Multiple distributions found for package optimum. Picked distribution: optimum [2026-07-09 23:19:49-INFO]: Found 77539 target latent files Set CUDA_VISIBLE_DEVICES to 0 [2026-07-09 23:19:49-INFO]: Image Num 77539, steps per epoch: 77539 [2026-07-09 23:19:55-INFO]: Model size: 0.471B [2026-07-09 23:19:55-INFO]: Loading checkpoint: ProMoE_B/ProMoE_TC_B/stage3/checkpoints/ckpt_step_2850000.pth [2026-07-09 23:19:59-INFO]: EMA model loaded [2026-07-09 23:19:59-INFO]: Successfully loaded checkpoint from step 2850000 [2026-07-09 23:19:59-INFO]: Resumed from step 2850000 [2026-07-09 23:19:59-INFO]: Gradient Accumulation Steps: 1 [2026-07-09 23:19:59-INFO]: Effective Batch Size: 1 [2026-07-09 23:19:59-INFO]: Start training loop (single GPU, Symbiosis) [2026-07-09 23:20:01-WARNING]: Abnormal loss skipped at step 2850003: 0.9877 [2026-07-09 23:20:02-WARNING]: Abnormal loss skipped at step 2850006: 0.9001 [2026-07-09 23:20:02-INFO]: epoch 0-step 2850010 loss: 0.403336 (accumulated 1 forwards) [2026-07-09 23:20:03-WARNING]: Abnormal loss skipped at step 2850011: 0.9313 [2026-07-09 23:20:04-WARNING]: Abnormal loss skipped at step 2850016: 0.8554 [2026-07-09 23:20:05-INFO]: epoch 0-step 2850020 loss: 0.669383 (accumulated 1 forwards) [2026-07-09 23:20:05-WARNING]: Abnormal loss skipped at step 2850021: 0.8017 [2026-07-09 23:20:05-WARNING]: Abnormal loss skipped at step 2850022: 0.9433 [2026-07-09 23:20:06-WARNING]: Abnormal loss skipped at step 2850023: 0.9063 [2026-07-09 23:20:06-WARNING]: Abnormal loss skipped at step 2850025: 0.8152 [2026-07-09 23:20:06-WARNING]: Abnormal loss skipped at step 2850026: 1.0535 [2026-07-09 23:20:06-WARNING]: Abnormal loss skipped at step 2850027: 0.9234 [2026-07-09 23:20:07-INFO]: epoch 0-step 2850030 loss: 0.694699 (accumulated 1 forwards) Multiple distributions found for package optimum. Picked distribution: optimum [2026-07-09 23:21:27-INFO]: Found 77539 target latent files Set CUDA_VISIBLE_DEVICES to 0 [2026-07-09 23:21:27-INFO]: Image Num 77539, steps per epoch: 77539 [2026-07-09 23:21:33-INFO]: Model size: 0.471B [2026-07-09 23:21:33-INFO]: Loading checkpoint: ProMoE_B/ProMoE_TC_B/stage3/checkpoints/ckpt_step_2850000.pth [2026-07-09 23:21:37-INFO]: EMA model loaded [2026-07-09 23:21:37-INFO]: Successfully loaded checkpoint from step 2850000 [2026-07-09 23:21:37-INFO]: Resumed from step 2850000 [2026-07-09 23:21:37-INFO]: Gradient Accumulation Steps: 1 [2026-07-09 23:21:37-INFO]: Effective Batch Size: 1 [2026-07-09 23:21:37-INFO]: Start training loop (single GPU, Symbiosis) [2026-07-09 23:21:40-INFO]: epoch 0-step 2850010 loss: 0.605335 (accumulated 1 forwards) [2026-07-09 23:21:43-INFO]: epoch 0-step 2850020 loss: 0.735790 (accumulated 1 forwards) [2026-07-09 23:21:45-INFO]: epoch 0-step 2850030 loss: 0.705630 (accumulated 1 forwards) [2026-07-09 23:21:47-INFO]: epoch 0-step 2850040 loss: 0.802988 (accumulated 1 forwards) [2026-07-09 23:21:50-INFO]: epoch 0-step 2850050 loss: 0.418516 (accumulated 1 forwards) [2026-07-09 23:21:52-INFO]: epoch 0-step 2850060 loss: 0.826379 (accumulated 1 forwards) [2026-07-09 23:21:54-INFO]: epoch 0-step 2850070 loss: 0.630274 (accumulated 1 forwards) [2026-07-09 23:21:56-INFO]: epoch 0-step 2850080 loss: 0.777455 (accumulated 1 forwards) [2026-07-09 23:21:59-INFO]: epoch 0-step 2850090 loss: 0.307366 (accumulated 1 forwards) [2026-07-09 23:22:01-INFO]: epoch 0-step 2850100 loss: 0.496776 (accumulated 1 forwards) [2026-07-09 23:22:03-INFO]: epoch 0-step 2850110 loss: 0.559018 (accumulated 1 forwards) [2026-07-09 23:22:06-INFO]: epoch 0-step 2850120 loss: 0.630785 (accumulated 1 forwards) [2026-07-09 23:22:08-INFO]: epoch 0-step 2850130 loss: 0.868380 (accumulated 1 forwards) [2026-07-09 23:22:10-INFO]: epoch 0-step 2850140 loss: 0.389605 (accumulated 1 forwards) [2026-07-09 23:22:12-INFO]: epoch 0-step 2850150 loss: 0.871568 (accumulated 1 forwards) [2026-07-09 23:22:15-INFO]: epoch 0-step 2850160 loss: 0.893016 (accumulated 1 forwards) [2026-07-09 23:22:17-INFO]: epoch 0-step 2850170 loss: 0.776400 (accumulated 1 forwards) [2026-07-09 23:22:19-INFO]: epoch 0-step 2850180 loss: 0.938013 (accumulated 1 forwards) [2026-07-09 23:22:21-INFO]: epoch 0-step 2850190 loss: 0.486453 (accumulated 1 forwards) [2026-07-09 23:22:24-INFO]: epoch 0-step 2850200 loss: 0.981067 (accumulated 1 forwards) [2026-07-09 23:22:26-INFO]: epoch 0-step 2850210 loss: 1.033741 (accumulated 1 forwards) [2026-07-09 23:22:28-INFO]: epoch 0-step 2850220 loss: 0.820033 (accumulated 1 forwards) [2026-07-09 23:22:30-INFO]: epoch 0-step 2850230 loss: 0.420432 (accumulated 1 forwards) [2026-07-09 23:22:32-INFO]: epoch 0-step 2850240 loss: 0.522842 (accumulated 1 forwards) [2026-07-09 23:22:35-INFO]: epoch 0-step 2850250 loss: 1.000804 (accumulated 1 forwards) [2026-07-09 23:22:37-INFO]: epoch 0-step 2850260 loss: 0.653976 (accumulated 1 forwards) [2026-07-09 23:22:39-INFO]: epoch 0-step 2850270 loss: 0.664723 (accumulated 1 forwards) [2026-07-09 23:22:41-INFO]: epoch 0-step 2850280 loss: 0.663410 (accumulated 1 forwards) Multiple distributions found for package optimum. Picked distribution: optimum [2026-07-09 23:22:58-INFO]: Found 77539 target latent files Set CUDA_VISIBLE_DEVICES to 0 [2026-07-09 23:22:58-INFO]: Image Num 77539, steps per epoch: 1212 [2026-07-09 23:23:04-INFO]: Model size: 0.471B [2026-07-09 23:23:04-INFO]: Loading checkpoint: ProMoE_B/ProMoE_TC_B/stage3/checkpoints/ckpt_step_2850000.pth [2026-07-09 23:23:09-INFO]: EMA model loaded [2026-07-09 23:23:09-INFO]: Successfully loaded checkpoint from step 2850000 [2026-07-09 23:23:09-INFO]: Resumed from step 2850000 [2026-07-09 23:23:09-INFO]: Gradient Accumulation Steps: 1 [2026-07-09 23:23:09-INFO]: Effective Batch Size: 64 [2026-07-09 23:23:09-INFO]: Start training loop (single GPU, Symbiosis) [2026-07-09 23:23:18-INFO]: epoch 0-step 2850010 loss: 0.694917 (accumulated 1 forwards) [2026-07-09 23:23:27-INFO]: epoch 0-step 2850020 loss: 0.667323 (accumulated 1 forwards) [2026-07-09 23:23:35-INFO]: epoch 0-step 2850030 loss: 0.696580 (accumulated 1 forwards) [2026-07-09 23:23:44-INFO]: epoch 0-step 2850040 loss: 0.677258 (accumulated 1 forwards) [2026-07-09 23:23:53-INFO]: epoch 0-step 2850050 loss: 0.650744 (accumulated 1 forwards) [2026-07-09 23:24:01-INFO]: epoch 0-step 2850060 loss: 0.666706 (accumulated 1 forwards) [2026-07-09 23:24:10-INFO]: epoch 0-step 2850070 loss: 0.623068 (accumulated 1 forwards) [2026-07-09 23:24:19-INFO]: epoch 0-step 2850080 loss: 0.675088 (accumulated 1 forwards) [2026-07-09 23:24:27-INFO]: epoch 0-step 2850090 loss: 0.710211 (accumulated 1 forwards) [2026-07-09 23:24:36-INFO]: epoch 0-step 2850100 loss: 0.669952 (accumulated 1 forwards) [2026-07-09 23:24:45-INFO]: epoch 0-step 2850110 loss: 0.695657 (accumulated 1 forwards) [2026-07-09 23:24:53-INFO]: epoch 0-step 2850120 loss: 0.635836 (accumulated 1 forwards) [2026-07-09 23:25:02-INFO]: epoch 0-step 2850130 loss: 0.691430 (accumulated 1 forwards) [2026-07-09 23:25:11-INFO]: epoch 0-step 2850140 loss: 0.628232 (accumulated 1 forwards) Multiple distributions found for package optimum. Picked distribution: optimum [2026-07-09 23:25:48-INFO]: Found 77539 target latent files Set CUDA_VISIBLE_DEVICES to 0 [2026-07-09 23:25:48-INFO]: Image Num 77539, steps per epoch: 1212 [2026-07-09 23:25:54-INFO]: Model size: 0.471B [2026-07-09 23:25:54-INFO]: Loading checkpoint: ProMoE_B/ProMoE_TC_B/stage3/checkpoints/ckpt_step_2850000.pth [2026-07-09 23:25:59-INFO]: EMA model loaded [2026-07-09 23:25:59-INFO]: Successfully loaded checkpoint from step 2850000 [2026-07-09 23:25:59-INFO]: Resumed from step 2850000 [2026-07-09 23:25:59-INFO]: Gradient Accumulation Steps: 1 [2026-07-09 23:25:59-INFO]: Effective Batch Size: 64 [2026-07-09 23:25:59-INFO]: Start training loop (single GPU, Symbiosis) Multiple distributions found for package optimum. Picked distribution: optimum [2026-07-09 23:26:42-INFO]: Found 77539 target latent files Set CUDA_VISIBLE_DEVICES to 0 [2026-07-09 23:26:42-INFO]: Image Num 77539, steps per epoch: 1212 [2026-07-09 23:26:48-INFO]: Model size: 0.471B [2026-07-09 23:26:48-INFO]: Loading checkpoint: ProMoE_B/ProMoE_TC_B/stage3/checkpoints/ckpt_step_2850000.pth [2026-07-09 23:26:53-INFO]: EMA model loaded [2026-07-09 23:26:53-INFO]: Successfully loaded checkpoint from step 2850000 [2026-07-09 23:26:53-INFO]: Resumed from step 2850000 [2026-07-09 23:26:53-INFO]: Gradient Accumulation Steps: 1 [2026-07-09 23:26:53-INFO]: Effective Batch Size: 64 [2026-07-09 23:26:53-INFO]: Start training loop (single GPU, Symbiosis) [2026-07-09 23:27:02-INFO]: epoch 0-step 2850010 loss: 0.713188 (accumulated 1 forwards) [2026-07-09 23:27:11-INFO]: epoch 0-step 2850020 loss: 0.704570 (accumulated 1 forwards) [2026-07-09 23:27:19-INFO]: epoch 0-step 2850030 loss: 0.633966 (accumulated 1 forwards) [2026-07-09 23:27:28-INFO]: epoch 0-step 2850040 loss: 0.684981 (accumulated 1 forwards) [2026-07-09 23:27:37-INFO]: epoch 0-step 2850050 loss: 0.691841 (accumulated 1 forwards) [2026-07-09 23:27:45-INFO]: epoch 0-step 2850060 loss: 0.654759 (accumulated 1 forwards) [2026-07-09 23:27:54-INFO]: epoch 0-step 2850070 loss: 0.682245 (accumulated 1 forwards) [2026-07-09 23:28:03-INFO]: epoch 0-step 2850080 loss: 0.646253 (accumulated 1 forwards) [2026-07-09 23:28:11-INFO]: epoch 0-step 2850090 loss: 0.702282 (accumulated 1 forwards) [2026-07-09 23:28:20-INFO]: epoch 0-step 2850100 loss: 0.669090 (accumulated 1 forwards) [2026-07-09 23:28:29-INFO]: epoch 0-step 2850110 loss: 0.685975 (accumulated 1 forwards) [2026-07-09 23:28:37-INFO]: epoch 0-step 2850120 loss: 0.734544 (accumulated 1 forwards) [2026-07-09 23:28:46-INFO]: epoch 0-step 2850130 loss: 0.652961 (accumulated 1 forwards) [2026-07-09 23:28:55-INFO]: epoch 0-step 2850140 loss: 0.649427 (accumulated 1 forwards) [2026-07-09 23:29:03-INFO]: epoch 0-step 2850150 loss: 0.636058 (accumulated 1 forwards) [2026-07-09 23:29:12-INFO]: epoch 0-step 2850160 loss: 0.719843 (accumulated 1 forwards) [2026-07-09 23:29:21-INFO]: epoch 0-step 2850170 loss: 0.691441 (accumulated 1 forwards) [2026-07-09 23:29:30-INFO]: epoch 0-step 2850180 loss: 0.705777 (accumulated 1 forwards) [2026-07-09 23:29:38-INFO]: epoch 0-step 2850190 loss: 0.700146 (accumulated 1 forwards) [2026-07-09 23:29:47-INFO]: epoch 0-step 2850200 loss: 0.660260 (accumulated 1 forwards) [2026-07-09 23:29:56-INFO]: epoch 0-step 2850210 loss: 0.667914 (accumulated 1 forwards) [2026-07-09 23:30:04-INFO]: epoch 0-step 2850220 loss: 0.674776 (accumulated 1 forwards) [2026-07-09 23:30:13-INFO]: epoch 0-step 2850230 loss: 0.677675 (accumulated 1 forwards) [2026-07-09 23:30:22-INFO]: epoch 0-step 2850240 loss: 0.660327 (accumulated 1 forwards) [2026-07-09 23:30:30-INFO]: epoch 0-step 2850250 loss: 0.669427 (accumulated 1 forwards) [2026-07-09 23:30:39-INFO]: epoch 0-step 2850260 loss: 0.689802 (accumulated 1 forwards) [2026-07-09 23:30:48-INFO]: epoch 0-step 2850270 loss: 0.695846 (accumulated 1 forwards) [2026-07-09 23:30:56-INFO]: epoch 0-step 2850280 loss: 0.630290 (accumulated 1 forwards) [2026-07-09 23:31:05-INFO]: epoch 0-step 2850290 loss: 0.694081 (accumulated 1 forwards) [2026-07-09 23:31:14-INFO]: epoch 0-step 2850300 loss: 0.685268 (accumulated 1 forwards) [2026-07-09 23:31:23-INFO]: epoch 0-step 2850310 loss: 0.668732 (accumulated 1 forwards) [2026-07-09 23:31:31-INFO]: epoch 0-step 2850320 loss: 0.674441 (accumulated 1 forwards) [2026-07-09 23:31:40-INFO]: epoch 0-step 2850330 loss: 0.611869 (accumulated 1 forwards) [2026-07-09 23:31:49-INFO]: epoch 0-step 2850340 loss: 0.656515 (accumulated 1 forwards) [2026-07-09 23:31:57-INFO]: epoch 0-step 2850350 loss: 0.674467 (accumulated 1 forwards) [2026-07-09 23:32:06-INFO]: epoch 0-step 2850360 loss: 0.690718 (accumulated 1 forwards) [2026-07-09 23:32:15-INFO]: epoch 0-step 2850370 loss: 0.673711 (accumulated 1 forwards) [2026-07-09 23:32:24-INFO]: epoch 0-step 2850380 loss: 0.674262 (accumulated 1 forwards) [2026-07-09 23:32:33-INFO]: epoch 0-step 2850390 loss: 0.693611 (accumulated 1 forwards) [2026-07-09 23:32:41-INFO]: epoch 0-step 2850400 loss: 0.707833 (accumulated 1 forwards) [2026-07-09 23:32:50-INFO]: epoch 0-step 2850410 loss: 0.653050 (accumulated 1 forwards) [2026-07-09 23:32:59-INFO]: epoch 0-step 2850420 loss: 0.656698 (accumulated 1 forwards) [2026-07-09 23:33:07-INFO]: epoch 0-step 2850430 loss: 0.684828 (accumulated 1 forwards) [2026-07-09 23:33:16-INFO]: epoch 0-step 2850440 loss: 0.700313 (accumulated 1 forwards) [2026-07-09 23:33:25-INFO]: epoch 0-step 2850450 loss: 0.691239 (accumulated 1 forwards) [2026-07-09 23:33:33-INFO]: epoch 0-step 2850460 loss: 0.642098 (accumulated 1 forwards) [2026-07-09 23:33:42-INFO]: epoch 0-step 2850470 loss: 0.703851 (accumulated 1 forwards) [2026-07-09 23:33:51-INFO]: epoch 0-step 2850480 loss: 0.693704 (accumulated 1 forwards) [2026-07-09 23:34:00-INFO]: epoch 0-step 2850490 loss: 0.703746 (accumulated 1 forwards) [2026-07-09 23:34:08-INFO]: epoch 0-step 2850500 loss: 0.677758 (accumulated 1 forwards) [2026-07-09 23:34:17-INFO]: epoch 0-step 2850510 loss: 0.681933 (accumulated 1 forwards) [2026-07-09 23:34:26-INFO]: epoch 0-step 2850520 loss: 0.718844 (accumulated 1 forwards) [2026-07-09 23:34:35-INFO]: epoch 0-step 2850530 loss: 0.677314 (accumulated 1 forwards) [2026-07-09 23:34:43-INFO]: epoch 0-step 2850540 loss: 0.623621 (accumulated 1 forwards) [2026-07-09 23:34:52-INFO]: epoch 0-step 2850550 loss: 0.679685 (accumulated 1 forwards) [2026-07-09 23:35:01-INFO]: epoch 0-step 2850560 loss: 0.673747 (accumulated 1 forwards) [2026-07-09 23:35:09-INFO]: epoch 0-step 2850570 loss: 0.653247 (accumulated 1 forwards) [2026-07-09 23:35:18-INFO]: epoch 0-step 2850580 loss: 0.705556 (accumulated 1 forwards) [2026-07-09 23:35:27-INFO]: epoch 0-step 2850590 loss: 0.678944 (accumulated 1 forwards) [2026-07-09 23:35:35-INFO]: epoch 0-step 2850600 loss: 0.651916 (accumulated 1 forwards) [2026-07-09 23:35:44-INFO]: epoch 0-step 2850610 loss: 0.677593 (accumulated 1 forwards) [2026-07-09 23:35:53-INFO]: epoch 0-step 2850620 loss: 0.687827 (accumulated 1 forwards) [2026-07-09 23:36:02-INFO]: epoch 0-step 2850630 loss: 0.645463 (accumulated 1 forwards) [2026-07-09 23:36:10-INFO]: epoch 0-step 2850640 loss: 0.631469 (accumulated 1 forwards) [2026-07-09 23:36:19-INFO]: epoch 0-step 2850650 loss: 0.679902 (accumulated 1 forwards) [2026-07-09 23:36:28-INFO]: epoch 0-step 2850660 loss: 0.675725 (accumulated 1 forwards) [2026-07-09 23:36:36-INFO]: epoch 0-step 2850670 loss: 0.677878 (accumulated 1 forwards) [2026-07-09 23:36:45-INFO]: epoch 0-step 2850680 loss: 0.687418 (accumulated 1 forwards) [2026-07-09 23:36:54-INFO]: epoch 0-step 2850690 loss: 0.659089 (accumulated 1 forwards) [2026-07-09 23:37:03-INFO]: epoch 0-step 2850700 loss: 0.662343 (accumulated 1 forwards) [2026-07-09 23:37:11-INFO]: epoch 0-step 2850710 loss: 0.681441 (accumulated 1 forwards) [2026-07-09 23:37:20-INFO]: epoch 0-step 2850720 loss: 0.647209 (accumulated 1 forwards) [2026-07-09 23:37:29-INFO]: epoch 0-step 2850730 loss: 0.664078 (accumulated 1 forwards) [2026-07-09 23:37:37-INFO]: epoch 0-step 2850740 loss: 0.647848 (accumulated 1 forwards) [2026-07-09 23:37:46-INFO]: epoch 0-step 2850750 loss: 0.677736 (accumulated 1 forwards) [2026-07-09 23:37:55-INFO]: epoch 0-step 2850760 loss: 0.601223 (accumulated 1 forwards) [2026-07-09 23:38:03-INFO]: epoch 0-step 2850770 loss: 0.677925 (accumulated 1 forwards) [2026-07-09 23:38:12-INFO]: epoch 0-step 2850780 loss: 0.662565 (accumulated 1 forwards) [2026-07-09 23:38:21-INFO]: epoch 0-step 2850790 loss: 0.679183 (accumulated 1 forwards) [2026-07-09 23:38:30-INFO]: epoch 0-step 2850800 loss: 0.680473 (accumulated 1 forwards) [2026-07-09 23:38:38-INFO]: epoch 0-step 2850810 loss: 0.658567 (accumulated 1 forwards) [2026-07-09 23:38:47-INFO]: epoch 0-step 2850820 loss: 0.658047 (accumulated 1 forwards) [2026-07-09 23:38:56-INFO]: epoch 0-step 2850830 loss: 0.689277 (accumulated 1 forwards) [2026-07-09 23:39:04-INFO]: epoch 0-step 2850840 loss: 0.700831 (accumulated 1 forwards) [2026-07-09 23:39:13-INFO]: epoch 0-step 2850850 loss: 0.671967 (accumulated 1 forwards) [2026-07-09 23:39:22-INFO]: epoch 0-step 2850860 loss: 0.687738 (accumulated 1 forwards) [2026-07-09 23:39:30-INFO]: epoch 0-step 2850870 loss: 0.657041 (accumulated 1 forwards) [2026-07-09 23:39:39-INFO]: epoch 0-step 2850880 loss: 0.675531 (accumulated 1 forwards) [2026-07-09 23:39:48-INFO]: epoch 0-step 2850890 loss: 0.662693 (accumulated 1 forwards) [2026-07-09 23:39:56-INFO]: epoch 0-step 2850900 loss: 0.668912 (accumulated 1 forwards) [2026-07-09 23:40:05-INFO]: epoch 0-step 2850910 loss: 0.667831 (accumulated 1 forwards) [2026-07-09 23:40:14-INFO]: epoch 0-step 2850920 loss: 0.679928 (accumulated 1 forwards) [2026-07-09 23:40:22-INFO]: epoch 0-step 2850930 loss: 0.656911 (accumulated 1 forwards) [2026-07-09 23:40:31-INFO]: epoch 0-step 2850940 loss: 0.638508 (accumulated 1 forwards) [2026-07-09 23:40:40-INFO]: epoch 0-step 2850950 loss: 0.680083 (accumulated 1 forwards) [2026-07-09 23:40:48-INFO]: epoch 0-step 2850960 loss: 0.678730 (accumulated 1 forwards) [2026-07-09 23:40:57-INFO]: epoch 0-step 2850970 loss: 0.645813 (accumulated 1 forwards) [2026-07-09 23:41:06-INFO]: epoch 0-step 2850980 loss: 0.681090 (accumulated 1 forwards) [2026-07-09 23:41:14-INFO]: epoch 0-step 2850990 loss: 0.649869 (accumulated 1 forwards) [2026-07-09 23:41:23-INFO]: epoch 0-step 2851000 loss: 0.701333 (accumulated 1 forwards) [2026-07-09 23:41:32-INFO]: epoch 0-step 2851010 loss: 0.661959 (accumulated 1 forwards) [2026-07-09 23:41:41-INFO]: epoch 0-step 2851020 loss: 0.665250 (accumulated 1 forwards) [2026-07-09 23:41:50-INFO]: epoch 0-step 2851030 loss: 0.663760 (accumulated 1 forwards) [2026-07-09 23:41:58-INFO]: epoch 0-step 2851040 loss: 0.698233 (accumulated 1 forwards) [2026-07-09 23:42:07-INFO]: epoch 0-step 2851050 loss: 0.692435 (accumulated 1 forwards) [2026-07-09 23:42:16-INFO]: epoch 0-step 2851060 loss: 0.666651 (accumulated 1 forwards) [2026-07-09 23:42:24-INFO]: epoch 0-step 2851070 loss: 0.706742 (accumulated 1 forwards) [2026-07-09 23:42:33-INFO]: epoch 0-step 2851080 loss: 0.659218 (accumulated 1 forwards) [2026-07-09 23:42:42-INFO]: epoch 0-step 2851090 loss: 0.683168 (accumulated 1 forwards) [2026-07-09 23:42:50-INFO]: epoch 0-step 2851100 loss: 0.625269 (accumulated 1 forwards) [2026-07-09 23:42:59-INFO]: epoch 0-step 2851110 loss: 0.682269 (accumulated 1 forwards) [2026-07-09 23:43:08-INFO]: epoch 0-step 2851120 loss: 0.657013 (accumulated 1 forwards) [2026-07-09 23:43:16-INFO]: epoch 0-step 2851130 loss: 0.671570 (accumulated 1 forwards) [2026-07-09 23:43:25-INFO]: epoch 0-step 2851140 loss: 0.672832 (accumulated 1 forwards) [2026-07-09 23:43:34-INFO]: epoch 0-step 2851150 loss: 0.671817 (accumulated 1 forwards) [2026-07-09 23:43:42-INFO]: epoch 0-step 2851160 loss: 0.659801 (accumulated 1 forwards) [2026-07-09 23:43:51-INFO]: epoch 0-step 2851170 loss: 0.615923 (accumulated 1 forwards) [2026-07-09 23:44:00-INFO]: epoch 0-step 2851180 loss: 0.696709 (accumulated 1 forwards) [2026-07-09 23:44:08-INFO]: epoch 0-step 2851190 loss: 0.673054 (accumulated 1 forwards) [2026-07-09 23:44:17-INFO]: epoch 0-step 2851200 loss: 0.704690 (accumulated 1 forwards) [2026-07-09 23:44:26-INFO]: epoch 0-step 2851210 loss: 0.653251 (accumulated 1 forwards) [2026-07-09 23:44:27-INFO]: Reload dataloader [2026-07-09 23:44:34-INFO]: epoch 1-step 2851220 loss: 0.649584 (accumulated 1 forwards) [2026-07-09 23:44:43-INFO]: epoch 1-step 2851230 loss: 0.701414 (accumulated 1 forwards) [2026-07-09 23:44:52-INFO]: epoch 1-step 2851240 loss: 0.642353 (accumulated 1 forwards) [2026-07-09 23:45:00-INFO]: epoch 1-step 2851250 loss: 0.686894 (accumulated 1 forwards) [2026-07-09 23:45:09-INFO]: epoch 1-step 2851260 loss: 0.657154 (accumulated 1 forwards) [2026-07-09 23:45:18-INFO]: epoch 1-step 2851270 loss: 0.671660 (accumulated 1 forwards) [2026-07-09 23:45:27-INFO]: epoch 1-step 2851280 loss: 0.671286 (accumulated 1 forwards) [2026-07-09 23:45:35-INFO]: epoch 1-step 2851290 loss: 0.682168 (accumulated 1 forwards) [2026-07-09 23:45:44-INFO]: epoch 1-step 2851300 loss: 0.657327 (accumulated 1 forwards) [2026-07-09 23:45:53-INFO]: epoch 1-step 2851310 loss: 0.666899 (accumulated 1 forwards) [2026-07-09 23:46:01-INFO]: epoch 1-step 2851320 loss: 0.689140 (accumulated 1 forwards) [2026-07-09 23:46:10-INFO]: epoch 1-step 2851330 loss: 0.655448 (accumulated 1 forwards) [2026-07-09 23:46:18-INFO]: epoch 1-step 2851340 loss: 0.669294 (accumulated 1 forwards) [2026-07-09 23:46:27-INFO]: epoch 1-step 2851350 loss: 0.676953 (accumulated 1 forwards) [2026-07-09 23:46:36-INFO]: epoch 1-step 2851360 loss: 0.679624 (accumulated 1 forwards) [2026-07-09 23:46:45-INFO]: epoch 1-step 2851370 loss: 0.663859 (accumulated 1 forwards) [2026-07-09 23:46:53-INFO]: epoch 1-step 2851380 loss: 0.628767 (accumulated 1 forwards) [2026-07-09 23:47:02-INFO]: epoch 1-step 2851390 loss: 0.666205 (accumulated 1 forwards) [2026-07-09 23:47:11-INFO]: epoch 1-step 2851400 loss: 0.691441 (accumulated 1 forwards) [2026-07-09 23:47:19-INFO]: epoch 1-step 2851410 loss: 0.659715 (accumulated 1 forwards) [2026-07-09 23:47:28-INFO]: epoch 1-step 2851420 loss: 0.670502 (accumulated 1 forwards) [2026-07-09 23:47:37-INFO]: epoch 1-step 2851430 loss: 0.669018 (accumulated 1 forwards) [2026-07-09 23:47:46-INFO]: epoch 1-step 2851440 loss: 0.658530 (accumulated 1 forwards) [2026-07-09 23:47:54-INFO]: epoch 1-step 2851450 loss: 0.679375 (accumulated 