Upload weights and configs - Run 20251012_032356
Browse files
weights/best_model.safetensors
CHANGED
|
@@ -1,3 +1,3 @@
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
-
oid sha256:
|
| 3 |
size 59515088
|
|
|
|
| 1 |
version https://git-lfs.github.com/spec/v1
|
| 2 |
+
oid sha256:cc187ab3f8538420f8a3440dd736f259e0c66a6587c78a5bbd13699917e2d8d9
|
| 3 |
size 59515088
|
weights/best_model_metadata.json
CHANGED
|
@@ -1,26 +1,26 @@
|
|
| 1 |
{
|
| 2 |
-
"epoch":
|
| 3 |
"optimizer_state_dict": {
|
| 4 |
"state": {
|
| 5 |
"0": {
|
| 6 |
-
"step": "tensor(
|
| 7 |
-
"exp_avg": "tensor([[
|
| 8 |
-
"exp_avg_sq": "tensor([[
|
| 9 |
},
|
| 10 |
"1": {
|
| 11 |
-
"step": "tensor(
|
| 12 |
-
"exp_avg": "tensor([
|
| 13 |
-
"exp_avg_sq": "tensor([
|
| 14 |
},
|
| 15 |
"2": {
|
| 16 |
-
"step": "tensor(
|
| 17 |
-
"exp_avg": "tensor([[
|
| 18 |
-
"exp_avg_sq": "tensor([[
|
| 19 |
}
|
| 20 |
},
|
| 21 |
"param_groups": [
|
| 22 |
{
|
| 23 |
-
"lr": 0.
|
| 24 |
"name": "scale_256",
|
| 25 |
"betas": [
|
| 26 |
0.9,
|
|
@@ -43,7 +43,7 @@
|
|
| 43 |
]
|
| 44 |
},
|
| 45 |
{
|
| 46 |
-
"lr": 0.
|
| 47 |
"name": "scale_512",
|
| 48 |
"betas": [
|
| 49 |
0.9,
|
|
@@ -66,7 +66,7 @@
|
|
| 66 |
]
|
| 67 |
},
|
| 68 |
{
|
| 69 |
-
"lr": 0.
|
| 70 |
"name": "scale_768",
|
| 71 |
"betas": [
|
| 72 |
0.9,
|
|
@@ -89,7 +89,7 @@
|
|
| 89 |
]
|
| 90 |
},
|
| 91 |
{
|
| 92 |
-
"lr": 0.
|
| 93 |
"name": "scale_1024",
|
| 94 |
"betas": [
|
| 95 |
0.9,
|
|
@@ -112,7 +112,7 @@
|
|
| 112 |
]
|
| 113 |
},
|
| 114 |
{
|
| 115 |
-
"lr": 0.
|
| 116 |
"name": "scale_1280",
|
| 117 |
"betas": [
|
| 118 |
0.9,
|
|
@@ -135,7 +135,7 @@
|
|
| 135 |
]
|
| 136 |
},
|
| 137 |
{
|
| 138 |
-
"lr": 0.
|
| 139 |
"name": "fusion",
|
| 140 |
"betas": [
|
| 141 |
0.9,
|
|
@@ -195,7 +195,7 @@
|
|
| 195 |
"T_i": 10,
|
| 196 |
"T_mult": 2,
|
| 197 |
"eta_min": 1e-06,
|
| 198 |
-
"T_cur":
|
| 199 |
"base_lrs": [
|
| 200 |
0.01,
|
| 201 |
0.01,
|
|
@@ -204,24 +204,24 @@
|
|
| 204 |
0.01,
|
| 205 |
0.005
|
| 206 |
],
|
| 207 |
-
"last_epoch":
|
| 208 |
"_step_count": 0,
|
| 209 |
"_is_initial": false,
|
| 210 |
"_get_lr_called_within_step": false,
|
| 211 |
"_last_lr": [
|
| 212 |
-
0.
|
| 213 |
-
0.
|
| 214 |
-
0.
|
| 215 |
-
0.
|
| 216 |
-
0.
|
| 217 |
-
0.
