Upload checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins

checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/wandb/offline-run-20260118_210409-checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins-run0/files/output.log

@@ -1145,6 +1145,9 @@ wandb: For more information, check out the docs at: https://weave-docs.wandb.ai/
 [2026-01-18 23:17:06] (step=0000964) Train Loss mse: 0.0000, Train Loss ce: 0.0772, Train Steps/Sec: 0.13,
 [2026-01-18 23:17:14] (step=0000965) Train Loss mse: 0.0000, Train Loss ce: 0.0769, Train Steps/Sec: 0.13,
 [2026-01-18 23:17:21] (step=0000966) Train Loss mse: 0.0000, Train Loss ce: 0.0778, Train Steps/Sec: 0.13,
+[2026-01-18 23:17:29] (step=0000967) Train Loss mse: 0.0000, Train Loss ce: 0.0770, Train Steps/Sec: 0.13,
+[2026-01-18 23:17:37] (step=0000968) Train Loss mse: 0.0000, Train Loss ce: 0.0793, Train Steps/Sec: 0.13,
+[2026-01-18 23:17:45] (step=0000969) Train Loss mse: 0.0000, Train Loss ce: 0.0751, Train Steps/Sec: 0.12,
 base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step1000
 Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
 [eval debug] first 3 batch fingerprints:
@@ -1166,9 +1169,6 @@ Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap
   fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
   fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
 ce_avg: 0.1431104689836502, mse_avg: 0.0
-[2026-01-18 23:17:29] (step=0000967) Train Loss mse: 0.0000, Train Loss ce: 0.0770, Train Steps/Sec: 0.13,
-[2026-01-18 23:17:37] (step=0000968) Train Loss mse: 0.0000, Train Loss ce: 0.0793, Train Steps/Sec: 0.13,
-[2026-01-18 23:17:45] (step=0000969) Train Loss mse: 0.0000, Train Loss ce: 0.0751, Train Steps/Sec: 0.12,
 [2026-01-18 23:17:53] (step=0000970) Train Loss mse: 0.0000, Train Loss ce: 0.0761, Train Steps/Sec: 0.13,
 [2026-01-18 23:18:01] (step=0000971) Train Loss mse: 0.0000, Train Loss ce: 0.0746, Train Steps/Sec: 0.13,
 [2026-01-18 23:18:08] (step=0000972) Train Loss mse: 0.0000, Train Loss ce: 0.0773, Train Steps/Sec: 0.13,
@@ -2472,20 +2472,6 @@ ce_avg: 0.1431104689836502, mse_avg: 0.0
 [2026-01-19 02:08:28] (step=0002270) Train Loss mse: 0.0000, Train Loss ce: 0.0758, Train Steps/Sec: 0.13,
 [2026-01-19 02:08:36] (step=0002271) Train Loss mse: 0.0000, Train Loss ce: 0.0735, Train Steps/Sec: 0.12,
 [2026-01-19 02:08:44] (step=0002272) Train Loss mse: 0.0000, Train Loss ce: 0.0754, Train Steps/Sec: 0.13,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step2500
-Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-ce_avg: 0.15242451429367065, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step3000
-Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-ce_avg: 0.07420127838850021, mse_avg: 0.0
 [2026-01-19 02:08:52] (step=0002273) Train Loss mse: 0.0000, Train Loss ce: 0.0747, Train Steps/Sec: 0.13,
 [2026-01-19 02:09:00] (step=0002274) Train Loss mse: 0.0000, Train Loss ce: 0.0752, Train Steps/Sec: 0.13,
 [2026-01-19 02:09:08] (step=0002275) Train Loss mse: 0.0000, Train Loss ce: 0.0736, Train Steps/Sec: 0.13,
@@ -2616,6 +2602,20 @@ ce_avg: 0.07420127838850021, mse_avg: 0.0
 [2026-01-19 02:25:33] (step=0002400) Train Loss mse: 0.0000, Train Loss ce: 0.0746, Train Steps/Sec: 0.13,
 [2026-01-19 02:25:41] (step=0002401) Train Loss mse: 0.0000, Train Loss ce: 0.0732, Train Steps/Sec: 0.13,
 [2026-01-19 02:25:49] (step=0002402) Train Loss mse: 0.0000, Train Loss ce: 0.0755, Train Steps/Sec: 0.13,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step2500
+Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+ce_avg: 0.15242451429367065, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step3000
+Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+ce_avg: 0.07420127838850021, mse_avg: 0.0
 [2026-01-19 02:25:57] (step=0002403) Train Loss mse: 0.0000, Train Loss ce: 0.0755, Train Steps/Sec: 0.13,
 [2026-01-19 02:26:05] (step=0002404) Train Loss mse: 0.0000, Train Loss ce: 0.0755, Train Steps/Sec: 0.13,
 [2026-01-19 02:26:13] (step=0002405) Train Loss mse: 0.0000, Train Loss ce: 0.0739, Train Steps/Sec: 0.13,

checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/wandb/offline-run-20260119_053756-checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins-run0/files/output.log

@@ -1213,6 +1213,20 @@ wandb: For more information, check out the docs at: https://weave-docs.wandb.ai/
 [2026-01-19 08:05:14] (step=0001025) Train Loss mse: 0.0000, Train Loss ce: 0.0710, Train Steps/Sec: 0.12,
 [2026-01-19 08:05:22] (step=0001026) Train Loss mse: 0.0000, Train Loss ce: 0.0719, Train Steps/Sec: 0.12,
 [2026-01-19 08:05:31] (step=0001027) Train Loss mse: 0.0000, Train Loss ce: 0.0701, Train Steps/Sec: 0.12,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step1500
+Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+ce_avg: 0.1267234981060028, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step2000
+Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+ce_avg: 0.1393442451953888, mse_avg: 0.0
 [2026-01-19 08:05:39] (step=0001028) Train Loss mse: 0.0000, Train Loss ce: 0.0715, Train Steps/Sec: 0.12,
 [2026-01-19 08:05:47] (step=0001029) Train Loss mse: 0.0000, Train Loss ce: 0.0707, Train Steps/Sec: 0.13,
 [2026-01-19 08:05:55] (step=0001030) Train Loss mse: 0.0000, Train Loss ce: 0.0728, Train Steps/Sec: 0.12,
@@ -1229,20 +1243,6 @@ wandb: For more information, check out the docs at: https://weave-docs.wandb.ai/
 [2026-01-19 08:07:25] (step=0001041) Train Loss mse: 0.0000, Train Loss ce: 0.0711, Train Steps/Sec: 0.12,
 [2026-01-19 08:07:34] (step=0001042) Train Loss mse: 0.0000, Train Loss ce: 0.0709, Train Steps/Sec: 0.12,
 [2026-01-19 08:07:42] (step=0001043) Train Loss mse: 0.0000, Train Loss ce: 0.0700, Train Steps/Sec: 0.12,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step1500
-Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-ce_avg: 0.1267234981060028, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step2000
-Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-ce_avg: 0.1393442451953888, mse_avg: 0.0
 [2026-01-19 08:07:50] (step=0001044) Train Loss mse: 0.0000, Train Loss ce: 0.0709, Train Steps/Sec: 0.12,
 [2026-01-19 08:07:59] (step=0001045) Train Loss mse: 0.0000, Train Loss ce: 0.0697, Train Steps/Sec: 0.12,
 [2026-01-19 08:08:07] (step=0001046) Train Loss mse: 0.0000, Train Loss ce: 0.0718, Train Steps/Sec: 0.12,
@@ -2593,27 +2593,6 @@ ce_avg: 0.1393442451953888, mse_avg: 0.0
 [2026-01-19 11:12:34] (step=0002391) Train Loss mse: 0.0000, Train Loss ce: 0.0692, Train Steps/Sec: 0.12,
 [2026-01-19 11:12:43] (step=0002392) Train Loss mse: 0.0000, Train Loss ce: 0.0692, Train Steps/Sec: 0.12,
 [2026-01-19 11:12:51] (step=0002393) Train Loss mse: 0.0000, Train Loss ce: 0.0686, Train Steps/Sec: 0.12,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step2500
-Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-ce_avg: 0.14870576560497284, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step3000
-Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-ce_avg: 0.07034339010715485, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step3500
-Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-ce_avg: 0.07060375064611435, mse_avg: 0.0
