-16, -1.5946e-16, -3.6794e-16, 1.1030e-16, -8.5434e-17,\n -2.6028e-16, -4.8366e-16, 7.0344e-17, -2.6260e-17, -6.5938e-16,\n 1.0351e-16, -5.7042e-17, 9.6833e-16, -7.6201e-17, -3.9897e-16,\n 7.2249e-16, 4.1411e-16, -8.2088e-17, 1.0956e-16, -6.1657e-17,\n -1.0132e-16, 6.4527e-16, -1.0503e-15, -3.8782e-17, 2.4657e-17,\n 6.5541e-16, 7.5778e-16, 7.6268e-16, 1.2816e-16, 3.6932e-16,\n 7.0480e-16, -9.0736e-16, -3.2552e-17, 8.8145e-17, 5.9729e-16,\n 6.1155e-17, 1.9338e-16, -1.8623e-16, 8.5142e-16, 4.6903e-17,\n -1.1368e-16, -6.6265e-16, 4.1084e-16, 1.8427e-16, -1.5993e-17,\n -9.3791e-16, 9.9122e-17, -1.6288e-16, -5.2911e-16, 5.5322e-16,\n 1.9773e-16, -2.3398e-16, 6.1923e-16, -5.4302e-17, 4.5700e-16,\n 3.4645e-16, -2.0068e-16, -2.7919e-16, 2.7881e-16, 1.0685e-16,\n -8.7827e-16, 3.5945e-17, 4.8833e-17, 1.0466e-16, 6.4262e-16,\n 3.4138e-18, -6.9712e-16, -4.5457e-16, 5.1524e-17, -3.8432e-17,\n -2.4981e-17, -3.7778e-16, -6.5803e-17, 1.5052e-16, 6.2066e-16,\n 2.4851e-17, -1.0489e-15, -3.5996e-16, 1.3860e-15, -4.8995e-17,\n 5.1559e-17, -1.6412e-15, 3.1517e-16, -2.7162e-16, -6.2522e-16,\n -6.9583e-16, -1.8388e-17, -2.7296e-17, -7.0102e-16, 1.0310e-15,\n 3.2545e-16, -3.6035e-16, 4.8452e-16, 9.9045e-16, 1.4239e-16,\n -7.5512e-16, 1.1389e-16, 5.9261e-16, 5.8860e-16, 4.0392e-17,\n -1.8993e-16, 2.4274e-16, 3.7092e-17, 7.8826e-16, 1.0217e-15,\n 3.4133e-16, 9.1568e-17, 5.5919e-16, 7.1156e-17, -2.9408e-16,\n -6.5692e-16, -8.6859e-17, -7.0056e-17, 3.8254e-16, 6.4796e-18,\n -2.1983e-16, 7.5824e-17, -1.2040e-16, 1.1386e-16, 1.8211e-16,\n 8.9922e-17, -4.0539e-17, 9.5525e-17, -1.4645e-15, 6.4047e-16,\n 1.9461e-16, -5.2903e-16, 3.3910e-16, -1.5820e-15, -1.3094e-15,\n 5.0413e-16, -1.1075e-15, 1.9473e-16, 1.2795e-16, -8.1622e-16,\n 1.0097e-15, 2.0783e-16, -7.8006e-17, 6.1211e-17, 3.1321e-16,\n 1.8180e-15, 7.6695e-16, -4.2560e-16, 2.0411e-16, -7.2066e-17,\n 3.7066e-16, -3.3477e-16, 2.6291e-17, -1.2505e-15, -5.3002e-16,\n -8.4844e-16, -5.9567e-17, 2.1995e-16, -6.5073e-17, 3.5192e-16,\n 4.0171e-16, 3.8533e-16, -4.0397e-16, -9.7852e-16, -1.4035e-16,\n 2.6990e-16, 7.1813e-16, 5.7891e-16, -5.0358e-16, -1.4260e-16,\n 6.2481e-17, 2.7835e-16, -5.3135e-16, 5.1093e-16, 7.1044e-16,\n 3.0849e-16, -1.5182e-15, 3.5014e-16, -5.7669e-16, -6.8281e-17,\n -3.8376e-17, 1.0212e-16, 3.5273e-17, 1.3002e-16, -5.4320e-16,\n 3.0968e-16], device='cuda:0')", | |
| "exp_avg_sq": "tensor([5.3117e-07, 5.1252e-09, 1.3314e-08, 1.0391e-07, 2.3640e-10, 2.0393e-07,\n 1.4818e-09, 5.6220e-08, 5.5922e-09, 1.9848e-07, 4.8285e-08, 8.4401e-07,\n 4.1963e-07, 2.2507e-08, 3.8459e-08, 1.1630e-07, 2.0378e-08, 2.4816e-08,\n 1.6838e-07, 3.4307e-09, 1.4090e-10, 7.8567e-11, 2.9784e-08, 8.4946e-08,\n 8.2403e-11, 2.1748e-06, 4.0651e-08, 1.3286e-06, 4.2936e-08, 1.0604e-07,\n 3.1919e-08, 3.4822e-07, 7.0035e-08, 2.2543e-07, 5.7949e-08, 1.1479e-07,\n 2.2664e-08, 1.2351e-06, 1.3725e-10, 1.3710e-08, 1.5555e-10, 3.3514e-09,\n 7.3737e-10, 1.1720e-10, 1.1840e-07, 9.5393e-07, 6.3198e-09, 9.1135e-07,\n 8.5478e-07, 1.1073e-08, 2.1414e-07, 1.0010e-07, 2.3845e-09, 2.1840e-08,\n 5.3956e-08, 1.1518e-07, 1.0645e-09, 5.1547e-09, 1.4629e-07, 9.6191e-08,\n 1.5878e-07, 9.0300e-10, 6.8221e-07, 7.4637e-07, 6.2093e-10, 1.2268e-09,\n 1.3059e-08, 5.0759e-07, 1.0622e-08, 2.3386e-07, 4.5544e-09, 1.3771e-08,\n 1.0656e-08, 5.5094e-08, 6.7968e-07, 2.3728e-11, 9.0880e-10, 2.1359e-08,\n 2.2666e-07, 1.8523e-08, 5.8280e-09, 3.2961e-09, 4.2941e-07, 3.9938e-10,\n 1.0058e-08, 2.8014e-08, 1.3901e-07, 6.8629e-07, 1.3378e-07, 6.2516e-09,\n 9.6881e-07, 2.1031e-07, 6.9000e-10, 4.0139e-08, 3.7243e-09, 5.3096e-09,\n 8.7232e-07, 9.2776e-08, 7.1552e-09, 1.7527e-09, 1.4580e-07, 5.6694e-08,\n 7.6369e-07, 1.6076e-08, 6.5163e-07, 7.0210e-07, 7.2206e-10, 8.4253e-08,\n 9.1710e-09, 1.5069e-08, 1.7974e-07, 8.6792e-11, 1.3145e-09, 1.8611e-07,\n 7.6749e-10, 1.7015e-07, 4.6566e-07, 6.7436e-10, 4.3676e-08, 2.2761e-08,\n 1.4522e-07, 2.3322e-08, 2.2197e-09, 5.1384e-09, 1.0687e-07, 4.2951e-07,\n 8.5619e-08, 5.6484e-07, 1.3634e-07, 1.6128e-06, 2.4926e-08, 2.7254e-09,\n 2.1769e-08, 3.5718e-07, 6.1786e-11, 3.1838e-07, 9.0652e-09, 9.7577e-11,\n 9.7676e-08, 1.3533e-07, 2.3773e-07, 7.8598e-07, 2.3660e-09, 1.4680e-09,\n 3.2553e-10, 7.4118e-09, 5.2440e-07, 3.7504e-08, 4.8217e-08, 6.6560e-08,\n 1.7894e-10, 2.3722e-06, 5.4934e-09, 1.7334e-08, 9.8020e-11, 4.0745e-07,\n 1.8405e-06, 9.3115e-09, 1.2040e-08, 1.5136e-07, 2.7920e-07, 7.6440e-10,\n 1.0388e-10, 8.3224e-07, 2.0808e-07, 2.4240e-08, 8.2646e-09, 3.7699e-07,\n 2.1007e-07, 1.9152e-07, 4.1648e-07, 1.4487e-07, 8.1406e-07, 6.8062e-09,\n 4.8829e-08, 2.2672e-08, 7.3216e-09, 2.1800e-07, 1.7364e-08, 2.2502e-07,\n 1.0272e-06, 2.8553e-10, 2.6806e-07, 7.6729e-08, 6.6011e-08, 4.8682e-07,\n 6.4287e-08, 3.4350e-08, 5.4337e-08, 4.9906e-07, 1.7249e-08, 3.4268e-08,\n 4.9817e-08, 5.4287e-07, 3.8017e-07, 4.0047e-09, 1.3334e-09, 9.5358e-08,\n 5.5097e-08, 1.2588e-07, 4.6687e-08, 2.8499e-06, 4.6743e-08, 4.0495e-07,\n 1.9921e-07, 6.2957e-07, 3.1992e-07, 2.2091e-10, 1.9294e-07, 3.7519e-07,\n 1.2046e-06, 3.2955e-10, 2.2477e-08, 5.9936e-10, 9.1385e-09, 1.9584e-07,\n 3.5885e-08, 1.7279e-09, 3.0906e-10, 8.0751e-10, 5.5031e-07, 3.0135e-07,\n 7.7475e-08, 8.2236e-07, 9.3503e-08, 3.4106e-07, 3.7646e-09, 9.7724e-08,\n 1.9830e-08, 7.9728e-08, 8.0219e-07, 4.1830e-07, 9.8330e-08, 1.2989e-07,\n 1.9025e-08, 4.1326e-08, 5.2006e-09, 1.8146e-06, 4.9977e-08, 7.3822e-08,\n 9.4979e-08, 1.1959e-06, 8.9215e-08, 5.0871e-09, 1.1525e-08, 2.6325e-09,\n 1.1941e-07, 5.2586e-09, 5.4431e-09, 4.0373e-08, 1.8256e-08, 4.1769e-07,\n 6.8662e-09, 2.1836e-08, 2.7347e-07, 3.1024e-07], device='cuda:0')" | |
| }, | |
| "38": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([-2.6052e-18, -5.1943e-18, -3.8413e-20, -1.0068e-16, -6.7946e-19,\n -6.0663e-19, -3.2696e-17, -7.8490e-19, -1.2062e-17, -1.3296e-16,\n -3.5815e-17, 1.6289e-21, -1.5668e-16, -1.7571e-17, 7.1716e-21,\n -5.8658e-17, -1.7911e-17, 4.6857e-19, -8.1878e-19, -6.2797e-19,\n -1.5919e-18, 2.6597e-20, -9.3794e-20, -3.7217e-17, -8.5485e-19,\n -1.8491e-16, 9.4949e-19, -1.3940e-16, -1.0956e-16, -5.2725e-19,\n -3.2016e-18, -1.1033e-16, -8.4388e-18, -6.0585e-19, 2.3708e-18,\n -2.2867e-19, -4.1848e-19, -4.9268e-19, -6.9613e-19, -6.5221e-20,\n 6.5319e-19, -2.0631e-17, -8.7368e-19, -1.4916e-18, -1.2795e-17,\n 4.1587e-19, -7.1301e-17, -1.4205e-16, -3.7275e-18, -3.1708e-17,\n -6.0553e-17, -1.0753e-19, -1.3338e-17, -7.1293e-19, -4.5329e-17,\n -1.1328e-17, 1.2341e-19, -1.4342e-18, -5.0747e-17, -1.3726e-16,\n -6.2413e-20, 3.7677e-19, -7.7200e-17, -1.5028e-16, -3.9996e-18,\n 8.3578e-20, -1.4555e-19, 9.0250e-20, -2.3528e-17, 7.6447e-20,\n -1.1159e-17, 1.2039e-19, -5.9467e-19, -2.1992e-18, -1.1875e-16,\n 4.2098e-19, -3.6354e-18, 3.2795e-18, 1.4977e-19, -4.0079e-18,\n -3.9002e-18, -6.8356e-18, -2.8813e-18, -1.6249e-18, -3.4384e-17,\n 7.1255e-19, -1.3938e-17, 1.2562e-19, 2.6546e-18, -3.5494e-17,\n -2.5880e-17, -6.8562e-20, -6.0004e-18, -1.2106e-19, -2.4253e-18,\n -1.0014e-18, 2.8956e-19, -8.6149e-17, -1.9593e-18, -1.1413e-18,\n -1.5842e-17, -3.2665e-18, -8.3173e-17, -2.4947e-19, -1.9161e-19,\n -9.5063e-17, -2.7931e-17, 7.4537e-20, 1.7074e-19, -3.7765e-19,\n -8.2755e-17, 1.9657e-19, -2.4688e-18, -3.2614e-17, 3.1205e-19,\n -5.5098e-18, -1.8399e-16, -7.3709e-19, -1.3970e-17, -8.7775e-17,\n -5.8828e-17, -1.9653e-19, -4.6975e-18, -1.0272e-17, -3.6446e-19,\n -9.0231e-17, 7.8885e-19, -7.0146e-19, -4.7740e-20, -3.2455e-19,\n 1.1065e-19, -3.8219e-19, -1.0393e-18, 7.5914e-20, 4.4502e-19,\n -1.3216e-16, 8.0479e-19, 1.5170e-18, 8.2382e-19, 7.7076e-19,\n -1.9363e-18, -1.0215e-16, 4.8812e-18, -3.6498e-19, 5.8023e-20,\n -2.7192e-19, -1.0866e-16, -4.1693e-18, -2.7500e-19, -1.2940e-20,\n -2.8260e-19, -1.7741e-16, 4.9468e-19, 5.2633e-20, -3.1914e-18,\n 1.6136e-19, -1.4228e-16, -5.5887e-19, 2.7551e-18, -1.0574e-16,\n -2.3484e-17, -9.5840e-18, -7.8956e-19, -1.6154e-16, -1.2750e-17,\n 7.8466e-20, -5.3850e-17, -9.6963e-18, -6.6838e-17, 6.6092e-20,\n -9.9710e-17, -7.9981e-17, -2.9154e-18, 8.1112e-20, -2.1992e-17,\n 1.5666e-18, 5.7703e-20, -5.1443e-17, 2.2138e-19, -9.1908e-18,\n -3.1685e-18, 3.8051e-19, -1.2863e-19, -1.2403e-17, 1.3257e-18,\n -5.9509e-18, -1.2839e-18, -6.1282e-18, 2.0056e-19, -2.2923e-18,\n 1.1832e-18, -1.5503e-19, -1.7207e-19, -7.1481e-17, 4.2781e-19,\n -1.2834e-19, -3.8791e-19, -9.0001e-17, -3.0285e-17, -6.1034e-17,\n 2.3965e-19, -1.4156e-16, -1.2057e-17, -9.4085e-17, -7.7527e-17,\n -2.0409e-19, -1.3801e-16, -1.1976e-19, -7.3187e-17, -5.4783e-17,\n -4.6075e-19, 2.2985e-19, -1.6253e-18, 2.4483e-19, -2.6423e-18,\n -1.6409e-18, -1.8589e-17, 4.5813e-18, -4.7345e-20, -5.8559e-19,\n -2.3929e-20, -2.6053e-18, -1.3134e-16, -1.1858e-16, -1.7234e-17,\n -2.2762e-17, -7.8663e-19, -1.6071e-19, -2.5142e-21, -2.5728e-19,\n -1.8833e-19, -6.4025e-17, 3.2958e-18, -4.0394e-17, -2.9035e-18,\n -1.6419e-17, -6.9506e-19, -1.2413e-17, -2.4620e-17, -4.4475e-19,\n 1.7036e-19, -1.9504e-19, 7.6237e-18, -1.4088e-19, -1.6095e-19,\n -3.7900e-19, -4.9461e-17, 2.2079e-19, -3.5447e-17, -2.8938e-18,\n -1.0279e-18, -2.6733e-19, 1.4082e-19, 1.6195e-19, -4.0991e-17,\n -1.8626e-19], device='cuda:0')", | |
| "exp_avg_sq": "tensor([1.1365e-10, 1.4180e-11, 1.3388e-13, 4.6139e-10, 1.4038e-12, 1.1877e-12,\n 5.6058e-11, 2.8982e-13, 1.2440e-13, 8.0441e-10, 1.1199e-11, 6.1113e-11,\n 1.4684e-09, 4.9813e-11, 2.4377e-13, 1.9697e-10, 1.7328e-13, 2.7396e-13,\n 3.0397e-13, 1.2648e-12, 3.3118e-14, 1.0742e-14, 8.5824e-13, 1.0557e-10,\n 1.7489e-13, 4.0760e-09, 6.5204e-14, 1.4222e-09, 5.5715e-10, 1.0489e-12,\n 5.9029e-14, 3.9293e-10, 3.8861e-13, 1.1528e-11, 1.2995e-14, 3.1708e-13,\n 1.1923e-12, 1.5181e-10, 1.8684e-13, 1.7359e-12, 8.8380e-14, 2.6145e-13,\n 1.6603e-13, 4.0782e-14, 5.8554e-11, 5.8231e-12, 6.2348e-10, 1.5777e-09,\n 1.1395e-10, 4.9630e-11, 5.8859e-10, 4.3544e-13, 5.4098e-14, 3.5451e-12,\n 2.0074e-10, 4.1743e-11, 4.5527e-13, 1.5933e-14, 8.2657e-11, 9.5507e-10,\n 3.4546e-12, 3.4598e-14, 1.2637e-10, 9.0654e-10, 3.0679e-13, 4.2003e-13,\n 1.8440e-12, 1.3801e-11, 1.2589e-10, 4.0144e-12, 1.3028e-13, 6.8321e-13,\n 3.6385e-15, 2.0035e-13, 1.6630e-09, 5.5539e-14, 5.4782e-15, 1.6299e-14,\n 4.7317e-13, 1.6080e-11, 6.6930e-15, 4.2896e-14, 9.3272e-11, 7.2588e-15,\n 3.0668e-12, 3.0425e-13, 1.1423e-10, 2.3189e-11, 6.5261e-12, 2.2626e-10,\n 1.6049e-10, 1.6571e-11, 1.5397e-13, 4.2671e-14, 6.4260e-12, 4.6899e-12,\n 5.5187e-11, 9.8381e-11, 2.1900e-14, 3.8626e-14, 2.9208e-11, 1.8564e-13,\n 1.5403e-09, 3.1850e-13, 2.1528e-11, 1.5599e-09, 1.4789e-13, 1.9998e-14,\n 1.6343e-12, 9.8862e-13, 4.3778e-10, 3.1054e-13, 1.9733e-15, 1.4480e-10,\n 2.5123e-12, 6.6568e-11, 2.6392e-09, 1.6122e-12, 7.5332e-11, 2.7114e-10,\n 4.4266e-10, 2.6391e-13, 1.0462e-14, 7.9907e-14, 1.4448e-12, 8.7889e-10,\n 4.2547e-14, 5.5163e-11, 1.7282e-13, 1.7059e-10, 2.5804e-13, 1.2651e-12,\n 8.0483e-15, 2.3607e-11, 1.5426e-13, 7.2833e-10, 2.7789e-12, 8.6459e-14,\n 1.4356e-14, 1.2823e-13, 1.2006e-11, 6.2610e-10, 8.7957e-15, 4.3166e-13,\n 6.2524e-13, 1.9795e-12, 7.3182e-10, 7.8401e-12, 4.3926e-12, 1.8711e-12,\n 2.4376e-13, 4.3741e-09, 1.1713e-15, 5.4312e-12, 5.8822e-14, 4.8773e-12,\n 1.2903e-09, 5.1426e-14, 6.9238e-13, 2.7368e-10, 1.4611e-10, 2.1331e-13,\n 1.3743e-13, 2.2037e-09, 4.6228e-11, 5.1944e-13, 4.0365e-11, 1.5960e-10,\n 7.3237e-10, 6.8067e-12, 4.3022e-10, 2.6239e-10, 5.1702e-11, 1.3069e-12,\n 4.6887e-11, 1.2872e-15, 4.7842e-13, 1.3939e-10, 1.6065e-12, 3.0829e-11,\n 1.9193e-10, 1.6562e-13, 9.2777e-13, 1.0078e-10, 2.7673e-14, 1.0223e-10,\n 7.4746e-13, 1.8905e-11, 2.8775e-12, 6.7839e-12, 1.0296e-14, 2.4809e-15,\n 5.8885e-14, 3.4093e-10, 4.7265e-12, 7.3209e-13, 2.2723e-12, 3.5975e-10,\n 9.3000e-12, 4.5917e-10, 1.6055e-15, 3.6196e-09, 1.3003e-10, 1.5022e-10,\n 6.3105e-11, 4.2956e-11, 6.4101e-10, 1.2466e-14, 2.5750e-10, 4.1895e-11,\n 1.5413e-10, 2.0765e-13, 2.4978e-15, 1.9702e-14, 2.9290e-12, 3.6999e-11,\n 3.7960e-11, 4.0143e-14, 3.7383e-13, 4.8179e-15, 2.0598e-11, 3.0527e-11,\n 1.0529e-09, 6.1592e-10, 9.3603e-11, 1.6406e-10, 1.2946e-14, 2.1633e-12,\n 1.1899e-13, 6.5237e-12, 7.9855e-11, 3.6903e-10, 2.2004e-14, 3.6531e-12,\n 6.1556e-16, 1.3534e-10, 2.5707e-12, 8.0012e-10, 2.4843e-13, 2.4294e-12,\n 2.9927e-12, 5.8841e-11, 1.8087e-13, 1.5619e-12, 2.3064e-12, 5.6613e-13,\n 3.4078e-12, 3.0351e-13, 7.7760e-12, 5.3215e-12, 3.4279e-15, 5.0893e-12,\n 2.1068e-12, 4.4617e-13, 1.3928e-11, 4.8595e-12], device='cuda:0')" | |
| }, | |
| "39": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([-1.3229e-17, 1.3582e-18, -1.7083e-19, -5.8190e-17, -4.2157e-18,\n -5.3806e-18, -3.4923e-17, -5.8476e-19, -4.2383e-17, -7.8853e-17,\n -5.6654e-17, 3.5212e-19, -8.3443e-17, -2.6934e-17, -2.5377e-18,\n -4.1701e-17, -4.2212e-17, 3.3071e-18, 7.1970e-18, -1.3269e-17,\n 5.6091e-19, -1.3631e-18, -7.6555e-18, -2.9840e-17, -3.2404e-19,\n -9.0661e-17, 1.5508e-18, -8.6186e-17, -7.1029e-17, -9.0866e-19,\n 2.4446e-18, -8.1773e-17, -1.7918e-17, -8.2246e-19, -1.5875e-18,\n 4.7305e-20, -6.3547e-18, -1.0022e-17, -1.5696e-18, -7.0147e-19,\n -5.4109e-18, -3.8049e-17, 3.9654e-18, 6.9988e-18, -4.0374e-17,\n 3.0964e-18, -4.4374e-17, -8.1452e-17, -1.2223e-17, -3.2083e-17,\n -3.7667e-17, 2.5468e-19, -2.9781e-17, -5.7500e-18, -4.1213e-17,\n -2.3716e-17, -5.3394e-19, 6.7203e-19, -4.5224e-17, -6.7010e-17,\n -1.9162e-18, 8.9203e-20, -7.3951e-17, -9.2553e-17, -2.6138e-17,\n -6.7470e-19, -8.0492e-18, 2.6459e-19, -3.6736e-17, 1.6156e-18,\n -2.8457e-17, 9.8657e-18, -1.5781e-18, 1.6007e-19, -6.2793e-17,\n -3.4924e-18, 2.1675e-18, -2.2796e-18, 3.7743e-19, 2.7228e-18,\n 2.7705e-18, -2.9962e-17, -1.8276e-17, 9.8667e-19, -4.9360e-17,\n -5.2205e-18, -2.9442e-17, 3.3141e-18, 5.3077e-19, -4.5403e-17,\n -2.6397e-17, -1.7502e-18, 4.4100e-18, -1.2932e-18, 1.0064e-18,\n -8.6691e-18, 3.6947e-18, -7.0593e-17, -7.8884e-19, -1.4812e-17,\n -2.0767e-17, -8.9265e-18, -4.6138e-17, -3.1986e-18, 4.6841e-20,\n -4.9950e-17, -4.9291e-17, 2.8665e-19, -7.7306e-18, 2.8506e-18,\n -5.4427e-17, 2.1119e-18, 1.9250e-18, -2.8907e-17, -8.4198e-18,\n -2.1980e-17, -8.4887e-17, 5.1047e-18, -2.5501e-17, -5.6063e-17,\n -6.1477e-17, -2.7945e-18, 3.2582e-18, -3.3754e-17, 3.5340e-18,\n -5.4971e-17, -5.0689e-19, -4.4597e-18, 4.9088e-20, -4.7367e-18,\n 2.4047e-18, -1.4922e-17, 9.2326e-19, -4.1076e-18, -3.4006e-18,\n -7.8258e-17, -9.1514e-18, 4.1965e-19, 6.7055e-19, 3.3376e-18,\n -1.6001e-17, -6.8691e-17, -3.2052e-18, -1.1314e-17, -3.1543e-19,\n -7.7769e-18, -6.6456e-17, 9.7256e-19, -2.0313e-18, -3.2750e-19,\n -1.8984e-18, -8.9699e-17, -3.7553e-19, 1.8132e-17, 2.6331e-18,\n -7.9154e-20, -9.1062e-17, -6.6423e-18, -1.9049e-18, -6.8698e-17,\n -4.4553e-17, 7.4145e-18, -8.8498e-19, -8.1671e-17, -1.5721e-17,\n -2.1615e-18, -5.1565e-17, -1.8203e-17, -3.9423e-17, -3.9869e-18,\n -6.9061e-17, -5.2573e-17, -1.0264e-17, 9.8817e-20, -3.2856e-17,\n -7.2604e-19, -1.8151e-18, -4.4946e-17, 7.8939e-18, -1.2393e-17,\n -1.3097e-17, -3.1257e-18, 4.8969e-18, -2.5831e-17, -8.4821e-19,\n -3.3652e-17, 2.2197e-19, -2.3260e-17, 4.1254e-18, 5.2713e-20,\n -9.3852e-19, -2.1687e-18, 5.4447e-20, -5.0709e-17, 1.0281e-18,\n -5.0675e-18, -5.7998e-18, -5.5555e-17, -6.1873e-17, -4.1273e-17,\n 1.3787e-18, -7.4924e-17, -2.1157e-17, -7.9979e-17, -7.1818e-17,\n 8.3808e-19, -8.2598e-17, 2.0815e-19, -5.0613e-17, -5.8146e-17,\n -2.0495e-18, -5.0409e-18, 6.8791e-19, -1.2977e-18, -1.3504e-17,\n -1.7650e-18, -2.2101e-17, -3.1491e-18, 1.4744e-18, 1.3763e-19,\n -4.8315e-19, -2.1070e-17, -6.5753e-17, -8.0606e-17, -3.8434e-17,\n -4.6040e-17, -4.2003e-19, 2.0995e-19, -5.8246e-18, -4.1871e-18,\n -4.1109e-18, -4.4622e-17, -2.1313e-18, -5.5984e-17, 2.0013e-18,\n -2.6691e-17, 6.6849e-18, -2.0061e-17, -4.2147e-17, 6.2711e-20,\n -4.5175e-18, -4.5786e-19, -5.0383e-18, 4.5546e-19, 4.5717e-18,\n -4.6883e-18, -6.7414e-17, -2.0878e-18, -4.7870e-17, 7.0258e-19,\n -3.9554e-19, -3.4555e-19, -5.7750e-18, -4.1293e-18, -4.9113e-17,\n 5.2092e-19], device='cuda:0')", | |