|
| 218 |
]
|
| 219 |
},
|
| 220 |
"metrics": {
|
| 221 |
-
"best_val_acc":
|
| 222 |
-
"best_epoch":
|
| 223 |
"scale_accuracies": {
|
| 224 |
-
"256":
|
| 225 |
}
|
| 226 |
},
|
| 227 |
"train_config": {
|
|
|
|
| 1 |
{
|
| 2 |
+
"epoch": 1,
|
| 3 |
"optimizer_state_dict": {
|
| 4 |
"state": {
|
| 5 |
"0": {
|
| 6 |
+
"step": "tensor(2504.)",
|
| 7 |
+
"exp_avg": "tensor([[-2.9847e-05, -1.5561e-05, -3.2015e-05, ..., -3.4633e-05,\n 9.6170e-05, -4.8952e-05],\n [-5.1054e-05, -3.7014e-05, 2.2638e-05, ..., -1.1228e-04,\n -1.0936e-04, 3.1119e-05],\n [ 2.6958e-06, -1.2456e-04, 1.4858e-05, ..., -4.4155e-05,\n 6.4120e-05, 2.8139e-05],\n ...,\n [-5.4542e-05, 2.4455e-05, -1.6261e-05, ..., -3.6784e-05,\n 4.9569e-05, -5.8800e-05],\n [-4.4965e-23, 2.3582e-24, 2.9657e-23, ..., -1.0940e-23,\n 6.9415e-24, -2.3902e-23],\n [-5.8117e-05, 2.1739e-05, -4.2380e-05, ..., -5.3147e-05,\n 3.0465e-05, 8.0894e-05]], device='cuda:0')",
|
| 8 |
+
"exp_avg_sq": "tensor([[8.1019e-08, 1.0675e-07, 1.2200e-08, ..., 2.7520e-08, 2.9280e-08,\n 8.1349e-09],\n [2.0122e-07, 1.4826e-07, 4.4990e-08, ..., 8.7383e-08, 4.9043e-08,\n 6.1986e-08],\n [1.6635e-07, 2.2215e-07, 4.7403e-08, ..., 5.8297e-08, 3.8775e-08,\n 4.2303e-08],\n ...,\n [4.0419e-07, 2.7279e-07, 4.6405e-08, ..., 4.8686e-08, 4.1250e-08,\n 2.8963e-08],\n [6.1337e-11, 1.4214e-09, 2.2304e-10, ..., 9.4234e-11, 2.9866e-10,\n 1.7780e-10],\n [2.7880e-07, 1.7346e-07, 2.7628e-08, ..., 3.3208e-08, 6.5641e-08,\n 3.2027e-08]], device='cuda:0')"
|
| 9 |
},
|
| 10 |
"1": {
|
| 11 |
+
"step": "tensor(2504.)",
|
| 12 |
+
"exp_avg": "tensor([ 1.7719e-03, -2.7156e-03, 1.7750e-03, 6.0193e-04, -1.2922e-15,\n 2.1121e-04, 4.9933e-04, -2.6053e-03, 3.6154e-04, 7.6070e-04,\n 2.2209e-04, -3.5991e-03, 5.6052e-45, 2.6509e-04, 5.7084e-04,\n 1.4721e-05, -4.8943e-04, 4.4611e-03, 7.7828e-04, 5.6052e-45,\n 1.3613e-04, -3.5591e-04, 3.9319e-04, 5.0281e-04, 5.6052e-45,\n 5.8405e-04, -2.1400e-04, 2.3399e-03, 3.3233e-03, -1.1443e-03,\n -1.8175e-03, 1.7404e-03, -6.8076e-11, 5.6052e-45, -9.1142e-04,\n 1.8191e-04, -2.5487e-03, 2.2662e-03, 1.0145e-03, -1.1766e-03,\n -2.9495e-04, 3.0875e-03, 1.2050e-04, -6.2883e-03, -2.0681e-04,\n -1.0954e-03, -2.4520e-03, -7.5239e-05, 1.6260e-03, 1.2031e-03,\n -8.9359e-04, 5.6052e-45, 4.0436e-08, 5.6052e-45, 9.4041e-04,\n -8.9957e-03, 8.6776e-04, 1.6575e-03, 5.6052e-45, 2.1743e-03,\n -2.7675e-03, -1.0951e-03, -1.4413e-03, 2.6270e-03, 3.0545e-03,\n -5.3882e-04, -2.0457e-03, 5.6052e-45, -1.2317e-03, -2.6512e-03,\n -1.2740e-04, 5.6052e-45, 3.3783e-03, 8.1476e-04, 7.9997e-04,\n -1.2763e-05, 2.0720e-04, -1.9931e-03, -2.9498e-03, -4.0966e-05,\n 4.4024e-04, 2.7361e-04, -1.2984e-03, 2.0401e-03, 2.5797e-03,\n -1.8919e-04, -2.6561e-03, -2.7024e-03, 6.2633e-04, 2.1254e-03,\n 5.6052e-45, -3.3968e-03, -1.4859e-03, 3.0709e-04, 3.1062e-04,\n -1.3830e-03, -1.2511e-03, 1.5813e-04, 2.0279e-03, 1.6878e-03,\n 2.7028e-03, -1.5681e-05, 1.8112e-03, 2.0162e-03, 1.1955e-03,\n 3.8765e-04, -1.6990e-03, -3.1365e-04, -3.3308e-03, 