Hugging Face's logo

AbstractPhil
/
gated-david

Image Classification
TensorBoard
Safetensors
English
imagenet
multi-scale
feature-geometry
david
Eval Results (legacy)
Model card Files
xet
 Metrics Community
AbstractPhil commited on
Commit
aa176c4
·
verified ·
1 Parent(s): 04a049f

Upload weights and configs - David-fully_shared-weighted_sum - Run 20251012_132646

Browse files
Files changed (2) hide show
  1. weights/David-fully_shared-weighted_sum/20251012_132646/best_model.safetensors +1 -1
  2. weights/David-fully_shared-weighted_sum/20251012_132646/best_model_metadata.json +36 -36
weights/David-fully_shared-weighted_sum/20251012_132646/best_model.safetensors CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:44c36a8d013401b705455b653c1e45cbe58875c978facd5f0607f1840b082d11
3
  size 2628344
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:6b9f9ffbce2e08246debd2a87f260ab74a3c05a223225ebce9a82c3381305e71
3
  size 2628344
weights/David-fully_shared-weighted_sum/20251012_132646/best_model_metadata.json CHANGED
@@ -1,46 +1,46 @@
1
  {
2
- "epoch": 1,
3
  "optimizer_state_dict": {
4
  "state": {
5
  "0": {
6
- "step": "tensor(2504.)",
7
- "exp_avg": "tensor([[-3.7776e-05, 2.5930e-03, -1.0904e-03, ..., 8.1735e-04,\n 5.7645e-04, 1.1802e-03],\n [-4.4963e-04, 7.0748e-04, 8.2980e-04, ..., 8.9101e-04,\n 7.2794e-05, 1.7938e-04],\n [-1.3904e-04, -2.6577e-03, -2.2862e-03, ..., 5.3059e-04,\n 1.2360e-04, 1.9591e-03],\n ...,\n [ 3.6896e-04, -4.1577e-04, -1.8343e-03, ..., -1.2041e-03,\n 1.5108e-03, 1.1555e-04],\n [ 7.3712e-04, -1.3868e-03, -1.2426e-03, ..., -5.3006e-04,\n -6.5094e-04, 2.5752e-04],\n [ 2.5474e-03, -3.7966e-04, 3.6154e-04, ..., -1.4336e-04,\n 2.6799e-05, 9.5708e-04]], device='cuda:0')",
8
- "exp_avg_sq": "tensor([[4.2325e-06, 3.5893e-05, 1.3479e-05, ..., 2.9146e-06, 2.3005e-06,\n 4.8253e-06],\n [8.1619e-06, 3.2394e-05, 1.6405e-05, ..., 3.8712e-06, 2.2472e-06,\n 4.6732e-06],\n [7.6940e-06, 5.5376e-05, 2.1418e-05, ..., 5.2286e-06, 3.0685e-06,\n 6.1451e-06],\n ...,\n [6.6711e-06, 5.0896e-05, 2.0446e-05, ..., 4.6117e-06, 3.5392e-06,\n 3.9067e-06],\n [3.9209e-06, 3.0676e-05, 1.2810e-05, ..., 2.9420e-06, 2.1341e-06,\n 2.6537e-06],\n [3.5377e-06, 1.8884e-05, 9.9787e-06, ..., 2.6884e-06, 1.6161e-06,\n 3.0642e-06]], device='cuda:0')"
9
  },
10
  "1": {
11
- "step": "tensor(2504.)",
12
- "exp_avg": "tensor([-1.2330e-03, -2.7070e-02, 3.9827e-03, 9.3708e-03, 3.9574e-03,\n -1.4923e-03, 2.4006e-02, 6.8119e-03, -2.3198e-02, 5.3953e-03,\n 4.2220e-02, 2.8956e-02, -3.0642e-03, -6.0452e-03, -9.1474e-03,\n 4.8018e-03, 3.7408e-03, 3.5073e-02, -7.8004e-03, 1.4369e-02,\n -3.3174e-03, -3.6713e-03, 7.3876e-03, 8.5043e-03, 2.8027e-02,\n -4.2522e-02, -3.8627e-03, -9.6676e-03, 1.5123e-02, -2.5726e-03,\n -2.1395e-02, 7.5408e-03, -2.9613e-02, -1.4218e-02, -1.9304e-02,\n 5.1801e-02, -6.6549e-03, 3.2683e-02, 1.0758e-02, -6.7370e-02,\n -4.8374e-02, -3.6733e-02, -9.5829e-03, 9.6607e-04, 2.8840e-03,\n 1.3213e-02, 1.3037e-02, 2.3417e-02, 3.2458e-02, 4.7392e-02,\n 7.8220e-04, -1.7166e-02, -1.4304e-03, -2.5363e-03, 3.2036e-02,\n 1.8568e-03, 1.3650e-02, 1.0755e-02, -1.5066e-02, 2.0273e-02,\n 6.4988e-03, 2.4395e-03, -2.6666e-02, -2.0115e-02, 2.1696e-02,\n -5.0570e-02, 2.2125e-02, -2.5530e-02, 8.5193e-03, 3.7800e-02,\n 1.4399e-02, 1.1608e-02, -1.9595e-02, 2.1609e-02, 6.8934e-04,\n -5.7166e-03, 2.6736e-02, 1.9745e-02, 1.5563e-02, 1.8687e-02,\n 1.6490e-02, -1.5713e-02, 1.3756e-02, 1.2605e-02, -1.0862e-03,\n -2.1507e-02, 1.8133e-02, 8.5814e-03, 2.2317e-03, -1.5408e-03,\n 3.7848e-02, -1.4847e-02, -3.0453e-03, -6.1651e-03, 1.1152e-02,\n 9.9882e-03, -3.0142e-02, 1.6645e-02, 4.6051e-03, 1.0073e-02,\n -3.5688e-02, 1.9460e-02, -1.7704e-02, -9.5606e-03, 6.2358e-03,\n 1.6609e-02, -2.2032e-02, -2.0797e-02, -2.8564e-03, 1.9164e-02,\n 9.7848e-03, 1.8900e-02, -2.2223e-02, 2.8903e-03, -2.3619e-02,\n -2.6077e-02, -9.9089e-04, 4.1012e-03, -1.1445e-02, 1.3650e-02,\n -1.7735e-03, 6.5409e-03, 1.1635e-02, -1.4130e-02, -2.3996e-02,\n 1.7638e-03, -5.5136e-02, 7.1711e-03, 2.8370e-02, 3.7838e-03,\n 2.4583e-02, -2.0575e-03, 1.8590e-02, 1.7808e-02, 2.3429e-03,\n -2.1123e-02, 3.0591e-02, 1.3352e-03, 7.3173e-03, 2.4353e-02,\n 3.8441e-03, 8.0498e-03, -4.3501e-03, 4.2925e-03, 2.1275e-03,\n -1.7449e-03, 9.5893e-03, -2.2752e-02, -1.0851e-02, 1.6564e-02,\n -3.3432e-03, -3.8581e-03, 1.0811e-02, 1.7190e-02, 1.7373e-02,\n -9.6422e-03, 1.1875e-02, 7.8579e-03, -1.3094e-03, -1.7253e-02,\n -2.5058e-02, 2.4905e-02, -3.3923e-03, -1.9406e-02, 3.4043e-03,\n -3.4419e-02, -8.0399e-03, 2.8188e-03, 3.9913e-03, 1.4462e-02,\n 1.4958e-02, 1.3767e-02, -9.5508e-03, 6.1402e-03, 1.1605e-02,\n -1.1909e-02, -6.1030e-03, 2.0123e-02, 8.0192e-03, 3.2700e-02,\n 6.2663e-03, 6.3462e-04, -1.8584e-03, 3.1119e-02, -9.8209e-03,\n 2.0296e-02, 3.3016e-02, 9.6425e-03, -1.4983e-02, 5.2626e-02,\n -3.4124e-04, -2.7290e-02, -1.6658e-02, 3.5102e-02, 2.3067e-02,\n 2.0097e-03, 3.4720e-02, 3.3226e-02, 1.2385e-03, -1.1241e-02,\n -5.4617e-02, 3.9409e-03, -1.5658e-02, 1.9346e-02, 2.8294e-02,\n -3.1394e-02, 1.1144e-02, -9.5524e-03, -1.2033e-02, 5.9579e-02,\n -1.5568e-02, 1.8908e-02, -5.5759e-03, 9.3324e-03, -1.8129e-02,\n -3.2248e-02, 2.9446e-03, -1.1270e-02, 1.5257e-02, -1.4520e-02,\n 3.1499e-02, -4.0112e-02, 1.5260e-02, 7.0770e-03, 1.1640e-03,\n -4.8305e-03, -4.9477e-02, -2.9144e-02, 2.6074e-02, -3.3924e-04,\n -1.4495e-02, 2.0956e-02, -2.3297e-02, -3.7419e-02, 1.5591e-02,\n 8.0628e-03, -2.1227e-02, -6.3550e-03, -2.1143e-02, -8.2961e-03,\n 3.0128e-02, 5.5892e-03, 1.8274e-02, -2.9007e-02, 2.8672e-02,\n -4.5433e-03, -1.8216e-02, -1.1997e-02, 1.3217e-02, 1.0889e-02,\n 2.1695e-02, -1.3489e-02, -3.2850e-03, -9.0139e-03, 1.1833e-02,\n -1.4557e-03, -7.8825e-03, -2.4029e-02, 3.1379e-03, -3.6006e-02,\n -2.6077e-03, -3.4478e-02, 1.0368e-02, 2.0008e-02, 9.0006e-03,\n 1.0138e-02, -3.7877e-02, -1.3889e-04, -2.4990e-05, -1.3655e-02,\n 4.2798e-02, -3.0341e-02, -1.0600e-02, 4.5082e-03, -5.1520e-03,\n 5.1242e-03, 1.5821e-03, 3.3316e-02, -1.4875e-02, -1.3199e-02,\n -1.8809e-02, -2.6036e-02, 3.4903e-02, 6.0485e-03, -1.2642e-02,\n 1.5351e-02, 2.4742e-02, 1.7476e-02, -6.6802e-03, 2.6129e-02,\n -2.0877e-04, 3.0710e-02, 1.9546e-02, -1.0710e-02, -1.0403e-02,\n -2.3132e-02, -3.5512e-02, -2.2684e-03, -8.4065e-03, 1.8485e-03,\n 3.9970e-02, 2.3412e-02, -4.1306e-02, 1.5390e-02, 4.6104e-03,\n -1.2060e-02, -7.9751e-04, 2.5443e-02, 1.2390e-02, -1.1933e-02,\n 1.5920e-02, -1.2368e-02, -3.5070e-03, -2.2955e-02, 1.7739e-02,\n -4.9563e-03, -4.4710e-02, -2.3131e-02, 2.0270e-02, -1.6853e-02,\n -1.3333e-02, -1.7003e-03, -2.6828e-02, 1.4197e-03, -4.6157e-03,\n 2.4602e-02, -3.3468e-02, -9.3479e-04, -1.8459e-03, 9.5125e-03,\n -7.4425e-03, -2.4931e-02, 1.3876e-02, -1.3937e-02, 1.3066e-02,\n 7.5824e-03, -6.6333e-04, -2.2951e-02, -3.6546e-03, -2.6906e-02,\n 5.9309e-03, 2.8210e-02, -7.2312e-03, -2.2280e-02, 6.7978e-03,\n 2.1084e-02, -1.2148e-02, -1.0728e-02, 1.6352e-02, 8.7777e-03,\n -7.8558e-03, -7.1589e-03, 1.2567e-02, 2.5498e-02, -4.2565e-04,\n 1.6786e-02, -5.9548e-02, -1.6657e-02, -2.6834e-02, -3.0901e-03,\n -4.9622e-03, -5.3421e-02, -1.0188e-02, -1.3243e-02, -4.1177e-02,\n -5.9646e-03, -4.8138e-03, -2.5619e-02, 2.2552e-03, 2.6494e-02,\n -5.4214e-03, -1.5973e-02, 2.7951e-04, -8.3979e-03, -1.6260e-02,\n -3.4576e-02, 3.0719e-02, -2.3516e-02, -2.4781e-03, 1.3934e-02,\n 4.7991e-02, 6.0208e-03, -7.4620e-03, -1.0103e-02, 5.5539e-03,\n -3.6788e-02, 3.6558e-02, 8.1194e-03, 3.7960e-02, -1.8259e-02,\n 1.4764e-03, 3.7082e-02, 1.0007e-02, -9.7297e-03, 1.4271e-02,\n -1.1326e-02, 2.4006e-02, -2.1545e-02, 8.6374e-03, 6.7155e-05,\n 1.9362e-02, 1.3444e-03, 1.7759e-03, 2.3905e-02, 4.2675e-03,\n 1.1374e-02, 1.9397e-02, -3.7005e-03, -1.1010e-02, -5.5736e-03,\n -1.3492e-02, -1.6431e-02, -1.3105e-02, -1.7507e-02, -3.7421e-02,\n -1.9095e-02, 3.5540e-03, -1.4820e-02, 6.3984e-03, -3.7961e-02,\n -2.9090e-02, 2.3057e-02, 1.0703e-03, 4.7136e-03, 3.2055e-02,\n 8.5376e-03, -1.1659e-02, -2.3225e-02, 3.0870e-02, 3.9144e-03,\n 1.2954e-02, -2.9102e-02, 1.2537e-02, 1.1885e-02, -1.5428e-02,\n 1.9845e-02, -1.6284e-02, 7.4325e-05, -1.2026e-02, 1.9487e-03,\n 8.6088e-03, -6.3218e-03, 1.5938e-02, 3.6728e-02, 1.6517e-02,\n -3.7753e-02, -5.8896e-03, 3.3164e-02, 3.0111e-02, -9.5126e-03,\n 1.1595e-03, 1.8039e-02, -2.2862e-02, 2.7915e-02, 1.5405e-02,\n -1.0270e-02, 3.8747e-02, 1.5766e-03, 4.5951e-03, -4.8094e-03,\n -1.4664e-02, -5.5251e-03, -1.6012e-02, 3.6013e-02, -4.3686e-03,\n 3.7566e-02, -2.7546e-03, -1.8684e-02, 3.2119e-02, 2.5149e-02,\n 1.8258e-02, -2.6646e-02, -7.1241e-03, -8.1233e-03, 7.8597e-03,\n 3.9407e-02, 1.2632e-02, 4.6449e-03, -1.3416e-02, -4.0604e-02,\n 2.5577e-02, -5.7391e-03, -2.4587e-02, -1.4031e-02, -2.7482e-04,\n 2.0312e-02, -1.9620e-02, -1.6740e-03, -3.5871e-03, 9.9431e-03,\n 2.4370e-02, -2.1187e-02, -1.2404e-03, -7.5375e-03, -4.9552e-02,\n -2.0648e-02, -2.7956e-03, -3.1480e-02, 1.1984e-02, -2.4418e-02,\n 1.1194e-02, -1.1966e-03, -3.1280e-02, -2.6200e-02, 2.9787e-02,\n -7.4888e-03, 2.9570e-02, -9.9626e-03, -7.1159e-02, -3.0023e-02,\n -1.0086e-02, 2.3418e-02], device='cuda:0')",
13
- "exp_avg_sq": "tensor([0.0042, 0.0045, 0.0064, 0.0063, 0.0034, 0.0032, 0.0070, 0.0049, 0.0054,\n 0.0041, 0.0059, 0.0039, 0.0039, 0.0050, 0.0052, 0.0054, 0.0061, 0.0033,\n 0.0033, 0.0044, 0.0068, 0.0037, 0.0039, 0.0038, 0.0059, 0.0049, 0.0040,\n 0.0043, 0.0058, 0.0065, 0.0051, 0.0064, 0.0051, 0.0060, 0.0041, 0.0098,\n 0.0048, 0.0038, 0.0038, 0.0064, 0.0051, 0.0045, 0.0048, 0.0071, 0.0027,\n 0.0056, 0.0022, 0.0050, 0.0043, 0.0033, 0.0048, 0.0036, 0.0057, 0.0057,\n 0.0048, 0.0038, 0.0033, 0.0035, 0.0038, 0.0044, 0.0030, 0.0033, 0.0054,\n 0.0080, 0.0046, 0.0084, 0.0040, 0.0036, 0.0087, 0.0066, 0.0046, 0.0095,\n 0.0049, 0.0055, 0.0038, 0.0050, 0.0049, 0.0021, 0.0029, 0.0062, 0.0037,\n 0.0057, 0.0049, 0.0034, 0.0010, 0.0052, 0.0056, 0.0062, 0.0037, 0.0055,\n 0.0053, 0.0041, 0.0064, 0.0041, 0.0064, 0.0059, 0.0044, 0.0050, 0.0043,\n 0.0055, 0.0042, 0.0049, 0.0054, 0.0051, 0.0027, 0.0058, 0.0027, 0.0057,\n 0.0058, 0.0043, 0.0045, 0.0052, 0.0046, 0.0044, 0.0030, 0.0039, 0.0062,\n 0.0025, 0.0024, 0.0059, 0.0043, 0.0051, 0.0022, 0.0035, 0.0054, 0.0060,\n 0.0079, 0.0047, 0.0054, 0.0057, 0.0049, 0.0019, 0.0030, 0.0047, 0.0036,\n 0.0068, 0.0071, 0.0021, 0.0086, 0.0045, 0.0031, 0.0060, 0.0041, 0.0039,\n 0.0023, 0.0071, 0.0081, 0.0076, 0.0056, 0.0071, 0.0045, 0.0051, 0.0045,\n 0.0046, 0.0065, 0.0066, 0.0030, 0.0037, 0.0027, 0.0041, 0.0060, 0.0057,\n 0.0030, 0.0047, 0.0077, 0.0072, 0.0037, 0.0030, 0.0053, 0.0073, 0.0034,\n 0.0040, 0.0037, 0.0033, 0.0068, 0.0055, 0.0062, 0.0049, 0.0080, 0.0052,\n 0.0074, 0.0042, 0.0081, 0.0057, 0.0059, 0.0068, 0.0053, 0.0040, 0.0039,\n 0.0054, 0.0041, 0.0036, 0.0046, 0.0041, 0.0060, 0.0075, 