 [2026-01-19 11:12:59] (step=0002394) Train Loss mse: 0.0000, Train Loss ce: 0.0692, Train Steps/Sec: 0.12,
 [2026-01-19 11:13:07] (step=0002395) Train Loss mse: 0.0000, Train Loss ce: 0.0696, Train Steps/Sec: 0.12,
 [2026-01-19 11:13:16] (step=0002396) Train Loss mse: 0.0000, Train Loss ce: 0.0689, Train Steps/Sec: 0.12,
@@ -2664,6 +2643,27 @@ ce_avg: 0.07060375064611435, mse_avg: 0.0
 [2026-01-19 11:19:26] (step=0002441) Train Loss mse: 0.0000, Train Loss ce: 0.0691, Train Steps/Sec: 0.12,
 [2026-01-19 11:19:35] (step=0002442) Train Loss mse: 0.0000, Train Loss ce: 0.0688, Train Steps/Sec: 0.12,
 [2026-01-19 11:19:43] (step=0002443) Train Loss mse: 0.0000, Train Loss ce: 0.0692, Train Steps/Sec: 0.12,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step2500
+Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+ce_avg: 0.14870576560497284, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step3000
+Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+ce_avg: 0.07034339010715485, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step3500
+Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+ce_avg: 0.07060375064611435, mse_avg: 0.0
 [2026-01-19 11:19:51] (step=0002444) Train Loss mse: 0.0000, Train Loss ce: 0.0689, Train Steps/Sec: 0.12,
 [2026-01-19 11:19:59] (step=0002445) Train Loss mse: 0.0000, Train Loss ce: 0.0693, Train Steps/Sec: 0.12,
 [2026-01-19 11:20:08] (step=0002446) Train Loss mse: 0.0000, Train Loss ce: 0.0680, Train Steps/Sec: 0.12,
@@ -3624,6 +3624,17 @@ ce_avg: 0.07060375064611435, mse_avg: 0.0
 [2026-01-19 13:33:33] (step=0003398) Train Loss mse: 0.0000, Train Loss ce: 0.0686, Train Steps/Sec: 0.12,
 [2026-01-19 13:33:42] (step=0003399) Train Loss mse: 0.0000, Train Loss ce: 0.0682, Train Steps/Sec: 0.11,
 [2026-01-19 13:33:50] (step=0003400) Train Loss mse: 0.0000, Train Loss ce: 0.0678, Train Steps/Sec: 0.12,
+[2026-01-19 13:33:58] (step=0003401) Train Loss mse: 0.0000, Train Loss ce: 0.0673, Train Steps/Sec: 0.12,
+[2026-01-19 13:34:06] (step=0003402) Train Loss mse: 0.0000, Train Loss ce: 0.0683, Train Steps/Sec: 0.12,
+[2026-01-19 13:34:14] (step=0003403) Train Loss mse: 0.0000, Train Loss ce: 0.0683, Train Steps/Sec: 0.12,
+[2026-01-19 13:34:23] (step=0003404) Train Loss mse: 0.0000, Train Loss ce: 0.0686, Train Steps/Sec: 0.12,
+[2026-01-19 13:34:31] (step=0003405) Train Loss mse: 0.0000, Train Loss ce: 0.0682, Train Steps/Sec: 0.12,
+[2026-01-19 13:34:39] (step=0003406) Train Loss mse: 0.0000, Train Loss ce: 0.0702, Train Steps/Sec: 0.12,
+[2026-01-19 13:34:47] (step=0003407) Train Loss mse: 0.0000, Train Loss ce: 0.0688, Train Steps/Sec: 0.12,
+[2026-01-19 13:34:55] (step=0003408) Train Loss mse: 0.0000, Train Loss ce: 0.0679, Train Steps/Sec: 0.12,
+[2026-01-19 13:35:03] (step=0003409) Train Loss mse: 0.0000, Train Loss ce: 0.0682, Train Steps/Sec: 0.13,
+[2026-01-19 13:35:12] (step=0003410) Train Loss mse: 0.0000, Train Loss ce: 0.0680, Train Steps/Sec: 0.12,
+[2026-01-19 13:35:20] (step=0003411) Train Loss mse: 0.0000, Train Loss ce: 0.0678, Train Steps/Sec: 0.12,
 base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step4000
 Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
 [eval debug] first 3 batch fingerprints:
@@ -3638,49 +3649,6 @@ Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap
   fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
   fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
 ce_avg: 0.07141611725091934, mse_avg: 0.0
-[2026-01-19 13:33:58] (step=0003401) Train Loss mse: 0.0000, Train Loss ce: 0.0673, Train Steps/Sec: 0.12,
-[2026-01-19 13:34:06] (step=0003402) Train Loss mse: 0.0000, Train Loss ce: 0.0683, Train Steps/Sec: 0.12,