1 forwards) [2026-07-09 23:48:03-INFO]: epoch 1-step 2851460 loss: 0.688114 (accumulated 1 forwards) [2026-07-09 23:48:12-INFO]: epoch 1-step 2851470 loss: 0.657929 (accumulated 1 forwards) [2026-07-09 23:48:20-INFO]: epoch 1-step 2851480 loss: 0.682765 (accumulated 1 forwards) [2026-07-09 23:48:29-INFO]: epoch 1-step 2851490 loss: 0.718915 (accumulated 1 forwards) [2026-07-09 23:48:38-INFO]: epoch 1-step 2851500 loss: 0.695140 (accumulated 1 forwards) [2026-07-09 23:48:46-INFO]: epoch 1-step 2851510 loss: 0.667126 (accumulated 1 forwards) [2026-07-09 23:48:55-INFO]: epoch 1-step 2851520 loss: 0.702512 (accumulated 1 forwards) [2026-07-09 23:49:04-INFO]: epoch 1-step 2851530 loss: 0.685131 (accumulated 1 forwards) [2026-07-09 23:49:12-INFO]: epoch 1-step 2851540 loss: 0.670125 (accumulated 1 forwards) [2026-07-09 23:49:21-INFO]: epoch 1-step 2851550 loss: 0.605290 (accumulated 1 forwards) [2026-07-09 23:49:30-INFO]: epoch 1-step 2851560 loss: 0.679399 (accumulated 1 forwards) [2026-07-09 23:49:39-INFO]: epoch 1-step 2851570 loss: 0.627033 (accumulated 1 forwards) [2026-07-09 23:49:47-INFO]: epoch 1-step 2851580 loss: 0.625897 (accumulated 1 forwards) [2026-07-09 23:49:56-INFO]: epoch 1-step 2851590 loss: 0.644934 (accumulated 1 forwards) [2026-07-09 23:50:05-INFO]: epoch 1-step 2851600 loss: 0.691026 (accumulated 1 forwards) [2026-07-09 23:50:13-INFO]: epoch 1-step 2851610 loss: 0.635656 (accumulated 1 forwards) [2026-07-09 23:50:22-INFO]: epoch 1-step 2851620 loss: 0.692984 (accumulated 1 forwards) [2026-07-09 23:50:31-INFO]: epoch 1-step 2851630 loss: 0.658039 (accumulated 1 forwards) [2026-07-09 23:50:39-INFO]: epoch 1-step 2851640 loss: 0.698038 (accumulated 1 forwards) [2026-07-09 23:50:48-INFO]: epoch 1-step 2851650 loss: 0.659279 (accumulated 1 forwards) [2026-07-09 23:50:57-INFO]: epoch 1-step 2851660 loss: 0.672485 (accumulated 1 forwards) [2026-07-09 23:51:05-INFO]: epoch 1-step 2851670 loss: 0.666749 (accumulated 1 forwards) [2026-07-09 23:51:14-INFO]: epoch 1-step 2851680 loss: 0.666091 (accumulated 1 forwards) [2026-07-09 23:51:23-INFO]: epoch 1-step 2851690 loss: 0.660731 (accumulated 1 forwards) [2026-07-09 23:51:31-INFO]: epoch 1-step 2851700 loss: 0.694114 (accumulated 1 forwards) [2026-07-09 23:51:40-INFO]: epoch 1-step 2851710 loss: 0.661623 (accumulated 1 forwards) [2026-07-09 23:51:49-INFO]: epoch 1-step 2851720 loss: 0.672848 (accumulated 1 forwards) [2026-07-09 23:51:57-INFO]: epoch 1-step 2851730 loss: 0.660407 (accumulated 1 forwards) [2026-07-09 23:52:06-INFO]: epoch 1-step 2851740 loss: 0.686725 (accumulated 1 forwards) [2026-07-09 23:52:15-INFO]: epoch 1-step 2851750 loss: 0.693059 (accumulated 1 forwards) [2026-07-09 23:52:24-INFO]: epoch 1-step 2851760 loss: 0.692372 (accumulated 1 forwards) [2026-07-09 23:52:32-INFO]: epoch 1-step 2851770 loss: 0.680659 (accumulated 1 forwards) [2026-07-09 23:52:41-INFO]: epoch 1-step 2851780 loss: 0.645750 (accumulated 1 forwards) [2026-07-09 23:52:50-INFO]: epoch 1-step 2851790 loss: 0.642455 (accumulated 1 forwards) [2026-07-09 23:52:58-INFO]: epoch 1-step 2851800 loss: 0.658305 (accumulated 1 forwards) [2026-07-09 23:53:07-INFO]: epoch 1-step 2851810 loss: 0.642389 (accumulated 1 forwards) [2026-07-09 23:53:16-INFO]: epoch 1-step 2851820 loss: 0.679519 (accumulated 1 forwards) [2026-07-09 23:53:24-INFO]: epoch 1-step 2851830 loss: 0.668576 (accumulated 1 forwards) [2026-07-09 23:53:33-INFO]: epoch 1-step 2851840 loss: 0.680239 (accumulated 1 forwards) [2026-07-09 23:53:42-INFO]: epoch 1-step 2851850 loss: 0.650130 (accumulated 1 forwards) [2026-07-09 23:53:50-INFO]: epoch 1-step 2851860 loss: 0.674428 (accumulated 1 forwards) [2026-07-09 23:53:59-INFO]: epoch 1-step 2851870 loss: 0.666429 (accumulated 1 forwards) [2026-07-09 23:54:08-INFO]: epoch 1-step 2851880 loss: 0.680725 (accumulated 1 forwards) [2026-07-09 23:54:17-INFO]: epoch 1-step 2851890 loss: 0.695785 (accumulated 1 forwards) [2026-07-09 23:54:25-INFO]: epoch 1-step 2851900 loss: 0.685835 (accumulated 1 forwards) [2026-07-09 23:54:34-INFO]: epoch 1-step 2851910 loss: 0.656854 (accumulated 1 forwards) [2026-07-09 23:54:43-INFO]: epoch 1-step 2851920 loss: 0.663359 (accumulated 1 forwards) [2026-07-09 23:54:51-INFO]: epoch 1-step 2851930 loss: 0.670374 (accumulated 1 forwards) [2026-07-09 23:55:00-INFO]: epoch 1-step 2851940 loss: 0.638248 (accumulated 1 forwards) [2026-07-09 23:55:09-INFO]: epoch 1-step 2851950 loss: 0.692661 (accumulated 1 forwards) [2026-07-09 23:55:17-INFO]: epoch 1-step 2851960 loss: 0.651248 (accumulated 1 forwards) [2026-07-09 23:55:26-INFO]: epoch 1-step 2851970 loss: 0.641746 (accumulated 1 forwards) [2026-07-09 23:55:35-INFO]: epoch 1-step 2851980 loss: 0.657161 (accumulated 1 forwards) [2026-07-09 23:55:44-INFO]: epoch 1-step 2851990 loss: 0.649337 (accumulated 1 forwards) [2026-07-09 23:55:52-INFO]: epoch 1-step 2852000 loss: 0.666648 (accumulated 1 forwards) [2026-07-09 23:56:01-INFO]: epoch 1-step 2852010 loss: 0.694436 (accumulated 1 forwards) [2026-07-09 23:56:10-INFO]: epoch 1-step 2852020 loss: 0.683304 (accumulated 1 forwards) [2026-07-09 23:56:18-INFO]: epoch 1-step 2852030 loss: 0.667584 (accumulated 1 forwards) [2026-07-09 23:56:27-INFO]: epoch 1-step 2852040 loss: 0.659562 (accumulated 1 forwards) [2026-07-09 23:56:36-INFO]: epoch 1-step 2852050 loss: 0.653942 (accumulated 1 forwards) [2026-07-09 23:56:44-INFO]: epoch 1-step 2852060 loss: 0.710867 (accumulated 1 forwards) [2026-07-09 23:56:53-INFO]: epoch 1-step 2852070 loss: 0.690067 (accumulated 1 forwards) [2026-07-09 23:57:02-INFO]: epoch 1-step 2852080 loss: 0.665861 (accumulated 1 forwards) [2026-07-09 23:57:10-INFO]: epoch 1-step 2852090 loss: 0.621744 (accumulated 1 forwards) [2026-07-09 23:57:19-INFO]: epoch 1-step 2852100 loss: 0.652122 (accumulated 1 forwards) [2026-07-09 23:57:28-INFO]: epoch 1-step 2852110 loss: 0.633639 (accumulated 1 forwards) [2026-07-09 23:57:36-INFO]: epoch 1-step 2852120 loss: 0.638626 (accumulated 1 forwards) [2026-07-09 23:57:45-INFO]: epoch 1-step 2852130 loss: 0.690387 (accumulated 1 forwards) [2026-07-09 23:57:54-INFO]: epoch 1-step 2852140 loss: 0.631431 (accumulated 1 forwards) [2026-07-09 23:58:02-INFO]: epoch 1-step 2852150 loss: 0.672076 (accumulated 1 forwards) [2026-07-09 23:58:11-INFO]: epoch 1-step 2852160 loss: 0.677734 (accumulated 1 forwards) [2026-07-09 23:58:20-INFO]: epoch 1-step 2852170 loss: 0.663861 (accumulated 1 forwards) [2026-07-09 23:58:28-INFO]: epoch 1-step 2852180 loss: 0.644674 (accumulated 1 forwards) [2026-07-09 23:58:37-INFO]: epoch 1-step 2852190 loss: 0.667099 (accumulated 1 forwards) [2026-07-09 23:58:46-INFO]: epoch 1-step 2852200 loss: 0.630454 (accumulated 1 forwards) [2026-07-09 23:58:55-INFO]: epoch 1-step 2852210 loss: 0.667414 (accumulated 1 forwards) [2026-07-09 23:59:03-INFO]: epoch 1-step 2852220 loss: 0.647508 (accumulated 1 forwards) [2026-07-09 23:59:12-INFO]: epoch 1-step 2852230 loss: 0.681985 (accumulated 1 forwards) [2026-07-09 23:59:21-INFO]: epoch 1-step 2852240 loss: 0.682674 (accumulated 1 forwards) [2026-07-09 23:59:29-INFO]: epoch 1-step 2852250 loss: 0.656487 (accumulated 1 forwards) [2026-07-09 23:59:38-INFO]: epoch 1-step 2852260 loss: 0.676453 (accumulated 1 forwards) [2026-07-09 23:59:47-INFO]: epoch 1-step 2852270 loss: 0.634226 (accumulated 1 forwards) [2026-07-09 23:59:55-INFO]: epoch 1-step 2852280 loss: 0.654415 (accumulated 1 forwards) [2026-07-10 00:00:04-INFO]: epoch 1-step 2852290 loss: 0.717799 (accumulated 1 forwards) [2026-07-10 00:00:13-INFO]: epoch 1-step 2852300 loss: 0.631959 (accumulated 1 forwards) [2026-07-10 00:00:21-INFO]: epoch 1-step 2852310 loss: 0.646071 (accumulated 1 forwards) [2026-07-10 00:00:30-INFO]: epoch 1-step 2852320 loss: 0.683613 (accumulated 1 forwards) [2026-07-10 00:00:39-INFO]: epoch 1-step 2852330 loss: 0.632242 (accumulated 1 forwards) [2026-07-10 00:00:47-INFO]: epoch 1-step 2852340 loss: 0.677972 (accumulated 1 forwards) [2026-07-10 00:00:56-INFO]: epoch 1-step 2852350 loss: 0.637440 (accumulated 1 forwards) [2026-07-10 00:01:05-INFO]: epoch 1-step 2852360 loss: 0.664578 (accumulated 1 forwards) [2026-07-10 00:01:13-INFO]: epoch 1-step 2852370 loss: 0.678222 (accumulated 1 forwards) [2026-07-10 00:01:22-INFO]: epoch 1-step 2852380 loss: 0.669618 (accumulated 1 forwards) [2026-07-10 00:01:31-INFO]: epoch 1-step 2852390 loss: 0.708921 (accumulated 1 forwards) [2026-07-10 00:01:40-INFO]: epoch 1-step 2852400 loss: 0.653850 (accumulated 1 forwards) [2026-07-10 00:01:48-INFO]: epoch 1-step 2852410 loss: 0.627545 (accumulated 1 forwards) [2026-07-10 00:01:57-INFO]: epoch 1-step 2852420 loss: 0.637277 (accumulated 1 forwards) [2026-07-10 00:02:00-INFO]: Reload dataloader [2026-07-10 00:02:06-INFO]: epoch 2-step 2852430 loss: 0.641550 (accumulated 1 forwards) [2026-07-10 00:02:14-INFO]: epoch 2-step 2852440 loss: 0.680952 (accumulated 1 forwards) [2026-07-10 00:02:23-INFO]: epoch 2-step 2852450 loss: 0.688238 (accumulated 1 forwards) [2026-07-10 00:02:32-INFO]: epoch 2-step 2852460 loss: 0.682103 (accumulated 1 forwards) [2026-07-10 00:02:40-INFO]: epoch 2-step 2852470 loss: 0.703644 (accumulated 1 forwards) [2026-07-10 00:02:49-INFO]: epoch 2-step 2852480 loss: 0.659114 (accumulated 1 forwards) [2026-07-10 00:02:58-INFO]: epoch 2-step 2852490 loss: 0.677621 (accumulated 1 forwards) [2026-07-10 00:03:06-INFO]: epoch 2-step 2852500 loss: 0.650947 (accumulated 1 forwards) [2026-07-10 00:03:16-INFO]: epoch 2-step 2852510 loss: 0.664498 (accumulated 1 forwards) [2026-07-10 00:03:24-INFO]: epoch 2-step 2852520 loss: 0.701335 (accumulated 1 forwards) [2026-07-10 00:03:33-INFO]: epoch 2-step 2852530 loss: 0.674314 (accumulated 1 forwards) [2026-07-10 00:03:42-INFO]: epoch 2-step 2852540 loss: 0.692408 (accumulated 1 forwards) [2026-07-10 00:03:51-INFO]: epoch 2-step 2852550 loss: 0.681899 (accumulated 1 forwards) [2026-07-10 00:03:59-INFO]: epoch 2-step 2852560 loss: 0.629464 (accumulated 1 forwards) [2026-07-10 00:04:08-INFO]: epoch 2-step 2852570 loss: 0.669701 (accumulated 1 forwards) [2026-07-10 00:04:17-INFO]: epoch 2-step 2852580 loss: 0.681332 (accumulated 1 forwards) [2026-07-10 00:04:25-INFO]: epoch 2-step 2852590 loss: 0.713940 (accumulated 1 forwards) [2026-07-10 00:04:34-INFO]: epoch 2-step 2852600 loss: 0.655346 (accumulated 1 forwards) [2026-07-10 00:04:43-INFO]: epoch 2-step 2852610 loss: 0.688904 (accumulated 1 forwards) [2026-07-10 00:04:51-INFO]: epoch 2-step 2852620 loss: 0.676146 (accumulated 1 forwards) [2026-07-10 00:05:00-INFO]: epoch 2-step 2852630 loss: 0.688144 (accumulated 1 forwards) [2026-07-10 00:05:09-INFO]: epoch 2-step 2852640 loss: 0.664200 (accumulated 1 forwards) [2026-07-10 00:05:17-INFO]: epoch 2-step 2852650 loss: 0.673884 (accumulated 1 forwards) [2026-07-10 00:05:26-INFO]: epoch 2-step 2852660 loss: 0.701664 (accumulated 1 forwards) [2026-07-10 00:05:35-INFO]: epoch 2-step 2852670 loss: 0.646821 (accumulated 1 forwards) [2026-07-10 00:05:43-INFO]: epoch 2-step 2852680 loss: 0.656727 (accumulated 1 forwards) [2026-07-10 00:05:52-INFO]: epoch 2-step 2852690 loss: 0.643913 (accumulated 1 forwards) [2026-07-10 00:06:01-INFO]: epoch 2-step 2852700 loss: 0.640162 (accumulated 1 forwards) [2026-07-10 00:06:09-INFO]: epoch 2-step 2852710 loss: 0.642010 (accumulated 1 forwards) [2026-07-10 00:06:18-INFO]: epoch 2-step 2852720 loss: 0.676086 (accumulated 1 forwards) [2026-07-10 00:06:27-INFO]: epoch 2-step 2852730 loss: 0.686944 (accumulated 1 forwards) [2026-07-10 00:06:35-INFO]: epoch 2-step 2852740 loss: 0.720548 (accumulated 1 forwards) [2026-07-10 00:06:44-INFO]: epoch 2-step 2852750 loss: 0.632605 (accumulated 1 forwards) [2026-07-10 00:06:53-INFO]: epoch 2-step 2852760 loss: 0.680690 (accumulated 1 forwards) [2026-07-10 00:07:02-INFO]: epoch 2-step 2852770 loss: 0.674203 (accumulated 1 forwards) [2026-07-10 00:07:10-INFO]: epoch 2-step 2852780 loss: 0.660907 (accumulated 1 forwards) [2026-07-10 00:07:19-INFO]: epoch 2-step 2852790 loss: 0.640362 (accumulated 1 forwards) [2026-07-10 00:07:28-INFO]: epoch 2-step 2852800 loss: 0.621139 (accumulated 1 forwards) [2026-07-10 00:07:36-INFO]: epoch 2-step 2852810 loss: 0.699381 (accumulated 1 forwards) [2026-07-10 00:07:45-INFO]: epoch 2-step 2852820 loss: 0.689039 (accumulated 1 forwards) [2026-07-10 00:07:54-INFO]: epoch 2-step 2852830 loss: 0.665049 (accumulated 1 forwards) [2026-07-10 00:08:02-INFO]: epoch 2-step 2852840 loss: 0.675410 (accumulated 1 forwards) [2026-07-10 00:08:11-INFO]: epoch 2-step 2852850 loss: 0.643073 (accumulated 1 forwards) [2026-07-10 00:08:20-INFO]: epoch 2-step 2852860 loss: 0.623141 (accumulated 1 forwards) [2026-07-10 00:08:28-INFO]: epoch 2-step 2852870 loss: 0.674719 (accumulated 1 forwards) [2026-07-10 00:08:37-INFO]: epoch 2-step 2852880 loss: 0.667887 (accumulated 1 forwards) [2026-07-10 00:08:46-INFO]: epoch 2-step 2852890 loss: 0.657935 (accumulated 1 forwards) [2026-07-10 00:08:54-INFO]: epoch 2-step 2852900 loss: 0.696871 (accumulated 1 forwards) [2026-07-10 00:09:03-INFO]: epoch 2-step 2852910 loss: 0.641236 (accumulated 1 forwards) [2026-07-10 00:09:12-INFO]: epoch 2-step 2852920 loss: 0.656599 (accumulated 1 forwards) [2026-07-10 00:09:21-INFO]: epoch 2-step 2852930 loss: 0.660857 (accumulated 1 forwards) [2026-07-10 00:09:29-INFO]: epoch 2-step 2852940 loss: 0.645705 (accumulated 1 forwards) [2026-07-10 00:09:38-INFO]: epoch 2-step 2852950 loss: 0.695768 (accumulated 1 forwards) [2026-07-10 00:09:47-INFO]: epoch 2-step 2852960 loss: 0.687312 (accumulated 1 forwards) [2026-07-10 00:09:55-INFO]: epoch 2-step 2852970 loss: 0.637966 (accumulated 1 forwards) [2026-07-10 00:10:04-INFO]: epoch 2-step 2852980 loss: 0.658736 (accumulated 1 forwards) [2026-07-10 00:10:13-INFO]: epoch 2-step 2852990 loss: 0.682380 (accumulated 1 forwards) [2026-07-10 00:10:21-INFO]: epoch 2-step 2853000 loss: 0.634490 (accumulated 1 forwards) [2026-07-10 00:10:30-INFO]: epoch 2-step 2853010 loss: 0.642231 (accumulated 1 forwards) [2026-07-10 00:10:39-INFO]: epoch 2-step 2853020 loss: 0.657896 (accumulated 1 forwards) [2026-07-10 00:10:47-INFO]: epoch 2-step 2853030 loss: 0.676702 (accumulated 1 forwards) [2026-07-10 00:10:56-INFO]: epoch 2-step 2853040 loss: 0.677230 (accumulated 1 forwards) [2026-07-10 00:11:05-INFO]: epoch 2-step 2853050 loss: 0.655737 (accumulated 1 forwards) [2026-07-10 00:11:14-INFO]: epoch 2-step 2853060 loss: 0.665005 (accumulated 1 forwards) [2026-07-10 00:11:22-INFO]: epoch 2-step 2853070 loss: 0.709944 (accumulated 1 forwards) [2026-07-10 00:11:31-INFO]: epoch 2-step 2853080 loss: 0.679157 (accumulated 1 forwards) [2026-07-10 00:11:40-INFO]: epoch 2-step 2853090 loss: 0.630660 (accumulated 1 forwards) [2026-07-10 00:11:48-INFO]: epoch 2-step 2853100 loss: 0.697320 (accumulated 1 forwards) [2026-07-10 00:11:57-INFO]: epoch 2-step 2853110 loss: 0.693461 (accumulated 1 forwards) [2026-07-10 00:12:06-INFO]: epoch 2-step 2853120 loss: 0.659397 (accumulated 1 forwards) [2026-07-10 00:12:14-INFO]: epoch 2-step 2853130 loss: 0.604189 (accumulated 1 forwards) [2026-07-10 00:12:23-INFO]: epoch 2-step 2853140 loss: 0.643605 (accumulated 1 forwards) [2026-07-10 00:12:32-INFO]: epoch 2-step 2853150 loss: 0.652351 (accumulated 1 forwards) [2026-07-10 00:12:40-INFO]: epoch 2-step 2853160 loss: 0.673358 (accumulated 1 forwards) [2026-07-10 00:12:49-INFO]: epoch 2-step 2853170 loss: 0.653328 (accumulated 1 forwards) [2026-07-10 00:12:58-INFO]: epoch 2-step 2853180 loss: 0.679196 (accumulated 1 forwards) [2026-07-10 00:13:07-INFO]: epoch 2-step 2853190 loss: 0.693981 (accumulated 1 forwards) [2026-07-10 00:13:15-INFO]: epoch 2-step 2853200 loss: 0.615904 (accumulated 1 forwards) [2026-07-10 00:13:24-INFO]: epoch 2-step 2853210 loss: 0.659471 (accumulated 1 forwards) [2026-07-10 00:13:33-INFO]: epoch 2-step 2853220 loss: 0.680878 (accumulated 1 forwards) [2026-07-10 00:13:41-INFO]: epoch 2-step 2853230 loss: 0.668947 (accumulated 1 forwards) [2026-07-10 00:13:50-INFO]: epoch 2-step 2853240 loss: 0.690468 (accumulated 1 forwards) [2026-07-10 00:13:59-INFO]: epoch 2-step 2853250 loss: 0.683837 (accumulated 1 forwards) [2026-07-10 00:14:07-INFO]: epoch 2-step 2853260 loss: 0.673643 (accumulated 1 forwards) [2026-07-10 00:14:16-INFO]: epoch 2-step 2853270 loss: 0.692614 (accumulated 1 forwards) [2026-07-10 00:14:25-INFO]: epoch 2-step 2853280 loss: 0.651842 (accumulated 1 forwards) [2026-07-10 00:14:33-INFO]: epoch 2-step 2853290 loss: 0.649326 (accumulated 1 forwards) [2026-07-10 00:14:42-INFO]: epoch 2-step 2853300 loss: 0.621272 (accumulated 1 forwards) [2026-07-10 00:14:51-INFO]: epoch 2-step 2853310 loss: 0.697186 (accumulated 1 forwards) [2026-07-10 00:14:59-INFO]: epoch 2-step 2853320 loss: 0.649735 (accumulated 1 forwards) [2026-07-10 00:15:08-INFO]: epoch 2-step 2853330 loss: 0.693612 (accumulated 1 forwards) [2026-07-10 00:15:17-INFO]: epoch 2-step 2853340 loss: 0.651694 (accumulated 1 forwards) [2026-07-10 00:15:26-INFO]: epoch 2-step 2853350 loss: 0.670580 (accumulated 1 forwards) [2026-07-10 00:15:34-INFO]: epoch 2-step 2853360 loss: 0.633523 (accumulated 1 forwards) [2026-07-10 00:15:43-INFO]: epoch 2-step 2853370 loss: 0.689650 (accumulated 1 forwards) [2026-07-10 00:15:52-INFO]: epoch 2-step 2853380 loss: 0.698577 (accumulated 1 forwards) [2026-07-10 00:16:00-INFO]: epoch 2-step 2853390 loss: 0.648090 (accumulated 1 forwards) [2026-07-10 00:16:09-INFO]: epoch 2-step 2853400 loss: 0.642662 (accumulated 1 forwards) [2026-07-10 00:16:18-INFO]: epoch 2-step 2853410 loss: 0.677509 (accumulated 1 forwards) [2026-07-10 00:16:26-INFO]: epoch 2-step 2853420 loss: 0.662814 (accumulated 1 forwards) [2026-07-10 00:16:35-INFO]: epoch 2-step 2853430 loss: 0.645383 (accumulated 1 forwards) [2026-07-10 00:16:44-INFO]: epoch 2-step 2853440 loss: 0.690120 (accumulated 1 forwards) [2026-07-10 00:16:53-INFO]: epoch 2-step 2853450 loss: 0.679610 (accumulated 1 forwards) [2026-07-10 00:17:01-INFO]: epoch 2-step 2853460 loss: 0.636654 (accumulated 1 forwards) [2026-07-10 00:17:10-INFO]: epoch 2-step 2853470 loss: 0.658187 (accumulated 1 forwards) [2026-07-10 00:17:19-INFO]: epoch 2-step 2853480 loss: 0.695120 (accumulated 1 forwards) [2026-07-10 00:17:27-INFO]: epoch 2-step 2853490 loss: 0.698900 (accumulated 1 forwards) [2026-07-10 00:17:36-INFO]: epoch 2-step 2853500 loss: 0.645826 (accumulated 1 forwards) [2026-07-10 00:17:45-INFO]: epoch 2-step 2853510 loss: 0.686962 (accumulated 1 forwards) [2026-07-10 00:17:53-INFO]: epoch 2-step 2853520 loss: 0.681828 (accumulated 1 forwards) [2026-07-10 00:18:02-INFO]: epoch 2-step 2853530 loss: 0.626280 (accumulated 1 forwards) [2026-07-10 00:18:11-INFO]: epoch 2-step 2853540 loss: 0.703973 (accumulated 1 forwards) [2026-07-10 00:18:19-INFO]: epoch 2-step 2853550 loss: 0.644192 (accumulated 1 forwards) [2026-07-10 00:18:28-INFO]: epoch 2-step 2853560 loss: 0.656991 (accumulated 1 forwards) [2026-07-10 00:18:37-INFO]: epoch 2-step 2853570 loss: 0.601190 (accumulated 1 forwards) [2026-07-10 00:18:45-INFO]: epoch 2-step 2853580 loss: 0.633264 (accumulated 1 forwards) [2026-07-10 00:18:54-INFO]: epoch 2-step 2853590 loss: 0.665747 (accumulated 1 forwards) [2026-07-10 00:19:03-INFO]: epoch 2-step 2853600 loss: 0.667846 (accumulated 1 forwards) [2026-07-10 00:19:12-INFO]: epoch 2-step 2853610 loss: 0.659209 (accumulated 1 forwards) [2026-07-10 00:19:20-INFO]: epoch 2-step 2853620 loss: 0.693967 (accumulated 1 forwards) [2026-07-10 00:19:29-INFO]: epoch 2-step 2853630 loss: 0.676542 (accumulated 1 forwards) [2026-07-10 00:19:34-INFO]: Reload dataloader [2026-07-10 00:19:37-INFO]: epoch 3-step 2853640 loss: 0.685542 (accumulated 1 forwards) [2026-07-10 00:19:46-INFO]: epoch 3-step 2853650 loss: 0.691512 (accumulated 1 forwards) [2026-07-10 00:19:55-INFO]: epoch 3-step 2853660 loss: 0.680955 (accumulated 1 forwards) [2026-07-10 00:20:04-INFO]: epoch 3-step 2853670 loss: 0.651255 (accumulated 1 forwards) [2026-07-10 00:20:12-INFO]: epoch 3-step 2853680 loss: 0.676672 (accumulated 1 forwards) [2026-07-10 00:20:21-INFO]: epoch 3-step 2853690 loss: 0.622645 (accumulated 1 forwards) [2026-07-10 00:20:30-INFO]: epoch 3-step 2853700 loss: 0.679066 (accumulated 1 forwards) [2026-07-10 00:20:39-INFO]: epoch 3-step 2853710 loss: 0.656724 (accumulated 1 forwards) [2026-07-10 00:20:47-INFO]: epoch 3-step 2853720 loss: 0.613348 (accumulated 1 forwards) [2026-07-10 00:20:56-INFO]: epoch 3-step 2853730 loss: 0.637188 (accumulated 1 forwards) [2026-07-10 00:21:05-INFO]: epoch 3-step 2853740 loss: 0.660232 (accumulated 1 forwards) [2026-07-10 00:21:13-INFO]: epoch 3-step 2853750 loss: 0.650317 (accumulated 1 forwards) [2026-07-10 00:21:22-INFO]: epoch 3-step 2853760 loss: 0.666622 (accumulated 1 forwards) [2026-07-10 00:21:31-INFO]: epoch 3-step 2853770 loss: 0.623075 (accumulated 1 forwards) [2026-07-10 00:21:39-INFO]: epoch 3-step 2853780 loss: 0.664501 (accumulated 1 forwards) [2026-07-10 00:21:48-INFO]: epoch 3-step 2853790 loss: 0.653015 (accumulated 1 forwards) [2026-07-10 00:21:57-INFO]: epoch 3-step 2853800 loss: 0.667583 (accumulated 1 forwards) [2026-07-10 00:22:05-INFO]: epoch 3-step 2853810 loss: 0.640655 (accumulated 1 forwards) [2026-07-10 00:22:14-INFO]: epoch 3-step 2853820 loss: 0.711021 (accumulated 1 forwards) [2026-07-10 00:22:23-INFO]: epoch 3-step 2853830 loss: 0.657505 (accumulated 1 forwards) [2026-07-10 00:22:31-INFO]: epoch 3-step 2853840 loss: 0.658787 (accumulated 1 forwards) [2026-07-10 00:22:40-INFO]: epoch 3-step 2853850 loss: 0.658737 (accumulated 1 forwards) [2026-07-10 