6.5740e-04,\n 3.3836e-03, 1.7968e-03, 5.6052e-45, 9.1714e-04, -3.7225e-11,\n -1.5972e-03, 3.3398e-03, -4.4637e-04, -3.6734e-19, 1.9321e-06,\n 5.6052e-45, -1.6496e-03, 5.6052e-45, -4.3220e-04, -1.1750e-05,\n -2.6207e-04, 2.5364e-03, 5.6052e-45, 2.4874e-04, -3.8771e-04,\n 3.1076e-03, -9.5760e-04, 6.0460e-04, 1.1282e-19, -9.6913e-04,\n 1.8747e-03, -2.3665e-03, 5.5657e-03, 3.6068e-04, 8.4913e-05,\n -3.0548e-05, -1.7480e-03, -1.6406e-03, -8.9434e-04, -1.3305e-04,\n -3.8913e-03, -5.6052e-45, 2.3141e-03, 1.2096e-03, 1.1739e-03,\n 3.7447e-19, 3.2070e-04, -2.2073e-03, 5.6052e-45, 2.6995e-04,\n 2.6495e-03, 1.7450e-16, -1.6806e-03, -7.4609e-04, -1.5372e-03,\n -1.7810e-03, 1.6644e-04, -7.2881e-03, -3.3304e-03, -5.3491e-04,\n -2.6843e-04, 1.1663e-03, 6.3363e-04, 1.8992e-03, -4.1072e-35,\n 5.6052e-45, -7.0372e-03, -7.7995e-04, 6.9158e-05, 3.1582e-04,\n -3.8041e-03, 2.7068e-03, 5.4373e-04, 5.6052e-45, -1.8535e-03,\n 2.0693e-03, -1.7930e-03, -7.1912e-04, 5.6052e-45, -1.6341e-03,\n -2.0785e-03, -1.5169e-03, -4.0041e-03, 6.6253e-03, -6.6289e-21,\n 5.6052e-45, -1.3885e-03, 1.3406e-03, 1.6914e-03, 5.4886e-04,\n 1.6931e-04, -4.8455e-04, -3.7299e-04, 5.6052e-45, 3.2643e-04,\n 1.9032e-03, 1.2197e-03, 1.2259e-03, 1.8186e-03, 1.2542e-03,\n 5.6052e-45, 2.3774e-03, -2.8570e-03, 6.9884e-04, 1.1143e-03,\n -1.3073e-04, -3.2651e-04, -1.5357e-03, -1.7092e-02, -8.1063e-04,\n 3.6122e-03, 7.0732e-04, 3.2542e-03, -4.1671e-04, -2.1162e-04,\n -8.7467e-05, 3.9853e-03, -1.5770e-03, 1.9945e-03, 5.1694e-04,\n -2.4642e-05, 1.9110e-03, -8.1750e-03, -4.4143e-03, 3.0264e-03,\n -5.1558e-05, 3.2541e-03, 2.5530e-03, 5.6052e-45, -4.2672e-04,\n 2.4149e-03, -1.5908e-03, 1.2887e-03, -2.3173e-03, 5.9383e-04,\n 1.4040e-04, -1.5054e-03, 1.8201e-03, -1.0404e-03, 4.2278e-03,\n 1.0953e-03, -8.5257e-04, -2.6585e-04, -3.4092e-04, -1.9889e-03,\n 1.5783e-03, 4.9310e-03, -3.4958e-04, -2.0176e-03, -4.1653e-05,\n 2.1225e-03, 5.6052e-45, -1.2615e-03, -1.5187e-03, -6.9453e-05,\n -1.2908e-03, 1.5175e-03, 3.8103e-04, 3.2755e-03, -7.3108e-04,\n 3.3495e-03, 2.6484e-03, 5.6052e-45, 1.3955e-04, 2.4664e-03,\n 3.0513e-05, 1.5540e-03, 2.7154e-03, 1.2035e-03, -1.2743e-04,\n 1.5686e-11, -2.2926e-03, 5.6052e-45, -1.1629e-03, -1.0426e-03,\n 4.4875e-03, 3.8890e-04, -1.2673e-03, -2.6926e-03, 5.8198e-04,\n -7.9894e-04, 5.6052e-45, -1.2130e-03, 1.6454e-03, -7.5367e-03,\n -1.0624e-05, -3.8889e-03, 1.2439e-03, -2.1511e-03, -1.6852e-03,\n 5.8289e-03, -2.4693e-03, -2.8744e-03, 1.4555e-03, 5.4154e-04,\n -3.0575e-03, 2.6792e-04, 1.5759e-07, -2.3129e-03, -2.6765e-03,\n -9.7535e-07, -7.1819e-04, 6.4001e-40, 5.6052e-45, -3.2492e-04,\n -1.8484e-03, 5.2411e-04, 4.2317e-03, -6.3674e-04, -1.5397e-03,\n 5.6048e-04, 2.3906e-03, -2.8164e-03, -1.1538e-03, 1.4445e-03,\n 3.8265e-03, -9.7848e-04, 4.0255e-04, -8.5947e-04, -8.6640e-05,\n -4.9727e-05, 1.1320e-03, -2.1908e-05, 3.3552e-03, 5.8964e-06,\n 4.2318e-04, -2.3774e-03, 4.0055e-03, -6.5512e-04, 5.6052e-45,\n 1.2157e-03, 