Upload weights and configs - David-fully_shared-weighted_sum - Run 20251012_135249
Browse files- weights/David-fully_shared-weighted_sum/20251012_135249/MODEL_SUMMARY.txt +55 -0
- weights/David-fully_shared-weighted_sum/20251012_135249/best_model_acc68.06.safetensors +3 -0
- weights/David-fully_shared-weighted_sum/20251012_135249/best_model_acc68.06_metadata.json +237 -0
- weights/David-fully_shared-weighted_sum/20251012_135249/david_config.json +29 -0
- weights/David-fully_shared-weighted_sum/20251012_135249/train_config.json +50 -0
- weights/David-fully_shared-weighted_sum/20251012_135249/training_history.json +25 -0
weights/David-fully_shared-weighted_sum/20251012_135249/MODEL_SUMMARY.txt
ADDED
|
@@ -0,0 +1,55 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
|
| 2 |
+
╔══════════════════════════════════════════════════════════════╗
|
| 3 |
+
║ DAVID MODEL SUMMARY ║
|
| 4 |
+
╠══════════════════════════════════════════════════════════════╣
|
| 5 |
+
║ ║
|
| 6 |
+
║ 🎯 VALIDATION ACCURACY: 68.06% ║
|
| 7 |
+
║ ║
|
| 8 |
+
╚══════════════════════════════════════════════════════════════╝
|
| 9 |
+
|
| 10 |
+
MODEL: David-fully_shared-weighted_sum
|
| 11 |
+
RUN ID: 20251012_135249
|
| 12 |
+
BEST EPOCH: 1/10
|
| 13 |
+
|
| 14 |
+
═══════════════════════════════════════════════════════════════
|
| 15 |
+
|
| 16 |
+
📊 PERFORMANCE BREAKDOWN
|
| 17 |
+
|
| 18 |
+
Final Training Accuracy: 63.40%
|
| 19 |
+
Best Validation Accuracy: 68.06%
|
| 20 |
+
|
| 21 |
+
Per-Scale Accuracies:
|
| 22 |
+
• Scale 256: 66.92%
|
| 23 |
+
• Scale 512: 67.88%
|
| 24 |
+
|
| 25 |
+
═══════════════════════════════════════════════════════════════
|
| 26 |
+
|
| 27 |
+
🏗️ ARCHITECTURE
|
| 28 |
+
|
| 29 |
+
Preset: small_fast
|
| 30 |
+
Sharing Mode: fully_shared
|
| 31 |
+
Fusion Mode: weighted_sum
|
| 32 |
+
Scales: 2 scales - [256, 512]
|
| 33 |
+
Feature Dim: 512
|
| 34 |
+
Parameters: 656,898
|
| 35 |
+
|
| 36 |
+
═══════════════════════════════════════════════════════════════
|
| 37 |
+
|
| 38 |
+
📈 TRAINING CURVE
|
| 39 |
+
|
| 40 |
+
Epoch | Train Acc | Val Acc | Learning Rate
|
| 41 |
+
------|-----------|----------|--------------
|
| 42 |
+
1 | 63.40% | 68.06% 👑 | 9.76e-04
|
| 43 |
+
|
| 44 |
+
═══════════════════════════════════════════════════════════════
|
| 45 |
+
|
| 46 |
+
📁 FILES
|
| 47 |
+
|
| 48 |
+
Best Model: best_model_acc68.06.safetensors
|
| 49 |
+
Config: david_config.json
|
| 50 |
+
Training Cfg: train_config.json
|
| 51 |
+
History: training_history.json
|
| 52 |
+
|
| 53 |
+
═══════════════════════════════════════════════════════════════
|
| 54 |
+
|
| 55 |
+
Generated: 2025-10-12 13:54:33
|
weights/David-fully_shared-weighted_sum/20251012_135249/best_model_acc68.06.safetensors
ADDED
|
@@ -0,0 +1,3 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:5c5ba240557bbcf26ceb2729a0756c153289d89c6bd79156e81493dec68cee96
|
| 3 |
+
size 2628344
|
weights/David-fully_shared-weighted_sum/20251012_135249/best_model_acc68.06_metadata.json
ADDED
|
@@ -0,0 +1,237 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epoch": 0,
|
| 3 |
+
"optimizer_state_dict": {
|
| 4 |
+
"state": {
|
| 5 |
+
"0": {
|
| 6 |
+
"step": "tensor(1252.)",
|
| 7 |
+
"exp_avg": "tensor([[-9.1365e-04, 7.1314e-04, 1.2978e-03, ..., -9.5756e-05,\n -4.9500e-04, 3.5325e-04],\n [ 1.2111e-03, 1.7542e-03, -7.3941e-04, ..., -1.2323e-04,\n -1.0179e-04, 1.1365e-03],\n [ 1.0063e-03, -4.4049e-05, 2.7084e-04, ..., 1.1131e-03,\n -2.2065e-04, 8.2217e-04],\n ...,\n [ 1.0660e-03, 1.4973e-03, -1.3250e-03, ..., -1.4943e-04,\n -3.8458e-04, -2.8948e-05],\n [-2.1601e-04, -2.6428e-03, -1.1565e-03, ..., -1.2976e-03,\n 6.9112e-04, -1.0415e-03],\n [ 8.5588e-04, -1.8722e-03, 1.1945e-03, ..., -1.0715e-04,\n 6.9357e-05, -7.8243e-05]], device='cuda:0')",
|
| 8 |
+
"exp_avg_sq": "tensor([[5.7551e-06, 2.7967e-05, 1.2876e-05, ..., 3.3045e-06, 2.5263e-06,\n 3.9946e-06],\n [7.3851e-06, 2.8976e-05, 1.9962e-05, ..., 7.1677e-06, 2.4347e-06,\n 1.0261e-05],\n [4.4791e-06, 2.6614e-05, 1.1532e-05, ..., 3.0015e-06, 1.6246e-06,\n 3.2705e-06],\n ...,\n [4.2794e-06, 2.3460e-05, 1.3150e-05, ..., 2.3952e-06, 1.4492e-06,\n 2.6654e-06],\n [4.6014e-06, 2.1044e-05, 1.0615e-05, ..., 3.3452e-06, 1.8650e-06,\n 4.3011e-06],\n [5.3457e-06, 3.5602e-05, 1.3041e-05, ..., 2.7265e-06, 1.9714e-06,\n 2.7347e-06]], device='cuda:0')"
|
| 9 |
+
},
|
| 10 |
+
"1": {
|
| 11 |
+
"step": "tensor(1252.)",
|
| 12 |
+
"exp_avg": "tensor([-1.0345e-03, 1.9513e-02, 1.0113e-02, -1.6058e-02, -5.1214e-03,\n -3.7942e-02, 1.6158e-02, 2.6687e-02, -1.1527e-02, -2.3177e-02,\n -1.6727e-02, -1.8606e-02, -3.2081e-02, 6.4268e-03, -8.2847e-03,\n 4.2176e-02, 6.0668e-03, 2.3389e-02, -2.1579e-02, 1.7081e-02,\n 8.0487e-03, -1.6720e-02, -4.2467e-03, 7.9349e-03, -1.1023e-02,\n 5.4805e-03, 6.0537e-03, -3.4957e-03, 2.3468e-02, -1.9929e-02,\n -1.7329e-02, 6.2308e-03, 4.7578e-03, 1.3289e-02, 1.1283e-02,\n 2.8744e-02, 8.4083e-03, 2.3479e-02, -1.9094e-02, -1.4374e-02,\n 3.7284e-02, 2.7442e-03, -5.7156e-03, 1.3952e-02, 2.8565e-04,\n 2.7523e-03, -4.9754e-03, -1.7910e-03, -3.1806e-02, 1.9222e-03,\n 9.6773e-04, -1.2753e-03, -8.1731e-03, -3.0203e-02, -4.6546e-03,\n 2.5028e-02, -8.0213e-03, -7.1162e-03, 1.9671e-02, -4.8543e-02,\n -4.0187e-03, -1.9892e-02, 7.3711e-03, 2.5200e-03, 1.3144e-02,\n 1.2958e-02, -1.9485e-04, 3.3053e-02, -2.4892e-02, -1.5203e-02,\n 6.4576e-02, -2.7599e-02, -5.7916e-02, -2.3090e-02, 2.2997e-02,\n 6.4076e-02, 5.4546e-02, 4.8258e-02, -3.1591e-02, -4.2180e-02,\n 4.3268e-03, 6.6534e-02, 3.3978e-03, 3.3881e-02, 2.5067e-02,\n 1.9308e-02, -1.2058e-02, 3.6164e-02, -5.8642e-03, 3.3834e-02,\n 4.3491e-03, 2.6756e-02, 1.0382e-02, -1.1091e-02, -1.4554e-02,\n 1.8379e-03, -1.4473e-02, 2.0115e-02, 3.0133e-02, -1.8080e-02,\n -4.2258e-02, 3.0788e-02, 1.5329e-02, -1.0580e-03, -1.7321e-02,\n 5.6011e-03, -2.9575e-02, 8.0558e-03, -1.4180e-02, -1.9074e-02,\n 1.4400e-02, 1.2149e-02, -8.5364e-03, -4.4423e-03, -2.3430e-02,\n -2.0834e-02, -1.6281e-02, 2.2583e-03, 1.8610e-02, 5.9026e-03,\n 1.5505e-02, 1.9835e-02, 2.5463e-02, 3.5602e-03, 3.8203e-03,\n -1.2053e-02, 1.4359e-02, 3.0717e-03, -1.5091e-02, -2.8394e-02,\n 1.0281e-02, 1.2756e-02, 3.7822e-03, -6.3951e-03, 6.9287e-03,\n -1.9281e-03, 1.1289e-02, 2.1973e-02, -3.4391e-02, 2.1849e-02,\n 2.7867e-03, 5.0467e-03, 1.7006e-03, 3.6517e-02, -1.1329e-02,\n 1.5013e-02, -1.7978e-02, 2.9485e-02, 3.7734e-02, -3.9237e-03,\n 2.9891e-02, -1.2818e-02, -1.5981e-02, 2.0484e-03, 2.1014e-02,\n 4.7011e-03, -4.2237e-03, -5.2642e-03, -5.2205e-03, 2.0118e-03,\n -1.5149e-03, -2.4206e-02, -4.4254e-02, 2.1544e-03, 3.4052e-03,\n -2.1403e-02, -1.0162e-03, -2.3011e-02, -2.1230e-02, -2.4202e-02,\n -2.5358e-02, -2.1525e-02, -1.5618e-02, 