00:22:49-INFO]: epoch 3-step 2853860 loss: 0.646469 (accumulated 1 forwards) [2026-07-10 00:22:57-INFO]: epoch 3-step 2853870 loss: 0.620184 (accumulated 1 forwards) [2026-07-10 00:23:06-INFO]: epoch 3-step 2853880 loss: 0.668243 (accumulated 1 forwards) [2026-07-10 00:23:15-INFO]: epoch 3-step 2853890 loss: 0.690348 (accumulated 1 forwards) [2026-07-10 00:23:24-INFO]: epoch 3-step 2853900 loss: 0.675704 (accumulated 1 forwards) [2026-07-10 00:23:32-INFO]: epoch 3-step 2853910 loss: 0.679673 (accumulated 1 forwards) [2026-07-10 00:23:41-INFO]: epoch 3-step 2853920 loss: 0.650238 (accumulated 1 forwards) [2026-07-10 00:23:50-INFO]: epoch 3-step 2853930 loss: 0.659775 (accumulated 1 forwards) [2026-07-10 00:23:58-INFO]: epoch 3-step 2853940 loss: 0.699015 (accumulated 1 forwards) [2026-07-10 00:24:07-INFO]: epoch 3-step 2853950 loss: 0.651520 (accumulated 1 forwards) [2026-07-10 00:24:16-INFO]: epoch 3-step 2853960 loss: 0.684399 (accumulated 1 forwards) [2026-07-10 00:24:24-INFO]: epoch 3-step 2853970 loss: 0.672646 (accumulated 1 forwards) [2026-07-10 00:24:33-INFO]: epoch 3-step 2853980 loss: 0.676235 (accumulated 1 forwards) [2026-07-10 00:24:42-INFO]: epoch 3-step 2853990 loss: 0.692912 (accumulated 1 forwards) [2026-07-10 00:24:50-INFO]: epoch 3-step 2854000 loss: 0.655088 (accumulated 1 forwards) [2026-07-10 00:24:59-INFO]: epoch 3-step 2854010 loss: 0.652171 (accumulated 1 forwards) [2026-07-10 00:25:08-INFO]: epoch 3-step 2854020 loss: 0.637971 (accumulated 1 forwards) [2026-07-10 00:25:16-INFO]: epoch 3-step 2854030 loss: 0.700007 (accumulated 1 forwards) [2026-07-10 00:25:25-INFO]: epoch 3-step 2854040 loss: 0.653277 (accumulated 1 forwards) [2026-07-10 00:25:34-INFO]: epoch 3-step 2854050 loss: 0.680266 (accumulated 1 forwards) [2026-07-10 00:25:43-INFO]: epoch 3-step 2854060 loss: 0.662854 (accumulated 1 forwards) [2026-07-10 00:25:51-INFO]: epoch 3-step 2854070 loss: 0.682226 (accumulated 1 forwards) [2026-07-10 00:26:00-INFO]: epoch 3-step 2854080 loss: 0.665325 (accumulated 1 forwards) [2026-07-10 00:26:09-INFO]: epoch 3-step 2854090 loss: 0.633324 (accumulated 1 forwards) [2026-07-10 00:26:17-INFO]: epoch 3-step 2854100 loss: 0.658331 (accumulated 1 forwards) [2026-07-10 00:26:26-INFO]: epoch 3-step 2854110 loss: 0.694964 (accumulated 1 forwards) [2026-07-10 00:26:35-INFO]: epoch 3-step 2854120 loss: 0.663226 (accumulated 1 forwards) [2026-07-10 00:26:43-INFO]: epoch 3-step 2854130 loss: 0.636079 (accumulated 1 forwards) [2026-07-10 00:26:52-INFO]: epoch 3-step 2854140 loss: 0.628322 (accumulated 1 forwards) [2026-07-10 00:27:01-INFO]: epoch 3-step 2854150 loss: 0.656669 (accumulated 1 forwards) [2026-07-10 00:27:09-INFO]: epoch 3-step 2854160 loss: 0.689410 (accumulated 1 forwards) [2026-07-10 00:27:18-INFO]: epoch 3-step 2854170 loss: 0.650138 (accumulated 1 forwards) [2026-07-10 00:27:27-INFO]: epoch 3-step 2854180 loss: 0.659211 (accumulated 1 forwards) [2026-07-10 00:27:35-INFO]: epoch 3-step 2854190 loss: 0.638071 (accumulated 1 forwards) [2026-07-10 00:27:44-INFO]: epoch 3-step 2854200 loss: 0.672839 (accumulated 1 forwards) [2026-07-10 00:27:53-INFO]: epoch 3-step 2854210 loss: 0.633229 (accumulated 1 forwards) [2026-07-10 00:28:01-INFO]: epoch 3-step 2854220 loss: 0.674135 (accumulated 1 forwards) [2026-07-10 00:28:10-INFO]: epoch 3-step 2854230 loss: 0.631549 (accumulated 1 forwards) [2026-07-10 00:28:19-INFO]: epoch 3-step 2854240 loss: 0.677416 (accumulated 1 forwards) [2026-07-10 00:28:28-INFO]: epoch 3-step 2854250 loss: 0.641300 (accumulated 1 forwards) [2026-07-10 00:28:36-INFO]: epoch 3-step 2854260 loss: 0.679384 (accumulated 1 forwards) [2026-07-10 00:28:45-INFO]: epoch 3-step 2854270 loss: 0.688937 (accumulated 1 forwards) [2026-07-10 00:28:54-INFO]: epoch 3-step 2854280 loss: 0.680465 (accumulated 1 forwards) [2026-07-10 00:29:02-INFO]: epoch 3-step 2854290 loss: 0.663593 (accumulated 1 forwards) [2026-07-10 00:29:11-INFO]: epoch 3-step 2854300 loss: 0.685417 (accumulated 1 forwards) [2026-07-10 00:29:20-INFO]: epoch 3-step 2854310 loss: 0.697477 (accumulated 1 forwards) [2026-07-10 00:29:28-INFO]: epoch 3-step 2854320 loss: 0.674898 (accumulated 1 forwards) [2026-07-10 00:29:37-INFO]: epoch 3-step 2854330 loss: 0.682722 (accumulated 1 forwards) [2026-07-10 00:29:46-INFO]: epoch 3-step 2854340 loss: 0.681333 (accumulated 1 forwards) [2026-07-10 00:29:54-INFO]: epoch 3-step 2854350 loss: 0.675255 (accumulated 1 forwards) [2026-07-10 00:30:03-INFO]: epoch 3-step 2854360 loss: 0.696865 (accumulated 1 forwards) [2026-07-10 00:30:12-INFO]: epoch 3-step 2854370 loss: 0.655999 (accumulated 1 forwards) [2026-07-10 00:30:21-INFO]: epoch 3-step 2854380 loss: 0.656460 (accumulated 1 forwards) [2026-07-10 00:30:29-INFO]: epoch 3-step 2854390 loss: 0.644361 (accumulated 1 forwards) [2026-07-10 00:30:38-INFO]: epoch 3-step 2854400 loss: 0.655217 (accumulated 1 forwards) [2026-07-10 00:30:47-INFO]: epoch 3-step 2854410 loss: 0.657204 (accumulated 1 forwards) [2026-07-10 00:30:56-INFO]: epoch 3-step 2854420 loss: 0.652695 (accumulated 1 forwards) [2026-07-10 00:31:04-INFO]: epoch 3-step 2854430 loss: 0.641149 (accumulated 1 forwards) [2026-07-10 00:31:13-INFO]: epoch 3-step 2854440 loss: 0.633215 (accumulated 1 forwards) [2026-07-10 00:31:22-INFO]: epoch 3-step 2854450 loss: 0.643309 (accumulated 1 forwards) [2026-07-10 00:31:30-INFO]: epoch 3-step 2854460 loss: 0.685451 (accumulated 1 forwards) [2026-07-10 00:31:39-INFO]: epoch 3-step 2854470 loss: 0.636323 (accumulated 1 forwards) [2026-07-10 00:31:48-INFO]: epoch 3-step 2854480 loss: 0.715108 (accumulated 1 forwards) [2026-07-10 00:31:56-INFO]: epoch 3-step 2854490 loss: 0.605080 (accumulated 1 forwards) [2026-07-10 00:32:05-INFO]: epoch 3-step 2854500 loss: 0.651419 (accumulated 1 forwards) [2026-07-10 00:32:14-INFO]: epoch 3-step 2854510 loss: 0.681072 (accumulated 1 forwards) [2026-07-10 00:32:22-INFO]: epoch 3-step 2854520 loss: 0.654462 (accumulated 1 forwards) [2026-07-10 00:32:31-INFO]: epoch 3-step 2854530 loss: 0.687473 (accumulated 1 forwards) [2026-07-10 00:32:40-INFO]: epoch 3-step 2854540 loss: 0.652642 (accumulated 1 forwards) [2026-07-10 00:32:48-INFO]: epoch 3-step 2854550 loss: 0.673329 (accumulated 1 forwards) [2026-07-10 00:32:57-INFO]: epoch 3-step 2854560 loss: 0.650323 (accumulated 1 forwards) [2026-07-10 00:33:06-INFO]: epoch 3-step 2854570 loss: 0.612895 (accumulated 1 forwards) [2026-07-10 00:33:14-INFO]: epoch 3-step 2854580 loss: 0.684497 (accumulated 1 forwards) [2026-07-10 00:33:23-INFO]: epoch 3-step 2854590 loss: 0.679064 (accumulated 1 forwards) [2026-07-10 00:33:32-INFO]: epoch 3-step 2854600 loss: 0.691902 (accumulated 1 forwards) [2026-07-10 00:33:41-INFO]: epoch 3-step 2854610 loss: 0.692721 (accumulated 1 forwards) [2026-07-10 00:33:49-INFO]: epoch 3-step 2854620 loss: 0.605248 (accumulated 1 forwards) [2026-07-10 00:33:58-INFO]: epoch 3-step 2854630 loss: 0.708938 (accumulated 1 forwards) [2026-07-10 00:34:07-INFO]: epoch 3-step 2854640 loss: 0.639049 (accumulated 1 forwards) [2026-07-10 00:34:15-INFO]: epoch 3-step 2854650 loss: 0.650126 (accumulated 1 forwards) [2026-07-10 00:34:24-INFO]: epoch 3-step 2854660 loss: 0.672905 (accumulated 1 forwards) [2026-07-10 00:34:33-INFO]: epoch 3-step 2854670 loss: 0.664446 (accumulated 1 forwards) [2026-07-10 00:34:41-INFO]: epoch 3-step 2854680 loss: 0.669977 (accumulated 1 forwards) [2026-07-10 00:34:50-INFO]: epoch 3-step 2854690 loss: 0.642424 (accumulated 1 forwards) [2026-07-10 00:34:59-INFO]: epoch 3-step 2854700 loss: 0.632084 (accumulated 1 forwards) [2026-07-10 00:35:07-INFO]: epoch 3-step 2854710 loss: 0.663994 (accumulated 1 forwards) [2026-07-10 00:35:16-INFO]: epoch 3-step 2854720 loss: 0.636517 (accumulated 1 forwards) [2026-07-10 00:35:25-INFO]: epoch 3-step 2854730 loss: 0.638823 (accumulated 1 forwards) [2026-07-10 00:35:33-INFO]: epoch 3-step 2854740 loss: 0.648192 (accumulated 1 forwards) [2026-07-10 00:35:42-INFO]: epoch 3-step 2854750 loss: 0.662799 (accumulated 1 forwards) [2026-07-10 00:35:51-INFO]: epoch 3-step 2854760 loss: 0.656116 (accumulated 1 forwards) [2026-07-10 00:36:00-INFO]: epoch 3-step 2854770 loss: 0.672181 (accumulated 1 forwards) [2026-07-10 00:36:08-INFO]: epoch 3-step 2854780 loss: 0.683895 (accumulated 1 forwards) [2026-07-10 00:36:17-INFO]: epoch 3-step 2854790 loss: 0.676131 (accumulated 1 forwards) [2026-07-10 00:36:26-INFO]: epoch 3-step 2854800 loss: 0.692323 (accumulated 1 forwards) [2026-07-10 00:36:34-INFO]: epoch 3-step 2854810 loss: 0.613000 (accumulated 1 forwards) [2026-07-10 00:36:43-INFO]: epoch 3-step 2854820 loss: 0.652018 (accumulated 1 forwards) [2026-07-10 00:36:52-INFO]: epoch 3-step 2854830 loss: 0.668757 (accumulated 1 forwards) [2026-07-10 00:37:00-INFO]: epoch 3-step 2854840 loss: 0.660451 (accumulated 1 forwards) [2026-07-10 00:37:07-INFO]: Reload dataloader [2026-07-10 00:37:09-INFO]: epoch 4-step 2854850 loss: 0.656480 (accumulated 1 forwards) [2026-07-10 00:37:18-INFO]: epoch 4-step 2854860 loss: 0.640531 (accumulated 1 forwards) [2026-07-10 00:37:26-INFO]: epoch 4-step 2854870 loss: 0.711283 (accumulated 1 forwards) [2026-07-10 00:37:35-INFO]: epoch 4-step 2854880 loss: 0.675584 (accumulated 1 forwards) [2026-07-10 00:37:44-INFO]: epoch 4-step 2854890 loss: 0.680149 (accumulated 1 forwards) [2026-07-10 00:37:52-INFO]: epoch 4-step 2854900 loss: 0.640643 (accumulated 1 forwards) [2026-07-10 00:38:01-INFO]: epoch 4-step 2854910 loss: 0.712351 (accumulated 1 forwards) [2026-07-10 00:38:10-INFO]: epoch 4-step 2854920 loss: 0.700118 (accumulated 1 forwards) [2026-07-10 00:38:18-INFO]: epoch 4-step 2854930 loss: 0.683910 (accumulated 1 forwards) [2026-07-10 00:38:27-INFO]: epoch 4-step 2854940 loss: 0.681506 (accumulated 1 forwards) [2026-07-10 00:38:36-INFO]: epoch 4-step 2854950 loss: 0.649865 (accumulated 1 forwards) [2026-07-10 00:38:44-INFO]: epoch 4-step 2854960 loss: 0.677072 (accumulated 1 forwards) [2026-07-10 00:38:53-INFO]: epoch 4-step 2854970 loss: 0.652273 (accumulated 1 forwards) [2026-07-10 00:39:02-INFO]: epoch 4-step 2854980 loss: 0.682746 (accumulated 1 forwards) [2026-07-10 00:39:10-INFO]: epoch 4-step 2854990 loss: 0.651416 (accumulated 1 forwards) [2026-07-10 00:39:19-INFO]: epoch 4-step 2855000 loss: 0.639555 (accumulated 1 forwards) [2026-07-10 00:39:28-INFO]: epoch 4-step 2855010 loss: 0.636621 (accumulated 1 forwards) [2026-07-10 00:39:37-INFO]: epoch 4-step 2855020 loss: 0.620042 (accumulated 1 forwards) [2026-07-10 00:39:45-INFO]: epoch 4-step 2855030 loss: 0.681491 (accumulated 1 forwards) [2026-07-10 00:39:54-INFO]: epoch 4-step 2855040 loss: 0.675460 (accumulated 1 forwards) [2026-07-10 00:40:03-INFO]: epoch 4-step 2855050 loss: 0.650280 (accumulated 1 forwards) [2026-07-10 00:40:11-INFO]: epoch 4-step 2855060 loss: 0.692545 (accumulated 1 forwards) [2026-07-10 00:40:20-INFO]: epoch 4-step 2855070 loss: 0.685167 (accumulated 1 forwards) [2026-07-10 00:40:29-INFO]: epoch 4-step 2855080 loss: 0.675329 (accumulated 1 forwards) [2026-07-10 00:40:37-INFO]: epoch 4-step 2855090 loss: 0.695993 (accumulated 1 forwards) [2026-07-10 00:40:46-INFO]: epoch 4-step 2855100 loss: 0.650539 (accumulated 1 forwards) [2026-07-10 00:40:55-INFO]: epoch 4-step 2855110 loss: 0.638927 (accumulated 1 forwards) [2026-07-10 00:41:03-INFO]: epoch 4-step 2855120 loss: 0.637462 (accumulated 1 forwards) [2026-07-10 00:41:12-INFO]: epoch 4-step 2855130 loss: 0.662638 (accumulated 1 forwards) [2026-07-10 00:41:21-INFO]: epoch 4-step 2855140 loss: 0.718151 (accumulated 1 forwards) [2026-07-10 00:41:30-INFO]: epoch 4-step 2855150 loss: 0.646225 (accumulated 1 forwards) [2026-07-10 00:41:38-INFO]: epoch 4-step 2855160 loss: 0.701481 (accumulated 1 forwards) [2026-07-10 00:41:47-INFO]: epoch 4-step 2855170 loss: 0.693642 (accumulated 1 forwards) [2026-07-10 00:41:56-INFO]: epoch 4-step 2855180 loss: 0.678971 (accumulated 1 forwards) [2026-07-10 00:42:04-INFO]: epoch 4-step 2855190 loss: 0.685586 (accumulated 1 forwards) [2026-07-10 00:42:13-INFO]: epoch 4-step 2855200 loss: 0.637791 (accumulated 1 forwards) [2026-07-10 00:42:22-INFO]: epoch 4-step 2855210 loss: 0.599334 (accumulated 1 forwards) [2026-07-10 00:42:30-INFO]: epoch 4-step 2855220 loss: 0.676704 (accumulated 1 forwards) [2026-07-10 00:42:39-INFO]: epoch 4-step 2855230 loss: 0.667503 (accumulated 1 forwards) [2026-07-10 00:42:48-INFO]: epoch 4-step 2855240 loss: 0.667624 (accumulated 1 forwards) [2026-07-10 00:42:56-INFO]: epoch 4-step 2855250 loss: 0.692389 (accumulated 1 forwards) [2026-07-10 00:43:05-INFO]: epoch 4-step 2855260 loss: 0.661316 (accumulated 1 forwards) [2026-07-10 00:43:14-INFO]: epoch 4-step 2855270 loss: 0.676286 (accumulated 1 forwards) [2026-07-10 00:43:22-INFO]: epoch 4-step 2855280 loss: 0.671956 (accumulated 1 forwards) [2026-07-10 00:43:31-INFO]: epoch 4-step 2855290 loss: 0.676917 (accumulated 1 forwards) [2026-07-10 00:43:40-INFO]: epoch 4-step 2855300 loss: 0.641119 (accumulated 1 forwards) [2026-07-10 00:43:48-INFO]: epoch 4-step 2855310 loss: 0.643831 (accumulated 1 forwards) [2026-07-10 00:43:57-INFO]: epoch 4-step 2855320 loss: 0.589448 (accumulated 1 forwards) [2026-07-10 00:44:06-INFO]: epoch 4-step 2855330 loss: 0.652488 (accumulated 1 forwards) [2026-07-10 00:44:14-INFO]: epoch 4-step 2855340 loss: 0.693292 (accumulated 1 forwards) [2026-07-10 00:44:23-INFO]: epoch 4-step 2855350 loss: 0.631829 (accumulated 1 forwards) [2026-07-10 00:44:32-INFO]: epoch 4-step 2855360 loss: 0.700133 (accumulated 1 forwards) [2026-07-10 00:44:40-INFO]: epoch 4-step 2855370 loss: 0.657576 (accumulated 1 forwards) [2026-07-10 00:44:49-INFO]: epoch 4-step 2855380 loss: 0.649328 (accumulated 1 forwards) [2026-07-10 00:44:58-INFO]: epoch 4-step 2855390 loss: 0.677672 (accumulated 1 forwards) [2026-07-10 00:45:07-INFO]: epoch 4-step 2855400 loss: 0.650086 (accumulated 1 forwards) [2026-07-10 00:45:15-INFO]: epoch 4-step 2855410 loss: 0.657787 (accumulated 1 forwards) [2026-07-10 00:45:24-INFO]: epoch 4-step 2855420 loss: 0.681653 (accumulated 1 forwards) [2026-07-10 00:45:33-INFO]: epoch 4-step 2855430 loss: 0.649519 (accumulated 1 forwards) [2026-07-10 00:45:41-INFO]: epoch 4-step 2855440 loss: 0.662051 (accumulated 1 forwards) [2026-07-10 00:45:50-INFO]: epoch 4-step 2855450 loss: 0.686394 (accumulated 1 forwards) [2026-07-10 00:45:59-INFO]: epoch 4-step 2855460 loss: 0.617608 (accumulated 1 forwards) [2026-07-10 00:46:07-INFO]: epoch 4-step 2855470 loss: 0.661696 (accumulated 1 forwards) [2026-07-10 00:46:16-INFO]: epoch 4-step 2855480 loss: 0.703311 (accumulated 1 forwards) [2026-07-10 00:46:25-INFO]: epoch 4-step 2855490 loss: 0.663422 (accumulated 1 forwards) [2026-07-10 00:46:34-INFO]: epoch 4-step 2855500 loss: 0.689613 (accumulated 1 forwards) [2026-07-10 00:46:42-INFO]: epoch 4-step 2855510 loss: 0.619864 (accumulated 1 forwards) [2026-07-10 00:46:51-INFO]: epoch 4-step 2855520 loss: 0.646716 (accumulated 1 forwards) [2026-07-10 00:47:00-INFO]: epoch 4-step 2855530 loss: 0.684084 (accumulated 1 forwards) [2026-07-10 00:47:08-INFO]: epoch 4-step 2855540 loss: 0.675883 (accumulated 1 forwards) [2026-07-10 00:47:17-INFO]: epoch 4-step 2855550 loss: 0.641805 (accumulated 1 forwards) [2026-07-10 00:47:26-INFO]: epoch 4-step 2855560 loss: 0.684481 (accumulated 1 forwards) [2026-07-10 00:47:34-INFO]: epoch 4-step 2855570 loss: 0.688123 (accumulated 1 forwards) [2026-07-10 00:47:43-INFO]: epoch 4-step 2855580 loss: 0.660881 (accumulated 1 forwards) [2026-07-10 00:47:52-INFO]: epoch 4-step 2855590 loss: 0.637609 (accumulated 1 forwards) [2026-07-10 00:48:01-INFO]: epoch 4-step 2855600 loss: 0.664054 (accumulated 1 forwards) [2026-07-10 00:48:09-INFO]: epoch 4-step 2855610 loss: 0.678103 (accumulated 1 forwards) [2026-07-10 00:48:18-INFO]: epoch 4-step 2855620 loss: 0.673687 (accumulated 1 forwards) [2026-07-10 00:48:27-INFO]: epoch 4-step 2855630 loss: 0.651555 (accumulated 1 forwards) [2026-07-10 00:48:35-INFO]: epoch 4-step 2855640 loss: 0.637417 (accumulated 1 forwards) [2026-07-10 00:48:44-INFO]: epoch 4-step 2855650 loss: 0.691928 (accumulated 1 forwards) [2026-07-10 00:48:53-INFO]: epoch 4-step 2855660 loss: 0.662513 (accumulated 1 forwards) [2026-07-10 00:49:02-INFO]: epoch 4-step 2855670 loss: 0.676799 (accumulated 1 forwards) [2026-07-10 00:49:10-INFO]: epoch 4-step 2855680 loss: 0.616653 (accumulated 1 forwards) [2026-07-10 00:49:19-INFO]: epoch 4-step 2855690 loss: 0.656934 (accumulated 1 forwards) [2026-07-10 00:49:28-INFO]: epoch 4-step 2855700 loss: 0.619270 (accumulated 1 forwards) [2026-07-10 00:49:36-INFO]: epoch 4-step 2855710 loss: 0.659291 (accumulated 1 forwards) [2026-07-10 00:49:45-INFO]: epoch 4-step 2855720 loss: 0.648984 (accumulated 1 forwards) [2026-07-10 00:49:54-INFO]: epoch 4-step 2855730 loss: 0.703113 (accumulated 1 forwards) [2026-07-10 00:50:02-INFO]: epoch 4-step 2855740 loss: 0.682621 (accumulated 1 forwards) [2026-07-10 00:50:11-INFO]: epoch 4-step 2855750 loss: 0.643167 (accumulated 1 forwards) [2026-07-10 00:50:20-INFO]: epoch 4-step 2855760 loss: 0.682701 (accumulated 1 forwards) [2026-07-10 00:50:28-INFO]: epoch 4-step 2855770 loss: 0.663213 (accumulated 1 forwards) [2026-07-10 00:50:37-INFO]: epoch 4-step 2855780 loss: 0.649856 (accumulated 1 forwards) [2026-07-10 00:50:46-INFO]: epoch 4-step 2855790 loss: 0.649961 (accumulated 1 forwards) [2026-07-10 00:50:54-INFO]: epoch 4-step 2855800 loss: 0.694322 (accumulated 1 forwards) [2026-07-10 00:51:03-INFO]: epoch 4-step 2855810 loss: 0.646304 (accumulated 1 forwards) [2026-07-10 00:51:12-INFO]: epoch 4-step 2855820 loss: 0.671977 (accumulated 1 forwards) [2026-07-10 00:51:21-INFO]: epoch 4-step 2855830 loss: 0.666252 (accumulated 1 forwards) [2026-07-10 00:51:29-INFO]: epoch 4-step 2855840 loss: 0.708608 (accumulated 1 forwards) [2026-07-10 00:51:38-INFO]: epoch 4-step 2855850 loss: 0.659300 (accumulated 1 forwards) [2026-07-10 00:51:47-INFO]: epoch 4-step 2855860 loss: 0.638587 (accumulated 1 forwards) [2026-07-10 00:51:55-INFO]: epoch 4-step 2855870 loss: 0.678199 (accumulated 1 forwards) [2026-07-10 00:52:04-INFO]: epoch 4-step 2855880 loss: 0.610147 (accumulated 1 forwards) [2026-07-10 00:52:13-INFO]: epoch 4-step 2855890 loss: 0.681900 (accumulated 1 forwards) [2026-07-10 00:52:21-INFO]: epoch 4-step 2855900 loss: 0.671407 (accumulated 1 forwards) [2026-07-10 00:52:30-INFO]: epoch 4-step 2855910 loss: 0.693978 (accumulated 1 forwards) [2026-07-10 00:52:39-INFO]: epoch 4-step 2855920 loss: 0.672480 (accumulated 1 forwards) [2026-07-10 00:52:47-INFO]: epoch 4-step 2855930 loss: 0.665117 (accumulated 1 forwards) [2026-07-10 00:52:56-INFO]: epoch 4-step 2855940 loss: 0.651652 (accumulated 1 forwards) [2026-07-10 00:53:05-INFO]: epoch 4-step 2855950 loss: 0.670682 (accumulated 1 forwards) [2026-07-10 00:53:13-INFO]: epoch 4-step 2855960 loss: 0.693395 (accumulated 1 forwards) [2026-07-10 00:53:22-INFO]: epoch 4-step 2855970 loss: 0.631043 (accumulated 1 forwards) [2026-07-10 00:53:31-INFO]: epoch 4-step 2855980 loss: 0.687358 (accumulated 1 forwards) [2026-07-10 00:53:39-INFO]: epoch 4-step 2855990 loss: 0.702733 (accumulated 1 forwards) [2026-07-10 00:53:48-INFO]: epoch 4-step 2856000 loss: 0.664823 (accumulated 1 forwards) [2026-07-10 00:53:57-INFO]: epoch 4-step 2856010 loss: 0.686582 (accumulated 1 forwards) [2026-07-10 00:54:05-INFO]: epoch 4-step 2856020 loss: 0.697290 (accumulated 1 forwards) [2026-07-10 00:54:14-INFO]: epoch 4-step 2856030 loss: 0.683034 (accumulated 1 forwards) [2026-07-10 00:54:23-INFO]: epoch 4-step 2856040 loss: 0.695246 (accumulated 1 forwards) [2026-07-10 00:54:31-INFO]: epoch 4-step 2856050 loss: 0.656215 (accumulated 1 forwards) [2026-07-10 00:54:40-INFO]: epoch 4-step 2856060 loss: 0.720070 (accumulated 1 forwards) [2026-07-10 00:54:40-INFO]: Reload dataloader [2026-07-10 00:54:49-INFO]: epoch 5-step 2856070 loss: 0.691999 (accumulated 1 forwards) [2026-07-10 00:54:57-INFO]: epoch 5-step 2856080 loss: 0.665644 (accumulated 1 forwards) [2026-07-10 00:55:06-INFO]: epoch 5-step 2856090 loss: 0.638643 (accumulated 1 forwards) [2026-07-10 00:55:15-INFO]: epoch 5-step 2856100 loss: 0.660644 (accumulated 1 forwards) [2026-07-10 00:55:24-INFO]: epoch 5-step 2856110 loss: 0.664203 (accumulated 1 forwards) [2026-07-10 00:55:32-INFO]: epoch 5-step 2856120 loss: 0.683209 (accumulated 1 forwards) [2026-07-10 00:55:41-INFO]: epoch 5-step 2856130 loss: 0.648684 (accumulated 1 forwards) [2026-07-10 00:55:49-INFO]: epoch 5-step 2856140 loss: 0.686462 (accumulated 1 forwards) [2026-07-10 00:55:58-INFO]: epoch 5-step 2856150 loss: 0.680897 (accumulated 1 forwards) [2026-07-10 00:56:07-INFO]: epoch 5-step 2856160 loss: 0.668486 (accumulated 1 forwards) [2026-07-10 00:56:16-INFO]: epoch 5-step 2856170 loss: 0.673029 (accumulated 1 forwards) [2026-07-10 00:56:24-INFO]: epoch 5-step 2856180 loss: 0.689304 (accumulated 1 forwards) [2026-07-10 00:56:33-INFO]: epoch 5-step 2856190 loss: 0.653955 (accumulated 1 forwards) [2026-07-10 00:56:42-INFO]: epoch 5-step 2856200 loss: 0.674787 (accumulated 1 forwards) [2026-07-10 00:56:50-INFO]: epoch 5-step 2856210 loss: 0.685492 (accumulated 1 forwards) [2026-07-10 00:56:59-INFO]: epoch 5-step 2856220 loss: 0.660065 (accumulated 1 forwards) [2026-07-10 00:57:08-INFO]: epoch 5-step 2856230 loss: 0.682953 (accumulated 1 forwards) [2026-07-10 00:57:16-INFO]: epoch 5-step 2856240 loss: 0.665484 (accumulated 1 forwards) [2026-07-10 00:57:25-INFO]: epoch 5-step 2856250 loss: 0.674994 (accumulated 1 forwards) [2026-07-10 00:57:33-INFO]: epoch 5-step 