Hugging Face's logo

AbstractPhil
/
david-shared-space

Image Classification
TensorBoard
English
imagenet
multi-scale
feature-geometry
david
Eval Results (legacy)
Model card Files
xet
 Metrics Community
david-shared-space / weights /David-decoupled-deep_efficiency /20251013_004438 /best_model_acc62.92_metadata.json
AbstractPhil's picture
AbstractPhil
Update best_model_acc62.92_metadata.json - Run 20251013_004438
bdc6cf6 verified
raw
history blame contribute delete
281 kB
{
 "epoch": 8,
 "optimizer_state_dict": {
 "state": {
 "0": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[-4.3774e-05, 9.6487e-05, -6.9877e-05, ..., -2.1661e-06,\n -6.2980e-05, -2.5716e-05],\n [-4.1603e-05, 4.8536e-05, -1.3514e-05, ..., -5.5728e-05,\n 1.1967e-05, -6.0179e-05],\n [-1.6100e-05, 3.8464e-05, 4.1490e-05, ..., -1.4689e-05,\n -4.5491e-06, -1.3797e-05],\n ...,\n [ 6.1222e-06, -1.0605e-05, 4.1516e-05, ..., 3.6144e-05,\n -2.7276e-06, -1.3136e-05],\n [-2.5508e-05, 1.8746e-04, 2.5521e-05, ..., -3.0096e-05,\n -1.8113e-05, -2.8629e-05],\n [ 2.6146e-05, 5.5631e-05, -8.0181e-05, ..., 1.8196e-05,\n -1.5131e-05, 1.9069e-05]], device='cuda:0')",
 "exp_avg_sq": "tensor([[1.5319e-07, 7.4411e-08, 4.3594e-08, ..., 5.2941e-08, 2.0757e-08,\n 2.7581e-08],\n [4.8658e-08, 8.7550e-08, 3.7333e-08, ..., 3.2001e-08, 1.2876e-08,\n 1.1442e-08],\n [3.1764e-08, 4.8351e-08, 2.2251e-08, ..., 8.6884e-09, 1.0930e-08,\n 8.8492e-09],\n ...,\n [3.3829e-08, 6.5826e-08, 2.1854e-08, ..., 1.1197e-08, 1.1559e-08,\n 1.1684e-08],\n [7.1738e-09, 4.6351e-08, 1.3503e-08, ..., 2.0445e-08, 5.6335e-09,\n 5.6490e-09],\n [1.5818e-08, 6.8950e-08, 3.0481e-08, ..., 3.3768e-08, 7.5503e-09,\n 1.1657e-08]], device='cuda:0')"
 },
 "1": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([-8.3494e-05, -9.9533e-04, -9.1317e-04, 3.2595e-04, 1.0327e-03,\n 7.1882e-04, 2.5600e-04, 2.0740e-05, -9.4271e-04, -1.6170e-03,\n -6.6830e-04, 4.0609e-04, -1.4920e-03, -5.2198e-04, -3.7610e-04,\n -1.1542e-03, 6.2631e-04, 9.8579e-04, 3.3468e-05, 9.6239e-04,\n -2.8524e-04, 1.2580e-03, 9.3349e-05, 2.1420e-04, 1.3951e-03,\n -7.7622e-04, 2.2000e-03, -3.2611e-04, -3.9057e-04, -8.7451e-05,\n -7.7386e-04, -1.9451e-03, 1.1765e-03, 2.0929e-05, -9.3680e-04,\n 2.4014e-04, 1.7412e-03, -8.3146e-04, -2.8940e-04, -1.8357e-03,\n -4.1561e-04, -2.4324e-04, 4.4015e-04, -1.4750e-03, -3.1477e-05,\n 3.8643e-05, -1.5054e-04, 7.3425e-04, -5.1117e-04, 6.8944e-04,\n -2.9122e-04, 8.0368e-04, 1.3314e-03, 1.3164e-03, -7.6602e-04,\n -7.4620e-04, 2.5289e-03, -6.2597e-05, -4.9384e-04, 6.8123e-04,\n -2.8774e-04, 6.3017e-04, 2.0659e-03, 4.5876e-04, -3.4174e-04,\n 2.3727e-03, -1.9204e-04, -1.0007e-03, 1.5987e-03, 9.5903e-04,\n -5.4579e-04, 3.1112e-04, 1.1018e-03, -1.3766e-03, -3.9275e-04,\n 4.2606e-05, -3.5768e-04, -9.4276e-04, 1.9816e-03, -4.2017e-04,\n 8.5700e-04, -5.1357e-04, 1.1290e-03, 2.4239e-03, 8.5671e-04,\n 1.1234e-04, -1.7621e-03, -5.3611e-04, 5.9463e-04, 2.3380e-04,\n -9.4281e-04, -9.9624e-04, 1.2809e-03, 9.6041e-04, -1.3686e-04,\n 2.0382e-04, -2.6517e-05, 4.4997e-04, -9.9882e-04, -9.3420e-04,\n -1.0565e-03, -1.3987e-03, -1.7220e-04, -1.6902e-03, 2.6943e-04,\n -4.6568e-04, 2.0446e-04, -4.0129e-05, 1.1577e-03, 4.5825e-04,\n -1.0112e-03, 1.8693e-03, 7.8788e-05, -1.1333e-03, 8.3701e-04,\n -1.0756e-03, 5.1301e-04, 4.6246e-04, -2.2672e-03, -3.4546e-03,\n 4.2457e-04, 9.8800e-04, 1.1882e-04, 7.2022e-04, 1.2526e-03,\n 9.0783e-04, -4.0299e-04, 1.0716e-03, 7.9160e-04, 7.6262e-04,\n 1.0487e-03, 2.9364e-04, 2.0547e-03, -3.8239e-04, 2.6059e-04,\n 9.8191e-04, -5.4663e-04, 1.0542e-04, -5.4379e-04, 1.6263e-03,\n 1.0816e-03, 4.3044e-05, 3.2337e-04, -6.6391e-05, -2.4723e-04,\n -3.9260e-04, -1.6075e-03, 3.9014e-04, 3.8160e-04, 4.3825e-04,\n 3.4103e-05, 6.9957e-05, -1.3398e-03, 9.1200e-04, 1.1188e-04,\n 2.1104e-04, -1.7768e-04, 1.1514e-03, -1.3551e-03, -1.7948e-03,\n -8.3785e-04, -9.1381e-04, -1.7352e-03, -6.1924e-04, -6.1661e-04,\n 7.6455e-04, 3.7628e-04, 3.2863e-04, -3.4044e-04, -1.1523e-03,\n -8.0465e-04, -5.2200e-04, 7.6657e-04, -2.8450e-05, 1.2619e-03,\n 5.5169e-04, 2.7957e-04, 6.8539e-04, -4.0417e-04, -2.4355e-04,\n -1.2135e-04, 6.7626e-04, 2.0304e-03, 1.6645e-04, 8.9022e-04,\n 1.2542e-04, -9.4288e-04, -2.1960e-04, -2.2184e-05, 2.3509e-04,\n -2.9011e-04, 3.8892e-04, -1.7689e-03, -9.0750e-04, 6.8000e-04,\n 3.7077e-04, -1.8480e-03, -1.2027e-03, 1.5801e-03, 8.9027e-04,\n 8.5956e-04, -8.0349e-05, 2.6018e-04, 3.6245e-04, 1.1733e-03,\n -3.7862e-04, 2.4241e-04, 1.1166e-03, -9.8069e-04, 6.8553e-04,\n 7.5863e-04, 1.0040e-03, -5.5819e-04, 1.0237e-03, 1.0339e-03,\n -9.0198e-04, 1.4921e-03, 1.8067e-03, -1.2147e-03, -8.1968e-04,\n -1.3152e-03, -2.2961e-03, -8.2194e-04, 2.9005e-04, -7.8560e-04,\n -5.7364e-05, -4.4187e-04, -4.5191e-04, 1.8254e-03, 1.2739e-04,\n 5.7469e-04, -4.8187e-04, 7.3317e-04, 5.8247e-04, -1.1572e-03,\n 1.4102e-03, 9.3719e-04, 2.3406e-05, 8.7637e-05, -1.6893e-03,\n 3.2474e-04, 5.9543e-04, 1.2513e-03, 2.1288e-03, 7.2107e-04,\n 5.4214e-04, 1.3099e-04, -9.2627e-04, 1.2961e-03, -2.0895e-03,\n 1.1698e-03, 6.3803e-04, 3.0437e-04, 2.1524e-03, -3.7316e-04,\n -8.1047e-04, 7.0224e-04, 8.2084e-04, -5.6673e-04, 1.0742e-03,\n 4.7384e-04, 1.4559e-04, 1.0487e-03, 4.4966e-04, 8.3830e-04,\n -5.7608e-04, -1.2281e-05, 1.4984e-03, 2.4469e-05, 1.0687e-03,\n 7.8321e-04, 1.4867e-03, -5.6820e-04, -7.3707e-04, -2.4074e-04,\n 1.4381e-04, 3.7705e-04, 8.2807e-04, -3.9000e-04, 1.6014e-03,\n 3.2390e-04, 3.2658e-04, 2.3380e-04, 6.6499e-04, -7.7247e-04,\n -1.3546e-03, 8.0459e-04, 3.2391e-04, 1.2300e-03, 5.4947e-05,\n -8.4873e-04, 7.4515e-04, 5.9636e-04, 5.0984e-04, -2.8469e-04,\n 4.9929e-04, -9.5671e-05, -9.9824e-05, 9.0178e-04, 3.8129e-05,\n -2.0118e-04, 6.0597e-05, 8.9774e-04, -1.1236e-03, -3.7691e-04,\n 3.1611e-04, 2.8782e-04, -6.4253e-04, 1.7664e-04, -7.2984e-04,\n 3.6352e-04, -6.9506e-04, 3.9488e-04, 7.0468e-04, 6.6516e-04,\n 6.9145e-04, -3.5652e-04, 8.4383e-04, -2.0423e-03, 9.0362e-04,\n 1.2322e-03, 1.4750e-04, 1.1455e-03, -6.3432e-04, 1.3324e-04,\n -3.8180e-04, 9.6468e-04, -8.0792e-05, -1.6481e-03, -7.2209e-04,\n 6.6461e-04, 5.1686e-04, 1.0802e-03, -3.8502e-03, -7.0078e-04,\n -2.0153e-04, 8.7592e-05, 3.3038e-04, -2.4322e-05, 5.2976e-04,\n -1.2465e-04, -1.9847e-03, -1.2253e-03, 2.0669e-03, 7.2627e-04,\n 8.3548e-04, 8.2080e-04, -7.0197e-04, 1.7440e-03, 4.3895e-04,\n -8.5514e-04, -2.8969e-03, -5.7657e-04, 6.5880e-04, -1.1310e-03,\n -2.0118e-03, 1.1429e-03, 1.4224e-03, -5.3339e-04, -9.8954e-04,\n 7.3284e-04, 3.1458e-04, -2.9991e-04, 9.2073e-04, -3.2933e-04,\n -7.5622e-05, -2.7749e-04, -7.0323e-04, 7.0765e-04, -4.9799e-04,\n -1.9902e-04, 3.6689e-04, -1.0157e-04, 7.9116e-04, -8.3229e-04,\n 1.5185e-04, -7.3784e-04, 3.4313e-04, 1.4078e-03, 8.8318e-04,\n 3.5959e-04, -1.2706e-03, -3.8135e-04, -1.7535e-04, -2.8152e-03,\n 2.4842e-04, 1.3821e-05, 4.2798e-04, -4.7474e-04, 9.9543e-04,\n 8.7410e-05, 1.1581e-04, 8.6874e-04, -4.7223e-04, -3.1425e-04,\n -5.8688e-04, 7.5871e-04, -1.1041e-03, 1.9387e-03, 1.0476e-04,\n -7.8023e-04, 8.0167e-04, -2.3299e-03, -9.5146e-04, -1.9047e-03,\n 1.2557e-04, 2.6365e-04, -7.6427e-04, 3.5738e-06, -1.3490e-03,\n 1.1707e-03, -1.4159e-05, -2.0083e-03, -5.8920e-04, -1.0940e-03,\n -7.1392e-04, 8.4966e-04, -7.2166e-04, -2.8395e-03, 1.2365e-03,\n 1.1335e-03, -1.2810e-03, -1.9765e-04, 6.3797e-04, 4.6675e-04,\n 5.3599e-04, -2.3345e-05, -2.3578e-04, -6.4262e-04, 8.1871e-04,\n 1.0625e-04, 8.9458e-05, 1.1325e-03, -1.1012e-03, 6.3657e-04,\n 1.1334e-04, 1.7789e-03, 1.3534e-03, 1.8378e-03, -1.8748e-03,\n -3.9243e-04, 4.9241e-04, -8.0517e-04, -1.3599e-03, 6.0894e-04,\n -8.1656e-04, -4.7685e-04, -3.7336e-04, 9.4009e-04, -9.2354e-04,\n -1.2128e-03, -2.6022e-05, 1.6671e-04, -3.1466e-04, 5.6113e-04,\n -4.2461e-04, -3.5025e-04, -3.9518e-04, -1.8420e-03, -8.1858e-04,\n -1.4160e-03, 6.1675e-04, 1.2212e-03, 2.1642e-03, -1.1609e-04,\n -5.2300e-04, -1.1450e-03, 1.2795e-03, 6.6566e-04, -3.5047e-04,\n -1.3001e-03, -7.0288e-04, 4.9406e-04, -5.4730e-04, 2.3518e-04,\n 1.0220e-04, -1.3724e-04, -3.4067e-03, 4.9827e-04, -9.9996e-05,\n 5.2542e-04, -2.7068e-04, -1.0690e-04, -3.4053e-04, -3.8496e-04,\n 2.4462e-04, 1.0336e-03, -1.7829e-03, 2.2178e-04, 9.7267e-04,\n 9.2276e-04, -5.0888e-04, -1.6702e-04, -1.3626e-03, -1.2482e-03,\n 9.5448e-04, -5.5065e-04, -4.2451e-04, -3.2383e-04, 1.1295e-03,\n -7.9954e-04, 7.6412e-04, -5.0951e-04, -6.3452e-04, -4.2879e-04,\n 1.9584e-04, 2.0151e-03, 2.0833e-03, 4.5977e-04, -1.0090e-04,\n -9.5655e-05, 1.0173e-03], device='cuda:0')",
 "exp_avg_sq": "tensor([3.5707e-05, 2.0437e-05, 1.0692e-05, 1.1588e-05, 1.5807e-05, 1.8226e-05,\n 1.2277e-05, 1.8867e-05, 1.1437e-05, 1.1031e-05, 1.8055e-05, 1.7829e-05,\n 9.7818e-06, 1.4193e-05, 1.0093e-05, 1.5074e-05, 9.2841e-06, 1.8846e-05,\n 6.8814e-06, 5.8608e-06, 8.7054e-06, 1.6264e-05, 9.0769e-06, 1.6108e-05,\n 1.6070e-05, 7.6395e-06, 2.6940e-05, 2.3167e-05, 1.1065e-05, 6.4886e-06,\n 1.3994e-05, 1.6891e-05, 1.5169e-05, 1.6696e-05, 2.7191e-05, 1.2949e-05,\n 1.8418e-05, 1.0941e-05, 2.8104e-05, 1.0043e-05, 2.4027e-05, 1.1942e-05,\n 1.1719e-05, 2.1454e-05, 1.3978e-05, 1.8776e-05, 1.1891e-05, 1.5111e-05,\n 2.0770e-05, 9.4472e-06, 8.8013e-06, 2.0757e-05, 1.6817e-05, 1.4046e-05,\n 7.3076e-06, 1.1579e-05, 1.9888e-05, 1.3221e-05, 4.1354e-05, 1.7324e-05,\n 1.7882e-05, 1.9455e-05, 2.7900e-05, 1.4849e-05, 9.0659e-06, 2.8520e-05,\n 1.4346e-05, 1.0334e-05, 1.4508e-05, 1.6230e-05, 1.3834e-05, 1.7275e-05,\n 1.0575e-05, 1.6116e-05, 8.8282e-06, 5.0692e-05, 1.7627e-05, 1.4375e-05,\n 1.7021e-05, 1.4683e-05, 1.2808e-05, 1.7179e-05, 1.8916e-05, 2.7023e-05,\n 1.3733e-05, 1.5113e-05, 1.8727e-05, 1.3027e-05, 1.0734e-05, 1.4093e-05,\n 1.1741e-05, 1.7261e-05, 1.3808e-05, 1.5518e-05, 1.3100e-05, 1.6424e-05,\n 9.3370e-06, 1.4177e-05, 8.0655e-06, 1.2533e-05, 2.6102e-05, 1.1595e-05,\n 3.4090e-06, 2.1996e-05, 2.1214e-05, 1.3336e-05, 3.0870e-05, 2.7225e-05,\n 1.0794e-05, 1.8988e-05, 1.1828e-05, 3.2459e-05, 1.9293e-05, 1.6634e-05,\n 1.2800e-05, 2.7701e-05, 1.6074e-05, 1.1321e-05, 2.1721e-05, 1.9223e-05,\n 7.3905e-06, 1.1902e-05, 3.0855e-06, 1.0772e-05, 1.8677e-05, 1.7408e-05,\n 1.0969e-05, 1.7079e-05, 1.8520e-05, 1.4587e-05, 1.3056e-05, 9.3212e-06,\n 1.0712e-05, 1.7605e-05, 1.2614e-05, 1.4920e-05, 1.2600e-05, 1.6054e-05,\n 1.1793e-05, 1.3669e-05, 1.2592e-05, 1.2705e-05, 1.3998e-05, 1.2102e-05,\n 1.9321e-05, 1.3169e-05, 1.8477e-05, 1.2563e-05, 2.5033e-05, 2.7382e-05,\n 1.2818e-05, 2.0960e-05, 3.0662e-05, 1.8391e-05, 1.6231e-05, 1.4540e-05,\n 3.5201e-05, 1.2933e-05, 2.2222e-05, 1.3932e-05, 1.2476e-05, 1.1312e-05,\n 2.8964e-05, 9.5763e-06, 1.3265e-05, 1.1284e-05, 2.1245e-05, 1.3659e-05,\n 1.5872e-05, 1.6625e-05, 1.9560e-05, 1.4686e-05, 1.2072e-05, 2.8102e-05,\n 2.0227e-05, 1.2974e-05, 1.1738e-05, 5.9517e-06, 1.0329e-05, 7.3722e-06,\n 1.5488e-05, 2.1486e-05, 1.9487e-05, 1.9636e-05, 2.3163e-05, 1.5494e-05,\n 2.6075e-05, 2.2374e-05, 8.0835e-06, 1.6508e-05, 9.4391e-06, 1.2280e-05,\n 1.2276e-05, 2.0394e-05, 1.7989e-05, 9.6844e-06, 9.9975e-06, 1.2336e-05,\n 1.9504e-05, 1.0941e-05, 1.5918e-05, 1.2683e-05, 1.1580e-05, 1.3688e-05,\n 1.5430e-05, 1.0444e-05, 9.0301e-06, 1.3223e-05, 2.6260e-05, 2.5173e-05,\n 1.9508e-05, 1.7360e-05, 2.6622e-05, 1.1618e-05, 2.2678e-05, 1.2839e-05,\n 1.0669e-05, 1.2996e-05, 1.4087e-05, 1.6485e-05, 2.1162e-05, 1.2692e-05,\n 1.6711e-05, 7.0241e-06, 2.2645e-05, 1.1873e-05, 1.4802e-05, 1.2098e-05,\n 1.3144e-05, 1.4835e-05, 1.9440e-05, 9.1273e-06, 1.4460e-05, 1.1722e-05,\n 1.3514e-05, 1.3654e-05, 8.2329e-06, 3.5608e-05, 1.3408e-05, 1.4811e-05,\n 1.1728e-05, 9.2874e-06, 1.2697e-05, 1.9620e-05, 5.1436e-06, 1.3067e-05,\n 1.1980e-05, 1.3937e-05, 2.3896e-05, 3.0803e-05, 2.2146e-05, 2.6906e-05,\n 7.0492e-06, 2.5903e-05, 1.6705e-05, 1.4545e-05, 1.6154e-05, 1.8418e-05,\n 7.8523e-06, 1.9490e-05, 1.2440e-05, 7.1390e-06, 1.3208e-05, 1.3716e-05,\n 3.2987e-05, 8.6863e-06, 1.9247e-05, 1.7188e-05, 9.6676e-06, 2.6917e-05,\n 8.9078e-06, 1.0829e-05, 1.4428e-05, 9.3380e-06, 8.1411e-06, 1.4742e-05,\n 8.2518e-06, 8.3512e-06, 1.8717e-05, 1.2754e-05, 7.1495e-06, 1.4191e-05,\n 1.3976e-05, 9.2250e-06, 1.1157e-05, 2.3371e-05, 6.9455e-06, 1.0135e-05,\n 1.1235e-05, 1.6555e-05, 1.0967e-05, 3.8234e-06, 9.8872e-06, 7.0155e-06,\n 1.1904e-05, 1.7541e-05, 1.4785e-05, 1.2403e-05, 2.0649e-05, 6.8113e-06,\n 2.0455e-05, 1.5924e-05, 2.9963e-05, 1.8011e-05, 2.2085e-05, 8.1127e-06,\n 3.3179e-05, 1.4077e-05, 1.2332e-05, 1.1314e-05, 9.1458e-06, 7.1714e-06,\n 1.9675e-05, 1.2101e-05, 2.3461e-05, 6.4144e-06, 1.1004e-05, 1.1621e-05,\n 2.0602e-05, 1.2924e-05, 1.5946e-05, 1.6811e-05, 1.4762e-05, 5.8586e-06,\n 1.0313e-05, 1.9138e-05, 1.1268e-05, 2.3011e-05, 7.3878e-06, 7.6416e-06,\n 1.1674e-05, 1.3170e-05, 1.4489e-05, 1.6125e-05, 1.2205e-05, 1.1415e-05,\n 1.9236e-05, 1.4415e-05, 3.2064e-05, 1.5585e-05, 1.4601e-05, 1.1208e-05,\n 2.0690e-05, 1.8084e-05, 1.1229e-05, 2.1510e-05, 1.0595e-05, 1.1169e-05,\n 1.7932e-05, 1.1793e-05, 1.5026e-05, 1.1694e-05, 1.6081e-05, 1.2941e-05,\n 1.0729e-05, 1.3560e-05, 3.9527e-05, 4.3386e-05, 8.3832e-06, 1.8100e-05,\n 2.1850e-05, 3.0680e-05, 1.7656e-05, 1.2229e-05, 2.1901e-05, 1.8098e-05,\n 1.5687e-05, 1.3059e-05, 6.4899e-06, 1.6058e-05, 8.0367e-06, 2.1577e-05,\n 1.0336e-05, 1.5517e-05, 1.8417e-05, 2.2028e-05, 5.7532e-06, 1.8901e-05,\n 1.3490e-05, 2.4930e-05, 1.4147e-05, 4.1056e-06, 1.0527e-05, 9.9501e-06,\n 4.5505e-05, 1.4881e-05, 9.6059e-06, 1.4916e-05, 1.2352e-05, 2.2713e-05,\n 1.9745e-05, 1.9171e-05, 1.5730e-05, 3.8003e-06, 1.2987e-05, 1.8712e-05,\n 2.1072e-05, 8.6121e-06, 2.2009e-05, 1.3683e-05, 1.6548e-05, 1.2448e-05,\n 2.7648e-05, 6.4604e-05, 2.0754e-05, 1.1232e-05, 6.2351e-06, 1.0861e-05,\n 5.1489e-06, 1.6830e-05, 2.0111e-05, 7.3856e-06, 1.9427e-05, 1.0567e-05,\n 1.4767e-05, 9.5168e-06, 7.4427e-06, 2.5674e-05, 3.7940e-05, 1.2963e-05,\n 1.0160e-05, 1.3216e-05, 1.4277e-05, 1.0785e-05, 1.2977e-05, 1.4433e-05,\n 8.4066e-06, 8.6275e-06, 1.0140e-05, 2.7509e-05, 1.4079e-05, 6.3240e-06,\n 1.8197e-05, 1.5383e-05, 1.0711e-05, 6.1583e-06, 1.6406e-05, 1.6478e-05,\n 1.9568e-05, 2.0753e-05, 1.3420e-05, 1.5598e-05, 9.4426e-06, 2.2911e-05,\n 7.7494e-06, 1.2652e-05, 1.6645e-05, 1.3993e-05, 1.5204e-05, 1.4336e-05,\n 1.3353e-05, 1.1057e-05, 1.0719e-05, 1.8013e-05, 7.6749e-06, 1.3929e-05,\n 1.0485e-05, 1.1150e-05, 1.5661e-05, 2.0676e-05, 2.3882e-05, 2.2186e-05,\n 1.6747e-05, 1.9542e-05, 6.9030e-06, 1.4501e-05, 1.2957e-05, 2.6930e-05,\n 1.9485e-05, 1.4721e-05, 1.5571e-05, 1.4024e-05, 2.7008e-05, 1.6178e-05,\n 1.0318e-05, 1.5087e-05, 2.0434e-05, 2.4942e-05, 1.2662e-05, 2.6954e-05,\n 1.7918e-05, 9.5767e-06, 1.2094e-05, 1.3672e-05, 1.5163e-05, 6.0257e-06,\n 1.0245e-05, 1.5073e-05, 1.5341e-05, 1.0499e-05, 3.1345e-05, 1.1665e-05,\n 7.7491e-06, 4.1505e-05, 1.4336e-05, 1.8099e-05, 1.5018e-05, 2.0358e-05,\n 1.1535e-05, 1.7492e-05, 1.0955e-05, 1.8588e-05, 1.4277e-05, 5.6200e-06,\n 1.3155e-05, 1.8358e-05, 2.9375e-05, 1.5190e-05, 1.0559e-05, 1.4244e-05,\n 8.1619e-06, 1.4023e-05], device='cuda:0')"
 },
 "2": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[-6.7545e-05, -1.2859e-05, -3.5510e-05, ..., -2.9536e-05,\n -1.0023e-05, -7.0491e-06],\n [-5.3563e-05, -1.9348e-05, 1.7793e-05, ..., -2.4749e-05,\n -6.8091e-06, -1.3526e-05],\n [-7.4350e-06, 1.4062e-05, -1.6148e-05, ..., 2.2035e-05,\n -1.1343e-05, -1.0824e-05],\n ...,\n [-7.7361e-07, -4.4765e-06, 4.6208e-05, ..., 1.0741e-05,\n -1.2373e-05, 7.9671e-06],\n [ 4.0612e-05, 1.1028e-05, -1.0965e-05, ..., -2.1975e-06,\n -2.3065e-06, -1.1625e-05],\n [-5.1235e-06, 2.5969e-05, 2.0479e-06, ..., -9.3428e-06,\n -7.0716e-06, 2.9497e-05]], device='cuda:0')",
 "exp_avg_sq": "tensor([[1.0557e-08, 2.7848e-09, 1.8594e-08, ..., 5.3875e-09, 5.5037e-09,\n 1.9164e-09],\n [1.6377e-08, 8.2395e-09, 6.8896e-09, ..., 1.0435e-08, 2.5514e-09,\n 7.3957e-09],\n [9.6567e-09, 2.0286e-08, 7.7015e-09, ..., 1.2082e-08, 2.9783e-09,\n 4.5733e-09],\n ...,\n [1.4849e-08, 8.2972e-09, 1.6483e-08, ..., 1.1595e-08, 5.0413e-09,\n 5.6713e-09],\n [1.7036e-08, 6.2054e-09, 6.3456e-09, ..., 2.1045e-08, 5.7870e-09,\n 4.7085e-09],\n [1.0395e-08, 6.4723e-09, 7.4996e-09, ..., 1.3613e-08, 4.5603e-09,\n 4.4207e-09]], device='cuda:0')"
 },
 "3": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[ 3.1020e-07, 7.8000e-07, -2.6976e-06, ..., -2.9908e-07,\n -1.5701e-06, -2.5762e-08],\n [-7.6638e-06, 7.5164e-06, -8.0157e-06, ..., 8.8531e-06,\n 2.5179e-06, 1.8346e-06],\n [ 8.6247e-07, 1.1337e-07, -4.4529e-07, ..., -2.0820e-06,\n -6.9590e-08, 9.0622e-07],\n ...,\n [ 1.2137e-06, -3.9453e-06, 2.0717e-06, ..., -3.6748e-06,\n -8.9125e-07, 2.4297e-06],\n [ 7.8327e-06, -6.0550e-06, -3.3032e-07, ..., -5.5061e-06,\n -1.1919e-06, 4.6989e-07],\n [ 6.5518e-07, 5.7767e-06, -2.0281e-06, ..., 3.3923e-06,\n -3.5973e-06, -3.3463e-06]], device='cuda:0')",
 "exp_avg_sq": "tensor([[1.8500e-10, 3.3083e-10, 1.6470e-10, ..., 6.2825e-11, 3.4719e-11,\n 5.6058e-11],\n [7.8853e-11, 3.8578e-10, 1.2351e-10, ..., 1.6888e-10, 4.1092e-11,\n 3.7922e-11],\n [3.8480e-11, 3.5541e-10, 9.4802e-11, ..., 2.5954e-11, 2.2075e-11,\n 2.6904e-11],\n ...,\n [8.9412e-11, 1.6949e-10, 2.0212e-10, ..., 1.0479e-10, 3.1882e-11,\n 5.1780e-11],\n [1.2330e-10, 2.8248e-10, 1.9995e-10, ..., 2.8406e-10, 5.7051e-11,\n 4.8430e-11],\n [7.7142e-11, 4.3432e-10, 1.0074e-10, ..., 5.2801e-11, 3.7201e-11,\n 4.4179e-11]], device='cuda:0')"
 },
 "4": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.5457e-05, -2.4987e-05, 3.2612e-06, ..., -8.1727e-05,\n -6.1699e-05, -2.4756e-05], device='cuda:0')",
 "exp_avg_sq": "tensor([6.8009e-08, 7.7258e-08, 2.9706e-08, ..., 6.2675e-08, 1.0706e-07,\n 6.0939e-08], device='cuda:0')"
 },
 "5": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[ 1.1995e-07, -1.4689e-07, 6.6974e-07, ..., 3.7067e-07,\n -7.4272e-08, 3.8360e-07],\n [ 4.5492e-07, -1.0890e-06, -9.5364e-07, ..., 3.1831e-07,\n 1.5086e-06, 7.9344e-07],\n [ 2.7676e-07, -6.9038e-07, 1.5712e-06, ..., -2.9120e-07,\n -3.8769e-08, -9.0052e-07],\n ...,\n [ 3.9921e-07, -2.0526e-07, -7.6037e-07, ..., -1.2937e-06,\n 5.0962e-07, 3.4841e-07],\n [ 7.8737e-07, 6.5238e-08, 7.4157e-07, ..., 4.9383e-09,\n 1.1411e-06, -5.1627e-07],\n [ 4.6305e-08, -6.0416e-08, 1.1775e-06, ..., -1.9029e-07,\n -1.4146e-06, -1.3278e-06]], device='cuda:0')",
 "exp_avg_sq": "tensor([[2.6413e-12, 6.4822e-12, 9.6194e-12, ..., 3.7884e-12, 1.3689e-11,\n 1.6987e-12],\n [5.4140e-12, 4.3020e-12, 1.7789e-11, ..., 8.2264e-12, 3.2056e-11,\n 1.1084e-11],\n [2.6175e-12, 3.3466e-12, 1.5662e-11, ..., 7.4916e-12, 1.8920e-11,\n 4.5280e-12],\n ...,\n [7.7274e-12, 6.1878e-12, 2.1955e-11, ..., 1.3119e-11, 2.3086e-11,\n 3.8655e-12],\n [8.0062e-12, 1.1154e-11, 2.2272e-11, ..., 8.8317e-12, 2.8991e-11,\n 3.0432e-12],\n [6.0868e-12, 2.5726e-12, 2.4960e-11, ..., 7.5013e-12, 3.0865e-11,\n 9.6121e-12]], device='cuda:0')"
 },