| "exp_avg_sq": "tensor([1.1527e-10, 6.1711e-12, 7.0175e-12, 2.8739e-10, 4.8150e-13, 3.0834e-12,\n 8.6793e-11, 1.0440e-13, 9.5844e-12, 4.1660e-10, 9.7045e-11, 1.3470e-10,\n 6.2874e-10, 7.2147e-11, 3.8132e-12, 2.3303e-10, 2.4488e-11, 1.0878e-11,\n 7.0383e-12, 6.3737e-13, 2.2947e-13, 2.7530e-15, 3.5504e-13, 1.7687e-10,\n 4.4818e-14, 1.5917e-09, 3.1487e-12, 1.0151e-09, 2.6049e-10, 1.4122e-11,\n 8.7874e-12, 4.2570e-10, 5.2245e-11, 5.0365e-11, 6.1383e-12, 9.1044e-12,\n 5.1656e-13, 2.6893e-10, 1.1010e-13, 8.1364e-13, 2.2909e-14, 2.7534e-11,\n 3.4772e-14, 2.8220e-15, 9.0253e-11, 2.0587e-10, 2.1245e-10, 8.6845e-10,\n 8.1373e-11, 9.2779e-11, 2.6667e-10, 2.4827e-11, 1.2837e-11, 1.6875e-12,\n 1.2139e-10, 2.8695e-11, 1.4848e-13, 9.9171e-13, 1.9261e-10, 3.4552e-10,\n 1.3895e-11, 1.2876e-13, 4.7450e-10, 6.9168e-10, 1.2691e-12, 2.6775e-13,\n 8.7488e-13, 8.8899e-11, 1.0850e-10, 3.0915e-11, 1.3224e-11, 3.5716e-13,\n 3.7667e-12, 1.3658e-11, 6.8140e-10, 3.7064e-15, 7.0973e-13, 7.6915e-13,\n 1.4521e-11, 6.2562e-12, 1.0795e-12, 6.6704e-12, 1.0145e-10, 6.2971e-14,\n 4.7166e-11, 5.5233e-12, 7.8347e-11, 4.0517e-11, 2.2064e-12, 1.2822e-10,\n 5.1048e-10, 4.1439e-11, 5.6130e-13, 1.8869e-14, 3.2964e-12, 2.1530e-12,\n 1.6100e-10, 1.9070e-10, 5.4446e-12, 1.4110e-11, 9.6143e-11, 4.4049e-11,\n 7.5011e-10, 1.5366e-13, 9.2669e-11, 6.9552e-10, 1.8911e-11, 6.7859e-15,\n 8.1711e-13, 5.0456e-13, 2.8319e-10, 1.3533e-14, 3.2109e-14, 1.6995e-10,\n 1.3353e-12, 8.6137e-11, 7.5915e-10, 1.5500e-13, 5.9206e-11, 1.9530e-10,\n 1.9626e-10, 5.7008e-12, 7.1816e-13, 9.6330e-12, 4.1587e-12, 5.4281e-10,\n 8.6754e-12, 5.8321e-11, 1.0600e-11, 3.1683e-10, 1.2399e-13, 6.2712e-13,\n 3.0176e-12, 7.0500e-11, 4.2337e-14, 4.7410e-10, 2.6277e-11, 5.4324e-15,\n 5.3906e-12, 5.2997e-13, 1.2357e-11, 6.0101e-10, 4.6227e-14, 1.9857e-13,\n 5.4289e-14, 1.0142e-12, 5.4213e-10, 2.6853e-12, 1.7471e-12, 8.3441e-12,\n 3.3692e-14, 1.6636e-09, 4.7649e-14, 2.4974e-12, 1.5409e-13, 6.5207e-11,\n 1.1905e-09, 8.9327e-12, 2.5603e-12, 2.6002e-10, 8.2201e-11, 4.8570e-13,\n 7.7373e-14, 8.7993e-10, 1.5316e-10, 2.5111e-13, 8.4696e-11, 8.4478e-11,\n 2.9969e-10, 3.3989e-11, 4.4084e-10, 2.5371e-10, 4.3382e-11, 6.8546e-13,\n 6.4859e-11, 1.3394e-13, 2.4027e-13, 2.3042e-10, 7.8574e-13, 1.2382e-10,\n 1.7862e-10, 5.6631e-14, 1.6946e-11, 7.6316e-11, 5.1805e-12, 1.6224e-10,\n 2.5502e-13, 3.4094e-11, 1.2446e-12, 1.1670e-10, 7.3100e-13, 2.5712e-12,\n 3.0533e-12, 4.7433e-10, 5.6769e-11, 3.5600e-13, 9.4074e-13, 2.4617e-10,\n 9.6966e-11, 2.0031e-10, 4.2156e-13, 1.6982e-09, 7.2621e-11, 3.7334e-10,\n 2.1426e-10, 1.0221e-10, 4.5206e-10, 9.2374e-16, 2.9154e-10, 2.7459e-10,\n 1.1596e-10, 7.6588e-14, 2.3684e-13, 3.5091e-14, 2.6872e-11, 1.7247e-10,\n 8.9561e-11, 4.0036e-14, 7.3285e-14, 3.0371e-14, 8.4405e-11, 5.7703e-11,\n 3.6632e-10, 6.1507e-10, 9.0920e-11, 1.8094e-10, 1.9407e-12, 1.3525e-11,\n 1.5431e-11, 1.9490e-11, 1.4554e-10, 3.7372e-10, 2.1034e-11, 1.2649e-10,\n 4.1049e-13, 7.5876e-11, 1.0673e-12, 3.3411e-10, 5.2744e-11, 7.7075e-13,\n 1.5248e-11, 2.5180e-10, 1.4752e-11, 8.2060e-13, 1.0940e-12, 2.6156e-13,\n 1.1956e-10, 1.4091e-13, 5.7413e-11, 1.8228e-12, 3.7708e-13, 7.2172e-11,\n 2.2170e-11, 2.0956e-13, 2.0031e-10, 2.5373e-11], device='cuda:0')" | |
| }, | |
| "40": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([[-4.9098e-20, -5.7874e-20, 3.0048e-20, ..., -1.3326e-20,\n -1.2023e-19, -1.8485e-19],\n [-3.6044e-20, -5.2637e-19, 3.0211e-19, ..., 1.6628e-20,\n -1.4716e-20, -4.5414e-20],\n [ 1.0671e-18, -2.8096e-18, 1.3841e-18, ..., 6.1389e-19,\n 3.0892e-19, 3.3590e-19],\n ...,\n [ 5.6518e-19, -1.4651e-18, 3.0809e-19, ..., 2.5908e-19,\n 6.8604e-20, 1.5865e-19],\n [ 1.9481e-19, -2.9369e-18, 7.7558e-19, ..., 6.2507e-19,\n 2.1952e-19, 1.6312e-19],\n [ 3.0723e-19, -3.7489e-19, 7.4636e-20, ..., 5.0146e-20,\n 3.0655e-20, 4.2396e-20]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[2.1496e-13, 5.9780e-14, 2.0019e-13, ..., 2.4081e-13, 2.5298e-13,\n 3.3270e-13],\n [7.7820e-15, 1.2347e-14, 6.4229e-15, ..., 7.2292e-15, 4.1218e-14,\n 1.7404e-14],\n [1.1758e-14, 1.6211e-15, 5.4776e-15, ..., 1.8095e-15, 1.1274e-14,\n 5.0793e-15],\n ...,\n [8.4350e-15, 9.2393e-16, 5.4897e-15, ..., 1.2306e-15, 6.6049e-15,\n 2.5105e-15],\n [6.6383e-13, 1.8053e-13, 4.7598e-13, ..., 5.3213e-13, 8.1630e-13,\n 9.1277e-13],\n [1.4266e-13, 4.2305e-14, 1.4118e-13, ..., 5.2735e-14, 1.9066e-13,\n 1.4890e-13]], device='cuda:0')" | |
| }, | |
| "41": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([-4.0158e-17, -9.3117e-18, 3.4753e-16, -4.5681e-17, -2.1243e-16,\n 4.7680e-17, 1.2574e-16, 3.2359e-16, -5.7696e-16, -4.9414e-16,\n 3.0465e-17, -1.1629e-16, -6.5427e-16, 6.8394e-17, 6.6641e-16,\n -5.9083e-16, -1.2911e-15, 3.4687e-17, 2.9725e-16, 7.5790e-17,\n -3.7277e-16, -9.3993e-18, 1.0677e-16, 2.1346e-17, -3.2766e-18,\n -1.6662e-15, -1.8994e-16, -1.5462e-15, -1.1039e-15, -6.9722e-16,\n 3.5279e-16, -1.2699e-15, 9.4621e-16, 9.7102e-17, 2.1523e-18,\n 3.0582e-16, -1.4923e-16, -1.5846e-17, 4.2245e-16, -1.7566e-16,\n 5.0025e-17, -3.3181e-16, 3.6866e-16, 5.9940e-16, -5.8140e-17,\n 6.6038e-16, -2.2252e-17, -1.1802e-15, 3.5163e-16, -3.2921e-16,\n 1.6900e-17, -8.5061e-17, -5.3323e-16, 9.3194e-16, 1.6009e-17,\n 1.4713e-16, 1.0901e-15, -4.0934e-16, -1.4671e-16, 5.0373e-16,\n 3.2248e-16, 5.6936e-16, -8.7056e-16, -4.2135e-17, 7.6196e-18,\n 1.9374e-16, -1.5565e-16, -1.3231e-17, 2.6105e-17, -2.3142e-16,\n 1.8125e-17, -1.5315e-17, 7.1297e-16, -1.8513e-16, 6.6127e-16,\n -1.3929e-16, 2.7014e-18, 3.4771e-16, 1.1463e-15, -1.0347e-15,\n 2.0473e-16, 3.6907e-17, 1.0339e-17, -4.3223e-16, -2.6267e-16,\n -1.2835e-16, 4.7340e-17, -2.8559e-16, 7.7906e-16, 7.5539e-18,\n 1.2472e-15, -1.6175e-16, -6.7487e-16, 2.1033e-16, 1.6358e-16,\n -9.1379e-17, -1.4013e-16, 1.9158e-16, -6.5154e-16, 1.7932e-16,\n 1.8379e-17, -5.0515e-17, -1.1053e-16, 4.7650e-16, -3.4665e-17,\n -2.4138e-16, 3.4438e-17, -2.3402e-16, -9.1967e-16, -1.1760e-16,\n 4.7485e-17, -4.3380e-16, 5.0656e-17, 3.6997e-17, -3.3445e-17,\n -4.6875e-16, -1.8936e-16, 5.7021e-16, 4.1942e-18, -9.2632e-17,\n 6.3540e-17, 2.2867e-16, 5.2336e-16, -7.4929e-17, 3.9077e-17,\n -2.8720e-16, 1.0831e-15, -4.5119e-17, 2.9807e-16, 4.8183e-16,\n 7.2526e-17, -5.2033e-16, 2.3212e-16, -1.0678e-16, -4.8450e-17,\n -1.2280e-15, 6.9671e-16, 2.5554e-18, -6.3640e-17, -3.1732e-16,\n 4.2793e-16, 8.4132e-16, 6.0771e-16, -1.1077e-17, -1.1695e-16,\n -1.2464e-16, -5.1096e-16, -4.4189e-16, 1.1039e-16, 2.1919e-16,\n 8.8220e-17, -1.4269e-15, 3.2665e-16, 1.5916e-16, -7.0424e-17,\n 4.3320e-16, 1.7453e-17, 4.2012e-16, 3.8095e-16, -6.8817e-17,\n 1.7998e-16, -1.6262e-15, -9.5199e-18, -6.3931e-18, -4.1736e-17,\n 1.8643e-16, -3.7008e-17, 1.8558e-16, 7.0889e-17, 4.9940e-18,\n -6.3725e-17, -7.8146e-16, 2.4473e-16, 3.6819e-16, 2.6049e-18,\n 2.6302e-17, -1.3342e-17, -1.3195e-17, 1.9683e-16, 4.0668e-16,\n 1.2846e-16, 2.8448e-17, -2.1436e-16, 1.5671e-17, -1.0122e-16,\n 6.3762e-17, 4.6885e-17, 8.7337e-16, -1.1780e-16, 2.4853e-16,\n 6.5016e-16, 6.3108e-16, 4.5032e-16, -2.2545e-17, 9.4921e-17,\n -8.9501e-17, 2.0602e-16, 4.3242e-17, -9.5728e-16, 1.1447e-15,\n -2.0178e-16, -9.1314e-17, -1.1427e-16, -1.4860e-15, 5.7001e-18,\n -3.2320e-17, -5.4698e-16, 4.2848e-17, 1.5098e-16, -6.2228e-17,\n 7.1159e-17, 1.8602e-16, 4.8230e-16, -2.0998e-16, 3.6194e-17,\n 3.9201e-16, 2.4131e-16, 5.8720e-16, 4.9812e-16, 3.1159e-16,\n 4.8258e-17, 6.9186e-17, -4.3475e-17, -1.1166e-16, 5.2436e-17,\n -3.0938e-18, 2.7111e-17, 3.7064e-18, -1.2408e-15, -6.0820e-20,\n 2.3249e-16, -2.0730e-16, -2.7181e-16, 1.2033e-16, 9.1363e-17,\n 5.4123e-17, 4.0001e-16, -1.7190e-16, 3.8847e-16, 5.5551e-16,\n -2.0158e-16, 4.5225e-16, -4.6295e-16, 4.2262e-17, -1.2418e-16,\n 2.1931e-16, -5.1254e-17, 1.2840e-16, 9.4716e-17, -8.5937e-17,\n 1.3922e-16, 6.4447e-17, -3.3173e-17, 1.6373e-16, 1.3053e-16,\n 1.2484e-17], device='cuda:0')", | |
| "exp_avg_sq": "tensor([4.9856e-08, 3.9765e-09, 1.9264e-09, 1.1144e-09, 1.3543e-07, 3.1012e-08,\n 3.1244e-09, 1.1819e-07, 3.5352e-07, 6.6843e-08, 4.6089e-11, 2.6136e-07,\n 3.6151e-07, 9.5516e-09, 1.1748e-08, 4.8578e-07, 1.0914e-06, 3.3370e-10,\n 3.6488e-08, 8.8399e-10, 4.3232e-07, 2.7288e-09, 4.4847e-08, 8.9629e-09,\n 1.3023e-08, 1.9634e-06, 5.9464e-08, 9.3148e-08, 9.7858e-08, 4.2782e-08,\n 2.1956e-06, 4.5264e-07, 3.2080e-07, 1.6860e-08, 1.0566e-07, 5.0387e-07,\n 7.4508e-09, 1.0550e-06, 1.7650e-07, 3.5488e-08, 2.7111e-08, 8.8194e-09,\n 4.5672e-08, 5.6500e-09, 2.5600e-07, 2.4256e-06, 2.1752e-10, 4.8624e-07,\n 2.2433e-07, 2.9529e-07, 1.5168e-08, 6.9170e-10, 5.6500e-07, 1.3719e-06,\n 6.8431e-10, 7.7212e-08, 3.7573e-08, 7.8012e-08, 5.3785e-09, 3.1111e-07,\n 9.6913e-09, 1.8633e-08, 7.3616e-07, 7.5077e-09, 3.2011e-09, 7.6198e-07,\n 3.1831e-08, 1.1282e-07, 3.0969e-09, 1.0461e-07, 7.4941e-09, 5.5805e-08,\n 2.4174e-07, 5.8006e-09, 9.2709e-07, 1.4674e-08, 1.4805e-09, 2.7513e-08,\n 8.4932e-07, 5.6640e-07, 2.2076e-09, 9.8223e-10, 5.3518e-08, 1.0223e-08,\n 8.4921e-08, 1.5144e-08, 4.0307e-09, 2.7689e-09, 1.6657e-08, 1.8806e-08,\n 1.3476e-06, 3.7966e-10, 1.4184e-07, 1.0068e-09, 9.2641e-07, 2.2024e-08,\n 7.5302e-08, 3.2428e-08, 1.7041e-07, 7.3054e-08, 1.1024e-08, 6.1088e-09,\n 6.3751e-10, 1.9459e-08, 1.4387e-07, 1.6980e-07, 8.3076e-08, 5.9932e-09,\n 2.4658e-07, 1.4723e-08, 4.5971e-08, 7.8525e-09, 3.7978e-08, 2.7755e-10,\n 8.0524e-09, 5.5278e-08, 9.6769e-08, 1.3084e-08, 3.2902e-08, 4.5536e-09,\n 6.3792e-09, 1.7465e-07, 4.1383e-08, 2.4716e-07, 1.6410e-08, 2.6348e-10,\n 6.6881e-07, 6.2801e-10, 3.1968e-07, 1.4259e-06, 2.5729e-08, 2.4546e-07,\n 1.2718e-08, 2.7906e-08, 2.3799e-07, 7.1204e-08, 9.2613e-09, 1.8824e-08,\n 8.0668e-09, 7.0917e-09, 1.2642e-07, 3.5839e-07, 4.6977e-09, 5.0827e-10,\n 4.4960e-09, 7.9428e-10, 2.6276e-07, 3.7196e-07, 3.3087e-09, 5.3180e-09,\n 1.5685e-07, 7.6053e-07, 4.6818e-08, 1.3488e-09, 3.3095e-07, 3.1877e-07,\n 1.5819e-06, 7.0873e-08, 1.3634e-06, 9.5284e-10, 1.3115e-07, 6.3136e-07,\n 6.1105e-09, 7.3647e-10, 2.0551e-07, 1.7422e-07, 1.0490e-08, 1.0585e-07,\n 1.3158e-08, 3.7266e-10, 8.2626e-09, 7.2469e-08, 2.7017e-07, 4.2707e-07,\n 4.8049e-08, 1.7809e-08, 3.4430e-10, 1.7294e-08, 7.8424e-10, 3.5611e-07,\n 4.6348e-09, 3.6334e-09, 5.2729e-08, 1.4681e-09, 5.6828e-10, 1.7025e-07,\n 5.3578e-09, 7.7039e-08, 6.5051e-08, 9.4665e-07, 5.4131e-07, 2.1957e-07,\n 1.4286e-08, 4.9301e-09, 2.0835e-10, 7.4696e-08, 2.4009e-07, 1.1235e-07,\n 2.5509e-07, 9.7224e-08, 1.1769e-08, 3.2514e-08, 3.8968e-10, 7.5207e-07,\n 2.8003e-08, 5.4994e-10, 2.9022e-07, 4.0457e-10, 7.5983e-09, 1.9169e-08,\n 3.9440e-10, 4.8892e-09, 1.5307e-07, 2.5547e-08, 3.1244e-10, 1.1969e-09,\n 3.2344e-08, 3.6086e-07, 1.1000e-08, 3.1896e-09, 3.2615e-07, 4.2900e-07,\n 2.5731e-10, 8.4350e-07, 8.2417e-10, 1.0888e-07, 1.5140e-08, 3.2018e-09,\n 2.2103e-07, 4.1778e-10, 4.1312e-07, 2.7330e-09, 1.6084e-08, 8.2041e-10,\n 2.2092e-08, 3.2479e-08, 8.9453e-07, 1.1399e-06, 2.2144e-07, 5.3769e-08,\n 1.1218e-08, 1.4266e-06, 1.9017e-08, 1.0925e-10, 3.9260e-08, 5.6933e-10,\n 7.5340e-09, 1.3348e-08, 1.1238e-07, 1.3897e-09, 5.5655e-09, 2.2890e-07,\n 1.4509e-08, 6.0920e-10, 1.4375e-07, 2.4502e-08], device='cuda:0')" | |
| }, | |
| "42": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([-4.9251e-19, -2.0922e-18, 7.6568e-20, -3.4073e-19, -8.2776e-17,\n -3.5624e-19, 2.0627e-19, -4.1008e-18, -8.8077e-17, -9.5308e-17,\n -1.3016e-18, -4.6184e-19, -4.7576e-17, -3.4506e-19, -5.0503e-18,\n -1.1069e-16, -6.9186e-17, 3.2326e-19, 4.4333e-19, 6.2098e-19,\n -1.0000e-16, -1.4405e-19, 1.7337e-19, -1.7988e-17, -5.1188e-19,\n -1.1768e-16, 2.4849e-18, -3.3246e-17, -4.4731e-17, -4.6380e-17,\n -5.5957e-18, -7.2400e-17, -1.3622e-17, -7.3157e-20, 4.9047e-19,\n -4.4491e-20, -1.5725e-17, -2.7058e-18, -4.7412e-17, -6.8271e-19,\n 3.0915e-19, -2.8710e-17, -1.8303e-20, -3.8410e-19, -2.6723e-18,\n -1.3022e-20, -1.0190e-18, -1.4832e-16, -8.1423e-19, -8.8507e-17,\n -3.4452e-19, -5.6815e-19, -9.6755e-17, -1.2066e-18, 6.2772e-19,\n 3.6762e-19, -1.2706e-19, -4.3135e-17, -3.5532e-18, -8.4556e-17,\n 1.2007e-19, 3.4233e-19, -9.2570e-17, -2.7492e-18, -1.2509e-18,\n -2.8832e-18, -3.8568e-17, 1.7914e-19, -1.1223e-17, 9.1144e-19,\n -3.8184e-19, 8.1015e-19, -5.8643e-18, -3.3373e-18, -7.2606e-17,\n -4.7666e-19, -7.3958e-20, -2.4648e-19, 1.3302e-18, -6.7102e-17,\n 3.6730e-19, -5.8795e-19, 4.4100e-19, 2.8582e-18, -1.0569e-16,\n -1.3648e-18, -8.2501e-20, 1.9756e-18, -4.1452e-19, -4.5703e-17,\n -3.3532e-17, -5.3164e-19, -4.5791e-17, 4.0209e-19, -1.0665e-16,\n -1.2363e-17, -2.2207e-19, -8.5822e-17, -2.8399e-17, -6.7214e-17,\n -2.8661e-19, -1.1106e-18, -1.6253e-18, -1.9554e-17, -7.0559e-19,\n -3.2760e-17, -1.0010e-16, 6.6374e-19, -2.5129e-17, -1.5015e-18,\n -2.1282e-17, 5.4411e-18, -3.5704e-19, -2.0020e-19, -1.5401e-18,\n -2.1153e-17, -1.1539e-16, -1.0749e-18, -8.7768e-19, -1.0395e-18,\n 2.6502e-19, -1.7946e-19, -3.2470e-17, -3.5771e-17, -1.4399e-19,\n -2.9563e-18, 2.7139e-18, 8.6618e-20, -5.9028e-20, -3.2400e-19,\n 6.0843e-19, -1.1013e-16, -2.0211e-19, -2.2025e-18, -1.1275e-17,\n -6.2783e-17, -7.5782e-18, -1.8045e-20, 1.1234e-18, 2.5638e-18,\n 3.0062e-20, -5.8803e-17, 9.1724e-20, -7.3673e-19, -1.1270e-19,\n -5.0385e-18, -3.5851e-17, -1.0921e-16, 2.8718e-19, 4.5402e-19,\n -1.1386e-19, -6.8653e-17, 6.7533e-20, 1.0265e-19, -8.2541e-17,\n 4.0982e-20, -1.0404e-16, -6.2609e-18, -3.0602e-19, -4.3106e-20,\n 2.5359e-19, -1.3496e-16, -2.5908e-18, -9.7327e-19, -2.4977e-17,\n 4.3392e-20, -1.3892e-18, -9.3351e-19, -2.7954e-18, -2.4009e-19,\n -3.3383e-18, -4.1251e-17, 8.2223e-20, 8.0481e-20, -6.5894e-17,\n -6.6404e-20, 4.8451e-19, -1.2697e-18, -1.1062e-19, -2.4652e-17,\n 7.9965e-19, 5.0679e-19, 2.2936e-18, 7.6840e-19, 9.7334e-19,\n 9.2126e-20, 4.7528e-19, -7.7693e-18, 2.9900e-19, -3.7647e-19,\n -1.9670e-18, -2.2410e-17, 1.6595e-19, -4.9259e-18, 1.0997e-19,\n -2.8706e-17, -3.0329e-17, -4.5660e-17, -9.5475e-17, -4.2586e-17,\n 2.2125e-18, -7.1186e-18, -2.0636e-18, -1.7593e-16, -4.7511e-17,\n 8.9949e-19, -1.3718e-16, -1.1466e-19, 2.3185e-19, -1.0178e-17,\n 1.3202e-20, -7.4887e-18, -1.7406e-18, -1.0273e-18, -6.0551e-19,\n -3.7975e-19, 9.3551e-21, -8.3397e-20, -3.3135e-19, -8.4776e-20,\n 5.3210e-20, 2.8114e-19, -2.3490e-18, -5.7399e-17, 7.4413e-19,\n -8.6193e-19, -2.5280e-19, 1.2715e-19, -1.1912e-16, 8.8050e-19,\n 5.9595e-20, -2.2393e-18, 2.4085e-19, -4.1924e-19, -4.7718e-19,\n -3.8397e-19, -5.3206e-20, -1.2292e-17, -5.8889e-17, 1.0867e-19,\n -1.0025e-18, -8.6302e-21, 5.1352e-18, -1.8196e-19, 4.8209e-19,\n -2.3941e-19, -2.6428e-18, -3.6032e-19, -8.5247e-17, -7.1289e-18,\n 1.0352e-19, 1.1251e-20, -3.8501e-17, 2.3245e-19, -1.7667e-17,\n -3.3833e-19], device='cuda:0')", | |
| "exp_avg_sq": "tensor([3.7742e-13, 1.2501e-11, 1.6793e-14, 5.9058e-12, 5.9008e-10, 1.9106e-13,\n 2.5896e-13, 2.7793e-11, 6.7884e-10, 5.6685e-10, 1.2514e-13, 4.2964e-13,\n 7.9892e-11, 3.4402e-12, 9.5961e-11, 1.1260e-09, 1.8450e-10, 1.5435e-14,\n 9.7179e-13, 1.3108e-12, 1.0762e-09, 2.4794e-13, 1.2597e-11, 5.1555e-13,\n 1.2774e-11, 1.3630e-09, 4.0212e-14, 4.3400e-13, 1.1872e-11, 2.1112e-10,\n 9.8265e-10, 3.1448e-10, 7.4929e-11, 1.1156e-12, 2.6166e-13, 3.9776e-11,\n 1.2243e-10, 2.4948e-10, 3.0194e-10, 6.4727e-14, 6.9303e-13, 3.3618e-12,\n 5.5782e-12, 1.9020e-13, 7.6203e-11, 1.7269e-10, 3.7248e-14, 1.7859e-09,\n 7.9265e-12, 6.0630e-10, 3.0740e-12, 1.2649e-14, 9.5514e-10, 4.7933e-10,\n 1.8485e-12, 2.6435e-14, 1.1084e-11, 3.5644e-11, 1.3372e-11, 1.4844e-09,\n 6.8059e-13, 1.6287e-12, 6.9242e-10, 1.4439e-11, 7.9430e-12, 1.8560e-10,\n 2.8759e-11, 7.6948e-13, 1.3275e-13, 9.7037e-14, 5.7082e-12, 1.0894e-14,\n 1.2732e-11, 2.0780e-14, 2.4385e-09, 2.2918e-15, 2.7867e-14, 6.6716e-13,\n 8.6606e-11, 1.8923e-10, 3.3049e-13, 3.1985e-12, 2.1972e-12, 9.3018e-14,\n 7.3905e-10, 9.5446e-15, 1.6158e-13, 5.0169e-15, 1.6765e-12, 9.6884e-11,\n 1.7172e-09, 2.0169e-14, 3.7534e-10, 1.7118e-14, 2.7402e-09, 4.4728e-11,\n 4.1673e-14, 5.5417e-10, 2.8451e-10, 3.3232e-10, 2.9933e-12, 7.1956e-12,\n 4.5544e-12, 4.6705e-11, 7.9079e-13, 1.9927e-11, 8.1289e-10, 2.5535e-16,\n 1.8137e-10, 6.6064e-15, 5.5237e-12, 2.9719e-14, 1.2120e-12, 8.3218e-14,\n 5.5776e-12, 8.2998e-11, 9.2839e-10, 4.2756e-12, 2.3713e-15, 2.1335e-12,\n 1.0155e-12, 7.6061e-12, 2.9236e-10, 7.5058e-11, 2.2368e-13, 2.2192e-13,\n 4.4973e-11, 8.2415e-14, 9.2152e-12, 2.9768e-10, 8.2964e-13, 1.0904e-09,\n 1.6892e-12, 9.3252e-15, 1.0233e-10, 6.6246e-11, 8.2747e-11, 2.8687e-12,\n 7.2164e-15, 2.6200e-15, 2.1274e-12, 1.0677e-09, 1.8367e-13, 5.5032e-14,\n 7.5456e-14, 5.0447e-14, 2.8291e-11, 9.8274e-10, 7.9589e-13, 1.0056e-14,\n 7.5339e-12, 1.4459e-10, 5.5655e-14, 7.0949e-13, 8.4402e-10, 2.0075e-11,\n 3.2502e-09, 8.6526e-12, 4.3939e-11, 2.9164e-14, 1.0567e-11, 1.1616e-09,\n 1.9802e-13, 5.7806e-12, 2.7763e-11, 1.1700e-12, 2.9228e-12, 2.3875e-11,\n 6.6890e-14, 5.5042e-14, 1.6063e-11, 1.4106e-11, 6.4702e-12, 3.0980e-11,\n 2.8926e-10, 2.3651e-15, 3.9674e-13, 6.7970e-12, 2.2863e-13, 1.0807e-10,\n 1.5392e-12, 3.1189e-13, 1.6578e-13, 2.5462e-13, 5.0321e-16, 4.5689e-12,\n 3.3819e-15, 1.0360e-10, 4.1047e-14, 1.2134e-10, 2.6622e-11, 1.6679e-10,\n 1.0864e-12, 2.1468e-11, 2.0732e-14, 7.5409e-11, 1.0355e-10, 1.3932e-10,\n 5.6850e-10, 7.3479e-10, 1.4614e-14, 4.7867e-13, 7.1813e-15, 2.3308e-09,\n 1.9311e-10, 8.9597e-14, 1.4895e-09, 4.8890e-15, 4.9621e-14, 8.4088e-14,\n 7.2820e-14, 1.0388e-10, 1.8918e-12, 1.4760e-14, 6.7656e-15, 2.5919e-14,\n 5.7868e-14, 8.4521e-12, 1.7746e-12, 6.3313e-13, 1.1788e-11, 2.0486e-11,\n 8.0760e-15, 4.9384e-10, 1.8764e-13, 7.4982e-13, 1.8370e-15, 3.8173e-13,\n 8.9646e-10, 3.8672e-14, 2.7458e-11, 8.3046e-14, 1.3421e-13, 4.7186e-14,\n 1.6756e-13, 7.8477e-15, 1.0553e-10, 4.8736e-10, 5.6461e-10, 1.2257e-12,\n 1.8619e-14, 2.2267e-10, 1.5362e-13, 4.8708e-15, 3.0726e-14, 5.9954e-14,\n 1.0077e-11, 2.2765e-13, 7.0647e-10, 1.5446e-13, 1.2714e-13, 1.4580e-12,\n 3.5639e-11, 6.6431e-14, 8.8616e-12, 2.3243e-12], device='cuda:0')" | |
