| D:\Projects\lejepa_control\third_party\stable-worldmodel\stable_worldmodel\envs\__init__.py:215: UserWarning: ale-py not found
|
| from stable_worldmodel.envs import ale # noqa: F401
|
| [32m2026-08-04 00:04:20.853[0m | [1mINFO [0m | [36mstable_worldmodel.wm.utils[0m:[36m_resolve_hf[0m:[36m150[0m - [1mLoading quentinll/lewm-pusht from local cache...[0m
|
| [32m2026-08-04 00:04:20.854[0m | [1mINFO [0m | [36mstable_worldmodel.wm.utils[0m:[36m_resolve_folder[0m:[36m138[0m - [1mLoading checkpoint from folder C:\Users\omnap\.stable_worldmodel\checkpoints\models--quentinll--lewm-pusht...[0m
|
| W0804 00:04:30.115000 32636 site-packages\torch\utils\flop_counter.py:29] triton not found
|
| [32m00:04:30[0m | [1mINFO [0m | [36matomic_chec~[0m| [1m[atomic_save] installed crash-safe checkpoint plugin (write to sibling .tmp + fsync + atomic rename)[0m
|
| [32m00:04:30[0m | [1mINFO [0m | [36mutils.py [0m| [1mCreated ViT-tiny from scratch with config: {'hidden_size': 192, 'num_hidden_layers': 12, 'num_attention_heads': 3, 'intermediate_size': 768, 'image_size': 224, 'patch_size': 14}[0m
|
| C:\Users\omnap\miniconda3\envs\llmdyn\Lib\site-packages\torch\nn\modules\module.py:1370: UserWarning: expandable_segments not supported on this platform (Triggered internally at C:\actions-runner\_work\pytorch\pytorch\pytorch\c10/cuda/CUDAAllocatorConfig.h:39.)
|
| return t.to(
|
| controller params: 6.80M
|
| support model loaded, c95=1.5306
|
| curriculum: [(0, 2), (5000, 3), (10000, 5)]
|
| step 250 H_goal<=2 loss 0.2538 terminal 0.2291 gain +0.1028 viol 0.182 grad 5.19 3.8 it/s support 0.0150
|
| step 500 H_goal<=2 loss 0.1169 terminal 0.0924 gain +0.1885 viol 0.029 grad 4.09 3.8 it/s support 0.0013
|
| step 750 H_goal<=2 loss 0.0876 terminal 0.0630 gain +0.2115 viol 0.058 grad 2.38 3.8 it/s support 0.0043
|
| step 1000 H_goal<=2 loss 0.0656 terminal 0.0432 gain +0.2080 viol 0.073 grad 1.04 3.7 it/s support 0.0061
|
| step 1250 H_goal<=2 loss 0.0530 terminal 0.0331 gain +0.1938 viol 0.086 grad 0.62 3.7 it/s support 0.0077
|
| step 1500 H_goal<=2 loss 0.0454 terminal 0.0280 gain +0.1740 viol 0.101 grad 0.39 3.7 it/s support 0.0101
|
| step 1750 H_goal<=2 loss 0.0400 terminal 0.0255 gain +0.1423 viol 0.103 grad 0.33 3.7 it/s support 0.0093
|
| step 2000 H_goal<=2 loss 0.0373 terminal 0.0257 gain +0.1038 viol 0.109 grad 0.35 3.7 it/s support 0.0108
|
| [val] terminal K=3: 0.1271 K=0: 0.3758 gain +0.2487 steps [0.527, 0.491, 0.452]
|
| step 2250 H_goal<=2 loss 0.0313 terminal 0.0226 gain +0.0744 viol 0.122 grad 0.28 3.7 it/s support 0.0106
|
| step 2500 H_goal<=2 loss 0.0299 terminal 0.0221 gain +0.0626 viol 0.125 grad 0.25 3.7 it/s support 0.0096
|
| step 2750 H_goal<=2 loss 0.0292 terminal 0.0216 gain +0.0593 viol 0.132 grad 0.24 3.7 it/s support 0.0106
|