0.0065, 0.0068,\n 0.0031, 0.0054, 0.0062, 0.0040, 0.0063, 0.0071, 0.0043, 0.0055, 0.0033,\n 0.0057, 0.0032, 0.0079, 0.0047, 0.0043, 0.0066, 0.0055, 0.0061, 0.0041,\n 0.0030, 0.0044, 0.0049, 0.0068, 0.0042, 0.0069, 0.0046, 0.0033, 0.0060,\n 0.0045, 0.0045, 0.0071, 0.0049, 0.0041, 0.0026, 0.0052, 0.0068, 0.0074,\n 0.0052, 0.0034, 0.0039, 0.0085, 0.0052, 0.0065, 0.0029, 0.0047, 0.0051,\n 0.0065, 0.0081, 0.0088, 0.0020, 0.0057, 0.0081, 0.0032, 0.0057, 0.0039,\n 0.0032, 0.0011, 0.0024, 0.0053, 0.0048, 0.0048, 0.0056, 0.0038, 0.0038,\n 0.0045, 0.0053, 0.0059, 0.0043, 0.0050, 0.0054, 0.0049, 0.0026, 0.0047,\n 0.0072, 0.0039, 0.0045, 0.0066, 0.0052, 0.0055, 0.0041, 0.0073, 0.0050,\n 0.0064, 0.0044, 0.0080, 0.0068, 0.0047, 0.0037, 0.0028, 0.0068, 0.0045,\n 0.0045, 0.0059, 0.0051, 0.0049, 0.0029, 0.0063, 0.0054, 0.0100, 0.0052,\n 0.0051, 0.0055, 0.0062, 0.0059, 0.0044, 0.0066, 0.0061, 0.0043, 0.0048,\n 0.0047, 0.0052, 0.0054, 0.0061, 0.0069, 0.0064, 0.0039, 0.0062, 0.0057,\n 0.0041, 0.0043, 0.0047, 0.0051, 0.0031, 0.0044, 0.0027, 0.0043, 0.0055,\n 0.0042, 0.0070, 0.0045, 0.0059, 0.0042, 0.0053, 0.0042, 0.0031, 0.0026,\n 0.0062, 0.0042, 0.0035, 0.0062, 0.0044, 0.0050, 0.0045, 0.0054, 0.0038,\n 0.0057, 0.0045, 0.0028, 0.0055, 0.0048, 0.0044, 0.0034, 0.0054, 0.0033,\n 0.0041, 0.0049, 0.0057, 0.0048, 0.0091, 0.0051, 0.0051, 0.0039, 0.0054,\n 0.0052, 0.0066, 0.0054, 0.0035, 0.0066, 0.0060, 0.0056, 0.0040, 0.0034,\n 0.0062, 0.0058, 0.0058, 0.0032, 0.0045, 0.0046, 0.0042, 0.0045, 0.0066,\n 0.0047, 0.0058, 0.0067, 0.0044, 0.0062, 0.0045, 0.0039, 0.0042, 0.0065,\n 0.0042, 0.0060, 0.0051, 0.0058, 0.0050, 0.0032, 0.0071, 0.0063, 0.0055,\n 0.0057, 0.0050, 0.0052, 0.0044, 0.0024, 0.0041, 0.0037, 0.0076, 0.0030,\n 0.0070, 0.0071, 0.0061, 0.0050, 0.0033, 0.0044, 0.0035, 0.0020, 0.0067,\n 0.0067, 0.0055, 0.0034, 0.0056, 0.0033, 0.0095, 0.0044, 0.0070, 0.0086,\n 0.0028, 0.0049, 0.0050, 0.0036, 0.0067, 0.0040, 0.0039, 0.0043, 0.0051,\n 0.0058, 0.0041, 0.0042, 0.0051, 0.0040, 0.0052, 0.0034, 0.0043, 0.0029,\n 0.0081, 0.0068, 0.0052, 0.0028, 0.0054, 0.0052, 0.0053, 0.0070, 0.0065,\n 0.0038, 0.0042, 0.0052, 0.0047, 0.0062, 0.0041, 0.0055, 0.0020, 0.0025,\n 0.0086, 0.0052, 0.0039, 0.0082, 0.0060, 0.0059, 0.0072, 0.0064, 0.0043,\n 0.0044, 0.0053, 0.0072, 0.0040, 0.0058, 0.0029, 0.0060, 0.0036, 0.0036,\n 0.0071, 0.0043, 0.0065, 0.0090, 0.0045, 0.0049, 0.0023, 0.0037, 0.0035,\n 0.0047, 0.0060, 0.0055, 0.0066, 0.0038, 0.0045, 0.0061, 0.0038, 0.0088,\n 0.0058, 0.0042, 0.0065, 0.0038, 0.0062, 0.0041, 0.0069, 0.0039, 0.0031,\n 0.0049, 0.0033, 0.0077, 0.0059, 0.0126, 0.0057, 0.0038, 0.0034],\n device='cuda:0')"
14
  },
15
  "2": {
16
- "step": "tensor(2504.)",
17
- "exp_avg": "tensor([ 1.2790e-03, -4.2484e-03, 1.6406e-03, 5.0555e-04, 1.8122e-04,\n 2.2344e-04, 2.0652e-03, 1.4049e-03, -2.9664e-03, 1.1021e-04,\n 4.4328e-03, 5.6112e-03, -8.6295e-04, -4.2636e-04, -1.8929e-03,\n 1.4027e-03, -8.2183e-04, 6.8354e-03, -1.1494e-03, 2.0638e-03,\n -1.2572e-03, -1.0911e-03, 1.0842e-03, 5.2755e-04, 1.5163e-03,\n -4.5268e-03, 6.0199e-04, -1.4617e-03, 1.4594e-03, 2.2350e-03,\n -2.9927e-03, 1.6324e-03, -2.3090e-03, -1.3903e-03, -7.6477e-04,\n 4.0510e-03, -8.5939e-04, 5.9945e-03, 2.6302e-03, -7.2995e-03,\n -5.2975e-03, -4.0528e-03, -2.2362e-03, 4.0155e-04, -1.0192e-04,\n 1.4388e-03, 3.6827e-03, 3.1243e-03, 5.0971e-03, 1.2525e-02,\n 9.2220e-04, -3.2902e-03, -1.6553e-03, -5.4094e-05, 4.0173e-03,\n -6.0995e-04, 2.3448e-03, 1.8635e-03, -2.0523e-03, 3.8647e-03,\n 2.8919e-03, 1.2832e-04, -3.6796e-03, -1.7574e-03, 3.7876e-03,\n -7.7616e-04, 2.9351e-03, -3.9637e-03, 9.8028e-04, 3.9020e-03,\n 1.8482e-03, 1.5702e-03, -2.0220e-03, 2.2531e-03, -1.1153e-03,\n -2.7992e-04, 4.2058e-03, 7.8005e-03, 2.3004e-03, 4.3761e-03,\n 2.9816e-03, -1.7060e-03, 5.7172e-04, 1.5981e-03, 3.2893e-03,\n -2.4894e-03, 1.8977e-03, 6.8948e-04, -3.9546e-04, -4.4735e-05,\n 3.4465e-03, -2.5500e-03, 1.3797e-04, -2.2315e-04, 1.0230e-03,\n 1.2288e-03, -5.7364e-03, 2.2616e-03, 8.2095e-04, 8.2322e-04,\n -4.9253e-03, 2.1418e-03, -3.6236e-03, -1.3637e-03, 3.2142e-03,\n 2.0785e-03, -4.6794e-03, -1.8576e-03, 1.4222e-04, 2.3148e-03,\n 2.0601e-03, 1.9982e-03, -3.4795e-03, 2.2888e-04, -3.5825e-03,\n -2.9464e-03, 6.0175e-04, 9.3494e-04, -3.5889e-03, 9.2820e-04,\n -3.3415e-04, 6.7745e-04, 5.7719e-03, -2.6046e-03, -2.7963e-03,\n -1.8435e-04, -1.4826e-03, 6.7832e-04, 1.3810e-03, 1.0691e-03,\n 2.2451e-03, -1.3684e-03, 3.6860e-03, 2.3206e-03, 3.5070e-04,\n -1.9631e-03, 1.7600e-03, 4.0018e-04, -8.2718e-04, 4.0060e-03,\n 1.3607e-03, 7.9929e-04, -1.7836e-04, 1.6332e-03, 1.0917e-03,\n -7.7408e-04, -2.3408e-04, -3.1471e-03, -5.9394e-04, 3.4904e-04,\n -1.3622e-03, -5.4131e-04, 1.7557e-03, 1.9031e-03, 1.5289e-03,\n -8.8077e-04, 2.9074e-03, 1.7656e-03, 1.8265e-04, -2.3185e-03,\n -1.2591e-03, 1.4722e-03, -1.7392e-03, -2.6327e-03, 5.0952e-04,\n -1.6860e-03, -7.2682e-04, 2.6305e-03, 1.3235e-03, 1.6998e-03,\n 3.8517e-03, 1.7787e-03, -1.0206e-03, 2.2493e-03, 6.3151e-04,\n -8.7286e-04, -6.7347e-04, 1.9214e-03, 6.6003e-04, 4.1254e-03,\n 6.9470e-05, -3.6713e-04, 2.4631e-04, 2.3392e-03, -1.6695e-03,\n 2.1976e-03, 2.6656e-03, -3.4137e-04, -2.3179e-03, 8.0005e-03,\n -6.9609e-04, -4.2545e-03, -1.7999e-03, 5.5861e-03, 1.6920e-03,\n 2.5788e-04, 2.3342e-03, 4.1709e-03, -3.3818e-06, -1.1217e-03,\n -4.6050e-03, 8.0839e-04, -1.8031e-03, 1.0724e-03, 1.9516e-03,\n -3.1839e-03, 8.7772e-04, -9.0838e-04, -2.8305e-03, 7.5855e-03,\n -1.8721e-03, 2.7642e-03, 2.5275e-04, 1.3991e-03, -2.9632e-03,\n -3.9958e-03, 7.9649e-04, -1.7456e-03, 1.6816e-03, 4.1697e-04,\n 4.7466e-03, -4.4316e-03, 1.1224e-03, 4.6692e-04, -2.3326e-06,\n -7.3693e-04, -5.2584e-03, -1.9178e-03, 1.8070e-03, 6.1201e-04,\n -3.2813e-03, 1.0893e-03, -2.6443e-03, -3.2658e-03, 2.2343e-03,\n 1.8169e-03, -4.4794e-03, -6.3571e-04, -2.3097e-03, -9.2412e-04,\n 7.8728e-03, 4.4402e-04, 2.0728e-03, -2.1065e-03, 3.0905e-03,\n 1.6858e-03, -8.7179e-03, -7.8336e-04, 1.0887e-03, 2.0208e-03,\n 2.2217e-03, -2.2284e-03, -8.1902e-04, -2.2593e-02, 2.6082e-03,\n 3.0420e-04, -5.8974e-04, -2.7672e-03, 3.2016e-04, -4.9472e-03,\n -7.2202e-04, -4.1475e-03, 1.0347e-03, 1.7835e-03, 1.9244e-03,\n 7.0917e-04, -4.8568e-03, 5.7119e-05, -5.1342e-04, -1.4261e-03,\n 3.2196e-03, -4.5744e-03, -1.0753e-03, 9.7476e-04, -4.0697e-04,\n 9.1177e-04, -1.0167e-03, 2.1779e-03, 8.0611e-05, -7.7584e-04,\n -1.4534e-03, -2.5266e-04, 2.1049e-03, 9.5170e-04, -1.2403e-03,\n 3.7413e-03, 2.2690e-03, 1.3262e-03, -8.9037e-04, 5.2042e-03,\n 5.6993e-04, 3.4228e-03, 3.5605e-03, -1.1342e-03, -8.8261e-04,\n -1.2546e-03, -4.5897e-03, 7.8635e-04, -1.1382e-03, -7.8536e-04,\n 3.4728e-03, 2.7894e-03, -3.6221e-03, 1.5110e-03, 1.2349e-03,\n -7.3677e-04, -2.5023e-04, 3.7674e-03, 1.1963e-03, -8.5041e-04,\n 1.1423e-03, -4.6913e-04, 3.7446e-04, -1.7061e-03, 1.6113e-03,\n 3.9741e-05, -6.1088e-03, -4.4270e-03, 1.0531e-03, -3.5592e-03,\n -2.7697e-03, -2.0445e-04, -3.0272e-03, 5.7398e-04, -1.2212e-03,\n 2.2466e-03, -3.8581e-03, 3.3973e-04, -5.4694e-04, -1.1898e-03,\n -1.3123e-04, -4.2530e-03, 4.8025e-03, -1.8760e-03, 1.8303e-03,\n 1.3452e-03, -4.7252e-04, -2.6312e-03, -7.7662e-04, -2.5494e-03,\n -7.4736e-04, 3.9667e-03, -5.8474e-04, -2.6305e-03, 9.3808e-04,\n 3.0981e-03, -2.8541e-03, -2.3027e-03, 2.3195e-03, 1.7742e-03,\n -1.3724e-03, -1.7603e-03, 1.7141e-03, 1.8940e-03, -8.2953e-04,\n 2.3942e-03, -9.4485e-03, -9.2281e-04, -3.6860e-03, 3.1648e-04,\n -3.6929e-04, -6.1181e-03, -1.6434e-03, -3.3540e-03, -3.4408e-03,\n -1.6314e-03, -9.1500e-04, -7.0155e-03, -5.4846e-04, 2.8126e-03,\n -7.4333e-04, -1.9405e-03, -3.2880e-05, -9.1282e-04, -2.3536e-03,\n -4.4223e-03, 3.1731e-03, -2.3925e-03, -6.4044e-04, 5.4111e-04,\n 3.3706e-03, 1.1672e-03, -6.5368e-04, -4.0866e-03, -3.1053e-05,\n -6.8132e-03, 3.4750e-03, 1.6849e-03, 2.2245e-03, -4.0527e-03,\n -7.8841e-05, 4.5716e-03, 2.7270e-03, -1.6207e-03, 1.2466e-03,\n -2.2339e-03, 1.8132e-03, -2.6257e-03, 6.0999e-04, 3.7132e-04,\n 4.5695e-03, 3.8663e-04, 1.0062e-03, 2.4301e-03, 2.0671e-04,\n 9.3171e-04, 2.7000e-03, 2.1378e-04, -1.0670e-03, -1.9897e-03,\n -1.0718e-03, -4.3459e-03, -8.4579e-03, -1.9016e-03, -3.4133e-03,\n -1.8527e-03, 4.9172e-05, -2.9636e-03, -5.3156e-05, -3.5039e-03,\n -4.3795e-03, 2.2596e-03, -4.6240e-04, 9.7416e-04, 3.4318e-03,\n 4.6988e-04, -2.1275e-03, -1.7066e-03, 4.6863e-03, 1.2256e-03,\n 1.9117e-03, -3.5641e-03, 2.1721e-04, 9.3837e-04, -2.0482e-03,\n 3.1870e-03, -1.8575e-03, 9.3930e-04, -2.6366e-03, 1.0659e-03,\n 2.0990e-03, 3.6482e-04, 1.1081e-03, 3.3554e-03, 3.7696e-03,\n -4.0759e-03, -1.0079e-03, 2.7841e-03, 1.8712e-03, -1.8022e-03,\n 4.2440e-04, 2.9884e-03, -1.7413e-03, 2.4906e-03, 1.2307e-03,\n -1.5759e-03, 5.8349e-03, 9.5376e-04, 1.8399e-03, -7.9760e-05,\n -1.0186e-03, -1.3060e-03, -1.3542e-03, 2.7161e-03, -5.9848e-04,\n 2.4873e-03, -8.6191e-05, -1.6870e-03, 5.1337e-03, 2.0545e-03,\n 1.4447e-03, -1.6558e-03, 2.1468e-04, -2.1253e-03, 1.3301e-03,\n 7.8942e-03, 2.1062e-03, -4.9350e-05, -2.3801e-03, -8.7209e-04,\n 3.0553e-03, -1.9568e-03, -3.4258e-03, -4.7832e-03, 3.5436e-04,\n 1.9804e-03, -3.5075e-03, -4.7697e-04, -1.3647e-04, 1.3686e-03,\n 5.0263e-03, -1.5575e-03, -3.0308e-04, -1.2371e-03, -3.3537e-03,\n -1.9197e-03, -8.7787e-04, -1.8275e-03, 3.3428e-03, -1.6492e-03,\n 1.4650e-03, 2.6495e-04, -4.3373e-03, -4.7413e-03, 4.8653e-03,\n -2.7373e-03, 2.0926e-03, -6.2167e-04, -3.6870e-03, -2.3643e-03,\n -1.2862e-03, 5.7621e-03], device='cuda:0')",
18