6.3504e-04, 3.2873e-03, -6.5067e-04, 1.6978e-04,\n -3.8350e-04, -6.3015e-04, -5.1111e-04, 6.1542e-04, 3.4668e-03,\n 8.5541e-03, 5.6052e-45, 7.0538e-04, 1.7820e-03, 1.1006e-04,\n -6.4889e-04, -3.8285e-04, 4.7411e-04, 4.0403e-03, -1.2390e-04,\n 1.4954e-03, 4.5684e-04, -4.0117e-04, -2.5432e-03, 5.4513e-28,\n 1.0953e-03, 1.2679e-03, -7.2141e-03, 2.2937e-03, 1.0306e-03,\n -1.4254e-03, -3.4962e-04, -3.8257e-04, 1.2993e-03, -1.0368e-04,\n 1.0880e-03, 2.0327e-03, -7.5284e-04, 7.0328e-04, 1.6928e-03,\n 1.1690e-03, 2.2969e-04, -1.1355e-03, 5.6052e-45, 5.6912e-04,\n 6.9722e-04, 4.9105e-07, -4.4864e-04, 1.5886e-03, -2.5353e-04,\n 1.1741e-03, -5.4882e-03, 4.8552e-04, -2.3815e-03, -4.0432e-04,\n 3.2380e-03, 1.9961e-03, 5.5193e-04, -1.7605e-33, 5.6052e-45,\n 2.6521e-03, 2.4160e-04, 2.9053e-03, -2.0351e-04, -1.7770e-03,\n 2.2448e-03, -1.8172e-03, 2.3603e-03, -9.3427e-04, 4.6545e-04,\n 2.4674e-03, -1.9138e-03, 5.6052e-45, 2.6908e-04, 6.4993e-05,\n 5.6052e-45, -1.0407e-07, 5.6052e-45, 1.9614e-03, 1.4860e-03,\n 1.0679e-04, -7.2474e-04, 5.6052e-45, 8.8939e-04, -2.2507e-03,\n -2.0403e-04, 2.4861e-04, 1.5738e-04, 4.1969e-03, -9.0950e-04,\n -4.3497e-07, -4.5330e-03, -4.3736e-16, 2.6728e-03, -3.4864e-12,\n 6.6541e-04, -3.8415e-03, -4.9617e-03, 5.6052e-45, -2.8454e-04,\n 3.2791e-03, 2.8651e-05, 5.0693e-03, 1.8879e-03, 1.4716e-03,\n -4.9795e-03, -1.4910e-03, -2.4324e-03, 2.8446e-03, -1.5682e-03,\n -2.8933e-09, 1.4629e-03, 9.7059e-20, -7.9898e-04, 8.4393e-04,\n 5.6052e-45, -1.2870e-03, -1.4444e-03, -3.7080e-04, 2.6747e-03,\n 7.7871e-10, -2.4555e-03, -1.6801e-04, 1.6547e-03, 5.6052e-45,\n 1.5095e-04, 8.6976e-04, 5.6052e-45, -1.3492e-03, 7.0371e-17,\n 1.9926e-03, -1.2434e-05, 5.6052e-45, 4.2391e-03, 1.6444e-03,\n -2.2978e-04, 1.0726e-03, 2.7664e-03, 6.5306e-04, 3.4669e-03,\n 1.4955e-03, 5.6052e-45, -6.9197e-04, 5.6052e-45, 9.3500e-04,\n 5.7621e-04, 6.1615e-04, -1.7161e-03, 1.6078e-04, 5.0908e-03,\n -1.7193e-03, 1.2305e-03, -1.9431e-03, -1.9954e-04, 8.6554e-04,\n 6.1657e-44, 3.6795e-03, -2.0908e-03, 4.7156e-04, 3.5966e-04,\n 5.7927e-04, 5.7433e-05, -2.0558e-03, 2.1328e-03, -1.0483e-03,\n 2.0550e-03, 5.6052e-45, 4.8020e-06, 3.4740e-03, -9.1481e-04,\n -2.3186e-03, -3.0433e-03, 2.2122e-03, 1.7908e-03, 2.8714e-03,\n -5.2433e-04, -6.5450e-04, -5.8112e-03, -6.1113e-04, 1.2991e-09,\n 5.6052e-45, -3.7407e-04, -8.6636e-04, -7.9024e-04, -1.2939e-03,\n 1.6390e-03, -6.4169e-04, -1.7501e-03, 2.5247e-03, 1.5380e-03,\n 5.4311e-04, -1.9798e-03, -1.3643e-03, 3.7215e-04, -3.3887e-04,\n 1.2315e-12, -6.6939e-04, 4.9933e-03, -3.3018e-03, -5.6052e-45,\n -9.1870e-03, -2.3953e-03, 9.9825e-34, 5.6052e-45, -1.8366e-03,\n 4.0818e-03, -7.4626e-05, 1.8173e-03, 3.3083e-03, -2.8544e-04,\n -2.2003e-05, 8.8995e-04, -2.6031e-03, -2.9603e-03, 1.7691e-03,\n 1.1856e-03, -1.0590e-03, -1.7087e-03, -4.7046e-03, -2.6489e-03,\n 9.6220e-04, 1.1795e-03, -6.1423e-03, 2.8327e-03, -1.9739e-04,\n 9.0532e-04, -2.3302e-03, -1.0570e-03, 2.9605e-03, 9.4562e-05,\n -2.4800e-03, -1.9176e-03, 