-[2026-01-19 13:34:14] (step=0003403) Train Loss mse: 0.0000, Train Loss ce: 0.0683, Train Steps/Sec: 0.12,
-[2026-01-19 13:34:23] (step=0003404) Train Loss mse: 0.0000, Train Loss ce: 0.0686, Train Steps/Sec: 0.12,
-[2026-01-19 13:34:31] (step=0003405) Train Loss mse: 0.0000, Train Loss ce: 0.0682, Train Steps/Sec: 0.12,
-[2026-01-19 13:34:39] (step=0003406) Train Loss mse: 0.0000, Train Loss ce: 0.0702, Train Steps/Sec: 0.12,
-[2026-01-19 13:34:47] (step=0003407) Train Loss mse: 0.0000, Train Loss ce: 0.0688, Train Steps/Sec: 0.12,
-[2026-01-19 13:34:55] (step=0003408) Train Loss mse: 0.0000, Train Loss ce: 0.0679, Train Steps/Sec: 0.12,
-[2026-01-19 13:35:03] (step=0003409) Train Loss mse: 0.0000, Train Loss ce: 0.0682, Train Steps/Sec: 0.13,
-[2026-01-19 13:35:12] (step=0003410) Train Loss mse: 0.0000, Train Loss ce: 0.0680, Train Steps/Sec: 0.12,
-[2026-01-19 13:35:20] (step=0003411) Train Loss mse: 0.0000, Train Loss ce: 0.0678, Train Steps/Sec: 0.12,
-[2026-01-19 13:35:28] (step=0003412) Train Loss mse: 0.0000, Train Loss ce: 0.0677, Train Steps/Sec: 0.12,
-[2026-01-19 13:35:36] (step=0003413) Train Loss mse: 0.0000, Train Loss ce: 0.0674, Train Steps/Sec: 0.12,
-[2026-01-19 13:35:45] (step=0003414) Train Loss mse: 0.0000, Train Loss ce: 0.0678, Train Steps/Sec: 0.12,
-[2026-01-19 13:35:53] (step=0003415) Train Loss mse: 0.0000, Train Loss ce: 0.0684, Train Steps/Sec: 0.12,
-[2026-01-19 13:36:01] (step=0003416) Train Loss mse: 0.0000, Train Loss ce: 0.0669, Train Steps/Sec: 0.12,
-[2026-01-19 13:36:10] (step=0003417) Train Loss mse: 0.0000, Train Loss ce: 0.0692, Train Steps/Sec: 0.12,
-[2026-01-19 13:36:18] (step=0003418) Train Loss mse: 0.0000, Train Loss ce: 0.0684, Train Steps/Sec: 0.12,
-[2026-01-19 13:36:26] (step=0003419) Train Loss mse: 0.0000, Train Loss ce: 0.0675, Train Steps/Sec: 0.12,
-[2026-01-19 13:36:34] (step=0003420) Train Loss mse: 0.0000, Train Loss ce: 0.0674, Train Steps/Sec: 0.12,
-[2026-01-19 13:36:43] (step=0003421) Train Loss mse: 0.0000, Train Loss ce: 0.0682, Train Steps/Sec: 0.12,
-[2026-01-19 13:36:51] (step=0003422) Train Loss mse: 0.0000, Train Loss ce: 0.0681, Train Steps/Sec: 0.12,
-[2026-01-19 13:36:59] (step=0003423) Train Loss mse: 0.0000, Train Loss ce: 0.0686, Train Steps/Sec: 0.12,
-[2026-01-19 13:37:07] (step=0003424) Train Loss mse: 0.0000, Train Loss ce: 0.0695, Train Steps/Sec: 0.12,
-[2026-01-19 13:37:15] (step=0003425) Train Loss mse: 0.0000, Train Loss ce: 0.0680, Train Steps/Sec: 0.12,
-[2026-01-19 13:37:23] (step=0003426) Train Loss mse: 0.0000, Train Loss ce: 0.0681, Train Steps/Sec: 0.12,
-[2026-01-19 13:37:32] (step=0003427) Train Loss mse: 0.0000, Train Loss ce: 0.0674, Train Steps/Sec: 0.12,
-[2026-01-19 13:37:40] (step=0003428) Train Loss mse: 0.0000, Train Loss ce: 0.0667, Train Steps/Sec: 0.12,
-[2026-01-19 13:37:48] (step=0003429) Train Loss mse: 0.0000, Train Loss ce: 0.0689, Train Steps/Sec: 0.12,
-[2026-01-19 13:37:56] (step=0003430) Train Loss mse: 0.0000, Train Loss ce: 0.0693, Train Steps/Sec: 0.13,
-[2026-01-19 13:38:04] (step=0003431) Train Loss mse: 0.0000, Train Loss ce: 0.0684, Train Steps/Sec: 0.12,
-[2026-01-19 13:38:13] (step=0003432) Train Loss mse: 0.0000, Train Loss ce: 0.0675, Train Steps/Sec: 0.12,
-[2026-01-19 13:38:21] (step=0003433) Train Loss mse: 0.0000, Train Loss ce: 0.0680, Train Steps/Sec: 0.12,
-[2026-01-19 13:38:29] (step=0003434) Train Loss mse: 0.0000, Train Loss ce: 0.0679, Train Steps/Sec: 0.12,
-[2026-01-19 13:38:37] (step=0003435) Train Loss mse: 0.0000, Train Loss ce: 0.0683, Train Steps/Sec: 0.12,