- "exp_avg_sq": "tensor([3.3338e-05, 4.0891e-05, 3.0563e-05, 3.0485e-05, 6.9688e-05, 6.0591e-05,\n 3.2258e-05, 2.7332e-05, 3.7099e-05, 4.1517e-05, 4.2407e-05, 8.6262e-05,\n 2.7901e-05, 4.0574e-05, 4.5581e-05, 3.9991e-05, 2.5208e-05, 8.2164e-05,\n 4.8659e-05, 5.2465e-05, 3.2466e-05, 4.9771e-05, 6.6870e-05, 6.5937e-05,\n 2.8035e-05, 4.1039e-05, 6.5331e-05, 3.5991e-05, 2.7179e-05, 4.1282e-05,\n 3.2037e-05, 2.2403e-05, 4.5448e-05, 2.5293e-05, 4.1587e-05, 4.8831e-05,\n 4.1625e-05, 6.0342e-05, 6.2297e-05, 5.2218e-05, 4.7676e-05, 4.4968e-05,\n 4.8250e-05, 4.0550e-05, 5.0440e-05, 4.4700e-05, 1.2086e-04, 4.6460e-05,\n 6.1279e-05, 1.2768e-04, 5.7797e-05, 4.3994e-05, 4.2931e-05, 3.3957e-05,\n 4.7253e-05, 3.8708e-05, 4.4073e-05, 7.4644e-05, 3.4958e-05, 4.9656e-05,\n 9.4695e-05, 4.8545e-05, 5.0398e-05, 3.8489e-05, 6.0528e-05, 3.0334e-05,\n 5.0380e-05, 6.3561e-05, 4.0139e-05, 4.4756e-05, 3.1186e-05, 4.7083e-05,\n 3.4496e-05, 3.6185e-05, 4.4828e-05, 5.1213e-05, 3.4582e-05, 1.2076e-04,\n 7.2060e-05, 2.7983e-05, 8.5515e-05, 4.1125e-05, 4.0997e-05, 4.1566e-05,\n 1.2052e-02, 2.8541e-05, 4.8259e-05, 2.7154e-05, 4.6101e-05, 4.2488e-05,\n 4.4813e-05, 4.3855e-05, 2.7527e-05, 8.1484e-05, 3.4683e-05, 4.4326e-05,\n 3.6619e-05, 3.7344e-05, 4.4350e-05, 4.5455e-05, 3.9298e-05, 3.2230e-05,\n 7.5612e-05, 3.4458e-05, 1.3283e-04, 3.1123e-05, 6.5756e-05, 4.5245e-05,\n 3.6632e-05, 3.9008e-05, 4.7731e-05, 3.9026e-05, 4.0048e-05, 4.4801e-05,\n 5.8117e-05, 4.5462e-05, 6.1690e-05, 1.0130e-04, 9.5146e-05, 4.2296e-05,\n 4.2426e-05, 2.4857e-05, 1.6508e-04, 4.3614e-05, 2.3850e-05, 3.0486e-05,\n 3.5641e-05, 4.0828e-05, 2.1999e-05, 3.5855e-05, 3.3825e-05, 1.8847e-04,\n 6.1550e-05, 3.6296e-05, 4.6309e-05, 2.9321e-05, 3.7758e-05, 1.3528e-04,\n 3.6601e-05, 5.3066e-05, 6.8005e-05, 2.6744e-05, 4.2724e-05, 6.5461e-05,\n 1.1547e-04, 3.8390e-05, 4.6391e-05, 4.3947e-05, 2.9390e-05, 3.0396e-05,\n 4.2064e-05, 4.1142e-05, 3.8589e-05, 3.8398e-05, 2.9110e-05, 3.9323e-05,\n 1.0418e-04, 5.2453e-05, 7.4684e-05, 4.3834e-05, 3.6885e-05, 4.1897e-05,\n 8.1258e-05, 3.9346e-05, 3.4214e-05, 4.8065e-05, 5.3357e-05, 7.2953e-05,\n 3.8760e-05, 4.1486e-05, 9.4745e-05, 4.7805e-05, 4.2539e-05, 7.3890e-05,\n 4.5774e-05, 3.7551e-05, 3.0696e-05, 3.2199e-05, 3.2666e-05, 3.6424e-05,\n 4.2887e-05, 4.9799e-05, 2.9519e-05, 2.7926e-05, 3.0067e-05, 4.1482e-05,\n 4.5119e-05, 4.0187e-05, 4.1558e-05, 5.1724e-05, 5.6956e-05, 6.1405e-05,\n 3.2051e-05, 4.1723e-05, 2.3355e-05, 3.5190e-05, 2.5219e-05, 3.7900e-05,\n 8.0215e-05, 4.7114e-05, 2.2117e-05, 4.5706e-05, 4.1451e-05, 3.1068e-05,\n 3.7383e-05, 2.7782e-05, 6.7053e-05, 3.8231e-05, 7.3848e-05, 6.8268e-05,\n 5.3507e-05, 6.4395e-05, 3.1594e-05, 3.1054e-05, 4.5699e-05, 5.8661e-05,\n 8.6892e-05, 2.9140e-05, 4.5533e-05, 4.0321e-05, 4.9775e-05, 4.2838e-05,\n 4.0789e-05, 6.5369e-05, 3.2006e-05, 4.8253e-05, 5.3961e-05, 2.9935e-05,\n 4.6455e-05, 4.2322e-05, 8.3053e-05, 4.8413e-05, 3.9313e-05, 3.9599e-05,\n 2.5295e-05, 6.6184e-05, 5.3513e-05, 4.4788e-05, 3.0427e-05, 3.0114e-05,\n 8.4766e-05, 5.5500e-05, 5.2937e-05, 2.5761e-05, 4.4927e-05, 4.2793e-05,\n 1.5413e-04, 3.8229e-05, 4.2327e-05, 8.5221e-05, 3.1582e-05, 4.4596e-05,\n 9.5355e-05, 1.5166e-03, 1.2044e-04, 4.8433e-05, 5.8765e-05, 5.6992e-05,\n 2.9083e-05, 4.6632e-05, 4.9326e-05, 5.6918e-05, 3.0022e-05, 3.9067e-05,\n 7.0922e-05, 3.1308e-05, 4.7772e-05, 2.9556e-05, 9.9351e-05, 3.3195e-05,\n 3.1044e-05, 5.7117e-05, 2.9549e-05, 3.8830e-05, 4.1554e-05, 2.8307e-05,\n 8.0694e-05, 3.0243e-05, 2.7085e-05, 4.2038e-05, 4.0116e-05, 2.7766e-05,\n 2.7337e-05, 5.5668e-05, 3.2775e-05, 8.7629e-05, 3.6763e-05, 5.1122e-05,\n 3.3466e-05, 5.1182e-05, 3.2806e-05, 4.4120e-05, 8.3699e-05, 5.5729e-05,\n 4.0231e-05, 4.3969e-05, 4.6422e-05, 2.9838e-05, 4.5303e-05, 2.7713e-05,\n 3.0975e-05, 3.9743e-05, 3.9954e-05, 3.5866e-05, 2.8956e-05, 2.5493e-05,\n 3.0681e-05, 5.0311e-05, 3.3731e-05, 4.1652e-05, 4.5238e-05, 3.4083e-05,\n 5.0303e-05, 2.9165e-05, 3.0727e-05, 7.0022e-05, 6.6552e-05, 4.6149e-05,\n 2.8913e-05, 6.2164e-05, 4.0083e-05, 9.8311e-05, 4.2723e-05, 4.0819e-05,\n 6.3002e-05, 4.0683e-05, 4.6628e-05, 3.3454e-05, 6.4425e-05, 5.0348e-05,\n 8.1974e-05, 6.2407e-05, 1.3249e-04, 3.5290e-05, 4.5062e-05, 5.4576e-05,\n 4.2159e-05, 4.9313e-05, 3.8148e-05, 4.7337e-05, 3.4497e-05, 7.6771e-05,\n 2.6173e-05, 4.5920e-05, 7.1593e-05, 3.5169e-05, 4.1971e-05, 4.1090e-05,\n 6.5914e-05, 3.7957e-05, 6.1370e-05, 4.8985e-05, 6.2919e-05, 3.2000e-05,\n 4.4119e-05, 4.8057e-05, 4.8254e-05, 3.1362e-05, 4.2126e-05, 2.6403e-05,\n 3.6094e-05, 3.6483e-05, 3.1101e-05, 7.8146e-05, 4.0186e-05, 6.6611e-05,\n 3.1327e-05, 1.2269e-04, 6.3248e-05, 3.1688e-05, 4.7406e-05, 5.7320e-05,\n 6.3094e-05, 4.9147e-05, 4.7798e-05, 4.5567e-05, 2.8825e-05, 2.7819e-05,\n 4.4674e-05, 4.0217e-05, 3.3449e-05, 6.0402e-05, 3.1742e-05, 6.2640e-05,\n 6.5805e-05, 6.9759e-05, 3.6058e-05, 5.5496e-05, 4.1946e-05, 4.7565e-05,\n 3.0564e-05, 5.2293e-05, 9.8895e-05, 5.3707e-05, 2.4360e-05, 5.1943e-05,\n 2.8603e-05, 5.7368e-05, 2.6586e-05, 6.6887e-05, 7.0007e-05, 4.0471e-05,\n 8.0214e-05, 4.6103e-05, 1.0529e-04, 3.7867e-05, 4.1558e-05, 3.6434e-05,\n 3.8410e-05, 5.8122e-05, 5.4547e-05, 7.0832e-05, 2.0009e-04, 3.2906e-05,\n 3.3789e-05, 3.0751e-05, 3.1551e-05, 4.5373e-05, 7.5756e-05, 3.9010e-05,\n 3.9137e-05, 3.6021e-05, 2.8622e-05, 1.1387e-04, 4.3064e-05, 3.9989e-05,\n 6.4569e-05, 3.5607e-05, 7.9118e-05, 5.4332e-05, 5.6026e-05, 3.5431e-05,\n 4.5171e-05, 3.7549e-05, 4.2373e-05, 3.9477e-05, 6.7538e-05, 2.1209e-05,\n 9.0285e-05, 2.8906e-05, 7.1810e-05, 3.7499e-05, 2.2367e-05, 2.7204e-05,\n 1.2068e-04, 4.3939e-05, 3.4320e-05, 2.1081e-05, 4.2563e-05, 2.9532e-05,\n 5.1960e-05, 4.3447e-05, 3.0266e-05, 4.1085e-05, 3.4545e-05, 5.0849e-05,\n 4.2884e-05, 2.1820e-04, 1.1029e-04, 3.3312e-05, 2.8168e-05, 3.0761e-05,\n 3.7158e-05, 3.5103e-05, 4.6120e-05, 2.4299e-05, 3.3562e-05, 6.1417e-05,\n 4.7847e-05, 2.6099e-05, 5.1415e-05, 2.7950e-05, 2.6324e-05, 6.1369e-05,\n 3.1025e-05, 5.6170e-05, 6.9819e-05, 3.3820e-05, 5.7520e-05, 3.0971e-05,\n 4.4657e-05, 4.1271e-05, 3.7743e-05, 1.4952e-04, 8.0749e-05, 5.6809e-05,\n 4.1760e-05, 4.0568e-05, 2.8321e-05, 3.3051e-05, 8.1581e-05, 6.9316e-05,\n 4.2774e-05, 6.8282e-05, 5.0559e-05, 3.6206e-05, 7.4123e-05, 2.7334e-05,\n 7.4851e-05, 4.0867e-05, 5.1206e-05, 3.4468e-05, 3.7989e-05, 6.8144e-05,\n 3.9566e-05, 7.2167e-05, 4.1969e-05, 4.5329e-05, 4.8923e-05, 5.1374e-05,\n 1.0516e-04, 6.6304e-05], device='cuda:0')"
19
  },
20
  "3": {
21
- "step": "tensor(2504.)",
22
- "exp_avg": "tensor([ 8.9917e-05, -3.0420e-03, 1.2294e-03, 1.1298e-03, 4.4695e-04,\n -1.5896e-04, 2.3489e-03, 5.1218e-04, -2.5857e-03, 6.4830e-04,\n 5.0049e-03, 4.8277e-03, -3.6673e-04, -5.9071e-04, -1.3323e-03,\n 7.0735e-04, 5.4100e-04, 5.4484e-03, -7.8413e-04, 1.6795e-03,\n -4.5967e-04, -4.6225e-04, 1.0743e-03, 9.1317e-04, 3.1566e-03,\n -4.9301e-03, -6.0988e-04, -1.0070e-03, 1.5133e-03, 5.3110e-04,\n -2.0917e-03, 1.0889e-03, -3.6228e-03, -1.0310e-03, -1.7813e-03,\n 5.0781e-03, -1.2648e-03, 4.5659e-03, 1.8726e-03, -8.2488e-03,\n -5.3662e-03, -4.5577e-03, -1.2795e-03, 8.6574e-05, 6.7374e-04,\n 1.4969e-03, 2.4582e-03, 3.1623e-03, 4.6868e-03, 9.1626e-03,\n 4.3317e-04, -2.7841e-03, -6.4129e-05, -3.1500e-04, 4.4490e-03,\n 2.6210e-04, 1.7395e-03, 1.8131e-03, -1.4469e-03, 3.0151e-03,\n 1.4076e-03, 7.6831e-05, -3.5573e-03, -2.3533e-03, 3.2342e-03,\n -4.0964e-03, 2.7848e-03, -4.0300e-03, 6.2388e-04, 4.4607e-03,\n 2.0500e-03, 6.0079e-04, -2.8188e-03, 2.6516e-03, -5.7826e-04,\n -9.8583e-04, 3.2623e-03, 3.9197e-03, 2.3227e-03, 2.0366e-03,\n 2.6013e-03, -2.1265e-03, 1.0606e-03, 1.5666e-03, 3.4473e-04,\n -2.2694e-03, 2.0563e-03, 9.0035e-04, 2.8518e-04, -1.4978e-04,\n 4.5770e-03, -2.0506e-03, -2.1597e-04, -1.1216e-03, 1.2157e-03,\n 1.0271e-03, -4.1984e-03, 1.8447e-03, 9.2633e-04, 1.3075e-03,\n -4.8616e-03, 1.9222e-03, -2.5959e-03, -1.2561e-03, 1.9927e-03,\n 1.5062e-03, -3.2645e-03, -2.5219e-03, -9.0800e-04, 2.3978e-03,\n 1.5769e-03, 2.2524e-03, -2.4335e-03, 1.6519e-04, -3.4093e-03,\n -3.4911e-03, 7.3580e-05, 9.0052e-04, -1.9897e-03, 1.5414e-03,\n -8.6098e-05, 5.7703e-04, 2.9754e-03, -2.1184e-03, -2.6288e-03,\n 3.3847e-05, -4.0541e-03, 4.5579e-04, 2.6660e-03, 6.9652e-04,\n 3.0925e-03, -2.4794e-04, 3.0848e-03, 2.4311e-03, 5.4496e-04,\n -2.0944e-03, 3.0986e-03, 6.6124e-04, 7.0823e-05, 3.5859e-03,\n 9.0701e-04, 9.3801e-04, -2.4942e-04, 1.3442e-03, 4.6447e-04,\n -2.2296e-04, 1.0089e-03, -2.4931e-03, -7.9820e-04, 1.2268e-03,\n -6.6358e-04, -1.1688e-03, 1.4302e-03, 2.2288e-03, 1.8461e-03,\n -1.0743e-03, 2.0535e-03, 1.1896e-03, -1.2377e-04, -2.4584e-03,\n -2.5335e-03, 2.8203e-03, -3.9421e-04, -1.6993e-03, 5.2307e-04,\n -2.6612e-03, -8.3595e-04, 8.8755e-04, 3.5700e-04, 1.2482e-03,\n 2.8439e-03, 1.6290e-03, -1.1897e-03, 1.2838e-03, 1.0668e-03,\n -1.2075e-03, -9.0450e-04, 2.4950e-03, 6.8186e-04, 4.3664e-03,\n 8.6564e-04, 2.7691e-04, -2.5230e-04, 3.6434e-03, -7.5357e-04,\n 2.1755e-03, 4.2306e-03, 1.1231e-03, -1.5554e-03, 7.2665e-03,\n 3.7809e-05, -3.5652e-03, -1.9000e-03, 4.5750e-03, 2.2088e-03,\n 3.6224e-04, 3.2673e-03, 4.4100e-03, 2.7837e-04, -1.9723e-03,\n -4.5632e-03, 6.6723e-04, -1.7103e-03, 1.9036e-03, 3.2725e-03,\n -3.4711e-03, 1.3666e-03, -1.4475e-03, -1.6423e-03, 6.7216e-03,\n -2.0613e-03, 2.9919e-03, -7.7183e-04, 1.2272e-03, -2.3518e-03,\n -4.0081e-03, 4.3316e-04, -1.1995e-03, 2.1883e-03, -8.7992e-04,\n 4.4694e-03, -4.6982e-03, 1.9549e-03, 9.7641e-04, 2.2499e-04,\n -7.7782e-04, -6.7043e-03, -2.8313e-03, 2.9655e-03, 2.1560e-04,\n -2.9077e-03, 2.4158e-03, -2.7133e-03, -4.2326e-03, 1.9694e-03,\n 1.2418e-03, -2.8491e-03, -8.5384e-04, -1.5333e-03, -8.7761e-04,\n 5.0640e-03, 6.3565e-04, 2.7060e-03, -3.1588e-03, 3.0271e-03,\n 1.8391e-04, -3.9307e-03, -1.7444e-03, 1.3427e-03, 2.0490e-03,\n 2.5167e-03, -1.7926e-03, -5.3525e-04, -4.8386e-03, 2.2028e-03,\n -1.8276e-04, -1.1485e-03, -3.0334e-03, 5.5356e-04, -5.2031e-03,\n -5.1544e-05, -4.7181e-03, 1.3528e-03, 2.1601e-03, 1.9454e-03,\n 1.1115e-03, -4.4911e-03, -1.2219e-04, 5.9228e-05, -1.4596e-03,\n 5.2705e-03, -4.1427e-03, -1.1032e-03, 3.6561e-04, -3.4662e-04,\n 7.0439e-04, -1.6487e-04, 3.4376e-03, -1.5846e-03, -1.5308e-03,\n -2.0193e-03, -1.6362e-03, 3.3395e-03, 8.1598e-04, -1.3840e-03,\n 2.3317e-03, 2.7890e-03, 2.3051e-03, -1.1115e-03, 4.3508e-03,\n -6.1658e-04, 3.7616e-03, 2.8322e-03, -1.3327e-03, -1.3437e-03,\n -2.1145e-03, -5.0888e-03, 7.1882e-05, -1.1873e-03, 1.4272e-04,\n 4.7769e-03, 3.2790e-03, -5.3356e-03, 1.2796e-03, 8.9122e-04,\n -1.2598e-03, 4.3833e-05, 3.0462e-03, 8.9773e-04, -1.1985e-03,\n 1.9996e-03, -1.2123e-03, -1.4262e-04, -2.6297e-03, 2.0249e-03,\n -4.8508e-04, -6.5914e-03, -2.5845e-03, 2.4243e-03, -2.5863e-03,\n -1.2570e-03, -4.6150e-04, -3.3606e-03, 3.3374e-04, -7.2765e-04,\n 2.3603e-03, -4.7575e-03, 5.8740e-05, -4.3198e-04, 1.4595e-03,\n -7.3505e-04, -3.9129e-03, 3.0226e-03, -2.3129e-03, 1.7259e-03,\n 8.9348e-04, -1.3704e-04, -2.5004e-03, -4.7108e-04, -3.2472e-03,\n 7.1299e-04, 4.3262e-03, -8.1627e-04, -3.0621e-03, 1.0154e-03,\n 2.2113e-03, -1.7223e-03, -1.2426e-03, 2.0201e-03, 1.1639e-03,\n -1.0558e-03, -1.3026e-03, 1.4910e-03, 2.6188e-03, -1.4889e-04,\n 1.9519e-03, -7.7098e-03, -1.8506e-03, -3.7976e-03, 1.7584e-05,\n -5.8030e-04, -5.9859e-03, -1.1342e-03, -2.2973e-03, -4.3251e-03,\n -1.1177e-03, -2.0530e-04, -4.0849e-03, 3.6634e-04, 2.5533e-03,\n -2.3194e-04, -2.3337e-03, 2.8768e-04, -1.5326e-03, -2.3506e-03,\n -4.3306e-03, 3.5664e-03, -2.8101e-03, -2.9573e-04, 1.3998e-03,\n 5.0543e-03, 1.0788e-03, -7.5632e-04, -1.3395e-03, 4.8853e-04,\n -5.1814e-03, 4.0470e-03, 1.3514e-03, 4.1408e-03, -2.5601e-03,\n 1.4906e-05, 4.7026e-03, 1.8089e-03, -6.2092e-04, 1.3581e-03,\n -1.7140e-03, 2.0595e-03, -2.7918e-03, 9.3435e-04, -1.2953e-04,\n 3.3657e-03, 5.0090e-04, 3.0030e-04, 2.6848e-03, 5.0510e-04,\n 1.3778e-03, 2.2635e-03, -2.9580e-04, -9.5765e-04, -5.7443e-04,\n -1.7492e-03, -2.9054e-03, -3.3299e-03, -2.0697e-03, -3.9704e-03,\n -2.1614e-03, 5.5759e-04, -1.8767e-03, 6.7070e-04, -3.5040e-03,\n -4.0416e-03, 2.2358e-03, -2.4251e-04, 1.0916e-03, 4.0181e-03,\n 1.3981e-03, -1.7760e-03, -2.0537e-03, 4.3788e-03, 7.9743e-04,\n 1.8602e-03, -3.7468e-03, 1.1496e-03, 1.2845e-03, -2.3780e-03,\n 2.4694e-03, -1.9479e-03, -1.9612e-06, -1.9247e-03, 1.6869e-04,\n 1.3277e-03, -4.3989e-04, 1.3341e-03, 3.7281e-03, 2.9744e-03,\n -3.7963e-03, -8.1581e-04, 3.3452e-03, 3.4099e-03, -1.5293e-03,\n 3.1634e-04, 2.4335e-03, -2.4368e-03, 3.2621e-03, 1.4959e-03,\n -1.5010e-03, 4.6088e-03, 7.8593e-04, 1.0325e-03, -5.4917e-04,\n -1.5238e-03, -5.0696e-04, -1.5362e-03, 4.0826e-03, -4.5589e-04,\n 3.6351e-03, -2.7325e-04, -2.2669e-03, 4.5547e-03, 2.9368e-03,\n 2.2025e-03, -2.8731e-03, -4.1435e-06, -1.0801e-03, 9.2627e-04,\n 5.8449e-03, 2.0699e-03, 3.3410e-04, -1.8674e-03, -4.4818e-03,\n 1.8903e-03, -6.3200e-04, -3.0064e-03, -2.4544e-03, -1.9081e-04,\n 2.7982e-03, -2.8229e-03, -1.6074e-04, -5.1889e-04, 1.2354e-03,\n 3.8189e-03, -2.6756e-03, -2.2398e-04, -1.4240e-03, -4.2980e-03,\n -2.1273e-03, -7.5733e-04, -2.9767e-03, 2.0242e-03, -2.5289e-03,\n 1.7065e-03, -2.6333e-04, -4.1181e-03, -3.8257e-03, 3.6514e-03,\n -9.0975e-04, 3.8315e-03, -1.1475e-03, -7.7796e-03, -3.0676e-03,\n -1.6739e-03, 3.8984e-03], device='cuda:0')",