2.6401e-03, -6.7864e-04, 9.1049e-04,\n 2.5613e-03, 1.8557e-03, 2.7468e-03, 1.7356e-04, 3.0913e-07,\n -1.6716e-09, 5.6052e-45, -3.9027e-03, -1.2063e-03, 3.5184e-03,\n 4.1062e-03, 4.3985e-22, -1.2497e-03, 4.5070e-04, 1.4553e-03,\n 3.4370e-03, 3.0550e-04, 2.5467e-03, -6.4545e-04, -2.6825e-03,\n 1.6909e-03, -1.3134e-03, 2.0747e-03, -2.3815e-03, -8.7160e-03,\n -5.2320e-03, 1.6452e-03, 6.0151e-04, -1.9492e-04, -4.6760e-04,\n -8.1648e-03, -2.6751e-03, 5.5363e-04, -9.6228e-05, -1.5683e-03,\n 4.8903e-04, 2.4498e-03, -2.3832e-03, 1.7399e-03, 1.2952e-03,\n 8.9709e-04, -2.6526e-04, -5.7862e-04, 4.5445e-04, 6.4842e-04,\n -9.3869e-05, -8.8490e-04, 4.7145e-03, 7.8731e-04, 3.4545e-12,\n 3.1115e-13, 1.7237e-07, -4.2924e-04, 2.8269e-03, 1.4195e-03,\n -4.3912e-03, 5.2732e-03, -9.6656e-04, -2.4521e-04, 2.0737e-03,\n -2.2138e-03, 2.4862e-28, 6.3193e-39, -1.6646e-03, 4.4695e-03,\n 1.0923e-03, -1.2242e-03, -7.8823e-04, -5.8591e-22, -2.6610e-04],\n device='cuda:0')",
|
| 13 |
+
"exp_avg_sq": "tensor([2.6185e-05, 1.0394e-04, 6.1769e-05, 4.7187e-05, 5.4188e-07, 1.1386e-05,\n 8.2271e-05, 8.5614e-05, 5.1013e-05, 8.3434e-05, 8.1024e-05, 1.4614e-04,\n 1.8464e-06, 7.5689e-05, 8.3315e-05, 8.7148e-05, 1.2984e-04, 1.1138e-04,\n 1.7976e-05, 4.8074e-07, 5.8047e-05, 6.4854e-05, 5.8102e-05, 8.5291e-05,\n 1.7548e-06, 1.3080e-04, 7.9323e-05, 1.1112e-04, 7.1867e-05, 1.0145e-04,\n 8.9637e-05, 1.0222e-04, 3.5326e-06, 3.0992e-07, 6.2278e-05, 8.4035e-05,\n 9.4529e-05, 6.0741e-05, 3.8722e-05, 8.5220e-05, 9.3767e-05, 4.0801e-05,\n 8.1706e-05, 6.9468e-05, 5.5784e-05, 9.4584e-05, 8.3837e-05, 1.4525e-04,\n 1.2973e-04, 3.2615e-05, 1.0640e-04, 1.8409e-07, 2.9872e-07, 2.7956e-07,\n 9.5597e-05, 1.5516e-04, 5.8881e-05, 5.9887e-05, 9.0280e-07, 1.1803e-04,\n 7.7816e-05, 7.0783e-05, 1.2705e-04, 4.5680e-05, 1.0234e-04, 1.2013e-04,\n 8.0419e-05, 3.4217e-07, 1.0126e-04, 8.9997e-05, 3.2717e-06, 1.3977e-06,\n 1.1270e-04, 6.5817e-05, 1.3604e-05, 8.7232e-05, 9.4244e-05, 1.0724e-04,\n 3.5342e-05, 9.8258e-05, 1.2847e-04, 7.9171e-05, 1.0172e-04, 6.4004e-05,\n 8.7273e-05, 6.1845e-05, 9.3616e-05, 3.6464e-05, 8.5587e-05, 7.1024e-05,\n 2.1820e-06, 1.1252e-04, 7.3912e-05, 5.1679e-05, 1.2169e-04, 2.4996e-05,\n 6.7809e-05, 4.7224e-05, 6.6435e-05, 5.3782e-05, 2.7094e-05, 1.7723e-06,\n 1.0651e-04, 1.1359e-04, 1.4858e-04, 2.8774e-05, 7.7910e-05, 1.2017e-04,\n 7.7663e-06, 3.1601e-05, 1.3684e-04, 1.3643e-04, 1.4752e-07, 6.1557e-05,\n 9.7825e-07, 5.3982e-05, 1.0140e-04, 1.3039e-04, 6.5776e-08, 3.2824e-07,\n 2.3927e-07, 6.9650e-05, 1.9053e-07, 1.7587e-04, 7.5585e-05, 1.2484e-04,\n 9.4132e-05, 3.5841e-08, 2.9235e-05, 1.1996e-04, 6.7474e-05, 6.5300e-05,\n 1.0678e-04, 6.3938e-08, 1.0664e-04, 8.1291e-05, 3.6267e-05, 8.1214e-05,\n 5.7613e-05, 8.0973e-05, 1.4920e-04, 9.8473e-05, 1.0549e-04, 8.5102e-05,\n 9.9494e-05, 8.1772e-05, 2.8215e-06, 9.8957e-05, 7.6632e-05, 6.0016e-05,\n 2.4200e-06, 5.4160e-05, 6.3808e-05, 1.2984e-06, 7.0577e-05, 1.0513e-04,\n 2.5717e-06, 1.1417e-04, 