 "6": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[-6.8097e-07, -1.2282e-06, 8.1915e-07, ..., -2.7865e-06,\n 4.7366e-07, -1.0252e-06],\n [ 5.4704e-08, 1.7155e-06, -2.6226e-06, ..., 7.7504e-07,\n 8.7008e-07, 3.2919e-07],\n [-1.9545e-06, 6.6635e-07, 1.2140e-07, ..., 3.4887e-07,\n 2.0571e-06, 1.9322e-06],\n ...,\n [ 4.6455e-07, 8.4287e-07, -1.0551e-06, ..., 4.1031e-07,\n 7.6085e-07, 3.2687e-07],\n [-1.4083e-07, -2.6414e-06, -1.6953e-06, ..., -1.2742e-06,\n 1.1095e-06, 6.1754e-07],\n [ 1.7109e-07, 6.3923e-08, 2.0554e-06, ..., 1.9703e-06,\n -1.9318e-07, 3.2643e-07]], device='cuda:0')",
 "exp_avg_sq": "tensor([[4.6030e-11, 2.2151e-10, 8.2474e-11, ..., 1.8587e-10, 2.5019e-11,\n 3.6389e-11],\n [5.5889e-11, 1.8783e-10, 7.4555e-11, ..., 7.8905e-11, 2.8755e-11,\n 4.9893e-11],\n [1.4052e-10, 3.2792e-10, 1.1834e-10, ..., 7.3369e-11, 1.1461e-10,\n 6.9942e-11],\n ...,\n [1.3693e-10, 1.1601e-10, 7.2944e-11, ..., 2.5508e-11, 2.2533e-11,\n 1.0367e-11],\n [3.4471e-11, 2.1563e-10, 5.1076e-11, ..., 1.9478e-11, 1.3386e-11,\n 1.4302e-11],\n [4.5469e-11, 1.8404e-10, 5.3087e-11, ..., 2.4244e-11, 2.3261e-11,\n 2.0877e-11]], device='cuda:0')"
 },
 "7": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([-3.3567e-05, -2.1180e-05, 4.9331e-05, ..., 3.0072e-05,\n -3.0421e-05, -1.3027e-05], device='cuda:0')",
 "exp_avg_sq": "tensor([5.4476e-08, 4.8405e-08, 7.9008e-08, ..., 2.3969e-08, 1.7294e-08,\n 3.4884e-08], device='cuda:0')"
 },
 "8": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[ 2.2533e-07, -3.6491e-07, 1.9544e-08, ..., -1.1606e-07,\n -3.8561e-08, -1.0691e-07],\n [-2.8501e-09, 2.0736e-07, 2.8191e-07, ..., -5.6303e-08,\n 4.4420e-08, -7.2615e-08],\n [-3.3458e-08, 3.7879e-08, -2.9230e-07, ..., -2.5236e-07,\n 1.6828e-07, -4.2740e-07],\n ...,\n [ 3.4880e-07, 5.9937e-07, 1.9081e-07, ..., 2.9882e-08,\n 2.7635e-07, 1.0506e-07],\n [ 1.9565e-07, 2.0306e-07, 1.3273e-07, ..., 8.1434e-08,\n 2.1102e-08, 1.8008e-07],\n [-9.6042e-08, -5.2339e-07, 1.4413e-07, ..., 1.1263e-07,\n 4.8820e-07, -5.3852e-07]], device='cuda:0')",
 "exp_avg_sq": "tensor([[1.3493e-12, 6.9531e-12, 1.3923e-12, ..., 7.9550e-13, 3.0363e-12,\n 8.0940e-13],\n [2.5693e-12, 2.0265e-11, 4.0327e-12, ..., 9.3017e-13, 3.3385e-12,\n 1.9184e-12],\n [3.1923e-12, 6.7546e-12, 6.9115e-12, ..., 1.2127e-12, 4.3784e-12,\n 1.1987e-12],\n ...,\n [2.9255e-12, 7.8155e-12, 3.5360e-12, ..., 2.8633e-12, 4.7495e-12,\n 2.4638e-12],\n [3.8279e-12, 7.0421e-12, 4.5877e-12, ..., 4.5972e-13, 4.3932e-12,\n 2.2380e-12],\n [2.7744e-12, 7.5196e-12, 4.6620e-12, ..., 1.2298e-12, 2.7215e-12,\n 2.7425e-12]], device='cuda:0')"
 },
 "9": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[ 1.1257e-06, -1.1332e-06, 3.5403e-07, ..., -1.3819e-06,\n -1.9834e-07, -9.3720e-08],\n [ 2.1521e-06, 8.3455e-07, 1.2749e-07, ..., 2.6951e-06,\n -1.5589e-06, 1.3844e-06],\n [ 2.9979e-07, -4.3879e-07, 9.8788e-07, ..., 1.7778e-06,\n -1.5447e-07, 6.0906e-07],\n ...,\n [-1.0944e-06, 2.7900e-06, 4.8924e-07, ..., -1.9629e-06,\n 1.4278e-06, -1.1931e-06],\n [-1.1666e-06, 2.2008e-06, -8.2252e-07, ..., 1.2749e-06,\n 6.1993e-07, 5.2456e-07],\n [-3.5820e-06, -3.4922e-07, 1.6070e-06, ..., -5.7839e-07,\n -7.3094e-07, -1.5359e-06]], device='cuda:0')",
 "exp_avg_sq": "tensor([[2.5271e-11, 2.5173e-11, 3.1615e-11, ..., 1.3822e-10, 1.1235e-11,\n 1.4908e-11],\n [7.7332e-11, 1.6856e-10, 3.4966e-11, ..., 1.2107e-10, 2.8331e-11,\n 3.4973e-11],\n [2.8069e-11, 1.6841e-10, 2.7036e-11, ..., 6.5007e-11, 1.4384e-11,\n 2.5877e-11],\n ...,\n [5.4526e-11, 1.0417e-10, 7.7903e-11, ..., 6.7396e-11, 1.4786e-11,\n 1.9563e-11],\n [7.1632e-11, 1.6595e-10, 5.9283e-11, ..., 1.6477e-10, 3.4341e-11,\n 3.3941e-11],\n [5.5315e-11, 6.3071e-11, 2.2357e-11, ..., 9.8236e-11, 2.6334e-11,\n 2.7719e-11]], device='cuda:0')"
 },
 "10": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([-9.8412e-06, 5.3439e-05, 3.3113e-05, ..., -4.0688e-05,\n 4.6506e-05, -6.0510e-05], device='cuda:0')",
 "exp_avg_sq": "tensor([3.8579e-08, 5.6686e-08, 2.7208e-08, ..., 3.2055e-08, 6.4011e-08,\n 4.3472e-08], device='cuda:0')"
 },
 "11": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[ 4.2249e-07, -2.4821e-08, 3.2338e-08, ..., 4.5546e-08,\n 7.0483e-07, -8.7041e-07],\n [ 3.4628e-08, -1.8100e-07, -4.3863e-08, ..., 3.2786e-07,\n -2.5009e-07, -5.8800e-07],\n [ 1.9084e-07, 1.9355e-07, -2.0835e-08, ..., -9.8138e-07,\n 1.5909e-06, -3.5732e-07],\n ...,\n [ 2.4751e-07, -2.0029e-07, -1.5109e-07, ..., -1.1197e-08,\n 3.3307e-07, -3.8360e-08],\n [-1.7090e-07, -1.4031e-08, 5.6548e-09, ..., 5.3941e-07,\n 8.7255e-08, 6.6126e-07],\n [ 5.1168e-08, 8.4341e-08, 1.9904e-07, ..., -5.3894e-07,\n 1.2016e-06, 7.0667e-07]], device='cuda:0')",
 "exp_avg_sq": "tensor([[2.3817e-12, 6.4329e-13, 2.7273e-12, ..., 4.5908e-13, 3.4347e-12,\n 3.8494e-12],\n [1.5762e-12, 8.9623e-13, 3.5688e-12, ..., 9.7062e-13, 1.0719e-11,\n 7.8520e-12],\n [1.9753e-12, 1.8680e-12, 3.5836e-12, ..., 4.0390e-12, 6.9893e-12,\n 6.5553e-12],\n ...,\n [1.6169e-12, 1.7135e-12, 5.4932e-12, ..., 6.4126e-13, 8.1325e-12,\n 8.9316e-12],\n [2.6448e-12, 1.3045e-12, 2.8187e-12, ..., 2.8704e-12, 6.1268e-12,\n 9.6076e-12],\n [1.3561e-12, 4.2382e-12, 8.6832e-12, ..., 2.0852e-12, 1.1035e-11,\n 6.4457e-12]], device='cuda:0')"
 },
 "12": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[ 5.9796e-09, 9.3499e-07, 1.5579e-07, ..., -2.0944e-06,\n -1.7651e-07, -6.5781e-07],\n [-4.7834e-07, 1.2536e-06, -3.6740e-07, ..., 6.9910e-07,\n -5.3269e-07, -5.5645e-07],\n [ 9.7025e-07, -2.3489e-06, 1.2687e-06, ..., -1.6225e-07,\n -1.1727e-07, 1.5192e-07],\n ...,\n [-6.7156e-07, -3.0253e-06, -1.5944e-06, ..., -3.3353e-07,\n -2.4778e-07, 3.9353e-07],\n [ 1.1003e-06, 7.6515e-08, -2.5050e-06, ..., 2.3794e-06,\n 7.4912e-07, -1.6795e-09],\n [ 2.4902e-06, -1.5137e-06, 9.4483e-07, ..., 1.0053e-06,\n -8.4765e-07, 3.1075e-07]], device='cuda:0')",
 "exp_avg_sq": "tensor([[1.8581e-11, 3.4814e-11, 5.3236e-12, ..., 1.8445e-11, 7.0485e-12,\n 8.8569e-12],\n [3.3785e-11, 1.3048e-10, 2.9300e-11, ..., 4.8534e-11, 1.8430e-11,\n 1.4938e-11],\n [3.2481e-11, 1.8975e-10, 4.3548e-11, ..., 2.3014e-11, 1.3983e-11,\n 1.4622e-11],\n ...,\n [1.2379e-11, 1.6500e-10, 3.8471e-11, ..., 8.0947e-12, 6.5772e-12,\n 1.1458e-11],\n [5.0638e-11, 9.9707e-11, 1.1152e-10, ..., 4.4787e-11, 1.2355e-11,\n 2.7127e-11],\n [8.5764e-11, 1.4805e-10, 4.6131e-11, ..., 5.2011e-11, 3.0423e-11,\n 2.9652e-11]], device='cuda:0')"
 },
 "13": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([-5.1822e-05, 5.6124e-06, 2.6697e-05, ..., -7.4379e-06,\n 5.9517e-05, -5.1954e-05], device='cuda:0')",
 "exp_avg_sq": "tensor([1.4144e-08, 2.4997e-08, 2.3853e-08, ..., 1.2832e-08, 3.1191e-08,\n 3.9853e-08], device='cuda:0')"
 },
 "14": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[-4.2178e-08, -1.1083e-07, -2.5184e-07, ..., -1.7218e-08,\n -2.2947e-07, -3.4386e-07],\n [-2.0890e-07, -1.4055e-07, -6.2237e-08, ..., -2.3796e-08,\n -1.5353e-07, 1.9388e-07],\n [ 6.7514e-08, -4.3120e-08, 3.9489e-08, ..., -1.5436e-07,\n 2.9775e-07, 3.9818e-07],\n ...,\n [-1.2529e-07, 1.4647e-07, 6.4380e-07, ..., 1.5665e-07,\n 7.9047e-08, -3.7899e-07],\n [-1.1293e-07, 1.0757e-07, -5.1403e-07, ..., 3.6701e-08,\n 2.5578e-08, -4.7015e-08],\n [-2.0512e-07, -2.8718e-07, -7.2149e-07, ..., -1.6093e-07,\n 1.5262e-07, 8.0219e-08]], device='cuda:0')",
 "exp_avg_sq": "tensor([[3.2388e-13, 2.3331e-13, 3.2312e-13, ..., 5.4951e-13, 1.1010e-12,\n 7.6875e-13],\n [6.2549e-13, 5.5610e-13, 3.8753e-13, ..., 1.1791e-12, 2.1127e-12,\n 1.7556e-12],\n [7.1624e-13, 5.7462e-13, 5.9286e-13, ..., 1.1938e-12, 2.7129e-12,\n 1.2372e-12],\n ...,\n [6.6445e-13, 4.8972e-13, 5.1089e-13, ..., 8.6552e-13, 5.1683e-12,\n 1.2847e-12],\n [8.0223e-13, 3.8840e-13, 6.3711e-13, ..., 8.2360e-13, 2.0274e-12,\n 1.2632e-12],\n [8.5375e-13, 8.1285e-13, 1.7481e-12, ..., 1.1044e-12, 1.7617e-12,\n 2.0416e-12]], device='cuda:0')"
 },
 "15": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[ 3.7821e-07, -7.3981e-08, -4.9920e-07, ..., 1.1770e-07,\n 4.1264e-07, -5.7333e-07],\n [ 3.4794e-07, -4.3413e-07, -3.4370e-07, ..., -1.9659e-07,\n -2.6613e-07, -6.1895e-08],\n [ 6.0840e-07, -8.4273e-07, -1.9395e-06, ..., 7.4171e-07,\n 5.3638e-08, 3.1610e-07],\n ...,\n [-5.1594e-07, -2.3084e-07, -2.1965e-07, ..., 7.3815e-07,\n 2.3543e-07, 1.6551e-07],\n [-1.7175e-06, 2.0776e-06, 3.2459e-07, ..., -3.3949e-08,\n 2.1164e-08, -3.4401e-07],\n [-9.5494e-08, -3.8307e-07, -2.4904e-07, ..., -1.0327e-06,\n -5.2967e-08, -4.7399e-07]], device='cuda:0')",
 "exp_avg_sq": "tensor([[1.3716e-11, 5.3200e-11, 2.7633e-11, ..., 1.2020e-11, 9.3705e-12,\n 1.1783e-11],\n [4.7791e-12, 9.7607e-11, 1.4380e-11, ..., 5.3763e-12, 2.2790e-12,\n 6.6963e-12],\n [3.5261e-11, 7.3866e-11, 4.8970e-11, ..., 8.5353e-11, 1.7001e-11,\n 3.5389e-11],\n ...,\n [4.8595e-11, 4.9754e-11, 9.0852e-12, ..., 1.2308e-11, 1.1053e-11,\n 8.4468e-12],\n [1.6299e-11, 2.4861e-11, 1.2925e-11, ..., 5.6340e-11, 1.2803e-11,\n 1.5946e-11],\n [2.7997e-11, 7.0570e-11, 1.6080e-11, ..., 5.5588e-11, 1.3956e-11,\n 2.4377e-11]], device='cuda:0')"
 },
 "16": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([-7.3340e-06, 5.9770e-06, -1.0880e-05, ..., -3.3156e-06,\n -2.1691e-05, -1.0948e-05], device='cuda:0')",
 "exp_avg_sq": "tensor([1.2978e-08, 5.3314e-09, 3.3466e-08, ..., 2.2294e-08, 2.2754e-08,\n 2.5445e-08], device='cuda:0')"
 },
 "17": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[ 5.7374e-08, -1.4840e-08, -4.4788e-08, ..., 1.7042e-07,\n 3.0582e-07, 9.3595e-08],\n [ 5.3102e-08, -1.5809e-07, 5.1671e-07, ..., -1.6341e-07,\n 1.6436e-07, -1.6660e-07],\n [ 5.3036e-08, -1.2327e-07, -1.3706e-07, ..., 1.9600e-07,\n 7.2264e-08, 2.2550e-07],\n ...,\n [-1.6482e-07, -7.1446e-08, -3.0863e-07, ..., 7.5750e-08,\n 1.3128e-07, -1.9444e-07],\n [ 8.2497e-09, -2.2606e-08, 1.0879e-06, ..., -1.0845e-07,\n 1.4117e-07, 2.7050e-07],\n [-4.3799e-08, -4.6149e-08, 2.8926e-07, ..., 2.0145e-07,\n -2.1611e-07, -1.2801e-07]], device='cuda:0')",
 "exp_avg_sq": "tensor([[2.0759e-13, 7.2847e-14, 3.2963e-12, ..., 3.8177e-13, 5.5922e-13,\n 7.6793e-13],\n [4.3584e-13, 1.0078e-13, 5.2098e-12, ..., 7.1577e-13, 5.1435e-13,\n 1.0121e-12],\n [3.5449e-13, 8.5065e-14, 4.6900e-12, ..., 7.6771e-13, 3.5907e-13,\n 6.2767e-13],\n ...,\n [2.8823e-13, 5.3099e-13, 1.3776e-12, ..., 2.1839e-12, 6.1615e-13,\n 8.8780e-13],\n [2.6832e-13, 1.0761e-13, 1.5335e-11, ..., 3.1643e-12, 5.4328e-13,\n 1.1498e-12],\n [4.3822e-13, 2.1752e-13, 3.7081e-12, ..., 2.1308e-12, 5.2495e-13,\n 7.5493e-13]], device='cuda:0')"
 },
 "18": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[-1.4200e-06, 3.9513e-06, 1.4101e-06, ..., -3.4596e-07,\n -6.6080e-08, 4.2703e-07],\n [-5.4096e-08, 3.3109e-08, -3.7329e-09, ..., 5.5504e-08,\n -1.3638e-07, 5.1965e-08],\n [ 6.5129e-07, -2.1776e-06, 4.7292e-07, ..., -1.8807e-07,\n 2.3366e-07, 3.1243e-07],\n ...,\n [ 8.6138e-07, 2.1089e-06, -1.4898e-06, ..., -4.1246e-07,\n 1.0266e-07, -7.8098e-08],\n [-6.8555e-07, 1.6120e-06, 1.1980e-06, ..., -2.6128e-06,\n -1.7880e-06, -9.9117e-07],\n [-6.0787e-07, 5.9868e-07, -1.1578e-08, ..., -3.9902e-07,\n -4.6130e-08, -4.3637e-07]], device='cuda:0')",
 "exp_avg_sq": "tensor([[6.0617e-11, 6.7220e-11, 3.3627e-11, ..., 1.5765e-11, 7.0272e-12,\n 2.5205e-11],\n [2.9699e-12, 6.0459e-12, 1.6371e-12, ..., 2.1198e-11, 1.8036e-12,\n 2.4030e-12],\n [4.2330e-11, 9.4012e-11, 2.4975e-11, ..., 9.3038e-11, 1.7284e-11,\n 2.8360e-11],\n ...,\n [8.2017e-11, 1.6030e-10, 6.7723e-11, ..., 3.2793e-11, 2.9998e-11,\n 1.9570e-11],\n [3.7448e-11, 1.4493e-10, 3.4588e-11, ..., 1.7635e-11, 1.6769e-11,\n 1.6112e-11],\n [8.1883e-11, 8.2047e-11, 3.2170e-11, ..., 3.7603e-11, 1.9865e-11,\n 1.8104e-11]], device='cuda:0')"
 },
 "19": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([-2.7514e-05, 1.8098e-06, 6.0252e-06, ..., 1.8500e-05,\n -4.0251e-05, -1.1021e-05], device='cuda:0')",
 "exp_avg_sq": "tensor([1.6803e-08, 5.6127e-09, 3.8600e-08, ..., 3.1972e-08, 2.6589e-08,\n 2.3604e-08], device='cuda:0')"
 },
 "20": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[-2.6969e-08, -2.7534e-08, 8.0372e-08, ..., 6.7182e-08,\n -6.3088e-08, 1.0196e-07],\n [-2.6521e-08, 2.1468e-08, 6.3594e-08, ..., -7.6155e-07,\n 9.5530e-08, 3.5637e-08],\n [-1.6642e-07, -2.1566e-08, -2.5673e-08, ..., -3.1602e-07,\n -9.5380e-08, 8.4636e-08],\n ...,\n [-5.0042e-08, 4.9138e-08, 1.4474e-07, ..., -1.0605e-07,\n -3.5191e-07, -6.1000e-08],\n [-8.8281e-09, 3.4481e-08, 2.0243e-07, ..., -4.4518e-08,\n -2.2817e-07, -4.0505e-08],\n [-1.8497e-08, -1.1059e-08, 2.8440e-07, ..., 6.9356e-09,\n 3.9773e-07, -4.3965e-07]], device='cuda:0')",
 "exp_avg_sq": "tensor([[4.2779e-13, 3.2385e-14, 1.6073e-12, ..., 1.7964e-12, 9.1189e-13,\n 6.4447e-13],\n [3.0758e-13, 7.8709e-14, 1.3436e-12, ..., 7.5955e-13, 1.4246e-12,\n 8.2253e-13],\n [4.8537e-13, 3.8691e-14, 8.6824e-13, ..., 1.6551e-12, 2.3192e-12,\n 9.9392e-13],\n ...,\n [4.5729e-13, 5.0852e-14, 1.5986e-12, ..., 1.3794e-12, 2.1937e-12,\n 1.2689e-12],\n [4.0272e-13, 4.5470e-14, 1.4629e-12, ..., 1.7628e-12, 1.9400e-12,\n 7.8826e-13],\n [4.1588e-13, 6.3669e-14, 1.2466e-12, ..., 9.0235e-13, 2.8169e-12,\n 2.0460e-12]], device='cuda:0')"
 },
 "21": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[-9.2264e-07, 9.7451e-07, -5.7227e-07, ..., -1.8680e-07,\n -4.6338e-07, -1.8832e-07],\n [-1.8448e-06, -1.5227e-06, -5.9326e-09, ..., -1.3713e-05,\n 6.4138e-07, -6.2495e-06],\n [-1.8261e-07, 3.4495e-07, 4.9182e-08, ..., -1.9561e-08,\n 1.0297e-07, -4.0577e-07],\n ...,\n [-5.3073e-07, -2.3422e-07, -1.5449e-07, ..., 8.9343e-07,\n 5.3517e-07, 2.8172e-07],\n [-3.3119e-06, -6.8422e-07, 1.6122e-06, ..., -5.7382e-07,\n -2.5795e-07, 2.5007e-06],\n [-3.1774e-06, 2.0617e-06, -1.6274e-07, ..., -2.6097e-06,\n 3.0713e-07, -1.6423e-06]], device='cuda:0')",
 "exp_avg_sq": "tensor([[6.0684e-11, 3.2876e-11, 5.7541e-12, ..., 4.7266e-12, 4.9226e-12,\n 5.3178e-12],\n [1.3890e-10, 1.0436e-10, 1.7261e-11, ..., 8.5036e-11, 1.8595e-11,\n 3.3957e-11],\n [5.9364e-11, 9.9784e-11, 1.2720e-11, ..., 1.3619e-11, 5.9732e-11,\n 2.1398e-11],\n ...,\n [7.8430e-12, 6.2142e-12, 1.0249e-11, ..., 4.6510e-11, 1.3840e-11,\n 7.1290e-12],\n [1.0389e-10, 1.7364e-10, 7.4957e-11, ..., 1.0941e-10, 4.2098e-11,\n 6.5392e-11],\n [3.5059e-11, 3.3718e-11, 9.7651e-12, ..., 2.6041e-11, 1.2271e-11,\n 1.2794e-11]], device='cuda:0')"
 },
 "22": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([-1.6798e-05, -1.7188e-04, -6.2626e-06, ..., 1.1163e-05,\n 1.7800e-05, -8.2458e-05], device='cuda:0')",
 "exp_avg_sq": "tensor([8.0889e-09, 3.9734e-08, 2.8274e-08, ..., 1.2901e-08, 6.2306e-08,\n 1.7412e-08], device='cuda:0')"
 },
 "23": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[ 4.8160e-08, 9.7232e-08, 6.1995e-09, ..., 1.1341e-08,\n 5.5137e-08, 3.7814e-09],\n [ 3.8829e-08, -2.9346e-07, 3.9520e-08, ..., 2.3462e-07,\n 3.8380e-07, 5.1998e-09],\n [-6.8992e-08, 4.1956e-08, -6.3252e-08, ..., -6.1109e-08,\n -5.7157e-07, 5.2603e-08],\n ...,\n [-8.2720e-08, 1.7356e-07, 8.8102e-08, ..., -8.0725e-08,\n 1.1247e-07, -2.1625e-08],\n [ 1.6160e-07, 1.5476e-07, 4.3431e-08, ..., -1.4741e-07,\n 8.2731e-08, 6.9379e-08],\n [-7.5909e-08, -1.4963e-07, -4.4853e-08, ..., 1.3661e-07,\n -7.9781e-07, -7.5500e-08]], device='cuda:0')",
 "exp_avg_sq": "tensor([[9.9975e-14, 8.7121e-14, 1.8312e-13, ..., 1.9079e-13, 2.0039e-13,\n 8.3543e-14],\n [1.7750e-13, 1.9955e-13, 7.2823e-13, ..., 4.4416e-13, 4.9392e-13,\n 1.3645e-13],\n [1.1694e-13, 2.1187e-13, 7.5666e-13, ..., 3.9137e-13, 8.5671e-13,\n 1.6829e-13],\n ...,\n [1.1867e-13, 3.0018e-13, 6.3095e-13, ..., 3.3006e-13, 4.5596e-13,\n 2.1084e-13],\n [1.5031e-13, 3.7845e-13, 8.5406e-13, ..., 3.4806e-13, 6.7384e-13,\n 9.1889e-14],\n [1.6262e-13, 2.4122e-13, 5.0382e-13, ..., 7.9451e-13, 1.0393e-12,\n 8.7368e-14]], device='cuda:0')"
 },
 "24": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[-1.9345e-06, -8.1528e-07, 3.5459e-07, ..., -3.7359e-07,\n 5.6320e-07, -4.0500e-07],\n [-4.5275e-07, 3.6220e-07, 6.6379e-07, ..., 3.8153e-07,\n -3.5411e-07, -1.9950e-07],\n [ 6.5596e-07, -3.4292e-08, 4.2077e-07, ..., -1.7336e-06,\n -3.6086e-07, 6.5141e-07],\n ...,\n [ 1.9383e-06, -1.4437e-06, -1.0149e-07, ..., 1.0160e-07,\n 5.1157e-07, -1.8201e-07],\n [ 4.3796e-07, -4.3321e-07, -1.7464e-06, ..., 6.6490e-07,\n 1.6888e-08, 5.6190e-09],\n [-1.1502e-07, 1.1608e-07, -1.1530e-07, ..., -5.2798e-07,\n -8.8619e-07, -8.1995e-07]], device='cuda:0')",
 "exp_avg_sq": "tensor([[4.3379e-11, 3.6635e-11, 8.0389e-12, ..., 1.7529e-11, 1.1370e-11,\n 5.1811e-12],\n [8.4566e-12, 1.3477e-11, 7.6721e-12, ..., 5.4455e-11, 8.6119e-12,\n 1.3624e-11],\n [1.3933e-11, 3.2377e-11, 1.2014e-11, ..., 2.7394e-11, 6.7498e-12,\n 8.2478e-12],\n ...,\n [9.1584e-11, 5.1899e-11, 7.0232e-12, ..., 1.9096e-11, 1.6022e-11,\n 8.1464e-12],\n [1.0773e-11, 3.1622e-11, 3.4997e-11, ..., 1.7988e-11, 5.5866e-12,\n 7.1113e-12],\n [1.2285e-11, 3.0416e-11, 3.8584e-11, ..., 2.7994e-11, 4.4026e-12,\n 6.0065e-12]], device='cuda:0')"
 },
 "25": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([-2.4717e-05, -1.3930e-06, -4.0413e-06, ..., 2.2503e-05,\n 4.5223e-05, 8.5674e-06], device='cuda:0')",
 "exp_avg_sq": "tensor([1.2556e-08, 1.7224e-08, 1.0642e-08, ..., 1.7508e-08, 9.8654e-09,\n 9.2997e-09], device='cuda:0')"
 },
 "26": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[ 4.1498e-08, 6.6717e-09, 2.0117e-09, ..., 3.7049e-08,\n 2.6488e-08, 4.8490e-08],\n [ 7.3546e-08, -5.1798e-08, 2.7361e-08, ..., -5.1817e-08,\n 9.6695e-09, -3.2698e-08],\n [ 1.4129e-07, 7.0925e-08, -5.2988e-08, ..., -3.4695e-08,\n -6.2380e-08, -9.8423e-08],\n ...,\n [ 1.6098e-07, 5.1665e-08, -8.1355e-08, ..., -1.6317e-08,\n 1.7053e-07, 1.0338e-07],\n [-4.6993e-08, 2.5313e-08, -1.4197e-07, ..., -5.5542e-09,\n -9.0433e-08, 3.9964e-08],\n [-1.1944e-07, -8.9079e-09, 7.5279e-08, ..., 7.8255e-08,\n -9.3269e-08, 1.3340e-07]], device='cuda:0')",
 "exp_avg_sq": "tensor([[1.1203e-13, 1.2254e-13, 7.5502e-14, ..., 1.0118e-13, 1.6917e-13,\n 1.5143e-13],\n [1.4490e-13, 1.4642e-13, 1.4111e-13, ..., 1.6598e-13, 2.8176e-13,\n 1.0873e-13],\n [2.5064e-13, 1.4403e-13, 1.8215e-13, ..., 3.6392e-13, 3.2958e-13,\n 2.5519e-13],\n ...,\n [1.3280e-13, 1.1923e-13, 1.3842e-13, ..., 1.7936e-13, 2.3590e-13,\n 1.9754e-13],\n [1.3693e-13, 1.8395e-13, 1.1762e-13, ..., 5.5443e-13, 1.4574e-13,\n 1.8353e-13],\n [1.6913e-13, 1.3227e-13, 1.3451e-13, ..., 2.8351e-13, 3.4921e-13,\n 1.5417e-13]], device='cuda:0')"
 },
 "27": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([-5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([1.6259e-20], device='cuda:0')"
 },
 "28": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([-5.6052e-45, 5.6052e-45, -5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([2.1864e-22, 1.4165e-22, 5.0668e-23], device='cuda:0')"
 },
 "29": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([1.6388e-21, 2.1687e-22, 4.5121e-22, 1.8944e-22, 8.6055e-23, 2.1356e-22,\n 1.3159e-21, 1.9580e-21, 9.0675e-23], device='cuda:0')"