| step 3000 H_goal<=2 loss 0.0274 terminal 0.0206 gain +0.0504 viol 0.138 grad 0.22 3.7 it/s support 0.0105
|
| step 3250 H_goal<=2 loss 0.0268 terminal 0.0204 gain +0.0471 viol 0.139 grad 0.21 3.7 it/s support 0.0100
|
| step 3500 H_goal<=2 loss 0.0257 terminal 0.0196 gain +0.0437 viol 0.150 grad 0.20 3.7 it/s support 0.0106
|
| step 3750 H_goal<=2 loss 0.0236 terminal 0.0182 gain +0.0375 viol 0.158 grad 0.16 3.7 it/s support 0.0109
|
| step 4000 H_goal<=2 loss 0.0239 terminal 0.0184 gain +0.0375 viol 0.151 grad 0.18 3.7 it/s support 0.0100
|
| [val] terminal K=3: 0.0806 K=0: 0.2419 gain +0.1613 steps [0.53, 0.485, 0.408]
|
| step 4250 H_goal<=2 loss 0.0232 terminal 0.0177 gain +0.0368 viol 0.163 grad 0.17 3.7 it/s support 0.0110
|
| step 4500 H_goal<=2 loss 0.0220 terminal 0.0171 gain +0.0304 viol 0.173 grad 0.14 3.7 it/s support 0.0112
|
| step 4750 H_goal<=2 loss 0.0229 terminal 0.0177 gain +0.0337 viol 0.169 grad 0.17 3.7 it/s support 0.0117
|
| step 5000 H_goal<=2 loss 0.0220 terminal 0.0170 gain +0.0310 viol 0.176 grad 0.16 3.7 it/s support 0.0114
|
| step 5250 H_goal<=3 loss 0.0451 terminal 0.0307 gain +0.1023 viol 0.146 grad 0.47 3.7 it/s support 0.0185
|
| step 5500 H_goal<=3 loss 0.0383 terminal 0.0258 gain +0.0902 viol 0.143 grad 0.35 3.7 it/s support 0.0164
|
| step 5750 H_goal<=3 loss 0.0381 terminal 0.0258 gain +0.0886 viol 0.144 grad 0.43 3.7 it/s support 0.0146
|
| step 6000 H_goal<=3 loss 0.0336 terminal 0.0235 gain +0.0710 viol 0.147 grad 0.31 3.7 it/s support 0.0153
|
| [val] terminal K=3: 0.0484 K=0: 0.1815 gain +0.1331 steps [0.537, 0.471, 0.361]
|
| step 6250 H_goal<=3 loss 0.0282 terminal 0.0199 gain +0.0575 viol 0.154 grad 0.19 3.7 it/s support 0.0142
|
| step 6500 H_goal<=3 loss 0.0261 terminal 0.0187 gain +0.0495 viol 0.157 grad 0.18 3.7 it/s support 0.0132
|
| step 6750 H_goal<=3 loss 0.0270 terminal 0.0188 gain +0.0560 viol 0.153 grad 0.20 3.7 it/s support 0.0120
|
| step 7000 H_goal<=3 loss 0.0273 terminal 0.0195 gain +0.0529 viol 0.162 grad 0.21 3.7 it/s support 0.0128
|
| step 7250 H_goal<=3 loss 0.0277 terminal 0.0196 gain +0.0537 viol 0.155 grad 0.22 3.7 it/s support 0.0127
|
| step 7500 H_goal<=3 loss 0.0254 terminal 0.0181 gain +0.0483 viol 0.161 grad 0.20 3.7 it/s support 0.0126
|
| step 7750 H_goal<=3 loss 0.0251 terminal 0.0180 gain +0.0463 viol 0.160 grad 0.18 3.7 it/s support 0.0113
|
| step 8000 H_goal<=3 loss 0.0243 terminal 0.0174 gain +0.0436 viol 0.167 grad 0.17 3.7 it/s support 0.0119
|
| [val] terminal K=3: 0.0348 K=0: 0.1564 gain +0.1216 steps [0.541, 0.465, 0.34]
|
| step 8250 H_goal<=3 loss 0.0248 terminal 0.0177 gain +0.0453 viol 0.169 grad 0.20 3.7 it/s support 0.0129
|
| step 8500 H_goal<=3 loss 0.0240 terminal 0.0173 gain +0.0414 viol 0.168 grad 0.16 3.7 it/s support 0.0125
|