| }, | |
| "43": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([-2.1277e-18, -6.6418e-19, -8.5801e-19, -3.4484e-18, -5.1043e-17,\n -7.0512e-19, -4.8806e-18, -1.1336e-17, -5.7272e-17, -5.7548e-17,\n -1.3267e-17, 1.7662e-19, -4.8043e-17, -9.8943e-19, -9.0936e-18,\n -6.1960e-17, -6.4616e-17, 1.9515e-19, 3.6655e-18, -6.7978e-18,\n -5.6649e-17, -8.3394e-18, -7.0396e-18, -2.6094e-17, -1.3511e-17,\n -8.0221e-17, -1.4685e-18, -5.9407e-17, -5.5028e-17, -4.8458e-17,\n -1.2640e-17, -6.4927e-17, -1.2750e-17, -4.1579e-19, 4.8486e-21,\n -3.1440e-18, -2.7340e-17, -1.5116e-17, -3.2766e-17, 4.3013e-19,\n -2.5478e-18, -3.6649e-17, 3.9290e-18, 6.4700e-18, -1.5889e-17,\n 8.0586e-18, -1.0187e-18, -7.7763e-17, -5.1398e-18, -5.3730e-17,\n -1.5621e-18, 4.1591e-19, -5.8483e-17, -2.6974e-18, -5.5276e-18,\n -2.3653e-18, 4.6474e-19, -4.2619e-17, 2.5166e-18, -4.1725e-17,\n -5.7108e-20, 6.0340e-18, -6.2198e-17, 1.4463e-18, -4.1560e-19,\n -1.1836e-17, -3.7427e-17, -1.2593e-19, -2.2241e-17, -6.8040e-19,\n -1.3469e-18, 3.8706e-20, -1.0048e-17, 2.5173e-18, -3.7351e-17,\n 3.2681e-19, -1.3444e-18, 2.8044e-18, 2.3648e-18, -5.9523e-17,\n -3.6070e-18, -2.0502e-18, -4.9826e-18, -2.0418e-18, -5.6375e-17,\n 7.6026e-19, -7.9087e-18, -1.4497e-18, 1.1298e-17, -3.7404e-17,\n -1.9584e-17, 4.0171e-19, -4.7900e-17, -2.5827e-18, -5.0777e-17,\n -2.5700e-17, 9.3332e-20, -4.5960e-17, -4.1254e-17, -4.1607e-17,\n -5.4397e-19, 3.7136e-19, -8.4883e-18, -2.0893e-17, 4.7815e-20,\n -3.6430e-17, -5.1119e-17, -5.2348e-19, -4.4359e-17, 1.0991e-18,\n -2.7607e-17, -3.6800e-18, -2.1415e-18, -5.6078e-19, -1.9302e-18,\n -3.5710e-17, -5.7272e-17, 5.7883e-18, 2.1892e-19, -9.1459e-18,\n -7.0091e-18, -5.9840e-18, -2.6474e-17, -3.5532e-17, -7.9362e-19,\n -1.8868e-17, 8.5146e-18, -1.2046e-19, -4.4672e-19, -2.2876e-18,\n 5.5989e-19, -6.1170e-17, -2.7416e-18, 1.0810e-18, -2.3385e-17,\n -6.1267e-17, -1.1708e-17, -1.5341e-19, -2.3498e-19, -1.8905e-18,\n 1.2442e-18, -3.1583e-17, 5.8046e-18, -7.0767e-19, 3.2678e-20,\n -2.0517e-17, -4.2040e-17, -5.9495e-17, -5.8032e-18, -1.6446e-18,\n -4.5105e-18, -6.6611e-17, 1.2814e-19, 1.8578e-18, -4.8452e-17,\n 1.2424e-18, -5.2323e-17, -1.2895e-17, 3.2212e-18, -1.0998e-17,\n -3.6304e-18, -8.3115e-17, -1.5203e-17, -2.0239e-18, -3.0638e-17,\n -1.4760e-18, -3.2978e-18, -7.6853e-18, -1.5580e-17, -1.2877e-18,\n 2.0702e-18, -4.8294e-17, -3.2599e-18, -1.1834e-18, -4.3625e-17,\n -3.1719e-20, -8.1685e-18, -6.7607e-19, 7.0717e-19, -2.4444e-17,\n -4.5912e-18, -8.1868e-18, -1.7308e-18, -7.1124e-18, -5.5770e-19,\n -3.0373e-18, -5.6978e-18, -1.0198e-17, -2.0338e-19, -3.9343e-18,\n -5.2960e-18, -2.0468e-17, 3.7482e-18, 2.9597e-18, 1.0489e-18,\n -3.2800e-17, -2.9295e-17, -3.7080e-17, -6.4995e-17, -2.3491e-17,\n -1.3726e-18, -2.0581e-17, 1.5635e-18, -8.6767e-17, -3.8423e-17,\n -3.0989e-19, -6.6714e-17, -8.5320e-20, -5.8254e-18, -2.3759e-17,\n -2.1542e-19, -1.7323e-17, -6.1751e-18, 7.1312e-19, -4.0869e-19,\n -2.0556e-18, -2.1243e-18, 5.3231e-18, 4.2167e-18, 7.3203e-19,\n -3.0829e-18, -3.1880e-18, 1.7523e-18, -4.3018e-17, -3.0801e-18,\n -8.2621e-19, -1.2744e-18, -8.5400e-18, -7.3964e-17, -4.8255e-18,\n -1.1959e-18, -1.8283e-17, -1.5856e-19, -7.7282e-18, -1.7255e-18,\n -9.3382e-20, -1.2604e-18, -2.5728e-17, -3.7214e-17, -8.5841e-20,\n 6.9205e-19, -1.4814e-18, -3.7924e-18, -9.5534e-20, -2.2829e-19,\n 1.6774e-18, 6.1202e-19, -2.2089e-18, -4.7088e-17, -2.0534e-17,\n -5.2308e-18, 8.8580e-20, -3.5765e-17, -3.4682e-18, -2.4814e-17,\n -8.2162e-18], device='cuda:0')", | |
| "exp_avg_sq": "tensor([1.8964e-11, 5.7195e-12, 1.0054e-14, 3.2533e-12, 2.4181e-10, 8.4640e-14,\n 1.4861e-13, 2.1260e-11, 4.9371e-10, 2.8230e-10, 5.7134e-12, 6.9067e-11,\n 3.3247e-10, 1.6109e-12, 7.1262e-11, 6.0495e-10, 7.1283e-10, 3.0100e-15,\n 4.6405e-13, 2.0678e-11, 5.8169e-10, 1.1218e-13, 3.1737e-11, 3.2166e-11,\n 4.1142e-11, 1.3489e-09, 6.0622e-12, 9.6802e-11, 1.3163e-10, 1.4766e-10,\n 5.5869e-10, 4.7372e-10, 2.5767e-10, 5.8753e-13, 1.3576e-11, 8.9162e-11,\n 9.2694e-11, 3.0528e-10, 2.3333e-10, 1.0897e-11, 6.5680e-12, 4.9020e-11,\n 2.5547e-12, 8.0547e-14, 1.0716e-10, 6.7776e-10, 1.2737e-12, 7.0558e-10,\n 9.7433e-12, 4.2120e-10, 1.4146e-12, 4.9514e-13, 6.0956e-10, 2.2127e-10,\n 1.0839e-12, 2.0867e-13, 4.7866e-11, 1.3782e-10, 6.8694e-12, 5.2009e-10,\n 3.6091e-13, 8.5854e-13, 6.7566e-10, 6.2643e-12, 4.0556e-12, 1.7677e-10,\n 9.8261e-11, 2.2408e-11, 2.4031e-11, 2.4388e-11, 2.8373e-12, 3.5124e-12,\n 6.5870e-12, 1.4984e-12, 8.7589e-10, 1.4373e-12, 1.0936e-12, 3.4716e-13,\n 5.4980e-11, 4.6304e-10, 1.7571e-13, 1.5837e-12, 2.0452e-11, 2.6610e-12,\n 3.0831e-10, 2.5786e-12, 8.1312e-14, 2.1245e-13, 9.2280e-13, 1.2285e-10,\n 8.4251e-10, 3.4445e-13, 2.3166e-10, 7.1038e-15, 9.2247e-10, 9.1652e-11,\n 2.1593e-11, 2.3846e-10, 1.4874e-10, 2.3029e-10, 1.5411e-12, 3.6124e-12,\n 2.2140e-12, 8.3452e-11, 2.4137e-11, 1.7214e-10, 3.2386e-10, 7.8019e-15,\n 1.5787e-10, 1.1273e-12, 7.3137e-11, 1.1173e-12, 1.0816e-11, 2.0771e-14,\n 2.5541e-12, 1.1867e-10, 3.5394e-10, 2.2048e-12, 3.5739e-13, 1.1176e-12,\n 5.5879e-13, 1.0939e-11, 1.2266e-10, 2.4820e-10, 6.3324e-12, 2.7370e-13,\n 5.5941e-11, 4.6699e-14, 3.7942e-11, 3.1079e-10, 3.7486e-13, 4.8585e-10,\n 8.8117e-13, 7.5581e-12, 5.9823e-11, 1.6082e-10, 9.4245e-11, 1.3613e-12,\n 3.5760e-13, 1.1314e-13, 4.4172e-12, 4.3705e-10, 8.2365e-14, 8.0080e-12,\n 2.3518e-12, 1.5022e-11, 2.3420e-10, 5.2335e-10, 4.3174e-13, 4.7182e-12,\n 3.5785e-11, 5.5284e-10, 5.2395e-12, 1.4204e-13, 4.7332e-10, 4.2501e-11,\n 1.2852e-09, 5.4196e-11, 3.4427e-10, 1.6381e-12, 3.5241e-11, 7.1477e-10,\n 8.5269e-12, 3.1044e-12, 1.7951e-10, 1.0694e-11, 1.3751e-12, 3.4377e-11,\n 1.8742e-11, 1.8051e-14, 7.1374e-12, 1.1303e-10, 1.7095e-11, 5.5780e-11,\n 1.7896e-10, 5.4938e-13, 1.5854e-13, 2.7608e-12, 5.1081e-14, 2.5771e-10,\n 7.9864e-13, 1.7141e-13, 6.7729e-12, 1.0647e-13, 2.5842e-16, 4.9778e-11,\n 5.8375e-12, 4.8021e-11, 9.0034e-12, 2.4635e-10, 2.1030e-11, 6.3951e-11,\n 5.5617e-13, 8.8317e-12, 1.2525e-13, 7.3831e-11, 2.5331e-10, 1.9423e-10,\n 4.1774e-10, 2.7727e-10, 5.4337e-13, 2.6608e-11, 1.0370e-13, 8.7249e-10,\n 1.3104e-10, 3.6236e-14, 5.5331e-10, 1.7060e-15, 7.3663e-12, 2.9446e-11,\n 9.8233e-15, 8.2781e-11, 2.0649e-12, 4.7939e-12, 5.9688e-14, 2.3046e-12,\n 6.5616e-12, 3.7145e-11, 9.7853e-13, 2.6896e-13, 7.2026e-11, 1.0447e-10,\n 1.0420e-13, 6.3254e-10, 6.2910e-14, 3.7158e-11, 9.0842e-13, 1.7314e-13,\n 4.3459e-10, 3.6639e-14, 9.4573e-11, 3.9613e-12, 4.9560e-12, 2.1513e-12,\n 7.3112e-14, 4.4009e-13, 2.0662e-10, 2.6257e-10, 3.2831e-10, 1.1098e-11,\n 2.2236e-12, 3.3938e-10, 4.5276e-12, 1.0639e-13, 5.7530e-12, 1.9389e-14,\n 4.5807e-12, 1.2218e-13, 3.3244e-10, 6.1704e-12, 6.6343e-14, 4.4900e-11,\n 9.2617e-11, 1.6558e-14, 1.3003e-10, 1.1818e-11], device='cuda:0')" | |
| }, | |
| "44": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([[-2.4554e-18, 4.3539e-18, -2.2824e-18, ..., -6.9929e-19,\n -8.1957e-19, -5.5429e-19],\n [ 1.9614e-19, -8.0539e-18, 2.5082e-19, ..., 9.0741e-20,\n -4.8904e-20, 5.1187e-20],\n [ 4.2463e-19, 3.2332e-19, -8.3394e-23, ..., 2.9339e-19,\n -1.7526e-21, 1.5272e-19],\n ...,\n [-1.1489e-19, 1.6015e-18, -4.7303e-20, ..., -1.3483e-19,\n -3.7796e-20, -1.1380e-19],\n [ 4.2254e-19, -9.3370e-19, 2.3584e-19, ..., 1.9772e-19,\n 8.7046e-21, 1.7242e-19],\n [ 2.7048e-18, -4.7042e-19, 2.8352e-18, ..., 1.1226e-18,\n 8.5141e-19, 7.3647e-19]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[8.6361e-13, 2.7420e-13, 5.6607e-13, ..., 1.0598e-12, 8.2505e-13,\n 1.1576e-12],\n [1.8957e-14, 1.0054e-14, 3.6343e-15, ..., 3.0096e-14, 1.0972e-14,\n 2.6782e-14],\n [1.2539e-15, 2.2654e-15, 2.4285e-15, ..., 4.0922e-15, 1.4745e-15,\n 7.4497e-16],\n ...,\n [1.0794e-13, 3.2676e-14, 7.2653e-14, ..., 6.2587e-14, 1.1673e-13,\n 1.4883e-13],\n [3.9667e-15, 5.3318e-15, 1.9045e-14, ..., 2.4847e-14, 4.1574e-14,\n 7.9364e-14],\n [2.5315e-13, 1.2917e-13, 1.8169e-13, ..., 2.1711e-13, 3.4416e-13,\n 4.8246e-13]], device='cuda:0')" | |
| }, | |
| "45": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([-7.1095e-16, 6.5551e-17, 4.3838e-17, 9.7025e-17, 1.4663e-16,\n 1.5839e-16, -1.1001e-16, 7.9454e-17, -1.1240e-15, -1.9663e-16,\n -9.8506e-17, 1.5253e-16, -3.6501e-16, -5.1934e-16, -4.2574e-18,\n -3.7997e-16, -5.4327e-16, 1.2563e-16, -1.6125e-16, 1.3915e-16,\n 9.1698e-18, 6.4887e-16, 2.3023e-17, 4.6821e-16, -7.1487e-16,\n -1.8040e-17, 4.5325e-17, -5.9314e-16, -1.8102e-16, -1.4192e-16,\n 2.5457e-16, -9.6877e-17, 9.3471e-17, 1.8166e-17, 3.9470e-17,\n 1.6851e-16, -1.0837e-16, 1.2902e-16, 5.1457e-16, -4.2191e-17,\n 3.0598e-16, 8.1061e-17, 4.0030e-16, 1.0772e-16, 1.0825e-16,\n 3.6337e-16, -3.7607e-16, -2.4552e-16, -1.2999e-16, 1.1572e-16,\n -1.4375e-16, 5.6454e-16, -2.2798e-16, -1.4774e-16, -3.8492e-16,\n 1.2103e-16, 3.4756e-16, 2.8660e-16, -5.3930e-17, -8.6746e-17,\n 1.4138e-16, -2.6742e-17, -1.2094e-15, -3.5099e-16, -7.0463e-16,\n 1.8442e-16, 1.8422e-17, 1.6652e-17, -1.8915e-17, 6.0776e-17,\n -6.6262e-16, 3.4650e-16, 1.4055e-16, -2.6372e-17, -1.1494e-15,\n -1.8513e-17, -3.9721e-16, -1.5212e-16, 6.8786e-16, -5.7975e-16,\n -9.8417e-18, -8.0050e-18, 5.9550e-17, 6.2694e-18, 3.8722e-16,\n 1.5372e-16, -6.7330e-19, -6.5958e-17, 1.0547e-15, -1.0416e-17,\n 1.2670e-16, 3.6412e-17, 8.3178e-17, 6.8549e-16, -1.7021e-16,\n -9.9893e-17, -2.8524e-16, -1.4455e-16, 2.5053e-16, -3.8993e-16,\n 1.0138e-16, 6.8791e-16, -8.3188e-16, -1.8473e-17, 3.4849e-16,\n 3.2921e-16, -7.7372e-16, 3.4158e-16, -3.1750e-17, 6.9734e-16,\n -5.1366e-16, 5.4148e-16, -6.5751e-18, 3.7255e-17, -1.4692e-15,\n -1.0425e-16, 3.1537e-17, -2.9408e-16, 2.1652e-16, -6.0841e-16,\n 9.2332e-17, 7.5922e-17, 7.2328e-17, 9.7496e-16, -1.7552e-16,\n -5.6095e-16, -1.2704e-16, 1.6080e-16, 1.5202e-16, 2.7504e-16,\n 1.8784e-16, -7.6247e-17, -1.5764e-16, 4.2085e-17, -1.0316e-16,\n -2.0496e-16, 6.1974e-18, 2.9381e-16, 2.5629e-16, -2.5343e-16,\n 2.4003e-16, -2.8259e-16, 2.3560e-16, -4.4532e-16, -5.1959e-16,\n -3.8785e-17, -1.1128e-16, -6.1822e-16, 5.5190e-16, -9.5130e-17,\n -2.9435e-16, -2.7464e-16, -2.3219e-17, 5.4921e-16, 2.5645e-16,\n -8.5335e-17, -1.0298e-16, -1.5188e-16, 1.8331e-16, -1.0072e-16,\n 2.4134e-16, -1.0403e-15, -8.5500e-17, -5.2206e-16, -9.2318e-17,\n 3.2792e-16, -3.6633e-16, -1.0040e-16, -3.8482e-16, 2.2060e-16,\n 3.7042e-16, 7.8509e-18, -4.9890e-17, -1.2317e-16, -1.9044e-17,\n 6.6250e-17, -2.1876e-17, -3.5783e-17, -1.9431e-16, 1.1946e-16,\n -1.0600e-16, 2.2261e-16, 5.2403e-16, -6.6796e-18, -1.7116e-16,\n 1.0678e-16, 9.3919e-17, 2.7972e-16, 1.5025e-16, -7.6492e-18,\n 5.7550e-16, 5.6506e-17, -1.3212e-16, -1.1486e-15, 4.0825e-16,\n 4.6980e-16, -9.2570e-18, 1.8099e-17, -1.9102e-16, 6.0066e-16,\n 3.5738e-16, -6.8579e-16, -1.4671e-16, 1.5482e-17, 3.8909e-16,\n 1.9706e-17, -3.2794e-16, 4.9848e-16, -4.2016e-17, -8.3839e-17,\n 2.7793e-16, -3.8493e-17, 1.0697e-16, 1.3632e-16, 1.2976e-16,\n -5.2409e-17, -6.1002e-17, 3.3644e-16, 6.3800e-16, 1.4939e-16,\n 4.9739e-16, 2.7043e-16, -1.1068e-16, -2.8487e-16, -5.0421e-17,\n -4.3750e-17, -1.6096e-16, -4.2159e-16, 6.7437e-18, 7.0701e-17,\n -4.6122e-17, 3.8519e-17, 5.9064e-16, 2.4156e-18, 2.6256e-16,\n 1.8036e-17, -3.6726e-17, 4.1924e-16, 2.1887e-18, -3.5306e-17,\n -5.8543e-17, 2.8532e-16, 3.5165e-16, 1.3831e-16, 2.9555e-16,\n 2.1742e-16, -4.4573e-16, 3.7433e-16, -3.7629e-16, -4.5914e-17,\n 3.9691e-16, 5.2601e-16, 8.8439e-17, -7.5060e-17, 8.2697e-17,\n 8.2882e-16], device='cuda:0')", | |
| "exp_avg_sq": "tensor([1.9217e-07, 4.9320e-09, 1.4838e-10, 5.4201e-07, 3.5660e-07, 7.6307e-09,\n 5.5555e-10, 4.8005e-08, 5.1502e-07, 7.8272e-08, 7.9364e-08, 5.8172e-07,\n 5.5081e-07, 6.8894e-09, 1.8384e-09, 2.2132e-09, 2.6187e-09, 3.2201e-09,\n 4.4299e-09, 7.0498e-08, 3.4889e-10, 5.0339e-08, 1.1207e-08, 1.7097e-08,\n 1.4915e-08, 4.7817e-09, 2.1989e-08, 6.2132e-07, 9.0972e-08, 1.3011e-09,\n 1.6403e-06, 5.6615e-10, 3.8096e-07, 1.3449e-08, 8.3977e-08, 1.1079e-06,\n 8.4876e-09, 1.1843e-08, 1.7350e-07, 3.2994e-10, 6.5492e-08, 1.3351e-07,\n 2.3169e-09, 1.8401e-08, 1.3377e-09, 8.8266e-10, 2.8012e-09, 3.6957e-07,\n 5.6240e-10, 1.1157e-07, 7.2089e-08, 6.0123e-08, 5.4726e-08, 1.0243e-08,\n 1.2093e-08, 4.8955e-08, 9.4105e-10, 8.9168e-09, 8.8726e-07, 1.3873e-08,\n 1.7866e-07, 1.0068e-07, 3.3996e-07, 1.1853e-06, 5.2582e-07, 5.1996e-07,\n 3.3398e-11, 3.7381e-10, 2.5408e-09, 1.4507e-07, 1.0234e-06, 1.0554e-07,\n 2.7950e-08, 1.3668e-08, 1.4935e-07, 5.4265e-09, 4.2970e-09, 1.3469e-08,\n 4.1509e-07, 6.2032e-07, 1.6880e-10, 1.1834e-09, 4.2176e-08, 1.5513e-07,\n 3.7534e-07, 1.6533e-08, 5.2727e-11, 1.6431e-10, 7.4145e-09, 1.4238e-07,\n 2.7334e-07, 3.5218e-08, 2.0552e-07, 3.9657e-07, 9.0853e-10, 5.3533e-10,\n 9.0773e-09, 1.9815e-07, 1.0215e-07, 4.4472e-08, 1.1728e-07, 6.6049e-07,\n 4.0672e-07, 1.5626e-08, 1.4582e-07, 1.5167e-07, 2.2291e-08, 2.1025e-08,\n 6.1921e-10, 5.4555e-07, 1.7234e-07, 1.0139e-08, 7.2725e-08, 1.0357e-08,\n 2.6343e-07, 8.2224e-09, 1.7983e-10, 4.8230e-09, 5.7875e-10, 3.9449e-09,\n 3.6895e-10, 1.2646e-08, 1.5397e-09, 5.9264e-07, 2.0203e-09, 1.5335e-07,\n 5.4918e-08, 3.5175e-07, 3.5738e-08, 1.7962e-06, 4.7188e-08, 4.3936e-10,\n 5.0740e-08, 2.9008e-09, 2.8437e-09, 3.9030e-07, 8.4811e-10, 9.9700e-09,\n 1.2960e-08, 4.6685e-10, 1.4827e-07, 1.0558e-08, 5.3471e-10, 2.4594e-08,\n 6.3690e-07, 2.1505e-10, 1.3984e-08, 2.3832e-08, 3.3943e-07, 1.0120e-08,\n 8.0094e-07, 6.6462e-09, 3.5193e-08, 3.7890e-09, 8.5596e-08, 6.2946e-08,\n 3.0898e-09, 1.6841e-07, 1.1497e-06, 1.0062e-08, 9.9729e-08, 1.4343e-07,\n 2.0387e-07, 8.4003e-07, 4.6495e-08, 2.7887e-07, 2.3779e-07, 3.1317e-09,\n 3.0313e-07, 2.7847e-07, 6.3436e-07, 1.2629e-09, 2.4678e-08, 1.9107e-08,\n 4.7289e-09, 3.6247e-08, 2.8695e-10, 6.8892e-09, 2.9613e-08, 1.9779e-08,\n 3.0934e-07, 4.3194e-07, 1.2836e-08, 1.8002e-09, 9.6327e-09, 1.1889e-06,\n 6.4730e-09, 3.9672e-09, 3.2035e-07, 2.2349e-07, 3.2190e-07, 2.2659e-09,\n 3.4561e-10, 5.4350e-07, 3.0273e-08, 1.0981e-07, 1.0749e-08, 1.2292e-07,\n 6.3567e-07, 8.7670e-08, 1.0754e-09, 3.3628e-06, 2.1906e-09, 4.1865e-07,\n 3.0225e-07, 1.2765e-07, 2.3918e-07, 1.4246e-07, 1.1109e-08, 5.1440e-09,\n 4.7387e-07, 1.9432e-10, 1.1573e-07, 2.1935e-10, 3.1969e-08, 2.9886e-09,\n 1.4629e-09, 2.1877e-08, 3.7291e-07, 2.0174e-09, 8.1750e-07, 1.6836e-08,\n 2.6776e-09, 9.5929e-07, 2.5369e-10, 1.1245e-08, 1.1482e-07, 5.2086e-08,\n 8.5998e-09, 8.1053e-09, 8.2062e-10, 6.8062e-08, 1.8866e-08, 4.1177e-09,\n 3.0284e-07, 3.0162e-08, 3.4652e-10, 1.8882e-06, 1.3673e-09, 1.5391e-09,\n 1.0899e-07, 1.9493e-06, 1.7039e-09, 1.3180e-07, 2.5514e-07, 1.7005e-09,\n 7.5290e-07, 1.2753e-07, 2.4558e-07, 9.1988e-11, 2.6204e-08, 9.7925e-07,\n 8.3357e-09, 3.0299e-08, 4.5938e-09, 9.0030e-08], device='cuda:0')" | |
| }, | |