2856260 loss: 0.650084 (accumulated 1 forwards) [2026-07-10 00:57:42-INFO]: epoch 5-step 2856270 loss: 0.676372 (accumulated 1 forwards) [2026-07-10 00:57:51-INFO]: epoch 5-step 2856280 loss: 0.691356 (accumulated 1 forwards) [2026-07-10 00:57:59-INFO]: epoch 5-step 2856290 loss: 0.649867 (accumulated 1 forwards) [2026-07-10 00:58:08-INFO]: epoch 5-step 2856300 loss: 0.684006 (accumulated 1 forwards) [2026-07-10 00:58:17-INFO]: epoch 5-step 2856310 loss: 0.697520 (accumulated 1 forwards) [2026-07-10 00:58:25-INFO]: epoch 5-step 2856320 loss: 0.673512 (accumulated 1 forwards) [2026-07-10 00:58:34-INFO]: epoch 5-step 2856330 loss: 0.669160 (accumulated 1 forwards) [2026-07-10 00:58:43-INFO]: epoch 5-step 2856340 loss: 0.674174 (accumulated 1 forwards) [2026-07-10 00:58:52-INFO]: epoch 5-step 2856350 loss: 0.693590 (accumulated 1 forwards) [2026-07-10 00:59:00-INFO]: epoch 5-step 2856360 loss: 0.670157 (accumulated 1 forwards) [2026-07-10 00:59:09-INFO]: epoch 5-step 2856370 loss: 0.643771 (accumulated 1 forwards) [2026-07-10 00:59:18-INFO]: epoch 5-step 2856380 loss: 0.673632 (accumulated 1 forwards) [2026-07-10 00:59:26-INFO]: epoch 5-step 2856390 loss: 0.685916 (accumulated 1 forwards) [2026-07-10 00:59:35-INFO]: epoch 5-step 2856400 loss: 0.684741 (accumulated 1 forwards) [2026-07-10 00:59:44-INFO]: epoch 5-step 2856410 loss: 0.670986 (accumulated 1 forwards) [2026-07-10 00:59:52-INFO]: epoch 5-step 2856420 loss: 0.656637 (accumulated 1 forwards) [2026-07-10 01:00:01-INFO]: epoch 5-step 2856430 loss: 0.647434 (accumulated 1 forwards) [2026-07-10 01:00:10-INFO]: epoch 5-step 2856440 loss: 0.680781 (accumulated 1 forwards) [2026-07-10 01:00:19-INFO]: epoch 5-step 2856450 loss: 0.683293 (accumulated 1 forwards) [2026-07-10 01:00:27-INFO]: epoch 5-step 2856460 loss: 0.651149 (accumulated 1 forwards) [2026-07-10 01:00:36-INFO]: epoch 5-step 2856470 loss: 0.660587 (accumulated 1 forwards) [2026-07-10 01:00:45-INFO]: epoch 5-step 2856480 loss: 0.672597 (accumulated 1 forwards) [2026-07-10 01:00:53-INFO]: epoch 5-step 2856490 loss: 0.635583 (accumulated 1 forwards) [2026-07-10 01:01:02-INFO]: epoch 5-step 2856500 loss: 0.668912 (accumulated 1 forwards) [2026-07-10 01:01:11-INFO]: epoch 5-step 2856510 loss: 0.684396 (accumulated 1 forwards) [2026-07-10 01:01:19-INFO]: epoch 5-step 2856520 loss: 0.669050 (accumulated 1 forwards) [2026-07-10 01:01:28-INFO]: epoch 5-step 2856530 loss: 0.652737 (accumulated 1 forwards) [2026-07-10 01:01:37-INFO]: epoch 5-step 2856540 loss: 0.678140 (accumulated 1 forwards) [2026-07-10 01:01:45-INFO]: epoch 5-step 2856550 loss: 0.678803 (accumulated 1 forwards) [2026-07-10 01:01:54-INFO]: epoch 5-step 2856560 loss: 0.689877 (accumulated 1 forwards) [2026-07-10 01:02:03-INFO]: epoch 5-step 2856570 loss: 0.680012 (accumulated 1 forwards) [2026-07-10 01:02:11-INFO]: epoch 5-step 2856580 loss: 0.694818 (accumulated 1 forwards) [2026-07-10 01:02:20-INFO]: epoch 5-step 2856590 loss: 0.684131 (accumulated 1 forwards) [2026-07-10 01:02:29-INFO]: epoch 5-step 2856600 loss: 0.646819 (accumulated 1 forwards) [2026-07-10 01:02:37-INFO]: epoch 5-step 2856610 loss: 0.647533 (accumulated 1 forwards) [2026-07-10 01:02:46-INFO]: epoch 5-step 2856620 loss: 0.658905 (accumulated 1 forwards) [2026-07-10 01:02:55-INFO]: epoch 5-step 2856630 loss: 0.657729 (accumulated 1 forwards) [2026-07-10 01:03:03-INFO]: epoch 5-step 2856640 loss: 0.652774 (accumulated 1 forwards) [2026-07-10 01:03:12-INFO]: epoch 5-step 2856650 loss: 0.607516 (accumulated 1 forwards) [2026-07-10 01:03:21-INFO]: epoch 5-step 2856660 loss: 0.677318 (accumulated 1 forwards) [2026-07-10 01:03:29-INFO]: epoch 5-step 2856670 loss: 0.666347 (accumulated 1 forwards) [2026-07-10 01:03:38-INFO]: epoch 5-step 2856680 loss: 0.699110 (accumulated 1 forwards) [2026-07-10 01:03:47-INFO]: epoch 5-step 2856690 loss: 0.663775 (accumulated 1 forwards) [2026-07-10 01:03:55-INFO]: epoch 5-step 2856700 loss: 0.694017 (accumulated 1 forwards) [2026-07-10 01:04:04-INFO]: epoch 5-step 2856710 loss: 0.682044 (accumulated 1 forwards) [2026-07-10 01:04:13-INFO]: epoch 5-step 2856720 loss: 0.680416 (accumulated 1 forwards) [2026-07-10 01:04:21-INFO]: epoch 5-step 2856730 loss: 0.614731 (accumulated 1 forwards) [2026-07-10 01:04:30-INFO]: epoch 5-step 2856740 loss: 0.644990 (accumulated 1 forwards) [2026-07-10 01:04:39-INFO]: epoch 5-step 2856750 loss: 0.637294 (accumulated 1 forwards) [2026-07-10 01:04:47-INFO]: epoch 5-step 2856760 loss: 0.669052 (accumulated 1 forwards) [2026-07-10 01:04:56-INFO]: epoch 5-step 2856770 loss: 0.671867 (accumulated 1 forwards) [2026-07-10 01:05:05-INFO]: epoch 5-step 2856780 loss: 0.681746 (accumulated 1 forwards) [2026-07-10 01:05:13-INFO]: epoch 5-step 2856790 loss: 0.698318 (accumulated 1 forwards) [2026-07-10 01:05:22-INFO]: epoch 5-step 2856800 loss: 0.645557 (accumulated 1 forwards) [2026-07-10 01:05:31-INFO]: epoch 5-step 2856810 loss: 0.670468 (accumulated 1 forwards) [2026-07-10 01:05:39-INFO]: epoch 5-step 2856820 loss: 0.660557 (accumulated 1 forwards) [2026-07-10 01:05:48-INFO]: epoch 5-step 2856830 loss: 0.666054 (accumulated 1 forwards) [2026-07-10 01:05:57-INFO]: epoch 5-step 2856840 loss: 0.674287 (accumulated 1 forwards) [2026-07-10 01:06:05-INFO]: epoch 5-step 2856850 loss: 0.704621 (accumulated 1 forwards) [2026-07-10 01:06:14-INFO]: epoch 5-step 2856860 loss: 0.685561 (accumulated 1 forwards) [2026-07-10 01:06:23-INFO]: epoch 5-step 2856870 loss: 0.631759 (accumulated 1 forwards) [2026-07-10 01:06:32-INFO]: epoch 5-step 2856880 loss: 0.651613 (accumulated 1 forwards) [2026-07-10 01:06:40-INFO]: epoch 5-step 2856890 loss: 0.627473 (accumulated 1 forwards) [2026-07-10 01:06:49-INFO]: epoch 5-step 2856900 loss: 0.648208 (accumulated 1 forwards) [2026-07-10 01:06:58-INFO]: epoch 5-step 2856910 loss: 0.668566 (accumulated 1 forwards) [2026-07-10 01:07:06-INFO]: epoch 5-step 2856920 loss: 0.675606 (accumulated 1 forwards) [2026-07-10 01:07:15-INFO]: epoch 5-step 2856930 loss: 0.664116 (accumulated 1 forwards) [2026-07-10 01:07:24-INFO]: epoch 5-step 2856940 loss: 0.638876 (accumulated 1 forwards) [2026-07-10 01:07:32-INFO]: epoch 5-step 2856950 loss: 0.626318 (accumulated 1 forwards) [2026-07-10 01:07:41-INFO]: epoch 5-step 2856960 loss: 0.657606 (accumulated 1 forwards) [2026-07-10 01:07:50-INFO]: epoch 5-step 2856970 loss: 0.687839 (accumulated 1 forwards) [2026-07-10 01:07:58-INFO]: epoch 5-step 2856980 loss: 0.643259 (accumulated 1 forwards) [2026-07-10 01:08:07-INFO]: epoch 5-step 2856990 loss: 0.660462 (accumulated 1 forwards) [2026-07-10 01:08:15-INFO]: epoch 5-step 2857000 loss: 0.689600 (accumulated 1 forwards) [2026-07-10 01:08:24-INFO]: epoch 5-step 2857010 loss: 0.666120 (accumulated 1 forwards) [2026-07-10 01:08:33-INFO]: epoch 5-step 2857020 loss: 0.691323 (accumulated 1 forwards) [2026-07-10 01:08:42-INFO]: epoch 5-step 2857030 loss: 0.655914 (accumulated 1 forwards) [2026-07-10 01:08:50-INFO]: epoch 5-step 2857040 loss: 0.648210 (accumulated 1 forwards) [2026-07-10 01:08:59-INFO]: epoch 5-step 2857050 loss: 0.694462 (accumulated 1 forwards) [2026-07-10 01:09:08-INFO]: epoch 5-step 2857060 loss: 0.654882 (accumulated 1 forwards) [2026-07-10 01:09:16-INFO]: epoch 5-step 2857070 loss: 0.644787 (accumulated 1 forwards) [2026-07-10 01:09:25-INFO]: epoch 5-step 2857080 loss: 0.661955 (accumulated 1 forwards) [2026-07-10 01:09:34-INFO]: epoch 5-step 2857090 loss: 0.653210 (accumulated 1 forwards) [2026-07-10 01:09:42-INFO]: epoch 5-step 2857100 loss: 0.611408 (accumulated 1 forwards) [2026-07-10 01:09:51-INFO]: epoch 5-step 2857110 loss: 0.649571 (accumulated 1 forwards) [2026-07-10 01:10:00-INFO]: epoch 5-step 2857120 loss: 0.685911 (accumulated 1 forwards) [2026-07-10 01:10:08-INFO]: epoch 5-step 2857130 loss: 0.627210 (accumulated 1 forwards) [2026-07-10 01:10:17-INFO]: epoch 5-step 2857140 loss: 0.665452 (accumulated 1 forwards) [2026-07-10 01:10:26-INFO]: epoch 5-step 2857150 loss: 0.707174 (accumulated 1 forwards) [2026-07-10 01:10:35-INFO]: epoch 5-step 2857160 loss: 0.684522 (accumulated 1 forwards) [2026-07-10 01:10:43-INFO]: epoch 5-step 2857170 loss: 0.674440 (accumulated 1 forwards) [2026-07-10 01:10:52-INFO]: epoch 5-step 2857180 loss: 0.691498 (accumulated 1 forwards) [2026-07-10 01:11:01-INFO]: epoch 5-step 2857190 loss: 0.668926 (accumulated 1 forwards) [2026-07-10 01:11:09-INFO]: epoch 5-step 2857200 loss: 0.660282 (accumulated 1 forwards) [2026-07-10 01:11:18-INFO]: epoch 5-step 2857210 loss: 0.655806 (accumulated 1 forwards) [2026-07-10 01:11:27-INFO]: epoch 5-step 2857220 loss: 0.639333 (accumulated 1 forwards) [2026-07-10 01:11:35-INFO]: epoch 5-step 2857230 loss: 0.649347 (accumulated 1 forwards) [2026-07-10 01:11:44-INFO]: epoch 5-step 2857240 loss: 0.638892 (accumulated 1 forwards) [2026-07-10 01:11:53-INFO]: epoch 5-step 2857250 loss: 0.649657 (accumulated 1 forwards) [2026-07-10 01:12:01-INFO]: epoch 5-step 2857260 loss: 0.636274 (accumulated 1 forwards) [2026-07-10 01:12:10-INFO]: epoch 5-step 2857270 loss: 0.663642 (accumulated 1 forwards) [2026-07-10 01:12:12-INFO]: Reload dataloader [2026-07-10 01:12:19-INFO]: epoch 6-step 2857280 loss: 0.680106 (accumulated 1 forwards) [2026-07-10 01:12:27-INFO]: epoch 6-step 2857290 loss: 0.655726 (accumulated 1 forwards) [2026-07-10 01:12:36-INFO]: epoch 6-step 2857300 loss: 0.625060 (accumulated 1 forwards) [2026-07-10 01:12:45-INFO]: epoch 6-step 2857310 loss: 0.693422 (accumulated 1 forwards) [2026-07-10 01:12:54-INFO]: epoch 6-step 2857320 loss: 0.685787 (accumulated 1 forwards) [2026-07-10 01:13:02-INFO]: epoch 6-step 2857330 loss: 0.653026 (accumulated 1 forwards) [2026-07-10 01:13:11-INFO]: epoch 6-step 2857340 loss: 0.701536 (accumulated 1 forwards) [2026-07-10 01:13:20-INFO]: epoch 6-step 2857350 loss: 0.665050 (accumulated 1 forwards) [2026-07-10 01:13:29-INFO]: epoch 6-step 2857360 loss: 0.652291 (accumulated 1 forwards) [2026-07-10 01:13:37-INFO]: epoch 6-step 2857370 loss: 0.665429 (accumulated 1 forwards) [2026-07-10 01:13:46-INFO]: epoch 6-step 2857380 loss: 0.633413 (accumulated 1 forwards) [2026-07-10 01:13:55-INFO]: epoch 6-step 2857390 loss: 0.675635 (accumulated 1 forwards) [2026-07-10 01:14:03-INFO]: epoch 6-step 2857400 loss: 0.668011 (accumulated 1 forwards) [2026-07-10 01:14:12-INFO]: epoch 6-step 2857410 loss: 0.703183 (accumulated 1 forwards) [2026-07-10 01:14:21-INFO]: epoch 6-step 2857420 loss: 0.660230 (accumulated 1 forwards) [2026-07-10 01:14:30-INFO]: epoch 6-step 2857430 loss: 0.666149 (accumulated 1 forwards) [2026-07-10 01:14:38-INFO]: epoch 6-step 2857440 loss: 0.645891 (accumulated 1 forwards) [2026-07-10 01:14:47-INFO]: epoch 6-step 2857450 loss: 0.685293 (accumulated 1 forwards) [2026-07-10 01:14:56-INFO]: epoch 6-step 2857460 loss: 0.647475 (accumulated 1 forwards) [2026-07-10 01:15:05-INFO]: epoch 6-step 2857470 loss: 0.673041 (accumulated 1 forwards) [2026-07-10 01:15:13-INFO]: epoch 6-step 2857480 loss: 0.647039 (accumulated 1 forwards) [2026-07-10 01:15:22-INFO]: epoch 6-step 2857490 loss: 0.617026 (accumulated 1 forwards) [2026-07-10 01:15:31-INFO]: epoch 6-step 2857500 loss: 0.630341 (accumulated 1 forwards) [2026-07-10 01:15:40-INFO]: epoch 6-step 2857510 loss: 0.685993 (accumulated 1 forwards) [2026-07-10 01:15:48-INFO]: epoch 6-step 2857520 loss: 0.661058 (accumulated 1 forwards) [2026-07-10 01:15:57-INFO]: epoch 6-step 2857530 loss: 0.602201 (accumulated 1 forwards) [2026-07-10 01:16:06-INFO]: epoch 6-step 2857540 loss: 0.653878 (accumulated 1 forwards) [2026-07-10 01:16:14-INFO]: epoch 6-step 2857550 loss: 0.672291 (accumulated 1 forwards) [2026-07-10 01:16:23-INFO]: epoch 6-step 2857560 loss: 0.663144 (accumulated 1 forwards) [2026-07-10 01:16:32-INFO]: epoch 6-step 2857570 loss: 0.670922 (accumulated 1 forwards) [2026-07-10 01:16:41-INFO]: epoch 6-step 2857580 loss: 0.671593 (accumulated 1 forwards) [2026-07-10 01:16:49-INFO]: epoch 6-step 2857590 loss: 0.680193 (accumulated 1 forwards) [2026-07-10 01:16:58-INFO]: epoch 6-step 2857600 loss: 0.675426 (accumulated 1 forwards) [2026-07-10 01:17:07-INFO]: epoch 6-step 2857610 loss: 0.674268 (accumulated 1 forwards) [2026-07-10 01:17:16-INFO]: epoch 6-step 2857620 loss: 0.673885 (accumulated 1 forwards) [2026-07-10 01:17:24-INFO]: epoch 6-step 2857630 loss: 0.703270 (accumulated 1 forwards) [2026-07-10 01:17:33-INFO]: epoch 6-step 2857640 loss: 0.645726 (accumulated 1 forwards) [2026-07-10 01:17:42-INFO]: epoch 6-step 2857650 loss: 0.675183 (accumulated 1 forwards) [2026-07-10 01:17:50-INFO]: epoch 6-step 2857660 loss: 0.649469 (accumulated 1 forwards) [2026-07-10 01:17:59-INFO]: epoch 6-step 2857670 loss: 0.636449 (accumulated 1 forwards) [2026-07-10 01:18:08-INFO]: epoch 6-step 2857680 loss: 0.687737 (accumulated 1 forwards) [2026-07-10 01:18:17-INFO]: epoch 6-step 2857690 loss: 0.680715 (accumulated 1 forwards) [2026-07-10 01:18:25-INFO]: epoch 6-step 2857700 loss: 0.659358 (accumulated 1 forwards) [2026-07-10 01:18:34-INFO]: epoch 6-step 2857710 loss: 0.661174 (accumulated 1 forwards) [2026-07-10 01:18:43-INFO]: epoch 6-step 2857720 loss: 0.621288 (accumulated 1 forwards) [2026-07-10 01:18:51-INFO]: epoch 6-step 2857730 loss: 0.604441 (accumulated 1 forwards) [2026-07-10 01:19:00-INFO]: epoch 6-step 2857740 loss: 0.670130 (accumulated 1 forwards) [2026-07-10 01:19:09-INFO]: epoch 6-step 2857750 loss: 0.703888 (accumulated 1 forwards) [2026-07-10 01:19:17-INFO]: epoch 6-step 2857760 loss: 0.640288 (accumulated 1 forwards) [2026-07-10 01:19:26-INFO]: epoch 6-step 2857770 loss: 0.665651 (accumulated 1 forwards) [2026-07-10 01:19:35-INFO]: epoch 6-step 2857780 loss: 0.638045 (accumulated 1 forwards) [2026-07-10 01:19:44-INFO]: epoch 6-step 2857790 loss: 0.674163 (accumulated 1 forwards) [2026-07-10 01:19:52-INFO]: epoch 6-step 2857800 loss: 0.675532 (accumulated 1 forwards) [2026-07-10 01:20:01-INFO]: epoch 6-step 2857810 loss: 0.642745 (accumulated 1 forwards) [2026-07-10 01:20:10-INFO]: epoch 6-step 2857820 loss: 0.697864 (accumulated 1 forwards) [2026-07-10 01:20:18-INFO]: epoch 6-step 2857830 loss: 0.660612 (accumulated 1 forwards) [2026-07-10 01:20:27-INFO]: epoch 6-step 2857840 loss: 0.657124 (accumulated 1 forwards) [2026-07-10 01:20:36-INFO]: epoch 6-step 2857850 loss: 0.624190 (accumulated 1 forwards) [2026-07-10 01:20:44-INFO]: epoch 6-step 2857860 loss: 0.662473 (accumulated 1 forwards) [2026-07-10 01:20:53-INFO]: epoch 6-step 2857870 loss: 0.693339 (accumulated 1 forwards) [2026-07-10 01:21:02-INFO]: epoch 6-step 2857880 loss: 0.639978 (accumulated 1 forwards) [2026-07-10 01:21:10-INFO]: epoch 6-step 2857890 loss: 0.699358 (accumulated 1 forwards) [2026-07-10 01:21:19-INFO]: epoch 6-step 2857900 loss: 0.620723 (accumulated 1 forwards) [2026-07-10 01:21:28-INFO]: epoch 6-step 2857910 loss: 0.659813 (accumulated 1 forwards) [2026-07-10 01:21:36-INFO]: epoch 6-step 2857920 loss: 0.696265 (accumulated 1 forwards) [2026-07-10 01:21:45-INFO]: epoch 6-step 2857930 loss: 0.710665 (accumulated 1 forwards) [2026-07-10 01:21:54-INFO]: epoch 6-step 2857940 loss: 0.611591 (accumulated 1 forwards) [2026-07-10 01:22:02-INFO]: epoch 6-step 2857950 loss: 0.659751 (accumulated 1 forwards) [2026-07-10 01:22:11-INFO]: epoch 6-step 2857960 loss: 0.677437 (accumulated 1 forwards) [2026-07-10 01:22:20-INFO]: epoch 6-step 2857970 loss: 0.682530 (accumulated 1 forwards) [2026-07-10 01:22:28-INFO]: epoch 6-step 2857980 loss: 0.664706 (accumulated 1 forwards) [2026-07-10 01:22:37-INFO]: epoch 6-step 2857990 loss: 0.633783 (accumulated 1 forwards) [2026-07-10 01:22:46-INFO]: epoch 6-step 2858000 loss: 0.651276 (accumulated 1 forwards) [2026-07-10 01:22:54-INFO]: epoch 6-step 2858010 loss: 0.647639 (accumulated 1 forwards) [2026-07-10 01:23:03-INFO]: epoch 6-step 2858020 loss: 0.643642 (accumulated 1 forwards) [2026-07-10 01:23:12-INFO]: epoch 6-step 2858030 loss: 0.662104 (accumulated 1 forwards) [2026-07-10 01:23:20-INFO]: epoch 6-step 2858040 loss: 0.668940 (accumulated 1 forwards) [2026-07-10 01:23:29-INFO]: epoch 6-step 2858050 loss: 0.643524 (accumulated 1 forwards) [2026-07-10 01:23:38-INFO]: epoch 6-step 2858060 loss: 0.651924 (accumulated 1 forwards) [2026-07-10 01:23:47-INFO]: epoch 6-step 2858070 loss: 0.684149 (accumulated 1 forwards) [2026-07-10 01:23:55-INFO]: epoch 6-step 2858080 loss: 0.641726 (accumulated 1 forwards) [2026-07-10 01:24:04-INFO]: epoch 6-step 2858090 loss: 0.679810 (accumulated 1 forwards) [2026-07-10 01:24:13-INFO]: epoch 6-step 2858100 loss: 0.642108 (accumulated 1 forwards) [2026-07-10 01:24:21-INFO]: epoch 6-step 2858110 loss: 0.681418 (accumulated 1 forwards) [2026-07-10 01:24:30-INFO]: epoch 6-step 2858120 loss: 0.666986 (accumulated 1 forwards) [2026-07-10 01:24:39-INFO]: epoch 6-step 2858130 loss: 0.687651 (accumulated 1 forwards) [2026-07-10 01:24:47-INFO]: epoch 6-step 2858140 loss: 0.654766 (accumulated 1 forwards) [2026-07-10 01:24:56-INFO]: epoch 6-step 2858150 loss: 0.656861 (accumulated 1 forwards) [2026-07-10 01:25:05-INFO]: epoch 6-step 2858160 loss: 0.663936 (accumulated 1 forwards) [2026-07-10 01:25:13-INFO]: epoch 6-step 2858170 loss: 0.632426 (accumulated 1 forwards) [2026-07-10 01:25:22-INFO]: epoch 6-step 2858180 loss: 0.665593 (accumulated 1 forwards) [2026-07-10 01:25:31-INFO]: epoch 6-step 2858190 loss: 0.650885 (accumulated 1 forwards) [2026-07-10 01:25:39-INFO]: epoch 6-step 2858200 loss: 0.634973 (accumulated 1 forwards) [2026-07-10 01:25:48-INFO]: epoch 6-step 2858210 loss: 0.655064 (accumulated 1 forwards) [2026-07-10 01:25:57-INFO]: epoch 6-step 2858220 loss: 0.665525 (accumulated 1 forwards) [2026-07-10 01:26:05-INFO]: epoch 6-step 2858230 loss: 0.684130 (accumulated 1 forwards) [2026-07-10 01:26:14-INFO]: epoch 6-step 2858240 loss: 0.665479 (accumulated 1 forwards) [2026-07-10 01:26:23-INFO]: epoch 6-step 2858250 loss: 0.663779 (accumulated 1 forwards) [2026-07-10 01:26:32-INFO]: epoch 6-step 2858260 loss: 0.689599 (accumulated 1 forwards) [2026-07-10 01:26:40-INFO]: epoch 6-step 2858270 loss: 0.655273 (accumulated 1 forwards) [2026-07-10 01:26:49-INFO]: epoch 6-step 2858280 loss: 0.673992 (accumulated 1 forwards) [2026-07-10 01:26:58-INFO]: epoch 6-step 2858290 loss: 0.666360 (accumulated 1 forwards) [2026-07-10 01:27:06-INFO]: epoch 6-step 2858300 loss: 0.664375 (accumulated 1 forwards) [2026-07-10 01:27:15-INFO]: epoch 6-step 2858310 loss: 0.640542 (accumulated 1 forwards) [2026-07-10 01:27:24-INFO]: epoch 6-step 2858320 loss: 0.658031 (accumulated 1 forwards) [2026-07-10 01:27:32-INFO]: epoch 6-step 2858330 loss: 0.680632 (accumulated 1 forwards) [2026-07-10 01:27:41-INFO]: epoch 6-step 2858340 loss: 0.683312 (accumulated 1 forwards) [2026-07-10 01:27:50-INFO]: epoch 6-step 2858350 loss: 0.615540 (accumulated 1 forwards) [2026-07-10 01:27:58-INFO]: epoch 6-step 2858360 loss: 0.667962 (accumulated 1 forwards) [2026-07-10 01:28:07-INFO]: epoch 6-step 2858370 loss: 0.647613 (accumulated 1 forwards) [2026-07-10 01:28:16-INFO]: epoch 6-step 2858380 loss: 0.654248 (accumulated 1 forwards) [2026-07-10 01:28:24-INFO]: epoch 6-step 2858390 loss: 0.646727 (accumulated 1 forwards) [2026-07-10 01:28:33-INFO]: epoch 6-step 2858400 loss: 0.642810 (accumulated 1 forwards) [2026-07-10 01:28:42-INFO]: epoch 6-step 2858410 loss: 0.637895 (accumulated 1 forwards) [2026-07-10 01:28:50-INFO]: epoch 6-step 2858420 loss: 0.687342 (accumulated 1 forwards) [2026-07-10 01:28:59-INFO]: epoch 6-step 2858430 loss: 0.697883 (accumulated 1 forwards) [2026-07-10 01:29:08-INFO]: epoch 6-step 2858440 loss: 0.701948 (accumulated 1 forwards) [2026-07-10 01:29:16-INFO]: epoch 6-step 2858450 loss: 0.650712 (accumulated 1 forwards) [2026-07-10 01:29:25-INFO]: epoch 6-step 2858460 loss: 0.664236 (accumulated 1 forwards) [2026-07-10 01:29:34-INFO]: epoch 6-step 2858470 loss: 0.708023 (accumulated 1 forwards) [2026-07-10 01:29:43-INFO]: epoch 6-step 2858480 loss: 0.673423 (accumulated 1 forwards) [2026-07-10 01:29:46-INFO]: Reload dataloader [2026-07-10 01:29:51-INFO]: epoch 7-step 2858490 loss: 0.674009 (accumulated 1 forwards) [2026-07-10 01:30:00-INFO]: epoch 7-step 2858500 loss: 0.686388 (accumulated 1 forwards) [2026-07-10 01:30:08-INFO]: epoch 7-step 2858510 loss: 0.665652 (accumulated 1 forwards) [2026-07-10 01:30:17-INFO]: epoch 7-step 2858520 loss: 0.663746 (accumulated 1 forwards) [2026-07-10 01:30:26-INFO]: epoch 7-step 2858530 loss: 0.625566 (accumulated 1 forwards) [2026-07-10 01:30:34-INFO]: epoch 7-step 2858540 loss: 0.674301 (accumulated 1 forwards) [2026-07-10 01:30:43-INFO]: epoch 7-step 2858550 loss: 0.671233 (accumulated 1 forwards) [2026-07-10 01:30:52-INFO]: epoch 7-step 2858560 loss: 0.703024 (accumulated 1 forwards) [2026-07-10 01:31:00-INFO]: epoch 7-step 2858570 loss: 0.626343 (accumulated 1 forwards) [2026-07-10 01:31:09-INFO]: epoch 7-step 2858580 loss: 0.648310 (accumulated 1 forwards) [2026-07-10 01:31:18-INFO]: epoch 7-step 2858590 loss: 0.663018 (accumulated 1 forwards) [2026-07-10 01:31:26-INFO]: epoch 7-step 2858600 loss: 0.640018 (accumulated 1 forwards) [2026-07-10 01:31:35-INFO]: epoch 7-step 2858610 loss: 0.683494 (accumulated 1 forwards) [2026-07-10 01:31:44-INFO]: epoch 7-step 2858620 loss: 0.693987 (accumulated 1 forwards) [2026-07-10 01:31:53-INFO]: epoch 7-step 2858630 loss: 0.641562 (accumulated 1 forwards) [2026-07-10 01:32:01-INFO]: epoch 7-step 2858640 loss: 0.683498 (accumulated 1 forwards) [2026-07-10 01:32:10-INFO]: epoch 7-step 2858650 loss: 0.664355 (accumulated 1 forwards) [2026-07-10 01:32:19-INFO]: epoch 7-step 2858660 loss: 0.673049 (accumulated 