| step 8750 H_goal<=3 loss 0.0233 terminal 0.0167 gain +0.0420 viol 0.172 grad 0.16 3.7 it/s support 0.0126
|
| step 9000 H_goal<=3 loss 0.0243 terminal 0.0169 gain +0.0470 viol 0.178 grad 0.19 3.7 it/s support 0.0145
|
| step 9250 H_goal<=3 loss 0.0231 terminal 0.0163 gain +0.0438 viol 0.175 grad 0.15 3.7 it/s support 0.0141
|
| step 9500 H_goal<=3 loss 0.0243 terminal 0.0170 gain +0.0472 viol 0.167 grad 0.19 3.7 it/s support 0.0127
|
| step 9750 H_goal<=3 loss 0.0217 terminal 0.0156 gain +0.0370 viol 0.174 grad 0.12 3.7 it/s support 0.0123
|
| step 10000 H_goal<=3 loss 0.0236 terminal 0.0166 gain +0.0452 viol 0.175 grad 0.18 3.7 it/s support 0.0134
|
| [val] terminal K=3: 0.0308 K=0: 0.1305 gain +0.0997 steps [0.543, 0.462, 0.329]
|
| step 10250 H_goal<=5 loss 0.0538 terminal 0.0324 gain +0.1425 viol 0.189 grad 0.58 3.7 it/s support 0.0296
|
| step 10500 H_goal<=5 loss 0.0462 terminal 0.0266 gain +0.1334 viol 0.170 grad 0.67 3.7 it/s support 0.0246
|
| step 10750 H_goal<=5 loss 0.0426 terminal 0.0247 gain +0.1226 viol 0.181 grad 0.58 3.7 it/s support 0.0246
|
| step 11000 H_goal<=5 loss 0.0379 terminal 0.0218 gain +0.1070 viol 0.180 grad 0.40 3.7 it/s support 0.0230
|
| step 11250 H_goal<=5 loss 0.0354 terminal 0.0206 gain +0.0977 viol 0.180 grad 0.31 3.7 it/s support 0.0218
|
| step 11500 H_goal<=5 loss 0.0361 terminal 0.0212 gain +0.0974 viol 0.180 grad 0.37 3.7 it/s support 0.0237
|
| step 11750 H_goal<=5 loss 0.0354 terminal 0.0212 gain +0.0924 viol 0.177 grad 0.34 3.7 it/s support 0.0224
|
| step 12000 H_goal<=5 loss 0.0342 terminal 0.0205 gain +0.0877 viol 0.182 grad 0.34 3.7 it/s support 0.0252
|
| [val] terminal K=3: 0.0177 K=0: 0.0862 gain +0.0686 steps [0.546, 0.455, 0.319]
|
| step 12250 H_goal<=5 loss 0.0327 terminal 0.0194 gain +0.0864 viol 0.179 grad 0.28 3.7 it/s support 0.0225
|
| step 12500 H_goal<=5 loss 0.0322 terminal 0.0193 gain +0.0829 viol 0.185 grad 0.34 3.7 it/s support 0.0237
|
| step 12750 H_goal<=5 loss 0.0311 terminal 0.0185 gain +0.0803 viol 0.187 grad 0.28 3.7 it/s support 0.0209
|
| step 13000 H_goal<=5 loss 0.0318 terminal 0.0192 gain +0.0796 viol 0.184 grad 0.43 3.7 it/s support 0.0228
|
| step 13250 H_goal<=5 loss 0.0295 terminal 0.0177 gain +0.0738 viol 0.185 grad 0.23 3.7 it/s support 0.0202
|
| step 13500 H_goal<=5 loss 0.0300 terminal 0.0179 gain +0.0779 viol 0.187 grad 0.64 3.7 it/s support 0.0215
|
| step 13750 H_goal<=5 loss 0.0293 terminal 0.0173 gain +0.0748 viol 0.188 grad 0.41 3.7 it/s support 0.0223
|
| step 14000 H_goal<=5 loss 0.0290 terminal 0.0174 gain +0.0728 viol 0.186 grad 0.26 3.7 it/s support 0.0205
|
| [val] terminal K=3: 0.0155 K=0: 0.0701 gain +0.0546 steps [0.548, 0.454, 0.314]
|
| step 14250 H_goal<=5 loss 0.0283 terminal 0.0170 gain +0.0700 viol 0.182 grad 0.31 3.7 it/s support 0.0202
|