6.7006e-05, 9.2386e-05, 6.7913e-06, 8.7186e-05,\n 1.9947e-04, 7.3766e-05, 7.8754e-05, 6.3501e-05, 1.4718e-04, 5.0104e-05,\n 7.4427e-05, 3.0491e-07, 4.7086e-07, 9.1470e-05, 1.1657e-04, 6.9807e-05,\n 6.6640e-05, 1.1796e-04, 3.5256e-05, 9.9528e-05, 8.5455e-08, 3.3643e-05,\n 7.4284e-05, 1.1369e-04, 6.7078e-06, 6.2876e-07, 1.0280e-04, 9.3019e-05,\n 1.2655e-04, 4.8381e-05, 6.5508e-05, 8.3013e-08, 2.0920e-06, 8.1143e-05,\n 4.1950e-05, 8.5873e-05, 9.3965e-05, 7.8254e-05, 5.4772e-05, 1.0383e-04,\n 3.2973e-07, 7.5674e-05, 5.0118e-05, 4.2554e-05, 1.3158e-04, 3.2573e-05,\n 1.7879e-04, 1.6264e-08, 8.9628e-05, 9.3593e-05, 7.0992e-05, 1.0373e-04,\n 4.2838e-05, 8.6818e-05, 9.0367e-05, 8.9597e-05, 4.0722e-05, 7.4222e-05,\n 5.1326e-05, 1.0765e-04, 1.0529e-04, 4.9929e-05, 6.4792e-05, 9.1554e-05,\n 6.7870e-05, 7.0876e-05, 7.7211e-05, 7.0628e-05, 1.3262e-04, 1.4059e-04,\n 5.7287e-05, 1.2511e-04, 8.2398e-05, 1.4213e-04, 1.0192e-04, 4.0426e-07,\n 9.4356e-05, 4.8902e-05, 7.8567e-05, 6.1416e-05, 1.8999e-04, 7.6075e-05,\n 4.4971e-05, 6.0220e-05, 8.8060e-05, 8.6001e-05, 1.3341e-04, 5.4736e-05,\n 3.7458e-05, 8.9258e-05, 5.9941e-05, 8.1535e-05, 7.6555e-05, 7.9841e-05,\n 6.3156e-05, 7.4600e-05, 1.0078e-04, 8.0634e-05, 1.7419e-06, 6.4350e-05,\n 4.5434e-05, 1.1492e-05, 8.4327e-05, 1.2536e-04, 9.3495e-05, 9.0823e-05,\n 8.5464e-05, 9.2246e-05, 6.1173e-05, 2.5820e-07, 8.8737e-05, 9.7515e-05,\n 6.1668e-05, 1.1619e-04, 7.8469e-05, 2.3436e-05, 1.1286e-04, 4.5058e-07,\n 8.2334e-05, 1.2733e-06, 5.7528e-05, 1.5404e-05, 7.8400e-05, 6.4173e-05,\n 1.3407e-04, 1.0631e-04, 1.0481e-04, 8.8875e-05, 5.1930e-08, 9.9428e-05,\n 1.1994e-04, 5.9341e-05, 1.1065e-06, 8.9699e-05, 1.1248e-04, 9.9404e-05,\n 7.6766e-05, 1.1760e-04, 1.3545e-04, 7.7120e-05, 8.2723e-05, 1.5759e-04,\n 6.8220e-05, 5.4418e-06, 3.6677e-06, 6.0936e-05, 8.8757e-05, 1.0897e-05,\n 9.4508e-05, 5.3596e-08, 1.0559e-06, 8.7500e-05, 1.1486e-04, 5.9288e-05,\n 9.1286e-05, 1.0201e-04, 7.0026e-05, 9.0122e-05, 1.0238e-04, 8.5591e-05,\n 5.7840e-05, 9.6669e-05, 1.0544e-04, 1.4472e-04, 1.2356e-04, 5.6766e-05,\n 2.4556e-05, 9.6125e-05, 7.5443e-05, 9.2146e-07, 6.3154e-05, 1.0093e-04,\n 1.0514e-05, 5.2251e-05, 1.1961e-04, 9.2645e-05, 1.2597e-10, 7.3657e-05,\n 4.6807e-05, 1.2449e-04, 5.7429e-05, 7.0788e-05, 1.1652e-05, 9.9527e-05,\n 7.5082e-05, 7.2245e-05, 7.6273e-05, 1.3233e-04, 1.8434e-06, 4.5481e-06,\n 8.8133e-05, 9.6553e-05, 5.2322e-05, 1.0399e-04, 7.0010e-05, 8.0560e-05,\n 1.3993e-04, 6.5921e-05, 7.4668e-05, 7.2204e-05, 1.0716e-04, 1.7566e-06,\n 4.2006e-05, 5.0364e-05, 9.5045e-05, 1.1070e-04, 9.8440e-05, 2.0420e-04,\n 1.8978e-05, 9.6445e-05, 1.0712e-04, 1.0500e-04, 1.0441e-04, 8.0089e-05,\n 7.3783e-05, 4.5673e-06, 7.9015e-05, 1.3546e-04, 1.8708e-04, 1.9370e-05,\n 8.0249e-07, 1.0890e-04, 4.9397e-06, 4.3785e-06, 7.1149e-05, 1.5961e-04,\n 5.1601e-05, 6.7308e-05, 8.8343e-05, 1.2766e-04, 5.8176e-05, 4.6119e-05,\n 1.1389e-04, 1.0290e-04, 6.7217e-05, 4.3518e-07, 1.4440e-06, 9.5064e-05,\n 9.4383e-05, 