 },
 "31": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[-5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, 5.6052e-45, ..., -5.6052e-45,\n -5.6052e-45, -5.6052e-45],\n [-5.6052e-45, 5.6052e-45, -5.6052e-45, ..., -5.6052e-45,\n 5.6052e-45, -5.6052e-45],\n ...,\n [ 5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, -5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45]], device='cuda:0')",
 "exp_avg_sq": "tensor([[5.5064e-27, 6.3416e-28, 1.1637e-27, ..., 7.7062e-28, 1.8289e-28,\n 2.3819e-27],\n [1.5297e-28, 1.9171e-27, 7.0857e-29, ..., 1.5980e-29, 2.1763e-29,\n 1.0146e-28],\n [1.1051e-26, 1.4486e-27, 1.6613e-26, ..., 1.1354e-27, 1.5583e-27,\n 1.4159e-27],\n ...,\n [1.9937e-28, 1.5072e-28, 5.4865e-29, ..., 1.0224e-28, 1.8392e-29,\n 6.9763e-29],\n [1.0794e-28, 1.6624e-28, 3.3187e-28, ..., 1.2635e-28, 8.7720e-29,\n 1.1804e-28],\n [7.8819e-27, 1.3501e-27, 3.2821e-27, ..., 2.4904e-27, 6.9494e-28,\n 2.4961e-27]], device='cuda:0')"
 },
 "32": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([-5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([7.3231e-24, 1.6534e-25, 1.4200e-23, 1.8967e-24, 2.6095e-25, 6.8914e-24,\n 8.5732e-26, 3.2692e-23, 1.3419e-25, 1.3624e-23, 1.3514e-23, 1.4620e-24,\n 8.6838e-24, 1.9893e-25, 9.0708e-25, 1.6680e-23, 5.0409e-25, 2.7245e-24,\n 1.8609e-23, 6.1837e-24, 1.8206e-23, 2.8363e-25, 2.3447e-24, 5.1340e-24,\n 1.8278e-25, 3.1147e-23, 1.1983e-25, 5.5273e-24, 1.4950e-23, 3.6164e-25,\n 2.1539e-24, 6.1439e-26, 9.8670e-24, 3.0081e-24, 2.2942e-24, 6.7090e-24,\n 2.2285e-25, 1.2046e-24, 7.0442e-25, 4.7851e-24, 3.5184e-25, 2.5528e-24,\n 3.2756e-24, 1.0114e-23, 3.7294e-24, 8.1939e-25, 5.0669e-24, 8.5159e-26,\n 2.8505e-24, 2.6466e-23, 5.5744e-24, 3.6304e-24, 2.8801e-24, 8.9080e-26,\n 8.4585e-24, 1.2619e-25, 1.2779e-25, 2.3092e-23, 4.6001e-23, 1.9085e-25,\n 9.1739e-25, 2.1959e-24, 1.4801e-23, 4.7628e-25, 2.2750e-25, 1.7984e-25,\n 1.6109e-25, 5.0110e-24, 1.4906e-24, 2.7876e-24, 3.0224e-25, 1.7779e-25,\n 5.6082e-23, 5.2148e-24, 1.6788e-25, 4.6720e-24, 2.2148e-24, 6.6602e-25,\n 2.2730e-25, 5.3428e-25, 3.6598e-24, 6.1308e-25, 3.3761e-24, 7.1794e-25,\n 4.4116e-25, 3.6694e-25, 9.4937e-25, 1.5700e-24, 5.4394e-24, 1.2655e-23,\n 3.7520e-24, 1.5046e-23, 3.0127e-25, 2.2844e-23, 3.3754e-26, 2.8644e-24,\n 2.1594e-23, 2.2860e-25, 3.4044e-24, 5.0137e-24, 1.3471e-24, 5.9033e-24,\n 2.3993e-24, 4.5488e-24, 1.1067e-24, 8.9561e-24, 4.4197e-25, 1.5856e-24,\n 4.8101e-24, 4.7490e-25, 6.2727e-24, 7.6262e-24, 2.7145e-24, 8.2445e-24,\n 3.8607e-23, 2.1498e-25, 3.6315e-25, 1.8423e-24, 1.1564e-24, 3.3358e-25,\n 7.1971e-24, 5.5115e-24, 2.4790e-24, 4.2027e-24, 8.9855e-24, 5.9350e-25,\n 3.4925e-24, 2.3408e-24, 1.1148e-23, 1.8419e-24, 4.3482e-25, 9.6362e-24,\n 9.1980e-26, 4.2927e-25, 1.5772e-24, 1.2856e-23, 1.7652e-24, 3.3268e-23,\n 1.3136e-24, 7.0199e-25, 7.6801e-25, 3.6451e-24, 1.3503e-24, 1.7510e-25,\n 5.8444e-25, 2.6680e-23, 6.1220e-25, 1.2830e-23, 8.0749e-24, 1.2957e-24,\n 4.3545e-23, 2.2710e-25, 6.1753e-24, 7.2778e-24, 1.5491e-24, 1.5531e-25,\n 2.6277e-26, 1.5534e-23, 6.0848e-23, 4.7373e-25, 1.2941e-23, 4.0897e-24,\n 1.5179e-24, 2.6275e-24, 7.5980e-24, 1.7182e-23, 5.3572e-23, 1.2674e-23,\n 3.8437e-23, 1.3912e-23, 1.8752e-23, 4.1263e-24, 7.8972e-24, 4.3213e-24,\n 2.0957e-24, 8.4693e-24, 2.3739e-24, 3.0901e-23, 4.8133e-24, 7.0464e-24,\n 7.0616e-24, 1.0406e-24, 1.0869e-24, 1.7961e-23, 8.4986e-24, 7.3692e-24,\n 1.1890e-23, 1.5557e-25, 3.9569e-25, 9.6398e-24, 3.6156e-24, 7.5321e-24,\n 3.0041e-23, 1.5455e-24, 6.6541e-25, 5.1084e-26, 4.4852e-25, 2.9987e-25,\n 1.6631e-23, 3.9703e-23, 6.3951e-24, 6.8999e-26, 1.3879e-25, 7.9800e-26,\n 1.9085e-25, 2.7612e-25, 5.9706e-24, 1.4511e-25, 8.7076e-24, 1.5954e-23,\n 3.4972e-24, 5.2194e-23, 7.7037e-24, 2.1378e-24, 4.0613e-24, 6.6593e-24,\n 6.5338e-25, 6.1420e-25, 1.0806e-24, 1.2226e-23, 2.0880e-23, 2.5957e-25,\n 4.9247e-25, 2.1694e-26, 6.3834e-23, 3.2406e-23, 4.0859e-24, 8.9255e-25,\n 1.0171e-23, 1.2971e-24, 2.2820e-24, 1.6914e-25, 1.1400e-23, 3.3898e-25,\n 9.7174e-25, 4.4595e-25, 5.5366e-24, 4.5507e-24, 3.5949e-24, 3.6146e-25,\n 5.7757e-24, 1.2595e-23, 2.4644e-23, 3.5675e-24, 5.9570e-25, 9.1975e-24,\n 9.6694e-24, 6.3704e-24, 3.0730e-24, 5.5054e-25, 2.2040e-24, 1.9697e-24,\n 1.8784e-23, 2.9793e-25, 1.4895e-25, 1.4861e-23], device='cuda:0')"
 },
 "33": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([1.5189e-26, 2.7214e-28, 1.8245e-26, 1.3218e-27, 3.6618e-28, 6.4278e-27,\n 2.0587e-28, 9.0836e-26, 4.5375e-29, 1.6443e-26, 4.5533e-26, 6.5377e-28,\n 2.0811e-26, 1.8214e-28, 1.0645e-28, 1.2611e-25, 1.2045e-27, 1.3758e-26,\n 2.9591e-25, 1.7016e-26, 1.9960e-26, 5.8460e-28, 5.6864e-28, 1.4001e-26,\n 8.5463e-29, 1.5555e-25, 2.8660e-28, 6.4625e-27, 1.5199e-26, 4.5598e-27,\n 1.4080e-27, 1.6529e-28, 1.0946e-26, 9.3981e-28, 4.2426e-27, 4.9978e-27,\n 3.0814e-28, 5.7732e-28, 1.7245e-28, 1.3361e-26, 9.7276e-28, 1.9698e-27,\n 1.0429e-26, 2.5894e-26, 8.6861e-27, 4.8866e-27, 3.9375e-27, 2.0715e-28,\n 6.6331e-27, 1.9998e-25, 2.4634e-27, 4.4847e-28, 3.0434e-27, 3.7663e-28,\n 6.2711e-27, 1.3119e-28, 2.7092e-28, 5.3664e-26, 2.6007e-26, 7.0152e-29,\n 9.3960e-27, 1.0948e-27, 1.6842e-26, 4.3041e-28, 2.0442e-28, 4.9348e-29,\n 4.5254e-28, 2.6198e-27, 2.1539e-28, 4.2677e-27, 3.4848e-28, 2.6937e-28,\n 8.1500e-26, 1.4854e-26, 2.6479e-28, 5.5411e-27, 3.0819e-27, 4.7394e-28,\n 4.5398e-28, 4.1439e-27, 5.5895e-26, 1.0220e-28, 8.3306e-27, 2.9477e-28,\n 2.7755e-28, 3.1010e-28, 9.9276e-28, 2.7287e-27, 2.4525e-26, 1.9370e-26,\n 4.2546e-27, 9.9509e-26, 4.8266e-28, 3.7066e-26, 1.9981e-28, 1.1820e-26,\n 6.3336e-26, 3.3745e-28, 3.4642e-27, 2.6966e-26, 1.5696e-27, 6.9705e-27,\n 7.1335e-27, 1.8638e-26, 1.3506e-28, 4.2961e-27, 1.5941e-27, 1.6374e-27,\n 9.4496e-28, 2.2553e-27, 2.4521e-26, 3.6613e-26, 6.1783e-27, 1.2649e-25,\n 1.2234e-25, 7.2441e-29, 8.9251e-28, 1.1123e-27, 6.0888e-27, 9.6994e-29,\n 9.7979e-27, 2.4603e-27, 5.3210e-27, 1.4015e-27, 2.0047e-26, 7.6168e-27,\n 5.1894e-27, 1.6372e-27, 3.3925e-26, 1.7274e-28, 1.5990e-28, 1.2983e-26,\n 8.0657e-28, 1.2818e-27, 3.3248e-28, 1.7398e-26, 3.9254e-27, 1.0970e-25,\n 1.3416e-27, 2.2621e-28, 5.4321e-28, 2.5690e-27, 5.4459e-28, 1.2867e-28,\n 1.5587e-28, 2.3774e-26, 2.6642e-28, 2.8524e-26, 3.5327e-27, 6.7005e-28,\n 4.8882e-26, 3.8443e-28, 3.6497e-27, 5.2615e-27, 6.3223e-27, 1.2972e-28,\n 5.4398e-29, 4.4079e-26, 8.2323e-25, 5.8435e-28, 1.5399e-25, 7.1823e-27,\n 1.2124e-27, 2.6973e-27, 1.8355e-26, 2.0847e-26, 6.3787e-26, 6.2998e-27,\n 4.0568e-26, 3.2214e-26, 6.4973e-26, 9.0729e-27, 1.3989e-26, 3.5275e-26,\n 1.4112e-27, 2.9762e-27, 3.2759e-27, 2.7536e-26, 3.1128e-27, 7.1274e-26,\n 5.4593e-27, 1.5976e-27, 2.3470e-27, 1.6952e-26, 6.0163e-26, 3.6065e-27,\n 6.4213e-26, 1.4453e-28, 1.2495e-27, 9.1467e-27, 1.9605e-27, 4.3305e-26,\n 5.3972e-26, 4.7536e-27, 3.5348e-28, 1.4267e-28, 3.6332e-28, 4.9336e-28,\n 4.0483e-26, 4.9096e-25, 3.7572e-27, 2.0055e-28, 4.5503e-28, 9.4422e-30,\n 5.7129e-28, 3.0056e-28, 2.8742e-27, 2.3496e-28, 7.4252e-27, 3.9885e-26,\n 2.5793e-26, 7.9929e-26, 2.3379e-26, 7.2452e-28, 8.0425e-27, 5.2899e-27,\n 3.6208e-28, 6.3231e-28, 3.8539e-28, 4.3655e-26, 4.1735e-26, 5.7727e-28,\n 1.8914e-27, 4.0686e-28, 1.1278e-25, 4.2952e-26, 2.5637e-27, 3.2636e-28,\n 2.0769e-26, 1.4071e-27, 5.3467e-27, 7.6360e-29, 4.2227e-26, 9.0930e-28,\n 2.6035e-28, 1.8005e-28, 1.7524e-26, 2.0619e-27, 2.2674e-27, 5.5958e-28,\n 5.0041e-27, 1.7882e-26, 3.4849e-26, 2.7588e-27, 7.4198e-28, 2.0490e-26,\n 6.1062e-26, 1.2202e-26, 6.5340e-27, 3.7138e-28, 4.7413e-27, 2.7386e-28,\n 1.0929e-25, 5.3604e-28, 2.1791e-28, 8.6620e-26], device='cuda:0')"
 },
 "34": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([1.2148e-26, 1.0310e-27, 2.1471e-26, 2.3592e-27, 1.1809e-27, 1.8736e-26,\n 2.3300e-27, 1.0084e-25, 1.7032e-27, 2.6350e-26, 2.6774e-26, 2.3428e-27,\n 3.9701e-26, 3.3605e-28, 1.2153e-27, 6.6381e-26, 4.1204e-27, 1.6854e-26,\n 7.0228e-26, 2.3470e-26, 6.0053e-26, 4.1464e-27, 2.8314e-27, 9.6680e-27,\n 5.3468e-29, 1.0591e-25, 1.3377e-28, 4.1536e-27, 1.6056e-26, 2.6222e-27,\n 1.0067e-26, 5.8334e-29, 1.4320e-26, 3.0151e-27, 1.1463e-26, 1.6969e-27,\n 1.1114e-27, 4.5118e-27, 3.4908e-27, 1.7842e-26, 2.3423e-27, 2.9725e-27,\n 9.4206e-27, 4.4834e-26, 1.3618e-26, 6.9557e-27, 1.5157e-26, 6.9335e-28,\n 7.7630e-27, 7.6302e-26, 5.3372e-27, 1.6255e-27, 7.2456e-27, 1.2053e-27,\n 9.8340e-27, 1.0527e-28, 4.2181e-28, 8.0018e-26, 6.0148e-26, 2.9798e-28,\n 2.9831e-27, 4.8851e-27, 5.0791e-26, 2.1213e-27, 1.0071e-27, 1.7208e-27,\n 9.7232e-29, 6.7416e-27, 1.8692e-27, 1.1412e-26, 1.7831e-27, 3.8288e-28,\n 6.5633e-26, 1.0620e-26, 3.3837e-28, 1.5843e-26, 6.8899e-27, 1.6155e-27,\n 3.3667e-27, 5.0050e-27, 2.8752e-26, 5.3651e-27, 7.8073e-27, 8.8168e-28,\n 8.7658e-29, 2.4924e-27, 7.3621e-28, 9.1946e-27, 2.8277e-26, 3.4009e-26,\n 4.6487e-27, 3.3961e-26, 1.4833e-27, 2.8227e-26, 2.9948e-29, 1.7421e-26,\n 2.4909e-26, 9.9345e-28, 3.9669e-27, 1.2179e-26, 4.8366e-27, 8.4889e-27,\n 6.5905e-27, 1.6752e-26, 1.0270e-27, 6.2344e-27, 3.3965e-27, 5.7104e-27,\n 2.9291e-27, 1.0130e-27, 2.0628e-26, 3.2447e-26, 1.2371e-26, 4.4666e-26,\n 4.2546e-26, 3.7863e-28, 1.4955e-27, 4.3242e-27, 8.5039e-27, 7.5749e-28,\n 1.3971e-26, 1.1129e-26, 8.8090e-27, 1.2332e-26, 3.1127e-26, 4.2062e-27,\n 9.0263e-27, 4.7482e-27, 1.6357e-26, 4.2270e-27, 9.7908e-28, 1.4415e-26,\n 1.4786e-27, 1.7046e-27, 5.2666e-27, 2.1284e-26, 5.4652e-27, 3.8023e-26,\n 6.1309e-27, 9.0148e-28, 1.9084e-27, 5.4292e-27, 1.2332e-27, 1.2181e-27,\n 1.1106e-27, 3.0020e-26, 1.8967e-27, 2.6117e-26, 1.0167e-26, 1.9954e-27,\n 4.5353e-26, 3.8564e-28, 3.6483e-27, 2.7771e-26, 1.3131e-26, 5.0031e-28,\n 2.5033e-29, 6.0973e-26, 2.0435e-25, 5.3403e-27, 6.4992e-26, 7.3216e-27,\n 6.5402e-27, 3.1278e-27, 1.4172e-26, 4.9003e-26, 1.4602e-25, 4.1895e-26,\n 4.0852e-26, 2.1303e-26, 5.8230e-26, 1.7244e-26, 3.4945e-26, 2.6273e-26,\n 3.9592e-27, 8.0259e-27, 3.2477e-27, 2.6257e-26, 1.8549e-26, 4.3281e-26,\n 4.5706e-27, 5.0982e-27, 5.7967e-27, 1.8502e-26, 3.6372e-26, 5.9280e-27,\n 5.8927e-26, 1.3535e-27, 5.2393e-28, 3.1900e-27, 7.1683e-27, 3.6234e-26,\n 3.3755e-26, 5.0253e-27, 2.8966e-27, 3.8071e-28, 6.6328e-28, 1.1882e-27,\n 4.9819e-26, 1.4570e-25, 1.5289e-27, 9.4343e-29, 1.8946e-27, 1.7149e-28,\n 2.0104e-27, 7.3159e-28, 6.6502e-27, 1.3587e-27, 9.8471e-27, 4.3398e-26,\n 6.6772e-27, 1.3173e-25, 2.6242e-26, 2.9190e-27, 2.1551e-26, 1.2742e-26,\n 7.2952e-27, 5.3891e-27, 5.7714e-27, 4.9069e-26, 5.3430e-26, 1.2139e-27,\n 1.6685e-27, 9.2448e-29, 1.6954e-25, 9.6740e-26, 4.3195e-27, 3.2606e-27,\n 2.4701e-26, 2.9269e-27, 8.8998e-27, 3.4565e-28, 4.6490e-26, 7.0369e-28,\n 8.3485e-28, 1.8943e-27, 2.5341e-26, 1.2912e-26, 1.0940e-26, 2.5350e-27,\n 2.5600e-26, 1.5269e-26, 2.2124e-26, 8.1187e-27, 8.4654e-28, 3.2879e-26,\n 1.5441e-26, 2.7206e-26, 1.4760e-26, 1.0103e-27, 8.2336e-27, 2.9142e-27,\n 7.9401e-26, 1.4900e-28, 8.7231e-28, 6.8038e-26], device='cuda:0')"
 },
 "35": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[ 5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n ...,\n [-5.6052e-45, -5.6052e-45, 5.6052e-45, ..., -5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n [-5.6052e-45, 5.6052e-45, -5.6052e-45, ..., -5.6052e-45,\n -5.6052e-45, -5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45]], device='cuda:0')",
 "exp_avg_sq": "tensor([[2.2414e-28, 9.5895e-28, 7.1889e-29, ..., 6.5977e-28, 1.3207e-28,\n 1.1053e-28],\n [3.2367e-28, 1.3464e-27, 2.3582e-28, ..., 3.7194e-28, 2.4475e-28,\n 9.7763e-28],\n [4.6166e-27, 1.1155e-27, 9.1065e-27, ..., 1.0266e-27, 6.7466e-28,\n 8.9855e-28],\n ...,\n [1.3968e-27, 8.7862e-29, 2.4695e-28, ..., 3.9513e-27, 1.5327e-28,\n 5.0227e-28],\n [8.6813e-27, 9.6357e-28, 4.2726e-27, ..., 2.7869e-27, 4.4112e-28,\n 1.6690e-27],\n [3.7870e-28, 5.0210e-28, 5.8890e-28, ..., 1.5995e-28, 5.4883e-29,\n 1.6646e-28]], device='cuda:0')"
 },
 "36": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([4.4206e-25, 4.2580e-25, 4.4914e-24, 7.6750e-24, 8.7278e-24, 7.2934e-24,\n 2.2397e-25, 1.6519e-23, 8.0999e-26, 5.7960e-25, 5.0161e-25, 1.6460e-24,\n 3.4625e-24, 6.4653e-25, 9.1407e-24, 5.7961e-24, 3.9236e-26, 8.3757e-25,\n 1.5847e-23, 1.2150e-23, 3.7314e-24, 1.0787e-25, 3.8960e-24, 3.1169e-24,\n 1.0148e-25, 2.0241e-23, 3.4956e-24, 1.0747e-23, 1.5174e-24, 4.4328e-25,\n 1.1108e-24, 2.8011e-25, 2.6625e-23, 1.2483e-24, 3.9417e-25, 2.7247e-24,\n 1.8858e-24, 8.4178e-25, 2.2047e-23, 8.5749e-25, 1.7692e-24, 7.9236e-24,\n 1.7443e-25, 2.2195e-23, 8.5342e-25, 1.0585e-24, 9.3847e-24, 5.6031e-26,\n 1.5320e-24, 1.6340e-23, 5.0699e-24, 2.9880e-24, 1.3518e-24, 3.9919e-24,\n 1.9470e-23, 4.4025e-25, 1.0741e-25, 7.3401e-24, 8.7731e-23, 3.0507e-24,\n 4.4034e-24, 3.4969e-24, 7.4845e-24, 1.1137e-24, 3.7239e-24, 1.1853e-25,\n 1.0872e-23, 1.0886e-23, 7.5020e-24, 2.5501e-24, 1.4574e-24, 4.6258e-24,\n 1.6132e-23, 2.9453e-24, 5.4148e-25, 5.7143e-26, 1.3962e-24, 1.2221e-24,\n 7.3197e-26, 1.2736e-24, 2.8454e-24, 1.9901e-24, 6.7845e-25, 3.0142e-25,\n 5.9675e-25, 8.5853e-24, 8.2582e-25, 2.5571e-25, 1.8840e-25, 1.2328e-23,\n 1.7321e-24, 1.3169e-23, 6.7516e-25, 6.2183e-24, 2.2948e-23, 1.9061e-24,\n 4.2298e-24, 1.3701e-25, 9.3377e-25, 5.1179e-24, 3.7925e-25, 7.7843e-25,\n 4.9869e-25, 2.9018e-25, 2.6609e-24, 1.3186e-23, 1.4756e-24, 2.5724e-25,\n 5.0323e-25, 3.2028e-24, 6.0770e-24, 2.1013e-24, 8.1767e-26, 4.4395e-26,\n 2.8519e-24, 1.3470e-24, 1.0359e-24, 2.5299e-24, 2.1362e-24, 3.0392e-24,\n 1.9271e-23, 1.5441e-23, 8.9735e-25, 4.0104e-25, 3.0012e-25, 1.5928e-23,\n 2.6124e-24, 1.4550e-25, 5.9863e-24, 2.1761e-23, 4.2018e-24, 8.5647e-24,\n 3.9299e-25, 1.6481e-24, 6.2751e-24, 7.4831e-25, 5.9047e-24, 1.5053e-23,\n 1.4766e-24, 5.3800e-26, 9.5830e-24, 4.4235e-26, 1.9168e-24, 3.0045e-24,\n 2.4141e-25, 1.0946e-23, 1.1216e-23, 4.6436e-25, 5.3245e-24, 8.3381e-24,\n 3.1207e-23, 4.4086e-25, 7.2995e-24, 6.7902e-24, 1.4935e-23, 2.5081e-24,\n 1.1725e-24, 1.8824e-24, 3.1854e-24, 4.5487e-26, 1.4748e-23, 9.1558e-25,\n 1.3161e-25, 1.8946e-24, 1.0743e-25, 7.9281e-24, 4.3995e-24, 5.6638e-24,\n 8.0934e-24, 1.9703e-26, 2.0890e-23, 1.8253e-25, 3.8783e-24, 5.4636e-24,\n 7.2179e-24, 3.7968e-24, 1.6146e-24, 7.1535e-26, 2.7533e-23, 2.8260e-26,\n 1.3016e-24, 1.9457e-24, 3.2940e-24, 4.2367e-24, 1.4113e-24, 4.9740e-25,\n 4.6034e-24, 1.8703e-24, 2.1462e-24, 1.2158e-23, 1.8549e-24, 6.4092e-24,\n 1.5982e-23, 7.4943e-24, 4.8867e-24, 6.5285e-24, 2.8549e-26, 9.6909e-24,\n 3.7716e-24, 1.1026e-23, 8.5362e-25, 6.8072e-25, 5.4864e-25, 5.2768e-25,\n 4.7753e-24, 6.9163e-24, 9.3814e-25, 1.0213e-24, 1.6315e-25, 5.3128e-25,\n 1.2220e-25, 7.4982e-25, 7.9437e-24, 3.3568e-24, 4.0523e-24, 3.1825e-24,\n 5.4696e-24, 2.9433e-24, 1.0068e-24, 6.9049e-26, 9.1277e-24, 4.1681e-25,\n 4.4444e-25, 3.8093e-25, 2.5773e-26, 8.1305e-24, 1.6118e-25, 1.3427e-23,\n 4.2728e-24, 2.4170e-24, 3.1370e-24, 9.0077e-25, 8.4739e-24, 5.3393e-25,\n 1.5387e-25, 2.6969e-25, 6.9583e-24, 2.1067e-24, 2.7539e-24, 2.1372e-24,\n 4.8027e-24, 5.0007e-24, 1.0221e-23, 4.3993e-24, 3.0672e-24, 1.9325e-23,\n 1.3049e-24, 2.5972e-24, 1.1342e-25, 2.0141e-26, 1.8337e-24, 3.8865e-24,\n 4.5412e-24, 2.6874e-24, 1.0132e-23, 5.7544e-25], device='cuda:0')"
 },
 "37": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([2.1609e-28, 7.3222e-28, 5.5805e-27, 1.3555e-26, 3.2705e-26, 1.5366e-26,\n 5.8270e-28, 1.4302e-26, 7.4178e-28, 2.9093e-28, 4.1378e-28, 4.9583e-28,\n 1.2686e-26, 7.0952e-28, 1.2874e-26, 1.6731e-26, 2.5063e-28, 5.8128e-27,\n 1.0760e-25, 7.1790e-26, 3.7396e-27, 3.8867e-28, 9.9340e-27, 1.8510e-26,\n 4.8786e-29, 6.0631e-26, 3.6711e-27, 3.3950e-26, 6.2474e-28, 2.6052e-28,\n 2.1585e-27, 1.1195e-28, 6.4263e-26, 6.5698e-27, 2.4371e-27, 7.3998e-28,\n 8.9673e-28, 1.9886e-28, 2.6883e-25, 3.2372e-27, 1.6325e-27, 5.0634e-26,\n 2.2008e-28, 6.7655e-26, 4.0909e-28, 1.3762e-27, 1.2159e-26, 1.4697e-28,\n 1.4824e-27, 8.9882e-26, 2.8789e-27, 6.0421e-29, 3.9486e-27, 9.2165e-27,\n 7.0771e-26, 9.2542e-28, 8.5828e-28, 4.3195e-27, 2.5723e-25, 4.6841e-27,\n 1.7826e-26, 2.1324e-27, 1.4494e-26, 2.4873e-27, 3.0030e-27, 1.4562e-28,\n 8.3750e-26, 1.7135e-25, 1.3880e-26, 1.7552e-26, 7.0575e-28, 4.0665e-27,\n 5.5470e-26, 3.3774e-27, 4.5108e-28, 2.2899e-28, 1.2369e-27, 1.1257e-27,\n 2.1819e-28, 1.9141e-26, 9.3401e-27, 3.6693e-27, 2.5586e-28, 1.1559e-28,\n 2.2388e-28, 4.4735e-26, 3.9679e-28, 2.1350e-28, 7.6215e-28, 4.8047e-26,\n 1.2916e-26, 1.0113e-25, 2.5955e-28, 8.0322e-26, 3.9659e-26, 8.1796e-28,\n 2.4364e-27, 2.3298e-28, 3.2555e-27, 1.8277e-26, 2.6957e-27, 4.9072e-28,\n 4.8691e-27, 6.9842e-29, 2.9311e-27, 1.7899e-26, 4.8547e-27, 8.1983e-28,\n 6.8619e-28, 1.6994e-27, 2.0154e-26, 1.1807e-26, 5.9675e-29, 2.1283e-28,\n 3.5910e-27, 3.2561e-27, 1.3702e-27, 2.8997e-27, 7.8912e-28, 3.7226e-27,\n 3.5681e-26, 4.5037e-26, 8.0493e-28, 2.0672e-27, 3.0472e-28, 7.0404e-26,\n 3.6204e-27, 1.4689e-27, 3.4602e-26, 8.4472e-26, 9.0638e-27, 2.0638e-26,\n 3.7995e-28, 3.3355e-26, 1.9908e-26, 3.3170e-28, 3.0040e-26, 4.2373e-26,\n 5.2420e-27, 3.1981e-28, 1.5235e-26, 2.0721e-28, 1.0298e-27, 4.4660e-27,\n 3.7057e-28, 1.4978e-26, 4.1956e-26, 2.1026e-27, 6.1539e-26, 2.8753e-26,\n 4.0461e-26, 4.4160e-28, 1.7936e-26, 9.2521e-27, 1.3832e-25, 7.1782e-28,\n 2.2146e-27, 1.5868e-27, 3.0037e-27, 1.7784e-28, 4.1177e-26, 1.5736e-28,\n 3.4340e-28, 6.5900e-26, 3.1822e-28, 1.0997e-26, 8.8687e-28, 4.6942e-27,\n 1.1466e-26, 1.7068e-28, 5.3137e-26, 1.7473e-27, 8.7419e-27, 2.7143e-26,\n 9.5126e-27, 1.3714e-27, 1.8478e-27, 1.1748e-28, 1.4044e-25, 3.7568e-28,\n 3.3717e-27, 2.2628e-26, 6.0573e-27, 8.8997e-27, 7.4919e-27, 6.0780e-29,\n 3.6770e-26, 5.5544e-27, 5.3579e-27, 1.3124e-26, 9.3028e-28, 2.6733e-26,\n 2.4446e-26, 1.5082e-26, 6.8402e-27, 5.0200e-26, 6.3111e-28, 1.3213e-26,\n 3.3531e-27, 2.2423e-26, 7.1900e-28, 3.1533e-27, 1.3697e-28, 5.8706e-27,\n 4.0677e-27, 8.2066e-27, 2.3254e-28, 2.6135e-27, 9.6576e-29, 1.3115e-28,\n 2.4251e-29, 2.7396e-27, 7.9380e-27, 1.6796e-27, 1.0959e-26, 6.6660e-27,\n 6.5172e-27, 1.3635e-26, 4.6404e-28, 2.5593e-28, 4.0970e-27, 6.5189e-28,\n 4.2806e-28, 2.4492e-28, 1.9619e-27, 6.6696e-27, 8.2381e-29, 4.0267e-26,\n 8.5656e-27, 3.3657e-27, 1.0943e-26, 7.5864e-28, 5.1024e-26, 2.3465e-27,\n 3.9422e-28, 2.0856e-28, 5.4022e-26, 8.2673e-28, 6.1207e-27, 3.2963e-26,\n 1.2631e-26, 1.5263e-26, 3.0173e-26, 1.3379e-26, 3.3539e-27, 8.5795e-26,\n 2.6309e-27, 2.0185e-27, 5.1469e-29, 7.3582e-29, 1.2109e-27, 4.8505e-27,\n 1.1530e-26, 1.8174e-27, 2.4307e-26, 1.5527e-27], device='cuda:0')"
 },