1 forwards) [2026-07-10 01:32:27-INFO]: epoch 7-step 2858670 loss: 0.675092 (accumulated 1 forwards) [2026-07-10 01:32:36-INFO]: epoch 7-step 2858680 loss: 0.654030 (accumulated 1 forwards) [2026-07-10 01:32:45-INFO]: epoch 7-step 2858690 loss: 0.664681 (accumulated 1 forwards) [2026-07-10 01:32:53-INFO]: epoch 7-step 2858700 loss: 0.628057 (accumulated 1 forwards) [2026-07-10 01:33:02-INFO]: epoch 7-step 2858710 loss: 0.684255 (accumulated 1 forwards) [2026-07-10 01:33:11-INFO]: epoch 7-step 2858720 loss: 0.675392 (accumulated 1 forwards) [2026-07-10 01:33:19-INFO]: epoch 7-step 2858730 loss: 0.629803 (accumulated 1 forwards) [2026-07-10 01:33:28-INFO]: epoch 7-step 2858740 loss: 0.665908 (accumulated 1 forwards) [2026-07-10 01:33:37-INFO]: epoch 7-step 2858750 loss: 0.670839 (accumulated 1 forwards) [2026-07-10 01:33:45-INFO]: epoch 7-step 2858760 loss: 0.685224 (accumulated 1 forwards) [2026-07-10 01:33:54-INFO]: epoch 7-step 2858770 loss: 0.686456 (accumulated 1 forwards) [2026-07-10 01:34:03-INFO]: epoch 7-step 2858780 loss: 0.648847 (accumulated 1 forwards) [2026-07-10 01:34:11-INFO]: epoch 7-step 2858790 loss: 0.713203 (accumulated 1 forwards) [2026-07-10 01:34:20-INFO]: epoch 7-step 2858800 loss: 0.677930 (accumulated 1 forwards) [2026-07-10 01:34:29-INFO]: epoch 7-step 2858810 loss: 0.659502 (accumulated 1 forwards) [2026-07-10 01:34:37-INFO]: epoch 7-step 2858820 loss: 0.636562 (accumulated 1 forwards) [2026-07-10 01:34:46-INFO]: epoch 7-step 2858830 loss: 0.677305 (accumulated 1 forwards) [2026-07-10 01:34:55-INFO]: epoch 7-step 2858840 loss: 0.680417 (accumulated 1 forwards) [2026-07-10 01:35:04-INFO]: epoch 7-step 2858850 loss: 0.663680 (accumulated 1 forwards) [2026-07-10 01:35:12-INFO]: epoch 7-step 2858860 loss: 0.657937 (accumulated 1 forwards) [2026-07-10 01:35:21-INFO]: epoch 7-step 2858870 loss: 0.641627 (accumulated 1 forwards) [2026-07-10 01:35:30-INFO]: epoch 7-step 2858880 loss: 0.684873 (accumulated 1 forwards) [2026-07-10 01:35:38-INFO]: epoch 7-step 2858890 loss: 0.683309 (accumulated 1 forwards) [2026-07-10 01:35:47-INFO]: epoch 7-step 2858900 loss: 0.693773 (accumulated 1 forwards) [2026-07-10 01:35:56-INFO]: epoch 7-step 2858910 loss: 0.662722 (accumulated 1 forwards) [2026-07-10 01:36:04-INFO]: epoch 7-step 2858920 loss: 0.680207 (accumulated 1 forwards) [2026-07-10 01:36:13-INFO]: epoch 7-step 2858930 loss: 0.639092 (accumulated 1 forwards) [2026-07-10 01:36:22-INFO]: epoch 7-step 2858940 loss: 0.603129 (accumulated 1 forwards) [2026-07-10 01:36:30-INFO]: epoch 7-step 2858950 loss: 0.683089 (accumulated 1 forwards) [2026-07-10 01:36:39-INFO]: epoch 7-step 2858960 loss: 0.661028 (accumulated 1 forwards) [2026-07-10 01:36:48-INFO]: epoch 7-step 2858970 loss: 0.655777 (accumulated 1 forwards) [2026-07-10 01:36:57-INFO]: epoch 7-step 2858980 loss: 0.657784 (accumulated 1 forwards) [2026-07-10 01:37:05-INFO]: epoch 7-step 2858990 loss: 0.718393 (accumulated 1 forwards) [2026-07-10 01:37:14-INFO]: epoch 7-step 2859000 loss: 0.622059 (accumulated 1 forwards) [2026-07-10 01:37:23-INFO]: epoch 7-step 2859010 loss: 0.642052 (accumulated 1 forwards) [2026-07-10 01:37:31-INFO]: epoch 7-step 2859020 loss: 0.665865 (accumulated 1 forwards) [2026-07-10 01:37:40-INFO]: epoch 7-step 2859030 loss: 0.656621 (accumulated 1 forwards) [2026-07-10 01:37:49-INFO]: epoch 7-step 2859040 loss: 0.682475 (accumulated 1 forwards) [2026-07-10 01:37:57-INFO]: epoch 7-step 2859050 loss: 0.667074 (accumulated 1 forwards) [2026-07-10 01:38:06-INFO]: epoch 7-step 2859060 loss: 0.671493 (accumulated 1 forwards) [2026-07-10 01:38:15-INFO]: epoch 7-step 2859070 loss: 0.674533 (accumulated 1 forwards) [2026-07-10 01:38:23-INFO]: epoch 7-step 2859080 loss: 0.684396 (accumulated 1 forwards) [2026-07-10 01:38:32-INFO]: epoch 7-step 2859090 loss: 0.653648 (accumulated 1 forwards) [2026-07-10 01:38:41-INFO]: epoch 7-step 2859100 loss: 0.626176 (accumulated 1 forwards) [2026-07-10 01:38:49-INFO]: epoch 7-step 2859110 loss: 0.698320 (accumulated 1 forwards) [2026-07-10 01:38:58-INFO]: epoch 7-step 2859120 loss: 0.645050 (accumulated 1 forwards) [2026-07-10 01:39:07-INFO]: epoch 7-step 2859130 loss: 0.647449 (accumulated 1 forwards) [2026-07-10 01:39:16-INFO]: epoch 7-step 2859140 loss: 0.610445 (accumulated 1 forwards) [2026-07-10 01:39:24-INFO]: epoch 7-step 2859150 loss: 0.667106 (accumulated 1 forwards) [2026-07-10 01:39:33-INFO]: epoch 7-step 2859160 loss: 0.666552 (accumulated 1 forwards) [2026-07-10 01:39:42-INFO]: epoch 7-step 2859170 loss: 0.690523 (accumulated 1 forwards) [2026-07-10 01:39:50-INFO]: epoch 7-step 2859180 loss: 0.653108 (accumulated 1 forwards) [2026-07-10 01:39:59-INFO]: epoch 7-step 2859190 loss: 0.699573 (accumulated 1 forwards) [2026-07-10 01:40:08-INFO]: epoch 7-step 2859200 loss: 0.655819 (accumulated 1 forwards) [2026-07-10 01:40:16-INFO]: epoch 7-step 2859210 loss: 0.661632 (accumulated 1 forwards) [2026-07-10 01:40:25-INFO]: epoch 7-step 2859220 loss: 0.662779 (accumulated 1 forwards) [2026-07-10 01:40:34-INFO]: epoch 7-step 2859230 loss: 0.676313 (accumulated 1 forwards) [2026-07-10 01:40:43-INFO]: epoch 7-step 2859240 loss: 0.630422 (accumulated 1 forwards) [2026-07-10 01:40:51-INFO]: epoch 7-step 2859250 loss: 0.645505 (accumulated 1 forwards) [2026-07-10 01:41:00-INFO]: epoch 7-step 2859260 loss: 0.655675 (accumulated 1 forwards) [2026-07-10 01:41:09-INFO]: epoch 7-step 2859270 loss: 0.644431 (accumulated 1 forwards) [2026-07-10 01:41:17-INFO]: epoch 7-step 2859280 loss: 0.708161 (accumulated 1 forwards) [2026-07-10 01:41:26-INFO]: epoch 7-step 2859290 loss: 0.639488 (accumulated 1 forwards) [2026-07-10 01:41:35-INFO]: epoch 7-step 2859300 loss: 0.643280 (accumulated 1 forwards) [2026-07-10 01:41:43-INFO]: epoch 7-step 2859310 loss: 0.704576 (accumulated 1 forwards) [2026-07-10 01:41:52-INFO]: epoch 7-step 2859320 loss: 0.652522 (accumulated 1 forwards) [2026-07-10 01:42:01-INFO]: epoch 7-step 2859330 loss: 0.702797 (accumulated 1 forwards) [2026-07-10 01:42:09-INFO]: epoch 7-step 2859340 loss: 0.668620 (accumulated 1 forwards) [2026-07-10 01:42:18-INFO]: epoch 7-step 2859350 loss: 0.698730 (accumulated 1 forwards) [2026-07-10 01:42:27-INFO]: epoch 7-step 2859360 loss: 0.645522 (accumulated 1 forwards) [2026-07-10 01:42:35-INFO]: epoch 7-step 2859370 loss: 0.636963 (accumulated 1 forwards) [2026-07-10 01:42:44-INFO]: epoch 7-step 2859380 loss: 0.639553 (accumulated 1 forwards) [2026-07-10 01:42:53-INFO]: epoch 7-step 2859390 loss: 0.652012 (accumulated 1 forwards) [2026-07-10 01:43:01-INFO]: epoch 7-step 2859400 loss: 0.655339 (accumulated 1 forwards) [2026-07-10 01:43:10-INFO]: epoch 7-step 2859410 loss: 0.665953 (accumulated 1 forwards) [2026-07-10 01:43:19-INFO]: epoch 7-step 2859420 loss: 0.666058 (accumulated 1 forwards) [2026-07-10 01:43:27-INFO]: epoch 7-step 2859430 loss: 0.650667 (accumulated 1 forwards) [2026-07-10 01:43:36-INFO]: epoch 7-step 2859440 loss: 0.678205 (accumulated 1 forwards) [2026-07-10 01:43:45-INFO]: epoch 7-step 2859450 loss: 0.619401 (accumulated 1 forwards) [2026-07-10 01:43:54-INFO]: epoch 7-step 2859460 loss: 0.695443 (accumulated 1 forwards) [2026-07-10 01:44:02-INFO]: epoch 7-step 2859470 loss: 0.617644 (accumulated 1 forwards) [2026-07-10 01:44:11-INFO]: epoch 7-step 2859480 loss: 0.684450 (accumulated 1 forwards) [2026-07-10 01:44:20-INFO]: epoch 7-step 2859490 loss: 0.668661 (accumulated 1 forwards) [2026-07-10 01:44:28-INFO]: epoch 7-step 2859500 loss: 0.698500 (accumulated 1 forwards) [2026-07-10 01:44:37-INFO]: epoch 7-step 2859510 loss: 0.675269 (accumulated 1 forwards) [2026-07-10 01:44:46-INFO]: epoch 7-step 2859520 loss: 0.645553 (accumulated 1 forwards) [2026-07-10 01:44:55-INFO]: epoch 7-step 2859530 loss: 0.634431 (accumulated 1 forwards) [2026-07-10 01:45:03-INFO]: epoch 7-step 2859540 loss: 0.667536 (accumulated 1 forwards) [2026-07-10 01:45:12-INFO]: epoch 7-step 2859550 loss: 0.678226 (accumulated 1 forwards) [2026-07-10 01:45:21-INFO]: epoch 7-step 2859560 loss: 0.615854 (accumulated 1 forwards) [2026-07-10 01:45:29-INFO]: epoch 7-step 2859570 loss: 0.639991 (accumulated 1 forwards) [2026-07-10 01:45:38-INFO]: epoch 7-step 2859580 loss: 0.641860 (accumulated 1 forwards) [2026-07-10 01:45:47-INFO]: epoch 7-step 2859590 loss: 0.645036 (accumulated 1 forwards) [2026-07-10 01:45:55-INFO]: epoch 7-step 2859600 loss: 0.648745 (accumulated 1 forwards) [2026-07-10 01:46:04-INFO]: epoch 7-step 2859610 loss: 0.633578 (accumulated 1 forwards) [2026-07-10 01:46:13-INFO]: epoch 7-step 2859620 loss: 0.689296 (accumulated 1 forwards) [2026-07-10 01:46:21-INFO]: epoch 7-step 2859630 loss: 0.670523 (accumulated 1 forwards) [2026-07-10 01:46:30-INFO]: epoch 7-step 2859640 loss: 0.682073 (accumulated 1 forwards) [2026-07-10 01:46:39-INFO]: epoch 7-step 2859650 loss: 0.707405 (accumulated 1 forwards) [2026-07-10 01:46:48-INFO]: epoch 7-step 2859660 loss: 0.693139 (accumulated 1 forwards) [2026-07-10 01:46:56-INFO]: epoch 7-step 2859670 loss: 0.693381 (accumulated 1 forwards) [2026-07-10 01:47:05-INFO]: epoch 7-step 2859680 loss: 0.646022 (accumulated 1 forwards) [2026-07-10 01:47:14-INFO]: epoch 7-step 2859690 loss: 0.709962 (accumulated 1 forwards) [2026-07-10 01:47:18-INFO]: Reload dataloader [2026-07-10 01:47:22-INFO]: epoch 8-step 2859700 loss: 0.663814 (accumulated 1 forwards) [2026-07-10 01:47:31-INFO]: epoch 8-step 2859710 loss: 0.655671 (accumulated 1 forwards) [2026-07-10 01:47:40-INFO]: epoch 8-step 2859720 loss: 0.608544 (accumulated 1 forwards) [2026-07-10 01:47:48-INFO]: epoch 8-step 2859730 loss: 0.647626 (accumulated 1 forwards) [2026-07-10 01:47:57-INFO]: epoch 8-step 2859740 loss: 0.694776 (accumulated 1 forwards) [2026-07-10 01:48:06-INFO]: epoch 8-step 2859750 loss: 0.739469 (accumulated 1 forwards) [2026-07-10 01:48:14-INFO]: epoch 8-step 2859760 loss: 0.681981 (accumulated 1 forwards) [2026-07-10 01:48:23-INFO]: epoch 8-step 2859770 loss: 0.615787 (accumulated 1 forwards) [2026-07-10 01:48:32-INFO]: epoch 8-step 2859780 loss: 0.702245 (accumulated 1 forwards) [2026-07-10 01:48:40-INFO]: epoch 8-step 2859790 loss: 0.661625 (accumulated 1 forwards) [2026-07-10 01:48:49-INFO]: epoch 8-step 2859800 loss: 0.699754 (accumulated 1 forwards) [2026-07-10 01:48:58-INFO]: epoch 8-step 2859810 loss: 0.660299 (accumulated 1 forwards) [2026-07-10 01:49:06-INFO]: epoch 8-step 2859820 loss: 0.677600 (accumulated 1 forwards) [2026-07-10 01:49:15-INFO]: epoch 8-step 2859830 loss: 0.663579 (accumulated 1 forwards) [2026-07-10 01:49:24-INFO]: epoch 8-step 2859840 loss: 0.654504 (accumulated 1 forwards) [2026-07-10 01:49:32-INFO]: epoch 8-step 2859850 loss: 0.668924 (accumulated 1 forwards) [2026-07-10 01:49:41-INFO]: epoch 8-step 2859860 loss: 0.715713 (accumulated 1 forwards) [2026-07-10 01:49:50-INFO]: epoch 8-step 2859870 loss: 0.711264 (accumulated 1 forwards) [2026-07-10 01:49:58-INFO]: epoch 8-step 2859880 loss: 0.687100 (accumulated 1 forwards) [2026-07-10 01:50:07-INFO]: epoch 8-step 2859890 loss: 0.652888 (accumulated 1 forwards) [2026-07-10 01:50:16-INFO]: epoch 8-step 2859900 loss: 0.620186 (accumulated 1 forwards) [2026-07-10 01:50:24-INFO]: epoch 8-step 2859910 loss: 0.653175 (accumulated 1 forwards) [2026-07-10 01:50:33-INFO]: epoch 8-step 2859920 loss: 0.672899 (accumulated 1 forwards) [2026-07-10 01:50:42-INFO]: epoch 8-step 2859930 loss: 0.683764 (accumulated 1 forwards) [2026-07-10 01:50:50-INFO]: epoch 8-step 2859940 loss: 0.646939 (accumulated 1 forwards) [2026-07-10 01:50:59-INFO]: epoch 8-step 2859950 loss: 0.674326 (accumulated 1 forwards) [2026-07-10 01:51:08-INFO]: epoch 8-step 2859960 loss: 0.705808 (accumulated 1 forwards) [2026-07-10 01:51:17-INFO]: epoch 8-step 2859970 loss: 0.682401 (accumulated 1 forwards) [2026-07-10 01:51:25-INFO]: epoch 8-step 2859980 loss: 0.669977 (accumulated 1 forwards) [2026-07-10 01:51:34-INFO]: epoch 8-step 2859990 loss: 0.696672 (accumulated 1 forwards) [2026-07-10 01:51:43-INFO]: epoch 8-step 2860000 loss: 0.664214 (accumulated 1 forwards) [2026-07-10 01:51:47-INFO]: Checkpoint saved at ProMoE_B/ProMoE_TC_B/stage3/checkpoints/ckpt_step_2860000.pth [2026-07-10 01:51:56-INFO]: epoch 8-step 2860010 loss: 0.656287 (accumulated 1 forwards) [2026-07-10 01:52:05-INFO]: epoch 8-step 2860020 loss: 0.676698 (accumulated 1 forwards) [2026-07-10 01:52:14-INFO]: epoch 8-step 2860030 loss: 0.697168 (accumulated 1 forwards) [2026-07-10 01:52:22-INFO]: epoch 8-step 2860040 loss: 0.682746 (accumulated 1 forwards) [2026-07-10 01:52:31-INFO]: epoch 8-step 2860050 loss: 0.678064 (accumulated 1 forwards) [2026-07-10 01:52:40-INFO]: epoch 8-step 2860060 loss: 0.648864 (accumulated 1 forwards) [2026-07-10 01:52:48-INFO]: epoch 8-step 2860070 loss: 0.672275 (accumulated 1 forwards) [2026-07-10 01:52:57-INFO]: epoch 8-step 2860080 loss: 0.669938 (accumulated 1 forwards) [2026-07-10 01:53:06-INFO]: epoch 8-step 2860090 loss: 0.641000 (accumulated 1 forwards) [2026-07-10 01:53:14-INFO]: epoch 8-step 2860100 loss: 0.659423 (accumulated 1 forwards) [2026-07-10 01:53:23-INFO]: epoch 8-step 2860110 loss: 0.668061 (accumulated 1 forwards) [2026-07-10 01:53:32-INFO]: epoch 8-step 2860120 loss: 0.660656 (accumulated 1 forwards) [2026-07-10 01:53:40-INFO]: epoch 8-step 2860130 loss: 0.686885 (accumulated 1 forwards) [2026-07-10 01:53:49-INFO]: epoch 8-step 2860140 loss: 0.632031 (accumulated 1 forwards) [2026-07-10 01:53:58-INFO]: epoch 8-step 2860150 loss: 0.667582 (accumulated 1 forwards) [2026-07-10 01:54:07-INFO]: epoch 8-step 2860160 loss: 0.669620 (accumulated 1 forwards) [2026-07-10 01:54:15-INFO]: epoch 8-step 2860170 loss: 0.665244 (accumulated 1 forwards) [2026-07-10 01:54:24-INFO]: epoch 8-step 2860180 loss: 0.622189 (accumulated 1 forwards) [2026-07-10 01:54:33-INFO]: epoch 8-step 2860190 loss: 0.633533 (accumulated 1 forwards) [2026-07-10 01:54:41-INFO]: epoch 8-step 2860200 loss: 0.672976 (accumulated 1 forwards) [2026-07-10 01:54:50-INFO]: epoch 8-step 2860210 loss: 0.656067 (accumulated 1 forwards) [2026-07-10 01:54:59-INFO]: epoch 8-step 2860220 loss: 0.643303 (accumulated 1 forwards) [2026-07-10 01:55:07-INFO]: epoch 8-step 2860230 loss: 0.631484 (accumulated 1 forwards) [2026-07-10 01:55:16-INFO]: epoch 8-step 2860240 loss: 0.674300 (accumulated 1 forwards) [2026-07-10 01:55:25-INFO]: epoch 8-step 2860250 loss: 0.686262 (accumulated 1 forwards) [2026-07-10 01:55:33-INFO]: epoch 8-step 2860260 loss: 0.649356 (accumulated 1 forwards) [2026-07-10 01:55:42-INFO]: epoch 8-step 2860270 loss: 0.648304 (accumulated 1 forwards) [2026-07-10 01:55:51-INFO]: epoch 8-step 2860280 loss: 0.693053 (accumulated 1 forwards) [2026-07-10 01:55:59-INFO]: epoch 8-step 2860290 loss: 0.625845 (accumulated 1 forwards) [2026-07-10 01:56:08-INFO]: epoch 8-step 2860300 loss: 0.640328 (accumulated 1 forwards) [2026-07-10 01:56:17-INFO]: epoch 8-step 2860310 loss: 0.646355 (accumulated 1 forwards) [2026-07-10 01:56:25-INFO]: epoch 8-step 2860320 loss: 0.672792 (accumulated 1 forwards) [2026-07-10 01:56:34-INFO]: epoch 8-step 2860330 loss: 0.662862 (accumulated 1 forwards) [2026-07-10 01:56:43-INFO]: epoch 8-step 2860340 loss: 0.635737 (accumulated 1 forwards) [2026-07-10 01:56:51-INFO]: epoch 8-step 2860350 loss: 0.647802 (accumulated 1 forwards) [2026-07-10 01:57:00-INFO]: epoch 8-step 2860360 loss: 0.644352 (accumulated 1 forwards) [2026-07-10 01:57:09-INFO]: epoch 8-step 2860370 loss: 0.647104 (accumulated 1 forwards) [2026-07-10 01:57:17-INFO]: epoch 8-step 2860380 loss: 0.649297 (accumulated 1 forwards) [2026-07-10 01:57:26-INFO]: epoch 8-step 2860390 loss: 0.672192 (accumulated 1 forwards) [2026-07-10 01:57:35-INFO]: epoch 8-step 2860400 loss: 0.673585 (accumulated 1 forwards) [2026-07-10 01:57:44-INFO]: epoch 8-step 2860410 loss: 0.664411 (accumulated 1 forwards) [2026-07-10 01:57:52-INFO]: epoch 8-step 2860420 loss: 0.667174 (accumulated 1 forwards) [2026-07-10 01:58:01-INFO]: epoch 8-step 2860430 loss: 0.669853 (accumulated 1 forwards) [2026-07-10 01:58:10-INFO]: epoch 8-step 2860440 loss: 0.671258 (accumulated 1 forwards) [2026-07-10 01:58:18-INFO]: epoch 8-step 2860450 loss: 0.675050 (accumulated 1 forwards) [2026-07-10 01:58:27-INFO]: epoch 8-step 2860460 loss: 0.645289 (accumulated 1 forwards) [2026-07-10 01:58:36-INFO]: epoch 8-step 2860470 loss: 0.697326 (accumulated 1 forwards) [2026-07-10 01:58:44-INFO]: epoch 8-step 2860480 loss: 0.660152 (accumulated 1 forwards) [2026-07-10 01:58:53-INFO]: epoch 8-step 2860490 loss: 0.683949 (accumulated 1 forwards) [2026-07-10 01:59:02-INFO]: epoch 8-step 2860500 loss: 0.674567 (accumulated 1 forwards) [2026-07-10 01:59:10-INFO]: epoch 8-step 2860510 loss: 0.674172 (accumulated 1 forwards) [2026-07-10 01:59:19-INFO]: epoch 8-step 2860520 loss: 0.655396 (accumulated 1 forwards) [2026-07-10 01:59:28-INFO]: epoch 8-step 2860530 loss: 0.673815 (accumulated 1 forwards) [2026-07-10 01:59:37-INFO]: epoch 8-step 2860540 loss: 0.676617 (accumulated 1 forwards) [2026-07-10 01:59:45-INFO]: epoch 8-step 2860550 loss: 0.626908 (accumulated 1 forwards) [2026-07-10 01:59:54-INFO]: epoch 8-step 2860560 loss: 0.676456 (accumulated 1 forwards) [2026-07-10 02:00:03-INFO]: epoch 8-step 2860570 loss: 0.701920 (accumulated 1 forwards) [2026-07-10 02:00:11-INFO]: epoch 8-step 2860580 loss: 0.686752 (accumulated 1 forwards) [2026-07-10 02:00:20-INFO]: epoch 8-step 2860590 loss: 0.692842 (accumulated 1 forwards) [2026-07-10 02:00:29-INFO]: epoch 8-step 2860600 loss: 0.698281 (accumulated 1 forwards) [2026-07-10 02:00:37-INFO]: epoch 8-step 2860610 loss: 0.656267 (accumulated 1 forwards) [2026-07-10 02:00:46-INFO]: epoch 8-step 2860620 loss: 0.676116 (accumulated 1 forwards) [2026-07-10 02:00:55-INFO]: epoch 8-step 2860630 loss: 0.660511 (accumulated 1 forwards) [2026-07-10 02:01:03-INFO]: epoch 8-step 2860640 loss: 0.684351 (accumulated 1 forwards) [2026-07-10 02:01:12-INFO]: epoch 8-step 2860650 loss: 0.689518 (accumulated 1 forwards) [2026-07-10 02:01:21-INFO]: epoch 8-step 2860660 loss: 0.695909 (accumulated 1 forwards) [2026-07-10 02:01:29-INFO]: epoch 8-step 2860670 loss: 0.667054 (accumulated 1 forwards) [2026-07-10 02:01:38-INFO]: epoch 8-step 2860680 loss: 0.671138 (accumulated 1 forwards) [2026-07-10 02:01:47-INFO]: epoch 8-step 2860690 loss: 0.683367 (accumulated 1 forwards) [2026-07-10 02:01:55-INFO]: epoch 8-step 2860700 loss: 0.686513 (accumulated 1 forwards) [2026-07-10 02:02:04-INFO]: epoch 8-step 2860710 loss: 0.641716 (accumulated 1 forwards) [2026-07-10 02:02:13-INFO]: epoch 8-step 2860720 loss: 0.672652 (accumulated 1 forwards) [2026-07-10 02:02:22-INFO]: epoch 8-step 2860730 loss: 0.696688 (accumulated 1 forwards) [2026-07-10 02:02:30-INFO]: epoch 8-step 2860740 loss: 0.609925 (accumulated 1 forwards) [2026-07-10 02:02:39-INFO]: epoch 8-step 2860750 loss: 0.660719 (accumulated 1 forwards) [2026-07-10 02:02:48-INFO]: epoch 8-step 2860760 loss: 0.711027 (accumulated 1 forwards) [2026-07-10 02:02:56-INFO]: epoch 8-step 2860770 loss: 0.672046 (accumulated 1 forwards) [2026-07-10 02:03:05-INFO]: epoch 8-step 2860780 loss: 0.666734 (accumulated 1 forwards) [2026-07-10 02:03:14-INFO]: epoch 8-step 2860790 loss: 0.620699 (accumulated 1 forwards) [2026-07-10 02:03:22-INFO]: epoch 8-step 2860800 loss: 0.658085 (accumulated 1 forwards) [2026-07-10 02:03:31-INFO]: epoch 8-step 2860810 loss: 0.642397 (accumulated 1 forwards) [2026-07-10 02:03:40-INFO]: epoch 8-step 2860820 loss: 0.665191 (accumulated 1 forwards) [2026-07-10 02:03:48-INFO]: epoch 8-step 2860830 loss: 0.721143 (accumulated 1 forwards) [2026-07-10 02:03:57-INFO]: epoch 8-step 2860840 loss: 0.658463 (accumulated 1 forwards) [2026-07-10 02:04:06-INFO]: epoch 8-step 2860850 loss: 0.681915 (accumulated 1 forwards) [2026-07-10 02:04:14-INFO]: epoch 8-step 2860860 loss: 0.658001 (accumulated 1 forwards) [2026-07-10 02:04:23-INFO]: epoch 8-step 2860870 loss: 0.708815 (accumulated 1 forwards) [2026-07-10 02:04:32-INFO]: epoch 8-step 2860880 loss: 0.667724 (accumulated 1 forwards) [2026-07-10 02:04:41-INFO]: epoch 8-step 2860890 loss: 0.670417 (accumulated 1 forwards) [2026-07-10 02:04:49-INFO]: epoch 8-step 2860900 loss: 0.649652 (accumulated 1 forwards) [2026-07-10 02:04:56-INFO]: Reload dataloader [2026-07-10 02:04:58-INFO]: epoch 9-step 2860910 loss: 0.614475 (accumulated 1 forwards) [2026-07-10 02:05:06-INFO]: epoch 9-step 2860920 loss: 0.657569 (accumulated 1 forwards) [2026-07-10 02:05:15-INFO]: epoch 9-step 2860930 loss: 0.658706 (accumulated 1 forwards) [2026-07-10 02:05:24-INFO]: epoch 9-step 2860940 loss: 0.662861 (accumulated 1 forwards) [2026-07-10 02:05:33-INFO]: epoch 9-step 2860950 loss: 0.690261 (accumulated 1 forwards) [2026-07-10 02:05:41-INFO]: epoch 9-step 2860960 loss: 0.655188 (accumulated 1 forwards) [2026-07-10 02:05:50-INFO]: epoch 9-step 2860970 loss: 0.649264 (accumulated 1 forwards) [2026-07-10 02:05:59-INFO]: epoch 9-step 2860980 loss: 0.701862 (accumulated 1 forwards) [2026-07-10 02:06:07-INFO]: epoch 9-step 2860990 loss: 0.646787 (accumulated 1 forwards) [2026-07-10 02:06:16-INFO]: epoch 9-step 2861000 loss: 0.652574 (accumulated 1 forwards) [2026-07-10 02:06:25-INFO]: epoch 9-step 2861010 loss: 0.629282 (accumulated 1 forwards) [2026-07-10 02:06:33-INFO]: epoch 9-step 2861020 loss: 0.671633 (accumulated 1 forwards) [2026-07-10 02:06:42-INFO]: epoch 9-step 2861030 loss: 0.656005 (accumulated 1 forwards) [2026-07-10 02:06:51-INFO]: epoch 9-step 2861040 loss: 0.648532 (accumulated 1 forwards) [2026-07-10 02:06:59-INFO]: epoch 9-step 2861050 loss: 0.682052 (accumulated 1 