3.8237e-02, -5.6486e-03,\n 2.0650e-02, 2.6039e-02, -6.5003e-04, 1.3047e-02, -1.2490e-02,\n -1.5246e-02, 1.6331e-02, 1.3839e-02, -5.7555e-03, -1.9315e-03,\n -1.3533e-02, -8.7072e-03, -1.0610e-04, -3.3873e-03, 5.9677e-03,\n -3.0142e-02, -1.4054e-02, -2.3672e-02, -3.2666e-02, 1.5289e-02,\n -1.0889e-02, 4.2117e-02, 1.3987e-03, 4.7731e-03, 1.8773e-02,\n 6.6232e-03, 1.8628e-03, 1.9037e-02, -1.2155e-02, -1.5573e-02,\n 3.1474e-02, -8.0962e-03, 8.3288e-03, 2.7768e-02, 3.7709e-03,\n 1.0615e-02, 2.2151e-02, -4.3566e-02, 9.1572e-04, 2.5451e-02,\n 1.1335e-02, 1.6184e-02, 4.6024e-02, -8.8379e-03, -2.2420e-02,\n -1.8083e-02, -3.2445e-03, 2.0148e-02, 1.7448e-02, -2.9068e-03,\n 4.0914e-02, -3.3368e-03, 1.5973e-02, 1.8882e-02, -3.7531e-02,\n -1.9620e-02, 1.0111e-02, -3.8294e-02, 9.9039e-03, -1.7152e-02,\n 2.0895e-02, -2.9584e-02, 1.6877e-03, 4.6363e-04, -1.6772e-03,\n 4.0323e-02, -1.4859e-02, 2.8846e-03, 9.4247e-03, -2.4728e-02,\n 3.4485e-02, 4.5092e-03, -1.9295e-03, 6.4870e-04, 2.5567e-02,\n -3.3639e-03, -2.3127e-02, -4.3041e-03, 4.1203e-04, 1.7587e-02,\n -3.5304e-03, 2.7396e-03, -7.7699e-03, -1.4269e-02, 7.4928e-03,\n -9.8451e-03, 3.1216e-02, -2.0391e-02, -1.1599e-02, 2.3838e-03,\n 1.9067e-02, -1.9385e-02, 2.4863e-02, 6.5175e-03, -2.0753e-02,\n -4.3915e-03, 8.5684e-03, -3.0143e-02, -4.0948e-02, 5.3038e-02,\n 1.3181e-02, 8.8739e-03, 2.6790e-02, -7.3524e-03, -5.9137e-03,\n -3.3235e-02, 5.5750e-03, 1.4614e-02, -4.6760e-02, -4.0284e-03,\n -1.3295e-02, 1.2139e-03, 4.8656e-03, -2.4471e-02, -2.2732e-03,\n -3.3758e-03, 1.0988e-02, -2.0542e-03, -7.2669e-03, 7.4211e-03,\n 3.7647e-03, 2.0904e-02, 3.8038e-06, 3.2961e-02, -1.5480e-02,\n -1.9996e-02, -1.4822e-02, 1.0667e-03, 1.3238e-02, -1.6433e-02,\n -1.1833e-02, -1.7020e-03, -1.1930e-02, -2.2586e-02, 3.7736e-02,\n -2.2706e-03, 7.1676e-03, 7.6457e-03, 3.3980e-03, -1.2568e-02,\n -1.4062e-02, 1.9759e-02, 2.9188e-02, -7.7580e-02, 2.1164e-02,\n 9.6864e-03, -5.4527e-03, -1.6083e-02, 2.3503e-02, -1.4662e-02,\n -4.0402e-02, -1.2838e-02, 2.2230e-02, 7.9791e-03, -2.9115e-02,\n -2.0610e-02, -2.0454e-02, 8.3565e-03, 9.3616e-03, -2.1483e-02,\n -1.4979e-02, -1.8036e-02, 1.0367e-02, 9.7804e-03, 5.1813e-02,\n -1.1124e-02, -5.0036e-02, -5.2014e-02, -1.2027e-02, 1.7251e-02,\n 7.3057e-03, 2.3335e-03, -1.2920e-02, 2.0217e-02, -6.1781e-02,\n 2.8478e-03, -1.5429e-02, -2.7365e-03, 4.9415e-02, -3.2768e-02,\n -1.9972e-02, -6.7088e-03, -9.6343e-03, 7.1084e-03, 5.0492e-02,\n -2.7585e-03, 4.3438e-02, -4.7207e-02, 1.3693e-04, -2.8103e-03,\n 6.7930e-03, 1.1639e-02, 3.9990e-02, 2.5401e-04, -2.3778e-02,\n 1.9834e-02, 3.3473e-02, 2.8521e-02, 1.4605e-02, -2.9707e-02,\n -7.2060e-03, 1.4748e-03, -3.1203e-02, 1.6537e-02, 9.4049e-03,\n -1.5118e-02, 2.1989e-02, 1.0663e-02, 2.7452e-03, 1.4000e-02,\n 2.2562e-02, 1.5475e-02, 3.8973e-02, 2.7099e-03, 7.4269e-03,\n 2.4185e-02, -2.8552e-02, -1.2434e-02, 5.8214e-03, 2.4508e-02,\n -2.5285e-03, 1.0422e-02, 1.1991e-02, 2.7905e-03, -3.1070e-03,\n 1.5817e-02, -2.4574e-02, 1.3334e-02, -3.3561e-03, 8.9805e-03,\n -1.1151e-02, 2.0020e-03, -6.9065e-04, -5.4094e-02, -6.5294e-04,\n 2.4417e-02, 2.3251e-02, -1.4618e-02, -2.1504e-02, -1.3897e-04,\n -2.1971e-02, 6.1102e-03, 1.2341e-02, -1.9056e-02, -6.1444e-03,\n -1.3419e-02, 1.7806e-03, -5.4216e-03, 2.2606e-02, -7.8131e-03,\n -4.9433e-02, -1.5525e-03, -2.0298e-02, -6.7116e-03, -2.1058e-02,\n -2.7041e-02, -2.5522e-02, -1.1122e-02, -2.7920e-03, 1.1766e-03,\n 2.6984e-02, 6.6903e-03, 1.3507e-03, -7.9892e-03, 9.3236e-03,\n 2.1629e-02, 2.3170e-03, 4.1470e-03, -1.8459e-03, -1.7042e-02,\n -1.3953e-02, 4.6930e-02, -1.0755e-02, -1.9298e-02, -4.8653e-03,\n -2.7014e-03, -5.9981e-02, 2.1422e-02, -8.0414e-03, -5.4148e-04,\n -4.8557e-02, 4.7861e-04, -5.0461e-03, -2.2092e-02, 4.1904e-03,\n -2.8695e-02, 1.6383e-03, 2.8598e-02, -1.1341e-02, 2.5723e-03,\n 3.5075e-02, -1.7383e-02, 1.0164e-02, 2.3384e-03, 9.6156e-03,\n -2.8520e-03, 1.1735e-02, 1.3666e-02, 1.4380e-02, -2.4961e-02,\n 1.7870e-02, -1.5116e-03, -2.6826e-02, -1.5368e-02, -2.3807e-02,\n 2.1702e-03, 1.4501e-02, -1.5726e-02, 4.7823e-03, 5.3529e-03,\n -5.5939e-02, 3.8301e-02, -4.6055e-03, -1.7109e-02, 3.3038e-02,\n -1.6546e-02, -2.2576e-02, 2.7513e-02, 2.9611e-02, -5.6201e-04,\n 1.1744e-02, -4.8179e-03, 1.2994e-02, -1.6764e-02, 7.4337e-03,\n 2.2403e-04, -4.6953e-03, 4.4981e-02, -2.9135e-03, -2.1863e-02,\n -1.6363e-02, -3.0946e-02, 1.7548e-02, -1.2270e-02, 1.9410e-02,\n -9.2116e-03, -1.2283e-02], device='cuda:0')",
|
| 13 |
+
"exp_avg_sq": "tensor([0.0046, 0.0062, 0.0037, 0.0036, 0.0046, 0.0044, 0.0047, 0.0043, 0.0048,\n 0.0048, 0.0043, 0.0044, 0.0064, 0.0042, 0.0056, 0.0053, 0.0025, 0.0037,\n 0.0038, 0.0035, 0.0040, 0.0042, 0.0043, 0.0054, 0.0050, 0.0044, 0.0044,\n 0.0036, 0.0051, 0.0032, 0.0044, 0.0056, 0.0042, 0.0052, 0.0029, 0.0036,\n 0.0034, 0.0041, 0.0052, 0.0042, 0.0055, 0.0043, 0.0054, 0.0062, 0.0042,\n 0.0040, 0.0048, 0.0044, 0.0036, 0.0045, 0.0034, 0.0036, 0.0054, 0.0035,\n 0.0059, 0.0040, 0.0033, 0.0046, 0.0044, 0.0037, 0.0037, 0.0053, 0.0056,\n 0.0033, 0.0026, 0.0040, 0.0044, 0.0055, 0.0067, 0.0048, 0.0066, 0.0043,\n 0.0061, 0.0046, 0.0053, 0.0051, 0.0052, 0.0049, 0.0041, 0.0052, 0.0061,\n 0.0058, 0.0049, 0.0044, 0.0047, 0.0040, 0.0021, 0.0063, 0.0053, 0.0045,\n 0.0042, 0.0047, 0.0035, 0.0054, 0.0054, 0.0053, 0.0057, 0.0052, 0.0037,\n 0.0052, 0.0047, 0.0046, 0.0052, 0.0032, 0.0038, 0.0043, 0.0052, 0.0030,\n 0.0055, 0.0049, 0.0029, 0.0051, 0.0051, 0.0047, 0.0036, 0.0050, 0.0050,\n 0.0065, 0.0044, 0.0042, 0.0042, 0.0037, 0.0042, 0.0049, 0.0054, 0.0053,\n 0.0054, 0.0042, 0.0054, 0.0039, 0.0056, 0.0043, 0.0038, 0.0034, 0.0045,\n 0.0038, 0.0047, 0.0032, 0.0067, 0.0056, 0.0045, 0.0038, 0.0042, 0.0061,\n 0.0044, 0.0046, 0.0045, 0.0052, 0.0050, 0.0058, 0.0041, 0.0038, 0.0036,\n 0.0043, 0.0045, 0.0042, 0.0042, 0.0053, 0.0030, 0.0058, 0.0031, 0.0043,\n 0.0050, 0.0041, 0.0047, 0.0054, 0.0033, 0.0035, 0.0040, 0.0052, 0.0043,\n 0.0041, 0.0030, 0.0041, 0.0048, 0.0040, 0.0041, 0.0043, 0.0041, 0.0050,\n 0.0056, 0.0061, 0.0029, 0.0045, 0.0049, 0.0041, 0.0027, 0.0031, 0.0053,\n 0.0036, 0.0031, 0.0028, 0.0044, 0.0046, 0.0053, 0.0046, 0.0050, 0.0039,\n 0.0052, 0.0072, 0.0035, 0.0041, 0.0040, 0.0057, 0.0049, 0.0033, 0.0033,\n 0.0035, 0.0042, 0.0060, 0.0040, 0.0053, 0.0045, 0.0054, 0.0032, 0.0039,\n 0.0040, 0.0046, 0.0038, 0.0047, 0.0046, 0.0055, 0.0055, 0.0049, 0.0035,\n 0.0052, 0.0040, 0.0057, 0.0042, 0.0048, 0.0058, 0.0035, 0.0053, 0.0037,\n 0.0049, 0.0052, 0.0047, 0.0050, 0.0044, 0.0058, 0.0058, 0.0044, 0.0048,\n 0.0049, 0.0030, 0.0053, 0.0047, 0.0048, 0.0039, 0.0044, 0.0039, 0.0045,\n 0.0028, 0.0049, 0.0049, 0.0050, 0.0048, 0.0051, 0.0044, 0.0046, 0.0047,\n 0.0055, 0.0036, 0.0032, 0.0027, 0.0042, 0.0040, 0.0054, 0.0039, 0.0055,\n 0.0033, 0.0036, 