 "38": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([7.4294e-28, 1.2223e-27, 1.1839e-26, 9.7546e-27, 3.2960e-26, 2.4072e-26,\n 1.3649e-27, 4.6431e-26, 2.3938e-28, 2.1173e-27, 9.6818e-28, 3.5622e-27,\n 1.6877e-26, 5.0202e-27, 1.5946e-26, 2.4736e-26, 4.1387e-29, 5.7126e-27,\n 6.1975e-26, 3.8836e-26, 1.3617e-26, 1.7018e-27, 9.2182e-27, 1.4666e-26,\n 1.7379e-28, 7.1144e-26, 1.6733e-26, 2.9394e-26, 1.8604e-27, 7.6074e-28,\n 4.1653e-27, 2.2462e-29, 3.5774e-26, 9.8677e-27, 9.0696e-28, 1.5545e-27,\n 3.5215e-27, 3.3617e-27, 9.2799e-26, 2.4378e-27, 5.3168e-27, 2.0808e-26,\n 3.6095e-28, 7.6795e-26, 3.6316e-27, 5.5886e-27, 2.5352e-26, 7.8724e-29,\n 3.2350e-27, 5.2161e-26, 7.3881e-27, 3.4910e-27, 5.3859e-27, 3.2069e-26,\n 3.8673e-26, 4.4455e-27, 3.0046e-28, 2.5713e-26, 1.3030e-25, 1.2197e-26,\n 6.6343e-27, 6.8431e-27, 2.8702e-26, 9.3943e-27, 4.3675e-27, 1.3551e-28,\n 3.6043e-26, 7.7608e-26, 1.4444e-26, 1.2118e-26, 4.0485e-27, 8.7802e-27,\n 2.7460e-26, 5.8223e-27, 2.7019e-27, 7.5093e-28, 4.2014e-27, 2.0878e-27,\n 2.2997e-28, 1.0140e-26, 6.4548e-27, 6.5756e-27, 1.0099e-27, 5.4990e-28,\n 4.9332e-28, 3.6053e-26, 1.4961e-27, 1.0558e-27, 3.3834e-27, 3.4941e-26,\n 1.1793e-26, 4.0756e-26, 1.6767e-27, 3.9500e-26, 3.0443e-26, 7.9673e-27,\n 7.6930e-27, 1.2463e-27, 2.4562e-27, 1.8100e-26, 3.0257e-27, 1.1707e-27,\n 7.4263e-27, 1.3579e-27, 7.6642e-27, 1.5089e-26, 9.6549e-27, 4.9834e-27,\n 1.4774e-27, 3.2060e-27, 2.0612e-26, 9.2998e-27, 3.7046e-28, 7.2354e-28,\n 1.0689e-26, 7.6320e-27, 3.9345e-27, 9.6371e-27, 1.0052e-26, 8.8930e-27,\n 2.4337e-26, 4.0965e-26, 3.0582e-27, 9.5725e-28, 5.1109e-28, 5.1410e-26,\n 6.7174e-27, 6.8237e-28, 3.2437e-26, 1.8499e-26, 2.8641e-26, 1.7833e-26,\n 5.6554e-27, 1.4348e-26, 2.5653e-26, 2.2831e-27, 3.0983e-26, 2.7694e-26,\n 9.4710e-27, 7.0898e-29, 3.3158e-26, 8.5585e-29, 3.3374e-27, 9.9134e-27,\n 3.9428e-28, 1.6691e-26, 2.0432e-26, 6.4268e-27, 1.9714e-26, 2.6429e-26,\n 3.7796e-26, 5.2953e-28, 7.5707e-27, 2.2946e-26, 6.3086e-26, 2.4495e-27,\n 1.2511e-27, 8.3138e-27, 1.0952e-26, 7.0309e-29, 4.4889e-26, 1.4757e-27,\n 4.6887e-28, 2.6303e-26, 2.9372e-28, 2.7995e-26, 1.2063e-26, 1.8535e-26,\n 1.1867e-26, 2.4228e-29, 6.1239e-26, 2.8617e-27, 1.4887e-26, 2.7494e-26,\n 1.6459e-26, 4.0646e-27, 4.8176e-27, 1.6000e-28, 9.0280e-26, 1.1618e-28,\n 2.1820e-27, 1.4591e-26, 1.3579e-26, 8.2079e-27, 9.8255e-27, 4.0414e-28,\n 2.3741e-26, 1.2043e-26, 3.0219e-27, 8.7789e-27, 3.1648e-27, 1.6854e-26,\n 1.5558e-26, 1.4633e-26, 1.8933e-26, 1.8797e-26, 1.8320e-28, 2.2679e-26,\n 1.2718e-26, 3.7171e-26, 9.2914e-28, 2.2554e-27, 2.8211e-27, 3.3764e-27,\n 1.1601e-26, 1.9935e-26, 1.4443e-27, 7.8397e-27, 3.6468e-28, 3.7758e-27,\n 1.9509e-28, 9.8784e-27, 2.7111e-26, 5.6921e-27, 2.0972e-26, 8.0749e-27,\n 1.7703e-26, 2.8845e-26, 4.2202e-27, 4.2242e-28, 2.1099e-26, 4.0406e-27,\n 8.8469e-27, 1.9377e-27, 2.0485e-28, 2.6693e-26, 1.3412e-27, 2.8146e-26,\n 1.2750e-26, 1.0907e-26, 9.5789e-27, 2.1584e-27, 4.3839e-26, 2.4566e-27,\n 2.3836e-27, 7.1328e-29, 3.1378e-26, 5.6300e-27, 1.2285e-26, 1.1370e-26,\n 2.1536e-26, 1.1204e-26, 1.9019e-26, 1.5606e-26, 7.3456e-27, 6.7038e-26,\n 7.0961e-27, 9.5951e-27, 1.6274e-29, 2.4975e-28, 4.5026e-27, 6.4700e-27,\n 1.9814e-26, 3.7918e-27, 2.7378e-26, 3.3657e-27], device='cuda:0')"
 },
 "39": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[ 5.6052e-45, -5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, -5.6052e-45],\n [ 5.6052e-45, -5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n ...,\n [-5.6052e-45, -5.6052e-45, 5.6052e-45, ..., -5.6052e-45,\n -5.6052e-45, -5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45]], device='cuda:0')",
 "exp_avg_sq": "tensor([[2.4865e-28, 7.0394e-28, 1.2916e-27, ..., 4.3904e-28, 1.0351e-28,\n 6.0176e-28],\n [2.9878e-27, 3.9016e-27, 7.5604e-28, ..., 1.7622e-27, 4.1266e-28,\n 2.4488e-27],\n [9.3755e-27, 3.1820e-27, 6.6249e-27, ..., 1.4222e-27, 8.9713e-28,\n 1.6270e-27],\n ...,\n [2.2874e-26, 1.8165e-27, 3.2772e-27, ..., 3.2818e-27, 1.1997e-27,\n 3.5638e-27],\n [8.4151e-28, 1.7919e-27, 3.3678e-27, ..., 4.3176e-28, 1.3084e-28,\n 2.9969e-28],\n [4.1349e-28, 1.9558e-28, 1.2900e-27, ..., 2.1222e-28, 2.6109e-28,\n 1.1575e-28]], device='cuda:0')"
 },
 "40": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([1.9257e-24, 2.7213e-24, 4.9168e-24, 5.8237e-24, 3.3710e-24, 5.2179e-24,\n 6.8122e-25, 7.9733e-25, 2.2644e-23, 1.2845e-24, 3.4909e-24, 6.4925e-24,\n 2.6018e-26, 1.4799e-23, 3.0547e-25, 2.8694e-24, 4.4527e-24, 9.1228e-25,\n 4.3791e-23, 1.0958e-23, 3.8999e-23, 9.6283e-25, 9.7298e-25, 6.3045e-24,\n 1.2484e-24, 4.1401e-24, 1.7372e-23, 1.4892e-23, 8.4410e-25, 6.2976e-24,\n 3.8871e-24, 6.1732e-24, 7.7126e-23, 6.0451e-25, 1.8123e-23, 4.7298e-24,\n 5.0935e-24, 1.5311e-24, 6.7373e-23, 4.2026e-24, 1.6795e-25, 9.3758e-24,\n 8.3901e-26, 2.0207e-24, 8.4757e-24, 2.7645e-25, 9.3774e-24, 5.0355e-24,\n 5.5336e-25, 1.5423e-23, 4.5214e-25, 2.0564e-23, 7.3352e-24, 5.8278e-26,\n 3.1796e-24, 6.9983e-24, 6.6834e-24, 1.3132e-23, 8.2921e-23, 5.8170e-24,\n 1.5522e-23, 5.1299e-24, 2.1597e-23, 2.2502e-24, 1.2986e-24, 4.8239e-25,\n 1.2956e-23, 6.1032e-25, 1.0101e-25, 1.7222e-23, 2.2952e-25, 7.3793e-25,\n 6.0662e-23, 7.8919e-24, 2.0382e-24, 1.7285e-24, 6.1433e-24, 6.9117e-24,\n 9.4318e-24, 4.9362e-25, 5.7306e-24, 8.0652e-25, 4.9505e-24, 9.8747e-24,\n 2.5459e-24, 2.4757e-24, 4.4450e-24, 3.5773e-24, 2.8286e-24, 7.1536e-24,\n 1.8750e-24, 1.1737e-24, 2.8400e-25, 1.8450e-23, 5.9990e-24, 5.2124e-24,\n 4.2395e-25, 1.4339e-24, 1.2356e-25, 2.1426e-25, 5.1383e-25, 1.4431e-25,\n 3.3017e-25, 4.1360e-24, 6.8870e-24, 8.5791e-24, 9.6041e-24, 4.1197e-25,\n 3.1946e-24, 3.0832e-24, 1.3109e-23, 6.2823e-24, 5.0433e-24, 9.1643e-24,\n 2.3017e-23, 3.2576e-24, 8.6751e-24, 6.8622e-24, 5.0129e-24, 2.5650e-25,\n 1.6151e-23, 4.4435e-24, 6.3619e-24, 2.3527e-25, 6.9621e-25, 7.5316e-24,\n 3.7861e-24, 2.0458e-24, 1.6224e-23, 3.0340e-23, 3.1095e-24, 8.5812e-24,\n 4.1368e-24, 1.7687e-23, 1.2909e-25, 1.2344e-24, 1.5649e-24, 2.2577e-23,\n 1.3904e-24, 1.9341e-25, 4.8140e-23, 3.5545e-24, 7.3560e-25, 1.2709e-24,\n 5.3463e-25, 2.1633e-23, 1.3775e-23, 1.3316e-23, 1.0885e-23, 3.4944e-24,\n 5.6128e-24, 4.9188e-24, 3.3972e-25, 1.6293e-23, 8.9789e-24, 4.7038e-24,\n 3.5578e-26, 1.4545e-25, 5.5473e-23, 2.7848e-25, 1.7086e-23, 2.7621e-24,\n 4.7168e-24, 7.0729e-25, 6.9567e-24, 3.2522e-23, 5.8145e-23, 1.3596e-23,\n 3.5642e-24, 5.7794e-25, 2.9004e-23, 7.5839e-24, 4.0043e-25, 1.5731e-24,\n 1.9346e-25, 4.6627e-25, 2.5962e-24, 1.1641e-23, 4.1076e-23, 1.6821e-24,\n 5.2017e-24, 3.4352e-24, 3.2171e-25, 1.2157e-24, 7.9731e-24, 1.0162e-23,\n 9.1842e-25, 1.3015e-24, 7.1322e-24, 8.7534e-24, 6.0537e-25, 1.3812e-24,\n 2.7502e-23, 4.6478e-24, 5.8361e-25, 6.4569e-25, 3.6052e-24, 1.3551e-23,\n 1.5169e-23, 2.8910e-23, 1.1333e-23, 4.1926e-25, 3.8244e-25, 1.2213e-23,\n 1.0557e-23, 4.4075e-24, 1.8842e-26, 2.2230e-25, 1.0321e-24, 2.8911e-24,\n 4.5958e-25, 1.5888e-25, 9.1748e-24, 2.2818e-23, 2.0188e-25, 9.2356e-24,\n 4.6270e-25, 3.5199e-25, 1.0570e-23, 7.5392e-24, 3.0804e-25, 7.5116e-24,\n 6.1128e-26, 1.3184e-23, 4.4145e-23, 1.5724e-24, 1.4894e-24, 7.2185e-24,\n 1.1284e-23, 1.2919e-23, 9.8624e-24, 9.4730e-25, 2.9308e-24, 3.4164e-24,\n 1.8114e-24, 8.1201e-25, 2.2990e-23, 1.2980e-23, 8.2391e-24, 6.3774e-25,\n 4.6406e-25, 6.1832e-24, 6.4322e-24, 1.7258e-25, 1.6025e-24, 3.3213e-24,\n 2.0049e-23, 4.1668e-25, 6.1884e-24, 5.4365e-24, 2.2216e-24, 6.8750e-25,\n 1.1977e-25, 2.1925e-23, 2.0250e-24, 6.6894e-25], device='cuda:0')"
 },
 "41": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([1.4972e-27, 1.7984e-27, 5.0022e-27, 5.8039e-27, 4.8110e-27, 2.4454e-26,\n 3.9893e-27, 8.6092e-28, 7.1339e-26, 7.7228e-28, 7.0585e-28, 8.1454e-27,\n 1.1045e-28, 6.8641e-27, 6.9380e-28, 6.7632e-27, 3.2058e-26, 7.3550e-28,\n 4.3879e-25, 1.0342e-25, 4.7986e-25, 4.2607e-27, 1.4890e-27, 1.7710e-25,\n 2.3585e-27, 2.4195e-27, 1.0523e-25, 3.7583e-27, 2.4215e-28, 3.6717e-26,\n 1.8836e-27, 2.5608e-27, 1.4493e-25, 3.7929e-27, 8.2612e-26, 1.3402e-26,\n 1.4891e-27, 3.1255e-28, 3.2343e-25, 5.3524e-27, 5.4683e-28, 8.7183e-26,\n 9.5898e-29, 1.2438e-27, 3.8228e-26, 9.7923e-29, 1.8457e-26, 4.4654e-27,\n 7.1413e-28, 2.1785e-26, 2.7020e-28, 3.5221e-26, 1.0152e-26, 1.5172e-27,\n 1.0839e-27, 1.9827e-27, 3.0836e-26, 1.1390e-26, 9.2992e-26, 1.1434e-26,\n 7.3551e-27, 8.5381e-28, 7.0850e-26, 5.2458e-27, 5.0459e-28, 1.0509e-27,\n 2.5853e-25, 5.2152e-27, 3.6102e-28, 9.8754e-26, 5.9380e-28, 3.3157e-28,\n 7.9686e-26, 3.6696e-27, 1.0310e-27, 2.9598e-28, 8.1779e-27, 7.8207e-27,\n 6.3392e-27, 3.0909e-28, 1.7375e-27, 7.8344e-28, 5.6288e-26, 9.0158e-27,\n 9.7386e-28, 6.3531e-28, 1.1867e-27, 3.9867e-27, 4.0016e-27, 1.5128e-26,\n 2.7969e-27, 9.0148e-27, 1.6996e-28, 1.7580e-26, 8.5924e-28, 1.6637e-26,\n 9.1373e-28, 1.4405e-26, 9.0773e-29, 3.0834e-28, 1.7139e-28, 8.6342e-28,\n 9.2871e-29, 3.3460e-27, 1.1785e-26, 1.9326e-27, 4.5870e-26, 7.7653e-28,\n 3.9826e-27, 3.7852e-27, 1.3039e-25, 5.6093e-26, 2.1968e-26, 1.5299e-26,\n 5.4476e-26, 3.2504e-26, 1.4912e-25, 4.2089e-27, 2.4006e-27, 5.1208e-28,\n 1.1395e-26, 1.0102e-27, 1.1661e-26, 1.6693e-27, 4.1311e-28, 9.5735e-27,\n 2.0992e-27, 8.8949e-28, 3.3310e-26, 5.7569e-26, 3.8805e-27, 1.9742e-28,\n 2.3225e-27, 1.8611e-25, 6.1921e-28, 4.6594e-28, 9.5362e-28, 1.6894e-25,\n 5.6295e-28, 7.8335e-29, 2.5201e-25, 3.2747e-27, 3.7872e-28, 8.0515e-28,\n 2.6558e-27, 2.8841e-26, 9.7914e-27, 4.6848e-26, 7.0097e-28, 6.3079e-29,\n 4.9793e-27, 1.5291e-26, 2.8293e-28, 1.0716e-25, 7.8200e-26, 6.0577e-27,\n 2.3490e-28, 2.7369e-28, 1.3242e-25, 1.6331e-28, 1.5573e-25, 1.9497e-28,\n 2.6294e-27, 4.0058e-28, 3.5179e-27, 5.6714e-26, 5.0627e-26, 4.1119e-27,\n 2.1390e-27, 2.5051e-28, 3.5132e-26, 4.1379e-26, 4.6194e-28, 2.2835e-26,\n 3.8185e-28, 5.5835e-28, 1.3616e-27, 1.1802e-27, 6.4493e-25, 9.2545e-27,\n 9.0680e-27, 1.1730e-27, 1.2575e-28, 2.6289e-29, 6.6435e-27, 2.0037e-26,\n 2.0915e-27, 8.3500e-28, 5.8449e-27, 2.5213e-27, 7.7646e-28, 2.5918e-27,\n 5.0808e-26, 2.5947e-27, 7.2393e-28, 2.0535e-28, 7.9618e-27, 1.7236e-26,\n 3.4338e-26, 3.0207e-25, 5.5202e-27, 1.6982e-28, 4.3506e-28, 1.2283e-27,\n 5.1211e-27, 5.9231e-27, 1.0445e-28, 3.6278e-28, 1.8770e-27, 9.2825e-28,\n 1.0916e-28, 9.8505e-28, 2.2143e-26, 2.1349e-26, 2.3863e-28, 7.1980e-27,\n 5.9004e-28, 1.4365e-27, 1.3301e-26, 8.3751e-27, 6.7872e-28, 1.2628e-26,\n 1.9082e-28, 1.2171e-26, 1.5209e-25, 1.3672e-28, 3.3080e-27, 2.7483e-27,\n 7.0694e-27, 1.2774e-25, 7.4533e-26, 9.0192e-28, 3.0284e-26, 1.3313e-27,\n 5.3631e-26, 1.2951e-28, 1.6073e-25, 4.2887e-27, 9.7862e-27, 2.1163e-26,\n 2.5510e-27, 1.9311e-26, 5.7051e-27, 2.4286e-28, 3.4900e-27, 1.3470e-27,\n 1.8044e-26, 6.6310e-28, 1.0850e-26, 1.7080e-26, 4.2952e-28, 9.9662e-28,\n 4.4089e-28, 1.5374e-26, 1.8260e-27, 6.4582e-28], device='cuda:0')"
 },
 "42": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([9.8798e-27, 8.7522e-27, 1.1381e-26, 1.5459e-26, 1.4953e-26, 1.6435e-26,\n 3.8094e-27, 1.2399e-27, 7.7829e-26, 1.6953e-27, 2.9867e-27, 2.8455e-26,\n 1.8015e-28, 7.9132e-27, 7.2507e-28, 1.5573e-26, 2.4772e-26, 1.7547e-27,\n 1.7256e-25, 7.0762e-26, 1.6476e-25, 8.6212e-27, 3.5066e-27, 5.2172e-26,\n 4.8036e-27, 1.5841e-26, 6.3053e-26, 7.3470e-27, 9.0640e-28, 3.6079e-26,\n 1.3326e-26, 5.9591e-27, 5.1853e-26, 5.7336e-27, 5.0669e-26, 8.3748e-27,\n 8.6108e-27, 3.8292e-27, 2.0488e-25, 1.4730e-26, 2.4311e-28, 5.6155e-26,\n 1.0428e-27, 9.2077e-27, 2.8235e-26, 6.3218e-28, 1.3110e-26, 8.9170e-27,\n 2.2029e-27, 4.5296e-26, 3.0977e-28, 2.3931e-26, 2.8588e-26, 1.9041e-27,\n 2.1678e-27, 3.4617e-27, 3.9165e-26, 4.4581e-26, 9.5692e-26, 1.9382e-26,\n 8.1655e-27, 1.5432e-27, 8.9877e-26, 1.1885e-26, 8.3402e-28, 3.3439e-27,\n 9.0390e-26, 5.5929e-27, 9.4981e-30, 7.6737e-26, 1.5814e-27, 3.5519e-28,\n 4.4508e-26, 8.0502e-27, 2.8476e-27, 7.9112e-27, 2.6892e-26, 9.6983e-27,\n 2.5267e-26, 3.1600e-27, 1.2070e-27, 1.3527e-27, 3.6555e-26, 1.1082e-26,\n 1.0912e-27, 7.1212e-27, 6.2712e-28, 1.0533e-26, 1.1070e-26, 2.3401e-26,\n 1.2558e-26, 7.3058e-27, 2.3279e-28, 3.0068e-26, 4.0235e-27, 2.9293e-26,\n 4.4132e-29, 7.0506e-27, 3.9327e-28, 1.0939e-28, 1.4706e-27, 1.1355e-27,\n 1.7577e-28, 3.4351e-27, 9.0043e-27, 3.1536e-27, 5.9503e-26, 6.6023e-28,\n 1.8587e-27, 2.1899e-27, 5.1004e-26, 4.2488e-26, 3.5100e-26, 4.3031e-26,\n 2.3687e-26, 2.7248e-26, 7.0886e-26, 2.7035e-26, 1.9650e-26, 2.1914e-28,\n 1.6137e-26, 1.5051e-26, 1.6333e-26, 3.5672e-28, 3.9787e-27, 1.9687e-26,\n 5.7468e-27, 5.9003e-27, 3.6121e-26, 2.4916e-26, 1.0829e-26, 1.1612e-26,\n 1.1443e-26, 7.1767e-26, 1.0355e-28, 3.4375e-27, 1.3459e-26, 6.3495e-26,\n 8.0471e-27, 3.0362e-27, 1.7535e-25, 8.3112e-27, 1.5127e-28, 3.8650e-27,\n 3.7833e-27, 2.9257e-26, 2.1088e-26, 1.9972e-26, 9.0873e-27, 2.6705e-27,\n 6.5323e-27, 1.4924e-26, 6.8976e-28, 8.0556e-26, 6.2715e-26, 9.0337e-27,\n 4.7377e-28, 2.5625e-28, 1.5327e-25, 1.0105e-27, 7.9100e-26, 1.4089e-27,\n 7.4216e-27, 4.1805e-28, 7.6075e-27, 1.0359e-25, 1.5236e-25, 3.4181e-26,\n 3.3298e-27, 9.5335e-28, 8.5430e-26, 4.8122e-26, 2.3065e-27, 2.3491e-26,\n 1.1235e-27, 3.1005e-28, 3.4490e-27, 6.2005e-27, 1.6984e-25, 9.4158e-27,\n 1.0382e-26, 2.3107e-27, 1.8037e-27, 5.5710e-28, 3.7785e-26, 2.4189e-26,\n 7.9837e-27, 4.7411e-27, 7.9516e-27, 7.3467e-27, 3.3268e-27, 4.7022e-27,\n 2.4504e-26, 6.2764e-27, 1.0507e-27, 2.3413e-27, 1.9627e-26, 2.8943e-26,\n 3.9058e-26, 1.1934e-25, 7.4244e-27, 3.3614e-28, 1.1946e-27, 5.0578e-28,\n 1.7377e-26, 9.9991e-27, 5.0731e-29, 3.4921e-27, 3.6292e-27, 4.9751e-27,\n 4.9708e-28, 1.3523e-27, 2.9005e-26, 2.8496e-26, 2.1897e-27, 1.1002e-26,\n 6.1116e-28, 5.3610e-27, 3.5410e-26, 2.9259e-26, 5.2173e-28, 3.3714e-26,\n 2.2728e-28, 4.8005e-26, 1.1907e-25, 4.6272e-27, 4.9887e-27, 4.1616e-27,\n 1.6596e-26, 7.5201e-26, 3.4752e-26, 2.1915e-27, 3.0495e-26, 1.1105e-26,\n 2.2790e-26, 9.4581e-28, 1.1756e-25, 3.6067e-26, 2.7255e-26, 1.0278e-26,\n 1.4865e-27, 2.7424e-26, 1.0545e-26, 5.5333e-28, 3.1738e-27, 7.8515e-27,\n 1.0332e-26, 6.5442e-28, 2.3462e-26, 1.5778e-26, 3.4243e-27, 9.6227e-28,\n 3.1280e-28, 2.4902e-26, 4.6971e-27, 4.9298e-27], device='cuda:0')"
 },
 "43": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[-5.6052e-45, -5.6052e-45, -5.6052e-45, ..., -5.6052e-45,\n -5.6052e-45, -5.6052e-45],\n [ 5.6052e-45, -5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, -5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, -5.6052e-45],\n ...,\n [-5.6052e-45, -5.6052e-45, 5.6052e-45, ..., -5.6052e-45,\n -5.6052e-45, -5.6052e-45],\n [-5.6052e-45, -5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, -5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45]], device='cuda:0')",
 "exp_avg_sq": "tensor([[8.4593e-28, 3.0367e-28, 9.9046e-29, ..., 2.2538e-27, 1.6113e-28,\n 9.3995e-28],\n [8.6847e-28, 2.9955e-27, 6.2474e-28, ..., 1.3436e-28, 3.6153e-28,\n 2.9765e-27],\n [1.2151e-26, 2.5103e-27, 1.0890e-26, ..., 1.2793e-27, 1.0530e-27,\n 8.1278e-29],\n ...,\n [6.2384e-27, 6.7576e-28, 2.8093e-27, ..., 4.9120e-27, 3.7041e-28,\n 9.1643e-28],\n [5.5503e-27, 7.8243e-27, 1.3470e-26, ..., 1.4976e-27, 1.6343e-28,\n 1.3420e-27],\n [1.1263e-27, 3.9889e-28, 1.3565e-27, ..., 3.2330e-28, 1.3616e-28,\n 4.3596e-28]], device='cuda:0')"
 },
 "44": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([2.5048e-24, 1.4892e-24, 7.3299e-24, 7.2566e-24, 5.6763e-25, 4.2343e-24,\n 8.1004e-25, 1.3485e-24, 2.8789e-24, 1.0766e-24, 1.0326e-23, 2.3424e-23,\n 1.3851e-23, 4.7663e-26, 3.3223e-25, 3.3020e-24, 2.8666e-24, 1.7794e-24,\n 1.5377e-23, 2.6972e-25, 2.6980e-23, 1.4473e-24, 7.4578e-24, 5.0999e-24,\n 1.3824e-24, 9.2179e-24, 6.5209e-25, 4.5811e-23, 8.8557e-24, 1.1332e-24,\n 1.8485e-23, 6.5230e-24, 4.1003e-24, 3.5547e-25, 8.0072e-24, 5.1727e-24,\n 3.6188e-25, 3.0449e-24, 1.3846e-23, 4.5012e-24, 2.3056e-24, 1.6208e-24,\n 1.8812e-24, 7.5504e-24, 9.3270e-25, 4.7385e-25, 2.3003e-24, 3.6017e-25,\n 1.3807e-23, 2.1985e-23, 1.6221e-23, 1.2271e-25, 4.5686e-24, 7.4867e-24,\n 1.1403e-23, 5.4920e-24, 1.3562e-23, 1.8525e-25, 2.1433e-24, 6.3274e-25,\n 5.3266e-24, 1.1710e-23, 1.4709e-25, 6.0884e-24, 1.8057e-25, 1.1321e-24,\n 8.5530e-24, 9.9142e-24, 4.0813e-24, 6.0191e-25, 2.1174e-24, 7.4916e-26,\n 3.7283e-23, 1.9240e-23, 9.9347e-26, 1.1033e-24, 1.0686e-24, 4.4015e-24,\n 4.3206e-25, 3.7661e-24, 5.0338e-26, 2.3582e-24, 8.0426e-25, 5.4205e-24,\n 3.0121e-24, 6.0006e-24, 4.1713e-24, 7.5622e-24, 1.1087e-23, 1.0805e-23,\n 4.0878e-25, 1.5035e-24, 5.8364e-24, 6.4939e-24, 9.3193e-24, 2.3387e-24,\n 3.7472e-24, 1.4722e-24, 2.7971e-24, 9.3815e-25, 1.7353e-24, 7.2503e-25,\n 4.5244e-25, 3.6047e-24, 1.6796e-24, 3.3068e-24, 2.1440e-24, 3.3086e-24,\n 3.9032e-24, 1.5478e-24, 5.8175e-24, 1.1495e-24, 6.1275e-24, 2.6768e-24,\n 5.2556e-25, 2.2599e-24, 2.5020e-24, 1.5046e-23, 4.8519e-24, 4.1418e-25,\n 5.5913e-24, 2.6297e-23, 5.5044e-24, 4.9384e-23, 5.5065e-24, 1.9521e-23,\n 4.3127e-25, 1.6230e-23, 1.3897e-24, 2.4387e-24, 8.6350e-25, 1.6190e-23,\n 2.5708e-24, 5.7515e-24, 8.2691e-26, 4.4922e-25, 8.7680e-24, 1.4696e-24,\n 6.7710e-24, 4.1952e-24, 3.8694e-24, 1.0660e-23, 5.3985e-25, 1.1738e-23,\n 1.6330e-24, 1.2073e-23, 7.1877e-24, 1.1900e-23, 2.6670e-24, 5.4259e-24,\n 2.3460e-23, 3.7465e-24, 4.2886e-24, 1.9856e-23, 7.3470e-26, 6.8697e-24,\n 1.0868e-24, 2.6130e-24, 1.6578e-24, 4.9430e-24, 1.7641e-23, 6.2214e-24,\n 9.4273e-24, 3.6307e-25, 3.2807e-24, 1.2530e-24, 2.1474e-23, 3.7769e-25,\n 1.8665e-23, 1.8816e-24, 9.0541e-24, 4.9275e-24, 3.3185e-24, 3.2767e-25,\n 5.5536e-24, 4.8099e-23, 4.3325e-25, 5.5860e-24, 1.5686e-23, 4.3609e-26,\n 1.9162e-25, 7.5420e-24, 2.3965e-25, 1.1577e-23, 1.4227e-23, 1.4774e-25,\n 7.4218e-25, 2.0604e-24, 1.5919e-23, 9.7740e-24, 7.5698e-25, 3.3251e-24,\n 2.8965e-25, 8.3564e-25, 4.1824e-24, 1.2523e-24, 1.0262e-25, 7.7452e-24,\n 2.6427e-24, 2.2516e-23, 8.7112e-25, 1.0343e-23, 3.3710e-25, 5.6656e-24,\n 8.0253e-24, 1.6682e-23, 1.7773e-25, 2.1359e-24, 1.1801e-24, 1.8406e-24,\n 1.3946e-25, 4.1514e-24, 3.5988e-24, 1.6041e-23, 1.2560e-23, 2.9283e-24,\n 1.1534e-24, 6.4940e-25, 1.8207e-24, 3.7836e-24, 3.6153e-25, 1.4401e-23,\n 4.7883e-25, 4.4175e-24, 9.0227e-24, 3.0233e-24, 4.7295e-25, 1.3486e-23,\n 4.6114e-25, 7.2590e-24, 1.3104e-25, 6.2507e-25, 1.3900e-25, 1.2543e-24,\n 2.1037e-24, 3.4516e-25, 9.9801e-24, 1.0220e-24, 4.3469e-25, 6.5010e-24,\n 2.8966e-24, 2.2836e-24, 1.2034e-24, 1.1639e-24, 1.4372e-24, 2.1067e-23,\n 1.1408e-23, 1.1150e-25, 3.7715e-24, 5.1036e-26, 2.6503e-24, 2.2589e-24,\n 1.4766e-23, 1.2523e-23, 7.1428e-24, 7.3517e-25], device='cuda:0')"
 },