9.7615e-05, 1.1845e-04, 5.2660e-05, 6.8345e-05, 6.1756e-05,\n 1.4186e-05, 5.7426e-05, 6.6783e-05, 1.2038e-04, 6.5310e-05, 5.3178e-07,\n 1.4157e-06, 2.1271e-05, 1.5319e-07, 1.8450e-06, 2.6727e-07, 1.1933e-04,\n 8.5703e-05, 2.4135e-05, 5.8756e-05, 7.2158e-07, 3.5566e-05, 7.8141e-05,\n 5.5834e-05, 9.3492e-05, 1.0075e-04, 1.3223e-04, 7.8316e-05, 1.8754e-06,\n 4.3172e-05, 5.4594e-07, 4.6574e-05, 1.0072e-07, 3.9667e-05, 9.7900e-05,\n 8.4505e-05, 4.6350e-07, 5.1821e-05, 7.8551e-05, 1.1745e-04, 1.1118e-04,\n 6.1050e-05, 3.2160e-05, 9.7376e-05, 4.0605e-05, 9.7365e-05, 1.2358e-04,\n 5.8186e-05, 1.2478e-06, 7.2198e-05, 1.5644e-07, 7.9856e-05, 4.7343e-05,\n 5.1903e-06, 2.6853e-05, 6.3829e-05, 5.3001e-06, 8.1505e-05, 2.4521e-07,\n 4.6788e-05, 5.6387e-06, 7.1344e-05, 7.0069e-06, 8.7032e-05, 1.2208e-04,\n 4.5556e-08, 7.1572e-05, 5.8981e-06, 9.8941e-05, 6.0820e-05, 3.7207e-07,\n 9.8656e-05, 5.5876e-05, 1.6058e-04, 1.2404e-04, 6.0918e-05, 1.4882e-05,\n 5.3910e-05, 8.8351e-05, 2.5861e-06, 1.0310e-04, 8.2110e-07, 6.7687e-06,\n 3.4286e-06, 3.0858e-06, 6.5671e-05, 9.9121e-05, 1.5407e-04, 8.6984e-05,\n 7.1729e-05, 6.9371e-05, 8.7477e-05, 6.8605e-05, 3.4451e-07, 8.3578e-05,\n 8.1260e-05, 1.5208e-04, 4.7259e-06, 2.0069e-05, 6.8325e-05, 5.7288e-05,\n 7.2662e-05, 6.0905e-06, 7.6995e-05, 2.5564e-06, 6.0025e-07, 1.2225e-04,\n 7.8866e-05, 1.3945e-04, 8.2698e-05, 7.7207e-05, 9.0479e-05, 6.0077e-05,\n 9.0126e-05, 1.6558e-04, 6.1826e-05, 8.6941e-05, 1.9133e-06, 2.1177e-06,\n 1.0096e-04, 1.9103e-04, 1.1656e-04, 7.5653e-05, 9.0906e-05, 9.2470e-05,\n 8.5884e-05, 8.2854e-05, 7.5234e-05, 1.4954e-05, 8.3693e-05, 9.7825e-05,\n 6.3007e-05, 5.4483e-05, 2.5304e-07, 5.8177e-05, 9.6390e-05, 5.2218e-05,\n 1.9278e-06, 7.9757e-05, 8.1680e-05, 1.5617e-07, 4.6986e-07, 9.3479e-05,\n 9.5239e-05, 1.0603e-04, 5.2933e-05, 1.0776e-04, 5.5441e-05, 9.0652e-05,\n 6.6375e-05, 4.8456e-05, 5.6797e-05, 8.7112e-05, 9.3751e-05, 9.1886e-05,\n 1.0527e-04, 1.0298e-04, 9.0333e-05, 4.2948e-05, 7.7361e-05, 2.2191e-04,\n 6.5279e-05, 7.3744e-05, 3.9408e-05, 7.9760e-05, 9.1075e-05, 1.4052e-04,\n 1.0290e-06, 6.9951e-05, 1.1078e-04, 4.7856e-05, 1.0981e-04, 8.3493e-05,\n 8.2707e-05, 8.2696e-05, 8.3315e-05, 1.2272e-04, 4.3915e-07, 2.4494e-07,\n 2.8034e-06, 8.8715e-05, 7.5586e-05, 1.0030e-04, 9.2822e-05, 5.3358e-06,\n 6.6156e-05, 8.0489e-05, 2.5207e-05, 7.7702e-05, 5.6717e-05, 6.4486e-05,\n 1.1964e-04, 9.9142e-05, 6.0166e-05, 1.1964e-04, 9.7669e-05, 1.2682e-04,\n 9.5410e-05, 7.4313e-05, 9.2821e-05, 9.2618e-05, 1.3747e-04, 1.4180e-04,\n 1.2441e-04, 5.9513e-05, 7.8540e-05, 7.1513e-05, 8.1343e-05, 6.1158e-05,\n 6.6370e-05, 8.8826e-05, 1.7511e-04, 1.3670e-04, 1.1294e-04, 1.2170e-04,\n 9.5316e-05, 2.8118e-05, 3.2750e-05, 9.4559e-05, 8.2394e-05, 8.8387e-05,\n 4.8002e-05, 9.3894e-08, 3.5055e-06, 2.6614e-07, 1.0201e-04, 7.1645e-05,\n 8.9279e-05, 1.1596e-04, 7.1520e-05, 5.8988e-05, 5.3306e-05, 1.1040e-04,\n 5.0714e-05, 8.9861e-07, 4.1866e-09, 1.1326e-04, 7.4199e-05, 1.3079e-04,\n 1.0399e-04, 8.0540e-05, 3.3352e-07, 6.1059e-05], device='cuda:0')"