0.0051, 0.0049, 0.0062, 0.0037, 0.0024, 0.0038, 0.0050,\n 0.0047, 0.0057, 0.0045, 0.0047, 0.0047, 0.0043, 0.0051, 0.0030, 0.0048,\n 0.0048, 0.0038, 0.0039, 0.0054, 0.0045, 0.0054, 0.0062, 0.0053, 0.0055,\n 0.0043, 0.0074, 0.0064, 0.0036, 0.0045, 0.0044, 0.0040, 0.0051, 0.0040,\n 0.0046, 0.0038, 0.0049, 0.0043, 0.0042, 0.0065, 0.0042, 0.0054, 0.0035,\n 0.0044, 0.0055, 0.0047, 0.0048, 0.0040, 0.0049, 0.0040, 0.0063, 0.0045,\n 0.0044, 0.0046, 0.0035, 0.0046, 0.0046, 0.0037, 0.0057, 0.0036, 0.0045,\n 0.0042, 0.0031, 0.0037, 0.0042, 0.0044, 0.0045, 0.0051, 0.0036, 0.0047,\n 0.0050, 0.0052, 0.0049, 0.0057, 0.0042, 0.0049, 0.0035, 0.0040, 0.0036,\n 0.0052, 0.0044, 0.0049, 0.0040, 0.0052, 0.0059, 0.0041, 0.0034, 0.0051,\n 0.0052, 0.0050, 0.0056, 0.0035, 0.0045, 0.0053, 0.0041, 0.0068, 0.0042,\n 0.0045, 0.0040, 0.0061, 0.0056, 0.0041, 0.0043, 0.0039, 0.0054, 0.0042,\n 0.0049, 0.0035, 0.0046, 0.0047, 0.0044, 0.0040, 0.0049, 0.0052, 0.0034,\n 0.0054, 0.0062, 0.0028, 0.0050, 0.0055, 0.0040, 0.0040, 0.0033, 0.0038,\n 0.0031, 0.0035, 0.0043, 0.0034, 0.0031, 0.0052, 0.0050, 0.0056, 0.0051,\n 0.0055, 0.0048, 0.0037, 0.0056, 0.0031, 0.0045, 0.0057, 0.0049, 0.0037,\n 0.0052, 0.0037, 0.0043, 0.0040, 0.0063, 0.0033, 0.0046, 0.0045, 0.0055,\n 0.0050, 0.0068, 0.0047, 0.0048, 0.0043, 0.0039, 0.0040, 0.0049, 0.0035,\n 0.0036, 0.0067, 0.0036, 0.0044, 0.0053, 0.0055, 0.0051, 0.0051, 0.0054,\n 0.0059, 0.0057, 0.0033, 0.0036, 0.0053, 0.0047, 0.0031, 0.0042, 0.0045,\n 0.0031, 0.0051, 0.0046, 0.0038, 0.0041, 0.0060, 0.0051, 0.0032, 0.0042,\n 0.0040, 0.0050, 0.0061, 0.0038, 0.0061, 0.0024, 0.0042, 0.0038, 0.0058,\n 0.0039, 0.0054, 0.0047, 0.0041, 0.0063, 0.0043, 0.0036, 0.0036, 0.0042,\n 0.0040, 0.0047, 0.0043, 0.0049, 0.0057, 0.0040, 0.0041, 0.0041, 0.0046,\n 0.0044, 0.0044, 0.0050, 0.0049, 0.0023, 0.0060, 0.0049, 0.0046, 0.0062,\n 0.0044, 0.0039, 0.0051, 0.0043, 0.0040, 0.0049, 0.0042, 0.0053, 0.0041,\n 0.0048, 0.0043, 0.0038, 0.0036, 0.0041, 0.0038, 0.0042, 0.0046],\n device='cuda:0')"
|
| 14 |
+
},
|
| 15 |
+
"2": {
|
| 16 |
+
"step": "tensor(1252.)",
|
| 17 |
+
"exp_avg": "tensor([ 2.7489e-04, 8.5788e-04, 5.5434e-04, -1.2148e-03, -3.0288e-04,\n -2.5640e-03, 1.7938e-03, 1.9798e-03, -1.7085e-04, -1.8919e-03,\n -1.2527e-03, -1.2757e-03, -2.7533e-03, -3.3177e-04, -3.8701e-04,\n 2.6513e-03, 2.8273e-03, 3.9187e-03, -2.2443e-03, 3.1732e-03,\n 6.4545e-04, -1.1814e-03, 4.5654e-04, -7.9659e-04, 4.5045e-04,\n -9.0495e-05, 3.8797e-04, 1.1772e-03, 1.2372e-03, -2.2443e-03,\n -6.6312e-04, -4.8422e-04, 7.9878e-04, 5.4743e-04, 1.1686e-03,\n 3.6091e-03, 8.0261e-04, 1.4709e-03, -5.5920e-04, -9.2071e-04,\n 2.3243e-03, 3.9200e-04, 3.2564e-04, 9.2308e-04, -9.6885e-04,\n 6.8041e-04, -1.0538e-05, 5.8704e-04, -2.8144e-03, 1.2684e-03,\n -4.0751e-04, -5.4753e-04, -8.0800e-04, -2.1936e-03, 1.2769e-04,\n 2.6036e-03, -1.4474e-04, -2.9826e-04, 1.0781e-03, -3.6488e-03,\n 2.2693e-04, -1.4562e-03, 8.0277e-04, 9.2633e-04, 2.8250e-03,\n 1.9512e-03, 5.5122e-04, 2.6434e-03, -3.6322e-04, -9.9319e-04,\n 2.1535e-03, -1.7670e-03, -3.5909e-03, 1.4244e-04, 2.2797e-03,\n 3.9584e-03, 5.0590e-03, 3.6174e-03, -2.7374e-03, -2.1527e-03,\n 3.6050e-04, 4.2081e-03, 1.5346e-04, 3.3643e-03, 1.2577e-03,\n 2.3854e-03, -4.9992e-03, 1.7993e-03, -6.5757e-04, 2.4871e-03,\n 7.9373e-04, 3.3292e-03, 7.8442e-04, -9.0649e-06, -1.4705e-03,\n 4.5914e-04, -2.0191e-04, 1.5126e-03, 1.3873e-03, -1.4366e-03,\n -3.5805e-03, 2.3424e-03, 2.5683e-04, -3.2781e-04, -1.9072e-03,\n 4.8551e-04, -2.3617e-03, 1.2050e-03, -7.8201e-04, -8.7830e-04,\n 2.3679e-03, -1.0770e-05, 1.3215e-04, -5.0389e-04, -1.9328e-03,\n -1.4101e-03, -9.0494e-04, 2.0136e-04, 1.9083e-03, 3.1983e-04,\n 4.6402e-04, 1.8559e-03, 1.5949e-03, 4.2611e-04, 4.5877e-04,\n 9.8964e-06, 1.4969e-03, -5.3407e-04, -3.6850e-04, -4.0410e-03,\n 4.8035e-04, 1.4167e-03, 1.3375e-03, -8.2613e-04, -1.6852e-05,\n -4.9800e-04, 9.9844e-04, 2.6185e-03, -1.7983e-03, 1.9949e-03,\n 1.2675e-03, 9.1593e-04, 2.5759e-04, 2.3463e-03, -2.4264e-03,\n 8.6327e-04, -1.6231e-03, 2.2409e-03, 2.4636e-03, 3.4836e-04,\n 9.7012e-04, -1.4849e-04, -1.2078e-03, 3.3251e-05, 1.9241e-03,\n 4.1494e-04, -1.2403e-03, -1.7760e-04, -3.1418e-04, -1.8414e-04,\n -6.3756e-05, -2.2824e-03, -3.4596e-03, 7.0894e-04, -7.5108e-05,\n -7.6189e-04, 2.0998e-04, -2.4605e-03, -1.0059e-03, -1.3216e-03,\n -1.4829e-03, -1.2487e-03, -2.3111e-03, 2.9463e-03, -2.5278e-04,\n 2.2551e-03, 2.9825e-03, -1.4352e-06, 5.9172e-04, 2.9526e-06,\n -6.8110e-04, 1.2682e-03, 2.3798e-03, -9.2602e-04, -4.4978e-04,\n -1.8226e-03, -8.4813e-04, 3.8561e-04, 1.2105e-03, 1.6279e-03,\n -3.2571e-03, -1.9804e-03, -1.9133e-03, -2.4077e-03, 9.0001e-04,\n -2.9004e-04, 1.8830e-03, 4.9564e-04, 9.6883e-06, 1.2203e-03,\n 1.2955e-03, 3.0034e-04, 1.5740e-03, -1.0692e-03, -1.4640e-03,\n 3.2246e-03, -4.7018e-04, 6.6310e-04, 2.3566e-03, -5.2875e-04,\n 8.7462e-04, 1.8417e-03, -1.7618e-03, -2.2411e-04, 2.5493e-03,\n 8.9122e-04, 1.1297e-03, 2.4005e-03, -1.8225e-03, -2.2349e-03,\n -1.3644e-03, -4.1333e-04, 5.3385e-04, 7.7282e-04, 5.1531e-04,\n 2.9346e-03, 3.2106e-04, 3.9698e-04, 2.0654e-03, -1.6182e-03,\n 2.4365e-04, 5.0997e-04, -1.3222e-03, 1.4099e-04, -1.1256e-03,\n 1.6137e-03, -2.8527e-03, 7.0430e-04, -2.9863e-04, -6.9797e-04,\n 2.1322e-03, -5.5546e-04, 1.5770e-04, 1.6043e-04, -2.8382e-03,\n 2.3250e-03, 1.2029e-03, 4.7858e-04, 1.2028e-03, 3.2272e-04,\n -2.1743e-04, -2.1784e-03, -1.5683e-03, 1.2643e-03, 2.1733e-03,\n 2.0119e-04, -1.3132e-03, 3.4366e-04, -7.7857e-04, -7.0426e-04,\n -3.0387e-04, 1.7062e-03, -2.0030e-03, -7.8686e-04, 1.8965e-04,\n 2.1725e-03, -1.8050e-03, 1.7942e-03, 1.1415e-03, -1.1856e-03,\n -1.5579e-03, 7.1500e-04, -4.7244e-04, -2.4624e-03, 3.1776e-03,\n 8.7298e-04, 3.2352e-04, 2.6199e-03, -7.8677e-04, -6.8274e-04,\n -1.7918e-03, 9.9736e-04, 1.1658e-03, -3.1052e-03, 8.5812e-04,\n -3.9995e-04, 1.2516e-05, 5.9637e-04, -1.6590e-03, -2.4806e-05,\n -5.7624e-04, 1.1087e-03, 5.1393e-04, 5.5657e-04, -1.1203e-04,\n 1.3195e-03, 8.7091e-04, 2.9630e-04, 1.8557e-03, -5.7805e-04,\n -1.6568e-03, -1.0496e-03, -2.1015e-04, 1.5732e-03, -6.6367e-04,\n 1.6618e-04, 1.1372e-04, -3.1450e-04, -1.7723e-03, 2.3197e-03,\n 8.2781e-05, 1.1876e-04, 1.5678e-03, 3.4171e-04, -1.9938e-03,\n -1.7338e-03, 2.3721e-03, 2.9404e-03, -3.2588e-03, 2.0260e-03,\n 1.1238e-03, 3.2950e-04, -9.4212e-04, 2.7985e-03, -3.3649e-04,\n -3.9658e-03, -1.4950e-03, 1.8349e-03, 1.1161e-03, -2.0485e-03,\n -1.9147e-03, -1.8984e-03, 1.7509e-03, 5.4262e-04, -3.9531e-03,\n -1.2244e-03, -8.8040e-04, 8.9640e-04, 5.4889e-04, 3.5309e-03,\n -1.5433e-03, -3.9268e-03, -3.3013e-03, -1.2317e-03, 1.1126e-03,\n 2.9239e-04, -7.1743e-04, -5.2685e-04, 2.8980e-03, -4.4808e-03,\n 8.6353e-04, -1.3845e-03, -7.5681e-04, 3.3841e-03, -3.0929e-03,\n -5.1730e-04, 1.0822e-03, -6.4489e-04, 1.1843e-03, 3.4817e-03,\n 2.5999e-04, 1.9316e-03, -3.2647e-03, 2.7599e-04, 1.7872e-05,\n 7.0990e-04, 1.2979e-03, 1.2486e-03, 1.6784e-04, -1.4107e-03,\n 1.4810e-03, 2.0645e-03, 2.4541e-03, 1.8642e-03, -3.1088e-03,\n -2.7197e-04, -8.9990e-05, -2.3802e-03, 1.0546e-03, 1.4537e-03,\n -3.7258e-04, 1.2954e-03, 4.5250e-04, 1.9834e-04, 9.8158e-04,\n 1.3597e-03, 1.7012e-03, 2.2476e-03, 1.1643e-05, 1.6158e-03,\n 2.3729e-03, -1.2264e-03, -1.1143e-03, -6.3326e-04, 1.4122e-03,\n 7.7072e-05, 1.2841e-03, 6.1061e-04, 2.4191e-05, 6.5063e-04,\n 2.8501e-03, -1.6582e-03, 9.3993e-04, -5.6766e-04, -6.3367e-04,\n -1.1751e-03, -3.2188e-04, 7.3917e-04, -2.2134e-03, -5.8142e-04,\n 2.0996e-03, 1.2368e-03, -9.8467e-04, -1.9734e-03, 4.5382e-04,\n -2.3836e-03, 1.0415e-03, 1.1676e-03, -1.2897e-03, -8.8565e-05,\n 6.2846e-04, 1.1800e-03, -1.8300e-04, 1.9165e-03, -7.9472e-05,\n -3.0051e-03, 7.0851e-04, 2.0721e-04, -3.5383e-04, -1.9741e-03,\n -2.1558e-03, -3.0362e-03, -1.1116e-03, -2.5106e-04, 9.9998e-05,\n 1.8242e-03, 7.3576e-04, -9.4350e-04, -8.9753e-04, 1.1076e-03,\n 1.0381e-03, 7.8988e-04, 2.7870e-04, -3.2995e-04, -6.4528e-04,\n -9.0054e-04, 2.3124e-03, -1.3633e-03, -1.1454e-03, 7.2923e-04,\n -9.5496e-04, -5.2481e-03, 1.7923e-03, -1.3764e-03, 8.0999e-04,\n -3.4137e-03, 3.2291e-04, 9.7261e-04, -2.2610e-03, 6.4759e-04,\n -1.8316e-03, -2.4252e-04, 2.5677e-03, -5.2207e-04, 1.1262e-03,\n 3.6471e-03, -1.2764e-03, 1.0657e-03, -1.8863e-04, -3.7269e-04,\n -2.7916e-04, 2.2260e-03, 1.4885e-03, 9.3488e-04, -2.1675e-03,\n 1.5444e-03, -4.7120e-04, -2.0182e-03, -4.6343e-04, -1.4432e-03,\n -2.6913e-04, 1.9434e-03, -1.3369e-03, 3.2471e-04, 7.9432e-04,\n -3.7064e-03, 2.6003e-03, -1.9404e-03, -8.7754e-04, 1.9813e-03,\n -3.2072e-03, -1.7308e-03, 3.1979e-03, 1.9819e-03, -3.1863e-04,\n 8.5676e-04, 9.7780e-05, 1.3452e-03, -5.7905e-04, 6.5145e-04,\n -8.9984e-06, -9.3405e-04, 2.1899e-03, -5.2431e-04, -1.0903e-03,\n -7.3144e-04, -1.7179e-03, 1.0927e-03, -1.8092e-03, 1.7866e-03,\n -1.9592e-04, -3.9919e-04], device='cuda:0')",
|
| 18 |
+
"exp_avg_sq": "tensor([1.3411e-05, 1.3749e-05, 1.2750e-05, 1.7121e-05, 1.0038e-05, 1.3509e-05,\n 1.3017e-05, 1.4824e-05, 1.0273e-05, 1.3730e-05, 1.4949e-05, 1.4382e-05,\n 1.3116e-05, 1.4380e-05, 1.0760e-05, 1.2488e-05, 5.7571e-05, 1.5363e-05,\n 1.3193e-05, 1.4697e-05, 1.7881e-05, 1.4675e-05, 1.7031e-05, 1.4105e-05,\n 1.1977e-05, 1.9487e-05, 1.7139e-05, 1.3983e-05, 1.2350e-05, 2.9440e-05,\n 1.2791e-05, 1.2364e-05, 1.6325e-05, 1.0950e-05, 1.6570e-05, 1.7723e-05,\n 1.6985e-05, 1.5308e-05, 1.2078e-05, 1.4452e-05, 1.5848e-05, 1.3939e-05,\n 9.6763e-06, 1.1301e-05, 1.3394e-05, 8.4984e-06, 1.4250e-05, 1.3320e-05,\n 2.0112e-05, 1.2840e-05, 1.8410e-05, 1.8110e-05, 1.1843e-05, 1.6004e-05,\n 1.4769e-05, 1.2530e-05, 2.2885e-05, 1.3701e-05, 1.4330e-05, 1.4219e-05,\n 1.5317e-05, 1.2714e-05, 1.3696e-05, 2.0386e-05, 2.5086e-05, 1.3532e-05,\n 1.7494e-05, 1.2792e-05, 1.2885e-05, 1.3762e-05, 1.2681e-05, 1.1494e-05,\n 1.2697e-05, 1.1288e-05, 1.7430e-05, 1.4592e-05, 1.4792e-05, 1.3955e-05,\n 1.4914e-05, 1.0886e-05, 1.3997e-05, 1.3310e-05, 1.4335e-05, 1.3346e-05,\n 1.3168e-05, 1.8197e-05, 4.9769e-05, 1.1022e-05, 1.1097e-05, 1.7167e-05,\n 1.4426e-05, 2.0328e-05, 1.4818e-05, 1.1305e-05, 1.3590e-05, 1.1216e-05,\n 1.4146e-05, 1.2224e-05, 1.0757e-05, 1.1087e-05, 1.1642e-05, 9.8141e-06,\n 1.1913e-05, 2.1720e-05, 1.9472e-05, 1.5533e-05, 1.0152e-05, 1.4585e-05,\n 1.2884e-05, 1.0970e-05, 2.4801e-05, 1.0269e-05, 1.3516e-05, 9.6619e-06,\n 2.2258e-05, 1.4742e-05, 1.1310e-05, 1.4081e-05, 1.9338e-05, 1.2515e-05,\n 1.2819e-05, 2.0426e-05, 1.0722e-05, 1.4548e-05, 1.3606e-05, 1.1287e-05,\n 1.0677e-05, 1.4172e-05, 1.3001e-05, 1.9343e-05, 1.1027e-05, 1.1312e-05,\n 2.0086e-05, 1.7222e-05, 1.4341e-05, 1.7481e-05, 1.1897e-05, 1.9403e-05,\n 1.1211e-05, 1.3449e-05, 1.3887e-05, 1.5890e-05, 1.3243e-05, 1.6011e-05,\n 1.3443e-05, 1.5033e-05, 1.2027e-05, 1.6537e-05, 1.3401e-05, 1.2996e-05,\n 9.3968e-06, 1.6779e-05, 1.2595e-05, 2.0321e-05, 1.5028e-05, 1.8359e-05,\n 1.2585e-05, 1.4685e-05, 1.9001e-05, 1.1087e-05, 1.3923e-05, 1.6102e-05,\n 1.5859e-05, 1.2119e-05, 1.4132e-05, 1.1767e-05, 2.2083e-05, 1.6502e-05,\n 7.9211e-06, 1.3985e-05, 1.5640e-05, 1.2594e-05, 1.9846e-05, 1.4368e-05,\n 1.2867e-05, 1.5653e-05, 1.2257e-05, 1.6601e-05, 1.3585e-05, 1.2379e-05,\n 1.2679e-05, 1.3707e-05, 2.8751e-05, 1.1190e-05, 1.2070e-05, 1.5107e-05,\n 2.7744e-05, 2.3469e-05, 1.2067e-05, 1.6499e-05, 2.0392e-05, 2.1314e-05,\n 1.1308e-05, 1.2344e-05, 1.7287e-05, 1.3221e-05, 1.1810e-05, 1.1768e-05,\n 1.1869e-05, 1.3621e-05, 1.7935e-05, 1.4545e-05, 9.1422e-06, 1.1402e-05,\n 1.3779e-05, 1.7547e-05, 1.9055e-05, 1.8611e-05, 1.5262e-05, 1.5088e-05,\n 1.4951e-05, 1.0846e-05, 1.2995e-05, 1.2876e-05, 1.6333e-05, 1.1146e-05,\n 1.6364e-05, 1.0637e-05, 1.4403e-05, 1.3568e-05, 1.7638e-05, 1.2265e-05,\n 1.0269e-05, 1.4997e-05, 1.7391e-05, 1.3734e-05, 1.6964e-05, 1.1359e-05,\n 2.0822e-05, 1.4332e-05, 1.1555e-05, 1.1801e-05, 1.0128e-05, 1.4101e-05,\n 1.5904e-05, 1.3336e-05, 1.8629e-05, 1.3609e-05, 1.5863e-05, 1.3234e-05,\n 1.2120e-05, 1.6312e-05, 1.5249e-05, 1.4084e-05, 2.0678e-05, 1.0460e-05,\n 1.1800e-05, 9.9060e-06, 1.4757e-05, 1.2023e-05, 1.5787e-05, 1.2084e-05,\n 2.3125e-05, 1.0928e-05, 1.4040e-05, 1.1469e-05, 1.6626e-05, 1.2552e-05,\n 1.3882e-05, 1.5951e-05, 1.0617e-05, 1.3055e-05, 1.9766e-05, 1.7895e-05,\n 3.1407e-05, 1.4383e-05, 1.3912e-05, 1.3007e-05, 1.8285e-05, 1.2160e-05,\n 1.8494e-05, 1.2648e-05, 1.1076e-05, 1.4004e-05, 1.2630e-05, 1.4082e-05,\n 2.6223e-05, 1.3142e-05, 1.3966e-05, 1.1923e-05, 9.9857e-06, 1.2920e-05,\n 1.3499e-05, 1.6330e-05, 1.3064e-05, 1.1918e-05, 2.8033e-05, 1.0121e-05,\n 1.1939e-05, 1.7580e-05, 1.3722e-05, 1.3072e-05, 1.3754e-05, 1.3919e-05,\n 1.5330e-05, 1.4048e-05, 1.3430e-05, 1.4980e-05, 1.5394e-05, 1.3119e-05,\n 1.5109e-05, 1.3708e-05, 1.1340e-05, 1.8370e-05, 1.4524e-05, 1.6820e-05,\n 1.1531e-05, 1.4894e-05, 1.4448e-05, 1.2645e-05, 1.0534e-05, 1.4424e-05,\n 1.5152e-05, 1.3992e-05, 2.2160e-05, 1.4538e-05, 1.3395e-05, 1.4382e-05,\n 1.0536e-05, 2.2646e-05, 1.1420e-05, 1.2661e-05, 1.3896e-05, 1.8322e-05,\n 9.9087e-06, 1.2491e-05, 1.6187e-05, 1.2068e-05, 9.8978e-06, 1.0216e-05,\n 1.3504e-05, 1.6695e-05, 1.1546e-05, 1.2509e-05, 2.6127e-05, 2.0025e-05,\n 1.1633e-05, 1.6815e-05, 1.4490e-05, 1.1651e-05, 1.6661e-05, 1.7400e-05,\n 1.5563e-05, 1.4692e-05, 1.3915e-05, 1.0816e-05, 1.5579e-05, 1.1118e-05,\n 2.1085e-05, 1.5110e-05, 1.2125e-05, 1.3685e-05, 8.9891e-06, 1.2460e-05,\n 1.6149e-05, 1.4397e-05, 1.3160e-05, 1.0757e-05, 2.3652e-05, 1.5040e-05,\n 1.4937e-05, 1.0062e-05, 1.3497e-05, 1.4755e-05, 1.0660e-05, 1.5230e-05,\n 1.2482e-05, 1.4038e-05, 1.4351e-05, 1.6665e-05, 