forwards) [2026-07-10 02:07:08-INFO]: epoch 9-step 2861060 loss: 0.685593 (accumulated 1 forwards) [2026-07-10 02:07:17-INFO]: epoch 9-step 2861070 loss: 0.704776 (accumulated 1 forwards) [2026-07-10 02:07:26-INFO]: epoch 9-step 2861080 loss: 0.653202 (accumulated 1 forwards) [2026-07-10 02:07:34-INFO]: epoch 9-step 2861090 loss: 0.643652 (accumulated 1 forwards) [2026-07-10 02:07:43-INFO]: epoch 9-step 2861100 loss: 0.647498 (accumulated 1 forwards) [2026-07-10 02:07:52-INFO]: epoch 9-step 2861110 loss: 0.665303 (accumulated 1 forwards) [2026-07-10 02:08:00-INFO]: epoch 9-step 2861120 loss: 0.692870 (accumulated 1 forwards) [2026-07-10 02:08:09-INFO]: epoch 9-step 2861130 loss: 0.659129 (accumulated 1 forwards) [2026-07-10 02:08:18-INFO]: epoch 9-step 2861140 loss: 0.644009 (accumulated 1 forwards) [2026-07-10 02:08:26-INFO]: epoch 9-step 2861150 loss: 0.642607 (accumulated 1 forwards) [2026-07-10 02:08:35-INFO]: epoch 9-step 2861160 loss: 0.609205 (accumulated 1 forwards) [2026-07-10 02:08:44-INFO]: epoch 9-step 2861170 loss: 0.712421 (accumulated 1 forwards) [2026-07-10 02:08:52-INFO]: epoch 9-step 2861180 loss: 0.657484 (accumulated 1 forwards) [2026-07-10 02:09:01-INFO]: epoch 9-step 2861190 loss: 0.662810 (accumulated 1 forwards) [2026-07-10 02:09:10-INFO]: epoch 9-step 2861200 loss: 0.642761 (accumulated 1 forwards) [2026-07-10 02:09:19-INFO]: epoch 9-step 2861210 loss: 0.622543 (accumulated 1 forwards) [2026-07-10 02:09:27-INFO]: epoch 9-step 2861220 loss: 0.628026 (accumulated 1 forwards) [2026-07-10 02:09:36-INFO]: epoch 9-step 2861230 loss: 0.688868 (accumulated 1 forwards) [2026-07-10 02:09:45-INFO]: epoch 9-step 2861240 loss: 0.668809 (accumulated 1 forwards) [2026-07-10 02:09:53-INFO]: epoch 9-step 2861250 loss: 0.672445 (accumulated 1 forwards) [2026-07-10 02:10:02-INFO]: epoch 9-step 2861260 loss: 0.666830 (accumulated 1 forwards) [2026-07-10 02:10:11-INFO]: epoch 9-step 2861270 loss: 0.642525 (accumulated 1 forwards) [2026-07-10 02:10:19-INFO]: epoch 9-step 2861280 loss: 0.633415 (accumulated 1 forwards) [2026-07-10 02:10:28-INFO]: epoch 9-step 2861290 loss: 0.700456 (accumulated 1 forwards) [2026-07-10 02:10:37-INFO]: epoch 9-step 2861300 loss: 0.633276 (accumulated 1 forwards) [2026-07-10 02:10:45-INFO]: epoch 9-step 2861310 loss: 0.692191 (accumulated 1 forwards) [2026-07-10 02:10:54-INFO]: epoch 9-step 2861320 loss: 0.698008 (accumulated 1 forwards) [2026-07-10 02:11:03-INFO]: epoch 9-step 2861330 loss: 0.658437 (accumulated 1 forwards) [2026-07-10 02:11:11-INFO]: epoch 9-step 2861340 loss: 0.649514 (accumulated 1 forwards) [2026-07-10 02:11:20-INFO]: epoch 9-step 2861350 loss: 0.642724 (accumulated 1 forwards) [2026-07-10 02:11:29-INFO]: epoch 9-step 2861360 loss: 0.690942 (accumulated 1 forwards) [2026-07-10 02:11:37-INFO]: epoch 9-step 2861370 loss: 0.650075 (accumulated 1 forwards) [2026-07-10 02:11:46-INFO]: epoch 9-step 2861380 loss: 0.653869 (accumulated 1 forwards) [2026-07-10 02:11:55-INFO]: epoch 9-step 2861390 loss: 0.646953 (accumulated 1 forwards) [2026-07-10 02:12:04-INFO]: epoch 9-step 2861400 loss: 0.639917 (accumulated 1 forwards) [2026-07-10 02:12:12-INFO]: epoch 9-step 2861410 loss: 0.694389 (accumulated 1 forwards) [2026-07-10 02:12:21-INFO]: epoch 9-step 2861420 loss: 0.656960 (accumulated 1 forwards) [2026-07-10 02:12:30-INFO]: epoch 9-step 2861430 loss: 0.611373 (accumulated 1 forwards) [2026-07-10 02:12:39-INFO]: epoch 9-step 2861440 loss: 0.669125 (accumulated 1 forwards) [2026-07-10 02:12:47-INFO]: epoch 9-step 2861450 loss: 0.689638 (accumulated 1 forwards) [2026-07-10 02:12:56-INFO]: epoch 9-step 2861460 loss: 0.620970 (accumulated 1 forwards) [2026-07-10 02:13:05-INFO]: epoch 9-step 2861470 loss: 0.665207 (accumulated 1 forwards) [2026-07-10 02:13:13-INFO]: epoch 9-step 2861480 loss: 0.740614 (accumulated 1 forwards) [2026-07-10 02:13:22-INFO]: epoch 9-step 2861490 loss: 0.695214 (accumulated 1 forwards) [2026-07-10 02:13:31-INFO]: epoch 9-step 2861500 loss: 0.632343 (accumulated 1 forwards) [2026-07-10 02:13:39-INFO]: epoch 9-step 2861510 loss: 0.700747 (accumulated 1 forwards) [2026-07-10 02:13:48-INFO]: epoch 9-step 2861520 loss: 0.671288 (accumulated 1 forwards) [2026-07-10 02:13:57-INFO]: epoch 9-step 2861530 loss: 0.675641 (accumulated 1 forwards) [2026-07-10 02:14:05-INFO]: epoch 9-step 2861540 loss: 0.618677 (accumulated 1 forwards) [2026-07-10 02:14:14-INFO]: epoch 9-step 2861550 loss: 0.675553 (accumulated 1 forwards) [2026-07-10 02:14:23-INFO]: epoch 9-step 2861560 loss: 0.663714 (accumulated 1 forwards) [2026-07-10 02:14:31-INFO]: epoch 9-step 2861570 loss: 0.661596 (accumulated 1 forwards) [2026-07-10 02:14:40-INFO]: epoch 9-step 2861580 loss: 0.628429 (accumulated 1 forwards) [2026-07-10 02:14:49-INFO]: epoch 9-step 2861590 loss: 0.655185 (accumulated 1 forwards) [2026-07-10 02:14:57-INFO]: epoch 9-step 2861600 loss: 0.633471 (accumulated 1 forwards) [2026-07-10 02:15:06-INFO]: epoch 9-step 2861610 loss: 0.659840 (accumulated 1 forwards) [2026-07-10 02:15:15-INFO]: epoch 9-step 2861620 loss: 0.669204 (accumulated 1 forwards) [2026-07-10 02:15:24-INFO]: epoch 9-step 2861630 loss: 0.650899 (accumulated 1 forwards) [2026-07-10 02:15:32-INFO]: epoch 9-step 2861640 loss: 0.665867 (accumulated 1 forwards) [2026-07-10 02:15:41-INFO]: epoch 9-step 2861650 loss: 0.662939 (accumulated 1 forwards) [2026-07-10 02:15:50-INFO]: epoch 9-step 2861660 loss: 0.703893 (accumulated 1 forwards) [2026-07-10 02:15:58-INFO]: epoch 9-step 2861670 loss: 0.669210 (accumulated 1 forwards) [2026-07-10 02:16:07-INFO]: epoch 9-step 2861680 loss: 0.699617 (accumulated 1 forwards) [2026-07-10 02:16:16-INFO]: epoch 9-step 2861690 loss: 0.682347 (accumulated 1 forwards) [2026-07-10 02:16:24-INFO]: epoch 9-step 2861700 loss: 0.681317 (accumulated 1 forwards) [2026-07-10 02:16:33-INFO]: epoch 9-step 2861710 loss: 0.664761 (accumulated 1 forwards) [2026-07-10 02:16:42-INFO]: epoch 9-step 2861720 loss: 0.677935 (accumulated 1 forwards) [2026-07-10 02:16:50-INFO]: epoch 9-step 2861730 loss: 0.655135 (accumulated 1 forwards) [2026-07-10 02:16:59-INFO]: epoch 9-step 2861740 loss: 0.659691 (accumulated 1 forwards) [2026-07-10 02:17:08-INFO]: epoch 9-step 2861750 loss: 0.627517 (accumulated 1 forwards) [2026-07-10 02:17:16-INFO]: epoch 9-step 2861760 loss: 0.681994 (accumulated 1 forwards) [2026-07-10 02:17:25-INFO]: epoch 9-step 2861770 loss: 0.661265 (accumulated 1 forwards) [2026-07-10 02:17:34-INFO]: epoch 9-step 2861780 loss: 0.661305 (accumulated 1 forwards) [2026-07-10 02:17:42-INFO]: epoch 9-step 2861790 loss: 0.654164 (accumulated 1 forwards) [2026-07-10 02:17:51-INFO]: epoch 9-step 2861800 loss: 0.620940 (accumulated 1 forwards) [2026-07-10 02:18:00-INFO]: epoch 9-step 2861810 loss: 0.638301 (accumulated 1 forwards) [2026-07-10 02:18:09-INFO]: epoch 9-step 2861820 loss: 0.678198 (accumulated 1 forwards) [2026-07-10 02:18:17-INFO]: epoch 9-step 2861830 loss: 0.628288 (accumulated 1 forwards) [2026-07-10 02:18:26-INFO]: epoch 9-step 2861840 loss: 0.645351 (accumulated 1 forwards) [2026-07-10 02:18:35-INFO]: epoch 9-step 2861850 loss: 0.661371 (accumulated 1 forwards) [2026-07-10 02:18:43-INFO]: epoch 9-step 2861860 loss: 0.693216 (accumulated 1 forwards) [2026-07-10 02:18:52-INFO]: epoch 9-step 2861870 loss: 0.654639 (accumulated 1 forwards) [2026-07-10 02:19:01-INFO]: epoch 9-step 2861880 loss: 0.663837 (accumulated 1 forwards) [2026-07-10 02:19:09-INFO]: epoch 9-step 2861890 loss: 0.645039 (accumulated 1 forwards) [2026-07-10 02:19:18-INFO]: epoch 9-step 2861900 loss: 0.659108 (accumulated 1 forwards) [2026-07-10 02:19:27-INFO]: epoch 9-step 2861910 loss: 0.651940 (accumulated 1 forwards) [2026-07-10 02:19:35-INFO]: epoch 9-step 2861920 loss: 0.660148 (accumulated 1 forwards) [2026-07-10 02:19:44-INFO]: epoch 9-step 2861930 loss: 0.621907 (accumulated 1 forwards) [2026-07-10 02:19:53-INFO]: epoch 9-step 2861940 loss: 0.650183 (accumulated 1 forwards) [2026-07-10 02:20:01-INFO]: epoch 9-step 2861950 loss: 0.655038 (accumulated 1 forwards) [2026-07-10 02:20:10-INFO]: epoch 9-step 2861960 loss: 0.702233 (accumulated 1 forwards) [2026-07-10 02:20:19-INFO]: epoch 9-step 2861970 loss: 0.642476 (accumulated 1 forwards) [2026-07-10 02:20:27-INFO]: epoch 9-step 2861980 loss: 0.648015 (accumulated 1 forwards) [2026-07-10 02:20:36-INFO]: epoch 9-step 2861990 loss: 0.647450 (accumulated 1 forwards) [2026-07-10 02:20:45-INFO]: epoch 9-step 2862000 loss: 0.633157 (accumulated 1 forwards) [2026-07-10 02:20:53-INFO]: epoch 9-step 2862010 loss: 0.647186 (accumulated 1 forwards) [2026-07-10 02:21:02-INFO]: epoch 9-step 2862020 loss: 0.635970 (accumulated 1 forwards) [2026-07-10 02:21:11-INFO]: epoch 9-step 2862030 loss: 0.673401 (accumulated 1 forwards) [2026-07-10 02:21:19-INFO]: epoch 9-step 2862040 loss: 0.684991 (accumulated 1 forwards) [2026-07-10 02:21:28-INFO]: epoch 9-step 2862050 loss: 0.657396 (accumulated 1 forwards) [2026-07-10 02:21:37-INFO]: epoch 9-step 2862060 loss: 0.652487 (accumulated 1 forwards) [2026-07-10 02:21:46-INFO]: epoch 9-step 2862070 loss: 0.636996 (accumulated 1 forwards) [2026-07-10 02:21:54-INFO]: epoch 9-step 2862080 loss: 0.678051 (accumulated 1 forwards) [2026-07-10 02:22:03-INFO]: epoch 9-step 2862090 loss: 0.683899 (accumulated 1 forwards) [2026-07-10 02:22:12-INFO]: epoch 9-step 2862100 loss: 0.674316 (accumulated 1 forwards) [2026-07-10 02:22:21-INFO]: epoch 9-step 2862110 loss: 0.675145 (accumulated 1 forwards) [2026-07-10 02:22:29-INFO]: epoch 9-step 2862120 loss: 0.696225 (accumulated 1 forwards) [2026-07-10 02:22:29-INFO]: Reload dataloader [2026-07-10 02:22:38-INFO]: epoch 10-step 2862130 loss: 0.668848 (accumulated 1 forwards) [2026-07-10 02:22:46-INFO]: epoch 10-step 2862140 loss: 0.722013 (accumulated 1 forwards) [2026-07-10 02:22:55-INFO]: epoch 10-step 2862150 loss: 0.616555 (accumulated 1 forwards) [2026-07-10 02:23:04-INFO]: epoch 10-step 2862160 loss: 0.684895 (accumulated 1 forwards) [2026-07-10 02:23:12-INFO]: epoch 10-step 2862170 loss: 0.657601 (accumulated 1 forwards) [2026-07-10 02:23:21-INFO]: epoch 10-step 2862180 loss: 0.637277 (accumulated 1 forwards) [2026-07-10 02:23:30-INFO]: epoch 10-step 2862190 loss: 0.707098 (accumulated 1 forwards) [2026-07-10 02:23:39-INFO]: epoch 10-step 2862200 loss: 0.652003 (accumulated 1 forwards) [2026-07-10 02:23:47-INFO]: epoch 10-step 2862210 loss: 0.650444 (accumulated 1 forwards) [2026-07-10 02:23:56-INFO]: epoch 10-step 2862220 loss: 0.666759 (accumulated 1 forwards) [2026-07-10 02:24:05-INFO]: epoch 10-step 2862230 loss: 0.620109 (accumulated 1 forwards) [2026-07-10 02:24:13-INFO]: epoch 10-step 2862240 loss: 0.614231 (accumulated 1 forwards) [2026-07-10 02:24:22-INFO]: epoch 10-step 2862250 loss: 0.635284 (accumulated 1 forwards) [2026-07-10 02:24:31-INFO]: epoch 10-step 2862260 loss: 0.655362 (accumulated 1 forwards) [2026-07-10 02:24:39-INFO]: epoch 10-step 2862270 loss: 0.616233 (accumulated 1 forwards) [2026-07-10 02:24:48-INFO]: epoch 10-step 2862280 loss: 0.642527 (accumulated 1 forwards) [2026-07-10 02:24:57-INFO]: epoch 10-step 2862290 loss: 0.639496 (accumulated 1 forwards) [2026-07-10 02:25:05-INFO]: epoch 10-step 2862300 loss: 0.684836 (accumulated 1 forwards) [2026-07-10 02:25:14-INFO]: epoch 10-step 2862310 loss: 0.598768 (accumulated 1 forwards) [2026-07-10 02:25:23-INFO]: epoch 10-step 2862320 loss: 0.618120 (accumulated 1 forwards) [2026-07-10 02:25:31-INFO]: epoch 10-step 2862330 loss: 0.627147 (accumulated 1 forwards) [2026-07-10 02:25:40-INFO]: epoch 10-step 2862340 loss: 0.684105 (accumulated 1 forwards) [2026-07-10 02:25:49-INFO]: epoch 10-step 2862350 loss: 0.686996 (accumulated 1 forwards) [2026-07-10 02:25:57-INFO]: epoch 10-step 2862360 loss: 0.646475 (accumulated 1 forwards) [2026-07-10 02:26:06-INFO]: epoch 10-step 2862370 loss: 0.683386 (accumulated 1 forwards) [2026-07-10 02:26:15-INFO]: epoch 10-step 2862380 loss: 0.672304 (accumulated 1 forwards) [2026-07-10 02:26:23-INFO]: epoch 10-step 2862390 loss: 0.638296 (accumulated 1 forwards) [2026-07-10 02:26:32-INFO]: epoch 10-step 2862400 loss: 0.658372 (accumulated 1 forwards) [2026-07-10 02:26:41-INFO]: epoch 10-step 2862410 loss: 0.655854 (accumulated 1 forwards) [2026-07-10 02:26:50-INFO]: epoch 10-step 2862420 loss: 0.662991 (accumulated 1 forwards) [2026-07-10 02:26:58-INFO]: epoch 10-step 2862430 loss: 0.655515 (accumulated 1 forwards) [2026-07-10 02:27:07-INFO]: epoch 10-step 2862440 loss: 0.648671 (accumulated 1 forwards) [2026-07-10 02:27:16-INFO]: epoch 10-step 2862450 loss: 0.612365 (accumulated 1 forwards) [2026-07-10 02:27:24-INFO]: epoch 10-step 2862460 loss: 0.641977 (accumulated 1 forwards) [2026-07-10 02:27:33-INFO]: epoch 10-step 2862470 loss: 0.676682 (accumulated 1 forwards) [2026-07-10 02:27:42-INFO]: epoch 10-step 2862480 loss: 0.714406 (accumulated 1 forwards) [2026-07-10 02:27:50-INFO]: epoch 10-step 2862490 loss: 0.657012 (accumulated 1 forwards) [2026-07-10 02:27:59-INFO]: epoch 10-step 2862500 loss: 0.677941 (accumulated 1 forwards) [2026-07-10 02:28:08-INFO]: epoch 10-step 2862510 loss: 0.634130 (accumulated 1 forwards) [2026-07-10 02:28:16-INFO]: epoch 10-step 2862520 loss: 0.655039 (accumulated 1 forwards) [2026-07-10 02:28:25-INFO]: epoch 10-step 2862530 loss: 0.653988 (accumulated 1 forwards) [2026-07-10 02:28:34-INFO]: epoch 10-step 2862540 loss: 0.630433 (accumulated 1 forwards) [2026-07-10 02:28:42-INFO]: epoch 10-step 2862550 loss: 0.654034 (accumulated 1 forwards) [2026-07-10 02:28:51-INFO]: epoch 10-step 2862560 loss: 0.655204 (accumulated 1 forwards) [2026-07-10 02:29:00-INFO]: epoch 10-step 2862570 loss: 0.613909 (accumulated 1 forwards) [2026-07-10 02:29:09-INFO]: epoch 10-step 2862580 loss: 0.686422 (accumulated 1 forwards) [2026-07-10 02:29:17-INFO]: epoch 10-step 2862590 loss: 0.678521 (accumulated 1 forwards) [2026-07-10 02:29:26-INFO]: epoch 10-step 2862600 loss: 0.659338 (accumulated 1 forwards) [2026-07-10 02:29:35-INFO]: epoch 10-step 2862610 loss: 0.671611 (accumulated 1 forwards) [2026-07-10 02:29:43-INFO]: epoch 10-step 2862620 loss: 0.681126 (accumulated 1 forwards) [2026-07-10 02:29:52-INFO]: epoch 10-step 2862630 loss: 0.681417 (accumulated 1 forwards) [2026-07-10 02:30:01-INFO]: epoch 10-step 2862640 loss: 0.634400 (accumulated 1 forwards) [2026-07-10 02:30:09-INFO]: epoch 10-step 2862650 loss: 0.666378 (accumulated 1 forwards) [2026-07-10 02:30:18-INFO]: epoch 10-step 2862660 loss: 0.675304 (accumulated 1 forwards) [2026-07-10 02:30:27-INFO]: epoch 10-step 2862670 loss: 0.696017 (accumulated 1 forwards) [2026-07-10 02:30:36-INFO]: epoch 10-step 2862680 loss: 0.705611 (accumulated 1 forwards) [2026-07-10 02:30:44-INFO]: epoch 10-step 2862690 loss: 0.668912 (accumulated 1 forwards) [2026-07-10 02:30:53-INFO]: epoch 10-step 2862700 loss: 0.683139 (accumulated 1 forwards) [2026-07-10 02:31:02-INFO]: epoch 10-step 2862710 loss: 0.661226 (accumulated 1 forwards) [2026-07-10 02:31:10-INFO]: epoch 10-step 2862720 loss: 0.696181 (accumulated 1 forwards) [2026-07-10 02:31:19-INFO]: epoch 10-step 2862730 loss: 0.644688 (accumulated 1 forwards) [2026-07-10 02:31:28-INFO]: epoch 10-step 2862740 loss: 0.630247 (accumulated 1 forwards) [2026-07-10 02:31:36-INFO]: epoch 10-step 2862750 loss: 0.669282 (accumulated 1 forwards) [2026-07-10 02:31:45-INFO]: epoch 10-step 2862760 loss: 0.675457 (accumulated 1 forwards) [2026-07-10 02:31:54-INFO]: epoch 10-step 2862770 loss: 0.664218 (accumulated 1 forwards) [2026-07-10 02:32:02-INFO]: epoch 10-step 2862780 loss: 0.666150 (accumulated 1 forwards) [2026-07-10 02:32:11-INFO]: epoch 10-step 2862790 loss: 0.671950 (accumulated 1 forwards) [2026-07-10 02:32:20-INFO]: epoch 10-step 2862800 loss: 0.669241 (accumulated 1 forwards) [2026-07-10 02:32:28-INFO]: epoch 10-step 2862810 loss: 0.667983 (accumulated 1 forwards) [2026-07-10 02:32:37-INFO]: epoch 10-step 2862820 loss: 0.657809 (accumulated 1 forwards) [2026-07-10 02:32:46-INFO]: epoch 10-step 2862830 loss: 0.654794 (accumulated 1 forwards) [2026-07-10 02:32:54-INFO]: epoch 10-step 2862840 loss: 0.619152 (accumulated 1 forwards) [2026-07-10 02:33:03-INFO]: epoch 10-step 2862850 loss: 0.628095 (accumulated 1 forwards) [2026-07-10 02:33:12-INFO]: epoch 10-step 2862860 loss: 0.651106 (accumulated 1 forwards) [2026-07-10 02:33:20-INFO]: epoch 10-step 2862870 loss: 0.666519 (accumulated 1 forwards) [2026-07-10 02:33:29-INFO]: epoch 10-step 2862880 loss: 0.613946 (accumulated 1 forwards) [2026-07-10 02:33:38-INFO]: epoch 10-step 2862890 loss: 0.664949 (accumulated 1 forwards) [2026-07-10 02:33:46-INFO]: epoch 10-step 2862900 loss: 0.632688 (accumulated 1 forwards) [2026-07-10 02:33:55-INFO]: epoch 10-step 2862910 loss: 0.641062 (accumulated 1 forwards) [2026-07-10 02:34:04-INFO]: epoch 10-step 2862920 loss: 0.669362 (accumulated 1 forwards) [2026-07-10 02:34:13-INFO]: epoch 10-step 2862930 loss: 0.680837 (accumulated 1 forwards) [2026-07-10 02:34:21-INFO]: epoch 10-step 2862940 loss: 0.603763 (accumulated 1 forwards) [2026-07-10 02:34:30-INFO]: epoch 10-step 2862950 loss: 0.618430 (accumulated 1 forwards) [2026-07-10 02:34:39-INFO]: epoch 10-step 2862960 loss: 0.664808 (accumulated 1 forwards) [2026-07-10 02:34:47-INFO]: epoch 10-step 2862970 loss: 0.688890 (accumulated 1 forwards) [2026-07-10 02:34:56-INFO]: epoch 10-step 2862980 loss: 0.642729 (accumulated 1 forwards) [2026-07-10 02:35:05-INFO]: epoch 10-step 2862990 loss: 0.624551 (accumulated 1 forwards) [2026-07-10 02:35:13-INFO]: epoch 10-step 2863000 loss: 0.699962 (accumulated 1 forwards) [2026-07-10 02:35:22-INFO]: epoch 10-step 2863010 loss: 0.705030 (accumulated 1 forwards) [2026-07-10 02:35:31-INFO]: epoch 10-step 2863020 loss: 0.667164 (accumulated 1 forwards) [2026-07-10 02:35:39-INFO]: epoch 10-step 2863030 loss: 0.637889 (accumulated 1 forwards) [2026-07-10 02:35:48-INFO]: epoch 10-step 2863040 loss: 0.662099 (accumulated 1 forwards) [2026-07-10 02:35:57-INFO]: epoch 10-step 2863050 loss: 0.632582 (accumulated 1 forwards) [2026-07-10 02:36:05-INFO]: epoch 10-step 2863060 loss: 0.671726 (accumulated 1 forwards) [2026-07-10 02:36:14-INFO]: epoch 10-step 2863070 loss: 0.679042 (accumulated 1 forwards) [2026-07-10 02:36:23-INFO]: epoch 10-step 2863080 loss: 0.634773 (accumulated 1 forwards) [2026-07-10 02:36:31-INFO]: epoch 10-step 2863090 loss: 0.712809 (accumulated 1 forwards) [2026-07-10 02:36:40-INFO]: epoch 10-step 2863100 loss: 0.677256 (accumulated 1 forwards) [2026-07-10 02:36:49-INFO]: epoch 10-step 2863110 loss: 0.672752 (accumulated 1 forwards) [2026-07-10 02:36:58-INFO]: epoch 10-step 2863120 loss: 0.663545 (accumulated 1 forwards) [2026-07-10 02:37:06-INFO]: epoch 10-step 2863130 loss: 0.605191 (accumulated 1 forwards) [2026-07-10 02:37:15-INFO]: epoch 10-step 2863140 loss: 0.653838 (accumulated 1 forwards) [2026-07-10 02:37:24-INFO]: epoch 10-step 2863150 loss: 0.679500 (accumulated 1 forwards) [2026-07-10 02:37:32-INFO]: epoch 10-step 2863160 loss: 0.680724 (accumulated 1 forwards) [2026-07-10 02:37:41-INFO]: epoch 10-step 2863170 loss: 0.621746 (accumulated 1 forwards) [2026-07-10 02:37:50-INFO]: epoch 10-step 2863180 loss: 0.591069 (accumulated 1 forwards) [2026-07-10 02:37:58-INFO]: epoch 10-step 2863190 loss: 0.677912 (accumulated 1 forwards) [2026-07-10 02:38:07-INFO]: epoch 10-step 2863200 loss: 0.667906 (accumulated 1 forwards) [2026-07-10 02:38:16-INFO]: epoch 10-step 2863210 loss: 0.633132 (accumulated 1 forwards) [2026-07-10 02:38:24-INFO]: epoch 10-step 2863220 loss: 0.663359 (accumulated 1 forwards) [2026-07-10 02:38:33-INFO]: epoch 10-step 2863230 loss: 0.658899 (accumulated 1 forwards) [2026-07-10 02:38:42-INFO]: epoch 10-step 2863240 loss: 0.695460 (accumulated 1 forwards) [2026-07-10 02:38:50-INFO]: epoch 10-step 2863250 loss: 0.684124 (accumulated 1 forwards) [2026-07-10 02:38:59-INFO]: epoch 10-step 2863260 loss: 0.678537 (accumulated 1 forwards) [2026-07-10 02:39:08-INFO]: epoch 10-step 2863270 loss: 0.649705 (accumulated 1 forwards) [2026-07-10 02:39:16-INFO]: epoch 10-step 2863280 loss: 0.623067 (accumulated 1 forwards) [2026-07-10 02:39:25-INFO]: epoch 10-step 2863290 loss: 0.622078 (accumulated 1 forwards) [2026-07-10 02:39:34-INFO]: epoch 10-step 2863300 loss: 0.650039 (accumulated 1 forwards) [2026-07-10 02:39:43-INFO]: epoch 10-step 2863310 loss: 0.685158 (accumulated 1 forwards) [2026-07-10 02:39:51-INFO]: epoch 10-step 2863320 loss: 0.678631 (accumulated 1 forwards) [2026-07-10 02:40:00-INFO]: epoch 10-step 2863330 loss: 0.666362 (accumulated 1 forwards) [2026-07-10 02:40:01-INFO]: Reload dataloader [2026-07-10 02:40:08-INFO]: epoch 11-step 2863340 loss: 0.637109 (accumulated 1 forwards) [2026-07-10 02:40:17-INFO]: epoch 11-step 2863350 loss: 0.657981 (accumulated 1 forwards) [2026-07-10 02:40:26-INFO]: epoch 11-step 2863360 loss: 0.676817 (accumulated 1 forwards) [2026-07-10 02:40:35-INFO]: epoch 11-step 2863370 loss: 0.664201 (accumulated 1 forwards) [2026-07-10 02:40:43-INFO]: epoch 11-step 2863380 loss: 0.684883 (accumulated 1 forwards) [2026-07-10 02:40:52-INFO]: epoch 11-step 2863390 loss: 0.654635 (accumulated 1 forwards) [2026-07-10 02:41:01-INFO]: epoch 11-step 2863400 loss: 0.645837 (accumulated 1 forwards) [2026-07-10 02:41:09-INFO]: epoch 11-step 2863410 loss: 0.630473 (accumulated 1 forwards) [2026-07-10 02:41:18-INFO]: epoch 11-step 2863420 loss: 0.660509 (accumulated 1 forwards) [2026-07-10 02:41:27-INFO]: epoch 11-step 2863430 loss: 0.642985 (accumulated 1 forwards) [2026-07-10 02:41:35-INFO]: epoch 11-step 2863440 loss: 0.687667 (accumulated 1 forwards) [2026-07-10 02:41:44-INFO]: epoch 11-step 2863450 loss: 0.657467 (accumulated 1 forwards) [2026-07-10 02:41:53-INFO]: epoch 11-step 2863460 loss: 0.670451 (accumulated 1 forwards) [2026-07-10 02:42:02-INFO]: epoch 11-step 2863470 loss: 0.667683 (accumulated 1 forwards) [2026-07-10 02:42:11-INFO]: epoch 11-step 2863480 loss: 0.656426 (accumulated 1 forwards) [2026-07-10 02:42:19-INFO]: epoch 11-step 2863490 loss: 0.665312 (accumulated 1 forwards) [2026-07-10 02:42:28-INFO]: epoch 11-step 2863500 loss: 0.665338 (accumulated 1 forwards) [2026-07-10 02:42:37-INFO]: epoch 11-step 2863510 loss: 0.701403 (accumulated 1 forwards) [2026-07-10 02:42:45-INFO]: epoch 11-step 2863520 loss: 0.694133 (accumulated 1 forwards) [2026-07-10 02:42:54-INFO]: epoch 11-step 2863530 loss: 0.630110 (accumulated 1 forwards) [2026-07-10 02:43:03-INFO]: epoch 11-step 2863540 loss: 0.662511 (accumulated 1 forwards) [2026-07-10 02:43:12-INFO]: epoch 11-step 2863550 loss: 0.629351 (accumulated 1 forwards) [2026-07-10 02:43:20-INFO]: epoch 11-step 2863560 loss: 0.631993 (accumulated 1 forwards) [2026-07-10 02:43:29-INFO]: epoch 11-step 2863570 loss: 0.691398 (accumulated 1 forwards) [2026-07-10 02:43:38-INFO]: epoch 11-step 2863580 loss: 0.700462 (accumulated 1 forwards) [2026-07-10 02:43:46-INFO]: epoch 11-step 2863590 loss: 0.691461 (accumulated 1 forwards) [2026-07-10 02:43:55-INFO]: epoch 11-step 2863600 loss: 0.703651 (accumulated 1 forwards) [2026-07-10 02:44:04-INFO]: epoch 11-step 2863610 loss: 0.670139 (accumulated 1 forwards) [2026-07-10 02:44:12-INFO]: epoch 11-step 2863620 loss: 0.670237 (accumulated 1 forwards) [2026-07-10 02:44:21-INFO]: epoch 11-step 2863630 loss: 0.666297 (accumulated 1 forwards) [2026-07-10 02:44:30-INFO]: epoch 11-step 2863640 loss: 0.653401 (accumulated 1 forwards) [2026-07-10 02:44:39-INFO]: epoch 11-step 2863650 loss: 0.691462 (accumulated 1 forwards) [2026-07-10 02:44:47-INFO]: epoch 11-step 2863660 loss: 0.628078 (accumulated 1 forwards) [2026-07-10 02:44:56-INFO]: epoch 11-step 2863670 loss: 0.655788 (accumulated 1 forwards) [2026-07-10 02:45:05-INFO]: epoch 11-step 2863680 loss: 0.687073 (accumulated 1 forwards) [2026-07-10 02:45:13-INFO]: epoch 11-step 2863690 loss: 0.699025 (accumulated 1 forwards) [2026-07-10 02:45:22-INFO]: epoch 11-step 2863700 loss: 0.664318 (accumulated 1 forwards) [2026-07-10 02:45:31-INFO]: epoch 11-step 2863710 loss: 0.664376 (accumulated 1 forwards) [2026-07-10 02:45:40-INFO]: epoch 11-step 2863720 loss: 0.681274 (accumulated 1 forwards) [2026-07-10 02:45:48-INFO]: epoch 11-step 2863730 loss: 0.666815 (accumulated 1 forwards) [2026-07-10 02:45:57-INFO]: epoch 11-step 2863740 loss: 0.666004 (accumulated 1 forwards) [2026-07-10 02:46:06-INFO]: epoch 11-step 2863750 loss: 0.698411 (accumulated 1 forwards) [2026-07-10 02:46:14-INFO]: epoch 11-step 2863760 loss: 0.654828 (accumulated 1 forwards) [2026-07-10 02:46:23-INFO]: epoch 11-step 2863770 loss: 0.649553 (accumulated 1 forwards) [2026-07-10 02:46:32-INFO]: epoch 11-step 2863780 loss: 0.633027 (accumulated 1 forwards) [2026-07-10 02:46:40-INFO]: epoch 11-step 2863790 loss: 0.632965 (accumulated 1 forwards) [2026-07-10 02:46:49-INFO]: epoch 11-step 2863800 loss: 0.653647 (accumulated 1 forwards) [2026-07-10 02:46:58-INFO]: epoch 11-step 2863810 loss: 0.682731 (accumulated 1 forwards) [2026-07-10 02:47:07-INFO]: epoch 11-step 2863820 loss: 0.671635 (accumulated 1 forwards) [2026-07-10 02:47:15-INFO]: epoch 11-step 2863830 loss: 0.663406 (accumulated 1 forwards) [2026-07-10 02:47:24-INFO]: epoch 11-step 2863840 loss: 0.640255 (accumulated 1 forwards) [2026-07-10 02:47:33-INFO]: epoch 11-step 2863850 loss: 0.692868 (accumulated 1 forwards) [2026-07-10 02:47:41-INFO]: epoch 11-step 2863860 loss: 0.636199 (accumulated 1 forwards) [2026-07-10 02:47:50-INFO]: epoch 11-step 2863870 loss: 0.658660 (accumulated 1 forwards) [2026-07-10 02:47:59-INFO]: epoch 11-step 2863880 loss: 0.676222 (accumulated 1 forwards) [2026-07-10 02:48:07-INFO]: epoch 11-step 2863890 loss: 0.656558 (accumulated 1 forwards) [2026-07-10 02:48:16-INFO]: epoch 11-step 2863900 loss: 0.674427 (accumulated 1 forwards) [2026-07-10 02:48:25-INFO]: epoch 11-step 2863910 loss: 0.628055 (accumulated 1 forwards) [2026-07-10 02:48:33-INFO]: epoch 11-step 2863920 loss: 0.635968 (accumulated 1 forwards) [2026-07-10 02:48:42-INFO]: epoch 11-step 2863930 loss: 0.679475 (accumulated 1 forwards) [2026-07-10 02:48:51-INFO]: epoch 11-step 2863940 loss: 0.681879 (accumulated 1 forwards) [2026-07-10 02:48:59-INFO]: epoch 11-step 2863950 loss: 0.651738 (accumulated 1 forwards) [2026-07-10 02:49:08-INFO]: epoch 11-step 2863960 loss: 0.647615 (accumulated 1 forwards) [2026-07-10 02:49:17-INFO]: epoch 11-step 2863970 loss: 0.645555 (accumulated 1 forwards) [2026-07-10 02:49:25-INFO]: epoch 11-step 2863980 loss: 0.607431 (accumulated 1 forwards) [2026-07-10 02:49:34-INFO]: epoch 11-step 2863990 loss: 0.656841 (accumulated 1 forwards) [2026-07-10 02:49:43-INFO]: epoch 11-step 2864000 loss: 0.665039 (accumulated 1 forwards) [2026-07-10 02:49:52-INFO]: epoch 11-step 2864010 loss: 0.683083 (accumulated 1 forwards) [2026-07-10 02:50:00-INFO]: epoch 11-step 2864020 loss: 0.681549 (accumulated 1 forwards) [2026-07-10 02:50:09-INFO]: epoch 11-step 2864030 loss: 0.617085 (accumulated 1 forwards) [2026-07-10 02:50:18-INFO]: epoch 11-step 2864040 loss: 0.662587 (accumulated 1 forwards) [2026-07-10 02:50:26-INFO]: epoch 11-step 2864050 loss: 0.674467 (accumulated 1 forwards) [2026-07-10 02:50:35-INFO]: epoch 11-step 2864060 loss: 0.675994 (accumulated 1 forwards) [2026-07-10 02:50:44-INFO]: epoch 11-step 2864070 loss: 0.676025 (accumulated 1 forwards) [2026-07-10 02:50:52-INFO]: epoch 11-step 2864080 loss: 0.623767 (accumulated 1 forwards) [2026-07-10 02:51:01-INFO]: epoch 11-step 2864090 loss: 0.614406 (accumulated 1 forwards) [2026-07-10 02:51:10-INFO]: epoch 11-step 2864100 loss: 0.677696 (accumulated 1 forwards) [2026-07-10 02:51:18-INFO]: epoch 11-step 2864110 loss: 0.676209 (accumulated 1 forwards) [2026-07-10 02:51:27-INFO]: epoch 11-step 2864120 loss: 0.640869 (accumulated 1 forwards) [2026-07-10 02:51:36-INFO]: epoch 11-step 2864130 loss: 0.678928 (accumulated 1 forwards) [2026-07-10 02:51:45-INFO]: epoch 11-step 2864140 loss: 0.619151 (accumulated 1 forwards) [2026-07-10 02:51:53-INFO]: epoch 11-step 2864150 loss: 0.661369 (accumulated 1 forwards) [2026-07-10 02:52:02-INFO]: epoch 11-step 2864160 loss: 0.671308 (accumulated 1 forwards) [2026-07-10 02:52:11-INFO]: epoch 11-step 2864170 loss: 0.695812 (accumulated 1 forwards) [2026-07-10 02:52:19-INFO]: epoch 11-step 2864180 loss: 0.658596 (accumulated 1 forwards) [2026-07-10 02:52:28-INFO]: epoch 11-step 2864190 loss: 0.690326 (accumulated 1 forwards) [2026-07-10 02:52:37-INFO]: epoch 11-step 2864200 loss: 0.703268 (accumulated 1 forwards) [2026-07-10 02:52:45-INFO]: epoch 11-step 2864210 loss: 0.640153 (accumulated 1 forwards) [2026-07-10 02:52:54-INFO]: epoch 11-step 2864220 loss: 0.655701 (accumulated 1 forwards) [2026-07-10 02:53:03-INFO]: epoch 11-step 2864230 loss: 0.640234 (accumulated 1 forwards) [2026-07-10 02:53:12-INFO]: epoch 11-step 2864240 loss: 0.731131 (accumulated 1 forwards) [2026-07-10 02:53:20-INFO]: epoch 11-step 2864250 loss: 0.687042 (accumulated 1 forwards) [2026-07-10 02:53:29-INFO]: epoch 11-step 2864260 loss: 0.635364 (accumulated 1 forwards) [2026-07-10 02:53:38-INFO]: epoch 11-step 2864270 loss: 0.667677 (accumulated 1 forwards) [2026-07-10 02:53:46-INFO]: epoch 11-step 2864280 loss: 0.688691 (accumulated 1 forwards) [2026-07-10 02:53:55-INFO]: epoch 11-step 2864290 loss: 0.681595 (accumulated 1 forwards) [2026-07-10 02:54:04-INFO]: epoch 11-step 2864300 loss: 0.670482 (accumulated 1 forwards) [2026-07-10 02:54:12-INFO]: epoch 11-step 2864310 loss: 0.650563 (accumulated 1 forwards) [2026-07-10 02:54:21-INFO]: epoch 11-step 2864320 loss: 0.647574 (accumulated 1 forwards) [2026-07-10 02:54:30-INFO]: epoch 11-step 2864330 loss: 0.707361 (accumulated 1 forwards) [2026-07-10 02:54:38-INFO]: epoch 11-step 2864340 loss: 0.623035 (accumulated 1 forwards) [2026-07-10 02:54:47-INFO]: epoch 11-step 2864350 loss: 0.686352 (accumulated 1 forwards) [2026-07-10 02:54:56-INFO]: epoch 11-step 2864360 loss: 0.674692 (accumulated 1 forwards) [2026-07-10 02:55:05-INFO]: epoch 11-step 2864370 loss: 0.653839 (accumulated 1 forwards) [2026-07-10 02:55:13-INFO]: epoch 11-step 2864380 loss: 0.690227 (accumulated 1 forwards) [2026-07-10 02:55:22-INFO]: epoch 11-step 2864390 loss: 0.711078 (accumulated 1 forwards) [2026-07-10 02:55:31-INFO]: epoch 11-step 2864400 loss: 0.700028 (accumulated 1 forwards) [2026-07-10 02:55:39-INFO]: epoch 11-step 2864410 loss: 0.688132 (accumulated 1 forwards) [2026-07-10 02:55:48-INFO]: epoch 11-step 2864420 loss: 0.820961 (accumulated 1 forwards) [2026-07-10 02:55:57-INFO]: epoch 11-step 2864430 loss: 0.803117 (accumulated 1 forwards) [2026-07-10 02:56:05-INFO]: epoch 11-step 2864440 loss: 0.938758 (accumulated 1 forwards) [2026-07-10 02:56:14-INFO]: epoch 11-step 2864450 loss: 1.003226 (accumulated 1 forwards) [2026-07-10 02:56:23-INFO]: epoch 11-step 2864460 loss: 1.001597 (accumulated 1 forwards) [2026-07-10 02:56:32-INFO]: epoch 11-step 2864470 loss: 1.430355 (accumulated 1 forwards) [2026-07-10 02:56:40-INFO]: epoch 11-step 2864480 loss: 1.598001 (accumulated 1 forwards) [2026-07-10 02:56:49-INFO]: epoch 11-step 2864490 loss: 2.502838 (accumulated 1 forwards) [2026-07-10 02:56:58-INFO]: epoch 11-step 2864500 loss: 1.704693 (accumulated 1 forwards) [2026-07-10 02:57:07-INFO]: epoch 11-step 2864510 loss: 1.424392 (accumulated 1 forwards) [2026-07-10 02:57:16-INFO]: epoch 11-step 2864520 loss: 1.327987 (accumulated 1 forwards) [2026-07-10 02:57:25-INFO]: epoch 11-step 2864530 loss: 1.102051 (accumulated 1 forwards) [2026-07-10 02:57:34-INFO]: epoch 11-step 2864540 loss: 1.102103 (accumulated 1 forwards) [2026-07-10 02:57:37-INFO]: Reload dataloader [2026-07-10 02:57:43-INFO]: epoch 12-step 2864550 loss: 1.013341 (accumulated 1 forwards) [2026-07-10 02:57:51-INFO]: epoch 12-step 2864560 loss: 0.972962 (accumulated 1 forwards) [2026-07-10 02:58:00-INFO]: epoch 12-step 2864570 loss: 0.947449 (accumulated 1 forwards) [2026-07-10 02:58:09-INFO]: epoch 12-step 2864580 loss: 0.950985 (accumulated 1 forwards) [2026-07-10 02:58:18-INFO]: epoch 12-step 2864590 loss: 0.915655 (accumulated 1 forwards) [2026-07-10 02:58:27-INFO]: epoch 12-step 2864600 loss: 0.945068 (accumulated 1 forwards) [2026-07-10 02:58:35-INFO]: epoch 12-step 2864610 loss: 0.907459 (accumulated 1 forwards) [2026-07-10 02:58:44-INFO]: epoch 12-step 2864620 loss: 0.908914 (accumulated 1 forwards) [2026-07-10 02:58:53-INFO]: epoch 12-step 2864630 loss: 0.835554 (accumulated 1 forwards) [2026-07-10 02:59:02-INFO]: epoch 12-step 2864640 loss: 0.898242 (accumulated 1 forwards) [2026-07-10 02:59:10-INFO]: epoch 12-step 2864650 loss: 0.951355 (accumulated 1 forwards) [2026-07-10 02:59:19-INFO]: epoch 12-step 2864660 loss: 0.918533 (accumulated 1 forwards) [2026-07-10 02:59:28-INFO]: epoch 12-step 2864670 loss: 0.867907 (accumulated 1 forwards) [2026-07-10 02:59:37-INFO]: epoch 12-step 2864680 loss: 0.955213 (accumulated 1 forwards) [2026-07-10 02:59:45-INFO]: epoch 12-step 2864690 loss: 0.893151 (accumulated 1 forwards) [2026-07-10 02:59:54-INFO]: epoch 12-step 2864700 loss: 1.144104 (accumulated 1 forwards) [2026-07-10 03:00:03-INFO]: epoch 12-step 2864710 loss: 1.431891 (accumulated 1 forwards) [2026-07-10 03:00:12-INFO]: epoch 12-step 2864720 loss: 1.674781 (accumulated 1 forwards) [2026-07-10 03:00:21-INFO]: epoch 12-step 2864730 loss: 1.542861 (accumulated 1 forwards) [2026-07-10 03:00:29-INFO]: epoch 12-step 2864740 loss: 1.513204 (accumulated 1 forwards) [2026-07-10 03:00:38-INFO]: epoch 12-step 2864750 loss: 1.394553 (accumulated 1 forwards) [2026-07-10 03:00:47-INFO]: epoch 12-step 2864760 loss: 1.309016 (accumulated 1 forwards) [2026-07-10 03:00:56-INFO]: epoch 12-step 2864770 loss: 1.345303 (accumulated 1 forwards) [2026-07-10 03:01:05-INFO]: epoch 12-step 2864780 loss: 1.097188 (accumulated 1 forwards) [2026-07-10 03:01:14-INFO]: epoch 12-step 2864790 loss: 1.066430 (accumulated 1 forwards) [2026-07-10 03:01:22-INFO]: epoch 12-step 2864800 loss: 0.992095 (accumulated 1 forwards) [2026-07-10 03:01:31-INFO]: epoch 12-step 2864810 loss: 0.942084 (accumulated 1 forwards) [2026-07-10 03:01:40-INFO]: epoch 12-step 2864820 loss: 0.958369 (accumulated 1 forwards) [2026-07-10 03:01:49-INFO]: epoch 12-step 2864830 loss: 0.923348 (accumulated 1 forwards) [2026-07-10 03:01:58-INFO]: epoch 12-step 2864840 loss: 0.934582 (accumulated 1 forwards) [2026-07-10 03:02:06-INFO]: epoch 12-step 2864850 loss: 0.914531 (accumulated 1 forwards) [2026-07-10 03:02:15-INFO]: epoch 12-step 2864860 loss: 0.909998 (accumulated 1 forwards) [2026-07-10 03:02:24-INFO]: epoch 12-step 2864870 loss: 0.930190 (accumulated 1 forwards) [2026-07-10 03:02:33-INFO]: epoch 12-step 2864880 loss: 0.863467 (accumulated 1 forwards) [2026-07-10 03:02:41-INFO]: epoch 12-step 2864890 loss: 0.951682 (accumulated 1 forwards) [2026-07-10 03:02:50-INFO]: epoch 12-step 2864900 loss: 0.895556 (accumulated 1 forwards) [2026-07-10 03:02:59-INFO]: epoch 12-step 2864910 loss: 0.891401 (accumulated 1 forwards) [2026-07-10 03:03:08-INFO]: epoch 12-step 2864920 loss: 0.906527 (accumulated 1 forwards) [2026-07-10 03:03:17-INFO]: epoch 12-step 2864930 loss: 0.917374 (accumulated 1 forwards) [2026-07-10 03:03:25-INFO]: epoch 12-step 2864940 loss: 0.903482 (accumulated 1 forwards) [2026-07-10 03:03:34-INFO]: epoch 12-step 2864950 loss: 0.884917 (accumulated 1 forwards) [2026-07-10 03:03:43-INFO]: epoch 12-step 2864960 loss: 0.893116 (accumulated 1 forwards) [2026-07-10 03:03:52-INFO]: epoch 12-step 2864970 loss: 0.921517 (accumulated 1 forwards) [2026-07-10 03:04:00-INFO]: epoch 12-step 2864980 loss: 0.927775 (accumulated 1 forwards) [2026-07-10 03:04:09-INFO]: epoch 12-step 2864990 loss: 0.876930 (accumulated 1 forwards) [2026-07-10 03:04:18-INFO]: epoch 12-step 2865000 loss: 0.890910 (accumulated 1 forwards) [2026-07-10 03:04:27-INFO]: epoch 12-step 2865010 loss: 0.959157 (accumulated 1 forwards) [2026-07-10 03:04:36-INFO]: epoch 12-step 2865020 loss: 0.921471 (accumulated 1 forwards) [2026-07-10 03:04:44-INFO]: epoch 12-step 2865030 loss: 0.925783 (accumulated 1 forwards) [2026-07-10 03:04:53-INFO]: epoch 12-step 2865040 loss: 0.895649 (accumulated 1 forwards) [2026-07-10 03:05:02-INFO]: epoch 12-step 2865050 loss: 0.922536 (accumulated 1 forwards) [2026-07-10 03:05:11-INFO]: epoch 12-step 2865060 loss: 0.936603 (accumulated 1 forwards) [2026-07-10 03:05:20-INFO]: epoch 12-step 2865070 loss: 0.961660 (accumulated 1 forwards) [2026-07-10 03:05:29-INFO]: epoch 12-step 2865080 loss: 0.902516 (accumulated 1 forwards) [2026-07-10 03:05:37-INFO]: epoch 12-step 2865090 loss: 0.944223 (accumulated 1 forwards) [2026-07-10 03:05:46-INFO]: epoch 12-step 2865100 loss: 0.953050 (accumulated 1 forwards) [2026-07-10 03:05:55-INFO]: epoch 12-step 2865110 loss: 0.950881 (accumulated 1 forwards) [2026-07-10 03:06:04-INFO]: epoch 12-step 2865120 loss: 1.029724 (accumulated 1 forwards) [2026-07-10 03:06:13-INFO]: epoch 12-step 2865130 loss: 0.957142 (accumulated 1 forwards) [2026-07-10 03:06:21-INFO]: epoch 12-step 2865140 loss: 0.928398 (accumulated 1 forwards) [2026-07-10 03:06:30-INFO]: epoch 12-step 2865150 loss: 0.980976 (accumulated 1 forwards) [2026-07-10 03:06:39-INFO]: epoch 12-step 2865160 loss: 0.961861 (accumulated 1 forwards) [2026-07-10 03:06:48-INFO]: epoch 12-step 2865170 loss: 1.015656 (accumulated 1 forwards) [2026-07-10 03:06:57-INFO]: epoch 12-step 2865180 loss: 0.984281 (accumulated 1 forwards) [2026-07-10 03:07:05-INFO]: epoch 12-step 2865190 loss: 1.027832 (accumulated 1 forwards) [2026-07-10 03:07:14-INFO]: epoch 12-step 2865200 loss: 1.008779 (accumulated 1 forwards) [2026-07-10 03:07:23-INFO]: epoch 12-step 2865210 loss: 0.977321 (accumulated 1 forwards) [2026-07-10 03:07:32-INFO]: epoch 12-step 2865220 loss: 0.979952 (accumulated 1 forwards) [2026-07-10 03:07:41-INFO]: epoch 12-step 2865230 loss: 0.955045 (accumulated 1 forwards) [2026-07-10 03:07:49-INFO]: epoch 12-step 2865240 loss: 0.955612 (accumulated 1 forwards) [2026-07-10 03:07:58-INFO]: epoch 12-step 2865250 loss: 0.927276 (accumulated 1 forwards) [2026-07-10 03:08:07-INFO]: epoch 12-step 2865260 loss: 0.976914 (accumulated 1 forwards) [2026-07-10 03:08:16-INFO]: epoch 12-step 2865270 loss: 0.986858 (accumulated 1 forwards) [2026-07-10 03:08:24-INFO]: epoch 12-step 2865280 loss: 0.960191 (accumulated 1 forwards) [2026-07-10 03:08:33-INFO]: epoch 12-step 2865290 loss: 0.938440 (accumulated 1 forwards) [2026-07-10 03:08:42-INFO]: epoch 12-step 2865300 loss: 0.916094 (accumulated 1 forwards) [2026-07-10 03:08:51-INFO]: epoch 12-step 2865310 loss: 0.956670 (accumulated 1 forwards) [2026-07-10 03:08:59-INFO]: epoch 12-step 2865320 loss: 0.942618 (accumulated 1 forwards) [2026-07-10 03:09:08-INFO]: epoch 12-step 2865330 loss: 0.930578 (accumulated 1 forwards) [2026-07-10 03:09:17-INFO]: epoch 12-step 2865340 loss: 0.954548 (accumulated 1 forwards) [2026-07-10 03:09:26-INFO]: epoch 12-step 2865350 loss: 0.969653 (accumulated 1 forwards) [2026-07-10 03:09:35-INFO]: epoch 12-step 2865360 loss: 0.913938 (accumulated 1 forwards) [2026-07-10 03:09:44-INFO]: epoch 12-step 2865370 loss: 0.928956 (accumulated 1 forwards) [2026-07-10 03:09:52-INFO]: epoch 12-step 2865380 loss: 0.904466 (accumulated 1 forwards) [2026-07-10 03:10:01-INFO]: epoch 12-step 2865390 loss: 0.951958 (accumulated 1 forwards) [2026-07-10 03:10:10-INFO]: epoch 12-step 2865400 loss: 1.004872 (accumulated 1 forwards) [2026-07-10 03:10:19-INFO]: epoch 12-step 2865410 loss: 0.939698 (accumulated 1 forwards) [2026-07-10 03:10:28-INFO]: epoch 12-step 2865420 loss: 0.955355 (accumulated 1 forwards) [2026-07-10 03:10:37-INFO]: epoch 12-step 2865430 loss: 0.941515 (accumulated 1 forwards) [2026-07-10 03:10:45-INFO]: epoch 12-step 2865440 loss: 0.983770 (accumulated 1 forwards) [2026-07-10 03:10:54-INFO]: epoch 12-step 2865450 loss: 0.929043 (accumulated 1 forwards) [2026-07-10 03:11:03-INFO]: epoch 12-step 2865460 loss: 0.969281 (accumulated 1 forwards) [2026-07-10 03:11:12-INFO]: epoch 12-step 2865470 loss: 0.919887 (accumulated 1 forwards) [2026-07-10 03:11:21-INFO]: epoch 12-step 2865480 loss: 0.976130 (accumulated 1 forwards) [2026-07-10 03:11:30-INFO]: epoch 12-step 2865490 loss: 0.948917 (accumulated 1 forwards) [2026-07-10 03:11:39-INFO]: epoch 12-step 2865500 loss: 0.983446 (accumulated 1 forwards) [2026-07-10 03:11:47-INFO]: epoch 12-step 2865510 loss: 0.927465 (accumulated 1 forwards) [2026-07-10 03:11:56-INFO]: epoch 12-step 2865520 loss: 0.943389 (accumulated 1 forwards) [2026-07-10 03:12:05-INFO]: epoch 12-step 2865530 loss: 0.956236 (accumulated 1 forwards) [2026-07-10 03:12:14-INFO]: epoch 12-step 2865540 loss: 0.967962 (accumulated 1 forwards) [2026-07-10 03:12:23-INFO]: epoch 12-step 2865550 loss: 0.886474 (accumulated 1 forwards) [2026-07-10 03:12:32-INFO]: epoch 12-step 2865560 loss: 0.975583 (accumulated 1 forwards) [2026-07-10 03:12:40-INFO]: epoch 12-step 2865570 loss: 0.946869 (accumulated 1 forwards) [2026-07-10 03:12:49-INFO]: epoch 12-step 2865580 loss: 0.961613 (accumulated 1 forwards) [2026-07-10 03:12:58-INFO]: epoch 12-step 2865590 loss: 0.906091 (accumulated 1 forwards) [2026-07-10 03:13:07-INFO]: epoch 12-step 2865600 loss: 0.954440 (accumulated 1 forwards) [2026-07-10 03:13:16-INFO]: epoch 12-step 2865610 loss: 0.934970 (accumulated 1 forwards) [2026-07-10 03:13:25-INFO]: epoch 12-step 2865620 loss: 0.931931 (accumulated 1 forwards) [2026-07-10 03:13:34-INFO]: epoch 12-step 2865630 loss: 1.001345 (accumulated 1 forwards) [2026-07-10 03:13:43-INFO]: epoch 12-step 2865640 loss: 0.935827 (accumulated 1 forwards) [2026-07-10 03:13:51-INFO]: epoch 12-step 2865650 loss: 0.956948 (accumulated 1 forwards) [2026-07-10 03:14:00-INFO]: epoch 12-step 2865660 loss: 0.971952 (accumulated 1 forwards) [2026-07-10 03:14:09-INFO]: epoch 12-step 2865670 loss: 1.020678 (accumulated 1 forwards) [2026-07-10 03:14:18-INFO]: epoch 12-step 2865680 loss: 1.034006 (accumulated 1 forwards) [2026-07-10 03:14:27-INFO]: epoch 12-step 2865690 loss: 0.997224 (accumulated 1 forwards) [2026-07-10 03:14:36-INFO]: epoch 12-step 2865700 loss: 0.997112 (accumulated 1 forwards) [2026-07-10 03:14:44-INFO]: epoch 12-step 2865710 loss: 0.958343 (accumulated 1 forwards) [2026-07-10 03:14:53-INFO]: epoch 12-step 2865720 loss: 0.954473 (accumulated 1 forwards) [2026-07-10 03:15:02-INFO]: epoch 12-step 2865730 loss: 1.002054 (accumulated 1 forwards) [2026-07-10 03:15:11-INFO]: epoch 12-step 2865740 loss: 0.987702 (accumulated 1 forwards) [2026-07-10 03:15:20-INFO]: epoch 12-step 2865750 loss: 0.947142 (accumulated 1 forwards) [2026-07-10 03:15:25-INFO]: Reload dataloader [2026-07-10 03:15:28-INFO]: epoch 13-step 2865760 loss: 0.952339 (accumulated 1 forwards) [2026-07-10 03:15:37-INFO]: epoch 13-step 2865770 loss: 1.011642 (accumulated 1 forwards) [2026-07-10 03:15:46-INFO]: epoch 13-step 2865780 loss: 0.980362 (accumulated 1 forwards) [2026-07-10 03:15:55-INFO]: epoch 13-step 2865790 loss: 1.035449 (accumulated 1 forwards) [2026-07-10 03:16:04-INFO]: epoch 13-step 2865800 loss: 1.011222 (accumulated 1 forwards) [2026-07-10 03:16:13-INFO]: epoch 13-step 2865810 loss: 1.036424 (accumulated 1 forwards) [2026-07-10 03:16:21-INFO]: epoch 