| "46": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([-1.3477e-17, -3.5905e-18, 9.6613e-20, -4.0287e-17, -4.0820e-17,\n -5.7482e-20, -9.8272e-19, -5.2755e-18, -7.7888e-17, -2.9416e-17,\n -3.6890e-17, 2.3096e-20, -1.0641e-16, -3.1736e-17, 3.9785e-19,\n -9.0887e-18, -4.8168e-18, -3.1497e-20, 1.9220e-18, -1.1259e-18,\n -8.3402e-18, -2.0328e-19, -5.3765e-18, -4.4435e-18, -4.0927e-17,\n -1.2882e-17, 1.5257e-18, -9.7664e-17, -7.3820e-17, -6.5123e-19,\n -3.3270e-18, -5.3281e-18, -2.3923e-17, 1.8076e-19, 3.7326e-19,\n -6.0721e-18, -7.9071e-19, 1.2380e-19, -4.6963e-17, 6.9692e-20,\n -2.2634e-18, -6.7674e-17, -2.3845e-19, 7.9366e-20, 1.4627e-19,\n -1.2412e-18, -1.2002e-17, -4.1759e-17, -1.2880e-19, -3.3639e-17,\n -3.1729e-17, 1.1002e-19, -1.2059e-17, -1.3196e-18, -2.2626e-17,\n -6.5398e-18, -2.3225e-19, -4.0470e-18, -5.9835e-17, -9.0676e-18,\n 3.0530e-19, 8.1635e-19, -4.5312e-17, -5.0664e-17, -6.7196e-17,\n 1.1934e-20, -6.1305e-19, 2.1342e-19, -4.1090e-19, 1.7818e-19,\n -9.4093e-17, -7.2550e-19, -3.5070e-19, -1.1583e-18, -4.9016e-17,\n -2.9816e-17, -9.3988e-18, 8.4576e-19, -7.9023e-19, -5.2224e-17,\n -1.9440e-18, -4.7411e-19, 2.5635e-19, -8.8678e-18, -6.9226e-17,\n -8.1512e-22, -7.9123e-19, 6.0175e-19, -1.0564e-18, -6.3544e-17,\n -1.4379e-17, 9.5434e-20, -2.6839e-17, 1.1662e-19, -4.0547e-18,\n 6.3934e-19, 2.6594e-18, -9.3672e-17, -9.3127e-18, -3.1760e-17,\n -1.7013e-17, -4.7673e-17, -9.0830e-17, -1.8813e-18, -4.3425e-20,\n -2.0579e-17, -2.9237e-17, 3.8965e-20, -2.0557e-18, 5.3208e-19,\n -3.3094e-17, -1.3662e-18, 3.2766e-19, -3.1362e-19, -8.5732e-17,\n -7.7497e-18, -2.0254e-19, 4.3785e-18, -7.6726e-20, -1.3446e-17,\n 2.9023e-19, -6.9479e-20, 9.7340e-20, -3.9824e-17, -1.2348e-19,\n -6.5863e-17, 1.1296e-18, 1.0091e-19, 5.3150e-21, -2.8944e-18,\n -1.2763e-19, -4.6491e-18, -3.0721e-18, -2.2459e-19, -3.3369e-18,\n -9.7192e-17, -1.2594e-19, -5.6341e-19, 4.3198e-19, 3.2363e-18,\n 2.7845e-20, -4.9426e-18, -1.3651e-19, -6.6667e-17, -1.6556e-17,\n -1.6491e-18, -7.5868e-18, -1.8830e-17, -2.9050e-19, 1.6849e-19,\n -1.0908e-17, -7.6984e-18, -8.4890e-19, -1.3695e-18, -7.4498e-17,\n 3.2820e-19, -1.0668e-17, -1.6436e-17, -1.8572e-19, -5.2849e-19,\n -8.0642e-18, -3.9902e-17, -3.5396e-17, -9.5629e-17, -9.3955e-18,\n -1.5536e-19, -8.9408e-17, -1.4363e-17, -4.5847e-17, -4.4849e-19,\n -4.1289e-17, -1.0521e-19, -5.9676e-19, -1.1368e-18, -1.4051e-18,\n -9.2811e-20, -2.4095e-18, -2.8058e-18, 2.7604e-18, -2.3355e-18,\n -1.3821e-17, -8.6405e-18, -6.5613e-19, -4.3994e-19, 5.9476e-19,\n -9.5775e-18, -1.8941e-20, -1.6316e-18, 3.5390e-19, -1.2066e-18,\n -7.8444e-18, -4.1010e-20, 4.4608e-19, -7.5990e-17, -1.0949e-19,\n -3.3642e-17, -2.6787e-18, -4.6315e-17, -8.5351e-17, -1.0221e-17,\n -3.1953e-19, -9.7224e-17, -2.1322e-18, -9.3942e-17, -4.6047e-17,\n 7.5499e-19, -3.7143e-17, -1.0636e-17, -1.2721e-18, -5.9016e-18,\n -1.4282e-19, -2.3008e-18, 2.6751e-19, 1.6690e-19, -3.4727e-19,\n -1.5124e-18, -9.2195e-19, -2.0516e-20, -5.1012e-19, 4.0842e-19,\n -4.4139e-22, -1.3654e-19, -1.0049e-18, -4.1976e-17, -1.7288e-18,\n -3.2479e-18, -3.7530e-17, -4.4269e-18, 6.2628e-19, -5.9191e-18,\n -1.5674e-18, -1.1510e-17, 4.7745e-19, -1.7999e-18, -6.8131e-18,\n -6.5222e-18, 4.2810e-19, -4.0665e-18, -1.0007e-18, -6.3823e-19,\n -2.2024e-18, -3.8738e-18, -8.2119e-19, -1.3512e-18, -4.9903e-19,\n -1.0483e-19, -7.2520e-17, -3.5807e-18, -9.2253e-17, -7.2242e-19,\n -6.0265e-20, -3.8822e-18, -3.3596e-17, 2.9502e-19, -3.1481e-18,\n -3.1569e-18], device='cuda:0')", | |
| "exp_avg_sq": "tensor([1.0699e-10, 1.2173e-13, 1.0600e-13, 3.6518e-10, 3.1874e-10, 3.0222e-13,\n 9.3270e-15, 5.3556e-11, 4.4575e-10, 2.1052e-10, 3.1678e-10, 6.6371e-12,\n 1.8328e-09, 1.6721e-10, 2.5730e-14, 2.1818e-13, 1.3204e-13, 1.0866e-12,\n 2.6864e-15, 1.7705e-11, 5.2131e-12, 6.9412e-12, 5.4652e-11, 2.7028e-12,\n 6.2259e-11, 2.9013e-11, 2.5012e-12, 1.4780e-09, 3.2920e-10, 1.0749e-14,\n 5.2113e-10, 2.8664e-12, 7.3185e-11, 2.1386e-14, 9.4822e-14, 3.5694e-10,\n 5.4506e-15, 6.2403e-12, 4.8792e-10, 1.4318e-13, 3.3033e-11, 5.5275e-10,\n 1.5127e-12, 3.8347e-15, 9.1881e-13, 1.9831e-12, 2.3141e-14, 1.2370e-10,\n 1.6560e-14, 6.9702e-11, 5.1587e-11, 8.7102e-12, 2.1054e-13, 8.0175e-14,\n 2.6657e-12, 3.2262e-11, 5.7818e-14, 2.9446e-12, 8.6443e-10, 1.7213e-13,\n 9.8672e-12, 7.7142e-14, 4.1610e-11, 4.6351e-10, 4.0584e-10, 5.5762e-11,\n 6.7870e-14, 8.5201e-14, 1.4196e-14, 1.8487e-13, 1.2355e-09, 2.2555e-14,\n 1.4135e-12, 1.4751e-14, 1.8264e-11, 1.8813e-10, 4.4030e-14, 2.2309e-14,\n 1.8111e-11, 2.2975e-10, 3.3105e-14, 2.4360e-12, 1.6809e-12, 6.6134e-11,\n 1.2558e-09, 1.8933e-12, 7.6837e-14, 1.2896e-13, 1.9648e-12, 4.4617e-10,\n 1.9560e-12, 3.1805e-13, 1.1225e-10, 2.0889e-11, 3.1468e-12, 8.4865e-13,\n 6.3652e-14, 1.1912e-09, 1.6684e-10, 9.5204e-12, 4.8844e-12, 1.3344e-09,\n 5.9415e-10, 2.6210e-12, 5.7707e-12, 4.6877e-11, 1.8723e-12, 8.4138e-14,\n 1.7429e-14, 6.4184e-11, 1.4958e-11, 2.4451e-12, 1.2549e-12, 1.2119e-12,\n 2.2210e-10, 1.2449e-12, 1.8555e-12, 4.4035e-13, 3.1459e-14, 2.4051e-13,\n 4.3053e-13, 8.1410e-13, 3.6368e-14, 1.5230e-09, 2.7454e-15, 2.8817e-10,\n 1.7788e-13, 8.5964e-12, 2.5075e-12, 5.0657e-10, 1.0847e-15, 2.3135e-12,\n 2.3895e-11, 1.1372e-12, 3.9662e-14, 1.4310e-09, 7.1631e-14, 5.5184e-14,\n 8.9012e-13, 2.4939e-14, 7.3670e-12, 1.8839e-11, 1.1849e-13, 3.5787e-10,\n 3.2943e-10, 4.2382e-14, 2.2687e-11, 7.9910e-14, 5.9192e-11, 6.0263e-15,\n 3.4691e-10, 1.3769e-11, 1.1299e-14, 3.6471e-12, 8.3570e-10, 1.4721e-13,\n 1.1210e-11, 2.9061e-11, 2.0715e-11, 4.8865e-12, 1.3123e-10, 1.7107e-11,\n 6.9597e-11, 1.6643e-09, 5.0793e-13, 6.9312e-12, 8.2753e-10, 8.5199e-11,\n 1.1660e-10, 4.7692e-11, 6.1658e-10, 2.0972e-12, 9.0238e-15, 8.3726e-14,\n 3.2744e-12, 4.8806e-14, 1.6247e-12, 6.5511e-12, 2.3211e-14, 2.3157e-13,\n 2.6175e-10, 1.8766e-10, 9.5191e-13, 4.8006e-14, 1.0255e-14, 6.7553e-10,\n 5.2465e-13, 1.2797e-11, 2.1828e-11, 1.3496e-13, 1.9717e-11, 2.6214e-14,\n 9.0374e-14, 2.6998e-10, 1.0128e-11, 2.1856e-10, 4.9166e-12, 2.2558e-10,\n 1.6245e-09, 4.8366e-11, 8.6155e-14, 3.2101e-09, 6.0893e-15, 1.6821e-09,\n 4.4176e-10, 1.0559e-13, 7.7500e-11, 3.4426e-11, 2.9495e-12, 1.0837e-11,\n 2.9269e-11, 4.3963e-14, 1.1766e-12, 8.8301e-14, 9.6697e-13, 3.9601e-12,\n 6.3521e-15, 3.1851e-12, 1.4968e-11, 3.0902e-14, 1.2256e-10, 6.1288e-12,\n 4.7833e-15, 2.0488e-10, 6.6605e-14, 7.4233e-14, 6.1826e-11, 1.3507e-11,\n 8.4264e-14, 2.3871e-11, 4.1874e-14, 3.0311e-12, 6.8451e-12, 4.9858e-12,\n 6.8675e-11, 1.2504e-11, 1.1953e-13, 8.5383e-10, 1.5742e-12, 9.4725e-15,\n 3.5977e-11, 6.6186e-10, 1.4472e-12, 3.5726e-11, 6.3337e-12, 9.3641e-13,\n 7.6348e-10, 5.1863e-12, 8.8584e-10, 3.0863e-14, 4.7199e-12, 3.9549e-10,\n 1.8602e-10, 2.3170e-13, 1.6842e-13, 4.7446e-11], device='cuda:0')" | |
| }, | |
| "47": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([-3.3255e-17, -1.3308e-17, 2.1368e-19, -3.0394e-17, -2.9874e-17,\n 4.9703e-19, 6.4443e-19, -1.4111e-17, -5.8862e-17, -3.1725e-17,\n -3.2377e-17, 1.1335e-19, -5.2230e-17, -3.7013e-17, -4.2067e-18,\n -2.5299e-17, -2.5788e-17, -6.3927e-19, -1.1794e-18, -8.1401e-18,\n 6.2607e-18, -4.3970e-19, -1.5370e-17, -8.5528e-18, -4.3384e-17,\n 8.8355e-18, 9.4317e-19, -5.4051e-17, -4.3054e-17, 5.0805e-19,\n -9.2631e-18, 3.9334e-18, -2.4499e-17, 8.5261e-20, 2.4565e-19,\n -1.3642e-17, 5.7788e-19, -3.1585e-18, -2.6861e-17, -4.3983e-20,\n -7.4860e-18, -3.7912e-17, 2.3004e-18, 5.9754e-21, -4.1262e-18,\n 5.7201e-18, -2.7259e-17, -3.5653e-17, -3.0016e-20, -2.8034e-17,\n -3.1514e-17, 4.2078e-18, -2.4808e-17, 1.0408e-18, -3.1611e-17,\n -1.5262e-17, 2.5580e-18, -1.0140e-17, -3.7838e-17, -2.0166e-17,\n -3.0015e-18, -2.8159e-19, -5.3276e-17, -3.9899e-17, -4.9713e-17,\n -3.8348e-18, -4.4283e-19, 2.0351e-19, -1.4730e-18, 4.5116e-19,\n -5.4556e-17, 4.0421e-18, -6.6831e-18, 2.8684e-19, -5.3487e-17,\n -2.8391e-17, -2.5986e-17, -5.8038e-19, -1.9981e-18, -4.3689e-17,\n 1.4897e-18, -6.7048e-19, -3.6108e-18, -1.8062e-17, -3.4370e-17,\n -1.1076e-18, -2.1194e-19, -3.3215e-19, 1.8429e-17, -3.8196e-17,\n -1.9592e-17, -2.3233e-18, -2.5975e-17, 7.6932e-20, 2.5430e-18,\n -9.5734e-18, -2.0511e-18, -4.6720e-17, -1.5284e-17, -3.4827e-17,\n -2.1424e-17, -2.5262e-17, -5.6600e-17, 8.9181e-19, 4.0372e-19,\n -2.0159e-17, -4.1308e-17, -1.0290e-18, -1.0827e-18, 2.3522e-18,\n -3.7269e-17, 7.3856e-18, -3.6158e-18, -5.8996e-19, -6.6724e-17,\n -1.9563e-17, -1.1846e-18, -3.3663e-18, -1.4893e-21, -3.2180e-17,\n -2.1245e-18, -1.1875e-18, -1.6921e-18, -1.9890e-17, 2.0980e-20,\n -4.7098e-17, -7.4276e-19, -8.4136e-19, 8.7702e-19, -9.1916e-18,\n 1.0723e-18, 3.6281e-18, -1.6554e-17, -5.8490e-18, 2.4657e-18,\n -4.8156e-17, -1.7598e-18, 3.5007e-18, 3.5470e-18, -2.5149e-18,\n -1.1478e-18, 3.3096e-18, 1.1224e-18, -4.5648e-17, -3.1123e-17,\n 1.1050e-18, 5.3506e-18, -3.4514e-17, -7.4162e-19, -2.3156e-19,\n -2.5125e-17, 4.9909e-18, 4.2721e-19, 8.1614e-18, -3.7041e-17,\n -1.2501e-19, 6.9564e-18, -2.5164e-17, 1.0600e-18, 3.9958e-19,\n -1.4111e-17, -4.9112e-17, -3.1300e-17, -5.2549e-17, -2.0223e-17,\n -2.3835e-18, -4.8701e-17, -2.3378e-17, -3.9312e-17, -6.6935e-18,\n -2.7275e-17, -1.9794e-18, 2.9498e-19, 8.0553e-19, 8.4312e-19,\n -5.4000e-19, 1.5172e-18, 1.9859e-18, -1.9497e-18, -1.0521e-17,\n -2.3154e-17, -1.4793e-17, 7.5833e-18, -9.7396e-18, -4.1727e-19,\n -1.6873e-17, -4.7130e-19, -7.0581e-18, -3.9010e-18, 6.9448e-19,\n -1.0065e-17, -6.9195e-19, -3.6180e-19, -5.8849e-17, 5.0227e-18,\n -2.3391e-17, 1.0532e-18, -3.3108e-17, -4.5642e-17, -1.1360e-17,\n 2.2500e-18, -5.5461e-17, 1.6627e-18, -4.4467e-17, -2.8070e-17,\n 1.3658e-19, -3.5795e-17, -1.2774e-17, 2.4780e-19, 4.3743e-18,\n -2.3067e-18, 1.5905e-18, -1.8345e-18, -8.3580e-19, -7.9704e-18,\n 1.4446e-18, 4.3617e-19, 4.4478e-18, 8.5953e-18, -1.8774e-18,\n 6.7763e-19, 7.4690e-19, 6.9663e-19, -3.6626e-17, 1.2862e-18,\n 2.2859e-18, -3.3206e-17, -2.2910e-17, -5.5109e-18, -1.5089e-17,\n 1.2113e-18, -1.9650e-17, 6.9662e-18, 7.6985e-19, -1.2685e-17,\n -1.6474e-17, -9.5204e-20, -8.4124e-18, 2.8485e-19, 2.2668e-19,\n -1.3894e-17, -9.6413e-18, 4.3175e-18, -8.5900e-18, 2.9826e-18,\n -9.7908e-19, -4.6672e-17, -8.4836e-18, -4.9897e-17, -5.2660e-18,\n -1.3875e-18, -7.2403e-18, -2.8213e-17, -5.2659e-18, -1.1881e-17,\n -3.7026e-18], device='cuda:0')", | |
| "exp_avg_sq": "tensor([1.1811e-10, 1.1135e-11, 5.8105e-14, 4.6325e-10, 3.1995e-10, 1.7282e-13,\n 3.0031e-14, 4.3899e-11, 5.2120e-10, 1.7959e-10, 2.0241e-10, 1.6330e-10,\n 6.9004e-10, 1.0505e-10, 5.2037e-12, 7.7100e-12, 5.6240e-12, 5.5804e-13,\n 1.1466e-13, 3.0943e-11, 2.8150e-12, 1.3103e-11, 5.3965e-11, 3.2985e-11,\n 9.9564e-11, 1.0545e-11, 1.2710e-12, 7.0360e-10, 2.3450e-10, 2.3224e-13,\n 4.7740e-10, 1.6594e-12, 2.5387e-10, 4.6193e-13, 1.7643e-11, 2.2096e-10,\n 6.6722e-13, 3.0315e-12, 2.2035e-10, 4.1293e-14, 4.0447e-11, 3.0813e-10,\n 7.7414e-13, 9.2260e-13, 4.9271e-13, 5.2884e-13, 2.2506e-11, 3.5084e-10,\n 1.8576e-15, 1.7363e-10, 1.1729e-10, 3.9678e-11, 5.9451e-11, 2.7387e-12,\n 4.0235e-11, 2.7076e-11, 2.4968e-14, 2.6484e-11, 6.5730e-10, 2.2652e-11,\n 4.5966e-11, 1.9055e-11, 2.8203e-10, 8.0642e-10, 4.8834e-10, 1.4102e-10,\n 2.3566e-14, 1.1427e-13, 3.7626e-12, 3.3703e-11, 8.2137e-10, 9.6280e-12,\n 8.0170e-12, 2.2538e-12, 1.5309e-10, 1.0216e-10, 1.5484e-11, 8.1353e-13,\n 1.3862e-11, 4.7606e-10, 3.0286e-14, 1.3394e-12, 2.2474e-11, 3.8421e-11,\n 4.9328e-10, 8.4942e-13, 2.6484e-14, 1.7165e-14, 1.0357e-12, 2.7429e-10,\n 1.7412e-10, 1.9918e-11, 2.6010e-10, 2.0188e-11, 1.0730e-12, 4.0249e-13,\n 2.2069e-12, 3.9984e-10, 9.8570e-11, 8.6410e-11, 9.5679e-11, 4.8353e-10,\n 4.5496e-10, 1.0304e-12, 1.8304e-11, 1.5404e-10, 5.4078e-11, 6.8063e-12,\n 2.1364e-12, 6.6775e-11, 1.5012e-10, 1.3064e-12, 1.1534e-11, 6.0755e-13,\n 3.2583e-10, 3.6980e-11, 9.9510e-13, 1.1368e-12, 1.1539e-14, 1.4938e-11,\n 2.2013e-13, 3.5666e-13, 2.1378e-12, 5.5228e-10, 2.0635e-14, 2.8070e-10,\n 1.0302e-11, 7.6061e-11, 1.0721e-12, 4.6492e-10, 1.0216e-12, 1.2847e-12,\n 5.3303e-11, 5.6555e-13, 8.4232e-13, 5.4988e-10, 5.9279e-12, 2.7442e-12,\n 4.9786e-13, 4.0966e-15, 2.8059e-11, 9.3318e-12, 6.2749e-14, 1.8443e-10,\n 2.1317e-10, 2.9183e-13, 8.1505e-12, 3.7866e-11, 6.2333e-11, 2.4285e-13,\n 2.1638e-10, 7.7666e-12, 1.8694e-12, 1.5845e-12, 2.9996e-10, 1.2988e-11,\n 5.8154e-12, 5.8801e-11, 3.1141e-10, 2.1318e-12, 8.8484e-11, 1.6623e-10,\n 1.8478e-10, 8.0537e-10, 3.6861e-11, 1.7771e-11, 3.7677e-10, 7.8949e-11,\n 2.4299e-10, 7.1570e-11, 5.2647e-10, 1.1411e-12, 2.1741e-12, 3.6431e-12,\n 1.7071e-12, 3.6913e-12, 8.3118e-13, 3.1828e-12, 3.4582e-12, 1.4081e-11,\n 1.3419e-10, 8.2953e-11, 5.2396e-13, 1.6850e-11, 4.6785e-13, 3.2923e-10,\n 2.6769e-13, 3.2727e-11, 4.7171e-11, 6.5990e-11, 1.1440e-11, 1.1680e-12,\n 2.3893e-14, 4.7799e-10, 4.8207e-12, 1.7180e-10, 2.0576e-12, 1.6060e-10,\n 7.2570e-10, 7.6150e-11, 4.0224e-14, 1.8879e-09, 1.7426e-13, 5.5896e-10,\n 3.1721e-10, 2.4190e-11, 2.4842e-10, 1.6555e-11, 1.3299e-12, 5.2606e-12,\n 9.0959e-11, 1.3957e-14, 1.4635e-11, 4.6055e-14, 7.1162e-12, 2.0246e-12,\n 1.9895e-13, 1.5964e-12, 7.1141e-11, 1.8684e-14, 1.8982e-10, 2.9138e-12,\n 2.5710e-13, 6.0897e-10, 5.5934e-15, 3.8472e-12, 1.5790e-10, 6.1810e-11,\n 1.5020e-11, 5.5520e-11, 2.0655e-13, 5.3838e-11, 3.5108e-12, 2.2354e-12,\n 3.4469e-11, 2.2317e-11, 1.0682e-13, 5.2149e-10, 9.1073e-13, 6.7702e-13,\n 5.3383e-11, 5.0842e-10, 6.3543e-13, 5.6515e-11, 5.5038e-11, 4.7185e-13,\n 5.9438e-10, 8.7127e-12, 4.2774e-10, 2.5801e-12, 1.3385e-11, 1.8230e-10,\n 1.2581e-10, 4.0232e-12, 8.6939e-12, 2.4692e-11], device='cuda:0')" | |
| }, | |
| "48": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([[-4.3027e-19, 5.2446e-18, -5.8394e-19, ..., -2.5318e-19,\n -2.3734e-20, -1.8527e-19],\n [-9.2368e-20, 5.4827e-18, -1.7427e-19, ..., 1.3106e-19,\n -1.0056e-19, 4.6250e-20],\n [ 5.6204e-19, -3.4646e-19, 5.1908e-19, ..., 3.4784e-19,\n 1.9798e-19, 2.2814e-19],\n ...,\n [ 2.4037e-19, 7.4623e-20, 1.7135e-19, ..., 2.1542e-19,\n 2.0760e-20, 1.0819e-19],\n [ 6.6195e-19, 4.1002e-18, -1.8989e-19, ..., 1.1596e-19,\n 2.5154e-19, 2.0602e-19],\n [-1.5205e-19, 1.9924e-18, -2.1720e-19, ..., -1.0730e-19,\n -1.0030e-20, -4.9317e-20]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[2.6249e-13, 7.0800e-14, 1.6615e-13, ..., 1.6996e-13, 1.8612e-13,\n 2.6320e-13],\n [4.2748e-12, 1.2540e-12, 3.2080e-12, ..., 1.9309e-12, 4.5230e-12,\n 5.1588e-12],\n [2.2538e-13, 4.2463e-14, 1.0154e-13, ..., 1.3909e-13, 1.6932e-13,\n 2.9921e-13],\n ...,\n [1.1301e-13, 5.8886e-14, 8.8891e-14, ..., 4.3549e-14, 1.4488e-13,\n 7.4780e-14],\n [5.1716e-12, 1.5237e-12, 3.3318e-12, ..., 3.1831e-12, 4.9947e-12,\n 5.8256e-12],\n [8.0128e-14, 4.3778e-14, 7.4641e-14, ..., 4.4551e-14, 1.6819e-13,\n 1.5563e-13]], device='cuda:0')" | |
| }, | |
| "49": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([-1.9993e-16, 4.0404e-17, 1.9780e-16, -4.3796e-16, 3.6314e-17,\n 1.7666e-16, 8.4962e-17, 1.7071e-16, -9.6122e-16, -1.3076e-16,\n 1.3243e-16, 2.9707e-16, -4.5056e-16, 6.0863e-18, 4.7951e-16,\n -8.5589e-16, -8.1524e-16, 2.7388e-17, -3.3787e-17, 7.2389e-17,\n -1.6627e-16, -2.7440e-16, 1.7072e-16, 1.9417e-16, 1.2941e-16,\n -1.2752e-15, 4.0845e-17, -9.9581e-16, -6.6911e-16, -1.1171e-16,\n -7.5349e-19, -1.9208e-16, -8.0474e-17, 1.4876e-16, 1.5661e-16,\n -2.5742e-17, 9.8378e-17, -3.3206e-16, 3.3294e-17, 2.4131e-16,\n 1.2324e-16, 1.3549e-17, 5.4782e-16, 3.8778e-17, -1.0302e-16,\n 4.8771e-16, 1.1105e-17, -9.1766e-16, 6.6800e-17, -4.8347e-16,\n 1.9311e-17, -1.8206e-16, 4.5236e-16, 4.5566e-16, 1.6639e-16,\n 1.0267e-15, 7.3188e-17, -1.6806e-16, -2.9803e-17, 8.1255e-18,\n 3.7413e-17, -1.0544e-16, -7.0011e-16, -5.8563e-16, 1.0192e-16,\n 3.7243e-16, -4.1846e-16, 4.0484e-16, -7.8385e-17, 1.4365e-16,\n -3.6530e-16, -8.9299e-17, 3.1382e-16, 3.1109e-16, -2.3564e-17,\n 7.2223e-16, 3.1563e-16, -1.5709e-16, 1.6390e-16, -6.2393e-16,\n 4.1910e-16, 1.1986e-15, 4.6520e-17, 3.9499e-16, -5.7679e-16,\n 4.3400e-16, 2.9836e-16, 1.6754e-16, 4.4368e-16, -9.0973e-17,\n -1.5818e-16, -1.0842e-16, -2.8469e-16, -1.4920e-16, -1.1406e-16,\n -4.8720e-16, 2.2707e-16, -1.5243e-16, 3.6579e-16, -2.6894e-17,\n 8.6867e-16, 1.3535e-16, -3.8656e-16, -1.0671e-16, 9.8438e-17,\n -2.1357e-16, -2.5675e-16, -1.7870e-16, -1.1146e-15, 1.6462e-16,\n -9.9617e-17, 6.0116e-16, 1.6533e-16, 3.3658e-16, -1.7411e-16,\n -1.3136e-16, -5.9550e-16, 1.3405e-16, 5.2165e-17, -8.2425e-16,\n 7.3610e-17, -1.1643e-16, 7.3084e-17, -1.2008e-16, -1.5397e-16,\n -1.8013e-16, 4.7152e-16, 2.7297e-16, 1.6667e-16, 4.1070e-16,\n 1.3642e-16, -9.8682e-17, 3.5456e-16, -1.6569e-16, -3.6128e-16,\n -4.6549e-17, 4.4329e-17, 1.1228e-16, 1.9233e-16, -4.5018e-17,\n 2.9107e-16, -2.4979e-16, 1.9202e-16, -1.1020e-16, 2.1599e-16,\n 6.9914e-17, 1.8893e-16, -8.7837e-16, 2.0597e-16, 2.5202e-16,\n -2.7639e-16, -8.4488e-16, 5.8073e-16, 2.5474e-16, -1.7069e-16,\n -1.5274e-16, -5.6658e-16, 2.3133e-17, 3.1499e-16, 1.6412e-16,\n -3.3998e-17, -1.0400e-17, 3.2516e-16, -5.6645e-17, 1.5628e-16,\n -3.8361e-17, -8.9537e-17, 1.2079e-17, 8.4381e-17, -1.0415e-16,\n -1.0751e-16, 2.1930e-17, 6.7014e-16, 3.1357e-16, -5.5823e-18,\n 4.0997e-16, -2.0938e-16, -8.5674e-17, 3.6471e-18, 6.5974e-16,\n -1.2273e-16, 8.4889e-17, -2.3206e-16, 2.7106e-17, -1.8377e-16,\n 5.9304e-17, -1.2251e-16, -2.2299e-17, 3.0146e-16, 6.5059e-16,\n 9.2382e-17, 5.2901e-16, 3.2248e-17, -9.2558e-16, -1.2594e-16,\n 4.5382e-17, -1.7930e-17, -6.4178e-17, -8.7951e-16, -4.3020e-17,\n 1.1398e-16, -3.1308e-16, -6.0567e-18, -9.3251e-16, -3.2462e-17,\n -2.0198e-16, -9.9286e-16, -8.8984e-17, 7.2691e-17, 2.5157e-16,\n 1.5003e-16, 3.5605e-17, 3.9843e-16, 1.5771e-16, 9.3876e-17,\n -2.7302e-16, 7.5772e-16, -9.9907e-17, 4.3234e-16, 5.8903e-16,\n 1.1130e-16, 3.0814e-16, -1.0784e-16, 7.5112e-17, 4.7246e-17,\n -8.5250e-17, -9.5373e-18, -9.2588e-16, -4.9960e-16, 3.7123e-18,\n 4.1701e-17, -4.0513e-17, 4.6751e-16, -2.1044e-16, 1.2051e-16,\n 6.0068e-18, -4.9294e-17, 7.3589e-17, 7.0390e-17, 9.4464e-16,\n 3.7663e-17, 1.4889e-16, -2.7354e-16, -9.9412e-17, -1.3196e-16,\n 1.1199e-16, -8.2474e-17, 2.9367e-16, 1.4084e-17, 6.9694e-16,\n 3.9159e-16, -5.3299e-17, 6.0864e-17, 1.2958e-16, 2.0826e-16,\n -2.3603e-17], device='cuda:0')", | |