1.8530e-05, 1.2670e-05,\n 1.4107e-05, 1.2813e-05, 1.2226e-05, 1.6555e-05, 1.3308e-05, 1.3742e-05,\n 1.1781e-05, 1.7827e-05, 9.4519e-06, 1.1899e-05, 1.4025e-05, 1.3609e-05,\n 1.0666e-05, 1.1890e-05, 1.3087e-05, 1.0734e-05, 1.0331e-05, 2.7101e-05,\n 1.2638e-05, 1.1057e-05, 1.7969e-05, 1.3733e-05, 1.6621e-05, 1.3323e-05,\n 1.7356e-05, 1.7348e-05, 1.0531e-05, 1.8803e-05, 1.9155e-05, 1.4579e-05,\n 1.1749e-05, 1.2910e-05, 1.5038e-05, 1.4456e-05, 1.4832e-05, 1.1317e-05,\n 1.2412e-05, 1.9479e-05, 1.3108e-05, 1.0964e-05, 1.2166e-05, 2.4714e-05,\n 1.0832e-05, 2.1834e-05, 1.2178e-05, 1.3752e-05, 1.3730e-05, 1.5717e-05,\n 1.7266e-05, 1.5162e-05, 1.2019e-05, 1.3557e-05, 1.3503e-05, 1.5950e-05,\n 1.2024e-05, 1.1900e-05, 1.7568e-05, 2.0945e-05, 1.1070e-05, 2.5574e-05,\n 1.5535e-05, 2.0952e-05, 1.5291e-05, 1.4866e-05, 1.1218e-05, 1.2043e-05,\n 1.4264e-05, 9.5460e-06, 1.3997e-05, 1.4226e-05, 1.4028e-05, 1.7587e-05,\n 1.2680e-05, 1.1724e-05, 1.1739e-05, 1.7899e-05, 1.6340e-05, 1.7711e-05,\n 2.3192e-05, 1.6961e-05, 1.0040e-05, 1.9120e-05, 1.5887e-05, 1.3441e-05,\n 8.4611e-06, 2.5195e-05, 1.2720e-05, 1.2159e-05, 1.5529e-05, 1.2541e-05,\n 1.7435e-05, 1.3409e-05, 2.9688e-05, 1.6050e-05, 1.1144e-05, 1.3692e-05,\n 1.5522e-05, 1.1337e-05, 1.1553e-05, 1.9779e-05, 1.4221e-05, 1.5446e-05,\n 1.6541e-05, 1.6422e-05, 1.3425e-05, 1.6321e-05, 1.0315e-05, 1.3648e-05,\n 1.1016e-05, 8.7994e-06, 1.5007e-05, 1.5491e-05, 9.6725e-06, 2.2158e-05,\n 1.2618e-05, 1.3238e-05, 1.3644e-05, 1.4214e-05, 3.8404e-04, 1.2813e-05,\n 1.3590e-05, 1.2847e-05, 1.5442e-05, 1.8541e-05, 1.6270e-05, 1.3929e-05,\n 1.1107e-05, 1.3278e-05, 1.2410e-05, 1.3600e-05, 1.4693e-05, 1.3768e-05,\n 1.1897e-05, 1.2175e-05, 1.2505e-05, 1.4123e-05, 1.5926e-05, 1.8030e-05,\n 1.3963e-05, 1.5550e-05], device='cuda:0')"
|
| 19 |
+
},
|
| 20 |
+
"3": {
|
| 21 |
+
"step": "tensor(1252.)",
|
| 22 |
+
"exp_avg": "tensor([-9.1585e-05, 1.1332e-03, 9.7939e-04, -1.2086e-03, -1.0672e-04,\n -2.4698e-03, 1.3611e-03, 2.1858e-03, -3.3356e-04, -1.3986e-03,\n -8.6483e-04, -1.2061e-03, -1.7172e-03, 6.3638e-04, -1.8465e-04,\n 2.7700e-03, 1.8900e-03, 2.1800e-03, -1.0895e-03, 1.7343e-03,\n 7.7181e-04, -8.3755e-04, -9.0681e-05, 7.3535e-04, -4.3096e-04,\n 6.5944e-04, 7.5415e-04, -2.4865e-04, 1.2114e-03, -1.4416e-03,\n -8.7159e-04, 4.0565e-04, 5.8238e-04, 1.3210e-03, 1.2765e-03,\n 2.7150e-03, 1.1751e-03, 2.0125e-03, -1.0056e-03, -9.3896e-04,\n 2.4877e-03, 4.4789e-04, -2.7485e-04, 9.3235e-04, 4.6253e-05,\n 3.9892e-04, -1.0113e-04, 5.7069e-05, -2.5275e-03, 4.1161e-04,\n 2.5201e-04, 1.2399e-04, -5.0457e-04, -2.3190e-03, -8.9405e-05,\n 2.2623e-03, -2.6905e-04, -2.9603e-04, 1.6576e-03, -3.4013e-03,\n -1.4178e-04, -1.0155e-03, 9.4792e-04, 3.2272e-04, 1.5775e-03,\n 1.1654e-03, 2.4701e-04, 2.4068e-03, -1.2965e-03, -9.5364e-04,\n 3.3126e-03, -1.5068e-03, -3.1931e-03, -1.1673e-03, 2.0102e-03,\n 3.9089e-03, 4.0825e-03, 3.4887e-03, -2.3607e-03, -2.4147e-03,\n 3.3310e-04, 3.9442e-03, 2.0756e-04, 2.5574e-03, 2.1594e-03,\n 2.1135e-03, -9.1363e-04, 2.1583e-03, -1.3370e-04, 2.4987e-03,\n 4.8433e-04, 2.2620e-03, 1.2528e-03, -6.7973e-04, -6.7316e-04,\n 5.3079e-04, -6.7459e-04, 1.7386e-03, 2.3673e-03, -1.0176e-03,\n -2.5107e-03, 2.0834e-03, 1.0706e-03, 1.7935e-04, -1.4112e-03,\n 8.1702e-04, -1.7935e-03, 7.9942e-04, -7.3539e-04, -1.2965e-03,\n 1.8050e-03, 9.2076e-04, -6.7511e-04, -2.1321e-04, -1.5112e-03,\n -1.4267e-03, -6.9700e-04, 1.8177e-04, 1.7887e-03, 5.3589e-04,\n 1.0790e-03, 1.9937e-03, 2.0729e-03, 3.4663e-04, 4.1018e-04,\n -6.6374e-04, 1.1277e-03, 3.8300e-04, -1.0546e-03, -2.3117e-03,\n 5.7523e-04, 1.2418e-03, 7.3759e-04, -1.2241e-04, 5.5268e-04,\n -2.1138e-04, 6.8918e-04, 2.2883e-03, -1.8951e-03, 1.4327e-03,\n 4.4106e-04, 5.2004e-04, 3.9564e-04, 2.7370e-03, -5.8012e-04,\n 1.4457e-03, -1.0211e-03, 2.3922e-03, 2.5492e-03, -4.4268e-05,\n 1.9679e-03, -8.6613e-04, -1.2325e-03, 4.4315e-04, 1.7422e-03,\n 4.4284e-04, -4.2864e-05, -1.3128e-04, -9.6661e-05, 3.2240e-04,\n 5.6509e-05, -1.6175e-03, -2.9765e-03, 4.2194e-04, 3.3982e-04,\n -1.4111e-03, 2.9816e-04, -1.6992e-03, -1.0327e-03, -1.7044e-03,\n -1.6254e-03, -1.3002e-03, -1.1470e-03, 2.8855e-03, -1.9737e-04,\n 1.7830e-03, 2.0100e-03, 3.8513e-04, 9.5911e-04, -6.1874e-04,\n -8.4439e-04, 1.0947e-03, 1.9271e-03, -2.1512e-04, -8.6921e-05,\n -8.6640e-04, -3.8005e-04, 2.2826e-04, -3.1726e-04, 6.5968e-04,\n -2.2230e-03, -9.1058e-04, -1.6599e-03, -2.1856e-03, 1.1439e-03,\n -4.4276e-04, 2.8115e-03, 3.6695e-04, 3.2982e-04, 9.8660e-04,\n 1.1180e-03, 4.6605e-04, 1.4253e-03, -4.2639e-04, -7.3634e-04,\n 3.2567e-03, -2.7501e-04, 9.8736e-04, 2.2923e-03, 3.9972e-04,\n 9.2131e-04, 1.6063e-03, -2.5182e-03, 1.6209e-04, 2.1743e-03,\n 1.1215e-03, 1.5371e-03, 2.9958e-03, -6.0116e-04, -1.3817e-03,\n -1.0561e-03, -3.7988e-05, 1.3565e-03, 9.9692e-04, 1.6378e-04,\n 2.9194e-03, 5.2716e-05, 1.0722e-03, 2.1581e-03, -2.4252e-03,\n -1.1639e-03, 9.3229e-04, -2.2170e-03, 1.1492e-03, -9.4675e-04,\n 1.2935e-03, -1.9219e-03, 7.2945e-05, 4.7227e-04, 4.3796e-05,\n 2.1640e-03, -9.7744e-04, 1.4534e-04, 7.4135e-04, -1.9102e-03,\n 2.1200e-03, 6.8058e-04, -8.4371e-05, 3.4624e-04, 1.6477e-03,\n -2.9572e-04, -1.5844e-03, -5.7856e-06, 1.4648e-04, 1.5125e-03,\n -1.9937e-04, 3.1988e-04, -4.1799e-04, -5.7315e-04, 6.9034e-04,\n -2.9521e-04, 2.0336e-03, -1.3513e-03, -4.9565e-04, 6.7006e-04,\n 1.8487e-03, -1.2223e-03, 1.7272e-03, 8.2190e-04, -1.2135e-03,\n 4.8972e-05, 8.8882e-04, -1.4034e-03, -2.3506e-03, 3.3288e-03,\n 1.1042e-03, 1.4838e-03, 2.2774e-03, -2.9216e-04, -2.0104e-04,\n -1.6795e-03, 7.2790e-04, 9.0919e-04, -3.2460e-03, 1.5348e-05,\n -5.3130e-04, 6.5279e-04, 4.6950e-04, -1.3080e-03, 2.2047e-04,\n -3.1240e-04, 7.7168e-04, 1.2426e-04, -2.1938e-04, 9.3674e-04,\n 5.6179e-04, 1.4262e-03, 2.0088e-04, 1.8372e-03, -7.9571e-04,\n -1.5437e-03, -6.9886e-04, 2.8067e-04, 1.3455e-03, -7.8558e-04,\n -4.7589e-04, 1.8021e-04, -7.9818e-04, -1.2685e-03, 2.7677e-03,\n 1.4205e-04, 5.4675e-04, 7.1274e-04, 3.4187e-04, -7.9863e-04,\n -6.7120e-04, 1.7778e-03, 2.3001e-03, -4.5745e-03, 2.1383e-03,\n 6.6218e-04, -2.6596e-04, -6.7287e-04, 2.2289e-03, -6.3910e-04,\n -2.5973e-03, -8.4345e-04, 1.7870e-03, 8.1605e-04, -1.8826e-03,\n -1.1721e-03, -1.2290e-03, 7.6908e-04, 8.8100e-04, -1.7431e-03,\n -9.4580e-04, -9.7469e-04, 1.0084e-03, 8.3952e-04, 3.8487e-03,\n -7.1312e-04, -3.3964e-03, -3.3721e-03, -7.5933e-04, 1.2488e-03,\n 6.6832e-04, -1.1821e-05, -5.8494e-04, 2.0349e-03, -3.2997e-03,\n 6.8437e-04, -8.4105e-04, -7.5562e-05, 3.4691e-03, -2.2931e-03,\n -1.2544e-03, 6.0755e-06, -4.7812e-04, 1.1839e-03, 3.3491e-03,\n 1.4714e-04, 2.3325e-03, -2.5985e-03, 2.3615e-04, 6.3248e-05,\n 6.6388e-04, 1.0878e-03, 2.4027e-03, 3.1039e-04, -1.5515e-03,\n 1.9088e-03, 2.1022e-03, 2.1342e-03, 