23
- "exp_avg_sq": "tensor([3.3736e-05, 3.6434e-05, 4.1709e-05, 3.7326e-05, 3.8954e-05, 4.0823e-05,\n 4.6864e-05, 3.0956e-05, 4.1726e-05, 3.8032e-05, 4.6580e-05, 5.6509e-05,\n 2.7508e-05, 4.3925e-05, 4.0546e-05, 4.2849e-05, 3.3014e-05, 4.6234e-05,\n 3.8647e-05, 4.6490e-05, 4.4832e-05, 3.8594e-05, 4.9757e-05, 4.3338e-05,\n 3.5136e-05, 4.1215e-05, 5.2571e-05, 3.3211e-05, 3.2871e-05, 4.4251e-05,\n 3.5612e-05, 3.4185e-05, 5.1387e-05, 2.9258e-05, 3.4760e-05, 6.5154e-05,\n 4.5559e-05, 4.4510e-05, 4.7246e-05, 5.2719e-05, 4.6747e-05, 4.1524e-05,\n 4.2328e-05, 5.6047e-05, 3.2630e-05, 4.7100e-05, 4.3567e-05, 4.8464e-05,\n 5.3490e-05, 6.0090e-05, 4.9307e-05, 3.9952e-05, 5.1035e-05, 4.3162e-05,\n 5.0659e-05, 3.6601e-05, 3.5794e-05, 4.7385e-05, 3.1246e-05, 4.7045e-05,\n 4.8189e-05, 3.7541e-05, 4.4120e-05, 5.8228e-05, 4.4229e-05, 4.9910e-05,\n 4.4350e-05, 4.3282e-05, 5.4539e-05, 5.2884e-05, 3.7264e-05, 6.6123e-05,\n 4.1163e-05, 4.8005e-05, 3.8516e-05, 3.9600e-05, 3.7953e-05, 4.2824e-05,\n 3.9241e-05, 3.1625e-05, 5.2310e-05, 4.7871e-05, 4.3754e-05, 3.4968e-05,\n 4.5685e-05, 3.5284e-05, 4.8719e-05, 3.5001e-05, 4.1217e-05, 4.7535e-05,\n 4.8739e-05, 3.8508e-05, 3.2665e-05, 5.7398e-05, 4.0627e-05, 5.0222e-05,\n 3.9793e-05, 3.8483e-05, 4.1388e-05, 5.1579e-05, 4.1830e-05, 3.3876e-05,\n 6.0335e-05, 3.9258e-05, 5.1080e-05, 3.4070e-05, 3.8036e-05, 4.8108e-05,\n 4.8421e-05, 3.9490e-05, 4.5648e-05, 4.4042e-05, 4.5612e-05, 4.1139e-05,\n 3.7587e-05, 3.8999e-05, 5.6516e-05, 4.4622e-05, 3.7583e-05, 4.9762e-05,\n 4.3106e-05, 3.3949e-05, 4.9991e-05, 3.6409e-05, 2.9639e-05, 4.1084e-05,\n 4.4581e-05, 4.3976e-05, 2.9839e-05, 4.6098e-05, 3.6895e-05, 4.6748e-05,\n 3.7251e-05, 3.5404e-05, 3.7890e-05, 4.1340e-05, 4.8282e-05, 4.5316e-05,\n 4.7608e-05, 4.6685e-05, 4.0505e-05, 3.6900e-05, 3.7792e-05, 5.3334e-05,\n 4.6062e-05, 4.8727e-05, 6.1526e-05, 5.4980e-05, 3.7449e-05, 4.1446e-05,\n 4.3177e-05, 4.6680e-05, 4.0604e-05, 3.9226e-05, 4.2400e-05, 4.5929e-05,\n 4.8589e-05, 3.7972e-05, 4.1099e-05, 4.0337e-05, 4.2936e-05, 4.3175e-05,\n 4.5672e-05, 3.2920e-05, 4.6721e-05, 4.9957e-05, 4.5714e-05, 4.0654e-05,\n 4.2467e-05, 5.7131e-05, 5.5352e-05, 4.0332e-05, 3.4274e-05, 4.5767e-05,\n 5.7389e-05, 3.5324e-05, 4.0264e-05, 3.8352e-05, 5.1005e-05, 4.4270e-05,\n 4.5728e-05, 4.1488e-05, 4.5895e-05, 3.9379e-05, 4.1569e-05, 4.6940e-05,\n 4.8529e-05, 3.1628e-05, 3.5177e-05, 4.8667e-05, 4.5456e-05, 3.8144e-05,\n 3.7025e-05, 3.8066e-05, 2.8926e-05, 4.7132e-05, 3.4906e-05, 5.2302e-05,\n 4.4911e-05, 5.0082e-05, 3.1965e-05, 3.5833e-05, 4.6918e-05, 4.0833e-05,\n 3.7879e-05, 3.6977e-05, 4.3264e-05, 4.2743e-05, 4.4139e-05, 7.1391e-05,\n 4.7637e-05, 4.3952e-05, 4.8760e-05, 3.6682e-05, 5.3621e-05, 4.3647e-05,\n 4.5712e-05, 3.2405e-05, 4.5507e-05, 5.0719e-05, 4.3420e-05, 5.1040e-05,\n 4.0879e-05, 4.0967e-05, 4.5448e-05, 4.1189e-05, 4.4245e-05, 4.2507e-05,\n 4.0928e-05, 3.6604e-05, 4.0612e-05, 4.8287e-05, 5.2083e-05, 5.3437e-05,\n 3.2612e-05, 4.6040e-05, 4.0749e-05, 5.8173e-05, 3.6049e-05, 3.6807e-05,\n 3.9965e-05, 4.9990e-05, 5.0752e-05, 4.1424e-05, 5.7443e-05, 5.9286e-05,\n 4.3949e-05, 4.2623e-05, 4.6117e-05, 4.9153e-05, 3.8776e-05, 4.1748e-05,\n 4.5661e-05, 8.0837e-05, 4.5475e-05, 5.0255e-05, 5.4884e-05, 4.8668e-05,\n 3.9271e-05, 4.0738e-05, 3.8562e-05, 4.5313e-05, 3.6318e-05, 5.0843e-05,\n 5.2422e-05, 3.7703e-05, 4.6778e-05, 3.2242e-05, 4.1812e-05, 3.7624e-05,\n 4.7966e-05, 4.4998e-05, 3.3613e-05, 5.0112e-05, 4.2086e-05, 3.7900e-05,\n 5.6099e-05, 4.6134e-05, 3.4342e-05, 5.3476e-05, 4.1961e-05, 4.0835e-05,\n 4.0187e-05, 4.7058e-05, 3.2191e-05, 4.3556e-05, 4.7064e-05, 4.9042e-05,\n 4.0130e-05, 5.8467e-05, 3.4754e-05, 4.2983e-05, 4.5356e-05, 5.6286e-05,\n 4.2784e-05, 6.0034e-05, 4.8141e-05, 3.8484e-05, 4.7390e-05, 3.6471e-05,\n 4.6066e-05, 3.9216e-05, 4.8280e-05, 3.9756e-05, 3.0531e-05, 3.1449e-05,\n 3.4472e-05, 4.9294e-05, 3.2542e-05, 4.4331e-05, 5.7695e-05, 3.9121e-05,\n 3.8296e-05, 4.2099e-05, 4.1264e-05, 5.1192e-05, 5.6671e-05, 4.3284e-05,\n 3.4434e-05, 3.8229e-05, 3.6066e-05, 4.6237e-05, 3.9243e-05, 4.3876e-05,\n 4.7931e-05, 4.4079e-05, 4.1089e-05, 3.6351e-05, 4.3073e-05, 4.7317e-05,\n 5.7516e-05, 4.3183e-05, 5.2985e-05, 4.3430e-05, 3.9058e-05, 3.6664e-05,\n 5.1573e-05, 5.0111e-05, 3.9704e-05, 4.1086e-05, 3.6613e-05, 4.8327e-05,\n 3.7092e-05, 4.1743e-05, 3.7607e-05, 4.5090e-05, 4.3875e-05, 4.0641e-05,\n 4.3031e-05, 4.8300e-05, 3.9341e-05, 4.2960e-05, 4.9949e-05, 4.1170e-05,\n 4.3380e-05, 7.2107e-05, 4.4648e-05, 3.4634e-05, 3.9042e-05, 3.6508e-05,\n 4.2057e-05, 5.0828e-05, 3.8524e-05, 4.9109e-05, 4.8777e-05, 6.2724e-05,\n 3.8703e-05, 6.3575e-05, 4.1336e-05, 3.3405e-05, 5.0081e-05, 5.6096e-05,\n 4.0749e-05, 4.9569e-05, 4.6237e-05, 3.6991e-05, 3.5160e-05, 4.6144e-05,\n 4.2750e-05, 4.2850e-05, 4.2430e-05, 5.0142e-05, 4.4393e-05, 4.2688e-05,\n 4.2952e-05, 5.1299e-05, 4.8739e-05, 4.5900e-05, 4.4586e-05, 4.6682e-05,\n 3.9295e-05, 4.8560e-05, 4.5711e-05, 6.1421e-05, 4.1188e-05, 5.2910e-05,\n 3.4918e-05, 4.6601e-05, 3.5615e-05, 5.4021e-05, 3.6371e-05, 4.0008e-05,\n 5.0484e-05, 6.0602e-05, 5.2016e-05, 4.6905e-05, 4.9344e-05, 5.0744e-05,\n 3.9600e-05, 3.7122e-05, 4.5651e-05, 4.5818e-05, 5.0732e-05, 4.8824e-05,\n 4.2425e-05, 3.4385e-05, 2.9117e-05, 4.4424e-05, 4.5195e-05, 5.1555e-05,\n 4.2542e-05, 5.3421e-05, 4.2313e-05, 5.1024e-05, 4.3822e-05, 4.0749e-05,\n 4.2974e-05, 4.4677e-05, 5.3972e-05, 4.0085e-05, 4.9099e-05, 4.0850e-05,\n 4.6226e-05, 3.5169e-05, 4.1771e-05, 4.1796e-05, 5.1961e-05, 3.1743e-05,\n 5.3680e-05, 3.4728e-05, 4.1830e-05, 4.1502e-05, 3.3666e-05, 3.1955e-05,\n 4.9287e-05, 4.1483e-05, 3.6275e-05, 3.1774e-05, 5.2560e-05, 4.6278e-05,\n 4.3703e-05, 3.9741e-05, 3.6813e-05, 4.3888e-05, 4.5666e-05, 4.6355e-05,\n 4.3711e-05, 5.2794e-05, 4.5557e-05, 5.4337e-05, 3.4692e-05, 2.9921e-05,\n 4.8077e-05, 4.4557e-05, 4.9606e-05, 3.6232e-05, 4.2805e-05, 4.6788e-05,\n 4.4933e-05, 3.7779e-05, 5.8835e-05, 2.9105e-05, 3.8335e-05, 3.5966e-05,\n 3.7206e-05, 4.0431e-05, 4.9238e-05, 5.0712e-05, 5.1801e-05, 4.1612e-05,\n 4.6747e-05, 4.3679e-05, 3.9390e-05, 4.8508e-05, 4.5384e-05, 4.2665e-05,\n 4.3727e-05, 4.4248e-05, 3.5352e-05, 4.3236e-05, 5.1203e-05, 5.7764e-05,\n 5.0794e-05, 5.0396e-05, 5.6937e-05, 4.2174e-05, 5.3101e-05, 4.5798e-05,\n 5.0420e-05, 5.0624e-05, 4.5248e-05, 5.0828e-05, 3.4861e-05, 4.1301e-05,\n 4.0070e-05, 4.1750e-05, 5.8413e-05, 5.3766e-05, 8.5195e-05, 4.5674e-05,\n 4.9283e-05, 4.2098e-05], device='cuda:0')"
24
  },
25
  "4": {
26
- "step": "tensor(2504.)",
27
- "exp_avg": "tensor([[-3.0799e-04, 2.0483e-04, 2.4507e-04, ..., 3.1225e-04,\n -1.1319e-04, 6.7023e-04],\n [-2.1215e-04, -3.2028e-05, 5.8429e-04, ..., 5.1161e-04,\n 2.0803e-05, 2.5227e-04],\n [-3.2354e-06, 3.2014e-04, -9.3487e-05, ..., 1.7889e-04,\n 1.8077e-04, 1.0577e-04],\n ...,\n [-2.0836e-04, -8.7245e-05, -8.3901e-05, ..., -2.9493e-04,\n -4.0131e-04, 6.1174e-04],\n [ 3.1606e-04, 2.8501e-05, 1.6351e-04, ..., -3.8017e-04,\n -1.6401e-04, 1.7792e-04],\n [-2.6547e-04, -1.3828e-04, -7.1850e-04, ..., -7.3453e-04,\n -1.4188e-04, 4.1676e-04]], device='cuda:0')",
28
- "exp_avg_sq": "tensor([[1.0335e-06, 1.3667e-06, 7.3996e-07, ..., 9.6428e-07, 1.3700e-06,\n 1.5905e-06],\n [2.2564e-06, 1.5947e-06, 2.8067e-06, ..., 2.4318e-06, 2.9675e-06,\n 2.4193e-06],\n [1.7887e-06, 1.4784e-06, 1.6819e-06, ..., 2.5790e-06, 2.4780e-06,\n 1.9698e-06],\n ...,\n [2.7667e-06, 1.6971e-06, 1.7593e-06, ..., 2.3946e-06, 3.1129e-06,\n 3.1270e-06],\n [2.2088e-06, 1.7304e-06, 1.5778e-06, ..., 2.1998e-06, 2.9607e-06,\n 3.0924e-06],\n [2.1896e-06, 2.0372e-06, 2.7445e-06, ..., 2.4318e-06, 3.3994e-06,\n 3.0873e-06]], device='cuda:0')"
29
  },
30
  "5": {
31
- "step": "tensor(2504.)",
32
- "exp_avg": "tensor([[-2.6328e-04, -1.0174e-04, 1.1467e-04, ..., 1.9829e-04,\n -7.9036e-05, -1.4280e-04],\n [-2.9791e-04, 5.4211e-05, 1.7471e-04, ..., 2.6994e-04,\n -2.7813e-06, 1.0776e-04],\n [ 1.6449e-04, 2.9613e-04, -8.6117e-05, ..., -1.4828e-05,\n 2.2731e-04, 2.8958e-04],\n ...,\n [-8.2474e-05, -7.0737e-05, 9.5644e-05, ..., 1.0410e-04,\n 1.1953e-05, -2.0145e-04],\n [ 1.0294e-03, 8.6253e-05, -2.7206e-04, ..., -1.2304e-04,\n -4.3338e-05, 1.2365e-04],\n [-3.0715e-04, 1.7091e-05, 2.0290e-04, ..., -1.3995e-04,\n -2.6561e-04, 4.7502e-04]], device='cuda:0')",
33
- "exp_avg_sq": "tensor([[2.7553e-07, 2.0770e-07, 2.2084e-07, ..., 3.2068e-07, 3.7935e-07,\n 3.6687e-07],\n [5.8158e-07, 3.8378e-07, 4.5756e-07, ..., 6.0932e-07, 6.8479e-07,\n 6.5814e-07],\n [5.6275e-07, 4.8793e-07, 4.7595e-07, ..., 7.0090e-07, 6.6818e-07,\n 6.2589e-07],\n ...,\n [6.2341e-07, 5.1133e-07, 6.7648e-07, ..., 6.0365e-07, 8.8705e-07,\n 7.8835e-07],\n [5.3485e-07, 5.9973e-07, 5.0700e-07, ..., 6.0193e-07, 7.5677e-07,\n 7.4837e-07],\n [5.4395e-07, 5.3297e-07, 5.7980e-07, ..., 6.7589e-07, 8.2242e-07,\n 6.7653e-07]], device='cuda:0')"
34
  },
35
  "6": {
36
- "step": "tensor(2504.)",
37
- "exp_avg": "tensor([-0.0001, 0.0001], device='cuda:0')",
38
- "exp_avg_sq": "tensor([1.5169e-05, 1.5169e-05], device='cuda:0')"
39
  }
40
  },
41
  "param_groups": [
42
  {
43
- "lr": 0.0009046039886902864,
44
  "name": "shared",
45
  "betas": [
46
  0.9,
@@ -64,7 +64,7 @@
64
  ]
65
  },
66
  {
67
- "lr": 0.0009046039886902864,
68
  "name": "scale_256",
69
  "betas": [
70
  0.9,
@@ -85,7 +85,7 @@
85
  ]
86
  },
87
  {
88
- "lr": 0.0009046039886902864,
89
  "name": "scale_512",
90
  "betas": [
91
  0.9,
@@ -106,7 +106,7 @@
106
  ]
107
  },
108
  {
109
- "lr": 0.0004523497400965494,
110
  "name": "fusion",
111
  "betas": [
112
  0.9,
@@ -133,30 +133,30 @@
133
  "T_i": 10,
134
  "T_mult": 2,
135
  "eta_min": 1e-06,