| step 14500 H_goal<=5 loss 0.0282 terminal 0.0169 gain +0.0703 viol 0.187 grad 0.24 3.7 it/s support 0.0194
|
| step 14750 H_goal<=5 loss 0.0275 terminal 0.0166 gain +0.0669 viol 0.186 grad 0.55 3.7 it/s support 0.0188
|
| step 15000 H_goal<=5 loss 0.0270 terminal 0.0164 gain +0.0651 viol 0.190 grad 0.20 3.7 it/s support 0.0196
|
| step 15250 H_goal<=5 loss 0.0269 terminal 0.0163 gain +0.0629 viol 0.189 grad 0.24 3.7 it/s support 0.0193
|
| step 15500 H_goal<=5 loss 0.0262 terminal 0.0160 gain +0.0614 viol 0.187 grad 0.18 3.7 it/s support 0.0186
|
| step 15750 H_goal<=5 loss 0.0262 terminal 0.0160 gain +0.0615 viol 0.187 grad 0.26 3.7 it/s support 0.0183
|
| step 16000 H_goal<=5 loss 0.0269 terminal 0.0164 gain +0.0621 viol 0.195 grad 0.39 3.7 it/s support 0.0204
|
| [val] terminal K=3: 0.0169 K=0: 0.0670 gain +0.0501 steps [0.549, 0.452, 0.313]
|
| step 16250 H_goal<=5 loss 0.0262 terminal 0.0159 gain +0.0614 viol 0.194 grad 0.28 3.7 it/s support 0.0198
|
| step 16500 H_goal<=5 loss 0.0269 terminal 0.0163 gain +0.0622 viol 0.194 grad 0.24 3.7 it/s support 0.0216
|
| step 16750 H_goal<=5 loss 0.0261 terminal 0.0159 gain +0.0593 viol 0.196 grad 0.71 3.7 it/s support 0.0210
|
| step 17000 H_goal<=5 loss 0.0258 terminal 0.0157 gain +0.0590 viol 0.196 grad 0.24 3.7 it/s support 0.0210
|
| step 17250 H_goal<=5 loss 0.0263 terminal 0.0161 gain +0.0599 viol 0.197 grad 0.24 3.7 it/s support 0.0220
|
| step 17500 H_goal<=5 loss 0.0256 terminal 0.0157 gain +0.0583 viol 0.195 grad 0.16 3.7 it/s support 0.0211
|
| step 17750 H_goal<=5 loss 0.0254 terminal 0.0156 gain +0.0570 viol 0.192 grad 0.24 3.7 it/s support 0.0204
|
| step 18000 H_goal<=5 loss 0.0251 terminal 0.0154 gain +0.0564 viol 0.197 grad 0.21 3.7 it/s support 0.0207
|
| [val] terminal K=3: 0.0143 K=0: 0.0630 gain +0.0488 steps [0.55, 0.451, 0.312]
|
| step 18250 H_goal<=5 loss 0.0252 terminal 0.0154 gain +0.0567 viol 0.199 grad 0.48 3.7 it/s support 0.0206
|
| step 18500 H_goal<=5 loss 0.0253 terminal 0.0156 gain +0.0568 viol 0.198 grad 0.18 3.7 it/s support 0.0211
|
| step 18750 H_goal<=5 loss 0.0250 terminal 0.0153 gain +0.0561 viol 0.196 grad 0.36 3.7 it/s support 0.0209
|
| step 19000 H_goal<=5 loss 0.0247 terminal 0.0150 gain +0.0560 viol 0.196 grad 0.16 3.7 it/s support 0.0204
|
| step 19250 H_goal<=5 loss 0.0253 terminal 0.0156 gain +0.0559 viol 0.195 grad 0.19 3.7 it/s support 0.0204
|
| step 19500 H_goal<=5 loss 0.0252 terminal 0.0153 gain +0.0572 viol 0.193 grad 0.19 3.7 it/s support 0.0207
|
| step 19750 H_goal<=5 loss 0.0255 terminal 0.0157 gain +0.0567 viol 0.197 grad 0.17 3.7 it/s support 0.0208
|
| step 20000 H_goal<=5 loss 0.0247 terminal 0.0153 gain +0.0551 viol 0.195 grad 0.19 3.7 it/s support 0.0199
|
| [val] terminal K=3: 0.0147 K=0: 0.0582 gain +0.0435 steps [0.55, 0.451, 0.312]
|
| done in 89.2 min -> data\runs\controller
|
|
|