1.1624e-03, -1.8467e-03,\n -6.1224e-04, 2.4224e-04, -1.9942e-03, 1.2179e-03, 1.1398e-03,\n -6.1457e-04, 1.5118e-03, 8.9202e-04, 2.8121e-04, 9.4640e-04,\n 1.7817e-03, 1.5262e-03, 2.3263e-03, 1.9718e-04, 1.3221e-03,\n 2.0269e-03, -1.3136e-03, -8.3033e-04, 9.2672e-04, 2.3848e-03,\n 7.0495e-05, 1.2734e-03, 1.4875e-03, 3.9665e-04, -6.2198e-05,\n 1.7494e-03, -1.6174e-03, 9.1990e-04, -1.9851e-04, 6.4918e-04,\n -7.3439e-04, 2.4284e-04, 5.0150e-05, -3.4306e-03, 5.9497e-04,\n 1.9796e-03, 1.3361e-03, -8.7319e-04, -1.6728e-03, -1.1518e-04,\n -1.6199e-03, 6.1890e-04, 1.1492e-03, -8.3870e-04, 6.8196e-05,\n -4.0310e-04, 4.6717e-04, -1.0858e-04, 1.6693e-03, -3.1539e-04,\n -3.3175e-03, 9.7120e-05, -1.2167e-03, -1.0766e-04, -1.3886e-03,\n -1.6084e-03, -2.1178e-03, -6.6602e-04, 5.7403e-05, 2.4398e-04,\n 2.2251e-03, 5.5769e-04, 2.2275e-04, -3.1229e-04, 5.9977e-04,\n 1.4908e-03, 3.7770e-04, 4.6450e-04, 6.8836e-05, -1.0622e-03,\n -7.5532e-04, 2.8235e-03, -7.8280e-04, -1.1077e-03, 1.0267e-04,\n 1.0818e-04, -3.7749e-03, 1.3008e-03, -3.9134e-04, 2.8943e-04,\n -2.7698e-03, 2.1895e-04, 3.9902e-05, -1.4756e-03, 4.9292e-04,\n -1.7255e-03, 4.3032e-04, 2.6125e-03, -6.7220e-04, 8.9875e-04,\n 2.8553e-03, -7.4467e-04, 6.9790e-04, 3.8627e-04, 6.2909e-04,\n 1.1496e-06, 1.2740e-03, 1.3956e-03, 1.1863e-03, -1.8118e-03,\n 1.8963e-03, -2.1000e-05, -1.8673e-03, -7.2928e-04, -1.6424e-03,\n 2.9304e-04, 9.0124e-04, -1.0360e-03, 5.7687e-04, 5.0034e-04,\n -4.0264e-03, 2.7492e-03, -1.3452e-04, -9.2723e-04, 2.3310e-03,\n -1.0273e-04, -1.0043e-03, 2.3106e-03, 2.3348e-03, 1.1878e-04,\n 1.2220e-03, 6.7277e-05, 9.7099e-04, -9.7045e-04, 7.2235e-04,\n 1.8983e-04, -8.5336e-05, 3.1176e-03, 1.5648e-04, -1.0431e-03,\n -9.5870e-04, -2.3945e-03, 1.2914e-03, -1.0509e-03, 1.7300e-03,\n -3.3704e-04, -4.8447e-04], device='cuda:0')",
|
| 23 |
+
"exp_avg_sq": "tensor([1.0795e-05, 1.0012e-05, 9.6583e-06, 1.1745e-05, 7.8265e-06, 1.1078e-05,\n 1.0456e-05, 1.2129e-05, 7.7046e-06, 1.1181e-05, 1.1360e-05, 1.1477e-05,\n 1.1925e-05, 1.1081e-05, 8.9359e-06, 1.1221e-05, 1.4703e-05, 1.1394e-05,\n 1.0367e-05, 1.0738e-05, 1.2098e-05, 1.1153e-05, 1.2801e-05, 1.1874e-05,\n 9.9593e-06, 1.3892e-05, 1.2849e-05, 1.0381e-05, 7.7603e-06, 1.4134e-05,\n 1.0374e-05, 1.0829e-05, 1.2650e-05, 1.0726e-05, 1.0007e-05, 1.2076e-05,\n 1.1007e-05, 1.1157e-05, 1.0532e-05, 1.1056e-05, 1.2372e-05, 1.1363e-05,\n 9.6217e-06, 1.0948e-05, 1.0433e-05, 7.7508e-06, 1.1714e-05, 1.0109e-05,\n 1.1892e-05, 1.1062e-05, 1.0755e-05, 1.2017e-05, 1.0794e-05, 1.1399e-05,\n 1.2319e-05, 1.0249e-05, 1.2960e-05, 1.0735e-05, 1.1397e-05, 1.0663e-05,\n 1.1788e-05, 8.5892e-06, 1.1617e-05, 1.2041e-05, 1.1029e-05, 1.0909e-05,\n 1.2627e-05, 1.1846e-05, 9.2721e-06, 1.1512e-05, 1.0266e-05, 9.4971e-06,\n 1.1275e-05, 1.0575e-05, 1.4620e-05, 9.9334e-06, 1.2652e-05, 1.1074e-05,\n 1.1615e-05, 1.0202e-05, 1.1223e-05, 1.2093e-05, 1.0377e-05, 1.0649e-05,\n 1.1974e-05, 1.2673e-05, 1.2790e-05, 1.1003e-05, 9.6634e-06, 1.2021e-05,\n 1.1889e-05, 1.3597e-05, 1.0453e-05, 1.0822e-05, 9.9528e-06, 1.1290e-05,\n 1.1755e-05, 1.1231e-05, 9.3235e-06, 9.8829e-06, 9.7179e-06, 9.3812e-06,\n 1.0684e-05, 1.1620e-05, 1.2779e-05, 1.2239e-05, 9.3779e-06, 9.4818e-06,\n 9.5329e-06, 9.9162e-06, 1.1961e-05, 9.4019e-06, 1.2597e-05, 9.8471e-06,\n 1.3383e-05, 1.2668e-05, 9.3304e-06, 1.2571e-05, 1.4336e-05, 9.1924e-06,\n 9.9052e-06, 1.2413e-05, 9.8720e-06, 1.1805e-05, 1.0941e-05, 1.1065e-05,\n 1.0410e-05, 1.1208e-05, 1.1631e-05, 1.2449e-05, 1.1858e-05, 1.0095e-05,\n 1.2855e-05, 1.0905e-05, 1.2131e-05, 1.1746e-05, 1.0236e-05, 1.1764e-05,\n 1.2565e-05, 1.1167e-05, 1.1129e-05, 1.1758e-05, 1.0446e-05, 1.4084e-05,\n 9.5178e-06, 1.2290e-05, 1.1181e-05, 1.3119e-05, 1.0382e-05, 1.3226e-05,\n 7.9014e-06, 1.1085e-05, 9.7762e-06, 1.3284e-05, 1.1359e-05, 1.3060e-05,\n 1.0142e-05, 1.1755e-05, 1.0694e-05, 9.2917e-06, 9.5108e-06, 1.1697e-05,\n 1.2289e-05, 1.0187e-05, 1.1490e-05, 1.1931e-05, 1.2068e-05, 1.1466e-05,\n 7.3045e-06, 1.1635e-05, 1.1697e-05, 1.0312e-05, 1.0786e-05, 1.0087e-05,\n 1.1452e-05, 1.1011e-05, 1.0356e-05, 1.2060e-05, 1.0297e-05, 8.8117e-06,\n 1.1079e-05, 1.2266e-05, 1.2343e-05, 1.0972e-05, 9.8724e-06, 1.0211e-05,\n 1.1667e-05, 1.1847e-05, 1.1350e-05, 1.1123e-05, 1.1703e-05, 1.1086e-05,\n 9.5132e-06, 1.0629e-05, 1.2502e-05, 1.1487e-05, 1.0780e-05, 8.5968e-06,\n 1.0071e-05, 1.2748e-05, 1.1865e-05, 1.0936e-05, 8.9485e-06, 1.1130e-05,\n 1.1887e-05, 1.1111e-05, 1.1514e-05, 1.1400e-05, 1.1789e-05, 1.3403e-05,\n 1.1596e-05, 1.0587e-05, 9.1720e-06, 1.0927e-05, 1.0284e-05, 1.0139e-05,\n 1.1575e-05, 9.8215e-06, 1.0155e-05, 1.1788e-05, 1.3018e-05, 1.1298e-05,\n 9.9689e-06, 9.5959e-06, 1.1818e-05, 1.2138e-05, 1.0185e-05, 1.0239e-05,\n 1.3758e-05, 1.0983e-05, 1.2174e-05, 9.0883e-06, 1.0133e-05, 1.0589e-05,\n 1.2786e-05, 1.0433e-05, 1.3180e-05, 1.0810e-05, 1.2248e-05, 1.0800e-05,\n 1.0252e-05, 1.2260e-05, 1.1949e-05, 1.2261e-05, 1.0770e-05, 1.0369e-05,\n 1.0123e-05, 8.0567e-06, 1.1446e-05, 9.8291e-06, 1.1989e-05, 1.1167e-05,\n 1.1026e-05, 8.8999e-06, 1.1207e-05, 9.0883e-06, 1.2500e-05, 1.2310e-05,\n 1.0515e-05, 1.2703e-05, 1.0191e-05, 1.0616e-05, 1.1934e-05, 1.0641e-05,\n 1.2920e-05, 1.1971e-05, 1.0688e-05, 1.1268e-05, 1.2863e-05, 1.1371e-05,\n 1.0936e-05, 9.5446e-06, 9.5933e-06, 8.7556e-06, 1.3267e-05, 9.2729e-06,\n 1.0550e-05, 1.0893e-05, 1.0416e-05, 1.0123e-05, 8.7055e-06, 1.0580e-05,\n 1.0412e-05, 1.3035e-05, 9.6488e-06, 1.0526e-05, 1.2297e-05, 9.9107e-06,\n 1.0156e-05, 1.2195e-05, 1.0828e-05, 1.1336e-05, 1.1999e-05, 1.2058e-05,\n 1.2682e-05, 1.1678e-05, 8.9235e-06, 1.1232e-05, 1.2473e-05, 1.2035e-05,\n 1.0858e-05, 1.1395e-05, 1.0253e-05, 1.2947e-05, 1.2822e-05, 1.2411e-05,\n 9.6429e-06, 1.1010e-05, 1.0843e-05, 1.0419e-05, 9.5663e-06, 1.1075e-05,\n 1.1684e-05, 1.2460e-05, 1.2012e-05, 1.1492e-05, 1.3568e-05, 1.1657e-05,\n 9.5094e-06, 1.3646e-05, 9.5277e-06, 9.9370e-06, 1.3656e-05, 1.3244e-05,\n 9.1914e-06, 1.0642e-05, 1.1022e-05, 1.0452e-05, 1.0220e-05, 9.0775e-06,\n 1.0853e-05, 1.1179e-05, 9.9287e-06, 1.1139e-05, 1.2788e-05, 1.2357e-05,\n 1.0452e-05, 1.3402e-05, 1.1010e-05, 1.1414e-05, 1.0756e-05, 1.3289e-05,\n 1.3187e-05, 1.2483e-05, 1.1063e-05, 1.2226e-05, 1.1448e-05, 1.0203e-05,\n 1.2154e-05, 6.8173e-06, 9.7099e-06, 9.4322e-06, 8.6864e-06, 1.1689e-05,\n 1.2863e-05, 1.1247e-05, 9.8324e-06, 9.3242e-06, 1.3026e-05, 1.1466e-05,\n 1.2076e-05, 7.9145e-06, 1.0294e-05, 1.0800e-05, 9.2742e-06, 1.3294e-05,\n 1.0003e-05, 1.2424e-05, 1.1405e-05, 1.3138e-05, 1.2371e-05, 1.1686e-05,\n 1.1936e-05, 1.0480e-05, 9.8413e-06, 1.2286e-05, 1.0985e-05, 1.1298e-05,\n 1.0985e-05, 1.1286e-05, 9.5373e-06, 1.0118e-05, 1.2113e-05, 1.0380e-05,\n 8.9639e-06, 1.1535e-05, 1.0227e-05, 1.0218e-05, 1.0998e-05, 1.2996e-05,\n 