| "exp_avg_sq": "tensor([5.7814e-08, 9.6946e-07, 4.3981e-08, 6.4890e-07, 9.9896e-08, 3.4679e-09,\n 4.3933e-09, 6.5183e-08, 3.4410e-07, 1.5017e-07, 1.8811e-08, 2.4014e-09,\n 6.4962e-08, 8.4368e-09, 3.6858e-08, 4.8232e-07, 1.0182e-06, 6.8824e-09,\n 2.8120e-10, 1.0423e-08, 8.7283e-08, 1.1260e-08, 1.2363e-08, 2.2072e-08,\n 5.9894e-09, 4.0539e-08, 2.1183e-08, 1.5189e-06, 3.8765e-08, 2.0834e-08,\n 1.8224e-07, 3.8926e-07, 2.2876e-08, 9.0629e-09, 1.0493e-07, 6.1366e-08,\n 4.7517e-08, 6.2583e-07, 3.1587e-09, 2.4175e-07, 2.1626e-08, 4.3681e-09,\n 5.2176e-09, 4.1457e-08, 1.2174e-08, 4.6533e-09, 2.0569e-09, 2.6797e-07,\n 5.5558e-08, 5.6027e-08, 1.2499e-09, 1.5585e-09, 1.4372e-06, 1.3336e-06,\n 6.9143e-08, 1.6186e-07, 1.9667e-10, 5.3430e-08, 2.9673e-08, 9.8386e-09,\n 1.1857e-07, 1.9411e-08, 8.2579e-07, 5.5668e-07, 5.3299e-07, 8.1159e-07,\n 1.2019e-08, 3.4455e-09, 2.6654e-10, 2.1033e-07, 6.9068e-07, 2.2107e-10,\n 5.1040e-08, 1.3124e-07, 1.0911e-08, 2.6026e-08, 1.8997e-08, 2.5442e-09,\n 1.1935e-07, 1.8512e-07, 6.3064e-08, 5.7188e-07, 5.9896e-08, 2.1337e-07,\n 3.6896e-09, 1.3708e-07, 5.9821e-08, 4.2541e-08, 2.6332e-07, 9.3672e-10,\n 7.6751e-09, 4.0797e-10, 1.4194e-07, 1.7680e-09, 2.8526e-09, 6.0769e-08,\n 2.9503e-09, 1.0550e-09, 2.6680e-08, 7.6204e-10, 4.9253e-07, 1.9879e-09,\n 1.0902e-10, 2.2971e-08, 2.0158e-07, 3.2624e-10, 4.2537e-08, 5.4274e-10,\n 2.2030e-07, 4.6927e-09, 3.3895e-10, 2.6732e-09, 1.3440e-07, 2.1484e-08,\n 2.4098e-07, 2.6314e-08, 2.1869e-07, 5.3361e-07, 1.1717e-08, 2.3958e-08,\n 1.5777e-07, 8.9440e-10, 3.9099e-09, 1.7868e-07, 1.5428e-09, 3.9394e-07,\n 7.6340e-07, 1.0108e-07, 1.2667e-07, 1.8455e-06, 3.5514e-08, 2.9913e-10,\n 9.9668e-08, 2.3649e-08, 2.0430e-07, 2.2314e-09, 1.0455e-09, 3.8767e-09,\n 2.5449e-08, 3.0332e-10, 2.4547e-09, 1.3160e-08, 6.2957e-10, 1.1893e-10,\n 1.0061e-06, 8.4799e-08, 9.2088e-07, 2.6129e-07, 2.9104e-07, 8.7983e-09,\n 4.1124e-07, 2.2174e-06, 1.0468e-07, 2.2134e-09, 1.1185e-07, 2.2299e-09,\n 1.0841e-06, 5.7840e-09, 8.9141e-07, 4.0842e-09, 6.0419e-08, 2.8386e-09,\n 1.8166e-07, 1.5922e-09, 5.8683e-09, 1.3694e-08, 1.5703e-08, 6.4305e-09,\n 7.6674e-09, 4.2148e-09, 1.0812e-09, 5.1892e-08, 5.5202e-07, 5.8451e-07,\n 4.2222e-10, 6.2620e-09, 3.2950e-07, 6.1382e-07, 7.3113e-11, 1.6570e-07,\n 3.5051e-07, 2.2087e-07, 1.6596e-08, 3.8082e-11, 2.4621e-09, 1.1287e-06,\n 1.7712e-07, 9.2800e-09, 5.0887e-07, 2.3524e-06, 1.0951e-08, 1.7024e-07,\n 3.9907e-08, 4.3130e-07, 7.6008e-08, 4.0129e-09, 6.0001e-09, 1.5080e-08,\n 3.4447e-07, 1.0600e-08, 1.2475e-10, 9.9656e-08, 1.6312e-09, 5.2876e-07,\n 2.0584e-10, 2.5220e-08, 6.7552e-08, 6.0841e-09, 5.7568e-09, 4.8091e-08,\n 1.2792e-08, 1.6841e-09, 2.8879e-07, 8.1885e-08, 2.9465e-08, 2.3733e-09,\n 5.8298e-08, 1.0435e-09, 5.0912e-09, 3.6268e-07, 3.0478e-07, 5.2063e-07,\n 1.7697e-08, 5.1451e-08, 7.8146e-09, 2.6217e-07, 6.9161e-09, 5.5517e-08,\n 6.3236e-08, 7.8377e-09, 1.0901e-07, 2.2232e-09, 9.7435e-09, 1.4628e-07,\n 7.3870e-08, 2.6621e-08, 1.3557e-07, 1.3886e-08, 1.8826e-07, 2.3773e-07,\n 8.0293e-08, 1.7793e-08, 1.2850e-08, 9.9536e-10, 1.7031e-09, 1.0789e-10,\n 1.6715e-09, 1.1905e-07, 8.6098e-09, 1.1839e-07, 2.9522e-08, 4.4684e-10,\n 1.7366e-09, 3.4007e-08, 1.0686e-06, 2.6781e-08], device='cuda:0')" | |
| }, | |
| "50": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([-7.4893e-20, -5.6621e-17, -5.4165e-21, -5.2819e-17, -1.8714e-17,\n 8.9230e-20, -1.6596e-17, -3.2082e-18, -8.2076e-17, -8.5931e-17,\n 8.1833e-19, -2.4057e-19, -1.8413e-17, -4.0185e-18, -7.5341e-18,\n -6.9541e-17, -4.4467e-17, 5.2799e-19, 2.2868e-19, 1.0869e-19,\n -3.6906e-17, -2.6683e-17, -3.0079e-18, -9.7544e-18, -4.8841e-17,\n -2.2249e-17, 6.5177e-19, -1.2495e-16, -5.8088e-17, 5.8306e-19,\n -4.7859e-19, -7.4213e-17, -6.6554e-18, -2.1929e-20, -9.1798e-20,\n -7.6472e-19, -8.7078e-19, -5.3673e-18, 4.9589e-20, -2.2723e-19,\n -2.9003e-18, -6.4759e-19, 2.4658e-19, 2.6036e-19, -1.1144e-18,\n -8.9445e-19, -3.9996e-19, -1.1767e-16, 2.2386e-19, -3.4878e-17,\n -2.2563e-20, 3.7597e-19, -5.8629e-17, -3.1712e-19, -1.5526e-17,\n 3.8977e-18, 2.0702e-19, -3.1691e-17, -1.0204e-17, -2.9547e-19,\n -2.9638e-19, 7.6222e-19, -7.1172e-17, -5.3389e-17, -4.2624e-17,\n -2.4698e-18, -3.1332e-17, -1.0402e-19, -2.9744e-18, 6.6645e-20,\n -4.2207e-17, 1.1318e-18, -9.7653e-19, -4.2383e-18, -3.9155e-18,\n -4.6112e-18, -3.9501e-17, 1.2364e-18, 8.5291e-21, -2.3405e-17,\n -5.9391e-18, -2.0820e-17, -1.7402e-19, -2.5884e-18, -1.2465e-17,\n -3.3293e-18, -2.3557e-19, 3.5051e-20, -6.8381e-19, -6.5015e-18,\n -2.0805e-18, -2.4094e-18, -6.8061e-17, -1.9841e-19, -4.6640e-18,\n -2.3196e-17, 3.3268e-20, -7.7856e-18, -1.0229e-17, 6.7346e-19,\n -1.6483e-17, -2.8218e-19, -3.2267e-18, -3.6455e-17, 9.9625e-20,\n -2.5438e-18, -2.3354e-17, 1.0773e-18, -6.0873e-17, 3.2215e-19,\n -3.9407e-18, -1.3962e-18, 3.2874e-19, 5.5231e-20, -6.6949e-17,\n -1.5732e-17, -5.0699e-17, 1.9062e-19, -9.6512e-20, -3.2822e-17,\n -1.3206e-18, -1.7965e-18, -7.1569e-20, -2.5338e-17, 1.3133e-19,\n -7.0870e-17, 1.3694e-19, -5.0622e-20, -1.3276e-20, -1.7550e-18,\n 2.6136e-20, -5.3602e-18, -3.4703e-18, -1.6786e-17, -1.9122e-17,\n -6.8635e-18, 3.8073e-19, -3.6058e-20, 2.0257e-19, 1.0321e-18,\n -1.3285e-19, -4.6170e-18, 1.3297e-19, -2.9843e-18, -6.1639e-18,\n -7.9464e-17, -5.8467e-17, -4.6066e-17, 1.4991e-19, -1.3686e-19,\n -5.8171e-18, -9.8244e-17, 4.5326e-19, -3.6599e-19, -4.3486e-17,\n 2.0202e-19, -7.2688e-17, -3.3752e-19, -5.2586e-19, -5.7196e-20,\n -4.7250e-19, -1.8731e-18, -4.8269e-17, -2.1940e-18, -1.1655e-18,\n -4.3006e-19, -1.5379e-17, -5.3853e-19, -4.7525e-18, -1.7198e-18,\n -6.2638e-18, -1.6197e-17, -1.4242e-18, -4.9910e-19, -2.0973e-18,\n -5.4647e-18, -8.4304e-17, -8.6254e-17, 8.8222e-20, -2.5042e-18,\n 1.2853e-19, -4.2051e-18, 3.7191e-18, -1.9386e-19, 1.2174e-18,\n -1.0138e-17, -6.8960e-18, -8.5137e-19, -2.8421e-19, 5.1364e-20,\n -1.1967e-19, -9.3475e-18, 1.4964e-19, -6.9044e-17, 1.9816e-18,\n 2.3653e-19, -1.1722e-18, -1.8819e-18, -8.3599e-17, -7.0863e-19,\n -3.7470e-20, -1.2638e-17, -6.6304e-19, -9.2308e-17, -2.4745e-18,\n 2.2533e-18, -3.5858e-17, -1.6670e-18, 1.1275e-19, -5.8313e-18,\n 1.3210e-19, 3.2996e-19, -5.8073e-21, -9.5364e-18, -1.3783e-17,\n -4.3980e-18, -3.6198e-18, 1.1637e-18, -7.4003e-21, -2.2882e-18,\n 1.6575e-19, -4.1992e-19, -1.4512e-17, -7.6225e-18, 2.3683e-19,\n -6.2665e-18, -7.3250e-19, -2.5685e-17, -2.2994e-17, -1.2561e-18,\n -1.1970e-19, -1.6775e-18, 1.4128e-19, -4.3907e-17, 1.4554e-20,\n -4.7611e-19, 1.3488e-19, 1.6280e-19, -2.8935e-17, -6.4736e-19,\n -8.8783e-18, 3.0740e-20, 3.0011e-18, -7.6147e-19, -1.0142e-19,\n 1.2579e-20, -3.8942e-18, -6.7276e-19, -2.2385e-18, -4.0228e-17,\n -3.1484e-20, 6.3025e-20, 1.9571e-20, -2.6007e-19, -5.3830e-17,\n -2.2070e-19], device='cuda:0')", | |
| "exp_avg_sq": "tensor([6.0876e-12, 8.9738e-10, 4.4514e-12, 2.6155e-10, 9.9146e-12, 6.6349e-14,\n 2.9925e-11, 1.8944e-11, 3.6221e-10, 1.1498e-09, 4.2590e-12, 1.8952e-12,\n 1.7584e-13, 1.7937e-12, 8.7253e-11, 2.8001e-10, 1.8370e-10, 1.7010e-14,\n 1.6916e-14, 4.5570e-12, 1.2177e-10, 5.0031e-11, 4.1107e-11, 8.4654e-13,\n 1.9839e-10, 3.3680e-13, 2.1750e-12, 2.9201e-09, 2.8779e-10, 2.3962e-12,\n 2.3973e-13, 9.7269e-10, 1.2167e-13, 1.9756e-13, 1.0484e-12, 2.0847e-14,\n 1.0256e-11, 2.0378e-10, 3.2293e-14, 4.1539e-11, 2.4709e-11, 3.3932e-12,\n 1.5901e-12, 2.0680e-12, 4.9130e-14, 4.3045e-12, 1.8339e-12, 1.1994e-09,\n 5.4644e-13, 2.8838e-11, 4.3092e-13, 1.5770e-14, 2.6201e-09, 2.3064e-10,\n 3.3033e-11, 1.2494e-12, 4.7642e-14, 3.2307e-11, 1.0198e-13, 1.5205e-12,\n 7.3581e-13, 7.0401e-14, 5.4270e-10, 1.5624e-10, 4.6807e-10, 2.8864e-10,\n 6.6009e-12, 1.4967e-12, 1.2375e-13, 2.0265e-12, 1.6763e-10, 5.6418e-14,\n 7.1450e-13, 3.9916e-11, 9.8174e-12, 1.0288e-10, 2.2388e-10, 1.5064e-14,\n 1.6029e-13, 3.4139e-12, 1.0434e-10, 1.2072e-09, 4.8936e-13, 4.8884e-11,\n 1.6624e-13, 3.1086e-11, 8.0545e-12, 4.5715e-12, 1.4068e-12, 4.5116e-14,\n 6.9222e-12, 2.0330e-14, 7.0209e-10, 1.2034e-16, 7.0492e-12, 1.1154e-10,\n 2.3816e-12, 7.3938e-14, 1.0735e-10, 1.5913e-14, 3.9913e-10, 5.3696e-14,\n 9.8216e-14, 5.4155e-11, 1.7698e-12, 8.3169e-14, 3.3782e-12, 9.1456e-15,\n 4.5605e-10, 2.5947e-13, 8.0238e-13, 8.8093e-13, 4.9688e-12, 2.7118e-14,\n 5.2421e-10, 1.1322e-11, 1.1515e-10, 2.5835e-12, 5.4657e-13, 3.3268e-12,\n 3.9868e-11, 3.5761e-15, 1.9703e-13, 1.9907e-11, 1.5724e-15, 7.8977e-10,\n 3.7773e-11, 8.1295e-13, 5.1627e-13, 6.8218e-10, 1.7151e-15, 7.2208e-13,\n 6.7321e-11, 1.2946e-11, 1.6183e-10, 1.3121e-11, 1.0689e-12, 9.0505e-13,\n 8.5870e-13, 1.2393e-13, 4.7022e-14, 2.1973e-11, 7.1959e-14, 4.1255e-13,\n 4.8211e-10, 7.3349e-10, 1.3697e-09, 3.2537e-11, 2.2836e-11, 4.4863e-13,\n 1.2491e-10, 2.3449e-09, 2.8401e-13, 2.6952e-12, 1.3627e-10, 7.4756e-14,\n 8.3784e-10, 1.6830e-12, 1.3328e-11, 7.6159e-14, 3.7436e-14, 7.6840e-12,\n 5.0136e-10, 5.3787e-12, 1.2221e-13, 2.4497e-15, 1.0883e-12, 1.4031e-12,\n 8.1150e-14, 2.4392e-14, 2.5449e-12, 5.2144e-12, 1.0945e-10, 7.5628e-11,\n 8.8797e-13, 2.8745e-12, 1.1685e-09, 1.6022e-09, 7.9304e-14, 3.6720e-12,\n 3.6266e-11, 5.9723e-11, 2.6141e-14, 3.5769e-13, 3.1783e-15, 7.4001e-10,\n 1.6422e-11, 2.3737e-15, 7.2054e-11, 9.2037e-10, 6.3429e-13, 5.2153e-11,\n 1.3645e-13, 3.1144e-10, 1.3977e-13, 5.1243e-14, 2.7841e-12, 5.6602e-12,\n 5.8483e-10, 4.2286e-12, 6.6613e-14, 9.0912e-13, 3.0318e-14, 7.8707e-10,\n 1.0678e-13, 8.8160e-14, 2.0505e-12, 3.3848e-15, 4.2095e-14, 1.7603e-13,\n 3.3114e-12, 1.5170e-14, 2.0778e-11, 4.5137e-11, 1.8716e-11, 7.1909e-12,\n 4.6378e-11, 1.6639e-13, 1.9634e-12, 1.2249e-11, 2.4547e-11, 8.2732e-11,\n 4.6292e-13, 2.2205e-13, 1.6473e-12, 1.1280e-10, 2.3885e-15, 1.2637e-10,\n 3.1709e-11, 2.5350e-14, 2.9303e-13, 2.9948e-12, 4.4215e-12, 1.4262e-10,\n 6.0571e-13, 1.9707e-14, 1.3067e-13, 8.1782e-12, 7.3539e-11, 5.2067e-11,\n 1.1973e-10, 8.7521e-12, 2.7711e-14, 7.8561e-15, 3.3212e-14, 3.7274e-14,\n 4.4791e-12, 1.5348e-13, 5.3138e-12, 5.6827e-10, 2.8600e-12, 4.0239e-13,\n 3.3802e-14, 1.2671e-11, 1.2210e-09, 1.3639e-14], device='cuda:0')" | |
| }, | |
| "51": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([-1.2758e-17, -3.3226e-17, -2.4579e-18, -3.9673e-17, -2.1407e-17,\n -5.5699e-19, -2.0206e-17, -9.1996e-18, -5.3916e-17, -4.2107e-17,\n -4.2964e-18, 1.2389e-18, -2.9285e-17, -1.3378e-17, -9.8472e-18,\n -4.9843e-17, -4.3308e-17, 2.3341e-19, -1.6030e-20, -6.5700e-18,\n -3.0919e-17, -2.9238e-17, -9.3834e-18, -1.4864e-17, -3.0090e-17,\n -4.5986e-17, 3.7558e-19, -6.1603e-17, -4.4068e-17, -9.6724e-18,\n -2.0656e-19, -4.0633e-17, -1.7496e-17, -1.6943e-19, 9.5487e-19,\n 3.9611e-19, -8.7902e-18, -2.1056e-17, -2.7504e-18, -3.6545e-18,\n -9.9402e-18, -1.2308e-18, 5.5253e-18, 5.3807e-19, 7.4740e-19,\n 6.2543e-18, -2.4023e-18, -5.9205e-17, -4.5679e-18, -3.5388e-17,\n -2.0234e-18, -2.7508e-19, -2.8875e-17, -4.5863e-19, -1.8077e-17,\n 7.7515e-18, 3.5220e-19, -2.9320e-17, -1.8125e-17, -7.0690e-19,\n -1.4428e-19, -4.6323e-19, -4.7231e-17, -4.1688e-17, -2.9114e-17,\n -6.2681e-18, -3.3174e-17, 2.4000e-18, 1.8135e-18, 9.3406e-19,\n -3.5219e-17, -6.1411e-19, -5.0630e-18, -8.2439e-18, 2.9763e-18,\n -5.2142e-18, -2.5156e-17, -9.3218e-19, -3.5062e-19, -3.4168e-17,\n -9.0613e-18, -1.0680e-17, -7.0316e-19, -6.0570e-18, -2.9065e-17,\n -6.7323e-18, -3.4052e-18, 1.2144e-18, 5.4837e-18, -1.6974e-17,\n 1.5488e-18, 1.6385e-18, -4.0590e-17, 1.3438e-19, 3.3041e-18,\n -3.1492e-17, 3.6459e-19, -1.9170e-17, -1.2948e-17, -5.8289e-18,\n -1.1294e-17, -6.0120e-18, -1.6715e-17, -2.9650e-17, -9.9062e-19,\n -1.3644e-17, -2.7736e-17, -8.3304e-19, -5.2207e-17, -1.8130e-18,\n 2.7451e-18, 7.9107e-18, -2.2705e-18, -4.3360e-19, -3.8580e-17,\n -2.2676e-17, -4.1130e-17, 1.3157e-18, -2.7309e-19, -4.0271e-17,\n -9.0630e-18, 1.2692e-18, -2.8323e-18, -2.6486e-17, -9.3868e-20,\n -3.9484e-17, 3.2822e-18, 1.4630e-18, 3.6348e-19, -4.9505e-18,\n 5.4581e-19, 3.7466e-18, -7.9206e-18, -2.4304e-17, -2.7337e-17,\n 5.0121e-18, -2.6731e-18, -8.9836e-20, 2.3101e-18, -6.1433e-19,\n 9.8090e-19, 2.8961e-18, 2.7535e-19, 2.1093e-18, -1.1693e-17,\n -3.8054e-17, -3.1975e-17, -4.4925e-17, -2.5123e-18, 9.6295e-19,\n -1.9324e-17, -5.4847e-17, 1.0396e-18, 3.1426e-18, -3.2758e-17,\n -1.3760e-19, -4.5645e-17, 6.5298e-20, 3.3259e-18, -3.8301e-18,\n 3.3469e-19, -1.2543e-18, -2.7252e-17, 1.6328e-18, -7.5366e-18,\n 1.8492e-19, -2.2056e-17, -2.1432e-18, -1.3161e-17, 1.1642e-18,\n 4.2961e-18, -2.0792e-17, -2.7830e-18, -3.4684e-18, 1.1085e-18,\n -8.8368e-18, -4.2598e-17, -4.1339e-17, 4.2103e-20, -3.8948e-18,\n -1.0203e-17, -1.2138e-17, -2.5438e-18, -3.9453e-19, -8.3462e-19,\n -1.7184e-17, -1.7273e-17, 4.1527e-19, -3.1230e-18, -1.7616e-20,\n -1.5764e-20, -1.0486e-17, 1.3234e-19, -5.0585e-17, -1.1720e-18,\n -1.5099e-18, -1.9165e-19, 9.8828e-19, -5.2704e-17, 2.2060e-19,\n 6.6734e-19, -2.4581e-17, 1.8098e-21, -5.5004e-17, 6.0428e-19,\n -1.5904e-18, -4.4176e-17, 1.1996e-18, -1.3394e-18, -1.1907e-17,\n -2.5198e-18, -2.8152e-18, -1.5238e-19, -1.5130e-17, -1.8678e-17,\n 2.8566e-18, -4.2844e-18, -8.9084e-19, 3.4743e-18, -4.7482e-18,\n -4.4055e-18, -3.6696e-18, -2.1414e-17, -1.5272e-17, -3.0674e-18,\n -1.5911e-17, 3.5170e-19, -4.0098e-17, -3.1659e-17, 6.5307e-19,\n -1.2728e-19, 1.0606e-18, 2.9630e-18, -3.3381e-17, -7.4414e-19,\n -4.2706e-19, -1.1401e-19, -4.9159e-18, -2.4892e-17, -5.7470e-19,\n -1.6365e-17, -3.3378e-18, -2.3681e-18, 5.7410e-19, 3.5424e-20,\n 7.3472e-19, 2.9851e-18, -4.5046e-18, 6.7912e-19, -2.1466e-17,\n -4.0919e-19, -7.7609e-20, -8.3395e-19, -6.7042e-18, -3.0131e-17,\n -1.5845e-20], device='cuda:0')", | |
| "exp_avg_sq": "tensor([4.0992e-11, 6.6888e-10, 1.4866e-11, 4.5221e-10, 9.6295e-11, 3.6531e-14,\n 4.9705e-11, 2.2685e-11, 3.7443e-10, 3.5922e-10, 2.4089e-11, 9.3928e-13,\n 6.5608e-11, 1.7126e-11, 4.2763e-11, 3.8994e-10, 5.9133e-10, 7.0446e-13,\n 2.6495e-15, 1.7501e-11, 1.7046e-10, 5.5320e-11, 4.5026e-11, 3.6036e-11,\n 1.1739e-10, 3.9049e-11, 1.0394e-12, 1.1457e-09, 1.6419e-10, 2.5480e-11,\n 5.1069e-11, 4.8581e-10, 2.3499e-11, 1.0842e-13, 1.8755e-11, 1.0230e-11,\n 1.5184e-11, 2.2220e-10, 2.1594e-12, 5.9218e-11, 3.1771e-11, 1.5143e-12,\n 7.7719e-13, 7.8841e-13, 3.3295e-12, 1.8096e-12, 9.4564e-13, 4.4505e-10,\n 2.4240e-12, 1.0299e-10, 2.3563e-13, 3.1392e-13, 1.0055e-09, 2.9662e-10,\n 5.2828e-11, 1.0084e-12, 1.0592e-14, 9.9503e-11, 3.2900e-11, 6.6557e-13,\n 2.8004e-11, 3.3406e-12, 5.9807e-10, 3.9066e-10, 4.3707e-10, 1.8029e-10,\n 4.8469e-11, 7.7520e-13, 3.6357e-14, 4.3764e-11, 4.5061e-10, 5.1423e-15,\n 1.1664e-11, 2.4578e-11, 3.9253e-12, 5.8713e-11, 1.1999e-10, 7.9143e-14,\n 6.5582e-12, 1.3821e-10, 5.5533e-11, 4.7552e-10, 1.6247e-11, 3.0664e-11,\n 1.0723e-11, 1.6632e-11, 1.1199e-11, 1.8757e-12, 3.0234e-11, 9.5511e-12,\n 3.8286e-12, 3.0137e-13, 3.0960e-10, 3.9437e-15, 3.5721e-12, 1.1879e-10,\n 1.2579e-12, 1.4508e-11, 9.9039e-11, 7.9602e-12, 3.2236e-10, 4.4598e-12,\n 1.1785e-12, 7.5488e-11, 3.3198e-11, 1.9223e-12, 6.1597e-11, 6.1404e-15,\n 2.7768e-10, 1.2405e-13, 2.1920e-13, 4.0956e-13, 1.8151e-11, 6.0162e-12,\n 3.3593e-10, 6.7119e-11, 2.3872e-10, 1.3388e-10, 2.4790e-13, 5.4935e-11,\n 6.0782e-11, 4.9675e-14, 1.0546e-13, 1.5770e-10, 2.2855e-13, 4.6318e-10,\n 1.3829e-10, 5.9983e-12, 1.3178e-11, 4.6376e-10, 8.8227e-13, 3.0281e-13,\n 4.5914e-11, 6.6329e-11, 6.7668e-11, 5.6292e-12, 5.9444e-13, 4.6966e-13,\n 3.8068e-13, 3.7153e-14, 2.4038e-14, 8.9617e-12, 2.7951e-14, 1.0980e-13,\n 2.7936e-10, 2.4200e-10, 6.8680e-10, 2.1263e-10, 7.0949e-11, 2.2323e-13,\n 1.0863e-10, 1.3356e-09, 8.9335e-13, 8.9391e-13, 1.7018e-10, 4.8665e-13,\n 7.2392e-10, 8.9026e-13, 2.0993e-10, 4.1942e-12, 1.1493e-11, 3.1769e-12,\n 2.0316e-10, 2.6545e-12, 7.7687e-12, 1.3251e-13, 3.5246e-11, 6.1292e-13,\n 1.5850e-11, 1.3052e-12, 1.1121e-12, 6.3519e-11, 4.9076e-11, 9.0718e-11,\n 3.8927e-13, 2.6592e-11, 3.9080e-10, 5.3403e-10, 6.2478e-15, 1.0046e-10,\n 9.5122e-11, 5.0083e-11, 1.5416e-12, 1.7662e-13, 1.5491e-13, 2.9942e-10,\n 1.2062e-11, 4.6186e-14, 9.2026e-11, 6.3348e-10, 3.2467e-13, 7.2937e-11,\n 3.4754e-12, 3.8188e-10, 1.5252e-11, 1.3826e-12, 1.3295e-12, 2.2146e-12,\n 4.1268e-10, 1.8099e-12, 1.1736e-14, 6.8366e-11, 1.5047e-12, 5.0466e-10,\n 1.2646e-13, 5.8220e-12, 7.9782e-11, 1.6617e-13, 1.6743e-12, 4.4749e-11,\n 1.6135e-12, 4.5117e-12, 3.2271e-11, 2.2280e-11, 4.6867e-11, 3.8304e-12,\n 3.9361e-11, 1.4581e-13, 1.0210e-12, 6.0423e-12, 6.5637e-11, 1.2749e-10,\n 3.2551e-11, 4.4034e-11, 7.0893e-13, 1.0823e-10, 1.8676e-13, 8.6901e-11,\n 1.0873e-10, 3.7361e-12, 2.3971e-11, 1.4180e-12, 2.3146e-12, 1.8016e-10,\n 4.8238e-12, 1.6207e-12, 2.8624e-11, 4.3352e-12, 1.6954e-10, 8.2769e-11,\n 8.7007e-11, 4.5066e-12, 1.2774e-12, 2.4076e-13, 5.0088e-13, 4.5847e-16,\n 2.4197e-12, 3.5983e-13, 2.0351e-12, 2.5040e-10, 8.8717e-12, 3.8206e-14,\n 7.3754e-12, 2.5339e-11, 6.9250e-10, 2.3344e-12], device='cuda:0')" | |
| }, | |