-[2026-01-19 13:38:46] (step=0003436) Train Loss mse: 0.0000, Train Loss ce: 0.0670, Train Steps/Sec: 0.12,
-[2026-01-19 13:38:54] (step=0003437) Train Loss mse: 0.0000, Train Loss ce: 0.0668, Train Steps/Sec: 0.12,
-[2026-01-19 13:39:03] (step=0003438) Train Loss mse: 0.0000, Train Loss ce: 0.0673, Train Steps/Sec: 0.12,
-[2026-01-19 13:39:11] (step=0003439) Train Loss mse: 0.0000, Train Loss ce: 0.0693, Train Steps/Sec: 0.12,
-[2026-01-19 13:39:19] (step=0003440) Train Loss mse: 0.0000, Train Loss ce: 0.0674, Train Steps/Sec: 0.12,
-[2026-01-19 13:39:27] (step=0003441) Train Loss mse: 0.0000, Train Loss ce: 0.0682, Train Steps/Sec: 0.12,
-[2026-01-19 13:39:35] (step=0003442) Train Loss mse: 0.0000, Train Loss ce: 0.0674, Train Steps/Sec: 0.12,
-[2026-01-19 13:39:43] (step=0003443) Train Loss mse: 0.0000, Train Loss ce: 0.0679, Train Steps/Sec: 0.12,
 [2026-01-19 13:39:52] (step=0003444) Train Loss mse: 0.0000, Train Loss ce: 0.0683, Train Steps/Sec: 0.12,
 [2026-01-19 13:40:00] (step=0003445) Train Loss mse: 0.0000, Train Loss ce: 0.0679, Train Steps/Sec: 0.12,
 [2026-01-19 13:40:08] (step=0003446) Train Loss mse: 0.0000, Train Loss ce: 0.0681, Train Steps/Sec: 0.12,
@@ -4996,6 +4964,13 @@ ce_avg: 0.07141611725091934, mse_avg: 0.0
 [2026-01-19 16:40:06] (step=0004756) Train Loss mse: 0.0000, Train Loss ce: 0.0654, Train Steps/Sec: 0.12,
 [2026-01-19 16:40:14] (step=0004757) Train Loss mse: 0.0000, Train Loss ce: 0.0667, Train Steps/Sec: 0.12,
 [2026-01-19 16:40:22] (step=0004758) Train Loss mse: 0.0000, Train Loss ce: 0.0655, Train Steps/Sec: 0.12,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step5000
+Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+ce_avg: 0.07195709645748138, mse_avg: 0.0
 [2026-01-19 16:40:31] (step=0004759) Train Loss mse: 0.0000, Train Loss ce: 0.0645, Train Steps/Sec: 0.12,
 [2026-01-19 16:40:39] (step=0004760) Train Loss mse: 0.0000, Train Loss ce: 0.0656, Train Steps/Sec: 0.12,
 [2026-01-19 16:40:47] (step=0004761) Train Loss mse: 0.0000, Train Loss ce: 0.0654, Train Steps/Sec: 0.12,
@@ -5018,13 +4993,6 @@ ce_avg: 0.07141611725091934, mse_avg: 0.0
 [2026-01-19 16:43:06] (step=0004778) Train Loss mse: 0.0000, Train Loss ce: 0.0658, Train Steps/Sec: 0.12,
 [2026-01-19 16:43:14] (step=0004779) Train Loss mse: 0.0000, Train Loss ce: 0.0661, Train Steps/Sec: 0.12,
 [2026-01-19 16:43:22] (step=0004780) Train Loss mse: 0.0000, Train Loss ce: 0.0658, Train Steps/Sec: 0.12,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step5000
-Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-ce_avg: 0.07195709645748138, mse_avg: 0.0
 [2026-01-19 16:43:31] (step=0004781) Train Loss mse: 0.0000, Train Loss ce: 0.0664, Train Steps/Sec: 0.12,
 [2026-01-19 16:43:39] (step=0004782) Train Loss mse: 0.0000, Train Loss ce: 0.0653, Train Steps/Sec: 0.12,
 [2026-01-19 16:43:47] (step=0004783) Train Loss mse: 0.0000, Train Loss ce: 0.0657, Train Steps/Sec: 0.12,

checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/wandb/offline-run-20260122_153153-checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins-run0/files/output.log

@@ -871,6 +871,129 @@ wandb: For more information, check out the docs at: https://weave-docs.wandb.ai/
 [2026-01-22 17:18:30] (step=0000860) Train Loss mse: 0.0000, Train Loss ce: 0.0712, Train Steps/Sec: 0.15,
 [2026-01-22 17:18:37] (step=0000861) Train Loss mse: 0.0000, Train Loss ce: 0.0715, Train Steps/Sec: 0.14,
 [2026-01-22 17:18:44] (step=0000862) Train Loss mse: 0.0000, Train Loss ce: 0.0702, Train Steps/Sec: 0.15,
+[2026-01-22 17:18:51] (step=0000863) Train Loss mse: 0.0000, Train Loss ce: 0.0719, Train Steps/Sec: 0.14,