1.1013e-05, 8.5467e-06, 1.2080e-05, 1.0531e-05, 1.0375e-05, 1.0190e-05,\n 1.0857e-05, 1.1478e-05, 9.2783e-06, 1.2145e-05, 1.1333e-05, 1.1253e-05,\n 9.6089e-06, 1.0202e-05, 8.6623e-06, 9.6052e-06, 1.1604e-05, 9.6260e-06,\n 1.2158e-05, 1.1071e-05, 1.1351e-05, 1.0076e-05, 1.0116e-05, 1.4294e-05,\n 9.8832e-06, 1.2945e-05, 9.4776e-06, 1.1117e-05, 1.1195e-05, 1.0138e-05,\n 1.2270e-05, 1.1933e-05, 1.0769e-05, 9.9122e-06, 1.0319e-05, 1.1777e-05,\n 8.6559e-06, 8.5890e-06, 1.1941e-05, 1.2873e-05, 9.8630e-06, 1.2956e-05,\n 1.0796e-05, 9.3667e-06, 1.0608e-05, 1.1236e-05, 9.8122e-06, 1.1897e-05,\n 1.2324e-05, 8.9288e-06, 1.2303e-05, 1.2795e-05, 1.2678e-05, 1.0894e-05,\n 1.0221e-05, 1.1721e-05, 9.0089e-06, 1.0980e-05, 1.2045e-05, 1.3118e-05,\n 1.1989e-05, 1.1357e-05, 8.5664e-06, 1.1503e-05, 1.1855e-05, 1.2041e-05,\n 9.0023e-06, 1.2747e-05, 1.0162e-05, 1.0271e-05, 1.3007e-05, 1.2070e-05,\n 1.1809e-05, 1.1116e-05, 1.1380e-05, 1.2563e-05, 9.4793e-06, 1.1528e-05,\n 1.1031e-05, 9.4141e-06, 1.0349e-05, 1.3281e-05, 1.2710e-05, 1.0500e-05,\n 1.1287e-05, 1.1143e-05, 1.1350e-05, 1.1696e-05, 7.2177e-06, 1.0816e-05,\n 1.0497e-05, 9.7584e-06, 1.1136e-05, 1.0880e-05, 8.8688e-06, 1.3892e-05,\n 1.0454e-05, 1.1472e-05, 1.0362e-05, 1.1427e-05, 1.2705e-05, 1.1663e-05,\n 1.1852e-05, 1.1660e-05, 1.3704e-05, 1.3051e-05, 1.0876e-05, 1.1115e-05,\n 9.3744e-06, 1.0491e-05, 1.1086e-05, 1.0477e-05, 1.0511e-05, 1.0559e-05,\n 1.1001e-05, 1.0758e-05, 1.0928e-05, 1.0659e-05, 1.1193e-05, 1.1553e-05,\n 1.0778e-05, 1.2104e-05], device='cuda:0')"
|
| 24 |
+
},
|
| 25 |
+
"4": {
|
| 26 |
+
"step": "tensor(1252.)",
|
| 27 |
+
"exp_avg": "tensor([[ 3.4858e-04, 5.9108e-05, 8.6917e-05, ..., 1.0962e-04,\n -3.9781e-05, -7.7514e-05],\n [-7.7966e-04, -4.8924e-04, -4.9839e-04, ..., -5.8629e-04,\n 7.0869e-05, -2.8217e-04],\n [ 1.3609e-04, 2.2269e-04, -5.4436e-04, ..., 2.3072e-04,\n 6.3068e-05, 2.5053e-04],\n ...,\n [ 3.9298e-04, 2.7288e-04, 8.9543e-04, ..., -1.4091e-04,\n -1.7561e-04, -9.9732e-04],\n [-2.0894e-04, 1.0109e-04, -5.0877e-05, ..., -4.3691e-04,\n -4.2677e-04, -4.7517e-04],\n [ 1.2677e-04, -8.6394e-05, 3.3509e-04, ..., 1.7947e-05,\n 5.6016e-05, -1.2058e-03]], device='cuda:0')",
|
| 28 |
+
"exp_avg_sq": "tensor([[2.0854e-06, 1.1253e-06, 1.7678e-06, ..., 1.7276e-06, 1.9579e-06,\n 2.1001e-06],\n [3.6910e-06, 2.3843e-06, 3.6250e-06, ..., 3.9738e-06, 3.8502e-06,\n 3.3890e-06],\n [3.6743e-06, 1.9672e-06, 3.2226e-06, ..., 4.0701e-06, 3.3827e-06,\n 3.1446e-06],\n ...,\n [4.1163e-06, 2.6482e-06, 3.8001e-06, ..., 4.1236e-06, 4.1392e-06,\n 3.7272e-06],\n [3.5845e-06, 2.7213e-06, 3.1363e-06, ..., 4.8903e-06, 3.7485e-06,\n 3.6860e-06],\n [3.8560e-06, 1.9685e-06, 3.3872e-06, ..., 4.4567e-06, 3.6900e-06,\n 3.8294e-06]], device='cuda:0')"
|
| 29 |
+
},
|
| 30 |
+
"5": {
|
| 31 |
+
"step": "tensor(1252.)",
|
| 32 |
+
"exp_avg": "tensor([[ 3.1150e-04, -1.4786e-05, 5.4503e-05, ..., 1.9095e-04,\n 8.9207e-05, 1.6761e-05],\n [-3.6662e-04, -2.3368e-04, -1.3012e-04, ..., -5.3095e-05,\n 2.3148e-05, 5.9998e-05],\n [-2.4919e-05, 2.1884e-04, -3.3732e-04, ..., 5.8781e-05,\n 9.2706e-05, 7.5446e-05],\n ...,\n [-6.4370e-05, -1.6774e-04, 1.1126e-04, ..., 1.3579e-04,\n -8.0765e-05, -5.5162e-04],\n [ 2.1112e-04, 1.1904e-04, 2.6423e-04, ..., 3.8537e-05,\n 2.4745e-04, -1.2132e-04],\n [-1.1308e-04, 2.5055e-04, 1.2282e-04, ..., -6.5947e-04,\n 4.1654e-04, 4.6674e-04]], device='cuda:0')",
|
| 33 |
+
"exp_avg_sq": "tensor([[5.8788e-07, 2.6556e-07, 5.2166e-07, ..., 6.5772e-07, 5.3903e-07,\n 5.4572e-07],\n [1.0926e-06, 6.9036e-07, 1.0059e-06, ..., 9.4499e-07, 9.3674e-07,\n 9.2468e-07],\n [1.0850e-06, 7.0075e-07, 9.7682e-07, ..., 1.2583e-06, 1.0594e-06,\n 8.8148e-07],\n ...,\n [1.0090e-06, 7.3202e-07, 9.6877e-07, ..., 1.3070e-06, 1.3088e-06,\n 1.1241e-06],\n [1.2609e-06, 7.6325e-07, 1.0440e-06, ..., 1.2580e-06, 1.1442e-06,\n 1.0319e-06],\n [1.2104e-06, 5.0362e-07, 1.1328e-06, ..., 1.1233e-06, 1.0607e-06,\n 1.1244e-06]], device='cuda:0')"
|
| 34 |
+
},
|
| 35 |
+
"6": {
|
| 36 |
+
"step": "tensor(1252.)",
|
| 37 |
+
"exp_avg": "tensor([-4.0188e-05, 4.0209e-05], device='cuda:0')",
|
| 38 |
+
"exp_avg_sq": "tensor([2.3241e-05, 2.3241e-05], device='cuda:0')"
|
| 39 |
+
}
|
| 40 |
+
},
|
| 41 |
+
"param_groups": [
|
| 42 |
+
{
|
| 43 |
+
"lr": 0.0009755527298894294,
|
| 44 |
+
"name": "shared",
|
| 45 |
+
"betas": [
|
| 46 |
+
0.9,
|
| 47 |
+
0.999
|
| 48 |
+
],
|
| 49 |
+
"eps": 1e-08,
|
| 50 |
+
"weight_decay": 1e-05,
|
| 51 |
+
"amsgrad": false,
|
| 52 |
+
"maximize": false,
|
| 53 |
+
"foreach": null,
|
| 54 |
+
"capturable": false,
|
| 55 |
+
"differentiable": false,
|
| 56 |
+
"fused": null,
|
| 57 |
+
"decoupled_weight_decay": true,
|
| 58 |
+
"initial_lr": 0.001,
|
| 59 |
+
"params": [
|
| 60 |
+
0,
|
| 61 |
+
1,
|
| 62 |
+
2,
|
| 63 |
+
3
|
| 64 |
+
]
|
| 65 |
+
},
|
| 66 |
+
{
|
| 67 |
+
"lr": 0.0009755527298894294,
|
| 68 |
+
"name": "scale_256",
|
| 69 |
+
"betas": [
|
| 70 |
+
0.9,
|
| 71 |
+
0.999
|
| 72 |
+
],
|
| 73 |
+
"eps": 1e-08,
|
| 74 |
+
"weight_decay": 1e-05,
|
| 75 |
+
"amsgrad": false,
|
| 76 |
+
"maximize": false,
|
| 77 |
+
"foreach": null,
|
| 78 |
+
"capturable": false,
|
| 79 |
+
"differentiable": false,
|
| 80 |
+
"fused": null,
|
| 81 |
+
"decoupled_weight_decay": true,
|
| 82 |
+
"initial_lr": 0.001,
|
| 83 |
+
"params": [
|
| 84 |
+
4
|
| 85 |
+
]
|
| 86 |
+
},
|
| 87 |
+
{
|
| 88 |
+
"lr": 0.0009755527298894294,
|
| 89 |
+
"name": "scale_512",
|
| 90 |
+
"betas": [
|
| 91 |
+
0.9,
|
| 92 |
+
0.999
|
| 93 |
+
],
|
| 94 |
+
"eps": 1e-08,
|
| 95 |
+
"weight_decay": 1e-05,
|
| 96 |
+
"amsgrad": false,
|
| 97 |
+
"maximize": false,
|
| 98 |
+
"foreach": null,
|
| 99 |
+
"capturable": false,
|
| 100 |
+
"differentiable": false,
|
| 101 |
+
"fused": null,
|
| 102 |
+
"decoupled_weight_decay": true,
|
| 103 |
+
"initial_lr": 0.001,
|
| 104 |
+
"params": [
|
| 105 |
+
5
|
| 106 |
+
]
|
| 107 |
+
},
|
| 108 |
+
{
|
| 109 |
+
"lr": 0.00048778860081564085,
|
| 110 |
+
"name": "fusion",
|
| 111 |
+
"betas": [
|
| 112 |
+
0.9,
|
| 113 |
+
0.999
|
| 114 |
+
],
|
| 115 |
+
"eps": 1e-08,
|
| 116 |
+
"weight_decay": 1e-05,
|
| 117 |
+
"amsgrad": false,
|
| 118 |
+
"maximize": false,
|
| 119 |
+
"foreach": null,
|
| 120 |
+
"capturable": false,
|
| 121 |
+
"differentiable": false,
|
| 122 |
+
"fused": null,
|
| 123 |
+
"decoupled_weight_decay": true,
|
| 124 |
+
"initial_lr": 0.0005,
|
| 125 |
+
"params": [
|
| 126 |
+
6
|
| 127 |
+
]
|
| 128 |
+
}
|
| 129 |
+
]
|
| 130 |
+
},
|
| 131 |
+
"scheduler_state_dict": {
|
| 132 |
+
"T_0": 10,
|
| 133 |
+
"T_i": 10,
|
| 134 |
+
"T_mult": 2,
|
| 135 |
+
"eta_min": 1e-06,
|
| 136 |
+
"T_cur": 1,
|
| 137 |
+
"base_lrs": [
|
| 138 |
+
0.001,
|
| 139 |
+
0.001,
|
| 140 |
+