136
- "T_cur": 2,
137
  "base_lrs": [
138
  0.001,
139
  0.001,
140
  0.001,
141
  0.0005
142
  ],
143
- "last_epoch": 2,
144
  "_step_count": 0,
145
  "_is_initial": false,
146
  "_get_lr_called_within_step": false,
147
  "_last_lr": [
148
- 0.0009046039886902864,
149
- 0.0009046039886902864,
150
- 0.0009046039886902864,
151
- 0.0004523497400965494
152
  ]
153
  },
154
  "metrics": {
155
- "best_val_acc": 69.506,
156
- "best_epoch": 1,
157
  "scale_accuracies": {
158
- "256": 68.852,
159
- "512": 69.342
160
  }
161
  },
162
  "train_config": {
 
1
  {
2
+ "epoch": 2,
3
  "optimizer_state_dict": {
4
  "state": {
5
  "0": {
6
+ "step": "tensor(3756.)",
7
+ "exp_avg": "tensor([[ 6.7759e-04, 2.4889e-04, -6.7273e-04, ..., 7.0414e-05,\n -2.8108e-04, -1.9840e-04],\n [-1.2107e-03, 1.6546e-03, 5.5194e-04, ..., -5.9277e-05,\n 2.6390e-04, -1.2140e-03],\n [ 9.7542e-05, -6.3067e-04, 1.6014e-03, ..., -6.8758e-04,\n 7.3871e-04, -1.0380e-03],\n ...,\n [ 7.1288e-05, -3.0128e-03, 3.0054e-03, ..., -1.0032e-04,\n -9.0897e-05, -5.2043e-04],\n [ 1.0580e-03, 2.9030e-03, -4.8911e-04, ..., 4.8776e-04,\n -3.4841e-04, 5.6367e-04],\n [-3.5929e-05, 1.1549e-03, -1.9907e-03, ..., -3.2192e-04,\n -7.5701e-06, 6.1974e-04]], device='cuda:0')",
8
+ "exp_avg_sq": "tensor([[3.9302e-06, 3.3157e-05, 1.4172e-05, ..., 3.0597e-06, 2.2477e-06,\n 5.1279e-06],\n [1.0176e-05, 3.3417e-05, 1.7713e-05, ..., 4.0512e-06, 2.3869e-06,\n 5.4166e-06],\n [8.1370e-06, 5.2533e-05, 2.2265e-05, ..., 5.8253e-06, 3.4015e-06,\n 7.3562e-06],\n ...,\n [6.6892e-06, 5.2534e-05, 1.9725e-05, ..., 5.1486e-06, 3.1852e-06,\n 3.8046e-06],\n [4.3659e-06, 3.6172e-05, 1.3145e-05, ..., 3.3067e-06, 2.2730e-06,\n 2.7602e-06],\n [3.1946e-06, 1.5721e-05, 1.1272e-05, ..., 2.6413e-06, 1.5627e-06,\n 2.3424e-06]], device='cuda:0')"
9
  },
10
  "1": {
11
+ "step": "tensor(3756.)",
12
+ "exp_avg": "tensor([ 1.7411e-02, -1.1874e-02, -1.9065e-03, -5.2130e-02, -1.2272e-02,\n 1.0265e-02, 6.7949e-04, 2.8304e-02, 2.4067e-02, -2.3813e-02,\n 1.0505e-02, 1.4477e-02, 7.8596e-03, 1.9951e-02, -1.5036e-02,\n 1.3441e-02, 5.6562e-02, 5.6740e-03, -2.0264e-02, -2.2209e-02,\n -1.0204e-02, 7.4248e-03, -1.6354e-02, 1.6518e-02, 8.5704e-03,\n -8.1644e-03, 1.4745e-02, -3.7807e-02, 1.1805e-02, -1.2838e-02,\n -3.0237e-02, -1.3807e-03, -2.0089e-02, -1.6855e-02, -3.7908e-02,\n -4.8549e-03, 1.5170e-02, 3.5482e-03, 5.3220e-03, -1.3672e-04,\n 1.3323e-02, -2.7579e-02, -7.3242e-03, -2.7761e-03, 9.2570e-03,\n -9.6880e-03, 1.2684e-02, -2.7631e-02, 6.8908e-03, 4.2355e-03,\n 2.0213e-02, 1.6986e-02, 4.5031e-02, -5.9697e-03, 8.3972e-03,\n -1.6652e-02, -1.0651e-03, -6.8880e-03, 1.6510e-02, -9.1885e-03,\n -2.1874e-02, -2.2878e-03, -2.9805e-02, -2.6707e-02, 3.1669e-02,\n 2.7287e-02, 3.6599e-02, 1.1398e-02, -1.4077e-02, 5.7601e-02,\n 1.0350e-02, -3.2286e-02, 2.5639e-02, 2.8799e-03, -4.2079e-02,\n 1.3095e-02, -2.2711e-02, 1.2157e-03, -1.9258e-02, 2.9977e-02,\n -4.2100e-03, -2.2425e-02, -1.5842e-02, 2.1389e-02, -7.4605e-03,\n 6.2837e-03, 8.2695e-03, 1.7845e-02, -1.3893e-02, -7.0633e-03,\n 2.1715e-02, -1.7408e-03, -1.1715e-02, -9.9260e-03, -1.2619e-02,\n 1.3420e-02, 6.8752e-03, -1.1958e-02, 1.1197e-02, 3.8789e-02,\n 4.1452e-03, -7.5388e-02, 5.7724e-02, -5.0349e-02, -3.8428e-02,\n -6.9296e-03, -3.9047e-03, -6.8285e-03, -1.2564e-02, 8.2814e-03,\n -5.9261e-03, 2.5579e-02, 9.6076e-03, -6.5807e-03, -4.2545e-03,\n -1.5576e-02, -1.4667e-03, 1.0124e-02, -3.6799e-03, -6.3001e-03,\n 3.7747e-02, 2.1118e-02, -2.1845e-02, 2.3380e-02, 1.6891e-02,\n 6.9585e-03, -1.9818e-03, -2.7730e-02, 3.7806e-03, -1.4143e-02,\n 1.7361e-02, 9.5980e-03, -2.1022e-02, -1.6021e-02, 1.2501e-02,\n 2.1900e-02, 1.9070e-02, -1.6674e-02, -1.1469e-02, 3.7149e-04,\n -1.2773e-02, 1.7138e-02, 6.3488e-03, 2.5025e-02, 1.3616e-02,\n -5.1356e-02, 3.7676e-03, 3.0618e-04, 8.2975e-03, -9.8148e-03,\n 2.3626e-02, 1.2816e-02, -6.9709e-03, -6.5145e-03, 1.1106e-02,\n 2.0044e-02, 3.7376e-03, 8.4615e-03, -7.1695e-03, -5.9623e-03,\n 3.6773e-03, 1.8339e-02, 1.2664e-02, 1.2980e-02, -2.0862e-02,\n 6.1726e-02, 8.5959e-03, 5.6376e-03, 1.9665e-03, -1.6294e-02,\n -1.8069e-02, -3.5302e-02, 3.3266e-02, 5.8151e-03, 3.9958e-04,\n -1.3710e-02, -1.9611e-02, -1.1977e-02, 3.0796e-02, 2.8114e-02,\n -2.7367e-02, -6.0602e-03, 2.1910e-02, -1.2944e-02, -4.6425e-02,\n -6.2400e-03, -2.2634e-02, 8.1232e-03, -7.6569e-03, -3.0810e-02,\n 1.7233e-02, 1.1559e-03, 1.8185e-02, 1.1797e-02, 7.9183e-03,\n -1.3294e-02, 8.9931e-03, 3.7951e-02, -2.0761e-02, -8.6475e-03,\n -3.7930e-02, 1.9567e-02, 8.4255e-03, 7.6728e-03, 3.3809e-02,\n 3.2640e-02, 1.8014e-02, -4.7719e-03, -2.5308e-02, 2.8988e-03,\n -2.3128e-02, -3.1888e-02, -1.9161e-02, -8.4491e-02, -3.9199e-02,\n -1.5904e-02, 2.0504e-02, 1.3673e-02, -6.2343e-03, -1.9469e-02,\n 7.4697e-03, -1.0698e-02, 9.5473e-03, -1.7708e-02, 2.8500e-02,\n -4.1380e-02, 2.5171e-02, 3.8445e-02, -1.9921e-02, -9.7629e-03,\n -2.3881e-02, 2.9523e-02, 5.2499e-03, 6.6042e-04, 4.5049e-03,\n 2.6912e-03, 5.0633e-03, -3.4267e-02, 1.2434e-02, 1.1670e-02,\n 1.1430e-02, 5.3995e-03, -2.0165e-02, 8.2320e-03, 1.8068e-02,\n -1.3855e-02, -1.5324e-03, -1.4178e-02, 1.2484e-02, -2.0311e-02,\n -1.4352e-02, 1.3633e-02, 1.3931e-02, 8.9267e-03, -1.5198e-02,\n -3.5871e-03, 7.1717e-02, 9.9890e-03, 1.9387e-02, -1.9858e-02,\n 1.5297e-04, 2.7845e-02, 1.3955e-02, -1.0555e-02, 1.9291e-02,\n 8.4865e-03, 8.7312e-03, -1.5420e-02, -5.2619e-03, -3.7042e-03,\n -3.1753e-02, -1.7973e-02, 6.0104e-03, -4.5360e-03, -1.6389e-02,\n -1.8280e-02, -2.0305e-03, -1.3018e-02, 1.0146e-02, 2.5714e-02,\n -1.8377e-02, -5.2177e-02, 1.7294e-02, -2.3993e-02, 2.9097e-03,\n -6.8114e-03, -7.7761e-03, 3.1384e-02, 2.2878e-02, 1.0246e-02,\n -9.5864e-03, 3.8713e-02, -2.7545e-02, 1.1753e-02, 1.6151e-02,\n 3.8893e-02, 5.8228e-03, 2.9968e-02, 1.6263e-02, -2.9204e-02,\n -9.4987e-03, 1.5473e-02, -3.9827e-02, -5.7825e-03, 1.0673e-02,\n -1.9801e-02, 4.9887e-03, 1.0815e-03, 3.4477e-02, -1.6073e-03,\n -4.6376e-02, 8.8496e-03, -1.5458e-02, -9.0136e-02, 9.5835e-03,\n 7.3576e-03, 1.9505e-02, -1.8352e-02, -4.8987e-03, -2.2272e-02,\n -1.3036e-02, -1.0235e-02, 1.4999e-03, -1.8184e-02, -3.1290e-02,\n -6.8669e-02, 6.9400e-02, 2.2298e-02, -1.0844e-02, -3.5607e-04,\n 9.0220e-03, 1.7932e-02, -9.0903e-03, -4.6994e-02, -2.4301e-03,\n -2.0550e-03, -6.9805e-03, 2.4629e-02, -2.0913e-02, 1.2914e-02,\n 4.8193e-03, 4.5746e-03, 2.2739e-04, -2.4903e-02, -1.1087e-02,\n -2.9622e-02, 2.6174e-02, -2.0518e-02, -1.5619e-02, -6.7053e-03,\n -1.9947e-02, -1.3378e-02, -9.2290e-03, -3.2152e-02, 1.9759e-02,\n -2.1919e-02, 1.4672e-02, -1.8781e-02, -1.1288e-02, -1.9312e-03,\n 6.6657e-03, 3.7305e-02, -3.1557e-03, 8.2836e-03, 3.8670e-02,\n 3.5093e-02, 2.0330e-02, -1.0824e-02, -8.0056e-03, 4.4080e-02,\n 2.5198e-04, 1.7818e-02, -2.0541e-02, 1.5721e-02, -6.1238e-03,\n 4.0029e-02, 1.6332e-02, 9.2802e-03, 2.4458e-02, -1.4858e-03,\n 1.1421e-02, 1.0243e-02, -2.7684e-03, 5.5198e-03, 1.5162e-02,\n 9.7073e-03, -2.6397e-02, -6.5156e-03, 1.0377e-02, -7.1147e-03,\n 1.6549e-02, 7.0057e-03, 1.1966e-02, -1.0627e-02, -2.7904e-02,\n 3.6448e-03, 2.0647e-02, -4.6876e-02, 1.9154e-04, 1.4044e-02,\n 2.4143e-02, 1.4448e-03, 6.1524e-03, -2.2319e-02, 1.2886e-02,\n 3.0407e-02, -4.4209e-02, -9.0659e-03, -2.2936e-02, -4.2978e-02,\n 1.6259e-02, -5.0802e-03, 1.7799e-02, -2.1898e-02, -9.7146e-03,\n -2.7653e-02, 1.8320e-02, 1.3034e-02, 1.6945e-02, 1.7643e-02,\n -1.7133e-02, 1.1562e-02, -3.8009e-03, 1.0081e-02, -1.8771e-02,\n 8.5282e-03, -1.8645e-02, 3.8515e-02, -3.0562e-02, -2.2493e-03,\n 2.8429e-02, -1.6940e-03, -2.2865e-02, 2.9162e-02, -2.0802e-03,\n -1.0765e-02, 1.9013e-02, 9.0248e-03, 4.1727e-03, -2.1759e-02,\n 5.6422e-03, -1.6406e-02, -3.3454e-03, 1.2452e-02, 6.5810e-03,\n 2.2445e-02, -3.0383e-02, -9.0387e-03, 1.3146e-02, -2.4963e-04,\n 1.6126e-02, 5.8823e-03, 4.0198e-02, -1.3609e-02, 1.3401e-02,\n 2.3551e-02, 3.2280e-02, 1.0536e-02, 2.7952e-02, -9.0190e-03,\n -4.7719e-02, 4.6740e-03, 1.7746e-02, -2.5993e-03, -2.7065e-02,\n 2.1810e-02, 3.0216e-02, -5.9034e-03, -1.6624e-02, 2.6740e-03,\n -3.0003e-02, 7.3891e-03, -1.3231e-02, -9.6170e-03, 1.3551e-02,\n 5.8899e-03, 8.3775e-03, -2.5827e-02, 7.4239e-03, 1.5367e-02,\n -2.9455e-02, 1.8182e-02, -9.5729e-03, 1.5095e-02, -2.3909e-02,\n 5.7085e-03, -3.0599e-02, -3.0470e-03, -2.6016e-02, 2.5829e-02,\n -2.8883e-03, 5.0216e-03, -3.2515e-03, -2.2396e-03, 7.8913e-03,\n -2.5995e-03, 3.2806e-02, 1.5623e-02, 5.1704e-05, -4.3319e-03,\n 7.3423e-03, -3.7631e-02, -1.0225e-02, 1.4883e-02, 2.7596e-02,\n -2.6141e-02, 2.4416e-02, 4.7635e-03, -2.5872e-02, -3.0397e-03,\n 2.4736e-02, 2.9566e-02], device='cuda:0')",
13
+ "exp_avg_sq": "tensor([0.0038, 0.0045, 0.0068, 0.0056, 0.0032, 0.0028, 0.0061, 0.0049, 0.0045,\n 0.0039, 0.0055, 0.0039, 0.0035, 0.0052, 0.0056, 0.0049, 0.0057, 0.0033,\n 0.0032, 0.0039, 0.0061, 0.0037, 0.0033, 0.0036, 0.0057, 0.0046, 0.0038,\n 0.0039, 0.0053, 0.0064, 0.0051, 0.0063, 0.0049, 0.0051, 0.0041, 0.0101,\n 0.0042, 0.0034, 0.0033, 0.0061, 0.0057, 0.0043, 0.0044, 0.0070, 0.0022,\n 0.0060, 0.0020, 0.0044, 0.0036, 0.0026, 0.0040, 0.0030, 0.0051, 0.0049,\n 0.0042, 0.0035, 0.0029, 0.0031, 0.0034, 0.0044, 0.0024, 0.0033, 0.0052,\n 0.0068, 0.0042, 0.0074, 0.0038, 0.0030, 0.0080, 0.0070, 0.0038, 0.0093,\n 0.0040, 0.0043, 0.0036, 0.0049, 0.0039, 0.0018, 0.0030, 0.0068, 0.0034,\n 0.0047, 0.0041, 0.0030, 0.0007, 0.0044, 0.0056, 0.0063, 0.0033, 0.0047,\n 0.0061, 0.0038, 0.0067, 0.0035, 0.0059, 0.0051, 0.0042, 0.0047, 0.0038,\n 0.0052, 0.0041, 0.0047, 0.0055, 0.0051, 0.0026, 0.0058, 0.0025, 0.0054,\n 0.0055, 0.0042, 0.0044, 0.0048, 0.0038, 0.0038, 0.0026, 0.0034, 0.0061,\n 0.0024, 0.0020, 0.0054, 0.0037, 0.0043, 0.0019, 0.0032, 0.0053, 0.0055,\n 0.0076, 0.0042, 0.0050, 0.0044, 0.0050, 0.0019, 0.0027, 0.0040, 0.0027,\n 0.0057, 0.0063, 0.0018, 0.0085, 0.0044, 0.0026, 0.0056, 0.0038, 0.0037,\n 0.0018, 0.0069, 0.0100, 0.0082, 0.0047, 0.0063, 0.0047, 0.0046, 0.0044,\n 0.0041, 0.0053, 0.0061, 0.0030, 0.0038, 0.0021, 0.0038, 0.0057, 0.0062,\n 0.0027, 0.0050, 0.0072, 0.0090, 0.0036, 0.0032, 0.0047, 0.0068, 0.0027,\n 0.0035, 0.0039, 0.0031, 0.0058, 0.0052, 0.0058, 0.0039, 0.0070, 0.0045,\n 0.0070, 0.0037, 0.0085, 0.0050, 0.0057, 0.0062, 0.0047, 0.0039, 0.0035,\n 0.0049, 0.0044, 0.0035, 0.0039, 0.0039, 0.0055, 0.0069, 0.0064, 0.0061,\n 0.0029, 0.0052, 0.0054, 0.0039, 0.0053, 0.0061, 0.0037, 0.0047, 0.0031,\n 0.0054, 0.0031, 0.0094, 0.0041, 0.0042, 0.0062, 0.0053, 0.0062, 0.0040,\n 0.0028, 0.0037, 0.0046, 0.0066, 0.0040, 0.0064, 0.0044, 0.0030, 0.0058,\n 0.0045, 0.0042, 0.0069, 0.0050, 0.0038, 0.0024, 0.0046, 0.0058, 0.0075,\n 0.0051, 0.0032, 0.0033, 0.0072, 0.0047, 0.0062, 0.0025, 0.0044, 0.0048,\n 0.0059, 0.0082, 0.0086, 0.0017, 0.0048, 0.0081, 0.0026, 0.0048, 0.0035,\n 0.0033, 0.0008, 0.0020, 0.0048, 