+[2026-01-22 17:18:58] (step=0000864) Train Loss mse: 0.0000, Train Loss ce: 0.0714, Train Steps/Sec: 0.15,
+[2026-01-22 17:19:04] (step=0000865) Train Loss mse: 0.0000, Train Loss ce: 0.0717, Train Steps/Sec: 0.15,
+[2026-01-22 17:19:12] (step=0000866) Train Loss mse: 0.0000, Train Loss ce: 0.0730, Train Steps/Sec: 0.13,
+[2026-01-22 17:19:19] (step=0000867) Train Loss mse: 0.0000, Train Loss ce: 0.0715, Train Steps/Sec: 0.15,
+[2026-01-22 17:19:25] (step=0000868) Train Loss mse: 0.0000, Train Loss ce: 0.0706, Train Steps/Sec: 0.15,
+[2026-01-22 17:19:32] (step=0000869) Train Loss mse: 0.0000, Train Loss ce: 0.0715, Train Steps/Sec: 0.15,
+[2026-01-22 17:19:39] (step=0000870) Train Loss mse: 0.0000, Train Loss ce: 0.0710, Train Steps/Sec: 0.15,
+[2026-01-22 17:19:46] (step=0000871) Train Loss mse: 0.0000, Train Loss ce: 0.0730, Train Steps/Sec: 0.15,
+[2026-01-22 17:19:54] (step=0000872) Train Loss mse: 0.0000, Train Loss ce: 0.0704, Train Steps/Sec: 0.13,
+[2026-01-22 17:20:01] (step=0000873) Train Loss mse: 0.0000, Train Loss ce: 0.0713, Train Steps/Sec: 0.14,
+[2026-01-22 17:20:07] (step=0000874) Train Loss mse: 0.0000, Train Loss ce: 0.0723, Train Steps/Sec: 0.15,
+[2026-01-22 17:20:14] (step=0000875) Train Loss mse: 0.0000, Train Loss ce: 0.0707, Train Steps/Sec: 0.15,
+[2026-01-22 17:20:20] (step=0000876) Train Loss mse: 0.0000, Train Loss ce: 0.0735, Train Steps/Sec: 0.15,
+[2026-01-22 17:20:27] (step=0000877) Train Loss mse: 0.0000, Train Loss ce: 0.0719, Train Steps/Sec: 0.15,
+[2026-01-22 17:20:34] (step=0000878) Train Loss mse: 0.0000, Train Loss ce: 0.0719, Train Steps/Sec: 0.14,
+[2026-01-22 17:20:41] (step=0000879) Train Loss mse: 0.0000, Train Loss ce: 0.0729, Train Steps/Sec: 0.13,
+[2026-01-22 17:20:49] (step=0000880) Train Loss mse: 0.0000, Train Loss ce: 0.0727, Train Steps/Sec: 0.13,
+[2026-01-22 17:20:56] (step=0000881) Train Loss mse: 0.0000, Train Loss ce: 0.0714, Train Steps/Sec: 0.14,
+[2026-01-22 17:21:03] (step=0000882) Train Loss mse: 0.0000, Train Loss ce: 0.0722, Train Steps/Sec: 0.14,
+[2026-01-22 17:21:10] (step=0000883) Train Loss mse: 0.0000, Train Loss ce: 0.0699, Train Steps/Sec: 0.14,
+[2026-01-22 17:21:17] (step=0000884) Train Loss mse: 0.0000, Train Loss ce: 0.0729, Train Steps/Sec: 0.14,
+[2026-01-22 17:21:24] (step=0000885) Train Loss mse: 0.0000, Train Loss ce: 0.0717, Train Steps/Sec: 0.14,
+[2026-01-22 17:21:31] (step=0000886) Train Loss mse: 0.0000, Train Loss ce: 0.0698, Train Steps/Sec: 0.15,
+[2026-01-22 17:21:38] (step=0000887) Train Loss mse: 0.0000, Train Loss ce: 0.0709, Train Steps/Sec: 0.15,
+[2026-01-22 17:21:44] (step=0000888) Train Loss mse: 0.0000, Train Loss ce: 0.0713, Train Steps/Sec: 0.15,
+[2026-01-22 17:21:51] (step=0000889) Train Loss mse: 0.0000, Train Loss ce: 0.0704, Train Steps/Sec: 0.15,
+[2026-01-22 17:21:58] (step=0000890) Train Loss mse: 0.0000, Train Loss ce: 0.0721, Train Steps/Sec: 0.14,
+[2026-01-22 17:22:06] (step=0000891) Train Loss mse: 0.0000, Train Loss ce: 0.0711, Train Steps/Sec: 0.13,
+[2026-01-22 17:22:12] (step=0000892) Train Loss mse: 0.0000, Train Loss ce: 0.0741, Train Steps/Sec: 0.15,
+[2026-01-22 17:22:19] (step=0000893) Train Loss mse: 0.0000, Train Loss ce: 0.0709, Train Steps/Sec: 0.15,
+[2026-01-22 17:22:26] (step=0000894) Train Loss mse: 0.0000, Train Loss ce: 0.0720, Train Steps/Sec: 0.14,
+[2026-01-22 17:22:33] (step=0000895) Train Loss mse: 0.0000, Train Loss ce: 0.0705, Train Steps/Sec: 0.15,
+[2026-01-22 17:22:40] (step=0000896) Train Loss mse: 0.0000, Train Loss ce: 0.0720, Train Steps/Sec: 0.13,