 "45": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([4.5789e-27, 2.6103e-27, 1.1119e-26, 1.9390e-26, 8.1211e-28, 2.7244e-27,\n 3.3496e-27, 8.2585e-28, 4.5933e-28, 3.8051e-28, 5.0407e-27, 6.3342e-26,\n 5.8525e-26, 4.2756e-29, 2.7979e-30, 1.0520e-26, 1.1163e-26, 7.8744e-27,\n 6.0045e-26, 6.6674e-28, 9.0378e-26, 1.2480e-27, 7.7143e-27, 3.8087e-26,\n 5.8416e-27, 1.7882e-26, 6.0046e-28, 8.4071e-26, 1.0474e-26, 1.0396e-27,\n 4.5455e-26, 2.8145e-27, 9.0384e-28, 2.6246e-28, 1.7810e-26, 2.9403e-27,\n 4.3793e-28, 2.6854e-27, 6.1844e-26, 1.5611e-26, 3.8513e-28, 7.1353e-27,\n 1.6291e-27, 1.1067e-26, 4.0915e-28, 1.4408e-27, 9.6457e-28, 5.2352e-28,\n 3.9822e-26, 7.8455e-26, 2.8721e-26, 1.4884e-28, 1.4645e-26, 1.0177e-26,\n 1.0413e-26, 1.0170e-26, 1.9184e-25, 3.7445e-28, 5.1992e-27, 2.2730e-28,\n 6.0551e-27, 2.9727e-26, 1.1532e-27, 2.2356e-26, 1.6334e-29, 9.5791e-29,\n 1.9229e-26, 9.2579e-26, 7.6067e-28, 1.4764e-27, 1.9316e-27, 2.9791e-28,\n 5.6238e-26, 4.7846e-26, 8.2717e-29, 9.6779e-29, 6.0538e-28, 3.1616e-27,\n 9.2341e-28, 2.3738e-26, 2.6850e-28, 3.0285e-27, 2.5113e-27, 1.5332e-27,\n 1.5419e-27, 4.5665e-27, 4.3081e-27, 2.1535e-26, 2.6278e-26, 1.2500e-26,\n 1.8675e-27, 1.4704e-27, 5.2967e-26, 9.2629e-27, 1.3890e-26, 4.6088e-27,\n 2.5278e-27, 2.1810e-27, 4.7381e-27, 1.2111e-28, 8.1344e-27, 7.5334e-28,\n 1.6839e-28, 4.3438e-27, 4.4673e-28, 2.6354e-27, 3.9951e-27, 1.3032e-27,\n 3.5655e-27, 1.6597e-27, 3.2931e-26, 1.3276e-27, 2.2259e-26, 1.5682e-26,\n 7.5196e-29, 1.9428e-26, 1.5328e-26, 1.4305e-26, 2.9952e-27, 2.3467e-28,\n 5.2516e-27, 6.3929e-26, 7.7524e-27, 1.9478e-25, 6.5473e-27, 8.8260e-26,\n 2.6832e-27, 2.9128e-26, 5.1894e-27, 1.4289e-27, 1.2827e-28, 4.3056e-26,\n 9.2404e-28, 2.7949e-26, 4.6944e-28, 1.5431e-28, 3.9260e-26, 4.5688e-27,\n 1.0333e-26, 2.3829e-27, 1.0986e-26, 2.9230e-26, 6.8912e-28, 2.4536e-26,\n 2.0439e-27, 6.8146e-27, 5.1417e-27, 8.0256e-26, 9.9524e-28, 1.3908e-26,\n 3.5198e-26, 4.3038e-27, 4.6781e-27, 6.5049e-26, 9.6032e-28, 5.2660e-27,\n 3.6309e-28, 1.0974e-26, 1.4664e-27, 1.0480e-27, 8.2361e-26, 7.5150e-27,\n 2.1070e-26, 1.2460e-28, 1.9805e-26, 1.4682e-27, 2.0194e-26, 1.6862e-27,\n 2.8874e-26, 3.0130e-28, 5.0805e-27, 1.4558e-27, 2.4524e-27, 2.1480e-28,\n 4.5645e-27, 6.0197e-26, 4.3378e-28, 4.4979e-27, 1.3627e-25, 1.1996e-28,\n 7.8091e-29, 5.4119e-26, 8.0099e-28, 1.2550e-26, 2.3007e-25, 1.5397e-28,\n 1.2434e-27, 5.8132e-27, 1.5084e-26, 7.7285e-27, 4.9192e-28, 2.0096e-26,\n 3.2832e-28, 1.1478e-28, 1.7101e-27, 1.5997e-27, 6.2640e-28, 5.6591e-27,\n 7.9744e-27, 1.7451e-25, 1.3700e-27, 1.1734e-26, 3.8677e-28, 5.7329e-27,\n 6.8204e-27, 3.7649e-26, 2.2035e-28, 5.7682e-27, 1.3027e-27, 3.0672e-28,\n 4.8979e-28, 3.0566e-27, 6.5059e-27, 2.0240e-26, 2.5026e-26, 4.6543e-27,\n 1.0176e-27, 6.6831e-28, 1.4144e-27, 3.4465e-27, 3.0300e-28, 3.6109e-26,\n 3.9027e-27, 4.9303e-27, 7.5368e-27, 8.4545e-28, 4.2271e-28, 9.4059e-27,\n 1.7603e-27, 1.4930e-26, 5.6505e-28, 1.3582e-28, 2.5781e-28, 4.8930e-28,\n 3.4661e-27, 6.9715e-29, 6.0185e-26, 1.6713e-28, 1.0899e-28, 8.8047e-27,\n 2.0753e-27, 7.6655e-28, 1.0979e-27, 8.6557e-27, 9.8688e-29, 6.5688e-26,\n 1.8135e-26, 3.4096e-28, 7.1960e-27, 4.1006e-28, 1.0150e-27, 2.2112e-27,\n 1.6581e-26, 1.0143e-26, 1.1465e-26, 2.6934e-28], device='cuda:0')"
 },
 "46": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([7.4200e-27, 4.5365e-27, 9.6878e-27, 1.3973e-26, 2.3084e-27, 1.0722e-26,\n 2.4592e-27, 2.1309e-27, 1.0252e-26, 2.2598e-27, 1.4677e-26, 5.5110e-26,\n 5.8230e-26, 1.0383e-28, 5.2380e-28, 1.5203e-26, 1.0169e-26, 1.0341e-26,\n 4.6542e-26, 2.3732e-27, 7.1986e-26, 4.7148e-27, 1.2260e-26, 1.2943e-26,\n 1.1050e-26, 3.2230e-26, 2.4148e-27, 6.2292e-26, 1.5541e-26, 1.7087e-27,\n 6.4325e-26, 2.7669e-27, 2.7863e-27, 5.7417e-28, 2.4848e-26, 3.0776e-27,\n 6.8737e-28, 1.1976e-26, 3.6708e-26, 1.8371e-26, 5.0885e-27, 5.7596e-27,\n 3.1181e-27, 2.9820e-26, 3.4608e-27, 4.3238e-27, 4.7152e-27, 1.3939e-27,\n 1.5812e-26, 5.7415e-26, 1.9460e-26, 4.5970e-28, 9.7839e-27, 1.1152e-26,\n 1.8232e-26, 7.0327e-27, 6.7204e-26, 1.4220e-27, 3.5483e-27, 2.1869e-27,\n 8.1352e-27, 1.2869e-26, 2.0675e-28, 1.6886e-26, 5.7960e-29, 5.3063e-27,\n 3.3325e-26, 3.0329e-26, 6.8129e-27, 2.0510e-27, 5.4765e-27, 1.0254e-28,\n 5.8903e-26, 3.4228e-26, 3.2161e-28, 2.6550e-27, 4.9519e-27, 7.9677e-27,\n 7.0009e-28, 9.8774e-27, 2.6641e-29, 5.2251e-27, 4.0758e-27, 8.0065e-27,\n 3.5743e-27, 1.5281e-26, 4.1850e-27, 2.7439e-26, 3.2252e-26, 2.4355e-26,\n 1.0080e-27, 2.0388e-27, 1.4880e-26, 1.1839e-26, 1.3274e-26, 1.0244e-26,\n 3.8727e-27, 4.0207e-27, 2.7977e-27, 1.2030e-27, 1.0652e-26, 3.1558e-28,\n 7.7088e-28, 1.0286e-26, 1.2984e-27, 1.8789e-27, 1.2337e-26, 6.7622e-27,\n 3.1897e-27, 2.1119e-27, 2.6481e-26, 4.9766e-27, 2.5655e-26, 1.4506e-26,\n 1.0618e-27, 1.4880e-26, 1.5955e-26, 2.4454e-26, 1.5212e-26, 7.1389e-28,\n 6.8253e-27, 6.5061e-26, 1.4097e-26, 1.2519e-25, 1.3038e-26, 5.1518e-26,\n 6.4137e-27, 3.3835e-26, 5.0039e-27, 6.5346e-27, 2.2485e-27, 2.4033e-26,\n 4.3684e-27, 1.4421e-26, 8.0714e-29, 2.3685e-27, 1.7582e-26, 5.0132e-27,\n 2.6056e-26, 5.2623e-27, 1.1406e-26, 1.5156e-26, 5.6860e-28, 2.5721e-26,\n 5.6632e-27, 1.4387e-26, 1.3165e-26, 3.3059e-26, 3.1127e-27, 1.5507e-26,\n 2.8340e-26, 7.0752e-27, 4.5162e-27, 6.6210e-26, 3.6692e-28, 6.8232e-27,\n 1.2510e-27, 1.4058e-26, 5.5876e-27, 1.0409e-26, 4.0689e-26, 7.9321e-27,\n 1.8572e-26, 5.1886e-28, 1.7106e-26, 3.2741e-27, 5.5198e-26, 1.2650e-27,\n 1.1016e-26, 2.4954e-27, 2.2488e-26, 1.4524e-26, 1.0186e-26, 1.9044e-27,\n 9.2953e-27, 4.2508e-26, 2.2234e-27, 5.6486e-27, 4.7254e-26, 6.3801e-28,\n 4.7742e-28, 1.2865e-26, 1.9706e-27, 1.2059e-26, 4.9853e-26, 2.6186e-28,\n 4.2097e-27, 1.1179e-26, 2.4964e-26, 6.7323e-27, 1.0381e-27, 1.8196e-26,\n 1.2536e-28, 1.2119e-27, 1.1165e-26, 2.3556e-27, 3.6580e-28, 1.7449e-26,\n 6.9686e-27, 6.2661e-26, 1.1138e-27, 8.3374e-27, 2.7364e-27, 6.2634e-27,\n 1.6632e-26, 5.0162e-26, 1.4678e-28, 8.1622e-27, 1.8600e-27, 4.4346e-27,\n 2.6754e-28, 9.2468e-27, 1.1563e-26, 2.4251e-26, 4.5201e-26, 4.7844e-27,\n 6.8795e-27, 1.0379e-27, 5.3099e-27, 1.2957e-26, 2.1290e-27, 4.2950e-26,\n 4.4464e-27, 9.3973e-27, 2.2100e-26, 8.3574e-27, 1.0830e-27, 2.0813e-26,\n 1.8769e-27, 2.6155e-26, 7.8247e-28, 1.4423e-27, 6.7256e-28, 2.9162e-27,\n 4.7466e-27, 9.0480e-28, 3.9995e-26, 3.0410e-27, 2.3167e-27, 1.1940e-26,\n 1.2604e-26, 2.4385e-27, 1.0911e-27, 5.3134e-27, 2.0910e-27, 6.5555e-26,\n 1.6912e-26, 3.0003e-28, 1.3501e-26, 1.2824e-28, 4.6624e-27, 4.8730e-27,\n 5.0358e-26, 1.5475e-26, 6.9276e-27, 3.8541e-27], device='cuda:0')"
 },
 "47": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[ 5.6052e-45, 5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, -5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n ...,\n [-5.6052e-45, -5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, -5.6052e-45],\n [-5.6052e-45, -5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, -5.6052e-45, ..., -5.6052e-45,\n 5.6052e-45, -5.6052e-45]], device='cuda:0')",
 "exp_avg_sq": "tensor([[1.8337e-27, 1.6172e-27, 1.0015e-27, ..., 1.0688e-27, 4.0788e-29,\n 2.1043e-28],\n [1.6998e-28, 5.3730e-28, 2.6691e-28, ..., 1.9295e-28, 2.2803e-29,\n 1.1080e-28],\n [5.4827e-27, 1.3307e-27, 1.4685e-26, ..., 2.9443e-27, 5.2226e-28,\n 2.8037e-27],\n ...,\n [9.1375e-27, 4.1331e-27, 3.6073e-27, ..., 1.5934e-27, 7.0137e-29,\n 1.5892e-27],\n [3.2402e-26, 5.2426e-27, 1.7065e-26, ..., 4.6552e-27, 1.4235e-27,\n 4.3814e-27],\n [8.3962e-28, 1.0633e-27, 2.9842e-28, ..., 1.4251e-27, 2.0679e-28,\n 9.8165e-29]], device='cuda:0')"
 },
 "48": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([-5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([2.8017e-24, 8.3204e-26, 1.2954e-23, 1.2911e-23, 2.4413e-25, 6.8902e-24,\n 4.3717e-25, 1.9691e-23, 4.1435e-24, 9.7737e-24, 1.9944e-23, 8.5550e-24,\n 2.4179e-24, 6.5682e-25, 1.1660e-24, 4.2088e-24, 1.9285e-24, 1.2654e-23,\n 1.5901e-25, 4.7570e-23, 5.2856e-23, 2.0556e-25, 6.2809e-24, 1.8953e-23,\n 4.0729e-24, 9.9781e-25, 5.7826e-24, 2.3676e-23, 2.0789e-23, 2.0696e-24,\n 1.4371e-23, 7.9925e-25, 1.9634e-24, 2.0195e-25, 4.7371e-23, 7.8411e-24,\n 4.5268e-24, 1.3346e-25, 3.0142e-23, 1.3946e-24, 6.2728e-24, 9.6214e-25,\n 1.4320e-25, 9.0036e-25, 1.4575e-23, 1.1093e-24, 2.1374e-24, 9.4740e-25,\n 4.8242e-24, 1.4322e-24, 3.8237e-23, 2.0924e-23, 6.7307e-24, 3.1607e-25,\n 1.7886e-23, 1.5219e-23, 1.6466e-24, 2.6559e-25, 1.8203e-23, 1.2059e-24,\n 5.1873e-24, 4.6192e-24, 3.4679e-23, 2.4947e-24, 1.7678e-23, 1.3878e-24,\n 1.4866e-25, 1.1898e-23, 1.4022e-23, 3.2184e-25, 3.4096e-25, 3.1318e-24,\n 2.5890e-23, 2.8868e-23, 6.5111e-26, 1.8388e-25, 2.2831e-25, 5.6888e-25,\n 5.0353e-25, 7.4838e-24, 6.7315e-24, 6.8476e-24, 1.1342e-23, 2.1256e-23,\n 1.6511e-25, 2.1648e-24, 1.3790e-24, 5.0887e-25, 1.7738e-23, 5.4615e-24,\n 1.2438e-24, 5.6718e-24, 8.6914e-24, 4.8306e-23, 6.8562e-25, 7.8916e-25,\n 2.5061e-24, 1.5821e-25, 6.1380e-24, 5.9605e-24, 1.1269e-25, 1.7295e-25,\n 3.5364e-23, 5.8914e-24, 4.8026e-24, 1.9089e-24, 2.1390e-23, 9.2676e-26,\n 1.2989e-23, 6.2971e-24, 1.6434e-24, 1.2244e-23, 1.3838e-24, 4.5824e-26,\n 4.1412e-23, 5.4181e-24, 2.6782e-25, 1.5732e-24, 1.4339e-23, 3.0320e-24,\n 3.2150e-23, 4.1401e-23, 4.3034e-24, 4.7267e-23, 4.8876e-25, 2.3586e-25,\n 2.6864e-25, 2.1650e-24, 2.2897e-25, 9.8456e-23, 2.0985e-24, 7.6123e-24,\n 3.9751e-25, 2.7509e-23, 5.3220e-23, 6.9625e-26, 5.4314e-24, 3.3683e-23,\n 3.0104e-23, 3.1031e-24, 3.2871e-25, 6.1472e-24, 5.7739e-25, 1.1021e-24,\n 1.1865e-25, 5.7454e-24, 6.3771e-24, 6.1770e-26, 2.7130e-25, 5.9586e-26,\n 2.2670e-23, 5.1378e-23, 7.8962e-25, 5.7103e-24, 5.5088e-24, 1.0946e-23,\n 5.3106e-24, 1.6354e-23, 4.8750e-23, 1.9696e-25, 1.3471e-23, 1.8686e-23,\n 5.7460e-25, 3.0049e-24, 1.0105e-23, 1.2778e-22, 2.2384e-24, 4.5890e-25,\n 3.4137e-24, 7.8031e-24, 1.1476e-24, 1.8621e-23, 5.2959e-26, 3.8366e-24,\n 4.6490e-26, 5.2046e-23, 1.4891e-24, 1.1375e-23, 2.4727e-23, 2.8432e-24,\n 1.1614e-24, 7.5586e-24, 2.6042e-25, 1.0668e-24, 1.5267e-23, 1.2753e-25,\n 3.3281e-23, 8.6850e-25, 1.1693e-23, 1.5024e-24, 1.7878e-23, 1.0763e-25,\n 1.5129e-24, 8.2132e-24, 1.7007e-24, 4.1705e-24, 3.3492e-25, 2.4591e-23,\n 3.7771e-23, 3.7546e-23, 1.2148e-24, 2.9472e-23, 2.6634e-25, 2.4785e-24,\n 1.0933e-24, 4.4461e-25, 1.8850e-24, 5.7547e-24, 6.3145e-24, 1.5692e-25,\n 6.1012e-25, 5.1675e-25, 2.6950e-23, 1.8125e-25, 8.6439e-24, 1.0234e-23,\n 7.0143e-24, 8.2061e-24, 1.8619e-23, 1.9021e-24, 2.0548e-25, 7.9458e-24,\n 3.0767e-24, 2.1282e-23, 8.2163e-24, 2.4443e-24, 4.6381e-24, 1.8093e-23,\n 2.1563e-23, 1.4861e-23, 1.1666e-23, 3.7212e-25, 1.7094e-24, 1.6185e-23,\n 2.4372e-24, 1.2125e-23, 5.4573e-24, 6.1220e-24, 9.6499e-24, 9.0993e-24,\n 3.4921e-24, 4.1913e-24, 2.7036e-23, 2.6375e-24, 4.2536e-25, 3.8095e-25,\n 1.8730e-23, 2.6262e-25, 1.6480e-23, 2.5060e-26, 1.4577e-24, 1.7364e-23,\n 4.1075e-23, 1.5594e-23, 3.4271e-23, 9.7375e-25], device='cuda:0')"
 },
 "49": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([9.0110e-28, 2.4787e-28, 8.6622e-27, 2.4504e-26, 2.8928e-28, 1.5089e-26,\n 1.1065e-27, 9.8641e-27, 4.7931e-27, 6.1448e-27, 2.7015e-26, 3.4204e-27,\n 5.6688e-26, 2.0547e-27, 9.5029e-28, 7.2611e-27, 9.2492e-28, 6.8426e-26,\n 1.2737e-27, 8.2884e-26, 2.8558e-25, 3.8889e-28, 5.1672e-27, 2.6003e-25,\n 3.9537e-27, 1.0523e-28, 1.7044e-27, 9.5631e-26, 1.5298e-26, 9.8003e-28,\n 4.8465e-26, 1.2908e-27, 8.1613e-27, 2.8562e-27, 2.5619e-25, 4.2597e-26,\n 2.9266e-27, 5.9604e-29, 1.6698e-25, 3.2036e-27, 7.7118e-27, 3.0785e-27,\n 1.0165e-28, 2.1940e-28, 2.8226e-26, 1.4256e-27, 6.4991e-28, 2.4588e-28,\n 4.9198e-26, 8.6270e-28, 9.3382e-26, 2.9106e-26, 4.6962e-27, 8.8813e-29,\n 1.1555e-26, 2.2406e-26, 1.6461e-27, 1.0542e-27, 2.8294e-27, 6.1942e-28,\n 4.6049e-27, 3.0192e-27, 3.8890e-25, 4.2150e-27, 4.9259e-26, 4.0620e-27,\n 1.1635e-27, 4.9905e-27, 6.2860e-27, 3.5276e-28, 1.1959e-27, 9.4221e-28,\n 8.7146e-27, 2.0465e-26, 5.7035e-28, 1.5264e-28, 2.6722e-28, 2.6845e-28,\n 3.6361e-28, 9.1291e-27, 1.0311e-26, 3.6601e-27, 1.3156e-26, 1.4137e-26,\n 7.1383e-28, 1.3064e-27, 6.5163e-28, 9.8271e-28, 5.3675e-26, 1.5553e-26,\n 3.5965e-27, 5.7071e-27, 1.5495e-26, 9.5849e-26, 2.2684e-29, 1.1540e-27,\n 7.7588e-28, 1.1149e-28, 4.1059e-26, 3.6381e-26, 3.6775e-28, 2.7505e-28,\n 7.7844e-26, 6.2474e-27, 9.2439e-27, 1.1665e-27, 6.1293e-26, 6.6612e-28,\n 9.5988e-27, 1.8677e-27, 2.9799e-28, 1.1336e-25, 4.1731e-27, 1.5178e-28,\n 1.8869e-25, 3.4372e-26, 5.4672e-28, 4.6260e-28, 3.4348e-26, 1.2632e-27,\n 3.8710e-26, 1.9089e-25, 3.1131e-27, 6.8978e-26, 2.4305e-27, 8.1412e-28,\n 3.1268e-28, 3.0067e-28, 5.1183e-27, 4.6677e-25, 2.2069e-27, 4.6852e-27,\n 3.9817e-28, 1.1723e-25, 4.8806e-26, 2.4406e-28, 8.4196e-26, 2.2619e-25,\n 4.9422e-26, 8.4537e-28, 3.3281e-27, 2.0522e-27, 5.3788e-28, 1.3939e-27,\n 3.7938e-28, 1.5778e-27, 3.5396e-27, 2.4135e-28, 4.0193e-28, 5.2460e-28,\n 1.2933e-26, 4.7968e-25, 3.4298e-28, 1.1135e-26, 3.8422e-27, 7.6260e-27,\n 1.3324e-26, 5.1872e-26, 1.2659e-25, 3.5904e-28, 4.9834e-27, 2.4321e-26,\n 1.1767e-27, 1.3179e-27, 8.2915e-26, 4.2144e-25, 1.6502e-27, 9.8667e-28,\n 2.2652e-27, 9.8654e-27, 2.8149e-28, 1.6586e-26, 2.0886e-28, 1.0533e-26,\n 3.0773e-28, 9.7084e-26, 1.0423e-27, 2.2890e-27, 7.8253e-26, 5.4974e-27,\n 5.7128e-28, 3.1488e-26, 1.6840e-28, 3.7672e-28, 1.7074e-25, 1.1852e-28,\n 3.8143e-25, 4.9405e-28, 2.3852e-26, 7.9891e-28, 8.5380e-27, 1.7050e-28,\n 1.2710e-27, 3.3465e-27, 5.7908e-28, 2.3635e-27, 1.6691e-28, 4.0856e-26,\n 1.6505e-25, 8.7356e-26, 2.8856e-28, 2.5974e-26, 2.6578e-28, 3.0040e-28,\n 2.7175e-29, 2.6875e-28, 1.9862e-27, 1.9990e-26, 7.7003e-27, 6.7220e-28,\n 8.5413e-28, 1.9294e-27, 7.5600e-26, 4.1521e-28, 2.3358e-26, 1.2284e-26,\n 8.7631e-27, 4.6985e-26, 1.2963e-25, 1.7852e-27, 5.9939e-28, 2.2445e-26,\n 3.9312e-27, 4.4041e-26, 3.9363e-27, 9.1336e-29, 2.4504e-27, 1.7442e-26,\n 3.8472e-26, 7.7163e-26, 4.1757e-26, 9.7942e-29, 1.0690e-27, 1.8612e-26,\n 9.5179e-28, 9.9660e-27, 1.8330e-26, 2.5332e-27, 4.6976e-26, 3.9232e-26,\n 3.8199e-27, 4.5100e-27, 3.2419e-26, 8.1702e-28, 2.2914e-28, 6.0448e-28,\n 2.3016e-26, 3.8583e-28, 5.1964e-26, 7.2932e-28, 2.1270e-27, 4.7665e-26,\n 3.5818e-26, 2.7018e-26, 3.1638e-25, 1.4177e-27], device='cuda:0')"
 },
 "50": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([6.0139e-27, 1.3248e-28, 4.3713e-26, 2.0524e-26, 8.8970e-28, 1.0456e-26,\n 3.9638e-27, 4.7538e-26, 1.8685e-26, 2.3151e-26, 3.7235e-26, 2.4173e-26,\n 1.8586e-26, 7.6314e-28, 1.9199e-27, 1.2033e-26, 5.9341e-27, 5.0907e-26,\n 4.1716e-28, 1.6311e-25, 1.7100e-25, 3.6904e-28, 1.0049e-26, 5.7314e-26,\n 7.2134e-27, 4.0867e-27, 4.3581e-27, 4.9792e-26, 2.9418e-26, 1.1463e-26,\n 6.2315e-26, 5.0687e-28, 6.0819e-27, 4.5482e-27, 1.7766e-25, 2.6356e-26,\n 1.2109e-26, 8.8620e-28, 9.0057e-26, 5.0496e-27, 2.0592e-26, 6.8961e-27,\n 6.4353e-28, 4.5809e-27, 5.8040e-26, 4.8837e-27, 5.2702e-27, 2.5053e-27,\n 2.6663e-26, 5.7881e-27, 5.6106e-26, 1.7141e-26, 2.8839e-26, 2.5004e-27,\n 1.8653e-26, 2.5648e-26, 5.5354e-27, 6.1430e-28, 2.2776e-26, 3.4050e-27,\n 1.9320e-26, 1.1596e-26, 1.5068e-25, 7.9194e-27, 1.3155e-26, 3.7724e-27,\n 2.9130e-28, 2.2786e-26, 2.1764e-26, 3.6200e-28, 3.4751e-27, 3.9700e-27,\n 4.3213e-26, 6.0204e-26, 3.4596e-28, 1.0280e-27, 1.2612e-27, 7.5003e-28,\n 1.2956e-27, 3.0937e-26, 7.3871e-27, 8.0569e-27, 1.4147e-26, 2.5881e-26,\n 1.7607e-28, 6.7950e-27, 1.0860e-27, 2.1124e-27, 6.7398e-26, 2.2831e-26,\n 9.0009e-27, 1.0607e-26, 2.4473e-26, 5.0677e-26, 1.5177e-27, 2.9556e-27,\n 7.8698e-27, 9.2676e-29, 2.2755e-26, 3.0678e-26, 3.8068e-28, 4.3436e-28,\n 4.3592e-26, 1.4004e-26, 1.0326e-26, 4.5540e-27, 8.9693e-26, 2.5378e-28,\n 7.7947e-27, 1.0887e-26, 3.2353e-27, 6.4448e-26, 8.5840e-27, 1.2310e-28,\n 7.4547e-26, 3.5791e-26, 3.1489e-28, 4.1822e-27, 4.8492e-26, 4.4659e-27,\n 6.7876e-26, 9.6367e-26, 8.2359e-27, 1.3149e-25, 1.1510e-27, 9.9188e-28,\n 1.8951e-27, 7.2761e-27, 4.7748e-28, 1.9478e-25, 7.7875e-27, 8.8734e-27,\n 1.9513e-27, 5.3438e-26, 1.2390e-25, 2.2877e-28, 3.3381e-26, 1.0843e-25,\n 1.0599e-25, 5.4913e-27, 2.2051e-27, 1.0827e-26, 1.1604e-27, 1.7101e-27,\n 3.4291e-28, 5.4823e-27, 9.2969e-27, 1.8333e-28, 3.5235e-28, 1.3837e-28,\n 3.1883e-26, 1.8224e-25, 7.9358e-28, 2.4820e-26, 1.2366e-26, 6.3787e-27,\n 1.8910e-26, 7.2779e-26, 1.2477e-25, 6.2797e-28, 5.1257e-26, 1.6523e-26,\n 1.6984e-27, 4.0103e-27, 4.1113e-26, 3.7468e-25, 4.6699e-27, 1.5265e-27,\n 3.9610e-27, 1.4433e-26, 4.5812e-27, 5.4917e-26, 1.1269e-28, 1.9508e-26,\n 1.4434e-28, 5.0895e-26, 3.3413e-27, 1.1307e-26, 9.0733e-26, 1.0423e-26,\n 1.3094e-27, 1.8869e-26, 3.8133e-28, 1.7408e-27, 6.8174e-26, 1.9789e-28,\n 1.4263e-25, 4.0066e-27, 1.9600e-26, 3.7075e-27, 5.5633e-26, 1.1647e-27,\n 4.2359e-27, 9.7652e-27, 2.5763e-27, 1.2542e-26, 1.0379e-27, 5.1938e-26,\n 9.9097e-26, 9.3153e-26, 4.4148e-28, 4.2926e-26, 2.2214e-27, 4.3277e-27,\n 1.5385e-27, 2.3659e-27, 7.4501e-27, 2.7857e-26, 1.3401e-26, 3.5842e-28,\n 4.3131e-27, 1.7172e-27, 7.5446e-26, 3.6882e-28, 3.2992e-26, 1.9536e-26,\n 1.4189e-26, 3.5523e-26, 5.8737e-26, 6.3061e-27, 6.4275e-28, 2.8780e-26,\n 5.7134e-27, 8.0025e-26, 2.0330e-26, 7.7726e-27, 5.5247e-27, 1.8889e-26,\n 4.9311e-26, 5.6417e-26, 3.9675e-26, 5.5290e-28, 4.2857e-27, 5.0946e-26,\n 6.1522e-28, 1.9860e-26, 2.4229e-26, 1.4792e-26, 4.1432e-26, 1.2969e-26,\n 1.5298e-26, 1.6149e-26, 3.4271e-26, 9.7992e-27, 2.1889e-27, 1.6333e-27,\n 3.4472e-26, 4.0818e-28, 4.9642e-26, 3.2131e-28, 3.4051e-27, 4.2340e-26,\n 1.2964e-25, 2.0541e-26, 8.2030e-26, 3.4706e-27], device='cuda:0')"
 },
 "51": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[ 5.6052e-45, -5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n ...,\n [-5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45]], device='cuda:0')",
 "exp_avg_sq": "tensor([[1.5589e-26, 1.9026e-27, 7.2553e-27, ..., 7.7244e-27, 5.1520e-28,\n 5.9042e-27],\n [2.6763e-27, 5.8093e-28, 1.2184e-28, ..., 3.2760e-27, 1.0463e-27,\n 1.4347e-27],\n [2.4185e-26, 7.5008e-27, 4.6241e-26, ..., 6.9217e-27, 7.4471e-28,\n 4.9223e-27],\n ...,\n [4.0618e-28, 1.1745e-27, 1.7437e-28, ..., 4.3389e-28, 2.3767e-28,\n 1.4796e-28],\n [2.4967e-26, 1.8198e-26, 2.6479e-26, ..., 3.6911e-26, 1.4143e-27,\n 2.2537e-26],\n [5.1003e-27, 8.9374e-27, 8.1091e-27, ..., 1.6688e-27, 4.6130e-28,\n 3.5721e-27]], device='cuda:0')"
 },