|
| 14 |
},
|
| 15 |
"2": {
|
| 16 |
+
"step": "tensor(2504.)",
|
| 17 |
+
"exp_avg": "tensor([[ 3.0606e-06, -7.0439e-06, -9.3668e-06, ..., 1.9523e-05,\n -6.8146e-25, 3.6510e-07],\n [-1.7220e-05, -8.9561e-06, 4.2140e-05, ..., 6.9116e-05,\n -5.2215e-25, 2.2923e-04],\n [ 7.1131e-06, 2.5793e-05, 5.8272e-05, ..., -9.9563e-06,\n -6.9443e-25, 4.2099e-05],\n ...,\n [ 3.9873e-05, 6.0743e-06, 1.5934e-05, ..., -4.3467e-06,\n -9.9725e-25, -9.5190e-05],\n [-4.2656e-05, -2.4305e-05, -5.0580e-05, ..., 3.3702e-05,\n -1.3243e-25, -4.7871e-05],\n [-2.6036e-06, -3.1546e-05, 1.0162e-04, ..., 7.3760e-05,\n 5.4446e-25, -1.6388e-04]], device='cuda:0')",
|
| 18 |
+
"exp_avg_sq": "tensor([[9.3977e-09, 2.8265e-08, 2.0304e-08, ..., 2.6207e-08, 1.8794e-13,\n 2.6969e-08],\n [1.5989e-08, 6.8773e-08, 4.2667e-08, ..., 5.1190e-08, 3.4671e-11,\n 4.2784e-08],\n [1.3324e-08, 4.7240e-08, 3.7844e-08, ..., 3.8673e-08, 2.3181e-10,\n 5.1588e-08],\n ...,\n [1.1073e-08, 5.1956e-08, 4.5597e-08, ..., 4.7573e-08, 1.6871e-10,\n 5.4408e-08],\n [2.2141e-08, 6.1669e-08, 4.6652e-08, ..., 5.7102e-08, 1.5353e-12,\n 1.2886e-07],\n [1.5296e-08, 5.8622e-08, 4.2407e-08, ..., 7.2614e-08, 2.1330e-10,\n 3.7079e-08]], device='cuda:0')"
|
| 19 |
}
|
| 20 |
},
|
| 21 |
"param_groups": [
|
| 22 |
{
|
| 23 |
+
"lr": 0.00904518046337755,
|
| 24 |
"name": "scale_256",
|
| 25 |
"betas": [
|
| 26 |
0.9,
|
|
|
|
| 43 |
]
|
| 44 |
},
|
| 45 |
{
|
| 46 |
+
"lr": 0.00904518046337755,
|
| 47 |
"name": "scale_512",
|
| 48 |
"betas": [
|
| 49 |
0.9,
|
|
|
|
| 66 |
]
|
| 67 |
},
|
| 68 |
{
|
| 69 |
+
"lr": 0.00904518046337755,
|
| 70 |
"name": "scale_768",
|
| 71 |
"betas": [
|
| 72 |
0.9,
|
|
|
|
| 89 |
]
|
| 90 |
},
|
| 91 |
{
|
| 92 |
+
"lr": 0.00904518046337755,
|
| 93 |
"name": "scale_1024",
|
| 94 |
"betas": [
|
| 95 |
0.9,
|
|
|
|
| 112 |
]
|
| 113 |
},
|
| 114 |
{
|
| 115 |
+
"lr": 0.00904518046337755,
|
| 116 |
"name": "scale_1280",
|
| 117 |
"betas": [
|
| 118 |
0.9,
|
|
|
|
| 135 |
]
|
| 136 |
},
|
| 137 |
{
|
| 138 |
+
"lr": 0.004522637977440181,
|
| 139 |
"name": "fusion",
|
| 140 |
"betas": [
|
| 141 |
0.9,
|
|
|
|
| 195 |
"T_i": 10,
|
| 196 |
"T_mult": 2,
|
| 197 |
"eta_min": 1e-06,
|
| 198 |
+
"T_cur": 2,
|
| 199 |
"base_lrs": [
|
| 200 |
0.01,
|
| 201 |
0.01,
|
|
|
|
| 204 |
0.01,
|
| 205 |
0.005
|
| 206 |
],
|
| 207 |
+
"last_epoch": 2,
|
| 208 |
"_step_count": 0,
|
| 209 |
"_is_initial": false,
|
| 210 |
"_get_lr_called_within_step": false,
|
| 211 |
"_last_lr": [
|
| 212 |
+
0.00904518046337755,
|
| 213 |
+
0.00904518046337755,
|
| 214 |
+
0.00904518046337755,
|
| 215 |
+
0.00904518046337755,
|
| 216 |
+
0.00904518046337755,
|
| 217 |
+
0.004522637977440181
|
| 218 |
]
|
| 219 |
},
|
| 220 |
"metrics": {
|
| 221 |
+
"best_val_acc": 73.398,
|
| 222 |
+
"best_epoch": 1,
|
| 223 |
"scale_accuracies": {
|
| 224 |
+
"256": 73.398
|
| 225 |
}
|
| 226 |
},
|
| 227 |
"train_config": {
|