+[2026-01-22 17:22:47] (step=0000897) Train Loss mse: 0.0000, Train Loss ce: 0.0715, Train Steps/Sec: 0.15,
+[2026-01-22 17:22:54] (step=0000898) Train Loss mse: 0.0000, Train Loss ce: 0.0719, Train Steps/Sec: 0.14,
+[2026-01-22 17:23:01] (step=0000899) Train Loss mse: 0.0000, Train Loss ce: 0.0713, Train Steps/Sec: 0.15,
+[2026-01-22 17:23:08] (step=0000900) Train Loss mse: 0.0000, Train Loss ce: 0.0717, Train Steps/Sec: 0.15,
+[2026-01-22 17:23:14] (step=0000901) Train Loss mse: 0.0000, Train Loss ce: 0.0731, Train Steps/Sec: 0.15,
+[2026-01-22 17:23:22] (step=0000902) Train Loss mse: 0.0000, Train Loss ce: 0.0701, Train Steps/Sec: 0.14,
+[2026-01-22 17:23:29] (step=0000903) Train Loss mse: 0.0000, Train Loss ce: 0.0726, Train Steps/Sec: 0.14,
+[2026-01-22 17:23:36] (step=0000904) Train Loss mse: 0.0000, Train Loss ce: 0.0711, Train Steps/Sec: 0.15,
+[2026-01-22 17:23:42] (step=0000905) Train Loss mse: 0.0000, Train Loss ce: 0.0736, Train Steps/Sec: 0.15,
+[2026-01-22 17:23:49] (step=0000906) Train Loss mse: 0.0000, Train Loss ce: 0.0708, Train Steps/Sec: 0.14,
+[2026-01-22 17:23:57] (step=0000907) Train Loss mse: 0.0000, Train Loss ce: 0.0712, Train Steps/Sec: 0.14,
+[2026-01-22 17:24:04] (step=0000908) Train Loss mse: 0.0000, Train Loss ce: 0.0736, Train Steps/Sec: 0.14,
+[2026-01-22 17:24:11] (step=0000909) Train Loss mse: 0.0000, Train Loss ce: 0.0709, Train Steps/Sec: 0.14,
+[2026-01-22 17:24:17] (step=0000910) Train Loss mse: 0.0000, Train Loss ce: 0.0711, Train Steps/Sec: 0.15,
+[2026-01-22 17:24:25] (step=0000911) Train Loss mse: 0.0000, Train Loss ce: 0.0730, Train Steps/Sec: 0.13,
+[2026-01-22 17:24:32] (step=0000912) Train Loss mse: 0.0000, Train Loss ce: 0.0745, Train Steps/Sec: 0.15,
+[2026-01-22 17:24:38] (step=0000913) Train Loss mse: 0.0000, Train Loss ce: 0.0722, Train Steps/Sec: 0.15,
+[2026-01-22 17:24:45] (step=0000914) Train Loss mse: 0.0000, Train Loss ce: 0.0705, Train Steps/Sec: 0.14,
+[2026-01-22 17:24:53] (step=0000915) Train Loss mse: 0.0000, Train Loss ce: 0.0704, Train Steps/Sec: 0.14,
+[2026-01-22 17:24:59] (step=0000916) Train Loss mse: 0.0000, Train Loss ce: 0.0720, Train Steps/Sec: 0.15,
+[2026-01-22 17:25:06] (step=0000917) Train Loss mse: 0.0000, Train Loss ce: 0.0714, Train Steps/Sec: 0.15,
+[2026-01-22 17:25:13] (step=0000918) Train Loss mse: 0.0000, Train Loss ce: 0.0727, Train Steps/Sec: 0.15,
+[2026-01-22 17:25:20] (step=0000919) Train Loss mse: 0.0000, Train Loss ce: 0.0725, Train Steps/Sec: 0.14,
+[2026-01-22 17:25:27] (step=0000920) Train Loss mse: 0.0000, Train Loss ce: 0.0723, Train Steps/Sec: 0.14,
+[2026-01-22 17:25:35] (step=0000921) Train Loss mse: 0.0000, Train Loss ce: 0.0734, Train Steps/Sec: 0.14,
+[2026-01-22 17:25:41] (step=0000922) Train Loss mse: 0.0000, Train Loss ce: 0.0729, Train Steps/Sec: 0.15,
+[2026-01-22 17:25:48] (step=0000923) Train Loss mse: 0.0000, Train Loss ce: 0.0705, Train Steps/Sec: 0.15,
+[2026-01-22 17:25:55] (step=0000924) Train Loss mse: 0.0000, Train Loss ce: 0.0713, Train Steps/Sec: 0.14,
+[2026-01-22 17:26:02] (step=0000925) Train Loss mse: 0.0000, Train Loss ce: 0.0730, Train Steps/Sec: 0.15,
+[2026-01-22 17:26:09] (step=0000926) Train Loss mse: 0.0000, Train Loss ce: 0.0706, Train Steps/Sec: 0.15,
+[2026-01-22 17:26:16] (step=0000927) Train Loss mse: 0.0000, Train Loss ce: 0.0732, Train Steps/Sec: 0.14,
+[2026-01-22 17:26:23] (step=0000928) Train Loss mse: 0.0000, Train Loss ce: 0.0733, Train Steps/Sec: 0.15,
+[2026-01-22 17:26:29] (step=0000929) Train Loss mse: 0.0000, Train Loss ce: 0.0700, Train Steps/Sec: 0.15,
+[2026-01-22 17:26:36] (step=0000930) Train Loss mse: 0.0000, Train Loss ce: 0.0698, Train Steps/Sec: 0.16,