 "52": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([2.1143e-23, 4.5476e-24, 4.1350e-23, 2.9063e-25, 1.2788e-23, 5.4085e-23,\n 8.2341e-25, 2.9141e-24, 1.4266e-24, 4.2342e-26, 7.3585e-24, 4.3481e-24,\n 3.3197e-23, 2.7678e-24, 4.9199e-24, 1.4375e-23, 2.2439e-25, 9.5637e-24,\n 4.3805e-23, 6.1678e-23, 1.8267e-25, 1.0187e-24, 1.6444e-24, 3.2987e-23,\n 3.2765e-23, 3.5434e-23, 9.2587e-24, 4.9621e-23, 7.3023e-24, 4.4945e-23,\n 9.8208e-23, 1.5152e-23, 9.0598e-23, 2.1335e-25, 1.1735e-23, 1.3514e-24,\n 1.9579e-24, 7.7422e-24, 3.3173e-23, 6.3388e-24, 3.1782e-24, 1.5307e-23,\n 2.6634e-24, 1.7983e-23, 3.0157e-23, 2.7391e-24, 1.0658e-23, 6.7516e-24,\n 1.7404e-23, 4.9355e-23, 6.7308e-24, 7.6983e-23, 1.1328e-25, 5.4754e-25,\n 5.4115e-23, 2.1135e-25, 8.6025e-25, 1.0320e-23, 2.0124e-23, 1.3145e-24,\n 1.8026e-25, 1.1243e-24, 3.7546e-23, 3.0688e-25, 2.6170e-23, 1.5075e-24,\n 3.5922e-24, 2.5138e-23, 3.7833e-23, 1.4661e-25, 4.5773e-25, 1.1632e-24,\n 7.7875e-23, 5.0454e-24, 2.0657e-24, 1.4979e-24, 2.6892e-24, 7.6444e-25,\n 1.4457e-23, 5.2172e-24, 3.3509e-25, 6.0435e-25, 6.7184e-24, 2.5362e-25,\n 1.2628e-24, 1.5207e-23, 1.0429e-23, 1.3469e-24, 6.6828e-24, 2.2549e-24,\n 2.9623e-23, 1.7137e-25, 9.2983e-25, 5.3353e-26, 3.3330e-23, 5.9533e-25,\n 4.5268e-24, 2.8555e-23, 1.4708e-23, 3.3597e-24, 2.9223e-24, 1.3108e-24,\n 8.3768e-24, 1.8011e-23, 5.3440e-24, 7.2238e-25, 6.4242e-24, 1.0456e-23,\n 3.8532e-23, 1.8132e-24, 4.5528e-24, 1.5078e-25, 1.4066e-24, 8.9040e-23,\n 1.0784e-22, 4.0439e-24, 7.5484e-25, 1.7226e-24, 2.9920e-25, 1.7411e-24,\n 6.4243e-24, 1.9590e-24, 6.8237e-24, 4.7516e-24, 9.1945e-24, 3.5402e-23,\n 8.3138e-25, 7.7693e-24, 3.1518e-23, 5.6763e-23, 3.1179e-23, 9.2977e-24,\n 4.9042e-24, 7.8423e-24, 5.9358e-25, 1.5219e-23, 1.5519e-24, 1.2303e-23,\n 1.3440e-25, 1.7210e-25, 5.1355e-23, 1.1193e-23, 6.4772e-24, 1.1672e-23,\n 4.4091e-24, 7.7233e-25, 1.4636e-24, 1.4404e-24, 1.5331e-24, 1.8382e-24,\n 4.9509e-24, 2.8502e-25, 6.4371e-24, 7.9686e-26, 2.3722e-23, 1.8443e-24,\n 2.4758e-25, 3.1954e-24, 9.8320e-23, 1.8166e-23, 4.2537e-23, 3.7239e-23,\n 1.8670e-23, 1.5686e-25, 8.7025e-25, 3.4569e-23, 1.0860e-23, 2.7232e-24,\n 1.1855e-22, 2.1142e-24, 3.5821e-23, 9.1993e-23, 6.3017e-23, 1.0371e-23,\n 6.3484e-23, 5.4041e-23, 2.9142e-23, 1.0732e-23, 1.2070e-23, 5.0005e-25,\n 7.2966e-24, 3.3838e-24, 2.5230e-24, 1.6008e-24, 7.5526e-24, 1.5473e-24,\n 9.8797e-25, 2.1081e-25, 3.4953e-23, 1.0723e-24, 1.3511e-23, 1.7619e-25,\n 6.0929e-23, 7.7475e-24, 1.0654e-24, 6.0479e-23, 6.4427e-24, 1.8138e-24,\n 2.2903e-23, 1.8508e-23, 2.2719e-23, 2.2036e-23, 1.8187e-24, 4.2147e-25,\n 6.8154e-25, 1.2358e-23, 1.1792e-23, 1.2114e-25, 4.6748e-24, 4.0072e-24,\n 1.7173e-22, 5.1460e-23, 6.2415e-24, 2.0217e-24, 8.6555e-24, 1.1531e-23,\n 7.8985e-24, 3.1472e-24, 4.3295e-23, 3.0411e-23, 1.0545e-23, 5.7641e-25,\n 1.2587e-24, 1.3738e-23, 6.5934e-23, 9.0680e-25, 2.0487e-24, 2.4856e-23,\n 2.7731e-23, 2.7399e-23, 2.1038e-23, 6.7221e-24, 2.8347e-24, 1.8262e-23,\n 1.7002e-23, 2.4080e-25, 5.4311e-23, 1.2096e-24, 1.2472e-23, 4.0504e-23,\n 1.4462e-23, 9.7874e-24, 3.9623e-23, 3.6179e-24, 5.1627e-24, 1.3136e-24,\n 3.7825e-23, 4.0615e-23, 1.0478e-23, 4.0240e-24, 5.0931e-24, 4.6143e-23,\n 1.5933e-23, 5.1781e-25, 1.0820e-22, 9.9901e-24], device='cuda:0')"
 },
 "53": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([7.9064e-26, 5.3986e-27, 1.4858e-25, 9.1159e-28, 9.6756e-27, 8.0356e-26,\n 1.1168e-27, 1.7521e-27, 9.9867e-28, 4.0460e-28, 3.9713e-27, 1.5862e-27,\n 8.1072e-26, 2.1289e-28, 3.4693e-27, 2.0346e-26, 9.0692e-28, 6.8979e-26,\n 1.9425e-25, 3.6326e-25, 1.4739e-27, 2.6656e-27, 1.0363e-27, 1.0119e-25,\n 5.8284e-26, 6.4703e-26, 2.4021e-26, 2.1376e-25, 2.0423e-27, 3.4219e-26,\n 1.2383e-24, 4.6151e-26, 3.8464e-25, 1.1091e-27, 4.6347e-26, 2.9174e-28,\n 2.6134e-27, 1.7266e-27, 6.3620e-26, 2.0265e-27, 3.3362e-27, 4.0391e-26,\n 3.8264e-27, 6.3447e-27, 1.2869e-26, 7.4606e-28, 5.9418e-26, 4.1391e-27,\n 6.0483e-26, 8.7841e-26, 2.1412e-27, 1.9135e-25, 8.0822e-28, 4.1799e-28,\n 1.0824e-25, 6.6725e-28, 4.9198e-28, 2.6497e-27, 1.2389e-26, 7.9574e-28,\n 4.4126e-29, 2.1066e-27, 1.5833e-25, 1.0113e-28, 2.6936e-26, 1.4162e-27,\n 1.3550e-26, 2.3272e-25, 4.4761e-26, 1.0598e-27, 4.3253e-28, 2.8686e-28,\n 3.3289e-25, 2.5603e-26, 1.4973e-26, 4.4061e-26, 1.8050e-27, 5.0620e-27,\n 1.8278e-26, 8.9418e-26, 3.2325e-27, 2.5893e-27, 3.4365e-27, 1.3307e-27,\n 7.6511e-28, 7.2151e-26, 2.7806e-26, 1.2443e-27, 3.2260e-27, 5.2994e-26,\n 6.4679e-26, 7.1885e-28, 1.1417e-27, 4.2833e-28, 4.5838e-26, 1.2511e-27,\n 1.1272e-26, 7.5107e-26, 1.3235e-26, 2.1597e-28, 3.5114e-27, 1.6230e-26,\n 4.4073e-26, 6.4143e-27, 7.1301e-27, 1.1504e-27, 2.5238e-26, 2.7555e-26,\n 5.9826e-26, 9.8676e-28, 4.8160e-27, 2.8047e-28, 1.6637e-27, 3.0479e-25,\n 1.6602e-24, 3.5690e-27, 1.3925e-27, 7.2002e-28, 9.4790e-28, 1.7411e-27,\n 1.8641e-26, 3.4340e-28, 5.2338e-27, 3.4911e-27, 2.7560e-27, 1.1307e-25,\n 1.1251e-27, 4.7148e-27, 4.7707e-26, 8.6631e-26, 1.1743e-26, 1.2970e-26,\n 1.1328e-26, 5.5160e-26, 2.8802e-28, 4.0075e-26, 8.6737e-27, 7.1680e-26,\n 1.2971e-27, 1.2212e-27, 1.1995e-25, 3.5595e-26, 3.6563e-27, 2.1285e-25,\n 1.5654e-27, 2.5605e-27, 1.9039e-27, 6.9432e-28, 2.8668e-27, 6.2439e-28,\n 5.5611e-26, 5.6351e-28, 3.5721e-27, 1.4705e-27, 6.7161e-26, 1.0310e-27,\n 9.8387e-28, 8.2007e-28, 6.4238e-25, 6.3313e-27, 1.6246e-25, 1.4188e-25,\n 1.3344e-26, 2.2372e-28, 1.2056e-27, 4.0840e-26, 3.7157e-27, 3.8230e-27,\n 1.7055e-25, 1.7447e-27, 9.7619e-26, 3.3287e-25, 1.7055e-25, 7.2575e-26,\n 1.0565e-25, 9.7090e-26, 4.5679e-26, 5.7811e-27, 2.4750e-26, 6.4436e-28,\n 3.0026e-26, 1.2991e-27, 5.6235e-27, 4.8733e-27, 6.2368e-26, 7.0622e-27,\n 1.0879e-27, 1.1451e-28, 1.1131e-25, 4.7640e-27, 1.8018e-26, 1.1608e-27,\n 1.7667e-25, 6.6590e-27, 1.4488e-27, 3.6945e-25, 2.2073e-27, 1.3667e-27,\n 7.1842e-26, 3.1210e-26, 8.0345e-26, 1.4663e-26, 2.2299e-27, 4.4405e-28,\n 5.3802e-28, 3.8033e-26, 6.2624e-26, 4.8277e-28, 8.9460e-28, 1.4091e-27,\n 5.5791e-25, 9.0965e-26, 1.6453e-26, 1.6163e-27, 1.6926e-26, 1.0231e-26,\n 2.0911e-26, 1.3543e-26, 3.4861e-26, 6.6410e-26, 8.3706e-28, 4.0472e-28,\n 7.6254e-28, 7.9180e-27, 2.1080e-25, 1.3267e-27, 3.4406e-26, 1.7330e-26,\n 1.8064e-25, 4.2542e-26, 6.1638e-26, 1.6839e-26, 1.7439e-27, 3.6021e-26,\n 4.9902e-26, 5.1854e-28, 1.8585e-25, 6.5867e-28, 3.9769e-26, 6.0209e-26,\n 1.1373e-26, 4.6246e-26, 8.6278e-26, 5.0757e-27, 1.2479e-27, 1.6935e-27,\n 2.5549e-25, 7.9261e-26, 1.1842e-26, 4.3974e-27, 1.1131e-27, 1.7082e-25,\n 1.4762e-25, 3.8455e-28, 1.0323e-24, 4.3826e-27], device='cuda:0')"
 },
 "54": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([5.0966e-26, 1.8072e-26, 1.2827e-25, 2.8904e-27, 3.5856e-26, 1.1045e-25,\n 3.4730e-27, 1.7172e-26, 9.6294e-27, 1.8507e-27, 9.6253e-27, 6.8831e-27,\n 1.3414e-25, 2.6275e-27, 1.2240e-26, 4.5128e-26, 2.6159e-27, 5.4099e-26,\n 1.4644e-25, 2.3337e-25, 1.8614e-27, 1.8643e-26, 6.1435e-27, 1.3069e-25,\n 3.7714e-26, 1.2027e-25, 1.4559e-26, 8.7134e-26, 1.9105e-26, 9.3071e-26,\n 3.8686e-25, 1.8467e-26, 1.5093e-25, 4.2202e-28, 5.8009e-26, 4.3390e-27,\n 3.9730e-26, 3.5719e-26, 1.0166e-25, 5.7584e-27, 1.4325e-26, 1.5990e-26,\n 7.3143e-27, 7.4071e-26, 1.1735e-25, 6.0613e-27, 5.4206e-26, 1.8916e-26,\n 3.1648e-26, 1.0168e-25, 3.7264e-27, 1.3926e-25, 7.8414e-28, 4.0053e-28,\n 1.2256e-25, 3.9509e-27, 5.6864e-27, 3.9171e-26, 3.5240e-26, 4.8699e-27,\n 1.7839e-27, 3.7048e-27, 1.3329e-25, 2.0644e-27, 1.0303e-25, 1.3428e-26,\n 1.9033e-26, 1.1380e-25, 7.3367e-26, 3.4667e-28, 3.2068e-27, 2.1012e-27,\n 1.0116e-25, 1.1044e-26, 3.0730e-26, 3.6222e-26, 6.0876e-27, 1.9417e-26,\n 4.6715e-26, 4.4587e-26, 1.6715e-27, 4.6433e-27, 3.3412e-26, 5.1005e-27,\n 1.3241e-27, 6.4045e-26, 2.1102e-26, 4.3747e-27, 3.1360e-26, 3.2437e-26,\n 5.9489e-26, 5.1269e-28, 4.5465e-27, 1.5925e-28, 5.5139e-26, 8.2812e-27,\n 2.2054e-26, 1.1778e-25, 1.4765e-26, 3.8977e-27, 1.7658e-26, 2.8002e-26,\n 2.7312e-26, 5.8363e-26, 1.2302e-26, 3.7829e-28, 3.4135e-26, 3.4501e-26,\n 5.9516e-26, 2.7092e-27, 1.3867e-26, 4.1052e-29, 9.8479e-27, 3.2907e-25,\n 3.6852e-25, 1.0780e-26, 6.8580e-27, 9.9297e-27, 3.6217e-27, 1.0159e-26,\n 2.6291e-26, 3.6114e-27, 1.9248e-26, 1.3377e-26, 3.0731e-26, 8.3479e-26,\n 4.0319e-27, 1.9484e-26, 5.1003e-26, 8.9850e-26, 5.6903e-26, 1.6016e-26,\n 1.6223e-26, 4.4463e-26, 1.7275e-27, 3.3868e-26, 2.5578e-26, 4.5284e-26,\n 1.6675e-27, 1.9089e-27, 1.6890e-25, 4.0395e-26, 2.6206e-26, 1.9695e-25,\n 1.3561e-26, 9.0877e-28, 5.0294e-27, 5.3818e-27, 2.6225e-27, 6.2335e-27,\n 2.9168e-26, 4.1347e-27, 2.8502e-26, 4.0928e-28, 8.5761e-26, 3.3913e-27,\n 4.6588e-27, 1.4222e-26, 3.1772e-25, 4.0180e-26, 1.3882e-25, 1.1023e-25,\n 5.8448e-26, 4.9390e-27, 2.9460e-27, 1.1892e-25, 3.2675e-26, 4.4334e-26,\n 2.3799e-25, 4.1390e-27, 1.0897e-25, 2.0696e-25, 2.1955e-25, 7.6723e-26,\n 2.0452e-25, 6.4242e-26, 3.5263e-26, 1.4713e-26, 5.0262e-26, 1.5011e-27,\n 1.5831e-26, 1.2877e-26, 1.2000e-26, 5.8264e-27, 4.5224e-26, 4.6158e-27,\n 1.0246e-26, 3.7301e-27, 3.8258e-26, 1.0547e-26, 3.0660e-26, 1.8641e-27,\n 6.8408e-26, 8.8303e-27, 1.6852e-26, 2.3409e-25, 8.6346e-27, 4.2739e-27,\n 6.4278e-26, 6.5734e-26, 7.9038e-26, 4.4232e-26, 9.7923e-27, 5.9365e-28,\n 6.9897e-27, 5.1466e-26, 7.2589e-26, 2.9438e-27, 6.2168e-27, 1.3947e-26,\n 4.5288e-25, 1.4081e-25, 2.3185e-26, 4.5807e-27, 3.3147e-26, 1.3556e-26,\n 3.8561e-26, 1.4820e-26, 1.4986e-25, 1.0086e-25, 1.9934e-26, 3.6261e-27,\n 2.5399e-26, 5.6770e-26, 1.7896e-25, 5.3925e-27, 3.5989e-26, 8.9303e-26,\n 8.2010e-26, 9.5955e-26, 2.7488e-26, 1.5983e-26, 1.5522e-26, 7.0187e-26,\n 2.8123e-26, 1.3580e-28, 2.1260e-25, 2.1665e-27, 4.7790e-26, 1.5339e-25,\n 5.6710e-26, 4.2638e-26, 4.6649e-26, 1.4187e-26, 8.8411e-27, 4.2088e-27,\n 8.1995e-26, 1.3394e-25, 2.9164e-26, 1.4789e-26, 1.2420e-26, 6.6663e-26,\n 8.9881e-26, 2.8030e-28, 4.0386e-25, 4.7647e-26], device='cuda:0')"
 },
 "55": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[-5.6052e-45, 5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n [-5.6052e-45, -5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n [-5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n ...,\n [-5.6052e-45, -5.6052e-45, 5.6052e-45, ..., -5.6052e-45,\n -5.6052e-45, -5.6052e-45],\n [-5.6052e-45, -5.6052e-45, 5.6052e-45, ..., -5.6052e-45,\n -5.6052e-45, -5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45]], device='cuda:0')",
 "exp_avg_sq": "tensor([[7.0494e-27, 4.2549e-27, 7.4097e-27, ..., 3.2759e-27, 2.7553e-28,\n 2.2509e-27],\n [9.8778e-28, 3.1447e-27, 3.4507e-28, ..., 8.6578e-28, 3.2994e-28,\n 2.2395e-28],\n [1.0388e-28, 2.0661e-28, 8.2265e-28, ..., 1.8773e-28, 8.7371e-29,\n 1.3637e-28],\n ...,\n [5.7907e-27, 3.7714e-27, 5.0489e-27, ..., 8.9662e-29, 4.0525e-28,\n 5.0892e-28],\n [1.3131e-26, 2.3367e-27, 9.0438e-27, ..., 4.1730e-27, 7.9927e-28,\n 5.0467e-27],\n [1.6350e-26, 1.0670e-26, 9.8161e-27, ..., 2.1745e-26, 3.2623e-28,\n 6.7167e-27]], device='cuda:0')"
 },
 "56": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([1.2940e-23, 1.5382e-24, 1.8893e-26, 2.1662e-24, 1.6560e-23, 1.8241e-23,\n 5.4324e-25, 2.3236e-24, 1.5262e-23, 8.5269e-24, 6.7715e-25, 1.8165e-23,\n 4.9305e-24, 8.1265e-24, 7.3072e-24, 3.1594e-24, 2.4743e-25, 3.6720e-24,\n 8.8687e-24, 2.3020e-23, 2.7876e-25, 2.4065e-24, 2.0589e-24, 2.1875e-23,\n 2.6687e-25, 2.2322e-23, 1.3842e-23, 7.4254e-24, 1.4169e-23, 9.5136e-25,\n 1.1343e-23, 1.2871e-23, 1.4051e-23, 9.4812e-25, 1.9456e-23, 2.6937e-24,\n 2.3449e-25, 6.5454e-25, 1.5399e-24, 1.3021e-23, 7.8270e-24, 1.6167e-24,\n 9.8549e-26, 1.0417e-23, 6.7795e-26, 1.0533e-25, 1.2122e-24, 1.9933e-23,\n 8.2749e-24, 1.1499e-23, 1.1742e-23, 1.2833e-23, 8.1561e-25, 1.4243e-23,\n 6.5578e-24, 4.6073e-24, 1.2635e-24, 2.1257e-25, 7.6477e-23, 5.9562e-24,\n 2.3553e-24, 1.0673e-23, 1.3804e-23, 1.8160e-24, 5.6075e-25, 1.9097e-24,\n 9.9630e-24, 1.2001e-23, 1.8216e-23, 8.4579e-25, 2.8300e-25, 1.1990e-24,\n 7.4935e-23, 7.2836e-24, 2.2497e-24, 4.7283e-24, 1.0331e-25, 1.1037e-23,\n 3.7448e-24, 4.7149e-24, 3.9440e-24, 5.6208e-24, 2.5353e-25, 2.8846e-24,\n 4.5718e-24, 6.9678e-24, 4.6136e-24, 7.8618e-25, 5.8048e-24, 7.2995e-24,\n 3.8364e-25, 4.1691e-25, 5.1483e-24, 2.5737e-23, 9.6433e-25, 2.0066e-24,\n 2.1074e-23, 1.9735e-25, 6.0522e-25, 8.9941e-24, 1.6438e-24, 1.2804e-24,\n 5.0106e-24, 3.4982e-24, 9.0813e-25, 2.3563e-23, 4.5439e-26, 2.6368e-23,\n 7.5719e-26, 6.9144e-24, 9.0786e-24, 3.3716e-24, 1.6906e-25, 9.0721e-24,\n 5.7623e-23, 1.7827e-25, 4.2079e-24, 1.4452e-23, 1.6459e-26, 7.0052e-25,\n 3.5442e-25, 1.2954e-24, 3.4844e-25, 3.3600e-23, 2.4360e-23, 4.4360e-25,\n 3.5665e-24, 1.0870e-24, 1.4015e-23, 3.8729e-24, 1.3362e-23, 1.7735e-24,\n 3.2916e-24, 2.9128e-23, 5.3924e-23, 1.3146e-25, 2.4908e-24, 1.0303e-23,\n 1.2762e-23, 3.3046e-26, 3.2572e-23, 1.9187e-24, 2.6359e-23, 7.6075e-24,\n 2.4308e-25, 1.7180e-25, 5.7130e-24, 8.5665e-24, 1.7115e-24, 1.2395e-24,\n 1.0625e-24, 7.9789e-24, 3.1099e-24, 2.4989e-23, 4.0884e-24, 1.2045e-23,\n 1.5997e-24, 6.0159e-24, 4.7106e-23, 3.4734e-25, 4.1754e-23, 7.6797e-25,\n 4.1966e-24, 1.4925e-24, 3.2902e-24, 7.1058e-23, 4.1441e-24, 1.4066e-23,\n 9.4775e-24, 1.6002e-25, 1.9687e-23, 1.0351e-23, 5.7442e-24, 8.5617e-25,\n 4.1541e-24, 1.2111e-23, 8.2994e-25, 1.3992e-23, 2.3581e-23, 4.8807e-24,\n 1.7522e-23, 4.0757e-25, 1.9284e-24, 2.1188e-25, 2.1420e-23, 5.4461e-25,\n 1.2242e-24, 3.1505e-24, 1.0822e-24, 1.5460e-24, 5.4830e-25, 9.0875e-25,\n 1.7935e-23, 6.1753e-25, 1.9238e-25, 5.3787e-25, 8.1068e-26, 4.6117e-24,\n 1.4638e-24, 1.0010e-25, 9.6565e-24, 1.0890e-24, 2.5105e-24, 4.8171e-24,\n 9.4393e-24, 2.4868e-24, 8.6049e-24, 4.0635e-25, 3.7301e-25, 3.9437e-25,\n 9.5328e-24, 8.1133e-23, 6.7786e-25, 4.7509e-24, 4.2211e-26, 1.7822e-23,\n 4.4398e-24, 1.8958e-25, 1.6467e-23, 1.6154e-23, 2.3038e-25, 4.5087e-24,\n 2.5983e-24, 2.0589e-23, 2.4527e-23, 2.2540e-23, 4.3898e-24, 2.9760e-25,\n 2.1131e-23, 1.2617e-23, 2.3273e-25, 3.9987e-24, 1.6284e-23, 4.3866e-24,\n 2.0956e-24, 4.5142e-24, 1.5150e-24, 6.9364e-25, 1.9382e-25, 1.0858e-25,\n 3.5822e-26, 6.4520e-24, 2.3265e-24, 1.1864e-23, 7.9352e-24, 1.5892e-25,\n 7.3643e-24, 8.5392e-25, 3.8723e-25, 4.4780e-26, 3.6345e-24, 6.2320e-24,\n 3.7344e-24, 1.0871e-23, 1.3282e-23, 5.1605e-23], device='cuda:0')"
 },
 "57": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([4.5724e-26, 9.7005e-28, 7.4374e-28, 1.4313e-27, 1.9867e-26, 1.8853e-26,\n 8.6878e-28, 3.2342e-27, 1.9546e-26, 6.2139e-27, 2.9640e-28, 1.5324e-26,\n 3.9930e-26, 6.6030e-27, 5.5727e-27, 6.2836e-27, 2.9923e-28, 8.1285e-27,\n 6.5444e-27, 6.1563e-26, 9.9986e-28, 5.0212e-27, 3.6468e-28, 1.8816e-25,\n 3.7027e-28, 5.4479e-26, 4.2410e-27, 2.9072e-27, 5.0860e-27, 2.8850e-28,\n 1.2494e-25, 6.6265e-27, 1.0481e-26, 2.0900e-28, 1.2379e-25, 1.0723e-26,\n 4.2895e-28, 1.5290e-28, 5.8483e-28, 1.5398e-26, 1.4464e-26, 6.0064e-27,\n 2.1645e-27, 2.9066e-26, 6.3567e-28, 1.5008e-27, 1.8439e-27, 5.9564e-26,\n 9.4957e-27, 7.3376e-26, 6.4380e-27, 1.0836e-26, 5.3559e-28, 2.6333e-26,\n 7.3339e-27, 3.7672e-27, 3.2442e-28, 1.0316e-27, 6.2458e-26, 1.0945e-26,\n 8.1175e-28, 1.1867e-26, 5.7175e-26, 4.6843e-28, 1.4775e-27, 5.0649e-26,\n 6.4232e-26, 1.2539e-26, 4.7529e-27, 3.1760e-27, 4.8042e-28, 1.3758e-28,\n 2.5906e-25, 2.0451e-27, 1.6608e-27, 7.6289e-27, 3.9757e-28, 2.2115e-26,\n 4.4987e-27, 2.7633e-26, 2.2413e-26, 1.4969e-26, 1.0301e-28, 2.5450e-28,\n 8.4504e-28, 9.2161e-27, 1.6784e-27, 6.6977e-29, 8.4364e-27, 2.8533e-26,\n 8.5598e-28, 6.6589e-28, 1.2791e-26, 3.3974e-26, 1.3312e-27, 4.3195e-27,\n 1.1648e-25, 1.6815e-29, 1.5026e-28, 2.8128e-27, 7.8913e-28, 2.9201e-28,\n 2.0202e-27, 6.8400e-27, 4.3854e-28, 2.4047e-26, 6.4980e-29, 8.8254e-26,\n 1.1114e-28, 1.2015e-26, 1.9518e-26, 5.8110e-27, 2.4508e-28, 1.1305e-25,\n 1.9530e-25, 3.5942e-28, 2.6867e-26, 1.2814e-26, 3.2758e-28, 3.4528e-28,\n 4.5341e-28, 4.4800e-28, 8.1994e-28, 2.5509e-26, 4.7746e-26, 6.7022e-28,\n 2.0842e-27, 1.3210e-27, 4.5483e-26, 1.0376e-26, 1.0339e-26, 4.0757e-28,\n 5.1389e-27, 1.5846e-25, 9.4820e-26, 3.6330e-28, 7.3732e-28, 3.4665e-26,\n 3.8289e-26, 2.8245e-28, 1.8113e-25, 4.2699e-27, 4.0501e-26, 8.7324e-27,\n 1.6646e-28, 3.2320e-28, 2.0109e-27, 8.3192e-27, 1.2809e-27, 5.6694e-28,\n 2.3686e-28, 1.4753e-26, 2.5532e-26, 3.3221e-25, 6.5776e-27, 4.1411e-26,\n 3.4953e-28, 1.5728e-26, 6.4620e-26, 1.3590e-27, 1.3615e-25, 6.2637e-28,\n 3.2277e-27, 8.8994e-27, 1.5959e-26, 2.1888e-25, 1.0940e-27, 1.0714e-26,\n 1.9450e-27, 4.1726e-28, 2.9752e-26, 3.0983e-26, 8.2438e-26, 6.6399e-28,\n 4.3418e-27, 6.4181e-27, 2.3141e-27, 7.1823e-27, 1.5484e-25, 2.4156e-27,\n 8.7445e-27, 1.1719e-27, 4.2454e-27, 1.7555e-28, 1.7139e-25, 4.0346e-28,\n 1.9848e-27, 4.8864e-27, 3.0320e-28, 3.3967e-27, 7.1355e-28, 8.4532e-28,\n 1.5578e-26, 5.3122e-28, 3.0796e-28, 5.2253e-28, 1.0948e-28, 2.0371e-27,\n 9.5578e-28, 4.6987e-28, 1.8846e-26, 2.2349e-28, 1.9670e-27, 1.2037e-27,\n 1.0725e-26, 9.5018e-27, 1.2886e-26, 3.3372e-28, 1.5615e-27, 6.3309e-28,\n 6.8158e-27, 2.6703e-25, 5.5692e-28, 2.4118e-27, 3.7480e-28, 1.8154e-26,\n 4.3892e-27, 1.4811e-28, 1.3792e-25, 4.6193e-26, 7.7988e-28, 9.0038e-27,\n 4.9401e-27, 1.6348e-25, 3.4312e-26, 2.1405e-26, 1.2727e-26, 4.3375e-28,\n 8.6366e-26, 5.3682e-26, 1.9206e-27, 7.6755e-27, 1.0387e-25, 1.0817e-26,\n 2.9439e-27, 1.4788e-27, 2.6154e-27, 1.2576e-27, 1.3157e-28, 2.4882e-28,\n 1.3809e-27, 2.4388e-26, 2.3484e-28, 1.9676e-26, 1.4951e-26, 4.5757e-28,\n 3.4740e-27, 4.5097e-28, 3.5255e-28, 4.1388e-28, 6.0931e-27, 5.5824e-27,\n 7.1052e-27, 2.5208e-27, 8.4672e-26, 2.2871e-25], device='cuda:0')"
 },
 "58": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([4.1832e-26, 4.8075e-27, 1.4772e-28, 7.5519e-27, 4.2822e-26, 3.0519e-26,\n 4.2786e-27, 4.0406e-27, 4.3975e-26, 2.4341e-26, 9.2256e-28, 2.5626e-26,\n 2.0039e-26, 8.9903e-27, 1.2065e-26, 1.4250e-26, 1.3422e-27, 1.1042e-26,\n 3.1142e-26, 9.3363e-26, 3.4715e-28, 1.3136e-26, 2.4940e-27, 6.0834e-26,\n 4.5743e-28, 8.1521e-26, 1.5072e-26, 7.5288e-27, 1.7928e-26, 3.5385e-27,\n 5.2724e-26, 2.2483e-27, 2.0483e-26, 1.4933e-27, 7.9534e-26, 8.1597e-27,\n 1.5609e-27, 3.2587e-27, 6.0429e-27, 9.4072e-27, 2.3438e-26, 1.1767e-26,\n 2.0937e-27, 4.8962e-26, 2.4325e-28, 4.3059e-27, 8.2214e-27, 5.3795e-26,\n 1.3994e-26, 3.8857e-26, 1.3853e-26, 9.2912e-27, 4.3637e-27, 4.2315e-26,\n 5.5606e-27, 7.2475e-27, 5.4937e-27, 3.6719e-28, 9.8181e-26, 1.2370e-26,\n 2.4209e-27, 1.3472e-26, 6.0892e-26, 2.9676e-27, 3.2925e-27, 1.6528e-26,\n 3.5591e-26, 1.7349e-26, 2.8383e-26, 6.2820e-27, 6.7104e-28, 1.5352e-27,\n 5.9839e-26, 9.8954e-27, 3.2825e-27, 1.4392e-26, 2.0427e-27, 2.5280e-26,\n 5.1968e-27, 3.3666e-26, 2.5884e-26, 1.9386e-26, 4.6994e-28, 2.1617e-27,\n 4.1773e-27, 1.7474e-26, 3.6575e-27, 2.2982e-27, 3.2597e-26, 2.8555e-26,\n 3.5522e-27, 1.6960e-27, 1.5824e-26, 2.5495e-26, 1.8920e-27, 1.3227e-26,\n 7.9924e-26, 5.9785e-28, 6.9254e-29, 4.7820e-27, 2.8429e-27, 2.0789e-27,\n 2.8899e-27, 1.2011e-26, 2.7494e-27, 1.6941e-26, 4.9254e-28, 7.3434e-26,\n 1.4441e-28, 8.6193e-27, 2.2448e-26, 1.8964e-26, 1.5381e-27, 4.2288e-26,\n 7.4095e-26, 1.8476e-27, 3.0209e-26, 3.1298e-26, 2.3780e-28, 2.1538e-27,\n 2.1518e-28, 2.7916e-27, 1.1674e-27, 8.9944e-26, 6.4280e-26, 6.2551e-28,\n 7.8861e-27, 3.4399e-27, 2.0901e-26, 9.2137e-27, 1.0413e-26, 1.3160e-28,\n 1.2811e-26, 9.5987e-26, 1.1083e-25, 9.3497e-29, 1.0148e-26, 3.4845e-26,\n 6.2481e-26, 2.2477e-28, 1.3083e-25, 9.3093e-27, 3.5509e-26, 1.8015e-26,\n 3.3739e-28, 7.1324e-28, 8.0399e-27, 2.0593e-26, 2.6620e-27, 2.6530e-27,\n 1.8602e-27, 3.2385e-26, 1.6983e-26, 1.0818e-25, 1.3370e-26, 2.7230e-26,\n 1.7741e-27, 3.0487e-26, 1.3088e-25, 1.2296e-27, 1.4183e-25, 3.2451e-28,\n 1.5487e-26, 7.1491e-27, 7.9361e-27, 2.1938e-25, 1.2182e-26, 4.1731e-26,\n 1.0136e-26, 4.0568e-28, 5.3558e-26, 4.2548e-26, 3.6209e-26, 5.7063e-27,\n 1.9173e-26, 1.7071e-26, 4.5799e-27, 1.3302e-26, 8.7126e-26, 1.5429e-26,\n 1.7183e-26, 2.7151e-27, 1.1034e-26, 5.1636e-29, 1.0559e-25, 8.0058e-28,\n 9.2055e-27, 1.2170e-26, 1.0401e-27, 5.1616e-27, 9.5680e-28, 2.9487e-27,\n 1.7370e-26, 2.9603e-27, 6.5024e-28, 2.2391e-27, 3.4465e-28, 7.4554e-27,\n 2.4058e-28, 2.1888e-28, 9.9840e-27, 1.4611e-27, 5.3452e-27, 2.7522e-27,\n 1.9084e-26, 1.5893e-26, 1.5045e-26, 1.1126e-28, 2.3064e-27, 1.4825e-27,\n 8.0455e-27, 2.1993e-25, 2.4796e-27, 7.1328e-27, 6.7421e-28, 1.9311e-26,\n 1.1287e-26, 1.6775e-27, 7.7595e-26, 6.0524e-26, 4.4202e-28, 2.1562e-26,\n 9.8347e-27, 1.0894e-25, 5.4383e-26, 6.5143e-26, 6.9928e-27, 1.3416e-27,\n 4.5448e-26, 5.9331e-26, 2.7076e-27, 1.3383e-26, 6.1565e-26, 2.2868e-26,\n 2.1040e-27, 5.7792e-27, 1.2094e-26, 1.3870e-27, 5.9556e-28, 5.8680e-29,\n 1.4448e-28, 1.4661e-26, 2.6711e-27, 3.5854e-26, 1.1882e-26, 4.4105e-28,\n 9.9855e-27, 3.8710e-27, 1.8208e-27, 4.1097e-29, 1.0315e-26, 8.9166e-27,\n 1.4988e-26, 1.6348e-26, 5.0471e-26, 1.9546e-25], device='cuda:0')"