| "52": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([[ 8.7405e-20, 4.0207e-18, 1.3902e-19, ..., -1.6627e-19,\n 2.5199e-18, -6.1094e-19],\n [ 2.2997e-19, 9.1986e-19, -2.4571e-20, ..., -1.2037e-19,\n 6.3299e-19, 1.7433e-20],\n [ 1.9355e-19, -1.8278e-18, -6.4870e-20, ..., 4.9001e-20,\n -2.0473e-18, 5.5638e-19],\n ...,\n [ 3.9990e-18, -7.6682e-17, 9.2636e-18, ..., 6.5017e-18,\n -1.0426e-16, 3.3363e-18],\n [ 6.5275e-19, -1.0237e-18, 1.0999e-18, ..., 1.1067e-18,\n -6.4769e-18, 3.2154e-19],\n [-2.1057e-18, 2.7834e-17, -5.1284e-18, ..., -4.1084e-18,\n 4.6035e-17, -8.5794e-19]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[1.2871e-11, 2.2219e-11, 7.2719e-12, ..., 2.9239e-12, 2.7215e-11,\n 8.6452e-11],\n [5.7876e-13, 4.4019e-13, 4.9820e-14, ..., 5.1535e-13, 5.9858e-13,\n 2.6566e-12],\n [2.4132e-12, 1.6417e-12, 1.2729e-14, ..., 1.2962e-12, 3.4345e-12,\n 7.2758e-12],\n ...,\n [3.1275e-11, 5.0328e-12, 4.3832e-13, ..., 8.2353e-13, 3.0866e-11,\n 3.5687e-11],\n [2.0125e-09, 2.5008e-10, 7.0046e-12, ..., 1.4714e-10, 8.8866e-10,\n 2.5101e-09],\n [4.4635e-10, 1.6224e-10, 1.3187e-12, ..., 4.0007e-11, 4.7553e-10,\n 7.7098e-10]], device='cuda:0')" | |
| }, | |
| "53": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([ 6.1043e-19, 4.9878e-19, -7.9664e-19, 1.0274e-18, 3.9213e-18,\n -1.9441e-19, 3.2070e-18, -8.6384e-20, 2.2035e-18, 9.4814e-19,\n -2.0256e-19, 9.4861e-19, -4.4442e-18, -5.2730e-19, -1.2972e-18,\n -4.1327e-18, 1.0605e-18, -1.5115e-18, -3.1660e-18, 1.0571e-18,\n -2.9584e-18, 1.2932e-19, 2.6988e-18, -1.0507e-19, -1.6432e-18,\n -1.6635e-18, 3.4481e-18, -4.6910e-18, -2.2903e-18, 3.4681e-19,\n 2.6664e-18, 2.2051e-18, -1.9551e-18, 2.1742e-18, -4.8881e-19,\n 2.1543e-18, -3.4220e-18, -1.9838e-18, -1.5422e-18, 3.4852e-18,\n -2.5016e-18, 1.8225e-18, 1.7517e-19, 2.7365e-18, -2.3234e-18,\n 2.5708e-18, 4.3907e-18, 1.2117e-18, -3.4693e-18, -1.5745e-18,\n -4.8485e-19, -2.5923e-18, -6.5356e-19, -1.0805e-18, -9.7811e-20,\n -3.7090e-18, 1.2391e-18, 2.8766e-18, -3.6215e-19, -4.8166e-20,\n 4.7005e-18, -4.9554e-19, 2.4770e-19, 1.6178e-18, -5.8034e-18,\n -1.4360e-18, 4.6515e-18, -4.6167e-18, 9.8628e-19, 5.1714e-18,\n 7.6750e-18, -4.9817e-18, -9.9075e-18, 2.4360e-18, -3.2801e-18,\n -2.7840e-18, -2.1871e-18, 3.4452e-18, 5.9263e-18, 2.4635e-18,\n 2.0165e-18, 1.1002e-18, -4.4342e-18, 7.4842e-18, 4.0018e-18,\n -5.9698e-18, -1.9271e-18, 6.0424e-18, -4.6254e-18, -3.2951e-18,\n 5.1313e-18, -4.7731e-18, 8.4653e-19, -1.4065e-19, -8.2931e-18,\n -5.1009e-18, -6.8371e-18, -2.6883e-19, -1.3327e-18, -1.9422e-18,\n 1.9521e-18, 1.0834e-18, 1.9671e-18, 7.8235e-19, 2.8126e-18,\n -1.7361e-18, 8.8153e-19, 1.7018e-18, -4.8304e-18, 2.0216e-18,\n -2.4738e-19, -1.7363e-18, 2.9507e-18, -7.0597e-19, 4.5088e-18,\n 5.9446e-18, 2.9233e-18, 2.6915e-18, -3.2966e-18, -3.6330e-18,\n 1.6622e-18, -3.7610e-20, -3.0026e-19, -9.4493e-19, -5.3760e-19,\n 1.1187e-18, -2.2235e-18, 2.9003e-18, 5.1274e-18, 4.7975e-18,\n -1.9907e-18, 1.1581e-18, 8.3302e-18, -2.4898e-18, -2.1588e-18,\n 1.2493e-18, -2.6218e-19, -5.7615e-18, 5.2922e-18, -1.4534e-18,\n -1.1120e-18, 3.7165e-20, 1.9003e-18, 1.9838e-18, 2.7997e-18,\n 7.5110e-18, -1.6465e-18, 8.5000e-19, 5.8157e-18, -3.5265e-18,\n 7.5772e-18, -1.5345e-18, 2.7985e-18, 2.0759e-19, 6.0534e-19,\n 2.5485e-18, 8.9251e-18, -3.1370e-18, -4.3755e-18, 7.8715e-18,\n 1.4903e-18, -2.0248e-18, -3.0146e-18, 1.6604e-18, -2.2921e-19,\n -5.4476e-19, 3.6608e-19, 4.7183e-18, 8.9970e-19, -2.2782e-18,\n -2.9476e-18, -2.4261e-18, -1.0013e-18, -3.1906e-18, -2.9893e-18,\n -3.1940e-18, -2.0360e-18, -1.2025e-18, -4.5483e-18, 2.7451e-18,\n 1.7424e-18, -2.1323e-18, 6.9387e-19, 1.9995e-18, -1.9185e-19,\n -2.2086e-18, -1.1186e-18, 3.0413e-18, -2.1205e-18, 2.4587e-18,\n 2.3484e-21, 1.0184e-18, 4.5859e-19, -3.8695e-19, 5.8423e-19,\n -2.9176e-18, -3.9277e-19, 1.5741e-18, 2.5544e-18, -4.8088e-19,\n -1.1984e-18, 5.7577e-19, 1.1448e-19, 2.8205e-18, 1.3880e-18,\n -3.0085e-18, -2.7281e-18, -3.6276e-20, -2.2702e-18, 2.2978e-18,\n -2.7828e-19, -2.4489e-19, 9.5547e-19, -5.6003e-19, -7.6854e-19,\n -3.3594e-18, 2.8417e-18, -9.9758e-19, -1.4197e-18, -3.8756e-19,\n 4.7709e-19, 2.2433e-18, 1.6040e-18, -6.2441e-19, 3.4350e-18,\n -2.4943e-18, 1.3725e-18, -3.2384e-19, -3.9118e-18, -3.6027e-18,\n -1.3412e-18, 2.1849e-18, -4.1590e-18, -5.6182e-19, 2.0756e-19,\n 1.8360e-18, 2.6066e-18, -8.6172e-19, 2.9965e-18, -4.2780e-19,\n 1.7392e-18, 1.5596e-19, -1.3689e-18, -5.8245e-19, -4.7896e-19,\n -2.1233e-18, -3.5797e-18, -1.4992e-18, -1.9931e-19, 4.0026e-18,\n 4.3126e-18, 3.0510e-19, -1.0831e-18, -2.5158e-18, 4.0122e-18,\n 1.3574e-18, 4.0373e-25, 9.3925e-25, -4.2765e-25, 1.7945e-26,\n 1.5174e-24, 3.1198e-26, 1.2306e-24, 2.0673e-25, 8.8660e-25,\n 5.1765e-25, -1.1120e-25, 3.3189e-25, -1.2998e-24, -5.6213e-25,\n -2.8445e-25, -1.2069e-24, -1.3626e-26, -1.1571e-24, -6.9730e-25,\n 5.1890e-25, -1.0932e-24, -9.5816e-26, 7.5747e-25, -6.9365e-25,\n -1.1168e-24, -9.0396e-25, 9.5995e-25, -3.2089e-25, -1.0428e-24,\n -3.8248e-25, 5.0128e-25, 1.2869e-24, -4.2849e-25, -1.5542e-25,\n 6.2647e-25, 1.7746e-25, -4.4040e-25, 6.7860e-25, -8.7574e-26,\n -2.6727e-25, 5.9696e-25, 2.1340e-25, 2.3381e-25, 2.3039e-26,\n 2.9180e-25, -4.3432e-25, -5.2018e-26, -1.0680e-25, -4.6410e-25,\n -2.3243e-25, 2.7013e-25, -4.3034e-28, -8.0323e-25, -2.2903e-25,\n -4.3458e-25, 6.8971e-25, -3.8142e-25, -4.4121e-25, 9.5349e-26,\n -1.0467e-25, -1.5999e-25, -6.1049e-26, 4.5744e-25, -5.5487e-25,\n -2.3748e-25, 2.5297e-25, -9.1473e-25, 2.7447e-25, 7.0860e-27,\n -4.1126e-25, -1.3300e-24, 9.1179e-25, -1.2106e-25, -4.5943e-25,\n 4.8444e-25, 2.1756e-25, 4.4885e-25, -1.9684e-25, -6.0268e-25,\n -1.1864e-24, 2.7133e-25, -3.1161e-25, 1.2169e-24, 1.5092e-26,\n 2.4275e-25, 5.3449e-25, 3.5173e-25, -1.5982e-26, 9.0832e-25,\n 4.8278e-25, -4.7066e-25, 7.4771e-25, -6.3576e-25, 3.5897e-25,\n 5.4934e-25, 2.2671e-26, 1.8983e-25, 1.5204e-25, 1.8916e-25,\n -1.9989e-25, -8.1031e-25, 3.5262e-25, 5.4539e-25, 3.9595e-25,\n -3.9745e-25, 1.5588e-25, 1.6919e-25, -6.1929e-26, -4.7236e-25,\n 5.7054e-25, 9.9266e-26, 5.9491e-25, -5.5858e-25, 1.1498e-25,\n -7.3361e-25, 8.5490e-26, 3.3605e-26, 4.3039e-25, -3.1377e-26,\n 7.3142e-25, 2.2649e-25, -2.8432e-25, -4.1408e-25, -1.0961e-26,\n -3.6946e-25, -5.4735e-26, -1.7163e-25, -1.4201e-26, 2.6399e-25,\n 3.6939e-25, -1.9475e-25, -1.0025e-24, 4.2126e-25, -3.1374e-25,\n -4.0540e-25, -6.1163e-25, -2.4245e-25, -1.1678e-25, 2.6885e-25,\n 2.4832e-25, 1.7973e-25, 1.7943e-25, -6.7169e-25, -9.6106e-26,\n -4.9966e-26, 1.4598e-25, 1.3393e-25, 1.0200e-24, 5.3906e-25,\n -5.9489e-26, 2.9561e-25, 6.1441e-26, 5.3201e-26, 9.3882e-25,\n -2.4909e-25, 1.9759e-25, 1.5388e-25, -4.8604e-25, 9.1327e-27,\n 2.3400e-25, 9.5078e-26, -2.4078e-25, -3.2496e-26, -1.8308e-25,\n 1.3658e-25, -6.0035e-26, -2.7125e-25, 1.4626e-25, -3.1510e-25,\n -1.8662e-25, -2.0055e-25, 8.1708e-26, -1.0145e-25, -1.2428e-25,\n -9.0535e-27, -4.4238e-25, -2.6668e-25, -3.5776e-26, -2.7548e-25,\n -1.8972e-25, 1.2458e-25, -1.8502e-25, -5.4097e-26, 2.1264e-25,\n -8.9758e-26, -3.4330e-25, -3.6470e-25, 2.8996e-25, -3.6890e-25,\n 8.7049e-26, 3.7389e-26, 1.9447e-25, -2.6517e-25, 3.1244e-25,\n 1.6135e-25, 2.3265e-25, -1.3264e-25, -3.7667e-25, 1.4019e-25,\n 1.1615e-25, -3.3195e-25, 1.9738e-25, -2.9481e-25, 1.3007e-25,\n -1.8215e-26, -1.0909e-25, 4.0610e-25, -2.5483e-25, -2.2853e-25,\n 1.1047e-25, -1.1828e-25, -1.1422e-28, 3.0362e-25, -7.7627e-26,\n 2.3839e-25, -5.0211e-25, 2.3326e-25, 3.7712e-25, 3.0925e-26,\n 1.1844e-25, 6.0371e-26, 3.1614e-26, -1.2454e-25, -9.6744e-26,\n 5.1094e-25, -1.0929e-24, -2.2926e-25, 7.0939e-25, -8.6011e-25,\n -2.3372e-25, -2.2262e-25, 1.0165e-24, -4.7200e-25, -2.3825e-25,\n -3.1184e-25, 8.0544e-25, 1.4707e-25, -1.0031e-24, 1.7060e-25,\n -8.1463e-25, 1.3529e-25, -3.6550e-25, -2.0760e-25, 2.3379e-25,\n 6.5248e-26, 2.5094e-25, -4.1160e-25, -5.0677e-25, -4.1498e-25,\n 2.1803e-25, 8.5780e-25, -7.9530e-25, -1.9039e-25, 1.1026e-25,\n 2.8320e-25, 2.3269e-25, 1.8849e-17, 6.9398e-17, -7.1658e-17,\n -2.7275e-17, 1.0077e-16, -9.8937e-17, -1.0054e-16, 3.2500e-17,\n -1.7816e-17, -8.1453e-18, -7.2072e-17, 5.9816e-17, -3.8986e-17,\n 6.8575e-17, 7.8015e-17, 3.2168e-17, -1.0016e-16, 6.1799e-17,\n -6.3579e-17, 8.6127e-17, -9.3504e-17, 8.0887e-17, -8.9465e-17,\n -8.7897e-17, -5.8590e-17, -1.9424e-17, -1.3402e-16, -7.1992e-17,\n 5.4331e-17, -8.8886e-17, 7.4883e-17, 5.5220e-17, -7.2015e-17,\n 6.9236e-17, -5.4196e-17, -3.2555e-17, 1.1254e-16, 3.4981e-17,\n -1.1486e-17, -9.1974e-17, 5.8166e-18, 1.3425e-16, 1.0705e-16,\n 6.1521e-17, -1.7995e-17, -8.5693e-17, 1.1308e-16, -1.3098e-16,\n -9.6151e-17, -6.4879e-17, 5.8151e-17, -2.8021e-17, 1.9511e-17,\n -8.5869e-17, -3.3810e-17, 7.1429e-17, -6.9798e-17, 8.0174e-17,\n -5.7479e-17, -2.5062e-17, 1.2834e-16, -6.7425e-17, 1.0752e-16,\n 4.4664e-17, 1.4151e-16, 2.3531e-17, -7.2834e-17, 7.8641e-17,\n -1.0938e-16, 1.5863e-16, 2.0598e-18, -1.0723e-16, 7.8158e-17,\n -4.2598e-17, 7.7347e-17, -7.1621e-17, 4.5329e-17, -3.8993e-17,\n 8.6607e-18, -5.5009e-17, 2.8660e-17, -6.0072e-17, 4.7685e-17,\n 7.8858e-17, 7.4414e-17, 3.2399e-17, -3.2851e-17, -3.1659e-17,\n 7.4898e-17, -5.1968e-17, -8.6173e-17, -7.1071e-17, 5.2709e-17,\n 1.1256e-16, -3.7932e-17, 5.7374e-17, -4.5078e-17, 1.4000e-16,\n -9.0328e-19, 6.6624e-17, 2.2207e-17, 6.0845e-17, 4.0485e-17,\n -1.3655e-16, 8.3571e-17, -1.6228e-17, 6.5714e-17, -1.8031e-17,\n -6.9953e-17, -6.0342e-17, 1.2860e-16, 4.1338e-17, 5.7041e-17,\n 8.2119e-17, 6.1936e-17, 6.7574e-17, 1.1874e-16, -1.0332e-16,\n 9.8079e-17, -5.7407e-17, 1.2909e-16, 5.9621e-17, 1.2711e-16,\n 2.5306e-17, 1.1459e-16, -1.2904e-16, -6.8040e-17, 1.7677e-17,\n -1.2625e-16, 2.4758e-17, -8.3904e-17, 7.9509e-17, 2.1792e-17,\n 6.3115e-17, -4.6702e-19, -8.6629e-17, 3.5060e-17, -9.9788e-17,\n 1.0648e-16, -6.7749e-17, -1.4705e-16, 3.1986e-17, -2.8178e-17,\n -5.7364e-17, -5.4130e-17, -7.0807e-17, 8.5983e-17, 6.0001e-17,\n -1.5694e-17, 3.2121e-17, 2.8494e-17, 7.7650e-17, -9.9874e-17,\n 1.2407e-17, -8.4836e-17, -5.5243e-17, -1.0260e-16, -7.3273e-17,\n -6.5683e-17, -9.9817e-17, 1.0518e-16, 6.0762e-17, -1.9001e-17,\n 3.8457e-17, -8.2524e-17, 2.4897e-17, -2.9891e-17, -4.0157e-17,\n 5.5495e-17, 1.2830e-16, -8.2930e-17, -1.9241e-18, 4.0965e-17,\n -7.5516e-17, -4.9963e-17, 2.0638e-17, -1.5288e-16, -9.5107e-17,\n 4.8241e-17, -9.8644e-17, -3.5251e-17, -3.6875e-17, 2.1419e-17,\n -9.1957e-17, 2.0140e-17, -4.0256e-17, 8.2277e-17, 3.7422e-17,\n -5.5998e-17, -6.5204e-17, 8.1976e-18, -8.3895e-17, -1.5319e-17,\n -9.5503e-17, 5.4733e-17, -9.1398e-17, 3.1668e-17, -1.0872e-16,\n 5.4408e-18, 1.7004e-17, 2.5542e-17, 1.1409e-16, 1.1479e-16,\n 8.6782e-17, -2.8165e-17, -7.7800e-17, -2.8408e-17, 1.1624e-16,\n -5.5762e-17, -9.8336e-17, 6.5224e-17, 9.3912e-17, -1.1395e-16,\n 4.5444e-17, 4.3699e-17, -6.0557e-17, 5.8700e-18, -1.0676e-17,\n -3.1348e-17, 1.4608e-17, 2.0375e-17, -6.3009e-17, 4.0490e-17,\n 1.0080e-16, -1.3888e-16, -2.7442e-18, -9.0770e-17, 8.3836e-17,\n 9.6716e-17, 2.7155e-17, -4.1766e-17, -1.0104e-16, -1.0955e-16,\n -8.0188e-17, 3.9317e-17, 1.2961e-17, 6.7594e-17, -1.0869e-16,\n 7.1919e-17, -7.0143e-17, -4.6331e-17, -1.0019e-16, -4.5802e-17,\n -1.0507e-16, -1.1511e-16, -2.1348e-17, 7.3329e-17, -1.0566e-16,\n -7.1485e-17, 4.6769e-17, -5.4865e-17, -8.7419e-17, 6.8831e-17,\n -9.6746e-17, -1.0091e-17, 5.0204e-17], device='cuda:0')", | |
| "exp_avg_sq": "tensor([3.1140e-11, 2.3963e-12, 1.8561e-11, 1.2830e-11, 1.6228e-11, 2.6515e-11,\n 2.9529e-12, 3.4222e-11, 4.4296e-11, 3.1264e-12, 1.7379e-11, 6.1114e-12,\n 4.3772e-11, 9.8848e-13, 5.0407e-11, 7.0156e-12, 1.0000e-11, 8.8150e-12,\n 1.0064e-11, 5.0432e-12, 5.8505e-12, 1.1881e-11, 1.0625e-12, 1.4054e-11,\n 1.8809e-11, 1.2620e-12, 1.9283e-11, 6.4558e-11, 3.7599e-12, 6.4339e-13,\n 1.8355e-11, 2.1116e-11, 8.0045e-11, 3.0809e-13, 6.8161e-12, 8.9536e-12,\n 5.7814e-11, 2.3107e-13, 1.1926e-11, 3.9336e-12, 5.7000e-12, 8.6022e-13,\n 8.0396e-11, 2.2612e-11, 5.0035e-11, 2.2949e-12, 1.2440e-10, 4.9056e-11,\n 1.4169e-11, 1.4068e-12, 3.9008e-12, 7.2964e-11, 2.0123e-12, 1.3713e-11,\n 5.6034e-13, 2.5399e-12, 4.7800e-11, 1.1187e-11, 6.2465e-12, 4.5046e-11,\n 3.5515e-11, 9.0925e-11, 7.7548e-11, 1.9858e-11, 4.6095e-11, 3.3586e-12,\n 8.6283e-12, 1.2659e-11, 1.0215e-12, 6.3816e-11, 6.7998e-11, 5.1420e-12,\n 5.5724e-11, 2.2068e-12, 5.2262e-12, 7.2156e-11, 8.4349e-12, 2.5032e-11,\n 1.0759e-11, 3.3386e-11, 1.5380e-11, 2.4487e-12, 1.4045e-11, 6.2594e-11,\n 5.6885e-11, 1.5992e-11, 7.9956e-12, 3.4278e-11, 1.4859e-11, 4.3785e-12,\n 2.6450e-11, 1.6578e-11, 1.7740e-11, 8.2490e-13, 4.8847e-11, 1.1496e-11,\n 5.6298e-12, 1.7970e-12, 1.7939e-11, 6.0210e-12, 2.0402e-12, 2.1817e-12,\n 7.0004e-12, 1.0109e-11, 4.4627e-12, 5.3574e-12, 2.3540e-12, 1.6965e-11,\n 4.0041e-12, 7.6865e-12, 4.8737e-13, 2.2029e-11, 3.1735e-11, 3.7102e-12,\n 5.8169e-12, 2.5732e-12, 1.0726e-11, 1.1186e-12, 1.2014e-12, 3.9806e-12,\n 1.2466e-12, 1.6178e-12, 4.9930e-12, 9.1053e-12, 6.5918e-12, 3.0069e-13,\n 1.0643e-11, 1.4309e-12, 5.1378e-12, 9.5443e-12, 5.7957e-11, 5.2559e-11,\n 3.6892e-11, 1.6621e-11, 1.9583e-11, 1.4171e-12, 8.5222e-11, 2.6827e-11,\n 1.5456e-11, 9.6226e-12, 2.1987e-11, 2.6332e-12, 3.3346e-12, 3.3232e-12,\n 2.2842e-12, 6.1094e-11, 4.8544e-12, 2.5905e-12, 1.6703e-10, 5.9385e-12,\n 1.5975e-11, 8.0300e-11, 3.9331e-11, 1.1001e-12, 5.1314e-12, 2.6804e-13,\n 8.6710e-11, 3.5114e-12, 2.9775e-11, 4.7848e-11, 1.4171e-13, 1.7410e-13,\n 1.6738e-12, 4.6638e-12, 4.5041e-13, 3.1012e-12, 5.5983e-13, 1.3132e-11,\n 2.2690e-12, 2.6353e-12, 1.5562e-12, 5.8786e-13, 3.3178e-12, 4.5556e-13,\n 3.4456e-13, 1.4262e-12, 1.4564e-13, 5.0271e-12, 1.1580e-12, 2.6034e-12,\n 5.7991e-12, 3.5858e-12, 1.0061e-12, 1.0108e-12, 2.0150e-13, 8.2851e-13,\n 5.7303e-12, 3.0642e-12, 2.1801e-12, 2.5987e-12, 4.0898e-12, 4.2404e-12,\n 6.0798e-12, 5.7587e-11, 1.3709e-10, 1.1133e-10, 5.5119e-11, 1.9927e-12,\n 1.6541e-11, 8.9545e-13, 2.8281e-11, 1.1338e-10, 1.5643e-11, 2.1331e-11,\n 2.1439e-11, 5.9684e-12, 2.0633e-11, 7.9384e-13, 8.8748e-11, 2.0961e-11,\n 6.6400e-11, 2.3932e-12, 3.5660e-11, 7.6232e-13, 1.5945e-13, 1.4688e-10,\n 1.1505e-11, 1.5800e-11, 2.5804e-12, 1.5221e-10, 3.5024e-12, 9.5814e-11,\n 4.6085e-11, 2.9750e-13, 1.0871e-11, 6.1876e-12, 3.3486e-12, 1.4477e-11,\n 3.5690e-12, 1.9648e-13, 2.3626e-13, 7.5599e-12, 5.2982e-12, 3.9704e-12,\n 2.3348e-11, 1.4570e-12, 1.1637e-11, 7.9217e-12, 2.9919e-13, 1.5994e-12,\n 1.1590e-11, 2.3788e-11, 7.4979e-12, 3.3820e-11, 1.7351e-12, 5.5314e-12,\n 5.3323e-11, 1.4682e-11, 1.1865e-12, 4.3782e-12, 9.4071e-12, 4.3118e-12,\n 4.8857e-12, 5.2649e-12, 3.7650e-13, 1.0350e-12, 2.1304e-28, 4.9443e-27,\n 3.0322e-27, 5.6454e-28, 8.2012e-28, 1.5172e-27, 9.1074e-28, 2.6340e-28,\n 5.3784e-28, 2.9354e-28, 1.9582e-27, 2.3241e-28, 1.8454e-27, 2.4646e-28,\n 3.6449e-28, 1.6608e-27, 6.8986e-28, 8.9852e-28, 8.8638e-28, 1.4484e-28,\n 4.2529e-27, 9.1502e-28, 4.0791e-27, 3.4943e-27, 3.2636e-27, 4.9883e-28,\n 2.6066e-27, 1.0938e-27, 2.4024e-27, 6.7919e-28, 3.3172e-28, 4.9442e-27,\n 1.9479e-27, 2.8966e-28, 1.5215e-27, 2.9692e-27, 4.9499e-27, 8.9265e-28,\n 2.1668e-27, 4.4894e-27, 6.8793e-28, 6.2482e-28, 3.2734e-27, 2.3788e-27,\n 1.6722e-27, 2.1738e-27, 4.0880e-27, 9.2812e-28, 1.6652e-27, 7.4751e-28,\n 4.7548e-28, 1.3706e-27, 3.8837e-28, 7.6630e-28, 1.9561e-27, 2.7021e-27,\n 5.6048e-27, 1.9228e-27, 2.6163e-27, 2.0287e-27, 3.2563e-27, 3.6949e-27,\n 1.0229e-27, 1.5887e-27, 8.4336e-27, 6.5493e-28, 4.9251e-27, 3.0154e-27,\n 2.2413e-27, 1.2585e-27, 8.5928e-28, 1.9248e-27, 2.8183e-27, 1.2175e-26,\n 4.4807e-28, 3.1987e-28, 3.2990e-28, 5.6177e-29, 8.8949e-28, 7.1868e-28,\n 3.7578e-28, 9.9212e-28, 3.9686e-27, 1.6054e-27, 6.5728e-28, 9.2386e-28,\n 1.3968e-27, 1.3600e-27, 6.2187e-28, 1.1629e-27, 1.5687e-28, 8.4487e-28,\n 2.1622e-28, 8.0979e-28, 1.1368e-27, 9.7595e-28, 1.2718e-27, 2.7615e-28,\n 7.0244e-28, 1.2493e-27, 6.6411e-28, 2.1753e-28, 3.1539e-28, 6.3941e-28,\n 1.3162e-27, 6.6009e-28, 1.8971e-28, 1.1581e-27, 2.8709e-28, 4.7951e-28,\n 6.4602e-28, 9.1488e-28, 9.5049e-28, 2.9406e-28, 2.4940e-27, 9.1586e-28,\n 4.3522e-28, 6.2242e-28, 4.3724e-28, 1.2664e-27, 5.5797e-28, 8.5897e-28,\n 5.2155e-28, 4.5287e-28, 4.7377e-28, 5.8862e-28, 3.4619e-28, 2.4617e-28,\n 5.2173e-27, 2.8225e-27, 1.2958e-27, 3.0793e-27, 7.3856e-27, 3.6923e-27,\n 7.6368e-28, 5.8386e-27, 8.7409e-29, 6.6658e-28, 1.7320e-27, 9.7185e-27,\n 9.1978e-28, 1.1963e-27, 4.1698e-27, 1.0843e-27, 4.3978e-27, 4.7610e-27,\n 9.4486e-28, 3.7168e-28, 3.2331e-27, 1.1624e-27, 2.3859e-27, 1.1074e-26,\n 7.6552e-28, 1.5767e-27, 1.2438e-27, 4.4971e-27, 7.9837e-27, 8.1330e-27,\n 3.1765e-27, 3.5519e-28, 6.4228e-28, 9.5526e-28, 3.2264e-28, 2.6581e-28,\n 3.5408e-28, 1.0974e-27, 7.3543e-29, 1.1918e-27, 4.2776e-28, 3.6725e-28,\n 7.3074e-29, 5.9396e-28, 