13-step 2865820 loss: 1.025873 (accumulated 1 forwards) [2026-07-10 03:16:30-INFO]: epoch 13-step 2865830 loss: 1.038956 (accumulated 1 forwards) [2026-07-10 03:16:39-INFO]: epoch 13-step 2865840 loss: 1.010174 (accumulated 1 forwards) [2026-07-10 03:16:48-INFO]: epoch 13-step 2865850 loss: 0.963498 (accumulated 1 forwards) [2026-07-10 03:16:57-INFO]: epoch 13-step 2865860 loss: 0.985170 (accumulated 1 forwards) [2026-07-10 03:17:06-INFO]: epoch 13-step 2865870 loss: 0.949895 (accumulated 1 forwards) [2026-07-10 03:17:14-INFO]: epoch 13-step 2865880 loss: 1.008086 (accumulated 1 forwards) [2026-07-10 03:17:23-INFO]: epoch 13-step 2865890 loss: 0.960713 (accumulated 1 forwards) [2026-07-10 03:17:32-INFO]: epoch 13-step 2865900 loss: 0.968074 (accumulated 1 forwards) [2026-07-10 03:17:41-INFO]: epoch 13-step 2865910 loss: 0.988524 (accumulated 1 forwards) [2026-07-10 03:17:50-INFO]: epoch 13-step 2865920 loss: 0.952197 (accumulated 1 forwards) [2026-07-10 03:17:59-INFO]: epoch 13-step 2865930 loss: 0.947625 (accumulated 1 forwards) [2026-07-10 03:18:07-INFO]: epoch 13-step 2865940 loss: 0.919820 (accumulated 1 forwards) [2026-07-10 03:18:16-INFO]: epoch 13-step 2865950 loss: 0.964476 (accumulated 1 forwards) [2026-07-10 03:18:25-INFO]: epoch 13-step 2865960 loss: 0.951130 (accumulated 1 forwards) [2026-07-10 03:18:34-INFO]: epoch 13-step 2865970 loss: 0.934255 (accumulated 1 forwards) [2026-07-10 03:18:43-INFO]: epoch 13-step 2865980 loss: 0.925009 (accumulated 1 forwards) [2026-07-10 03:18:51-INFO]: epoch 13-step 2865990 loss: 0.975870 (accumulated 1 forwards) [2026-07-10 03:19:00-INFO]: epoch 13-step 2866000 loss: 0.959966 (accumulated 1 forwards) [2026-07-10 03:19:09-INFO]: epoch 13-step 2866010 loss: 0.921770 (accumulated 1 forwards) [2026-07-10 03:19:18-INFO]: epoch 13-step 2866020 loss: 0.911374 (accumulated 1 forwards) [2026-07-10 03:19:26-INFO]: epoch 13-step 2866030 loss: 0.983636 (accumulated 1 forwards) [2026-07-10 03:19:35-INFO]: epoch 13-step 2866040 loss: 0.930418 (accumulated 1 forwards) [2026-07-10 03:19:44-INFO]: epoch 13-step 2866050 loss: 1.004404 (accumulated 1 forwards) [2026-07-10 03:19:53-INFO]: epoch 13-step 2866060 loss: 0.923037 (accumulated 1 forwards) [2026-07-10 03:20:02-INFO]: epoch 13-step 2866070 loss: 0.942131 (accumulated 1 forwards) [2026-07-10 03:20:10-INFO]: epoch 13-step 2866080 loss: 0.925263 (accumulated 1 forwards) [2026-07-10 03:20:19-INFO]: epoch 13-step 2866090 loss: 0.950697 (accumulated 1 forwards) [2026-07-10 03:20:28-INFO]: epoch 13-step 2866100 loss: 0.980778 (accumulated 1 forwards) [2026-07-10 03:20:37-INFO]: epoch 13-step 2866110 loss: 0.954342 (accumulated 1 forwards) [2026-07-10 03:20:45-INFO]: epoch 13-step 2866120 loss: 0.936609 (accumulated 1 forwards) [2026-07-10 03:20:54-INFO]: epoch 13-step 2866130 loss: 0.946859 (accumulated 1 forwards) [2026-07-10 03:21:03-INFO]: epoch 13-step 2866140 loss: 0.977240 (accumulated 1 forwards) [2026-07-10 03:21:12-INFO]: epoch 13-step 2866150 loss: 1.140457 (accumulated 1 forwards) [2026-07-10 03:21:21-INFO]: epoch 13-step 2866160 loss: 1.124674 (accumulated 1 forwards) [2026-07-10 03:21:29-INFO]: epoch 13-step 2866170 loss: 1.004546 (accumulated 1 forwards) [2026-07-10 03:21:38-INFO]: epoch 13-step 2866180 loss: 1.045526 (accumulated 1 forwards) [2026-07-10 03:21:47-INFO]: epoch 13-step 2866190 loss: 0.959539 (accumulated 1 forwards) [2026-07-10 03:21:56-INFO]: epoch 13-step 2866200 loss: 0.972806 (accumulated 1 forwards) [2026-07-10 03:22:05-INFO]: epoch 13-step 2866210 loss: 0.946429 (accumulated 1 forwards) [2026-07-10 03:22:14-INFO]: epoch 13-step 2866220 loss: 1.000758 (accumulated 1 forwards) [2026-07-10 03:22:22-INFO]: epoch 13-step 2866230 loss: 0.985828 (accumulated 1 forwards) [2026-07-10 03:22:31-INFO]: epoch 13-step 2866240 loss: 0.921044 (accumulated 1 forwards) [2026-07-10 03:22:40-INFO]: epoch 13-step 2866250 loss: 0.966502 (accumulated 1 forwards) [2026-07-10 03:22:49-INFO]: epoch 13-step 2866260 loss: 1.001352 (accumulated 1 forwards) [2026-07-10 03:22:58-INFO]: epoch 13-step 2866270 loss: 0.960270 (accumulated 1 forwards) [2026-07-10 03:23:07-INFO]: epoch 13-step 2866280 loss: 0.973079 (accumulated 1 forwards) [2026-07-10 03:23:15-INFO]: epoch 13-step 2866290 loss: 0.979018 (accumulated 1 forwards) [2026-07-10 03:23:24-INFO]: epoch 13-step 2866300 loss: 0.963383 (accumulated 1 forwards) [2026-07-10 03:23:33-INFO]: epoch 13-step 2866310 loss: 1.016267 (accumulated 1 forwards) [2026-07-10 03:23:42-INFO]: epoch 13-step 2866320 loss: 0.890792 (accumulated 1 forwards) [2026-07-10 03:23:51-INFO]: epoch 13-step 2866330 loss: 0.931171 (accumulated 1 forwards) [2026-07-10 03:23:59-INFO]: epoch 13-step 2866340 loss: 0.979465 (accumulated 1 forwards) [2026-07-10 03:24:08-INFO]: epoch 13-step 2866350 loss: 0.936363 (accumulated 1 forwards) [2026-07-10 03:24:17-INFO]: epoch 13-step 2866360 loss: 1.096466 (accumulated 1 forwards) [2026-07-10 03:24:26-INFO]: epoch 13-step 2866370 loss: 1.020811 (accumulated 1 forwards) [2026-07-10 03:24:35-INFO]: epoch 13-step 2866380 loss: 0.990120 (accumulated 1 forwards) [2026-07-10 03:24:43-INFO]: epoch 13-step 2866390 loss: 1.033625 (accumulated 1 forwards) [2026-07-10 03:24:52-INFO]: epoch 13-step 2866400 loss: 0.958678 (accumulated 1 forwards) [2026-07-10 03:25:01-INFO]: epoch 13-step 2866410 loss: 1.000609 (accumulated 1 forwards) [2026-07-10 03:25:10-INFO]: epoch 13-step 2866420 loss: 0.969263 (accumulated 1 forwards) [2026-07-10 03:25:19-INFO]: epoch 13-step 2866430 loss: 0.990115 (accumulated 1 forwards) [2026-07-10 03:25:27-INFO]: epoch 13-step 2866440 loss: 1.046910 (accumulated 1 forwards) [2026-07-10 03:25:36-INFO]: epoch 13-step 2866450 loss: 1.010770 (accumulated 1 forwards) [2026-07-10 03:25:45-INFO]: epoch 13-step 2866460 loss: 1.029960 (accumulated 1 forwards) [2026-07-10 03:25:54-INFO]: epoch 13-step 2866470 loss: 1.101449 (accumulated 1 forwards) [2026-07-10 03:26:02-INFO]: epoch 13-step 2866480 loss: 1.083097 (accumulated 1 forwards) [2026-07-10 03:26:11-INFO]: epoch 13-step 2866490 loss: 1.040974 (accumulated 1 forwards) [2026-07-10 03:26:20-INFO]: epoch 13-step 2866500 loss: 1.083853 (accumulated 1 forwards) [2026-07-10 03:26:29-INFO]: epoch 13-step 2866510 loss: 1.001706 (accumulated 1 forwards) [2026-07-10 03:26:38-INFO]: epoch 13-step 2866520 loss: 1.036452 (accumulated 1 forwards) [2026-07-10 03:26:47-INFO]: epoch 13-step 2866530 loss: 1.057894 (accumulated 1 forwards) [2026-07-10 03:26:55-INFO]: epoch 13-step 2866540 loss: 1.044173 (accumulated 1 forwards) [2026-07-10 03:27:04-INFO]: epoch 13-step 2866550 loss: 1.000996 (accumulated 1 forwards) [2026-07-10 03:27:13-INFO]: epoch 13-step 2866560 loss: 1.048201 (accumulated 1 forwards) [2026-07-10 03:27:22-INFO]: epoch 13-step 2866570 loss: 1.036114 (accumulated 1 forwards) [2026-07-10 03:27:31-INFO]: epoch 13-step 2866580 loss: 1.024121 (accumulated 1 forwards) [2026-07-10 03:27:40-INFO]: epoch 13-step 2866590 loss: 1.031604 (accumulated 1 forwards) [2026-07-10 03:27:48-INFO]: epoch 13-step 2866600 loss: 0.999865 (accumulated 1 forwards) [2026-07-10 03:27:57-INFO]: epoch 13-step 2866610 loss: 1.015218 (accumulated 1 forwards) [2026-07-10 03:28:06-INFO]: epoch 13-step 2866620 loss: 1.009570 (accumulated 1 forwards) [2026-07-10 03:28:15-INFO]: epoch 13-step 2866630 loss: 1.062805 (accumulated 1 forwards) [2026-07-10 03:28:24-INFO]: epoch 13-step 2866640 loss: 1.071350 (accumulated 1 forwards) [2026-07-10 03:28:33-INFO]: epoch 13-step 2866650 loss: 1.107397 (accumulated 1 forwards) [2026-07-10 03:28:42-INFO]: epoch 13-step 2866660 loss: 1.035894 (accumulated 1 forwards) [2026-07-10 03:28:50-INFO]: epoch 13-step 2866670 loss: 1.032077 (accumulated 1 forwards) [2026-07-10 03:28:59-INFO]: epoch 13-step 2866680 loss: 1.083001 (accumulated 1 forwards) [2026-07-10 03:29:08-INFO]: epoch 13-step 2866690 loss: 1.020142 (accumulated 1 forwards) [2026-07-10 03:29:17-INFO]: epoch 13-step 2866700 loss: 1.050653 (accumulated 1 forwards) [2026-07-10 03:29:26-INFO]: epoch 13-step 2866710 loss: 1.041644 (accumulated 1 forwards) [2026-07-10 03:29:35-INFO]: epoch 13-step 2866720 loss: 1.051301 (accumulated 1 forwards) [2026-07-10 03:29:43-INFO]: epoch 13-step 2866730 loss: 0.999843 (accumulated 1 forwards) [2026-07-10 03:29:52-INFO]: epoch 13-step 2866740 loss: 1.089908 (accumulated 1 forwards) [2026-07-10 03:30:01-INFO]: epoch 13-step 2866750 loss: 1.106449 (accumulated 1 forwards) [2026-07-10 03:30:10-INFO]: epoch 13-step 2866760 loss: 1.118018 (accumulated 1 forwards) [2026-07-10 03:30:19-INFO]: epoch 13-step 2866770 loss: 1.094472 (accumulated 1 forwards) [2026-07-10 03:30:28-INFO]: epoch 13-step 2866780 loss: 1.068860 (accumulated 1 forwards) [2026-07-10 03:30:37-INFO]: epoch 13-step 2866790 loss: 1.049212 (accumulated 1 forwards) [2026-07-10 03:30:45-INFO]: epoch 13-step 2866800 loss: 1.011310 (accumulated 1 forwards) [2026-07-10 03:30:54-INFO]: epoch 13-step 2866810 loss: 1.047766 (accumulated 1 forwards) [2026-07-10 03:31:03-INFO]: epoch 13-step 2866820 loss: 1.016813 (accumulated 1 forwards) [2026-07-10 03:31:12-INFO]: epoch 13-step 2866830 loss: 1.021302 (accumulated 1 forwards) [2026-07-10 03:31:21-INFO]: epoch 13-step 2866840 loss: 1.094163 (accumulated 1 forwards) [2026-07-10 03:31:30-INFO]: epoch 13-step 2866850 loss: 1.153972 (accumulated 1 forwards) [2026-07-10 03:31:39-INFO]: epoch 13-step 2866860 loss: 1.091812 (accumulated 1 forwards) [2026-07-10 03:31:48-INFO]: epoch 13-step 2866870 loss: 1.032770 (accumulated 1 forwards) [2026-07-10 03:31:56-INFO]: epoch 13-step 2866880 loss: 1.062868 (accumulated 1 forwards) [2026-07-10 03:32:05-INFO]: epoch 13-step 2866890 loss: 1.040028 (accumulated 1 forwards) [2026-07-10 03:32:14-INFO]: epoch 13-step 2866900 loss: 1.039536 (accumulated 1 forwards) [2026-07-10 03:32:23-INFO]: epoch 13-step 2866910 loss: 1.035966 (accumulated 1 forwards) [2026-07-10 03:32:32-INFO]: epoch 13-step 2866920 loss: 1.031637 (accumulated 1 forwards) [2026-07-10 03:32:41-INFO]: epoch 13-step 2866930 loss: 1.014475 (accumulated 1 forwards) [2026-07-10 03:32:50-INFO]: epoch 13-step 2866940 loss: 1.064152 (accumulated 1 forwards) [2026-07-10 03:32:59-INFO]: epoch 13-step 2866950 loss: 0.998678 (accumulated 1 forwards) [2026-07-10 03:33:07-INFO]: epoch 13-step 2866960 loss: 1.048195 (accumulated 1 forwards) [2026-07-10 03:33:14-INFO]: Reload dataloader [2026-07-10 03:33:16-INFO]: epoch 14-step 2866970 loss: 1.008601 (accumulated 1 forwards) [2026-07-10 03:33:25-INFO]: epoch 14-step 2866980 loss: 1.021483 (accumulated 1 forwards) [2026-07-10 03:33:34-INFO]: epoch 14-step 2866990 loss: 1.076815 (accumulated 1 forwards) [2026-07-10 03:33:43-INFO]: epoch 14-step 2867000 loss: 1.027893 (accumulated 1 forwards) [2026-07-10 03:33:51-INFO]: epoch 14-step 2867010 loss: 1.041954 (accumulated 1 forwards) [2026-07-10 03:34:00-INFO]: epoch 14-step 2867020 loss: 1.031747 (accumulated 1 forwards) [2026-07-10 03:34:09-INFO]: epoch 14-step 2867030 loss: 1.043853 (accumulated 1 forwards) [2026-07-10 03:34:18-INFO]: epoch 14-step 2867040 loss: 1.100734 (accumulated 1 forwards) [2026-07-10 03:34:27-INFO]: epoch 14-step 2867050 loss: 0.998882 (accumulated 1 forwards) [2026-07-10 03:34:36-INFO]: epoch 14-step 2867060 loss: 1.081841 (accumulated 1 forwards) [2026-07-10 03:34:44-INFO]: epoch 14-step 2867070 loss: 1.034599 (accumulated 1 forwards) [2026-07-10 03:34:53-INFO]: epoch 14-step 2867080 loss: 1.064412 (accumulated 1 forwards) [2026-07-10 03:35:02-INFO]: epoch 14-step 2867090 loss: 0.979172 (accumulated 1 forwards) [2026-07-10 03:35:11-INFO]: epoch 14-step 2867100 loss: 1.021803 (accumulated 1 forwards) [2026-07-10 03:35:19-INFO]: epoch 14-step 2867110 loss: 0.994795 (accumulated 1 forwards) [2026-07-10 03:35:28-INFO]: epoch 14-step 2867120 loss: 1.068124 (accumulated 1 forwards) [2026-07-10 03:35:37-INFO]: epoch 14-step 2867130 loss: 1.053542 (accumulated 1 forwards) [2026-07-10 03:35:46-INFO]: epoch 14-step 2867140 loss: 0.996568 (accumulated 1 forwards) [2026-07-10 03:35:55-INFO]: epoch 14-step 2867150 loss: 0.954893 (accumulated 1 forwards) [2026-07-10 03:36:03-INFO]: epoch 14-step 2867160 loss: 1.075989 (accumulated 1 forwards) [2026-07-10 03:36:12-INFO]: epoch 14-step 2867170 loss: 1.025952 (accumulated 1 forwards) [2026-07-10 03:36:21-INFO]: epoch 14-step 2867180 loss: 1.128234 (accumulated 1 forwards) [2026-07-10 03:36:30-INFO]: epoch 14-step 2867190 loss: 1.082051 (accumulated 1 forwards) [2026-07-10 03:36:39-INFO]: epoch 14-step 2867200 loss: 1.034937 (accumulated 1 forwards) [2026-07-10 03:36:48-INFO]: epoch 14-step 2867210 loss: 1.028039 (accumulated 1 forwards) [2026-07-10 03:36:56-INFO]: epoch 14-step 2867220 loss: 1.037358 (accumulated 1 forwards) [2026-07-10 03:37:05-INFO]: epoch 14-step 2867230 loss: 1.010737 (accumulated 1 forwards) [2026-07-10 03:37:14-INFO]: epoch 14-step 2867240 loss: 1.019890 (accumulated 1 forwards) [2026-07-10 03:37:23-INFO]: epoch 14-step 2867250 loss: 1.014592 (accumulated 1 forwards) [2026-07-10 03:37:32-INFO]: epoch 14-step 2867260 loss: 0.964891 (accumulated 1 forwards) [2026-07-10 03:37:40-INFO]: epoch 14-step 2867270 loss: 1.021183 (accumulated 1 forwards) [2026-07-10 03:37:49-INFO]: epoch 14-step 2867280 loss: 1.081133 (accumulated 1 forwards) [2026-07-10 03:37:58-INFO]: epoch 14-step 2867290 loss: 0.932163 (accumulated 1 forwards) [2026-07-10 03:38:07-INFO]: epoch 14-step 2867300 loss: 1.060915 (accumulated 1 forwards) [2026-07-10 03:38:16-INFO]: epoch 14-step 2867310 loss: 1.014917 (accumulated 1 forwards) [2026-07-10 03:38:25-INFO]: epoch 14-step 2867320 loss: 0.995692 (accumulated 1 forwards) [2026-07-10 03:38:33-INFO]: epoch 14-step 2867330 loss: 0.986999 (accumulated 1 forwards) [2026-07-10 03:38:42-INFO]: epoch 14-step 2867340 loss: 0.982917 (accumulated 1 forwards) [2026-07-10 03:38:51-INFO]: epoch 14-step 2867350 loss: 0.972698 (accumulated 1 forwards) [2026-07-10 03:39:00-INFO]: epoch 14-step 2867360 loss: 0.999780 (accumulated 1 forwards) [2026-07-10 03:39:09-INFO]: epoch 14-step 2867370 loss: 1.041015 (accumulated 1 forwards) [2026-07-10 03:39:17-INFO]: epoch 14-step 2867380 loss: 1.006393 (accumulated 1 forwards) [2026-07-10 03:39:26-INFO]: epoch 14-step 2867390 loss: 1.041078 (accumulated 1 forwards) [2026-07-10 03:39:35-INFO]: epoch 14-step 2867400 loss: 1.109598 (accumulated 1 forwards) [2026-07-10 03:39:44-INFO]: epoch 14-step 2867410 loss: 1.028440 (accumulated 1 forwards) [2026-07-10 03:39:53-INFO]: epoch 14-step 2867420 loss: 1.022282 (accumulated 1 forwards) [2026-07-10 03:40:02-INFO]: epoch 14-step 2867430 loss: 1.098255 (accumulated 1 forwards) [2026-07-10 03:40:11-INFO]: epoch 14-step 2867440 loss: 1.117181 (accumulated 1 forwards) [2026-07-10 03:40:19-INFO]: epoch 14-step 2867450 loss: 1.132418 (accumulated 1 forwards) [2026-07-10 03:40:28-INFO]: epoch 14-step 2867460 loss: 1.139877 (accumulated 1 forwards) [2026-07-10 03:40:37-INFO]: epoch 14-step 2867470 loss: 1.018098 (accumulated 1 forwards) [2026-07-10 03:40:46-INFO]: epoch 14-step 2867480 loss: 1.079531 (accumulated 1 forwards) [2026-07-10 03:40:55-INFO]: epoch 14-step 2867490 loss: 1.004678 (accumulated 1 forwards) [2026-07-10 03:41:04-INFO]: epoch 14-step 2867500 loss: 1.073306 (accumulated 1 forwards) [2026-07-10 03:41:13-INFO]: epoch 14-step 2867510 loss: 1.080803 (accumulated 1 forwards) [2026-07-10 03:41:22-INFO]: epoch 14-step 2867520 loss: 1.100217 (accumulated 1 forwards) [2026-07-10 03:41:31-INFO]: epoch 14-step 2867530 loss: 1.071235 (accumulated 1 forwards) [2026-07-10 03:41:40-INFO]: epoch 14-step 2867540 loss: 1.067691 (accumulated 1 forwards) [2026-07-10 03:41:48-INFO]: epoch 14-step 2867550 loss: 1.070234 (accumulated 1 forwards) [2026-07-10 03:41:57-INFO]: epoch 14-step 2867560 loss: 1.122271 (accumulated 1 forwards) [2026-07-10 03:42:06-INFO]: epoch 14-step 2867570 loss: 1.088083 (accumulated 1 forwards) [2026-07-10 03:42:15-INFO]: epoch 14-step 2867580 loss: 1.064022 (accumulated 1 forwards) [2026-07-10 03:42:24-INFO]: epoch 14-step 2867590 loss: 1.107424 (accumulated 1 forwards) [2026-07-10 03:42:33-INFO]: epoch 14-step 2867600 loss: 1.085509 (accumulated 1 forwards) [2026-07-10 03:42:42-INFO]: epoch 14-step 2867610 loss: 1.041309 (accumulated 1 forwards) [2026-07-10 03:42:51-INFO]: epoch 14-step 2867620 loss: 1.081152 (accumulated 1 forwards) [2026-07-10 03:43:00-INFO]: epoch 14-step 2867630 loss: 1.275393 (accumulated 1 forwards) [2026-07-10 03:43:09-INFO]: epoch 14-step 2867640 loss: 1.124215 (accumulated 1 forwards) [2026-07-10 03:43:18-INFO]: epoch 14-step 2867650 loss: 1.167610 (accumulated 1 forwards) [2026-07-10 03:43:27-INFO]: epoch 14-step 2867660 loss: 1.163427 (accumulated 1 forwards) [2026-07-10 03:43:35-INFO]: epoch 14-step 2867670 loss: 1.126938 (accumulated 1 forwards) [2026-07-10 03:43:44-INFO]: epoch 14-step 2867680 loss: 1.110536 (accumulated 1 forwards) [2026-07-10 03:43:53-INFO]: epoch 14-step 2867690 loss: 1.184370 (accumulated 1 forwards) [2026-07-10 03:44:02-INFO]: epoch 14-step 2867700 loss: 1.189444 (accumulated 1 forwards) [2026-07-10 03:44:11-INFO]: epoch 14-step 2867710 loss: 1.321916 (accumulated 1 forwards) [2026-07-10 03:44:20-INFO]: epoch 14-step 2867720 loss: 1.175803 (accumulated 1 forwards) [2026-07-10 03:44:29-INFO]: epoch 14-step 2867730 loss: 1.213647 (accumulated 1 forwards) [2026-07-10 03:44:38-INFO]: epoch 14-step 2867740 loss: 1.185459 (accumulated 1 forwards) [2026-07-10 03:44:46-INFO]: epoch 14-step 2867750 loss: 1.246192 (accumulated 1 forwards) [2026-07-10 03:44:55-INFO]: epoch 14-step 2867760 loss: 1.283425 (accumulated 1 forwards) [2026-07-10 03:45:04-INFO]: epoch 14-step 2867770 loss: 1.303285 (accumulated 1 forwards) [2026-07-10 03:45:13-INFO]: epoch 14-step 2867780 loss: 1.398857 (accumulated 1 forwards) [2026-07-10 03:45:22-INFO]: epoch 14-step 2867790 loss: 1.556573 (accumulated 1 forwards) [2026-07-10 03:45:31-INFO]: epoch 14-step 2867800 loss: 1.591814 (accumulated 1 forwards) [2026-07-10 03:45:40-INFO]: epoch 14-step 2867810 loss: 1.649999 (accumulated 1 forwards) [2026-07-10 03:45:49-INFO]: epoch 14-step 2867820 loss: 1.563783 (accumulated 1 forwards) [2026-07-10 03:45:58-INFO]: epoch 14-step 2867830 loss: 1.417940 (accumulated 1 forwards) [2026-07-10 03:46:07-INFO]: epoch 14-step 2867840 loss: 1.818210 (accumulated 1 forwards) [2026-07-10 03:46:16-INFO]: epoch 14-step 2867850 loss: 1.457593 (accumulated 1 forwards) [2026-07-10 03:46:24-INFO]: epoch 14-step 2867860 loss: 1.457617 (accumulated 1 forwards) [2026-07-10 03:46:33-INFO]: epoch 14-step 2867870 loss: 1.310428 (accumulated 1 forwards) [2026-07-10 03:46:42-INFO]: epoch 14-step 2867880 loss: 1.231807 (accumulated 1 forwards) [2026-07-10 03:46:51-INFO]: epoch 14-step 2867890 loss: 1.247895 (accumulated 1 forwards) [2026-07-10 03:47:00-INFO]: epoch 14-step 2867900 loss: 1.328918 (accumulated 1 forwards) [2026-07-10 03:47:09-INFO]: epoch 14-step 2867910 loss: 1.261137 (accumulated 1 forwards) [2026-07-10 03:47:18-INFO]: epoch 14-step 2867920 loss: 1.364934 (accumulated 1 forwards) [2026-07-10 03:47:27-INFO]: epoch 14-step 2867930 loss: 1.353586 (accumulated 1 forwards) [2026-07-10 03:47:36-INFO]: epoch 14-step 2867940 loss: 1.244631 (accumulated 1 forwards) [2026-07-10 03:47:45-INFO]: epoch 14-step 2867950 loss: 1.380893 (accumulated 1 forwards) [2026-07-10 03:47:54-INFO]: epoch 14-step 2867960 loss: 1.339791 (accumulated 1 forwards) [2026-07-10 03:48:03-INFO]: epoch 14-step 2867970 loss: 1.392343 (accumulated 1 forwards) [2026-07-10 03:48:12-INFO]: epoch 14-step 2867980 loss: 1.297990 (accumulated 1 forwards) [2026-07-10 03:48:20-INFO]: epoch 14-step 2867990 loss: 1.286975 (accumulated 1 forwards) [2026-07-10 03:48:29-INFO]: epoch 14-step 2868000 loss: 1.287859 (accumulated 1 forwards) [2026-07-10 03:48:38-INFO]: epoch 14-step 2868010 loss: 1.340903 (accumulated 1 forwards) [2026-07-10 03:48:47-INFO]: epoch 14-step 2868020 loss: 1.298156 (accumulated 1 forwards) [2026-07-10 03:48:56-INFO]: epoch 14-step 2868030 loss: 1.311315 (accumulated 1 forwards) [2026-07-10 03:49:05-INFO]: epoch 14-step 2868040 loss: 1.232444 (accumulated 1 forwards) [2026-07-10 03:49:14-INFO]: epoch 14-step 2868050 loss: 1.268517 (accumulated 1 forwards) [2026-07-10 03:49:23-INFO]: epoch 14-step 2868060 loss: 1.220671 (accumulated 1 forwards) [2026-07-10 03:49:32-INFO]: epoch 14-step 2868070 loss: 1.370002 (accumulated 1 forwards) [2026-07-10 03:49:41-INFO]: epoch 14-step 2868080 loss: 1.352007 (accumulated 1 forwards) [2026-07-10 03:49:50-INFO]: epoch 14-step 2868090 loss: 1.244271 (accumulated 1 forwards) [2026-07-10 03:49:59-INFO]: epoch 14-step 2868100 loss: 1.304112 (accumulated 1 forwards) [2026-07-10 03:50:08-INFO]: epoch 14-step 2868110 loss: 1.256199 (accumulated 1 forwards) [2026-07-10 03:50:17-INFO]: epoch 14-step 2868120 loss: 1.235333 (accumulated 1 forwards) [2026-07-10 03:50:25-INFO]: epoch 14-step 2868130 loss: 1.252192 (accumulated 1 forwards) [2026-07-10 03:50:34-INFO]: epoch 14-step 2868140 loss: 1.247524 (accumulated 1 forwards) [2026-07-10 03:50:43-INFO]: epoch 14-step 2868150 loss: 1.250457 (accumulated 1 forwards) [2026-07-10 03:50:52-INFO]: epoch 14-step 2868160 loss: 1.217262 (accumulated 1 forwards) [2026-07-10 03:51:01-INFO]: epoch 14-step 2868170 loss: 1.286111 (accumulated 1 forwards) [2026-07-10 03:51:10-INFO]: epoch 14-step 2868180 loss: 1.311347 (accumulated 1 forwards) [2026-07-10 03:51:10-INFO]: Reload dataloader [2026-07-10 03:51:19-INFO]: epoch 15-step 2868190 loss: 1.286911 (accumulated 1 forwards)