 },
 "59": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[-5.6052e-45, -5.6052e-45, -5.6052e-45, ..., -5.6052e-45,\n -5.6052e-45, -5.6052e-45],\n [ 5.6052e-45, -5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, -5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n ...,\n [-5.6052e-45, 5.6052e-45, 5.6052e-45, ..., -5.6052e-45,\n -5.6052e-45, -5.6052e-45],\n [ 5.6052e-45, -5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, -5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45]], device='cuda:0')",
 "exp_avg_sq": "tensor([[1.0917e-26, 6.5458e-27, 5.2497e-27, ..., 7.3736e-27, 6.5608e-28,\n 6.0575e-27],\n [2.4368e-28, 3.5802e-28, 9.0866e-28, ..., 3.3611e-28, 5.6170e-29,\n 2.9021e-28],\n [6.3684e-28, 1.7452e-28, 4.3994e-28, ..., 4.0382e-28, 1.6592e-28,\n 2.8858e-28],\n ...,\n [2.7833e-26, 6.0787e-27, 1.5496e-26, ..., 1.1171e-26, 1.6493e-27,\n 5.6792e-27],\n [1.3287e-27, 2.2180e-28, 3.2169e-27, ..., 9.1883e-28, 9.5256e-28,\n 1.1018e-27],\n [8.9873e-28, 2.0251e-27, 1.0501e-28, ..., 6.9802e-28, 5.9107e-28,\n 3.8854e-28]], device='cuda:0')"
 },
 "60": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([-5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([2.5733e-23, 3.2969e-25, 4.8970e-25, 2.8468e-25, 1.7475e-24, 1.8370e-22,\n 1.9938e-24, 4.1569e-24, 1.2310e-23, 4.1366e-24, 8.5235e-24, 2.5638e-23,\n 3.9143e-24, 3.0340e-24, 1.5930e-23, 2.2899e-23, 3.6727e-24, 2.6796e-23,\n 1.8541e-22, 2.3933e-24, 5.1128e-23, 3.2998e-25, 3.3473e-24, 6.7499e-23,\n 4.8429e-23, 6.0676e-24, 6.3903e-23, 6.8742e-23, 1.0227e-23, 7.7508e-24,\n 8.7825e-24, 6.3025e-24, 1.1550e-23, 4.1199e-24, 9.6056e-25, 8.8066e-24,\n 1.0057e-23, 1.0287e-24, 3.8877e-23, 1.5862e-23, 6.2735e-25, 3.7677e-23,\n 5.8247e-24, 1.8837e-24, 8.9455e-25, 1.1726e-23, 2.3117e-23, 5.6676e-23,\n 2.8014e-23, 6.6265e-23, 1.5509e-23, 1.2313e-23, 2.4798e-22, 2.1031e-24,\n 6.4800e-23, 9.4094e-24, 3.6722e-23, 4.1777e-23, 1.2254e-22, 7.0386e-24,\n 4.0804e-22, 6.1516e-24, 1.5937e-23, 1.8191e-23, 2.3016e-25, 2.6234e-24,\n 1.5418e-23, 3.0896e-23, 2.6576e-22, 1.3230e-24, 7.6978e-24, 3.5263e-25,\n 2.5390e-22, 1.3158e-22, 4.0788e-24, 3.2510e-24, 3.2642e-23, 7.4342e-24,\n 1.9960e-23, 2.6819e-24, 1.6172e-23, 4.3300e-25, 1.7606e-24, 5.6461e-24,\n 1.4820e-23, 5.8525e-25, 2.2572e-24, 2.3402e-23, 2.8962e-25, 5.1908e-25,\n 4.9017e-24, 1.9288e-23, 5.6292e-24, 9.9445e-24, 1.6690e-24, 4.3401e-25,\n 3.6386e-23, 2.9660e-23, 1.7974e-23, 2.6654e-23, 3.5462e-25, 7.7442e-24,\n 9.5397e-24, 1.6603e-23, 1.6630e-24, 1.2554e-23, 2.0282e-24, 1.6956e-24,\n 1.5798e-24, 9.0842e-25, 5.9589e-23, 1.3621e-24, 2.1623e-24, 1.0250e-23,\n 3.7695e-23, 1.6814e-23, 1.8736e-23, 2.3686e-23, 2.4571e-23, 1.4399e-23,\n 1.5607e-24, 9.1314e-24, 1.2946e-24, 1.4214e-23, 3.8985e-24, 8.3366e-24,\n 6.2461e-25, 1.8488e-24, 4.5383e-23, 2.0391e-23, 9.1771e-24, 2.2768e-23,\n 7.2299e-24, 6.6122e-23, 1.1832e-23, 7.8241e-26, 1.0826e-24, 9.3943e-23,\n 5.7878e-23, 2.3101e-24, 2.5607e-23, 1.5351e-23, 5.5483e-25, 3.3414e-24,\n 1.0142e-24, 1.5377e-23, 3.1992e-24, 1.6349e-23, 3.4631e-23, 2.6460e-24,\n 8.4849e-24, 2.0326e-24, 6.8184e-24, 5.1501e-25, 2.5174e-24, 2.8493e-23,\n 1.1598e-22, 2.4889e-24, 1.5565e-22, 4.7466e-24, 1.1488e-23, 5.7549e-25,\n 1.0816e-23, 2.5451e-25, 6.3776e-24, 9.7183e-24, 4.8819e-23, 1.0184e-23,\n 1.5606e-24, 9.8726e-24, 8.6695e-24, 2.2463e-23, 2.2461e-25, 5.1975e-25,\n 8.5226e-23, 9.3549e-24, 7.8909e-25, 1.9830e-25, 1.9592e-23, 3.0774e-25,\n 4.1948e-24, 5.0889e-25, 2.3266e-23, 6.4199e-24, 5.3285e-23, 4.6181e-23,\n 2.7498e-24, 2.2812e-24, 3.4260e-25, 6.0605e-23, 1.2679e-24, 2.2886e-23,\n 3.3008e-23, 2.8987e-25, 8.0908e-24, 2.8576e-24, 1.0893e-24, 2.9643e-23,\n 4.6828e-23, 3.4090e-23, 2.7148e-23, 1.6895e-22, 1.6355e-24, 5.5725e-23,\n 2.7103e-24, 2.2186e-23, 4.0460e-23, 2.2460e-24, 2.8559e-24, 9.3633e-24,\n 8.2195e-23, 1.6744e-24, 1.3881e-24, 2.0567e-23, 1.0616e-23, 1.0369e-23,\n 9.1755e-25, 9.5252e-25, 5.2018e-23, 4.0103e-24, 2.1484e-23, 5.9355e-24,\n 6.5537e-25, 8.3256e-23, 1.7211e-24, 1.7838e-23, 9.0358e-24, 3.1559e-24,\n 2.8864e-24, 3.5120e-24, 2.0712e-23, 1.1081e-23, 5.3876e-24, 2.8592e-24,\n 6.4743e-24, 1.1205e-23, 2.0901e-23, 7.8920e-25, 9.9316e-25, 5.3506e-25,\n 1.4115e-23, 1.0465e-23, 4.8841e-25, 8.2507e-24, 8.1958e-24, 8.3456e-24,\n 1.3405e-23, 3.2943e-24, 2.2000e-23, 2.4712e-23, 3.1175e-25, 5.0843e-24,\n 4.5974e-25, 4.9697e-23, 4.5689e-24, 7.1611e-25], device='cuda:0')"
 },
 "61": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([2.0296e-25, 8.7909e-28, 3.3745e-27, 6.3786e-28, 8.5727e-28, 7.2105e-25,\n 4.4848e-27, 3.8413e-27, 6.4692e-27, 2.6326e-27, 1.0010e-26, 1.6998e-25,\n 1.1609e-27, 1.8983e-27, 1.5035e-26, 6.4713e-26, 3.3211e-27, 7.2144e-26,\n 1.8111e-24, 3.5939e-27, 3.7905e-25, 4.3656e-28, 7.4641e-27, 1.0674e-24,\n 5.1567e-26, 3.8975e-27, 9.3277e-25, 9.2281e-26, 6.9544e-27, 2.2081e-26,\n 1.1318e-26, 1.9144e-25, 1.9854e-27, 1.2392e-26, 5.9875e-28, 1.0953e-25,\n 1.3845e-26, 1.5954e-27, 5.4391e-25, 3.5067e-25, 1.0623e-27, 2.0441e-25,\n 6.6739e-27, 2.6485e-27, 1.4737e-27, 3.3828e-26, 1.0494e-26, 4.6082e-26,\n 4.0104e-26, 1.3529e-25, 1.5434e-28, 1.4379e-25, 1.4234e-24, 5.7850e-28,\n 3.1906e-25, 6.7444e-26, 9.3983e-26, 3.7692e-26, 3.2374e-25, 2.5148e-27,\n 1.9328e-24, 9.0669e-27, 9.1084e-27, 1.7983e-25, 1.6675e-27, 2.5977e-27,\n 2.3296e-26, 4.7736e-25, 6.3529e-25, 1.0276e-27, 6.3495e-27, 4.1938e-27,\n 8.0429e-25, 5.4210e-25, 5.3762e-27, 2.4515e-27, 4.0251e-26, 5.1272e-26,\n 1.8087e-26, 6.0812e-28, 5.3196e-26, 1.2580e-27, 2.6457e-27, 5.9069e-27,\n 1.7943e-25, 5.0600e-27, 1.1492e-26, 5.6013e-26, 1.4069e-27, 3.4445e-28,\n 8.0506e-26, 3.4444e-25, 2.5907e-26, 2.1321e-26, 1.1824e-27, 9.6007e-28,\n 8.7784e-26, 7.4863e-26, 3.2240e-26, 4.0439e-27, 2.6279e-28, 4.4801e-27,\n 8.5540e-26, 3.2498e-25, 6.7571e-27, 5.4019e-26, 4.4436e-26, 3.8476e-27,\n 5.0812e-28, 3.1360e-28, 5.5916e-26, 1.6303e-27, 1.6947e-26, 3.1297e-25,\n 2.3413e-25, 2.6338e-25, 1.0304e-25, 5.2647e-26, 2.4455e-26, 3.6271e-27,\n 4.0767e-28, 2.6552e-27, 1.2860e-27, 2.5620e-27, 4.1489e-26, 7.0042e-27,\n 3.9677e-28, 1.7125e-27, 1.6363e-25, 8.5821e-27, 3.6175e-26, 1.2630e-26,\n 4.0991e-27, 1.6424e-24, 9.2984e-26, 9.8529e-28, 2.7155e-27, 1.4847e-24,\n 1.8010e-25, 1.8703e-27, 1.7517e-26, 7.0279e-27, 7.6549e-28, 2.5037e-27,\n 5.4066e-28, 1.6441e-27, 3.0049e-27, 6.9924e-26, 5.9698e-25, 2.0397e-27,\n 8.0967e-27, 5.5473e-27, 6.7175e-26, 1.1560e-27, 9.5012e-27, 2.9980e-25,\n 2.6091e-25, 4.0692e-27, 2.5528e-25, 1.5736e-27, 2.4951e-27, 1.2499e-27,\n 7.2162e-27, 1.4775e-28, 5.0989e-28, 7.7794e-27, 6.0481e-26, 4.1331e-27,\n 5.7870e-28, 9.8255e-27, 7.4191e-27, 6.4844e-26, 5.2643e-28, 1.0946e-27,\n 1.1934e-25, 3.8752e-27, 7.4769e-28, 1.0341e-27, 7.0489e-26, 4.2830e-28,\n 1.2899e-26, 8.2175e-28, 8.2158e-27, 3.0127e-26, 5.5707e-26, 1.1815e-25,\n 6.7566e-26, 1.2891e-27, 1.7447e-28, 8.8053e-26, 2.7477e-26, 8.0804e-27,\n 3.0717e-25, 3.1385e-27, 2.6286e-26, 4.3918e-26, 2.8001e-27, 5.5654e-26,\n 2.1024e-25, 3.7978e-26, 4.6125e-25, 5.5999e-25, 1.9053e-27, 5.8211e-26,\n 6.5420e-28, 1.5426e-25, 2.1622e-26, 2.3383e-27, 1.8254e-28, 3.2317e-26,\n 3.7232e-25, 2.0943e-27, 1.1064e-27, 1.2916e-26, 1.1689e-26, 2.1960e-26,\n 1.3991e-27, 3.4845e-27, 7.5464e-25, 1.2294e-27, 1.2561e-26, 8.3279e-27,\n 7.8370e-28, 5.1108e-25, 9.1059e-27, 1.0371e-26, 1.2475e-26, 2.4759e-26,\n 3.9544e-28, 1.8443e-26, 1.0023e-25, 3.3331e-25, 2.6126e-25, 2.6528e-27,\n 5.3250e-26, 7.1481e-27, 1.4027e-25, 6.4417e-27, 1.1625e-27, 8.1098e-27,\n 1.4394e-26, 4.2313e-26, 1.2239e-27, 1.7917e-26, 9.9084e-29, 1.1997e-26,\n 6.6713e-27, 6.1880e-27, 1.6780e-25, 2.4122e-26, 1.8663e-27, 2.6590e-27,\n 9.8327e-28, 2.3940e-25, 2.0651e-27, 2.8201e-27], device='cuda:0')"
 },
 "62": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([1.3973e-25, 5.8629e-27, 3.5802e-28, 1.8861e-28, 1.1961e-26, 2.9460e-25,\n 2.2036e-26, 2.6628e-26, 7.7985e-26, 2.9544e-26, 6.0548e-26, 1.0255e-25,\n 1.6387e-26, 7.1684e-27, 3.1904e-26, 9.1040e-26, 2.2924e-26, 4.8784e-26,\n 8.5368e-25, 1.8717e-26, 1.7572e-25, 1.4158e-27, 1.8193e-26, 4.0638e-25,\n 8.0785e-26, 4.5764e-26, 2.6661e-25, 7.1242e-26, 4.2976e-26, 4.8817e-26,\n 1.2489e-26, 6.0628e-26, 4.2780e-27, 6.6759e-26, 3.4440e-27, 3.9225e-26,\n 4.2055e-26, 1.7655e-26, 2.0918e-25, 1.3034e-25, 6.1457e-27, 2.3278e-25,\n 3.8227e-26, 1.7194e-26, 3.1556e-27, 7.7930e-26, 1.1121e-25, 7.6806e-26,\n 1.6559e-26, 1.3641e-25, 1.6470e-26, 8.2983e-26, 1.0659e-24, 1.2736e-26,\n 3.1083e-25, 8.1140e-26, 2.0758e-25, 1.8728e-25, 3.2561e-25, 2.7200e-26,\n 7.6970e-25, 1.6553e-26, 6.9944e-26, 1.2975e-25, 6.7606e-28, 2.6271e-26,\n 1.0549e-26, 1.8020e-25, 5.2869e-25, 1.7918e-26, 2.5858e-26, 1.2618e-27,\n 3.3853e-25, 2.7391e-25, 1.8108e-26, 3.0422e-26, 1.4226e-25, 9.0588e-26,\n 3.4116e-26, 2.7892e-26, 1.1638e-25, 8.6439e-27, 2.7416e-26, 4.9498e-27,\n 5.2928e-26, 8.1622e-27, 2.4845e-26, 1.2166e-25, 8.4291e-27, 1.7631e-27,\n 7.3721e-26, 8.8967e-26, 3.5970e-26, 3.2123e-26, 1.0844e-26, 4.5172e-27,\n 1.8387e-25, 1.7171e-25, 2.6843e-26, 3.4783e-26, 1.6138e-27, 1.3245e-26,\n 5.6698e-26, 1.7340e-25, 1.5702e-26, 3.1901e-26, 5.4051e-26, 1.3766e-26,\n 5.8017e-27, 1.6629e-26, 7.2665e-26, 1.3479e-26, 2.5516e-26, 1.1306e-25,\n 5.3568e-26, 1.6626e-25, 1.3361e-25, 8.1281e-26, 1.0182e-25, 7.8596e-26,\n 1.4096e-26, 5.8422e-26, 7.3619e-27, 4.6295e-26, 6.0421e-26, 2.0150e-26,\n 2.8149e-27, 2.6971e-26, 4.0189e-26, 2.0755e-26, 3.4693e-26, 5.2312e-26,\n 3.8503e-26, 4.2028e-25, 5.3918e-26, 4.3294e-28, 1.7084e-26, 4.5569e-25,\n 2.9802e-25, 1.9666e-26, 1.1786e-25, 1.8025e-26, 1.1595e-28, 1.1295e-26,\n 5.6768e-27, 1.7252e-26, 9.3749e-27, 6.1881e-26, 1.5535e-25, 1.0238e-26,\n 3.8206e-26, 3.1023e-26, 2.8837e-26, 1.0042e-26, 2.5407e-26, 1.6818e-25,\n 2.1788e-25, 1.1978e-26, 5.7339e-25, 1.2552e-26, 6.2381e-26, 1.3764e-28,\n 2.8640e-26, 2.3329e-28, 1.5315e-27, 3.5127e-26, 1.5251e-25, 3.3740e-26,\n 1.9798e-28, 2.4297e-26, 3.1435e-26, 1.2183e-25, 5.2651e-27, 5.1993e-27,\n 3.6675e-25, 4.4552e-26, 1.3732e-27, 7.0117e-28, 8.6838e-26, 1.7643e-27,\n 1.3347e-26, 4.0709e-27, 1.2469e-25, 8.8999e-26, 2.5034e-25, 2.2393e-25,\n 4.4681e-26, 2.8589e-26, 1.1573e-28, 3.7198e-26, 1.0488e-26, 4.7682e-26,\n 1.2125e-25, 6.1684e-27, 5.2764e-26, 2.0044e-26, 8.1741e-27, 8.1996e-26,\n 1.2277e-25, 1.3203e-25, 1.7706e-25, 2.1935e-25, 7.8911e-27, 4.9893e-26,\n 1.7429e-26, 1.5806e-25, 3.9389e-26, 5.9728e-27, 4.8519e-26, 7.3228e-26,\n 2.5918e-25, 1.8723e-26, 9.0196e-27, 8.1153e-26, 7.0221e-26, 1.7316e-26,\n 1.1143e-27, 1.3679e-26, 3.2611e-25, 1.6535e-26, 4.4945e-26, 3.6044e-26,\n 1.3955e-26, 4.2512e-25, 7.8756e-27, 5.1350e-26, 7.0828e-26, 2.1584e-26,\n 3.1590e-27, 5.7694e-26, 4.3346e-26, 1.4220e-25, 1.1211e-25, 6.7529e-27,\n 7.1001e-26, 1.0210e-25, 1.4259e-25, 8.3803e-27, 1.8904e-26, 2.0121e-26,\n 7.0846e-26, 8.6022e-26, 3.3173e-28, 2.5633e-26, 4.8933e-27, 2.3650e-26,\n 9.8552e-27, 1.3168e-26, 1.1878e-25, 1.0431e-25, 2.7585e-27, 3.1188e-26,\n 5.6293e-27, 2.0130e-25, 3.4711e-26, 4.8706e-27], device='cuda:0')"
 },
 "63": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[-5.6052e-45, -5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, -5.6052e-45],\n [-5.6052e-45, 5.6052e-45, -5.6052e-45, ..., -5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n [-5.6052e-45, 5.6052e-45, -5.6052e-45, ..., -5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n ...,\n [ 5.6052e-45, -5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, -5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, 5.6052e-45],\n [-5.6052e-45, 5.6052e-45, -5.6052e-45, ..., -5.6052e-45,\n 5.6052e-45, 5.6052e-45]], device='cuda:0')",
 "exp_avg_sq": "tensor([[4.0834e-27, 1.4822e-28, 9.4058e-28, ..., 1.4207e-27, 1.3117e-26,\n 2.4903e-28],\n [1.2679e-27, 2.3953e-28, 2.8271e-27, ..., 5.1746e-27, 6.2919e-27,\n 8.5589e-28],\n [4.5593e-28, 2.2110e-29, 1.2983e-27, ..., 2.1118e-27, 2.7554e-27,\n 1.2372e-27],\n ...,\n [1.2560e-24, 3.0536e-26, 9.5106e-25, ..., 6.6263e-25, 3.4440e-24,\n 7.4993e-25],\n [6.4935e-25, 4.5598e-26, 7.7528e-25, ..., 5.1764e-25, 2.1977e-24,\n 1.1895e-25],\n [4.3822e-25, 2.6109e-26, 2.6448e-25, ..., 3.5527e-25, 1.2475e-24,\n 4.6695e-26]], device='cuda:0')"
 },
 "64": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([-5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([7.4748e-27, 1.1298e-26, 5.3425e-27, 1.6958e-26, 6.2094e-27, 3.6746e-27,\n 1.6357e-26, 8.5526e-27, 6.3374e-27, 1.5139e-26, 7.2906e-27, 2.1704e-27,\n 2.2588e-26, 1.5329e-26, 6.0189e-27, 3.0312e-26, 2.9792e-26, 3.6498e-26,\n 5.4478e-27, 8.3163e-27, 7.4467e-27, 1.7384e-26, 1.2894e-26, 5.4853e-27,\n 5.8038e-27, 1.1889e-26, 3.1290e-27, 1.3472e-26, 1.3703e-26, 3.3597e-26,\n 1.1961e-26, 5.5306e-27, 2.4937e-26, 4.9956e-27, 3.5794e-27, 4.0165e-27,\n 1.1432e-26, 2.2634e-26, 6.1240e-26, 3.3527e-26, 2.8543e-27, 5.8835e-27,\n 1.2184e-26, 3.1786e-26, 5.0783e-27, 6.4494e-27, 1.0433e-26, 1.7480e-27,\n 6.9880e-27, 3.9316e-27, 6.4513e-27, 2.5378e-26, 6.5044e-28, 3.6018e-26,\n 7.2832e-27, 1.4467e-27, 8.7052e-27, 9.4675e-27, 1.8188e-26, 9.0332e-27,\n 9.4774e-27, 1.0216e-26, 7.3233e-28, 1.2473e-26, 7.0597e-27, 4.5022e-27,\n 1.2397e-26, 8.3295e-27, 6.3387e-27, 6.5585e-27, 3.7793e-27, 8.8765e-27,\n 6.1479e-28, 2.2041e-26, 1.1284e-26, 3.1317e-26, 4.7213e-27, 6.8760e-27,\n 1.3239e-26, 7.7002e-27, 1.7695e-27, 1.3318e-27, 7.2719e-27, 5.9788e-27,\n 9.9174e-27, 7.9435e-27, 4.7173e-27, 3.9487e-27, 3.0061e-26, 2.2189e-26,\n 4.4906e-27, 1.6147e-27, 1.9462e-26, 3.3639e-26, 5.4227e-27, 4.7485e-27,\n 7.5411e-27, 5.7435e-27, 1.0750e-26, 2.2140e-26, 3.7181e-27, 9.3550e-27,\n 4.5653e-28, 1.1240e-26, 4.9396e-27, 7.6488e-27, 6.0411e-27, 3.6758e-27,\n 4.9463e-27, 2.4529e-27, 1.1480e-26, 3.9760e-27, 3.0093e-27, 6.8613e-27,\n 2.9479e-27, 3.4704e-27, 6.9158e-27, 5.9024e-28, 3.0471e-27, 1.7741e-26,\n 1.2684e-26, 1.1382e-27, 1.1845e-26, 4.4511e-27, 6.7613e-27, 8.9535e-27,\n 6.5863e-27, 2.1982e-26, 1.1487e-26, 2.4686e-27, 8.1302e-27, 2.7930e-27,\n 1.7348e-26, 1.0842e-26, 3.2311e-27, 1.2731e-26, 1.0133e-26, 1.1172e-26,\n 8.1404e-27, 3.8881e-27, 1.8009e-27, 8.2301e-27, 4.1616e-27, 8.1558e-27,\n 8.8675e-28, 7.7215e-27, 2.4582e-27, 4.7271e-27, 2.7467e-27, 4.7021e-27,\n 5.2968e-27, 7.5945e-27, 3.7859e-27, 3.6775e-27, 4.2151e-28, 7.4372e-27,\n 6.7131e-27, 3.1329e-27, 1.1977e-26, 6.4490e-27, 8.3476e-27, 5.7314e-27,\n 1.2755e-26, 3.9982e-27, 1.5245e-27, 2.4855e-27, 9.5017e-27, 1.1375e-26,\n 5.6359e-27, 2.6749e-27, 5.6970e-27, 1.2734e-26, 9.4598e-27, 7.9890e-27,\n 4.2319e-27, 1.9679e-27, 4.4991e-27, 1.2284e-26, 6.2421e-27, 3.0752e-27,\n 5.8363e-27, 7.1149e-27, 6.9684e-27, 1.4281e-27, 3.7521e-27, 2.5422e-27,\n 2.9646e-27, 1.3246e-26, 2.5541e-27, 3.7545e-27, 5.8159e-27, 4.1996e-27,\n 9.8859e-27, 1.6239e-28, 1.6924e-27, 3.5447e-27, 2.6673e-26, 3.9380e-27,\n 2.5467e-27, 1.5030e-27, 5.1862e-26, 3.9339e-27, 1.4202e-27, 5.7434e-27,\n 2.1122e-26, 1.8554e-26, 2.5539e-26, 4.5297e-27, 6.8970e-27, 1.1799e-26,\n 3.2743e-27, 5.5963e-27, 2.5075e-26, 3.4573e-27, 7.8124e-27, 2.3457e-26,\n 4.4677e-27, 3.4468e-28, 6.8034e-27, 5.2494e-28, 3.6060e-27, 1.6920e-26,\n 1.2368e-26, 3.7405e-26, 3.6653e-27, 9.3700e-28, 5.8893e-27, 7.6217e-27,\n 3.0607e-27, 4.3406e-27, 4.0084e-28, 8.8425e-27, 3.7639e-27, 6.7107e-27,\n 3.3339e-27, 5.4767e-27, 3.4442e-27, 6.3417e-27, 1.7758e-26, 1.0854e-26,\n 6.0147e-27, 1.5025e-26, 1.8385e-27, 4.9376e-27, 6.9157e-27, 2.1138e-26,\n 1.3689e-26, 6.1512e-28, 6.5028e-27, 2.0646e-27, 1.0003e-26, 4.2756e-27,\n 4.8377e-27, 4.9496e-27, 8.5690e-27, 4.5195e-27, 1.1674e-34, 4.8614e-35,\n 1.1495e-34, 5.2619e-35, 1.2798e-34, 1.1643e-35, 2.2060e-35, 9.4367e-35,\n 5.7245e-35, 1.4979e-35, 8.6763e-35, 5.9197e-35, 1.0637e-34, 1.0208e-34,\n 1.1170e-34, 4.9121e-35, 1.2036e-34, 1.1517e-34, 9.0153e-35, 7.9756e-35,\n 6.3215e-35, 2.3795e-35, 2.2130e-35, 1.3290e-35, 2.0610e-35, 5.2910e-35,\n 2.8303e-35, 1.0949e-34, 7.7288e-35, 1.0095e-35, 5.0328e-35, 6.5626e-36,\n 3.3859e-35, 2.7208e-35, 3.6104e-35, 1.6355e-35, 4.3101e-35, 6.5800e-35,\n 1.7331e-34, 5.0685e-35, 1.3073e-35, 2.2822e-35, 5.1284e-35, 1.4741e-35,\n 1.5316e-35, 3.4508e-35, 3.3158e-35, 4.4196e-36, 2.9558e-35, 2.3273e-35,\n 3.3251e-35, 3.0097e-35, 5.2333e-35, 6.9247e-36, 3.8580e-36, 2.0473e-35,\n 2.7171e-36, 7.6593e-36, 4.0366e-35, 3.5256e-35, 2.4231e-35, 3.2480e-36,\n 1.6286e-35, 3.8582e-35, 3.4160e-35, 3.1474e-35, 1.2306e-35, 4.8126e-35,\n 5.9011e-35, 3.6104e-35, 1.1598e-34, 1.3265e-34, 5.6215e-35, 4.5867e-36,\n 1.2023e-34, 4.2230e-35, 2.7981e-34, 5.8516e-35, 1.2947e-34, 3.3048e-35,\n 6.3662e-35, 8.7726e-35, 3.4005e-35, 7.0760e-35, 7.1658e-35, 3.2677e-35,\n 6.8975e-35, 2.3322e-35, 6.0293e-35, 2.9508e-35, 8.2026e-35, 1.4016e-35,\n 1.5463e-35, 1.0759e-34, 9.2387e-35, 5.1763e-35, 1.4786e-35, 5.9342e-35,\n 1.7711e-35, 7.0109e-35, 3.4290e-35, 3.5746e-35, 4.0295e-35, 7.2782e-35,\n 6.4706e-35, 2.5676e-35, 6.5527e-35, 3.8257e-35, 2.4152e-35, 9.5624e-35,\n 4.1889e-35, 7.5908e-35, 7.4129e-35, 2.6178e-35, 8.2319e-36, 5.4920e-35,\n 7.4490e-35, 2.5211e-35, 1.8975e-35, 2.6005e-35, 5.4737e-36, 4.9170e-35,\n 7.4334e-35, 7.7408e-35, 2.0742e-35, 3.3885e-35, 5.4538e-35, 1.0823e-35,\n 1.7244e-34, 8.2800e-35, 7.6205e-35, 1.4241e-35, 7.3848e-35, 5.5361e-35,\n 1.0462e-34, 1.0967e-35, 7.8765e-35, 3.8862e-35, 2.7838e-36, 4.2798e-35,\n 4.0609e-35, 1.1412e-35, 3.1409e-35, 3.5564e-35, 3.9718e-35, 6.2989e-35,\n 1.1873e-34, 9.9520e-36, 2.2189e-35, 7.4690e-36, 3.7335e-35, 2.0361e-35,\n 6.8657e-35, 2.2543e-35, 9.1446e-35, 5.6348e-35, 3.8322e-35, 4.7737e-35,\n 1.3033e-35, 5.6441e-35, 6.3268e-36, 3.2872e-35, 5.0201e-35, 6.1957e-35,\n 2.6366e-35, 4.6205e-35, 4.0683e-35, 5.3215e-35, 1.0365e-34, 9.9308e-35,\n 5.1629e-35, 1.2128e-34, 4.5562e-35, 1.9686e-35, 4.7408e-35, 1.8954e-35,\n 8.4234e-36, 5.2878e-35, 3.7965e-35, 4.4341e-35, 4.6620e-35, 2.5867e-35,\n 8.7463e-35, 1.2206e-35, 7.2376e-35, 5.5923e-35, 3.4091e-35, 2.7296e-35,\n 3.0571e-35, 5.3669e-35, 3.4919e-35, 7.3569e-35, 2.5757e-35, 3.9045e-36,\n 4.0500e-35, 7.6624e-35, 1.4045e-34, 1.2276e-35, 7.1778e-35, 4.5911e-35,\n 1.4616e-35, 5.3945e-35, 1.0724e-34, 1.0977e-34, 3.6489e-35, 