1.9578e-28, 8.1496e-28, 4.5841e-28, 1.4116e-27,\n 3.3121e-28, 6.5607e-28, 1.6095e-28, 1.4990e-28, 5.3552e-29, 1.3096e-28,\n 3.9456e-28, 8.8556e-29, 7.5968e-28, 5.3444e-28, 1.6076e-27, 3.1500e-28,\n 1.7492e-27, 1.3684e-28, 1.6256e-28, 3.4792e-28, 2.4990e-28, 5.6812e-27,\n 3.5747e-28, 1.0900e-27, 8.7224e-28, 3.2360e-27, 7.3048e-28, 3.1584e-28,\n 1.0435e-27, 2.0523e-27, 3.8832e-28, 1.6992e-28, 3.8569e-27, 1.3625e-28,\n 4.1804e-28, 6.1953e-28, 4.3522e-28, 2.6236e-27, 2.0015e-27, 2.5152e-27,\n 2.1141e-28, 6.7136e-28, 1.3561e-28, 5.3915e-28, 4.6964e-28, 1.4660e-27,\n 7.9697e-28, 1.5154e-27, 6.7335e-28, 6.7146e-28, 6.9997e-28, 1.7043e-28,\n 2.8601e-27, 5.3738e-27, 2.9800e-27, 1.1212e-27, 6.3966e-27, 5.6312e-27,\n 2.1938e-27, 2.5401e-27, 8.8790e-28, 4.4672e-27, 3.5003e-27, 1.7670e-27,\n 1.2915e-27, 1.9505e-27, 2.4814e-27, 4.7808e-28, 2.8492e-27, 1.1243e-27,\n 1.7207e-27, 5.6564e-28, 7.0067e-28, 4.6593e-27, 1.6097e-27, 7.3337e-27,\n 1.1214e-27, 4.5803e-28, 3.2714e-27, 4.0396e-27, 3.6514e-27, 1.3065e-27,\n 3.8643e-28, 3.6901e-27, 6.0254e-09, 9.4151e-10, 2.5435e-09, 1.0914e-09,\n 1.5337e-08, 1.2674e-09, 1.0263e-08, 1.5727e-09, 2.7763e-09, 1.5638e-10,\n 1.9496e-10, 3.0144e-09, 2.8789e-09, 3.9628e-10, 1.0912e-09, 1.1965e-08,\n 3.6162e-09, 9.3543e-10, 6.7174e-09, 1.4593e-09, 6.6713e-09, 1.0131e-09,\n 1.6699e-09, 3.8085e-10, 8.4906e-10, 8.7371e-10, 9.5597e-09, 1.0697e-08,\n 7.0435e-09, 7.3911e-09, 2.6088e-09, 1.4074e-09, 2.2137e-09, 3.3687e-09,\n 5.2744e-09, 1.7781e-10, 2.2084e-08, 2.3207e-09, 3.9394e-10, 4.3545e-09,\n 8.3169e-10, 4.9032e-09, 1.0635e-08, 1.2500e-08, 5.6813e-09, 5.2012e-10,\n 1.5046e-08, 3.4878e-09, 4.7874e-09, 1.0931e-09, 1.0799e-09, 6.8610e-09,\n 1.3267e-08, 6.0398e-09, 2.4405e-09, 1.9703e-09, 3.4516e-09, 6.5884e-10,\n 1.2369e-09, 1.4221e-09, 2.1840e-08, 2.1725e-09, 1.5712e-09, 2.2752e-10,\n 2.4916e-09, 2.4405e-10, 4.4560e-09, 1.0623e-09, 5.4032e-09, 2.2436e-08,\n 9.8296e-10, 7.0374e-09, 9.7908e-09, 1.6001e-09, 9.9554e-09, 1.2251e-09,\n 1.3298e-10, 3.5000e-10, 1.0022e-10, 1.5368e-09, 1.4263e-09, 1.1420e-08,\n 5.7545e-10, 6.9708e-09, 2.3969e-09, 1.2290e-09, 1.7493e-09, 1.2782e-10,\n 2.9271e-09, 1.6603e-09, 1.5072e-09, 9.1275e-09, 1.3117e-09, 1.2394e-08,\n 1.0739e-08, 9.5292e-10, 3.8715e-09, 1.2264e-08, 9.4565e-09, 3.6422e-09,\n 5.3822e-11, 4.0385e-09, 1.0672e-09, 6.7864e-09, 1.7749e-08, 5.0388e-10,\n 2.8704e-10, 1.0107e-09, 5.7774e-10, 1.0802e-09, 1.9447e-08, 7.7355e-10,\n 3.7339e-10, 2.0011e-09, 3.9089e-09, 4.2978e-09, 1.0626e-08, 8.1373e-09,\n 1.2200e-09, 1.1616e-09, 1.1281e-08, 3.0039e-10, 1.2841e-08, 4.9970e-10,\n 1.0229e-09, 4.6304e-09, 9.1501e-10, 1.7679e-10, 9.5161e-09, 2.3805e-09,\n 5.9202e-09, 7.7072e-09, 3.0282e-10, 3.8331e-09, 3.4743e-10, 2.9404e-09,\n 2.1049e-09, 9.7418e-09, 7.4703e-09, 5.0040e-10, 8.3940e-09, 3.7675e-09,\n 3.9961e-10, 5.5797e-10, 9.1047e-09, 4.7776e-09, 8.1895e-10, 2.4047e-09,\n 3.8539e-09, 1.2262e-09, 2.6098e-10, 4.6273e-10, 1.2710e-08, 2.4815e-10,\n 2.8263e-09, 7.1126e-10, 1.0981e-08, 2.7817e-09, 1.0464e-09, 6.3462e-09,\n 4.0733e-09, 1.4307e-09, 7.4451e-09, 3.9390e-10, 6.2541e-09, 2.8152e-10,\n 1.1716e-09, 1.5513e-08, 8.3570e-10, 4.3619e-09, 8.1669e-10, 3.5754e-09,\n 1.9887e-09, 2.7465e-09, 4.5149e-09, 1.1331e-09, 6.2013e-09, 1.5086e-09,\n 9.6127e-10, 1.5787e-09, 7.1367e-10, 1.8002e-10, 1.9941e-11, 1.6223e-09,\n 3.9051e-09, 7.1302e-11, 3.1464e-09, 9.0153e-10, 6.1805e-09, 3.1567e-09,\n 5.3629e-09, 1.5584e-09, 1.3651e-09, 3.3559e-09, 7.4700e-10, 5.9758e-09,\n 6.1676e-10, 1.9600e-09, 1.0936e-10, 8.7262e-09, 6.7986e-10, 2.4559e-09,\n 4.9111e-09, 2.6799e-09, 3.6003e-09, 5.4742e-09, 7.3458e-09, 4.5320e-09,\n 2.6133e-09, 1.6766e-09, 2.3996e-09, 5.7263e-09, 2.1010e-08, 3.5071e-09,\n 2.1431e-09, 8.6488e-09, 4.1954e-09, 2.5598e-09, 3.8599e-09, 9.8779e-10,\n 1.6284e-08, 8.4985e-10, 4.0543e-09, 1.1973e-08, 3.5481e-09, 1.0868e-10,\n 2.1306e-09, 1.8631e-09, 1.9402e-08, 4.5774e-09, 1.1757e-10, 4.4551e-09,\n 1.0751e-08, 6.9173e-09, 3.9721e-09, 6.6857e-09, 9.7531e-10, 2.1298e-08,\n 5.9939e-10, 1.0243e-08, 2.4035e-09, 1.2622e-09, 2.6983e-10, 2.6307e-08,\n 2.2211e-09, 2.4154e-09, 7.6101e-09, 3.7478e-09, 9.7987e-09, 5.1288e-10,\n 8.8032e-09, 5.3845e-09, 7.9230e-09, 2.9622e-10, 9.6402e-09, 3.3086e-09],\n device='cuda:0')" | |
| }, | |
| "54": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([[-4.0109e-17, 3.5336e-17, -2.1082e-17, ..., -2.9808e-17,\n -6.1099e-17, 5.0642e-17],\n [-1.6411e-17, 1.2242e-17, -6.7757e-18, ..., -6.0089e-18,\n -1.1867e-17, 1.9449e-17],\n [ 4.1428e-17, -3.4704e-17, 1.7515e-17, ..., 2.5885e-17,\n 4.8464e-17, -5.0744e-17],\n ...,\n [ 4.2969e-17, -3.7083e-17, 1.7348e-17, ..., 3.2594e-17,\n 6.4548e-17, -5.2939e-17],\n [ 6.1421e-17, -5.4584e-17, 3.3635e-17, ..., 4.2042e-17,\n 7.8243e-17, -8.0192e-17],\n [ 1.5995e-17, -1.5265e-17, 9.3474e-18, ..., 1.2094e-17,\n 1.8353e-17, -2.3558e-17]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[9.3695e-11, 5.7403e-10, 1.5431e-09, ..., 7.8198e-12, 8.9398e-10,\n 1.0015e-09],\n [1.0930e-10, 8.2507e-10, 1.8562e-09, ..., 1.1915e-11, 5.7745e-10,\n 9.6402e-10],\n [7.7449e-11, 3.1551e-10, 4.5714e-10, ..., 7.4311e-12, 2.6487e-10,\n 3.9094e-10],\n ...,\n [1.2385e-10, 6.0850e-10, 2.1549e-09, ..., 1.4509e-11, 1.5385e-09,\n 1.1952e-09],\n [1.7923e-11, 1.1698e-10, 3.1261e-10, ..., 2.1382e-12, 2.5286e-10,\n 2.9886e-10],\n [8.2066e-11, 4.1484e-10, 1.1892e-09, ..., 2.1068e-12, 4.0633e-10,\n 5.5370e-10]], device='cuda:0')" | |
| }, | |
| "55": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([-6.3707e-17, -1.7075e-17, 5.6217e-17, -1.3613e-17, 6.0958e-17,\n -3.5958e-17, 5.0873e-17, -3.2798e-17, 1.1062e-16, -3.6101e-19,\n 1.9178e-17, -3.8767e-17, -3.9216e-17, -2.0972e-17, -1.0700e-16,\n -8.2327e-17, 3.0869e-17, -7.3505e-17, -4.8911e-18, 3.6039e-17,\n -7.4097e-17, -5.1379e-17, -1.1611e-16, 1.3635e-17, -7.7106e-17,\n -5.0435e-17, 3.6409e-17, -2.7661e-17, 3.1308e-17, 6.7451e-17,\n 8.0115e-17, 5.3395e-17, 6.7240e-17, -2.4169e-17, 6.8497e-17,\n 5.6713e-17, 1.0520e-16, 7.0843e-17, 5.2612e-18, 6.4697e-17,\n 2.1818e-17, 5.5787e-17, 3.2326e-17, -5.3297e-17, 8.0635e-17,\n -3.7225e-17, 1.6301e-17, -7.0542e-17, 2.5017e-17, -1.1938e-17,\n -6.9845e-17, 7.6054e-18, 5.5564e-17, 1.8432e-17, 1.1692e-16,\n -9.3196e-17, -1.6857e-17, 6.2076e-17, -7.4205e-17, 7.7936e-17,\n 6.9725e-17, -7.1921e-17, 3.3687e-17, 5.0976e-17, -5.4901e-17,\n -3.9019e-17, -1.0464e-16, -6.4359e-17, -5.0035e-17, -9.9651e-18,\n -9.7337e-17, 1.0385e-16, 8.7490e-17, 4.0176e-17, -7.9093e-17,\n -5.9321e-17, -5.7306e-17, -3.8385e-17, -6.3864e-17, 2.4086e-17,\n -9.0150e-17, -3.4003e-18, -4.6446e-17, 2.8499e-17, 9.6891e-17,\n -2.8692e-17, 1.6809e-17, 4.6182e-17, -9.3972e-18, -4.2842e-17,\n 3.2983e-17, 6.1812e-17, 9.4631e-18, -8.3366e-17, 3.6110e-17,\n -3.7094e-17, 2.8019e-17, -2.5960e-17, 7.8544e-17, 3.7489e-17,\n 6.8530e-18, -3.2560e-17, -4.3584e-17, 6.2654e-17, 9.1673e-17,\n 4.0137e-17, 5.9690e-17, -1.0331e-17, -4.1601e-17, 9.2430e-17,\n -5.6919e-17, -8.8555e-17, 5.8654e-17, -2.4149e-17, 4.9411e-18,\n -5.7823e-17, 4.0004e-17, -4.4708e-17, 4.6763e-17, 5.6783e-17,\n -4.2427e-17, -9.3185e-19, -7.1637e-17, 2.3900e-17, -1.0667e-16,\n -1.4232e-17, 7.9901e-17, -1.8400e-17, -6.8269e-17, -2.6865e-17,\n 9.3482e-17, -4.6481e-17, -7.5542e-17, -6.3225e-17, -7.7561e-17,\n 8.2151e-17, 6.2755e-17, -3.4563e-17, 5.6410e-17, 5.9169e-17,\n -1.0462e-16, 6.9995e-17, -4.7969e-17, -1.0328e-16, -3.5858e-17,\n 1.9190e-17, -1.4569e-17, 3.3762e-17, 1.0714e-16, 8.7992e-17,\n 3.1108e-17, -3.5583e-18, -9.3119e-17, 3.7003e-17, 8.4984e-17,\n 8.6515e-18, 6.6288e-17, -7.5520e-17, -3.2366e-17, 3.9358e-17,\n 5.5312e-17, -4.5400e-17, -4.2268e-17, 8.7668e-17, 8.9159e-18,\n 4.0661e-17, -1.0624e-16, 6.1957e-17, -4.1488e-18, 1.0165e-16,\n -1.2325e-17, 1.6074e-17, -1.4827e-16, 6.5558e-17, -8.8140e-17,\n 1.4749e-16, 6.5179e-17, 4.8757e-17, -3.7748e-18, -1.0563e-16,\n -2.3367e-17, -4.4339e-17, 6.7611e-17, 4.2160e-18, -5.1677e-17,\n 1.9994e-17, -1.1753e-16, 2.7552e-17, 5.2414e-17, -8.5418e-17,\n -5.2086e-17, -1.1254e-16, 2.1735e-17, -8.2007e-17, -7.1968e-17,\n 2.1344e-17, -2.1404e-17, 4.7535e-17, 6.3199e-17, 3.7900e-17,\n 2.8340e-17, 6.6347e-17, 7.9784e-17, -1.2346e-16, 7.8953e-17,\n 2.5162e-17, 3.7605e-17, 1.1588e-16, 8.7505e-17, 2.3282e-17,\n -7.3727e-17, -3.3000e-17, 5.8078e-17, 6.4655e-17, 5.9929e-17,\n 1.4426e-17, -8.5109e-17, 8.1272e-17, -4.2691e-17, 7.6618e-17,\n -1.4523e-17, 1.1733e-16, 4.8460e-17, 3.4366e-17, -4.1602e-17,\n 3.1188e-17, -6.3464e-17, -1.1119e-16, 7.6041e-17, 9.0331e-17,\n 3.3036e-17, -5.8505e-17, -7.0299e-17, -8.1871e-17, 7.2381e-17,\n -8.2535e-17, 5.1223e-17, 7.1051e-18, -7.1734e-17, -4.9915e-17,\n -7.6812e-17, -3.4438e-17, -1.3162e-17, -4.8098e-17, -7.5246e-17,\n 5.6203e-18, -4.3828e-17, 1.2911e-17, 3.5860e-17, 5.4144e-17,\n 2.8594e-17, 1.1451e-16, -2.0845e-17, 6.5899e-17, 9.1747e-17,\n 2.4345e-17], device='cuda:0')", | |
| "exp_avg_sq": "tensor([7.7720e-09, 1.0480e-08, 2.5552e-09, 4.3517e-08, 5.4991e-08, 1.9488e-09,\n 2.4164e-09, 5.8980e-10, 2.7231e-09, 3.1336e-10, 6.0546e-10, 2.3936e-09,\n 3.2650e-09, 1.1716e-08, 2.7353e-09, 6.2563e-10, 1.3222e-08, 1.3714e-08,\n 5.7194e-09, 8.1950e-09, 4.9303e-08, 3.3272e-08, 6.0572e-09, 2.5532e-08,\n 1.3090e-08, 3.1910e-08, 3.2832e-10, 1.8720e-08, 1.4971e-09, 1.5975e-09,\n 5.1534e-09, 3.4949e-09, 1.7989e-09, 1.8090e-09, 3.3612e-08, 1.9415e-09,\n 9.1793e-08, 8.3582e-09, 1.2683e-08, 4.9924e-09, 1.2166e-09, 2.4770e-09,\n 2.1615e-09, 9.6527e-09, 1.4212e-08, 2.2702e-08, 9.4996e-10, 5.8298e-09,\n 1.3200e-09, 2.7358e-09, 5.1848e-09, 1.4442e-09, 1.5918e-08, 2.1919e-09,\n 6.1564e-08, 5.6291e-09, 2.7325e-09, 4.0517e-08, 2.2887e-08, 4.7497e-10,\n 4.3596e-08, 2.5640e-08, 2.7560e-08, 8.4445e-09, 5.1275e-09, 3.4154e-08,\n 3.1459e-08, 4.7918e-10, 6.0687e-10, 2.7239e-08, 4.2839e-08, 6.2183e-09,\n 8.0240e-09, 1.4029e-09, 2.3266e-09, 2.2691e-08, 5.7162e-10, 7.5165e-09,\n 2.3159e-08, 1.4284e-09, 3.6112e-09, 2.5704e-09, 1.2896e-08, 5.5280e-09,\n 6.3761e-09, 1.4263e-09, 3.6416e-08, 4.6350e-10, 5.4227e-09, 2.7528e-09,\n 1.9021e-09, 1.1335e-08, 4.5175e-08, 6.3180e-09, 5.9071e-10, 5.1792e-09,\n 1.9140e-09, 4.5817e-09, 1.1255e-08, 3.6737e-10, 1.3273e-08, 6.0201e-10,\n 5.0950e-09, 2.1576e-08, 2.7475e-09, 5.7123e-09, 6.0859e-10, 5.6330e-09,\n 1.5042e-08, 4.1233e-09, 6.9671e-09, 4.5229e-09, 3.3321e-09, 4.1588e-10,\n 1.9511e-09, 3.1024e-09, 3.6923e-09, 1.2250e-08, 1.0454e-08, 6.6749e-09,\n 8.8718e-09, 2.5461e-09, 6.3056e-09, 1.4731e-09, 4.2144e-08, 2.8582e-09,\n 2.0352e-09, 2.4094e-09, 1.0159e-08, 1.1285e-08, 1.0922e-08, 3.2765e-08,\n 4.4714e-10, 2.4197e-08, 2.1735e-08, 1.9369e-09, 1.7766e-08, 2.4587e-09,\n 4.5208e-09, 8.0779e-09, 6.0530e-09, 8.4825e-09, 1.6102e-08, 2.1477e-08,\n 8.8817e-10, 4.1929e-09, 1.3031e-09, 1.5762e-10, 8.2186e-10, 2.8962e-09,\n 3.3546e-10, 3.2251e-08, 4.1661e-09, 9.0374e-09, 3.2905e-08, 6.0922e-09,\n 7.6446e-09, 4.0986e-08, 2.7014e-10, 2.1923e-09, 2.7134e-09, 9.3935e-10,\n 2.6942e-09, 2.7256e-08, 1.9333e-08, 9.3367e-09, 1.4880e-08, 2.3244e-08,\n 2.3975e-09, 6.3981e-09, 5.6286e-10, 7.3600e-09, 1.2597e-08, 5.8805e-09,\n 9.0869e-09, 7.7592e-09, 6.0365e-10, 1.2037e-08, 1.0169e-08, 5.7531e-10,\n 7.0082e-09, 4.3133e-09, 1.3425e-09, 2.3821e-08, 1.8116e-09, 4.2798e-08,\n 1.7994e-09, 5.3969e-10, 1.1688e-09, 8.9857e-10, 1.6722e-08, 1.1762e-08,\n 5.9606e-09, 4.9706e-09, 3.4676e-08, 4.6682e-09, 8.8037e-09, 3.0399e-10,\n 3.9794e-08, 2.3782e-09, 1.6431e-08, 6.0877e-09, 1.1765e-08, 7.1724e-09,\n 6.1211e-09, 4.5542e-10, 1.5196e-09, 1.9349e-08, 1.0927e-08, 2.8562e-08,\n 2.6138e-08, 2.8238e-08, 4.0938e-09, 9.0923e-09, 4.1735e-09, 3.0578e-08,\n 1.9840e-08, 1.8828e-08, 2.7392e-08, 4.9824e-10, 8.3352e-10, 1.4009e-09,\n 8.2420e-09, 2.8726e-08, 1.0619e-08, 2.2756e-09, 5.5602e-09, 4.0940e-08,\n 5.2656e-09, 1.7133e-08, 8.9904e-10, 1.3387e-08, 2.0073e-08, 1.5202e-09,\n 2.8397e-09, 5.9318e-09, 8.4652e-09, 2.7086e-10, 1.3364e-08, 4.4634e-10,\n 9.8934e-10, 2.9690e-09, 1.3971e-08, 5.6140e-10, 9.7181e-09, 2.9964e-09,\n 3.4173e-10, 4.1961e-10, 2.2005e-08, 4.9265e-10, 5.4290e-09, 2.0263e-08,\n 3.0200e-09, 1.1182e-08, 1.4868e-09, 6.2689e-09], device='cuda:0')" | |
| }, | |
| "56": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([[-5.6052e-45, -5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n [-5.0613e-17, -8.3266e-17, 6.8294e-17, ..., 1.2201e-16,\n 7.9579e-17, -5.1603e-17],\n [-1.6604e-17, -2.6433e-17, 2.1157e-17, ..., 3.8177e-17,\n 2.4863e-17, -1.5931e-17],\n ...,\n [ 9.7539e-17, 1.5699e-16, -1.2257e-16, ..., -2.2888e-16,\n -1.4330e-16, 9.2918e-17],\n [ 3.9188e-16, 6.3078e-16, -4.9688e-16, ..., -9.3549e-16,\n -5.9109e-16, 3.8188e-16],\n [ 1.1690e-16, 1.8748e-16, -1.4863e-16, ..., -2.7755e-16,\n -1.7612e-16, 1.1346e-16]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[7.6573e-09, 2.8801e-08, 7.2075e-08, ..., 2.5373e-07, 1.1422e-07,\n 1.7248e-08],\n [9.6145e-11, 4.1319e-10, 8.9049e-10, ..., 3.1275e-09, 1.4018e-09,\n 2.3474e-10],\n [8.8237e-11, 3.2193e-10, 8.2816e-10, ..., 2.9352e-09, 1.3188e-09,\n 1.9305e-10],\n ...,\n [1.0899e-10, 4.1659e-10, 1.0742e-09, ..., 3.6905e-09, 1.6902e-09,\n 2.6770e-10],\n [1.7065e-10, 5.8761e-10, 1.2682e-09, ..., 5.3233e-09, 2.1505e-09,\n 2.3753e-10],\n [8.9908e-11, 3.1364e-10, 7.8217e-10, ..., 2.8660e-09, 1.2611e-09,\n 1.7349e-10]], device='cuda:0')" | |
| }, | |
| "57": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([-5.6052e-45, -5.9367e-17, -1.8305e-17, -9.2125e-18, -3.7014e-18,\n 9.5823e-17, 1.1422e-16, 1.0924e-16, 4.4299e-16, 1.3217e-16],\n device='cuda:0')", | |
| "exp_avg_sq": "tensor([2.2591e-06, 2.7141e-08, 2.6113e-08, 2.0546e-08, 2.4686e-08, 2.2910e-08,\n 3.8406e-08, 3.4348e-08, 3.7103e-08, 2.4158e-08], device='cuda:0')" | |
| }, | |
| "58": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([[-5.6052e-45, -5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n [-5.0053e-17, -8.2345e-17, 6.7538e-17, ..., 1.2066e-16,\n 7.8699e-17, -5.1032e-17],\n [-1.6421e-17, -2.6140e-17, 2.0923e-17, ..., 3.7754e-17,\n 2.4588e-17, -1.5755e-17],\n ...,\n [ 9.6459e-17, 1.5526e-16, -1.2121e-16, ..., -2.2635e-16,\n -1.4171e-16, 9.1890e-17],\n [ 3.8755e-16, 6.2380e-16, -4.9138e-16, ..., -9.2513e-16,\n -5.8455e-16, 3.7765e-16],\n [ 1.1561e-16, 1.8540e-16, -1.4698e-16, ..., -2.7447e-16,\n -1.7418e-16, 1.1220e-16]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[7.6180e-09, 2.8683e-08, 7.1861e-08, ..., 2.5259e-07, 1.1382e-07,\n 1.7228e-08],\n [9.5616e-11, 4.1151e-10, 8.8786e-10, ..., 3.1126e-09, 1.3967e-09,\n 2.3450e-10],\n [8.7789e-11, 3.2061e-10, 8.2568e-10, ..., 2.9223e-09, 1.3142e-09,\n 1.9282e-10],\n ...,\n [1.0844e-10, 4.1496e-10, 1.0711e-09, ..., 3.6745e-09, 1.6846e-09,\n 2.6742e-10],\n [1.6972e-10, 5.8458e-10, 1.2635e-09, ..., 5.2967e-09, 2.1415e-09,\n 2.3707e-10],\n [8.9445e-11, 3.1227e-10, 7.7970e-10, ..., 2.8529e-09, 1.2565e-09,\n 1.7327e-10]], device='cuda:0')" | |
| }, | |
| "59": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([-5.6052e-45, -5.8710e-17, -1.8102e-17, -9.1105e-18, -3.6604e-18,\n 9.4763e-17, 1.1296e-16, 1.0803e-16, 4.3808e-16, 1.3071e-16],\n device='cuda:0')", | |
| "exp_avg_sq": "tensor([2.2536e-06, 2.7076e-08, 2.6048e-08, 2.0501e-08, 2.4632e-08, 2.2854e-08,\n 3.8318e-08, 3.4266e-08, 3.6987e-08, 2.4094e-08], device='cuda:0')" | |
| }, | |
| "60": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([[-5.6052e-45, -5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n [-5.0686e-17, -8.3386e-17, 6.8392e-17, ..., 1.2218e-16,\n 7.9694e-17, -5.1677e-17],\n [-1.6628e-17, -2.6471e-17, 2.1188e-17, ..., 3.8232e-17,\n 2.4899e-17, -1.5954e-17],\n ...,\n [ 9.7679e-17, 1.5722e-16, -1.2275e-16, ..., -2.2921e-16,\n -1.4351e-16, 9.3052e-17],\n [ 3.9245e-16, 6.3169e-16, -4.9759e-16, ..., -9.3684e-16,\n -5.9195e-16, 3.8243e-16],\n [ 1.1707e-16, 1.8775e-16, -1.4884e-16, ..., -2.7795e-16,\n -1.7638e-16, 1.1362e-16]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[7.6577e-09, 2.8803e-08, 7.2076e-08, ..., 2.5373e-07, 1.1422e-07,\n 1.7248e-08],\n [9.6151e-11, 4.1322e-10, 8.9051e-10, ..., 3.1276e-09, 1.4018e-09,\n 2.3475e-10],\n [8.8241e-11, 3.2194e-10, 8.2817e-10, ..., 2.9353e-09, 1.3188e-09,\n 1.9305e-10],\n ...,\n [1.0899e-10, 4.1661e-10, 1.0742e-09, ..., 3.6907e-09, 1.6903e-09,\n 2.6770e-10],\n [1.7067e-10, 5.8766e-10, 1.2682e-09, ..., 5.3235e-09, 2.1506e-09,\n 2.3754e-10],\n [8.9913e-11, 3.1366e-10, 7.8218e-10, ..., 2.8661e-09, 1.2611e-09,\n 1.7349e-10]], device='cuda:0')" | |
| }, | |
| "61": { | |
| "step": "tensor(5008.)", | |
| "exp_avg": "tensor([-5.6052e-45, -5.9453e-17, -1.8331e-17, -9.2258e-18, -3.7067e-18,\n 9.5961e-17, 1.1439e-16, 1.0940e-16, 4.4363e-16, 1.3236e-16],\n device='cuda:0')", | |