0.001,
|
| 141 |
+
0.0005
|
| 142 |
+
],
|
| 143 |
+
"last_epoch": 1,
|
| 144 |
+
"_step_count": 0,
|
| 145 |
+
"_is_initial": false,
|
| 146 |
+
"_get_lr_called_within_step": false,
|
| 147 |
+
"_last_lr": [
|
| 148 |
+
0.0009755527298894294,
|
| 149 |
+
0.0009755527298894294,
|
| 150 |
+
0.0009755527298894294,
|
| 151 |
+
0.00048778860081564085
|
| 152 |
+
]
|
| 153 |
+
},
|
| 154 |
+
"metrics": {
|
| 155 |
+
"best_val_acc": 68.056,
|
| 156 |
+
"best_epoch": 0,
|
| 157 |
+
"scale_accuracies": {
|
| 158 |
+
"256": 66.924,
|
| 159 |
+
"512": 67.882
|
| 160 |
+
},
|
| 161 |
+
"training_history": {
|
| 162 |
+
"epochs": [
|
| 163 |
+
1
|
| 164 |
+
],
|
| 165 |
+
"train_loss": [
|
| 166 |
+
5.603825211905824
|
| 167 |
+
],
|
| 168 |
+
"train_acc": [
|
| 169 |
+
63.39829233815732
|
| 170 |
+
],
|
| 171 |
+
"val_acc": [
|
| 172 |
+
68.056
|
| 173 |
+
],
|
| 174 |
+
"scale_accs": {
|
| 175 |
+
"256": [
|
| 176 |
+
66.924
|
| 177 |
+
],
|
| 178 |
+
"512": [
|
| 179 |
+
67.882
|
| 180 |
+
]
|
| 181 |
+
},
|
| 182 |
+
"lr": [
|
| 183 |
+
0.0009755527298894294
|
| 184 |
+
]
|
| 185 |
+
}
|
| 186 |
+
},
|
| 187 |
+
"train_config": {
|
| 188 |
+
"name": "david_training",
|
| 189 |
+
"run_id": "20251012_135249",
|
| 190 |
+
"dataset_name": "AbstractPhil/imagenet-clip-features-orderly",
|
| 191 |
+
"model_variant": "clip_vit_laion_b32",
|
| 192 |
+
"num_classes": 1000,
|
| 193 |
+
"preset": "small_fast",
|
| 194 |
+
"custom_config_path": null,
|
| 195 |
+
"num_classes_override": null,
|
| 196 |
+
"use_belly_override": null,
|
| 197 |
+
"belly_expand_override": null,
|
| 198 |
+
"progressive_training_override": false,
|
| 199 |
+
"scale_warmup_epochs_override": null,
|
| 200 |
+
"num_epochs": 10,
|
| 201 |
+
"batch_size": 1024,
|
| 202 |
+
"learning_rate": 0.001,
|
| 203 |
+
"weight_decay": 1e-05,
|
| 204 |
+
"warmup_epochs": 3,
|
| 205 |
+
"use_rose_loss": true,
|
| 206 |
+
"rose_initial_weight": 0.1,
|
| 207 |
+
"rose_max_weight": 0.5,
|
| 208 |
+
"rose_weight_schedule": "adaptive",
|
| 209 |
+
"use_cayley_loss": false,
|
| 210 |
+
"cayley_weight": 0.001,
|
| 211 |
+
"scale_loss_balance": null,
|
| 212 |
+
"use_mixed_precision": false,
|
| 213 |
+
"gradient_clip": 10.0,
|
| 214 |
+
"scheduler_type": "cosine_restarts",
|
| 215 |
+
"min_lr": 1e-06,
|
| 216 |
+
"freeze_strategy": "never",
|
| 217 |
+
"freeze_threshold": 90.0,
|
| 218 |
+
"unfreeze_on_plateau": true,
|
| 219 |
+
"patience": 10,
|
| 220 |
+
"track_gradients": true,
|
| 221 |
+
"gradient_scale_threshold": 1e-05,
|
| 222 |
+
"gradient_scale_multiplier": 10.0,
|
| 223 |
+
"log_interval": 50,
|
| 224 |
+
"val_interval": 1,
|
| 225 |
+
"save_interval": 5,
|
| 226 |
+
"log_fusion_weights": true,
|
| 227 |
+
"log_loss_components": true,
|
| 228 |
+
"save_format": "safetensors",
|
| 229 |
+
"hf_repo": "AbstractPhil/gated-david",
|
| 230 |
+
"upload_to_hub": true,
|
| 231 |
+
"base_dir": "./david_training",
|
| 232 |
+
"num_workers": 10,
|
| 233 |
+
"pin_memory": true,
|
| 234 |
+
"prefetch_factor": 4,
|
| 235 |
+
"persistent_workers": true
|
| 236 |
+
}
|
| 237 |
+
}
|
weights/David-fully_shared-weighted_sum/20251012_135249/david_config.json
ADDED
|
@@ -0,0 +1,29 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"name": "david_small_fast",
|
| 3 |
+
"uid": "c.david.small_fast",
|
| 4 |
+
"feature_dim": 512,
|
| 5 |
+
"num_classes": 1000,
|
| 6 |
+
"scales": [
|
| 7 |
+
256,
|
| 8 |
+
512
|
| 9 |
+
],
|
| 10 |
+
"sharing_mode": "fully_shared",
|
| 11 |
+
"fusion_mode": "weighted_sum",
|
| 12 |
+
"use_belly": false,
|
| 13 |
+
"belly_expand": 2.0,
|
| 14 |
+
"shared_feature_dim": 512,
|
| 15 |
+
"shared_layers": 1,
|
| 16 |
+
"shared_dropout": 0.1,
|
| 17 |
+
"fusion_temperature": 1.0,
|
| 18 |
+
"fusion_dropout": 0.1,
|
| 19 |
+
"tree_depth": 3,
|
| 20 |
+
"num_experts": 3,
|
| 21 |
+
"compression_ratio": 4,
|
| 22 |
+
"expert_dropout": 0.1,
|
| 23 |
+
"attention_dropout": 0.1,
|
| 24 |
+
"progressive_training": false,
|
| 25 |
+
"scale_warmup_epochs": {
|
| 26 |
+
"256": 0,
|
| 27 |
+
"512": 0
|
| 28 |
+
}
|
| 29 |
+
}
|
weights/David-fully_shared-weighted_sum/20251012_135249/train_config.json
ADDED
|
@@ -0,0 +1,50 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"name": "david_training",
|
| 3 |
+
"run_id": "20251012_135249",
|
| 4 |
+
"dataset_name": "AbstractPhil/imagenet-clip-features-orderly",
|
| 5 |
+
"model_variant": "clip_vit_laion_b32",
|
| 6 |
+
"num_classes": 1000,
|
| 7 |
+
"preset": "small_fast",
|
| 8 |
+
"custom_config_path": null,
|
| 9 |
+
"num_classes_override": null,
|
| 10 |
+
"use_belly_override": null,
|
| 11 |
+
"belly_expand_override": null,
|
| 12 |
+
"progressive_training_override": false,
|
| 13 |
+
"scale_warmup_epochs_override": null,
|
| 14 |
+
"num_epochs": 10,
|
| 15 |
+
"batch_size": 1024,
|
| 16 |
+
"learning_rate": 0.001,
|
| 17 |
+
"weight_decay": 1e-05,
|
| 18 |
+
"warmup_epochs": 3,
|
| 19 |
+
"use_rose_loss": true,
|
| 20 |
+
"rose_initial_weight": 0.1,
|
| 21 |
+
"rose_max_weight": 0.5,
|
| 22 |
+
"rose_weight_schedule": "adaptive",
|
| 23 |
+
"use_cayley_loss": false,
|
| 24 |
+
"cayley_weight": 0.001,
|
| 25 |
+
"scale_loss_balance": null,
|
| 26 |
+
"use_mixed_precision": false,
|
| 27 |
+
"gradient_clip": 10.0,
|
| 28 |
+
"scheduler_type": "cosine_restarts",
|
| 29 |
+
"min_lr": 1e-06,
|
| 30 |
+
"freeze_strategy": "never",
|
| 31 |
+
"freeze_threshold": 90.0,
|
| 32 |
+
"unfreeze_on_plateau": true,
|
| 33 |
+
"patience": 10,
|
| 34 |
+
"track_gradients": true,
|
| 35 |
+
"gradient_scale_threshold": 1e-05,
|
| 36 |
+
"gradient_scale_multiplier": 10.0,
|
| 37 |
+
"log_interval": 50,
|
| 38 |
+
"val_interval": 1,
|
| 39 |
+
"save_interval": 5,
|
| 40 |
+
"log_fusion_weights": true,
|
| 41 |
+
"log_loss_components": true,
|
| 42 |
+
"save_format": "safetensors",
|
| 43 |
+
"hf_repo": "AbstractPhil/gated-david",
|
| 44 |
+
"upload_to_hub": true,
|
| 45 |
+
"base_dir": "./david_training",
|
| 46 |
+
"num_workers": 10,
|
| 47 |
+
"pin_memory": true,
|
| 48 |
+
"prefetch_factor": 4,
|
| 49 |
+
"persistent_workers": true
|
| 50 |
+
}
|
weights/David-fully_shared-weighted_sum/20251012_135249/training_history.json
ADDED
|
@@ -0,0 +1,25 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
{
|
| 2 |
+
"epochs": [
|
| 3 |
+
1
|
| 4 |
+
],
|
| 5 |
+
"train_loss": [
|
| 6 |
+
5.603825211905824
|
| 7 |
+
],
|
| 8 |
+
"train_acc": [
|
| 9 |
+
63.39829233815732
|
| 10 |
+
],
|
| 11 |
+
"val_acc": [
|
| 12 |
+
68.056
|
| 13 |
+
],
|
| 14 |
+
"scale_accs": {
|
| 15 |
+
"256": [
|
| 16 |
+
66.924
|
| 17 |
+
],
|
| 18 |
+
"512": [
|
| 19 |
+
67.882
|
| 20 |
+
]
|
| 21 |
+
},
|
| 22 |
+
"lr": [
|
| 23 |
+
0.0009755527298894294
|
| 24 |
+
]
|
| 25 |
+
}
|