0.0044, 0.0043, 0.0051, 0.0036, 0.0030,\n 0.0041, 0.0051, 0.0050, 0.0039, 0.0050, 0.0056, 0.0044, 0.0025, 0.0041,\n 0.0071, 0.0038, 0.0041, 0.0064, 0.0056, 0.0043, 0.0038, 0.0065, 0.0043,\n 0.0055, 0.0036, 0.0072, 0.0060, 0.0046, 0.0039, 0.0025, 0.0067, 0.0035,\n 0.0040, 0.0060, 0.0048, 0.0046, 0.0026, 0.0060, 0.0045, 0.0099, 0.0048,\n 0.0047, 0.0047, 0.0053, 0.0051, 0.0039, 0.0063, 0.0060, 0.0039, 0.0041,\n 0.0046, 0.0051, 0.0049, 0.0059, 0.0063, 0.0058, 0.0033, 0.0065, 0.0057,\n 0.0036, 0.0039, 0.0044, 0.0042, 0.0026, 0.0048, 0.0022, 0.0041, 0.0057,\n 0.0040, 0.0076, 0.0048, 0.0055, 0.0043, 0.0054, 0.0037, 0.0027, 0.0022,\n 0.0058, 0.0049, 0.0035, 0.0056, 0.0049, 0.0054, 0.0045, 0.0052, 0.0035,\n 0.0053, 0.0039, 0.0027, 0.0050, 0.0049, 0.0046, 0.0032, 0.0050, 0.0028,\n 0.0040, 0.0044, 0.0053, 0.0043, 0.0080, 0.0051, 0.0051, 0.0033, 0.0048,\n 0.0048, 0.0064, 0.0050, 0.0030, 0.0067, 0.0060, 0.0052, 0.0037, 0.0032,\n 0.0067, 0.0057, 0.0062, 0.0027, 0.0038, 0.0042, 0.0042, 0.0040, 0.0059,\n 0.0044, 0.0056, 0.0062, 0.0041, 0.0064, 0.0047, 0.0035, 0.0046, 0.0055,\n 0.0034, 0.0061, 0.0056, 0.0055, 0.0047, 0.0030, 0.0075, 0.0059, 0.0056,\n 0.0050, 0.0058, 0.0046, 0.0042, 0.0022, 0.0035, 0.0034, 0.0071, 0.0026,\n 0.0068, 0.0072, 0.0049, 0.0046, 0.0031, 0.0038, 0.0030, 0.0019, 0.0060,\n 0.0063, 0.0052, 0.0032, 0.0049, 0.0031, 0.0102, 0.0036, 0.0066, 0.0098,\n 0.0024, 0.0048, 0.0048, 0.0037, 0.0063, 0.0040, 0.0037, 0.0039, 0.0051,\n 0.0058, 0.0039, 0.0034, 0.0044, 0.0037, 0.0044, 0.0027, 0.0040, 0.0025,\n 0.0076, 0.0062, 0.0045, 0.0025, 0.0051, 0.0046, 0.0044, 0.0077, 0.0069,\n 0.0036, 0.0042, 0.0045, 0.0050, 0.0050, 0.0041, 0.0046, 0.0019, 0.0023,\n 0.0094, 0.0046, 0.0036, 0.0077, 0.0052, 0.0052, 0.0062, 0.0064, 0.0045,\n 0.0037, 0.0050, 0.0072, 0.0039, 0.0051, 0.0032, 0.0056, 0.0028, 0.0030,\n 0.0067, 0.0038, 0.0051, 0.0079, 0.0039, 0.0050, 0.0022, 0.0036, 0.0032,\n 0.0045, 0.0064, 0.0042, 0.0059, 0.0035, 0.0040, 0.0057, 0.0031, 0.0097,\n 0.0059, 0.0039, 0.0063, 0.0033, 0.0066, 0.0040, 0.0060, 0.0035, 0.0027,\n 0.0050, 0.0032, 0.0073, 0.0048, 0.0111, 0.0050, 0.0039, 0.0030],\n device='cuda:0')"
14
  },
15
  "2": {
16
+ "step": "tensor(3756.)",
17
+ "exp_avg": "tensor([ 2.2412e-03, -1.7135e-03, 1.7414e-04, -4.3304e-03, -1.4381e-03,\n 2.2773e-03, 3.8135e-04, 2.8400e-03, 3.3537e-03, -4.8464e-03,\n 2.8367e-03, 2.6625e-03, 1.1428e-03, 2.5408e-03, -9.4593e-04,\n 2.4687e-03, 4.9804e-03, 1.2671e-03, -4.9175e-03, -6.1412e-03,\n -9.8762e-04, 1.3469e-04, -5.9047e-03, 3.7092e-03, -6.1670e-05,\n -1.8071e-03, 4.0642e-03, -5.5387e-03, 5.7623e-04, 6.4057e-04,\n -5.4331e-03, 6.6069e-04, -3.1341e-03, -2.0199e-03, -5.1051e-03,\n -8.0458e-04, 1.6603e-03, 2.2239e-03, 5.0428e-04, 1.2053e-04,\n 3.7555e-03, -5.0957e-03, -9.1718e-04, -2.9196e-05, 2.1659e-03,\n -1.1933e-03, 4.8316e-03, -4.0742e-03, 2.6366e-03, 1.4382e-03,\n 5.2703e-03, 3.5278e-03, 5.3773e-03, 1.3938e-04, 1.2351e-03,\n -1.6874e-03, -4.7025e-04, -2.8393e-03, 2.6994e-03, -1.4851e-03,\n -6.4587e-03, -1.8559e-03, -1.8142e-03, -2.2883e-03, 3.9893e-03,\n 2.9580e-03, 5.2687e-03, 1.4087e-03, -1.8676e-03, 8.5207e-03,\n 3.2102e-03, -1.8096e-03, 4.5923e-03, 2.4317e-04, -6.2930e-03,\n 1.2876e-03, -2.4752e-03, 7.2579e-05, -3.0093e-03, 3.4962e-03,\n 4.7789e-04, -3.6991e-03, -2.8246e-03, 3.0508e-03, 4.4410e-02,\n 1.4753e-03, 8.2979e-04, 1.1141e-03, -3.3824e-03, -5.7142e-04,\n 2.8538e-03, -3.9017e-04, -4.3236e-04, -1.0912e-03, -1.8318e-03,\n 2.5278e-03, 8.3838e-04, 3.5042e-04, 2.4786e-03, 5.0220e-03,\n 2.3841e-04, -1.1023e-02, 1.0064e-02, -6.0304e-03, -1.4359e-02,\n -1.4968e-03, -5.9049e-05, -1.1945e-03, 1.5137e-05, 1.3881e-03,\n -1.8327e-03, 2.9801e-03, 2.0266e-03, -6.8345e-04, -1.1472e-03,\n -2.8268e-03, 2.7150e-04, 3.0047e-03, -7.8555e-04, -1.3243e-03,\n 5.7360e-03, 1.6748e-03, -1.0727e-02, 4.0181e-03, 1.4181e-03,\n 1.1378e-03, -9.9631e-06, -4.0464e-03, 4.0542e-04, -2.1097e-03,\n 2.8582e-03, 6.8607e-03, -6.2325e-03, -2.7747e-03, 2.4745e-03,\n 2.4338e-03, 1.6621e-03, -8.0157e-03, -1.8352e-03, 8.7467e-04,\n -2.7310e-03, 1.8914e-03, 7.8300e-04, 5.9486e-03, 6.3792e-03,\n -6.2411e-03, -4.0980e-04, 2.5085e-04, 1.5017e-03, -1.6664e-03,\n 2.7440e-03, 1.8272e-03, -6.3675e-04, -2.0371e-03, 2.2415e-03,\n 1.6142e-03, 2.1835e-03, 2.1963e-03, -3.1199e-03, -2.6184e-03,\n 1.4482e-03, 8.9175e-04, 4.4728e-03, 2.2908e-03, -3.0687e-03,\n 7.3962e-03, 1.8102e-03, 1.5618e-03, -1.3161e-03, -6.0806e-04,\n -3.8076e-03, -6.6185e-03, 5.8626e-03, 2.7417e-03, -6.3661e-04,\n -1.9567e-03, -1.8094e-03, -2.3712e-03, 2.5291e-03, 4.6080e-03,\n -1.7453e-03, 1.0957e-04, 2.0464e-03, 3.3967e-04, -4.7387e-03,\n -1.1669e-03, -3.5051e-03, -3.7303e-04, -1.8864e-03, -4.3955e-03,\n 4.0612e-03, -6.8800e-04, 4.5782e-03, 4.5491e-04, 5.7239e-04,\n -1.1122e-03, 6.6922e-04, 3.7448e-03, -7.1595e-03, -1.5208e-03,\n -3.1904e-03, 1.8243e-03, 1.1170e-03, -3.4170e-04, 3.2902e-03,\n 2.8336e-03, 4.4310e-03, -9.5381e-04, -7.4804e-03, 1.3518e-03,\n -2.6924e-03, -4.2903e-03, -2.8415e-03, -7.2002e-03, -4.9347e-03,\n -3.6700e-03, 7.9178e-03, 1.3275e-03, -2.6430e-03, -2.6274e-03,\n 2.8503e-03, -7.2066e-04, 1.8085e-03, -4.9095e-03, 3.2414e-03,\n -3.3482e-03, 3.9858e-03, 6.6900e-03, -1.6051e-03, -1.3105e-03,\n -9.1046e-03, 4.5852e-03, -1.9213e-04, 1.2340e-03, 4.1343e-04,\n -1.7289e-04, -7.2067e-04, -5.3894e-03, 1.6155e-03, 4.1339e-05,\n 4.0963e-03, 3.9025e-04, -2.7697e-03, 2.9600e-04, 1.1189e-03,\n -1.6872e-03, -1.0500e-03, -1.8361e-03, 1.1936e-03, -5.5934e-03,\n -1.2943e-03, 2.8115e-03, 4.9239e-03, 4.0005e-02, -4.4574e-03,\n -1.8593e-03, 1.4525e-02, 3.2345e-03, 2.6853e-03, -2.5231e-03,\n -8.5564e-04, 4.1857e-03, 1.5528e-03, -1.7314e-03, 2.9448e-03,\n 1.8864e-03, 1.8238e-03, -3.8789e-04, -1.0849e-04, -5.4990e-04,\n -2.3847e-03, -3.2831e-03, 1.0048e-03, -2.4168e-03, -3.3254e-03,\n -1.9478e-03, -4.6220e-04, -9.0199e-04, -2.2809e-04, 3.1444e-03,\n -2.9999e-03, -3.6490e-03, 1.9346e-03, -4.2916e-03, 2.4314e-04,\n -2.4831e-03, -2.3518e-04, 7.1814e-03, 2.1665e-03, 8.3098e-04,\n -2.5579e-04, 5.1445e-03, -7.7625e-03, 5.9586e-04, 3.2712e-03,\n 4.0598e-03, 2.1439e-03, 3.6067e-03, 2.2015e-03, -3.1624e-03,\n -1.0527e-03, 2.3959e-03, -4.8800e-03, -7.8469e-04, 7.2287e-04,\n -1.5966e-03, 1.3113e-03, 4.2742e-04, 3.8884e-03, 5.2452e-04,\n -4.3255e-03, 1.0893e-03, -3.8631e-03, -8.3471e-03, 1.3907e-03,\n 1.2566e-03, 6.1187e-03, -3.3134e-03, -7.3867e-04, -5.3702e-03,\n -3.0016e-03, -3.2627e-03, 5.3090e-04, -2.6888e-03, -5.4431e-03,\n -7.9559e-03, 1.1285e-02, 3.0928e-03, -2.1883e-03, -5.2636e-04,\n 2.9540e-03, 3.9284e-03, -2.1602e-03, -5.1570e-03, 2.8134e-04,\n -2.9672e-03, -1.4579e-03, 3.6893e-03, -3.9067e-03, 2.0354e-03,\n -7.4523e-04, 8.1314e-04, 6.4396e-04, -4.3433e-03, -1.9355e-03,\n -3.3759e-03, 5.3959e-03, -2.8397e-03, -4.7204e-03, -2.2167e-03,\n -4.1177e-03, -1.9518e-03, -1.7475e-03, -1.7199e-03, 3.6560e-03,\n -2.8245e-03, 2.5957e-03, -2.7418e-03, -2.8444e-03, -3.4857e-04,\n 7.4086e-04, 3.8491e-03, -9.4503e-05, 1.8652e-03, 5.9095e-03,\n 6.6348e-03, 2.8447e-03, -2.4476e-03, -1.1085e-03, 5.7215e-03,\n 3.2211e-04, 2.5288e-03, -4.4411e-03, 1.6789e-03, 5.1900e-04,\n 6.9788e-03, 2.2084e-03, 3.0094e-04, 4.0592e-03, -3.6869e-04,\n -9.8965e-06, 2.4043e-03, -1.0744e-03, 2.9592e-04, 3.1814e-03,\n 2.3348e-03, -3.6615e-03, -3.5001e-04, 2.5392e-03, -6.5642e-04,\n 1.7816e-03, 1.1544e-03, 2.8410e-03, -1.0436e-03, -1.4853e-03,\n 8.4537e-04, 1.5618e-03, -9.2119e-03, 1.5251e-03, 3.9330e-03,\n 6.8915e-03, 1.1578e-03, 1.7283e-03, -3.8092e-03, 4.7025e-03,\n 3.4279e-03, -4.1997e-03, -1.9524e-03, -2.7501e-03, -7.7438e-03,\n 3.2323e-03, -3.0048e-03, 1.0423e-02, -3.0120e-03, -7.8937e-04,\n -2.3586e-03, 3.1767e-03, 2.3955e-03, 4.0481e-03, 2.2316e-03,\n -3.4405e-03, 1.7126e-03, -7.3341e-04, 5.6000e-03, -2.3987e-03,\n 6.5948e-04, -4.0300e-03, 3.8832e-03, -4.8191e-03, -7.2676e-04,\n 6.2522e-03, -1.2198e-03, -3.8399e-03, 4.6496e-03, 4.6605e-04,\n -1.4959e-03, 4.2936e-03, 1.3917e-03, 3.2749e-03, -2.5358e-03,\n 6.5342e-04, -2.0811e-04, -3.5312e-04, 2.6695e-03, 3.2043e-03,\n 5.6174e-04, -3.0822e-03, -6.4715e-04, 1.6246e-04, 1.8698e-04,\n 1.7960e-03, 1.4190e-03, 5.1615e-03, -1.5778e-03, 1.9907e-03,\n 4.2274e-03, 4.2330e-03, 4.7313e-03, 7.1710e-03, -5.2973e-04,\n -6.8373e-03, 2.2191e-03, 1.7422e-03, 1.2161e-04, -4.1333e-03,\n 1.2762e-03, 2.2669e-03, -1.5078e-03, -2.0603e-03, -4.3502e-05,\n -4.1420e-03, 3.6279e-04, -1.8044e-04, -2.6054e-03, 1.1233e-03,\n -3.5794e-04, 2.0161e-03, -2.0684e-03, 1.1285e-03, 2.2444e-04,\n -3.2994e-03, 2.9648e-03, -1.7433e-03, 6.8459e-03, -4.7469e-03,\n 1.6106e-03, -4.9114e-03, -6.3474e-04, -3.3919e-03, 2.5912e-03,\n -2.0158e-03, 5.3829e-05, -1.0945e-03, -5.7670e-04, 7.7105e-04,\n -1.8355e-03, 5.5802e-03, 1.6919e-03, 2.5590e-03, -8.2609e-04,\n 9.4579e-04, -4.2359e-03, -3.3171e-04, 5.0010e-03, 2.9104e-03,\n -3.8487e-03, 2.2766e-03, 8.9527e-04, -2.2935e-03, -6.5193e-04,\n 7.2461e-03, 6.0258e-03], device='cuda:0')",
18
+ "exp_avg_sq": "tensor([5.5446e-05, 6.4132e-05, 5.6399e-05, 4.6511e-05, 1.2707e-04, 1.0791e-04,\n 4.5738e-05, 4.7122e-05, 5.5230e-05, 7.4603e-05, 7.1506e-05, 1.9255e-04,\n 4.9323e-05, 7.9633e-05, 8.4222e-05, 6.4092e-05, 3.9996e-05, 1.7049e-04,\n 1.0788e-04, 8.4483e-05, 5.2609e-05, 1.0349e-04, 1.2437e-04, 1.0921e-04,\n 4.2751e-05, 6.6668e-05, 1.3386e-04, 6.0573e-05, 4.0787e-05, 6.9743e-05,\n 5.8580e-05, 3.4733e-05, 8.9943e-05, 3.6119e-05, 6.7910e-05, 8.0744e-05,\n 6.9886e-05, 1.0271e-04, 1.1208e-04, 7.9783e-05, 8.9811e-05, 8.3915e-05,\n 8.1075e-05, 6.9102e-05, 8.8261e-05, 8.0852e-05, 2.3186e-04, 7.7987e-05,\n 9.9830e-05, 2.2391e-04, 9.2092e-05, 7.8770e-05, 6.6982e-05, 5.3671e-05,\n 7.6479e-05, 7.1291e-05, 8.0652e-05, 1.3843e-04, 6.2977e-05, 9.8219e-05,\n 1.4893e-04, 9.9706e-05, 8.0226e-05, 5.5441e-05, 1.0266e-04, 4.6826e-05,\n 9.3978e-05, 1.1177e-04, 5.5899e-05, 8.3006e-05, 5.8068e-05, 7.6468e-05,\n 5.7955e-05, 5.3015e-05, 8.1328e-05, 8.9731e-05, 4.6919e-05, 2.5374e-04,\n 1.4616e-04, 4.9310e-05, 1.4999e-04, 6.7440e-05, 6.5267e-05, 7.0864e-05,\n 5.4232e-02, 4.6815e-05, 8.4769e-05, 3.9349e-05, 8.1684e-05, 6.8021e-05,\n 9.7640e-05, 7.9762e-05, 4.0023e-05, 1.4191e-04, 5.0675e-05, 6.8357e-05,\n 6.5486e-05, 6.0154e-05, 7.8052e-05, 8.9092e-05, 7.1649e-05, 5.9699e-05,\n 1.4064e-04, 5.9679e-05, 2.7588e-04, 5.0106e-05, 1.3404e-04, 7.6390e-05,\n 6.3836e-05, 7.5547e-05, 9.1545e-05, 6.3531e-05, 6.2582e-05, 6.8280e-05,\n 1.0431e-04, 8.3868e-05, 1.0332e-04, 2.0666e-04, 1.8027e-04, 7.8560e-05,\n 7.1190e-05, 3.8187e-05, 3.4004e-04, 7.3410e-05, 3.9404e-05, 4.9781e-05,\n 5.2473e-05, 6.8688e-05, 3.1789e-05, 4.9773e-05, 6.7704e-05, 4.5765e-04,\n 1.1654e-04, 5.4563e-05, 6.4924e-05, 4.4541e-05, 5.3853e-05, 2.5316e-04,\n 5.7468e-05, 1.0858e-04, 1.2075e-04, 4.2033e-05, 7.5276e-05, 