2.5282e-35,\n 2.7707e-35, 1.6866e-35, 8.1448e-35, 1.8994e-35, 1.3647e-35, 1.2692e-34,\n 6.5548e-35, 1.7804e-35, 1.8697e-35, 2.9647e-35, 4.4206e-36, 2.2466e-35,\n 2.5373e-35, 1.5387e-34, 5.8565e-35, 1.0143e-35, 2.2074e-35, 2.9427e-35,\n 6.3082e-35, 7.9779e-35, 1.2126e-34, 1.6427e-35, 1.1024e-34, 3.4530e-35,\n 2.4330e-35, 2.9402e-34, 1.4795e-34, 5.4908e-35, 1.4457e-34, 1.2240e-34,\n 3.9305e-35, 1.6623e-34, 6.7841e-35, 4.7143e-35, 2.6181e-34, 4.6783e-35,\n 3.7403e-35, 1.0509e-34, 1.0982e-34, 1.0211e-34, 3.3841e-34, 6.8407e-35,\n 1.0637e-34, 1.5551e-35, 3.3336e-34, 5.6758e-35, 7.2873e-35, 1.4250e-34,\n 2.7511e-35, 8.6297e-35, 1.6797e-23, 3.7121e-24, 2.8760e-24, 8.6997e-24,\n 9.7461e-24, 9.4230e-25, 5.9389e-24, 9.8330e-25, 1.0750e-23, 2.2155e-24,\n 6.0356e-24, 3.0776e-24, 6.7494e-24, 2.3995e-24, 4.8490e-24, 2.4081e-24,\n 6.5988e-25, 1.1666e-24, 7.8454e-25, 1.3816e-24, 1.6602e-23, 2.7485e-24,\n 6.8149e-24, 5.3021e-24, 6.6619e-25, 6.7036e-25, 3.8823e-24, 3.8416e-24,\n 8.7666e-25, 1.1661e-23, 5.9525e-24, 2.8653e-24, 6.0601e-25, 2.4961e-23,\n 4.2736e-24, 1.2740e-23, 9.3094e-25, 1.5771e-24, 6.0382e-24, 4.0953e-24,\n 1.9834e-24, 9.1812e-24, 2.8222e-24, 8.5843e-24, 1.7696e-24, 5.9887e-25,\n 2.6366e-24, 5.5874e-25, 5.1216e-24, 1.1289e-23, 2.1839e-24, 7.3299e-24,\n 4.2486e-24, 3.4966e-24, 5.0739e-24, 4.4253e-24, 8.0293e-25, 2.3413e-24,\n 8.6214e-25, 7.1324e-24, 1.1797e-23, 2.1612e-24, 3.9144e-24, 1.7666e-23,\n 2.4453e-24, 4.8938e-25, 3.8130e-24, 3.3400e-24, 2.4129e-24, 4.6281e-24,\n 5.5237e-24, 2.6323e-24, 1.0745e-23, 5.5623e-24, 2.7928e-24, 1.1193e-23,\n 1.0663e-23, 5.4619e-24, 3.1805e-24, 6.6345e-24, 2.3563e-24, 3.0674e-24,\n 3.0191e-24, 1.0835e-23, 5.4105e-24, 1.0412e-23, 6.6100e-25, 1.7455e-24,\n 2.2986e-24, 2.5509e-24, 1.7554e-24, 3.9537e-24, 6.0736e-24, 1.7911e-24,\n 1.2900e-24, 1.6244e-23, 1.1603e-24, 9.0968e-25, 9.4222e-24, 5.9648e-24,\n 4.3253e-24, 4.7655e-25, 7.3875e-24, 8.9678e-24, 3.6610e-24, 3.8428e-24,\n 2.5513e-24, 3.0111e-24, 2.3224e-24, 4.0181e-24, 2.1152e-24, 7.2876e-24,\n 2.8020e-24, 3.6608e-24, 3.5083e-24, 9.6263e-24, 7.8298e-25, 1.3055e-23,\n 4.4023e-24, 3.7070e-24, 1.1707e-23, 2.3992e-24, 8.2942e-24, 4.1160e-24,\n 3.5141e-24, 1.1431e-24, 9.5042e-24, 1.9757e-24, 1.4323e-24, 1.5816e-24,\n 7.0126e-24, 9.2675e-24, 3.0830e-24, 2.3637e-24, 7.1993e-24, 7.1535e-25,\n 7.6275e-25, 2.1418e-24, 8.2901e-24, 3.6317e-24, 2.3594e-24, 2.1496e-24,\n 5.9147e-24, 1.6285e-24, 5.4257e-24, 3.4829e-25, 3.6460e-24, 1.0596e-23,\n 2.2441e-24, 3.0670e-24, 8.2082e-24, 5.1787e-24, 3.3738e-24, 9.5891e-24,\n 8.9281e-24, 9.6757e-24, 8.8068e-25, 3.6381e-24, 1.6590e-24, 2.5927e-24,\n 2.8086e-24, 3.8570e-24, 6.0992e-24, 1.0686e-23, 1.8520e-24, 1.5429e-24,\n 7.6104e-25, 1.3121e-24, 5.5887e-24, 1.3324e-24, 1.1738e-23, 6.9858e-24,\n 2.1250e-24, 4.0819e-24, 1.8993e-24, 3.4524e-24, 2.8957e-24, 1.0980e-24,\n 7.8109e-24, 3.3378e-24, 5.4347e-24, 3.9800e-24, 2.2877e-24, 3.7584e-24,\n 4.7761e-24, 2.5101e-24, 2.0368e-24, 6.6905e-24, 1.3077e-25, 3.3522e-24,\n 6.1818e-24, 1.1241e-23, 2.0240e-24, 2.8990e-24, 9.7526e-24, 7.0068e-24,\n 5.7461e-25, 1.2740e-24, 6.6858e-25, 7.3395e-25, 4.2194e-24, 8.5264e-24,\n 3.5233e-25, 8.5983e-24, 3.8195e-24, 1.3262e-23, 1.1934e-23, 5.9309e-24,\n 8.4250e-24, 1.9546e-24, 2.8497e-24, 5.6343e-25, 5.1965e-24, 1.4530e-24,\n 1.2978e-24, 1.4915e-23, 5.6407e-24, 3.9019e-24, 1.3055e-23, 2.6100e-24,\n 2.7149e-24, 2.2660e-24, 4.9831e-24, 8.3108e-24, 6.8181e-24, 8.7143e-24,\n 5.6939e-24, 6.6160e-24, 2.7847e-25, 3.5621e-24, 2.8730e-24, 2.1793e-24,\n 6.5991e-25, 4.5098e-24, 5.5695e-24, 5.0674e-24, 1.8703e-24, 5.2012e-24,\n 5.5172e-24, 4.0417e-24, 3.3509e-24, 2.9313e-25, 4.4073e-24, 1.4603e-24,\n 4.3793e-24, 6.2101e-24, 6.6571e-24, 3.4843e-24, 9.7259e-25, 2.8287e-24,\n 4.6970e-24, 8.2606e-24, 4.5497e-24, 9.8795e-24, 4.0340e-24, 1.9304e-24],\n device='cuda:0')"
 },
 "65": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[-5.6052e-45, -5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, -5.6052e-45],\n [-5.6052e-45, 5.6052e-45, -5.6052e-45, ..., -5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, -5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, -5.6052e-45],\n ...,\n [ 5.6052e-45, -5.6052e-45, 5.6052e-45, ..., -5.6052e-45,\n -5.6052e-45, -5.6052e-45],\n [ 5.6052e-45, -5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n 5.6052e-45, -5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, -5.6052e-45, ..., -5.6052e-45,\n -5.6052e-45, 5.6052e-45]], device='cuda:0')",
 "exp_avg_sq": "tensor([[1.6578e-25, 1.4165e-24, 1.6140e-24, ..., 1.6045e-24, 6.2649e-25,\n 1.2809e-24],\n [2.2923e-25, 4.3249e-24, 5.4146e-24, ..., 7.4231e-24, 1.5251e-24,\n 3.5771e-24],\n [5.3871e-26, 1.4521e-24, 1.4885e-24, ..., 9.5422e-25, 2.8456e-25,\n 1.4063e-24],\n ...,\n [4.0309e-25, 6.9034e-24, 6.3734e-24, ..., 3.3754e-24, 1.5355e-24,\n 7.6274e-24],\n [1.2018e-24, 1.5003e-23, 1.4564e-23, ..., 1.0235e-23, 3.9559e-24,\n 1.5481e-23],\n [3.0802e-26, 9.8663e-25, 1.2242e-24, ..., 1.5508e-24, 3.1101e-25,\n 9.3186e-25]], device='cuda:0')"
 },
 "66": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, -5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45,\n 5.6052e-45, -5.6052e-45, 5.6052e-45, 5.6052e-45, 5.6052e-45,\n -5.6052e-45, 5.6052e-45, 5.6052e-45, -5.6052e-45, 5.6052e-45,\n -5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([1.8723e-24, 5.7553e-24, 1.3027e-24, 7.5839e-24, 1.3906e-23, 2.7640e-24,\n 7.2696e-25, 3.1943e-24, 6.3476e-25, 3.0497e-24, 7.8929e-24, 6.0264e-24,\n 1.0325e-24, 6.5194e-24, 3.4898e-24, 2.2126e-24, 1.0174e-23, 2.0337e-24,\n 6.4771e-24, 9.4692e-24, 4.0947e-24, 3.5371e-24, 5.1845e-24, 1.3406e-23,\n 3.0156e-25, 2.4730e-24, 6.6838e-24, 1.3530e-23, 9.1633e-24, 6.2538e-24,\n 7.5289e-24, 3.8999e-24, 6.8632e-24, 2.6026e-24, 8.2056e-24, 1.0190e-23,\n 7.5633e-24, 1.0451e-23, 5.7309e-24, 7.0944e-24, 3.1346e-23, 5.6886e-24,\n 6.4719e-24, 2.4498e-24, 7.4908e-24, 1.4696e-23, 2.9631e-24, 6.6864e-24,\n 2.8909e-24, 1.2360e-23, 2.4155e-23, 6.8642e-24, 7.3189e-24, 2.9332e-24,\n 7.4326e-24, 1.6729e-23, 9.3267e-24, 3.8397e-24, 5.4636e-24, 5.9348e-24,\n 7.3934e-24, 2.4916e-24, 9.9633e-24, 5.9302e-24, 2.1948e-24, 8.6382e-24,\n 1.1580e-24, 6.2862e-24, 4.9478e-24, 1.6417e-24, 1.4625e-24, 3.9888e-24,\n 9.7773e-24, 1.5844e-24, 2.6441e-24, 3.2497e-24, 6.4720e-24, 2.5760e-24,\n 1.0771e-23, 3.4264e-24, 6.8651e-24, 1.8858e-24, 7.1458e-24, 5.5147e-24,\n 2.1177e-24, 2.0028e-24, 1.0918e-23, 4.7265e-24, 5.8449e-24, 2.4413e-23,\n 8.4584e-24, 4.9500e-24, 3.4355e-24, 7.7454e-24, 1.6631e-23, 1.9771e-23,\n 9.6449e-25, 2.5688e-24, 1.1386e-23, 4.7883e-24, 3.4508e-24, 8.2803e-24,\n 6.1597e-24, 1.4280e-23, 4.0978e-24, 6.7846e-24, 3.9499e-24, 4.0210e-24,\n 3.4048e-24, 1.5153e-24, 4.4794e-24, 4.9629e-24, 5.2663e-24, 2.3907e-24,\n 2.8899e-24, 1.4588e-23, 9.4349e-25, 1.3301e-24, 4.9071e-24, 1.5249e-24,\n 1.2431e-23, 1.9453e-24, 1.2052e-23, 5.4796e-24, 6.3124e-24, 5.3035e-24,\n 4.3938e-24, 4.3202e-24, 1.1791e-24, 2.9287e-24, 2.7038e-24, 4.8613e-24,\n 4.6981e-24, 6.4254e-24, 2.3313e-24, 4.3591e-24, 2.1327e-25, 8.2241e-24,\n 5.4190e-24, 6.8438e-24, 2.1290e-24, 3.7813e-24, 7.7501e-24, 1.5939e-23,\n 5.0136e-24, 6.8326e-25, 9.1918e-25, 5.8980e-24, 5.1776e-24, 6.4216e-24,\n 8.5492e-24, 9.4281e-24, 1.3316e-23, 9.6935e-24, 4.1813e-24, 9.7211e-24,\n 6.8487e-24, 1.8178e-24, 5.0791e-24, 6.4727e-24, 1.6490e-23, 9.7821e-25,\n 9.7440e-24, 3.3393e-24, 6.2329e-24, 9.6970e-24, 2.5031e-24, 1.3537e-23,\n 1.2402e-23, 7.9411e-24, 6.3151e-24, 2.5013e-24, 3.2343e-24, 3.3939e-23,\n 7.2235e-24, 4.1319e-23, 2.6973e-24, 8.0275e-24, 3.6748e-24, 3.5663e-24,\n 2.9165e-24, 4.3862e-24, 5.0311e-24, 6.5058e-24, 5.2660e-24, 4.9208e-24,\n 3.6282e-24, 1.0486e-23, 3.0142e-24, 2.7000e-23, 6.7622e-24, 6.6074e-24,\n 1.0458e-23, 5.1676e-24, 2.3182e-23, 3.3804e-24, 2.7068e-24, 6.8537e-24,\n 2.0773e-24, 6.5914e-24, 6.0514e-25, 5.0596e-24, 5.8832e-24, 3.5234e-24,\n 5.5038e-24, 5.2155e-24, 7.4311e-24, 7.5439e-24, 5.6946e-24, 1.3689e-23,\n 8.6267e-24, 8.8296e-24, 1.7473e-24, 1.3684e-23, 2.3993e-23, 1.6099e-24,\n 3.7238e-24, 8.5844e-24, 1.9180e-23, 2.0070e-23, 4.8814e-24, 1.5122e-24,\n 4.9555e-24, 4.0046e-24, 9.8169e-24, 5.4177e-24, 2.5418e-24, 4.7713e-24,\n 4.9716e-24, 1.4693e-23, 6.2570e-24, 8.4763e-24, 1.1669e-23, 2.8748e-23,\n 1.0706e-23, 3.6999e-24, 3.7225e-24, 5.3917e-24, 8.9828e-24, 3.8525e-24,\n 9.4398e-24, 4.6522e-24, 1.8470e-23, 5.1357e-24, 2.2590e-23, 4.3277e-24,\n 6.9232e-25, 9.3690e-24, 8.2906e-24, 6.5990e-24, 1.6105e-24, 6.4556e-24,\n 1.2442e-23, 6.4689e-24, 1.2214e-23, 2.3525e-24], device='cuda:0')"
 },
 "67": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[ 5.6052e-45, -5.6052e-45, 5.6052e-45, ..., -5.6052e-45,\n 5.6052e-45, -5.6052e-45],\n [-5.6052e-45, -5.6052e-45, -5.6052e-45, ..., -5.6052e-45,\n 5.6052e-45, -5.6052e-45],\n [-5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n ...,\n [ 5.6052e-45, -5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n [-5.6052e-45, 5.6052e-45, 5.6052e-45, ..., -5.6052e-45,\n 5.6052e-45, -5.6052e-45]], device='cuda:0')",
 "exp_avg_sq": "tensor([[2.1017e-24, 3.3234e-22, 3.5146e-23, ..., 4.4620e-22, 1.6130e-24,\n 4.5362e-23],\n [2.1187e-23, 1.3068e-23, 1.0182e-23, ..., 8.8395e-23, 2.1186e-23,\n 1.5409e-23],\n [6.9811e-25, 3.4832e-22, 3.3380e-23, ..., 4.6750e-22, 9.8797e-25,\n 4.7813e-23],\n ...,\n [1.3579e-23, 3.9841e-24, 4.3459e-24, ..., 1.1805e-22, 2.8845e-24,\n 5.5003e-23],\n [1.0537e-22, 5.8836e-23, 5.1220e-23, ..., 4.8351e-22, 1.0335e-22,\n 1.1696e-22],\n [1.3252e-25, 1.5357e-25, 3.7186e-25, ..., 1.3114e-24, 1.1566e-25,\n 3.8197e-25]], device='cuda:0')"
 },
 "68": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([1.8344e-22, 2.4631e-23, 5.1552e-23, 2.1515e-23, 9.6265e-24, 2.4007e-23,\n 1.4796e-22, 2.2144e-22, 1.0086e-23], device='cuda:0')"
 },
 "69": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[ 5.6052e-45, -5.6052e-45, 5.6052e-45, ..., -5.6052e-45,\n 5.6052e-45, -5.6052e-45],\n [-5.6052e-45, -5.6052e-45, -5.6052e-45, ..., -5.6052e-45,\n 5.6052e-45, -5.6052e-45],\n [-5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n ...,\n [ 5.6052e-45, -5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n [-5.6052e-45, 5.6052e-45, 5.6052e-45, ..., -5.6052e-45,\n 5.6052e-45, -5.6052e-45]], device='cuda:0')",
 "exp_avg_sq": "tensor([[2.0100e-24, 3.1370e-22, 3.3322e-23, ..., 4.2111e-22, 1.5628e-24,\n 4.2808e-23],\n [2.0420e-23, 1.2592e-23, 9.8168e-24, ..., 8.5202e-23, 2.0421e-23,\n 1.4855e-23],\n [6.6193e-25, 3.2867e-22, 3.1510e-23, ..., 4.4112e-22, 9.3276e-25,\n 4.5115e-23],\n ...,\n [1.3315e-23, 3.9081e-24, 4.2673e-24, ..., 1.1572e-22, 2.8276e-24,\n 5.3915e-23],\n [1.0174e-22, 5.6783e-23, 4.9411e-23, ..., 4.6758e-22, 9.9674e-23,\n 1.1353e-22],\n [1.3023e-25, 1.5309e-25, 3.7150e-25, ..., 1.2890e-24, 1.1367e-25,\n 3.7458e-25]], device='cuda:0')"
 },
 "70": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([1.8089e-22, 2.3849e-23, 4.8880e-23, 2.0832e-23, 9.5301e-24, 2.3590e-23,\n 1.4535e-22, 2.1571e-22, 1.0070e-23], device='cuda:0')"
 },
 "71": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([[ 5.6052e-45, -5.6052e-45, 5.6052e-45, ..., -5.6052e-45,\n 5.6052e-45, -5.6052e-45],\n [-5.6052e-45, -5.6052e-45, -5.6052e-45, ..., -5.6052e-45,\n 5.6052e-45, -5.6052e-45],\n [-5.6052e-45, 5.6052e-45, -5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n ...,\n [ 5.6052e-45, -5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n [ 5.6052e-45, 5.6052e-45, 5.6052e-45, ..., 5.6052e-45,\n -5.6052e-45, 5.6052e-45],\n [-5.6052e-45, 5.6052e-45, 5.6052e-45, ..., -5.6052e-45,\n 5.6052e-45, -5.6052e-45]], device='cuda:0')",
 "exp_avg_sq": "tensor([[2.0545e-24, 3.2139e-22, 3.4079e-23, ..., 4.3148e-22, 1.5840e-24,\n 4.3863e-23],\n [2.0383e-23, 1.2578e-23, 9.7995e-24, ..., 8.5061e-23, 2.0384e-23,\n 1.4830e-23],\n [6.7692e-25, 3.3686e-22, 3.2288e-23, ..., 4.5211e-22, 9.5556e-25,\n 4.6238e-23],\n ...,\n [1.3315e-23, 3.9081e-24, 4.2673e-24, ..., 1.1572e-22, 2.8276e-24,\n 5.3915e-23],\n [1.0159e-22, 5.6697e-23, 4.9333e-23, ..., 4.6696e-22, 9.9513e-23,\n 1.1342e-22],\n [1.3022e-25, 1.5309e-25, 3.7150e-25, ..., 1.2890e-24, 1.1366e-25,\n 3.7458e-25]], device='cuda:0')"
 },
 "72": {
 "step": "tensor(33786.)",
 "exp_avg": "tensor([ 5.6052e-45, -5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45,\n 5.6052e-45, 5.6052e-45, -5.6052e-45, -5.6052e-45], device='cuda:0')",
 "exp_avg_sq": "tensor([1.8195e-22, 2.3817e-23, 4.9991e-23, 2.0804e-23, 9.5288e-24, 2.3590e-23,\n 1.4535e-22, 2.1556e-22, 1.0070e-23], device='cuda:0')"
 }
 },
 "param_groups": [
 {
 "lr": 0.00024569294678237997,
 "name": "scale_128",
 "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": 0.01,
 "params": [
 0,
 1,
 2
 ]
 },
 {
 "lr": 0.00024569294678237997,
 "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": 0.01,
 "params": [
 3,
 4,
 5
 ]
 },
 {
 "lr": 0.00024569294678237997,
 "name": "scale_384",
 "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": 0.01,
 "params": [
 6,
 7,
 8
 ]
 },
 {
 "lr": 0.00024569294678237997,
 "name": "scale_448",
 "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": 0.01,
 "params": [
 9,
 10,
 11
 ]
 },
 {
 "lr": 0.00024569294678237997,
 "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": 0.01,
 "params": [
 12,
 13,
 14
 ]
 },
 {
 "lr": 0.00024569294678237997,
 "name": "scale_576",
 "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": 0.01,
 "params": [
 15,
 16,
 17
 ]
 },
 {
 "lr": 0.00024569294678237997,
 "name": "scale_640",
 "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": 0.01,
 "params": [
 18,
 19,
 20
 ]
 },
 {
 "lr": 0.00024569294678237997,
 "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": 0.01,
 "params": [
 21,
 22,
 23
 ]
 },
 {
 "lr": 0.00024569294678237997,
 "name": "scale_896",
 "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": 0.01,
 "params": [
 24,
 25,
 26
 ]
 },
 {
 "lr": 0.00012333423752026375,
 "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": 0.005,
 "params": [
 27,
 28,
 29,
 30,
 31,
 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,
 62,
 63,
 64,
 65,
 66,
 67,
 68,
 69,
 70,
 71,
 72
 ]
 }
 ]
 },
 "scheduler_state_dict": {
 "T_0": 10,
 "T_i": 10,
 "T_mult": 2,
 "eta_min": 1e-06,
 "T_cur": 9,
 "base_lrs": [
 0.01,
 0.01,
 0.01,
 0.01,
 0.01,
 0.01,
 0.01,
 0.01,
 0.01,
 0.005
 ],
 "last_epoch": 9,
 "_step_count": 0,
 "_is_initial": false,
 "_get_lr_called_within_step": false,
 "_last_lr": [
 0.00024569294678237997,
 0.00024569294678237997,
 0.00024569294678237997,
 0.00024569294678237997,
 0.00024569294678237997,
 0.00024569294678237997,
 0.00024569294678237997,
 0.00024569294678237997,
 0.00024569294678237997,
 0.00012333423752026375
 ]
 },
 "metrics": {
 "best_val_acc": 62.92066666666667,
 "best_epoch": 8,
 "scale_accuracies": {
 "128": 62.92066666666667,
 "256": 71.04133333333333,
 "384": 73.38066666666667,
 "448": 74.20933333333333,
 "512": 74.59066666666666,
 "576": 75.03266666666667,
 "640": 75.182,
 "768": 75.57266666666666,
 "896": 75.86
 },
 "training_history": {
 "epochs": [
 1,
 2,
 3,
 4,
 5,
 6,
 7,
 8,
 9
 ],
 "train_loss": [
 4.900288952114611,
 4.0311296205286915,
 3.872760212415066,
 3.772298225203155,
 3.694623793321019,
 3.6294693627380346,
 3.5741883491542534,
 3.5266703713365737,
 3.490817522837497
 ],
 "train_acc": [
 51.657460216609806,
 56.824520144524485,
 57.987626385423084,
 58.73678711154231,
 59.30749074866899,
 59.79954213619302,
 60.232454733327764,
 60.59530100291375,
 60.873484877459376
 ],
 "val_acc": [
 58.398666666666664,
 60.18533333333333,
 60.89533333333333,
 61.468666666666664,
 61.802,
 62.246,
 62.55133333333333,
 62.774,
 62.92066666666667
 ],
 "scale_accs": {
 "128": [
 58.398666666666664,
 60.18533333333333,
 60.89533333333333,
 61.468666666666664,
 61.802,
 62.246,
 62.55133333333333,
 62.774,
 62.92066666666667
 ],
 "256": [
 67.03133333333334,
 68.56666666666666,
 69.374,
 69.826,
 70.146,
 70.442,
 70.74533333333333,
 70.86466666666666,
 71.04133333333333
 ],
 "384": [
 69.55466666666666,
 71.02133333333333,
 71.764,
 72.25133333333333,
 72.66,
 72.89,
 73.12933333333334,
 73.318,
 73.38066666666667
 ],
 "448": [
 70.34333333333333,
 71.97066666666667,
 72.66733333333333,
 73.1,
 73.44666666666667,
 73.616,
 73.936,
 74.12333333333333,
 74.20933333333333
 ],
 "512": [
 70.83533333333334,
 72.43066666666667,
 73.07733333333333,
 73.59866666666667,
 73.876,
 74.13,
 74.29866666666666,
 74.48333333333333,
 74.59066666666666
 ],
 "576": [
 71.29266666666666,
 72.69866666666667,
 73.488,
 73.876,
 74.25133333333333,
 74.53733333333334,
 74.74466666666666,
 74.91666666666667,
 75.03266666666667
 ],
 "640": [
 71.60266666666666,
 73.09866666666667,
 73.69266666666667,
 74.18333333333334,
 74.45733333333334,
 74.72066666666667,
 74.906,
 75.04866666666666,
 75.182
 ],
 "768": [
 72.03333333333333,
 73.37733333333334,
 74.138,
 74.52333333333333,
 74.82,
 75.086,
 75.31,
 75.43866666666666,
 75.57266666666666
 ],
 "896": [
 72.252,
 73.866,
 74.37,
 75.06266666666667,
 75.262,
 75.43733333333333,
 75.59266666666667,
 75.78266666666667,
 75.86
 ]
 },
 "lr": [
 0.00975530705321762,
 0.00904518046337755,
 0.00793913236883622,
 0.00654543046337755,
 0.005000500000000001,
 0.0034555695366224513,
 0.0020618676311637816,
 0.0009558195366224509,
 0.00024569294678237997
 ]
 }
 },
 "train_config": {
 "name": "david_training",
 "run_id": "20251013_004438",
 "dataset_name": "AbstractPhil/imagenet-clip-features-orderly",
 "model_variant": [
 "clip_vit_b16",
 "clip_vit_laion_b32",
 "clip_vit_b32"
 ],
 "num_classes": 1000,
 "preset": "gated_expert_team",
 "custom_config_path": null,
 "num_classes_override": null,
 "use_belly_override": null,
 "belly_expand_override": null,
 "progressive_training_override": true,
 "scale_warmup_epochs_override": {
 "128": 0,
 "256": 0,
 "384": 0,
 "448": 0,
 "512": 0,
 "576": 0,
 "640": 0,
 "768": 0,
 "896": 0
 },
 "num_epochs": 10,
 "batch_size": 1024,
 "learning_rate": 0.01,
 "weight_decay": 1e-05,
 "warmup_epochs": 3,
 "use_rose_loss": true,
 "rose_initial_weight": 0.1,
 "rose_max_weight": 0.8,
 "rose_weight_schedule": "adaptive",
 "use_cayley_loss": false,
 "cayley_weight": 0.01,
 "scale_loss_balance": null,
 "use_mixed_precision": false,
 "gradient_clip": 10.0,
 "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-05,
 "gradient_scale_multiplier": 10.0,
 "log_interval": 50,
 "val_interval": 1,
 "save_interval": 5,
 "log_fusion_weights": true,
 "log_loss_components": true,
 "save_format": "safetensors",
 "hf_repo": "AbstractPhil/david-shared-space",
 "upload_to_hub": true,
 "base_dir": "./david_training",
 "num_workers": 10,
 "pin_memory": true,
 "prefetch_factor": 4,
 "persistent_workers": true
 }
}