| "exp_avg_sq": "tensor([2.2592e-06, 2.7141e-08, 2.6113e-08, 2.0546e-08, 2.4686e-08, 2.2910e-08,\n 3.8406e-08, 3.4348e-08, 3.7103e-08, 2.4158e-08], device='cuda:0')" | |
| }, | |
| "8": { | |
| "step": "tensor(3756.)", | |
| "exp_avg": "tensor([[ 2.3408e-06, -4.4076e-07, 3.6113e-06, ..., -8.8469e-06,\n 1.6432e-06, -7.9093e-07],\n [ 4.8357e-06, 1.7113e-05, 4.5901e-07, ..., 1.2441e-06,\n -1.9744e-06, 1.0176e-06],\n [-1.0319e-05, -3.0483e-06, 8.3586e-07, ..., -1.2833e-06,\n 7.2613e-07, 1.5204e-06],\n ...,\n [-2.6922e-06, 8.7933e-07, -8.5255e-06, ..., 2.2636e-06,\n 2.6625e-08, -2.0135e-07],\n [ 3.7722e-06, -5.0037e-05, 2.1919e-05, ..., 9.2709e-07,\n -1.8431e-05, -2.2440e-07],\n [ 3.0550e-06, 1.5188e-06, 2.5970e-06, ..., 2.7710e-06,\n 2.3964e-06, -2.5696e-07]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[5.5349e-10, 4.5716e-10, 2.5724e-10, ..., 3.1153e-10, 4.4539e-10,\n 2.6719e-10],\n [1.8946e-09, 1.0468e-09, 5.8869e-11, ..., 6.0623e-10, 2.3425e-10,\n 2.9193e-10],\n [1.9738e-09, 3.1997e-10, 8.9530e-11, ..., 1.1827e-10, 1.7271e-10,\n 5.0219e-10],\n ...,\n [9.9602e-10, 2.8127e-10, 1.3937e-10, ..., 1.9612e-10, 1.6015e-10,\n 1.3582e-10],\n [8.0738e-10, 6.7157e-09, 5.1707e-10, ..., 2.1609e-09, 4.3988e-10,\n 2.2537e-10],\n [4.8331e-10, 9.2093e-11, 2.9637e-10, ..., 5.0955e-10, 3.0519e-10,\n 1.7842e-10]], device='cuda:0')" | |
| }, | |
| "9": { | |
| "step": "tensor(3756.)", | |
| "exp_avg": "tensor([ 0.0011, 0.0010, 0.0006, ..., -0.0012, 0.0006, 0.0002],\n device='cuda:0')", | |
| "exp_avg_sq": "tensor([4.5729e-06, 5.3169e-06, 6.4973e-06, ..., 5.3954e-06, 7.5306e-06,\n 5.2422e-06], device='cuda:0')" | |
| }, | |
| "10": { | |
| "step": "tensor(3756.)", | |
| "exp_avg": "tensor([[-1.3731e-06, 4.1984e-06, 1.0424e-06, ..., 7.0934e-06,\n 1.1541e-05, -5.8902e-07],\n [-1.4363e-07, 2.0176e-06, -5.5787e-06, ..., 5.1679e-06,\n 8.6322e-07, -2.1557e-06],\n [-5.9027e-06, 4.6774e-06, 5.7582e-06, ..., -8.4540e-06,\n 3.7427e-06, 1.9004e-06],\n ...,\n [ 8.6274e-07, -1.2614e-05, 3.8509e-06, ..., 1.1077e-06,\n -9.0771e-06, 4.7832e-06],\n [-1.4042e-06, 2.5466e-06, 6.4350e-06, ..., -5.8845e-07,\n 2.6080e-05, -1.1520e-06],\n [ 1.8947e-06, -3.9506e-06, -1.6682e-06, ..., 4.5621e-06,\n -5.5422e-06, 5.2093e-06]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[1.3193e-10, 1.7846e-10, 1.9963e-10, ..., 1.6523e-10, 1.3176e-10,\n 1.6331e-10],\n [2.5075e-10, 2.5249e-10, 3.4105e-10, ..., 2.4727e-10, 2.4920e-10,\n 2.0366e-10],\n [2.7342e-10, 2.9457e-10, 4.9393e-10, ..., 4.4647e-10, 3.3752e-10,\n 1.6528e-10],\n ...,\n [2.5067e-10, 2.8506e-10, 3.4854e-10, ..., 2.7197e-10, 2.5355e-10,\n 2.0678e-10],\n [2.9403e-10, 3.7680e-10, 5.1603e-10, ..., 3.4682e-10, 4.2829e-10,\n 2.5861e-10],\n [2.4841e-10, 3.8732e-10, 4.5181e-10, ..., 2.7507e-10, 2.3437e-10,\n 3.1526e-10]], device='cuda:0')" | |
| }, | |
| "11": { | |
| "step": "tensor(2504.)", | |
| "exp_avg": "tensor([[ 2.9899e-06, 6.0771e-06, 5.9951e-06, ..., -3.1146e-06,\n 1.6999e-07, 2.5472e-06],\n [-1.8563e-05, 3.2364e-06, -2.0522e-06, ..., 1.9069e-07,\n 1.3308e-06, 8.5951e-07],\n [-1.4405e-07, 4.2691e-05, -1.9207e-06, ..., -9.6256e-09,\n 3.0577e-07, 2.0083e-06],\n ...,\n [ 1.7840e-06, 1.4268e-06, 1.2129e-05, ..., 2.8027e-06,\n 2.7883e-06, 3.3274e-06],\n [-3.2223e-06, -2.1989e-06, -1.4747e-06, ..., -1.4368e-06,\n -3.7994e-06, 2.4822e-06],\n [-3.6316e-06, 4.7333e-06, 1.0054e-05, ..., 9.9365e-08,\n 7.3783e-06, 6.5952e-07]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[4.9164e-10, 2.5462e-10, 1.2643e-10, ..., 2.4979e-10, 1.7931e-10,\n 3.6967e-10],\n [1.1190e-09, 4.4031e-10, 1.2929e-10, ..., 3.9732e-10, 1.7782e-10,\n 1.7968e-10],\n [6.5130e-10, 9.2553e-10, 2.2866e-10, ..., 3.5372e-10, 2.2528e-10,\n 8.6751e-11],\n ...,\n [5.0820e-10, 2.0470e-10, 8.2508e-10, ..., 1.9359e-10, 2.3070e-10,\n 1.2800e-09],\n [8.9122e-10, 1.5829e-10, 2.6534e-10, ..., 1.6346e-10, 1.8256e-10,\n 6.4885e-10],\n [2.1672e-09, 7.7055e-10, 2.6815e-10, ..., 2.9814e-10, 3.4232e-10,\n 4.3202e-10]], device='cuda:0')" | |
| }, | |
| "12": { | |
| "step": "tensor(2504.)", | |
| "exp_avg": "tensor([ 0.0005, -0.0007, 0.0003, ..., 0.0008, -0.0001, 0.0002],\n device='cuda:0')", | |
| "exp_avg_sq": "tensor([4.9243e-06, 5.8197e-06, 4.3481e-06, ..., 6.8969e-06, 5.2874e-06,\n 6.4128e-06], device='cuda:0')" | |
| }, | |
| "13": { | |
| "step": "tensor(2504.)", | |
| "exp_avg": "tensor([[ 8.7628e-07, 9.6780e-07, -2.8217e-06, ..., -1.0238e-06,\n 2.4372e-06, -3.4876e-06],\n [ 4.5171e-06, -3.3805e-06, -2.6060e-06, ..., -1.4243e-06,\n 1.1542e-06, 6.8748e-06],\n [ 1.1686e-05, -3.3019e-06, 2.5983e-06, ..., 3.6526e-06,\n 7.8058e-07, 8.3269e-06],\n ...,\n [-2.2897e-06, 2.8611e-06, -4.3360e-06, ..., 1.1881e-06,\n -6.1881e-07, -1.4087e-05],\n [-4.1410e-06, 3.6013e-06, -1.3188e-06, ..., 1.8443e-06,\n -2.5329e-06, 9.2301e-06],\n [-8.6771e-06, -5.9636e-07, -1.5661e-07, ..., -2.6659e-06,\n 1.0153e-06, -9.6010e-07]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[9.7092e-11, 9.6133e-11, 9.9272e-11, ..., 1.0387e-10, 1.2096e-10,\n 1.2311e-10],\n [1.7807e-10, 1.9702e-10, 1.7574e-10, ..., 1.5991e-10, 1.8390e-10,\n 2.0540e-10],\n [1.8564e-10, 1.5692e-10, 2.1355e-10, ..., 1.7217e-10, 1.9091e-10,\n 2.4891e-10],\n ...,\n [2.0331e-10, 1.8378e-10, 1.9510e-10, ..., 1.9050e-10, 2.4024e-10,\n 2.9302e-10],\n [1.8140e-10, 1.8097e-10, 1.7673e-10, ..., 1.8567e-10, 1.8911e-10,\n 2.0720e-10],\n [2.5691e-10, 2.2681e-10, 1.9177e-10, ..., 2.1102e-10, 2.1868e-10,\n 2.6098e-10]], device='cuda:0')" | |
| }, | |
| "14": { | |
| "step": "tensor(1252.)", | |
| "exp_avg": "tensor([[-2.3241e-06, -2.4354e-06, -1.5035e-06, ..., -1.9910e-06,\n -1.3310e-06, -3.6656e-06],\n [ 4.0862e-06, 2.6900e-06, -2.7945e-06, ..., -2.5387e-07,\n 1.3357e-06, 7.2418e-06],\n [-3.7757e-06, 6.0759e-05, -7.7792e-07, ..., -9.0916e-07,\n -1.6565e-06, -6.2128e-07],\n ...,\n [-4.7018e-06, 6.2539e-05, -4.3217e-06, ..., -2.8579e-05,\n 2.2729e-07, -5.4295e-06],\n [ 7.2727e-06, -9.8645e-07, -3.8121e-06, ..., -5.1804e-07,\n -4.6194e-06, 9.7865e-07],\n [ 2.2263e-06, 9.2490e-06, 5.6031e-06, ..., -7.3038e-06,\n -2.2663e-06, -1.9168e-06]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[3.8565e-10, 1.0092e-10, 1.9969e-10, ..., 3.7874e-10, 4.7767e-10,\n 1.5746e-10],\n [5.3969e-10, 7.0202e-10, 1.3483e-10, ..., 5.7815e-10, 7.5775e-11,\n 3.0745e-10],\n [3.6335e-10, 2.1498e-09, 1.2835e-10, ..., 7.4070e-10, 9.3374e-10,\n 2.9598e-10],\n ...,\n [3.0553e-10, 1.0438e-09, 4.4592e-10, ..., 9.5753e-10, 3.0086e-10,\n 8.3615e-10],\n [3.4507e-09, 2.0365e-10, 3.5372e-10, ..., 1.4935e-10, 3.4193e-10,\n 1.6271e-10],\n [2.9787e-10, 8.7619e-11, 3.7083e-10, ..., 4.1479e-10, 3.2707e-10,\n 3.8803e-10]], device='cuda:0')" | |
| }, | |
| "15": { | |
| "step": "tensor(1252.)", | |
| "exp_avg": "tensor([ 0.0002, 0.0002, -0.0002, ..., 0.0002, 0.0009, 0.0002],\n device='cuda:0')", | |
| "exp_avg_sq": "tensor([5.2128e-06, 6.7659e-06, 5.7125e-06, ..., 5.9044e-06, 8.2738e-06,\n 5.9655e-06], device='cuda:0')" | |
| }, | |
| "16": { | |
| "step": "tensor(1252.)", | |
| "exp_avg": "tensor([[-2.2835e-08, 2.1704e-06, -2.4873e-06, ..., 6.9057e-07,\n 1.0074e-06, -1.5570e-06],\n [-5.9472e-06, 3.0554e-06, 4.6510e-06, ..., 7.0555e-07,\n 1.0359e-06, 3.4456e-06],\n [-1.5387e-06, -5.0648e-06, -2.0318e-06, ..., -5.2834e-06,\n 2.5951e-06, 8.4689e-07],\n ...,\n [ 3.6394e-06, -4.1955e-06, 3.7368e-06, ..., 2.0479e-06,\n 1.0550e-05, 3.2225e-07],\n [-2.0552e-06, 2.1075e-06, 2.4115e-06, ..., 3.8501e-06,\n 3.7201e-06, 5.3708e-06],\n [ 2.9456e-06, 1.9021e-06, 3.1334e-06, ..., 6.5462e-06,\n 1.2454e-06, 1.8990e-06]], device='cuda:0')", | |
| "exp_avg_sq": "tensor([[1.2831e-10, 1.3091e-10, 1.3910e-10, ..., 1.2642e-10, 1.5206e-10,\n 1.5052e-10],\n [1.6977e-10, 1.6662e-10, 1.3200e-10, ..., 2.4036e-10, 1.8997e-10,\n 1.7468e-10],\n [1.9562e-10, 1.8751e-10, 1.9258e-10, ..., 3.1713e-10, 2.3304e-10,\n 1.8993e-10],\n ...,\n [1.5170e-10, 2.1276e-10, 1.6629e-10, ..., 2.3036e-10, 3.0253e-10,\n 1.8065e-10],\n [2.1048e-10, 1.7563e-10, 2.0580e-10, ..., 3.5059e-10, 2.7610e-10,\n 2.5218e-10],\n [1.9779e-10, 1.9028e-10, 1.6880e-10, ..., 2.3218e-10, 2.6998e-10,\n 2.2546e-10]], device='cuda:0')" | |
| } | |
| }, | |
| "param_groups": [ | |
| { | |
| "lr": 0.0005005000000000001, | |
| "name": "shared", | |
| "betas": [ | |
| 0.9, | |
| 0.999 | |
| ], | |
| "eps": 1e-08, | |
| "weight_decay": 1e-05, | |
| "amsgrad": false, | |
| "maximize": false, | |
| "foreach": null, | |
| "capturable": false, | |
| "differentiable": false, | |
| "fused": null, | |
| "decoupled_weight_decay": true, | |
| "initial_lr": 0.001, | |
| "params": [ | |
| 0, | |
| 1 | |
| ] | |
| }, | |
| { | |
| "lr": 0.0005005000000000001, | |
| "name": "scale_256", | |
| "betas": [ | |
| 0.9, | |
| 0.999 | |
| ], | |
| "eps": 1e-08, | |
| "weight_decay": 1e-05, | |
| "amsgrad": false, | |
| "maximize": false, | |
| "foreach": null, | |
| "capturable": false, | |
| "differentiable": false, | |
| "fused": null, | |
| "decoupled_weight_decay": true, | |
| "initial_lr": 0.001, | |
| "params": [ | |
| 2, | |
| 3, | |
| 4 | |
| ] | |
| }, | |
| { | |
| "lr": 0.0005005000000000001, | |
| "name": "scale_512", | |
| "betas": [ | |
| 0.9, | |
| 0.999 | |
| ], | |
| "eps": 1e-08, | |
| "weight_decay": 1e-05, | |
| "amsgrad": false, | |
| "maximize": false, | |
| "foreach": null, | |
| "capturable": false, | |
| "differentiable": false, | |
| "fused": null, | |
| "decoupled_weight_decay": true, | |
| "initial_lr": 0.001, | |
| "params": [ | |
| 5, | |
| 6, | |
| 7 | |
| ] | |
| }, | |
| { | |
| "lr": 0.0005005000000000001, | |
| "name": "scale_768", | |
| "betas": [ | |
| 0.9, | |
| 0.999 | |
| ], | |
| "eps": 1e-08, | |
| "weight_decay": 1e-05, | |
| "amsgrad": false, | |
| "maximize": false, | |
| "foreach": null, | |
| "capturable": false, | |
| "differentiable": false, | |
| "fused": null, | |
| "decoupled_weight_decay": true, | |
| "initial_lr": 0.001, | |
| "params": [ | |
| 8, | |
| 9, | |
| 10 | |
| ] | |
| }, | |
| { | |
| "lr": 0.0005005000000000001, | |
| "name": "scale_1024", | |
| "betas": [ | |
| 0.9, | |
| 0.999 | |
| ], | |
| "eps": 1e-08, | |
| "weight_decay": 1e-05, | |
| "amsgrad": false, | |
| "maximize": false, | |
| "foreach": null, | |
| "capturable": false, | |
| "differentiable": false, | |
| "fused": null, | |
| "decoupled_weight_decay": true, | |
| "initial_lr": 0.001, | |
| "params": [ | |
| 11, | |
| 12, | |
| 13 | |
| ] | |
| }, | |
| { | |
| "lr": 0.0005005000000000001, | |
| "name": "scale_1280", | |
| "betas": [ | |
| 0.9, | |
| 0.999 | |
| ], | |
| "eps": 1e-08, | |
| "weight_decay": 1e-05, | |
| "amsgrad": false, | |
| "maximize": false, | |
| "foreach": null, | |
| "capturable": false, | |
| "differentiable": false, | |
| "fused": null, | |
| "decoupled_weight_decay": true, | |
| "initial_lr": 0.001, | |
| "params": [ | |
| 14, | |
| 15, | |
| 16 | |
| ] | |
| }, | |
| { | |
| "lr": 0.0005005000000000001, | |
| "name": "scale_1536", | |
| "betas": [ | |
| 0.9, | |
| 0.999 | |
| ], | |
| "eps": 1e-08, | |
| "weight_decay": 1e-05, | |
| "amsgrad": false, | |
| "maximize": false, | |
| "foreach": null, | |
| "capturable": false, | |
| "differentiable": false, | |
| "fused": null, | |
| "decoupled_weight_decay": true, | |
| "initial_lr": 0.001, | |
| "params": [ | |
| 17, | |
| 18, | |
| 19 | |
| ] | |
| }, | |
| { | |
| "lr": 0.0005005000000000001, | |
| "name": "scale_1792", | |
| "betas": [ | |
| 0.9, | |
| 0.999 | |
| ], | |
| "eps": 1e-08, | |
| "weight_decay": 1e-05, | |
| "amsgrad": false, | |
| "maximize": false, | |
| "foreach": null, | |
| "capturable": false, | |
| "differentiable": false, | |
| "fused": null, | |
| "decoupled_weight_decay": true, | |
| "initial_lr": 0.001, | |
| "params": [ | |
| 20, | |
| 21, | |
| 22 | |
| ] | |
| }, | |
| { | |
| "lr": 0.0005005000000000001, | |
| "name": "scale_2048", | |
| "betas": [ | |
| 0.9, | |
| 0.999 | |
| ], | |
| "eps": 1e-08, | |
| "weight_decay": 1e-05, | |
| "amsgrad": false, | |
| "maximize": false, | |
| "foreach": null, | |
| "capturable": false, | |
| "differentiable": false, | |
| "fused": null, | |
| "decoupled_weight_decay": true, | |
| "initial_lr": 0.001, | |
| "params": [ | |
| 23, | |
| 24, | |
| 25 | |
| ] | |
| }, | |
| { | |
| "lr": 0.0005005000000000001, | |
| "name": "scale_2304", | |
| "betas": [ | |
| 0.9, | |
| 0.999 | |
| ], | |
| "eps": 1e-08, | |
| "weight_decay": 1e-05, | |
| "amsgrad": false, | |
| "maximize": false, | |
| "foreach": null, | |
| "capturable": false, | |
| "differentiable": false, | |
| "fused": null, | |
| "decoupled_weight_decay": true, | |
| "initial_lr": 0.001, | |
| "params": [ | |
| 26, | |
| 27, | |
| 28 | |
| ] | |
| }, | |
| { | |
| "lr": 0.0005005000000000001, | |
| "name": "scale_2560", | |
| "betas": [ | |
| 0.9, | |
| 0.999 | |
| ], | |
| "eps": 1e-08, | |
| "weight_decay": 1e-05, | |
| "amsgrad": false, | |
| "maximize": false, | |
| "foreach": null, | |
| "capturable": false, | |
| "differentiable": false, | |
| "fused": null, | |
| "decoupled_weight_decay": true, | |
| "initial_lr": 0.001, | |
| "params": [ | |
| 29, | |
| 30, | |
| 31 | |
| ] | |
| }, | |
| { | |
| "lr": 0.0002505, | |
| "name": "fusion", | |
| "betas": [ | |
| 0.9, | |
| 0.999 | |
| ], | |
| "eps": 1e-08, | |
| "weight_decay": 1e-05, | |
| "amsgrad": false, | |
| "maximize": false, | |
| "foreach": null, | |
| "capturable": false, | |
| "differentiable": false, | |
| "fused": null, | |
| "decoupled_weight_decay": true, | |
| "initial_lr": 0.0005, | |
| "params": [ | |
| 32, | |
| 33, | |
| 34, | |
| 35, | |
| 36, | |
| 37, | |
| 38, | |
| 39, | |
| 40, | |
| 41, | |
| 42, | |
| 43, | |
| 44, | |
| 45, | |
| 46, | |
| 47, | |
| 48, | |
| 49, | |
| 50, | |
| 51, | |
| 52, | |
| 53, | |
| 54, | |
| 55, | |
| 56, | |
| 57, | |
| 58, | |
| 59, | |
| 60, | |
| 61 | |
| ] | |
| } | |
| ] | |
| }, | |
| "scheduler_state_dict": { | |
| "T_0": 10, | |
| "T_i": 10, | |
| "T_mult": 2, | |
| "eta_min": 1e-06, | |
| "T_cur": 5, | |
| "base_lrs": [ | |
| 0.001, | |
| 0.001, | |
| 0.001, | |
| 0.001, | |
| 0.001, | |
| 0.001, | |
| 0.001, | |
| 0.001, | |
| 0.001, | |
| 0.001, | |
| 0.001, | |
| 0.0005 | |
| ], | |
| "last_epoch": 5, | |
| "_step_count": 0, | |
| "_is_initial": false, | |
| "_get_lr_called_within_step": false, | |
| "_last_lr": [ | |
| 0.0005005000000000001, | |
| 0.0005005000000000001, | |
| 0.0005005000000000001, | |
| 0.0005005000000000001, | |
| 0.0005005000000000001, | |
| 0.0005005000000000001, | |
| 0.0005005000000000001, | |
| 0.0005005000000000001, | |
| 0.0005005000000000001, | |
| 0.0005005000000000001, | |
| 0.0005005000000000001, | |
| 0.0002505 | |
| ] | |
| }, | |
| "metrics": { | |
| "val_acc": 82.648 | |
| }, | |
| "train_config": { | |
| "name": "david_training", | |
| "run_id": "20251012_060013", | |
| "dataset_name": "AbstractPhil/imagenet-clip-features-orderly", | |
| "model_variant": "clip_vit_l14", | |
| "num_classes": 1000, | |
| "preset": "clip_vit_l14_deep", | |
| "custom_config_path": null, | |
| "num_classes_override": null, | |
| "use_belly_override": null, | |
| "belly_expand_override": null, | |
| "progressive_training_override": true, | |
| "scale_warmup_epochs_override": { | |
| "256": 0, | |
| "512": 1, | |
| "768": 2, | |
| "1024": 3, | |
| "1280": 4, | |
| "1536": 5, | |
| "1792": 6, | |
| "2048": 7, | |
| "2304": 8, | |
| "2560": 9 | |
| }, | |
| "num_epochs": 10, | |
| "batch_size": 1024, | |
| "learning_rate": 0.001, | |
| "weight_decay": 1e-05, | |
| "warmup_epochs": 0, | |
| "use_rose_loss": true, | |
| "rose_initial_weight": 0.1, | |
| "rose_max_weight": 0.5, | |
| "rose_weight_schedule": "adaptive", | |
| "use_cayley_loss": false, | |
| "cayley_weight": 0.001, | |
| "scale_loss_balance": null, | |
| "use_mixed_precision": false, | |
| "gradient_clip": 5.0, | |
| "scheduler_type": "cosine_restarts", | |
| "min_lr": 1e-06, | |
| "freeze_strategy": "never", | |
| "freeze_threshold": 90.0, | |
| "unfreeze_on_plateau": true, | |
| "patience": 10, | |
| "track_gradients": true, | |
| "gradient_scale_threshold": 1e-07, | |
| "gradient_scale_multiplier": 5.0, | |
| "log_interval": 50, | |
| "val_interval": 1, | |
| "save_interval": 5, | |
| "log_fusion_weights": true, | |
| "log_loss_components": true, | |
| "save_format": "safetensors", | |
| "hf_repo": "AbstractPhil/gated-david", | |
| "upload_to_hub": true, | |
| "base_dir": "./david_training", | |
| "num_workers": 10, | |
| "pin_memory": true, | |
| "prefetch_factor": 4, | |
| "persistent_workers": true | |
| } | |
| } |