1.4631e-04,\n 1.9843e-04, 6.7765e-05, 8.6803e-05, 7.8811e-05, 4.2527e-05, 4.0923e-05,\n 8.2128e-05, 6.6967e-05, 7.5650e-05, 6.4201e-05, 4.4222e-05, 6.3211e-05,\n 2.0454e-04, 1.0282e-04, 1.3514e-04, 7.4200e-05, 6.1881e-05, 8.2469e-05,\n 1.6406e-04, 6.7362e-05, 4.7482e-05, 8.8163e-05, 1.2432e-04, 1.5225e-04,\n 6.3945e-05, 6.6770e-05, 1.5806e-04, 7.6109e-05, 8.4151e-05, 1.3881e-04,\n 6.7831e-05, 6.0170e-05, 5.4812e-05, 5.1137e-05, 4.7145e-05, 6.0506e-05,\n 6.7781e-05, 8.3530e-05, 5.2330e-05, 4.3812e-05, 4.9201e-05, 6.3878e-05,\n 7.0716e-05, 7.4726e-05, 6.8775e-05, 8.4952e-05, 1.4471e-04, 1.2892e-04,\n 5.5337e-05, 7.7961e-05, 3.4819e-05, 5.3437e-05, 3.9895e-05, 5.9554e-05,\n 1.7291e-04, 8.9519e-05, 3.1590e-05, 8.1119e-05, 6.2007e-05, 4.5370e-05,\n 6.0922e-05, 4.5289e-05, 1.3641e-04, 6.3188e-05, 1.3748e-04, 1.4415e-04,\n 8.9278e-05, 1.1314e-04, 6.0811e-05, 4.6358e-05, 8.1932e-05, 1.0966e-04,\n 1.7981e-04, 4.9724e-05, 8.1919e-05, 6.2570e-05, 9.6432e-05, 6.4979e-05,\n 6.9389e-05, 1.2214e-04, 5.5688e-05, 8.6697e-05, 9.6204e-05, 4.9833e-05,\n 8.2200e-05, 7.4171e-05, 1.5341e-04, 7.5434e-05, 5.9514e-05, 6.3185e-05,\n 3.8851e-05, 1.2778e-04, 8.7470e-05, 6.0584e-05, 4.6198e-05, 4.0764e-05,\n 1.5707e-04, 9.7872e-05, 9.0519e-05, 4.2534e-05, 7.2300e-05, 6.7137e-05,\n 3.0606e-04, 5.9555e-05, 6.7416e-05, 1.4839e-04, 4.9552e-05, 7.7599e-05,\n 1.8491e-04, 3.9916e-03, 2.2390e-04, 7.8529e-05, 1.0600e-04, 9.4837e-05,\n 4.7665e-05, 9.4210e-05, 7.0152e-05, 9.8161e-05, 5.1627e-05, 5.6590e-05,\n 1.2276e-04, 5.5648e-05, 8.8651e-05, 4.2604e-05, 2.0395e-04, 4.7976e-05,\n 5.5407e-05, 1.1679e-04, 5.3577e-05, 6.1860e-05, 8.1109e-05, 4.0159e-05,\n 1.4890e-04, 4.5064e-05, 3.7903e-05, 6.1216e-05, 6.3164e-05, 3.7977e-05,\n 4.1421e-05, 1.0231e-04, 6.6525e-05, 1.6127e-04, 5.7824e-05, 8.0690e-05,\n 5.8176e-05, 9.7729e-05, 5.3551e-05, 7.5359e-05, 1.4710e-04, 9.1842e-05,\n 6.7127e-05, 6.4002e-05, 7.5696e-05, 5.1891e-05, 6.3699e-05, 4.4089e-05,\n 4.6159e-05, 5.9607e-05, 6.3256e-05, 6.4111e-05, 5.7587e-05, 4.2578e-05,\n 5.1245e-05, 8.9717e-05, 5.3103e-05, 6.7883e-05, 6.8497e-05, 5.0562e-05,\n 8.1577e-05, 5.4415e-05, 5.4882e-05, 1.2190e-04, 1.2503e-04, 8.4783e-05,\n 3.9527e-05, 1.1437e-04, 7.6957e-05, 1.7722e-04, 7.8181e-05, 8.2478e-05,\n 1.1803e-04, 7.0484e-05, 9.3804e-05, 5.1647e-05, 1.2228e-04, 1.0473e-04,\n 1.4041e-04, 1.1026e-04, 2.5453e-04, 5.5197e-05, 8.8876e-05, 1.0998e-04,\n 7.6670e-05, 1.0826e-04, 8.0951e-05, 9.3408e-05, 5.2477e-05, 1.4197e-04,\n 4.1826e-05, 6.8913e-05, 1.3685e-04, 5.7397e-05, 8.1774e-05, 8.0215e-05,\n 1.1461e-04, 7.3814e-05, 1.0045e-04, 8.1093e-05, 1.0607e-04, 4.9563e-05,\n 7.0981e-05, 7.2750e-05, 8.8279e-05, 5.5486e-05, 6.4468e-05, 4.2244e-05,\n 5.9107e-05, 5.8662e-05, 4.8839e-05, 1.3711e-04, 6.7313e-05, 1.2214e-04,\n 4.9914e-05, 2.1295e-04, 1.1759e-04, 5.1836e-05, 8.6552e-05, 1.1239e-04,\n 1.0987e-04, 8.0413e-05, 8.0804e-05, 8.6793e-05, 4.6665e-05, 4.1324e-05,\n 7.7176e-05, 6.7241e-05, 4.9253e-05, 1.1633e-04, 5.5365e-05, 1.1659e-04,\n 1.0838e-04, 1.4481e-04, 5.1352e-05, 8.3429e-05, 7.5449e-05, 1.0290e-04,\n 5.2795e-05, 9.1061e-05, 1.8421e-04, 1.0431e-04, 4.0243e-05, 1.0567e-04,\n 4.7256e-05, 1.3438e-04, 4.4296e-05, 1.3330e-04, 1.3447e-04, 6.7811e-05,\n 1.6271e-04, 7.4289e-05, 1.7912e-04, 5.9456e-05, 6.9112e-05, 4.9076e-05,\n 6.5650e-05, 1.1166e-04, 8.9616e-05, 1.2644e-04, 3.9920e-04, 5.5705e-05,\n 5.4856e-05, 4.8031e-05, 6.1140e-05, 7.1174e-05, 1.4039e-04, 5.1650e-05,\n 6.2376e-05, 6.3482e-05, 4.7972e-05, 2.2953e-04, 6.8583e-05, 7.2732e-05,\n 1.3039e-04, 5.5429e-05, 1.5717e-04, 9.4277e-05, 1.0426e-04, 5.8689e-05,\n 8.0745e-05, 6.8853e-05, 6.9083e-05, 6.0683e-05, 1.1935e-04, 3.2928e-05,\n 1.5923e-04, 5.2598e-05, 1.2275e-04, 4.7202e-05, 2.9086e-05, 4.0901e-05,\n 2.3439e-04, 7.1133e-05, 5.4041e-05, 3.0520e-05, 8.0903e-05, 5.2271e-05,\n 9.2661e-05, 8.0243e-05, 4.7343e-05, 8.2351e-05, 4.9579e-05, 1.0376e-04,\n 6.6373e-05, 4.5664e-04, 2.1615e-04, 5.9465e-05, 5.1457e-05, 5.7198e-05,\n 5.3993e-05, 5.7062e-05, 7.1492e-05, 3.0075e-05, 5.1284e-05, 1.3091e-04,\n 7.6941e-05, 4.7884e-05, 8.3514e-05, 4.7215e-05, 4.1200e-05, 1.3425e-04,\n 4.8361e-05, 8.5841e-05, 1.1048e-04, 5.6314e-05, 1.0450e-04, 4.7403e-05,\n 6.2005e-05, 7.4012e-05, 6.7158e-05, 3.0993e-04, 1.5304e-04, 1.1113e-04,\n 7.3731e-05, 7.4469e-05, 4.1666e-05, 5.0553e-05, 1.5470e-04, 1.1745e-04,\n 7.0925e-05, 1.0838e-04, 8.5005e-05, 6.5341e-05, 1.2265e-04, 4.4649e-05,\n 1.2706e-04, 6.7683e-05, 9.7837e-05, 5.2595e-05, 6.5802e-05, 1.2117e-04,\n 8.1331e-05, 1.4946e-04, 7.2662e-05, 6.9193e-05, 7.2229e-05, 7.7744e-05,\n 2.1255e-04, 1.3064e-04], device='cuda:0')"
19
  },
20
  "3": {
21
+ "step": "tensor(3756.)",
22
+ "exp_avg": "tensor([ 3.0635e-03, -1.9861e-03, -9.4052e-04, -6.8551e-03, -1.3992e-03,\n 1.7724e-03, -4.0755e-04, 4.1266e-03, 3.3579e-03, -4.1456e-03,\n 1.4877e-03, 2.3882e-03, 1.3644e-03, 2.8746e-03, -1.5333e-03,\n 2.0384e-03, 6.7964e-03, 1.9253e-03, -4.2738e-03, -4.1146e-03,\n -1.4686e-03, 4.1665e-04, -4.4689e-03, 2.6356e-03, 9.1327e-04,\n -1.2995e-03, 3.1245e-03, -5.4362e-03, 1.6028e-03, -1.9910e-03,\n -4.2968e-03, -5.9344e-04, -3.2506e-03, -2.1614e-03, -3.3544e-03,\n -8.8182e-04, 2.6627e-03, 1.0330e-03, 7.2511e-04, 2.4259e-04,\n 2.9966e-03, -5.4110e-03, -4.8079e-04, -9.0030e-04, 1.9219e-03,\n -1.3773e-03, 3.3264e-03, -3.5490e-03, 2.5106e-03, 9.2618e-04,\n 4.6565e-03, 2.9999e-03, 7.1533e-03, -7.0061e-04, 1.4664e-03,\n -2.7204e-03, 6.1207e-05, -2.0212e-03, 2.3404e-03, -1.3691e-03,\n -4.1453e-03, -6.2792e-04, -3.5221e-03, -3.4472e-03, 4.9964e-03,\n 3.4823e-03, 6.9411e-03, 2.0060e-03, -1.6492e-03, 8.8252e-03,\n 1.9255e-03, -3.8099e-03, 4.0052e-03, 1.0838e-03, -6.7481e-03,\n 1.9491e-03, -3.3329e-03, -4.7661e-05, -3.6133e-03, 3.4728e-03,\n -1.4024e-03, -3.2754e-03, -2.5890e-03, 3.3532e-03, 2.7508e-03,\n 1.2270e-03, 1.1252e-03, 2.2928e-03, -2.5464e-03, -1.1223e-03,\n 4.3943e-03, -4.5321e-04, -8.0929e-04, -1.5300e-03, -1.9327e-03,\n 1.9752e-03, 7.7886e-04, -5.7136e-04, 2.1995e-03, 6.5292e-03,\n 5.8801e-04, -1.1201e-02, 9.8411e-03, -6.0881e-03, -9.0481e-03,\n -1.1076e-03, -3.5725e-04, -1.2649e-03, -1.5373e-03, 1.4969e-03,\n -1.7543e-04, 3.8658e-03, 1.8722e-03, -1.2217e-03, -6.9723e-04,\n -2.7562e-03, 5.7324e-04, 2.0006e-03, -4.8141e-04, -1.1413e-03,\n 6.7301e-03, 2.7898e-03, -5.3258e-03, 4.0333e-03, 2.0983e-03,\n 1.3567e-03, -5.1343e-04, -4.8040e-03, 5.9798e-04, -2.3855e-03,\n 2.5707e-03, 3.2367e-03, -4.8481e-03, -2.3262e-03, 2.2225e-03,\n 2.7672e-03, 2.8142e-03, -3.8686e-03, -1.5575e-03, 3.1642e-04,\n -1.8861e-03, 1.8973e-03, 1.3121e-03, 5.2085e-03, 4.3428e-03,\n -6.8374e-03, 5.5278e-04, 1.0958e-04, 1.2738e-03, -1.0990e-03,\n 3.9662e-03, 2.0328e-03, -7.3299e-04, -3.5983e-04, 1.7263e-03,\n 2.2384e-03, 1.6913e-03, 1.5454e-03, -2.0404e-03, -1.2524e-03,\n 6.4269e-04, 1.6968e-03, 2.7551e-03, 1.8094e-03, -2.6116e-03,\n 7.2182e-03, 2.0176e-03, 1.0283e-03, -1.1125e-03, -2.4973e-03,\n -4.0231e-03, -6.7085e-03, 5.1633e-03, 1.6480e-03, 4.9078e-04,\n -1.2503e-03, -2.5853e-03, -2.0476e-03, 4.3006e-03, 5.1631e-03,\n -2.7732e-03, -6.9292e-04, 2.6587e-03, -1.3835e-03, -7.1646e-03,\n -1.3706e-03, -3.3198e-03, 1.1076e-03, -9.1956e-04, -4.5856e-03,\n 2.6703e-03, 9.0647e-05, 3.2604e-03, 1.3473e-03, 9.6050e-04,\n -1.9882e-03, 1.0131e-03, 5.6720e-03, -4.7512e-03, -1.6065e-03,\n -4.5577e-03, 2.7334e-03, 1.4091e-03, 8.7326e-04, 4.4031e-03,\n 5.0517e-03, 3.5589e-03, -4.4089e-04, -4.6530e-03, 1.8879e-04,\n -4.0591e-03, -5.1315e-03, -2.4968e-03, -9.1457e-03, -5.2014e-03,\n -2.9639e-03, 4.6425e-03, 2.3862e-03, -1.3374e-03, -3.0314e-03,\n 1.9171e-03, -1.3634e-03, 1.2566e-03, -4.0148e-03, 4.1147e-03,\n -4.0598e-03, 4.4757e-03, 5.1633e-03, -3.2027e-03, -1.0377e-03,\n -5.5657e-03, 4.6007e-03, 7.9854e-04, 7.3536e-04, 4.5224e-04,\n 4.3658e-04, 5.7845e-04, -5.3852e-03, 2.1520e-03, 1.2606e-03,\n 2.2583e-03, 1.0296e-03, -3.7393e-03, 7.3201e-04, 2.2330e-03,\n -1.8458e-03, -5.6234e-05, -2.0583e-03, 1.7502e-03, -4.0080e-03,\n -1.8866e-03, 2.5375e-03, 3.1242e-03, 8.1933e-03, -3.4721e-03,\n -4.4185e-04, 1.3221e-02, 2.3641e-03, 3.2023e-03, -2.1942e-03,\n -9.5125e-05, 4.4712e-03, 1.8799e-03, -1.8269e-03, 2.6930e-03,\n 1.4045e-03, 6.4966e-04, -1.9578e-03, -1.0029e-03, -8.8652e-04,\n -4.3043e-03, -2.8491e-03, 1.4734e-03, -1.5237e-03, -2.3068e-03,\n -2.2489e-03, 1.0189e-05, -1.7836e-03, 1.8488e-03, 3.8255e-03,\n -3.0502e-03, -5.3631e-03, 2.7422e-03, -3.9854e-03, 4.9860e-04,\n -1.0035e-03, -7.0601e-04, 6.0719e-03, 3.9114e-03, 2.3470e-03,\n -1.6819e-03, 6.1819e-03, -6.0942e-03, 2.0775e-03, 1.5665e-03,\n 4.0838e-03, 1.5915e-03, 5.0645e-03, 2.3797e-03, -4.1394e-03,\n -1.4515e-03, 2.0652e-03, -5.9057e-03, -7.2555e-04, 1.7022e-03,\n -2.9052e-03, 9.6378e-04, 9.8166e-05, 4.6818e-03, 1.0888e-04,\n -6.7935e-03, 9.9089e-04, -2.4687e-03, -1.1505e-02, 1.2198e-03,\n 1.7136e-03, 4.7233e-03, -3.4642e-03, -5.3167e-04, -4.2299e-03,\n -2.4076e-03, -2.3350e-03, 3.3399e-04, -2.7955e-03, -5.4495e-03,\n -1.0014e-02, 1.0200e-02, 2.8222e-03, -1.7014e-03, 1.3081e-04,\n 1.2816e-03, 3.8313e-03, -1.3833e-03, -6.2671e-03, -1.0140e-04,\n -7.1668e-04, -1.4104e-03, 4.6292e-03, -2.1599e-03, 2.5092e-03,\n 6.8392e-05, 1.0930e-03, 5.3854e-05, -4.1666e-03, -1.6033e-03,\n -3.8560e-03, 4.5929e-03, -3.0020e-03, -3.5071e-03, -1.4027e-03,\n -3.0637e-03, -2.2965e-03, -1.5402e-03, -3.7924e-03, 3.2121e-03,\n -3.8421e-03, 2.3435e-03, -2.2251e-03, -1.9467e-03, -2.8707e-05,\n 8.2852e-04, 5.9882e-03, -1.8725e-04, 1.4480e-03, 5.6261e-03,\n 6.1335e-03, 2.5131e-03, -2.1429e-03, -6.0423e-04, 5.2684e-03,\n 6.1458e-05, 3.2672e-03, -3.8139e-03, 2.7077e-03, -3.5169e-04,\n 6.4010e-03, 2.7040e-03, 9.6470e-04, 4.5959e-03, -5.5243e-05,\n 9.6883e-04, 1.9827e-03, -3.4303e-04, 9.6726e-04, 2.4136e-03,\n 1.8019e-03, -4.1580e-03, -1.1355e-03, 1.4680e-03, -5.1983e-04,\n 2.5715e-03, 9.3408e-04, 1.7994e-03, -1.2365e-03, -3.6763e-03,\n 5.7712e-04, 2.8488e-03, -7.3797e-03, -1.7441e-04, 3.3275e-03,\n 4.9972e-03, 4.5258e-04, 7.9507e-04, -3.3248e-03, 3.1472e-03,\n 4.5782e-03, -5.9514e-03, -1.8133e-03, -2.8582e-03, -7.2692e-03,\n 3.1107e-03, -9.4695e-04, 5.0172e-03, -2.9549e-03, -1.5187e-03,\n -3.1892e-03, 2.6704e-03, 2.1104e-03, 3.6205e-03, 2.5166e-03,\n -3.2336e-03, 2.1713e-03, -4.2148e-04, 3.4130e-03, -3.1914e-03,\n 9.1727e-04, -3.5759e-03, 5.1915e-03, -4.9941e-03, -1.3840e-04,\n 5.8908e-03, -5.6121e-04, -3.5299e-03, 4.3961e-03, -3.1667e-04,\n -1.4551e-03, 3.6778e-03, 1.1681e-03, 1.6906e-03, -3.5076e-03,\n 9.3956e-04, -1.2590e-03, -6.2212e-04, 1.4641e-03, 1.8272e-03,\n 2.5139e-03, -4.0894e-03, -1.4365e-03, 1.3786e-03, 2.7928e-04,\n 3.1579e-03, 1.0241e-03, 6.1064e-03, -3.3676e-03, 2.0733e-03,\n 4.5430e-03, 4.6333e-03, 3.4528e-03, 5.6793e-03, -7.5373e-04,\n -6.6768e-03, 8.7553e-04, 2.2048e-03, -4.5108e-04, -4.1387e-03,\n 2.2810e-03, 4.0264e-03, -1.2661e-03, -2.3642e-03, 5.1904e-04,\n -4.7711e-03, 8.0742e-04, -8.3147e-04, -2.0117e-03, 1.7358e-03,\n 1.0230e-03, 2.1090e-03, -3.4938e-03, 1.6084e-03, 2.0406e-03,\n -3.3305e-03, 2.7882e-03, -1.5883e-03, 3.8706e-03, -4.1698e-03,\n 1.4631e-03, -5.2684e-03, 3.4250e-05, -3.4205e-03, 3.3450e-03,\n -3.1402e-04, 4.0645e-04, -1.5899e-04, 8.1472e-05, 1.0232e-03,\n -3.6184e-04, 6.1906e-03, 2.2909e-03, 5.3375e-04, -4.3507e-04,\n 1.2563e-03, -5.8063e-03, -1.8233e-03, 3.9287e-03, 4.0902e-03,\n -4.5025e-03, 3.6272e-03, 6.3574e-04, -3.5152e-03, -8.6676e-04,\n 4.1573e-03, 5.7829e-03], device='cuda:0')",
23
+ "exp_avg_sq": "tensor([6.0834e-05, 6.5793e-05, 9.2561e-05, 6.6879e-05, 6.8785e-05, 7.4239e-05,\n 7.6099e-05, 5.7881e-05, 6.7822e-05, 7.3908e-05, 8.7127e-05, 1.2235e-04,\n 4.8012e-05, 9.4348e-05, 8.8311e-05, 7.6162e-05, 6.4876e-05, 8.9561e-05,\n 8.8322e-05, 7.6611e-05, 8.1757e-05, 8.2324e-05, 8.7883e-05, 7.3611e-05,\n 6.4298e-05, 7.5858e-05, 1.0628e-04, 6.2393e-05, 5.6462e-05, 8.7126e-05,\n 7.1471e-05, 6.5175e-05, 1.0228e-04, 4.7828e-05, 6.1129e-05, 1.3637e-04,\n 8.1554e-05, 7.9041e-05, 8.8118e-05, 9.2059e-05, 1.0108e-04, 8.1213e-05,\n 7.7058e-05, 1.1382e-04, 5.7289e-05, 1.0196e-04, 8.1267e-05, 8.8083e-05,\n 9.0421e-05, 1.0164e-04, 8.5605e-05, 7.2819e-05, 9.0387e-05, 7.4485e-05,\n 8.9011e-05, 7.2438e-05, 6.4429e-05, 9.0421e-05, 5.7645e-05, 9.5814e-05,\n 7.4039e-05, 7.6638e-05, 8.1094e-05, 9.5955e-05, 7.8677e-05, 8.4874e-05,\n 8.8072e-05, 7.4790e-05, 9.0508e-05, 1.0953e-04, 6.7421e-05, 1.3098e-04,\n 7.0891e-05, 7.6291e-05, 7.5750e-05, 7.4933e-05, 5.6367e-05, 8.4441e-05,\n 8.3688e-05, 6.8537e-05, 9.3435e-05, 7.5785e-05, 7.1543e-05, 5.8774e-05,\n 1.6682e-04, 5.7500e-05, 9.1815e-05, 6.8485e-05, 7.3267e-05, 8.0846e-05,\n 1.1637e-04, 7.3753e-05, 6.3681e-05, 9.8934e-05, 6.8585e-05, 8.6584e-05,\n 7.9001e-05, 6.9743e-05, 7.6759e-05, 1.0464e-04, 7.9571e-05, 6.8383e-05,\n 1.2211e-04, 7.2213e-05, 1.0599e-04, 6.5506e-05, 7.3396e-05, 9.0114e-05,\n 9.6428e-05, 7.8010e-05, 9.1807e-05, 7.7939e-05, 7.6842e-05, 6.8402e-05,\n 6.4268e-05, 6.8503e-05, 1.0579e-04, 8.9884e-05, 6.6973e-05, 9.7088e-05,\n 7.3179e-05, 5.6461e-05, 9.5700e-05, 6.6400e-05, 5.9170e-05, 7.5419e-05,\n 8.2268e-05, 7.8889e-05, 5.1525e-05, 6.8475e-05, 7.8440e-05, 1.0967e-04,\n 6.9869e-05, 5.8837e-05, 5.4389e-05, 6.9697e-05, 7.7583e-05, 8.2186e-05,\n 9.0512e-05, 9.9071e-05, 7.3221e-05, 6.5244e-05, 6.6412e-05, 1.1660e-04,\n 7.6357e-05, 9.5597e-05, 1.3771e-04, 1.1727e-04, 6.1555e-05, 6.8617e-05,\n 9.2715e-05, 7.7988e-05, 8.2645e-05, 7.0963e-05, 7.1068e-05, 8.1020e-05,\n 9.5576e-05, 7.5441e-05, 7.2869e-05, 7.3098e-05, 7.4034e-05, 9.4628e-05,\n 9.1677e-05, 6.4768e-05, 8.1106e-05, 1.0330e-04, 1.0022e-04, 8.7793e-05,\n 7.1874e-05, 1.0560e-04, 9.0404e-05, 6.8557e-05, 6.7871e-05, 8.9238e-05,\n 9.2286e-05, 5.8317e-05, 7.9830e-05, 6.2663e-05, 8.3620e-05, 7.8149e-05,\n 8.5944e-05, 6.9084e-05, 9.5227e-05, 6.6487e-05, 8.0305e-05, 8.1712e-05,\n 8.2511e-05, 6.3915e-05, 6.1141e-05, 8.3292e-05, 1.0207e-04, 7.1343e-05,\n 6.5020e-05, 7.2091e-05, 5.1742e-05, 8.5058e-05, 6.3585e-05, 9.7868e-05,\n 9.1490e-05, 9.9497e-05, 5.3711e-05, 6.7703e-05, 7.6611e-05, 6.6150e-05,\n 5.9130e-05, 6.3545e-05, 8.6123e-05, 7.9870e-05, 8.1166e-05, 1.6941e-04,\n 8.3141e-05, 8.4700e-05, 1.0173e-04, 6.4970e-05, 1.0639e-04, 8.4820e-05,\n 9.0221e-05, 5.3956e-05, 8.5816e-05, 9.0874e-05, 8.1447e-05, 9.0657e-05,\n 7.5157e-05, 8.0270e-05, 8.4183e-05, 7.7899e-05, 7.8708e-05, 8.3505e-05,\n 7.8443e-05, 6.6162e-05, 7.3741e-05, 8.1946e-05, 8.7127e-05, 9.9863e-05,\n 6.1553e-05, 9.0703e-05, 6.9988e-05, 9.9942e-05, 6.1316e-05, 6.5354e-05,\n 7.0351e-05, 9.3356e-05, 9.7646e-05, 7.4664e-05, 1.0573e-04, 1.1103e-04,\n 8.0231e-05, 7.1690e-05, 9.1827e-05, 8.2427e-05, 6.2492e-05, 7.0337e-05,\n 8.9927e-05, 1.8154e-04, 8.1747e-05, 8.5976e-05, 9.9535e-05, 8.8756e-05,\n 7.1424e-05, 7.9903e-05, 5.3677e-05, 8.1276e-05, 6.8675e-05, 8.0012e-05,\n 9.2465e-05, 7.4134e-05, 9.2149e-05, 5.4630e-05, 8.7574e-05, 6.0973e-05,\n 9.5834e-05, 8.9957e-05, 6.3466e-05, 9.3110e-05, 8.8340e-05, 5.7320e-05,\n 1.0265e-04, 7.7786e-05, 5.5678e-05, 8.6577e-05, 6.9418e-05, 6.6665e-05,\n 7.1249e-05, 9.1498e-05, 7.3993e-05, 8.0080e-05, 8.9907e-05, 7.8582e-05,\n 7.3805e-05, 1.1810e-04, 6.5748e-05, 7.8805e-05, 7.9566e-05, 1.0304e-04,\n 6.8995e-05, 1.0601e-04, 8.5454e-05, 7.2140e-05, 7.5285e-05, 6.2888e-05,\n 7.5692e-05, 6.3439e-05, 8.9816e-05, 7.9192e-05, 5.7944e-05, 5.7112e-05,\n 6.6068e-05, 9.6226e-05, 5.8398e-05, 7.9959e-05, 1.0117e-04, 6.4187e-05,\n 6.2107e-05, 9.1229e-05, 7.8129e-05, 9.2890e-05, 1.0368e-04, 8.0049e-05,\n 5.2279e-05, 6.7065e-05, 7.6999e-05, 8.0353e-05, 6.8660e-05, 8.6473e-05,\n 8.9579e-05, 9.4978e-05, 8.5223e-05, 6.5604e-05, 8.5023e-05, 9.6538e-05,\n 9.9599e-05, 7.4634e-05, 9.7656e-05, 7.8402e-05, 8.5238e-05, 7.3040e-05,\n 9.6489e-05, 1.1308e-04, 8.4670e-05, 8.4849e-05, 6.0422e-05, 8.9862e-05,\n 6.5811e-05, 6.9821e-05, 7.0601e-05, 7.9594e-05, 8.8806e-05, 8.3173e-05,\n 7.6582e-05, 9.4453e-05, 6.5436e-05, 7.8860e-05, 8.9605e-05, 7.2683e-05,\n 7.5471e-05, 1.2534e-04, 9.0010e-05, 7.1167e-05, 6.1294e-05, 6.3167e-05,\n 7.6312e-05, 1.0025e-04, 6.5798e-05, 8.9228e-05, 9.1657e-05, 1.2453e-04,\n 6.8987e-05, 1.1476e-04, 7.8456e-05, 7.0452e-05, 9.8432e-05, 1.2175e-04,\n 6.9026e-05, 8.4681e-05, 8.4602e-05, 6.8929e-05, 6.2035e-05, 7.6582e-05,\n 7.9701e-05, 8.1276e-05, 7.3286e-05, 1.0012e-04, 9.2251e-05, 8.5758e-05,\n 7.5400e-05, 1.1265e-04, 7.6971e-05, 7.1076e-05, 9.1112e-05, 1.0891e-04,\n 7.9465e-05, 8.9334e-05, 8.4763e-05, 1.3443e-04, 7.8676e-05, 1.1371e-04,\n 6.0353e-05, 1.1043e-04, 6.3244e-05, 1.1084e-04, 6.6983e-05, 7.0307e-05,\n 1.0113e-04, 1.1338e-04, 8.9911e-05, 8.6220e-05, 9.3854e-05, 7.6121e-05,\n 7.3335e-05, 7.0490e-05, 7.7074e-05, 8.2446e-05, 9.7714e-05, 8.5861e-05,\n 8.0054e-05, 6.2086e-05, 5.6873e-05, 7.4090e-05, 8.3842e-05, 1.0013e-04,\n 7.0491e-05, 1.0398e-04, 9.4761e-05, 9.9555e-05, 7.5219e-05, 7.8840e-05,\n 8.7417e-05, 7.8029e-05, 1.1169e-04, 7.4614e-05, 9.3823e-05, 7.4534e-05,\n 8.6591e-05, 6.5262e-05, 6.7971e-05, 6.9949e-05, 9.4687e-05, 5.6783e-05,\n 9.5330e-05, 6.4467e-05, 7.2713e-05, 7.2989e-05, 5.7483e-05, 5.2634e-05,\n 9.4237e-05, 7.2743e-05, 6.1976e-05, 5.1105e-05, 1.1844e-04, 9.6508e-05,\n 8.3919e-05, 7.8081e-05, 6.1649e-05, 9.5045e-05, 7.2898e-05, 9.9954e-05,\n 6.9288e-05, 1.0382e-04, 8.7828e-05, 1.1776e-04, 6.2349e-05, 5.6138e-05,\n 8.5009e-05, 7.7100e-05, 8.3858e-05, 5.5822e-05, 8.1764e-05, 9.9986e-05,\n 7.8230e-05, 7.2563e-05, 1.1527e-04, 5.6187e-05, 6.4458e-05, 8.0152e-05,\n 6.7075e-05, 6.0722e-05, 7.6461e-05, 9.3941e-05, 9.7532e-05, 5.9559e-05,\n 7.7522e-05, 8.1723e-05, 7.5089e-05, 9.7448e-05, 8.6352e-05, 8.5514e-05,\n 8.2527e-05, 8.2690e-05, 5.2719e-05, 7.4208e-05, 9.5420e-05, 1.0038e-04,\n 9.1589e-05, 7.9584e-05, 1.2014e-04, 8.4123e-05, 9.4772e-05, 8.6468e-05,\n 8.8833e-05, 1.0636e-04, 9.2120e-05, 8.5711e-05, 6.3368e-05, 7.5171e-05,\n 8.2695e-05, 8.3628e-05, 1.1828e-04, 8.0662e-05, 1.4772e-04, 7.4693e-05,\n 9.9198e-05, 8.1108e-05], device='cuda:0')"
24
  },
25
  "4": {
26
+ "step": "tensor(3756.)",
27
+ "exp_avg": "tensor([[-3.4391e-05, -3.5256e-04, -4.3252e-04, ..., -1.8171e-04,\n -2.5565e-04, -2.6654e-04],\n [-3.9092e-04, -2.4294e-04, -1.0229e-03, ..., -7.9166e-04,\n -1.0557e-04, -1.7626e-05],\n [-2.4305e-04, -1.8652e-04, -4.5418e-05, ..., 6.5509e-05,\n 8.2397e-04, 2.2071e-04],\n ...,\n [ 4.3741e-04, -1.1140e-04, -5.7404e-04, ..., 3.0648e-04,\n -1.7183e-04, 1.2139e-05],\n [-2.0305e-04, 3.1875e-04, -2.8852e-04, ..., 1.0013e-04,\n -4.2142e-04, 5.0660e-05],\n [ 2.6191e-05, 4.3236e-04, -3.2897e-04, ..., -8.7541e-05,\n -7.8050e-05, -1.2225e-04]], device='cuda:0')",
28
+ "exp_avg_sq": "tensor([[5.9783e-07, 1.0108e-06, 4.7227e-07, ..., 5.8347e-07, 9.1972e-07,\n 1.0363e-06],\n [1.4595e-06, 7.8790e-07, 2.1549e-06, ..., 1.6828e-06, 1.9239e-06,\n 1.4089e-06],\n [1.0708e-06, 7.7245e-07, 1.1206e-06, ..., 1.7251e-06, 1.6237e-06,\n 1.1080e-06],\n ...,\n [1.7025e-06, 8.5543e-07, 1.1475e-06, ..., 1.5056e-06, 2.0411e-06,\n 2.3060e-06],\n [1.2905e-06, 9.6499e-07, 1.1635e-06, ..., 1.3275e-06, 1.9544e-06,\n 2.0286e-06],\n [1.2500e-06, 1.2009e-06, 1.8497e-06, ..., 1.4400e-06, 2.3834e-06,\n 1.8763e-06]], device='cuda:0')"
29
  },
30
  "5": {
31
+ "step": "tensor(3756.)",
32
+ "exp_avg": "tensor([[-1.4355e-05, 3.4932e-05, -1.4439e-04, ..., -3.6623e-05,\n 1.1373e-04, -6.8252e-05],\n [-1.5023e-04, -1.0828e-04, 6.8987e-05, ..., -4.5354e-04,\n -3.1898e-04, -1.5809e-04],\n [-1.1518e-04, -1.0717e-04, -2.3383e-04, ..., 1.8232e-05,\n 1.1541e-04, 7.7687e-05],\n ...,\n [-1.3518e-04, 2.9120e-04, -5.6392e-05, ..., -1.1116e-04,\n 4.4884e-05, 1.1508e-04],\n [ 1.6986e-04, -1.4886e-04, -7.5104e-05, ..., 2.7816e-04,\n -6.6326e-05, 9.2276e-05],\n [-1.8153e-04, -1.2126e-04, -8.4295e-04, ..., -1.9631e-04,\n 1.8855e-04, -2.0120e-05]], device='cuda:0')",
33
+ "exp_avg_sq": "tensor([[1.6407e-07, 1.0444e-07, 1.4908e-07, ..., 2.4003e-07, 2.7275e-07,\n 2.0669e-07],\n [3.8078e-07, 1.9902e-07, 3.1786e-07, ..., 4.4809e-07, 4.8252e-07,\n 4.4533e-07],\n [3.4032e-07, 3.1527e-07, 3.5080e-07, ..., 4.8132e-07, 4.6884e-07,\n 3.8657e-07],\n ...,\n [4.1920e-07, 3.1640e-07, 4.9826e-07, ..., 3.9984e-07, 6.4873e-07,\n 5.7168e-07],\n [3.2181e-07, 4.4911e-07, 3.9630e-07, ..., 3.7836e-07, 5.4745e-07,\n 5.4970e-07],\n [3.6029e-07, 3.4294e-07, 4.7732e-07, ..., 5.0040e-07, 6.1011e-07,\n 4.4770e-07]], device='cuda:0')"
34
  },
35
  "6": {
36
+ "step": "tensor(3756.)",
37
+ "exp_avg": "tensor([ 0.0007, -0.0007], device='cuda:0')",
38
+ "exp_avg_sq": "tensor([1.1519e-05, 1.1519e-05], device='cuda:0')"
39
  }
40
  },
41
  "param_groups": [
42
  {
43
+ "lr": 0.0007940987335200904,
44
  "name": "shared",
45
  "betas": [
46
  0.9,
 
64
  ]
65
  },
66
  {
67
+ "lr": 0.0007940987335200904,
68
  "name": "scale_256",
69
  "betas": [
70
  0.9,
 
85
  ]
86
  },
87
  {
88
+ "lr": 0.0007940987335200904,
89
  "name": "scale_512",
90
  "betas": [
91
  0.9,
 
106
  ]
107
  },
108
  {
109
+ "lr": 0.00039715242044697206,
110
  "name": "fusion",
111
  "betas": [
112
  0.9,
 
133
  "T_i": 10,
134
  "T_mult": 2,
135
  "eta_min": 1e-06,
136
+ "T_cur": 3,
137
  "base_lrs": [
138
  0.001,
139
  0.001,
140
  0.001,
141
  0.0005
142
  ],
143
+ "last_epoch": 3,
144
  "_step_count": 0,
145
  "_is_initial": false,
146
  "_get_lr_called_within_step": false,
147
  "_last_lr": [
148
+ 0.0007940987335200904,
149
+ 0.0007940987335200904,
150
+ 0.0007940987335200904,
151
+ 0.00039715242044697206
152
  ]
153
  },
154
  "metrics": {
155
+ "best_val_acc": 70.132,
156
+ "best_epoch": 2,
157
  "scale_accuracies": {
158
+ "256": 69.576,
159
+ "512": 69.884
160
  }
161
  },
162
  "train_config": {