+[2026-01-22 17:26:43] (step=0000931) Train Loss mse: 0.0000, Train Loss ce: 0.0708, Train Steps/Sec: 0.15,
+[2026-01-22 17:26:50] (step=0000932) Train Loss mse: 0.0000, Train Loss ce: 0.0699, Train Steps/Sec: 0.13,
+[2026-01-22 17:26:58] (step=0000933) Train Loss mse: 0.0000, Train Loss ce: 0.0707, Train Steps/Sec: 0.13,
+[2026-01-22 17:27:05] (step=0000934) Train Loss mse: 0.0000, Train Loss ce: 0.0708, Train Steps/Sec: 0.14,
+[2026-01-22 17:27:12] (step=0000935) Train Loss mse: 0.0000, Train Loss ce: 0.0715, Train Steps/Sec: 0.14,
+[2026-01-22 17:27:19] (step=0000936) Train Loss mse: 0.0000, Train Loss ce: 0.0724, Train Steps/Sec: 0.13,
+[2026-01-22 17:27:26] (step=0000937) Train Loss mse: 0.0000, Train Loss ce: 0.0721, Train Steps/Sec: 0.15,
+[2026-01-22 17:27:32] (step=0000938) Train Loss mse: 0.0000, Train Loss ce: 0.0724, Train Steps/Sec: 0.15,
+[2026-01-22 17:27:40] (step=0000939) Train Loss mse: 0.0000, Train Loss ce: 0.0710, Train Steps/Sec: 0.14,
+[2026-01-22 17:27:46] (step=0000940) Train Loss mse: 0.0000, Train Loss ce: 0.0695, Train Steps/Sec: 0.15,
+[2026-01-22 17:27:53] (step=0000941) Train Loss mse: 0.0000, Train Loss ce: 0.0715, Train Steps/Sec: 0.14,
+[2026-01-22 17:28:00] (step=0000942) Train Loss mse: 0.0000, Train Loss ce: 0.0715, Train Steps/Sec: 0.15,
+[2026-01-22 17:28:07] (step=0000943) Train Loss mse: 0.0000, Train Loss ce: 0.0744, Train Steps/Sec: 0.15,
+[2026-01-22 17:28:14] (step=0000944) Train Loss mse: 0.0000, Train Loss ce: 0.0714, Train Steps/Sec: 0.15,
+[2026-01-22 17:28:21] (step=0000945) Train Loss mse: 0.0000, Train Loss ce: 0.0712, Train Steps/Sec: 0.14,
+[2026-01-22 17:28:29] (step=0000946) Train Loss mse: 0.0000, Train Loss ce: 0.0706, Train Steps/Sec: 0.13,
+[2026-01-22 17:28:35] (step=0000947) Train Loss mse: 0.0000, Train Loss ce: 0.0723, Train Steps/Sec: 0.15,
+[2026-01-22 17:28:42] (step=0000948) Train Loss mse: 0.0000, Train Loss ce: 0.0707, Train Steps/Sec: 0.15,
+[2026-01-22 17:28:49] (step=0000949) Train Loss mse: 0.0000, Train Loss ce: 0.0706, Train Steps/Sec: 0.14,
+[2026-01-22 17:28:56] (step=0000950) Train Loss mse: 0.0000, Train Loss ce: 0.0711, Train Steps/Sec: 0.15,
+[2026-01-22 17:29:03] (step=0000951) Train Loss mse: 0.0000, Train Loss ce: 0.0697, Train Steps/Sec: 0.14,
+[2026-01-22 17:29:11] (step=0000952) Train Loss mse: 0.0000, Train Loss ce: 0.0703, Train Steps/Sec: 0.13,
+[2026-01-22 17:29:18] (step=0000953) Train Loss mse: 0.0000, Train Loss ce: 0.0714, Train Steps/Sec: 0.13,
+[2026-01-22 17:29:26] (step=0000954) Train Loss mse: 0.0000, Train Loss ce: 0.0713, Train Steps/Sec: 0.14,
+[2026-01-22 17:29:32] (step=0000955) Train Loss mse: 0.0000, Train Loss ce: 0.0714, Train Steps/Sec: 0.15,
+[2026-01-22 17:29:39] (step=0000956) Train Loss mse: 0.0000, Train Loss ce: 0.0696, Train Steps/Sec: 0.15,
+[2026-01-22 17:29:46] (step=0000957) Train Loss mse: 0.0000, Train Loss ce: 0.0720, Train Steps/Sec: 0.14,
+[2026-01-22 17:29:53] (step=0000958) Train Loss mse: 0.0000, Train Loss ce: 0.0712, Train Steps/Sec: 0.14,
+[2026-01-22 17:30:01] (step=0000959) Train Loss mse: 0.0000, Train Loss ce: 0.0727, Train Steps/Sec: 0.13,
+[2026-01-22 17:30:08] (step=0000960) Train Loss mse: 0.0000, Train Loss ce: 0.0722, Train Steps/Sec: 0.13,
+[2026-01-22 17:30:15] (step=0000961) Train Loss mse: 0.0000, Train Loss ce: 0.0740, Train Steps/Sec: 0.15,
+[2026-01-22 17:30:22] (step=0000962) Train Loss mse: 0.0000, Train Loss ce: 0.0714, Train Steps/Sec: 0.14,
+[2026-01-22 17:30:30] (step=0000963) Train Loss mse: 0.0000, Train Loss ce: 0.0717, Train Steps/Sec: 0.13,
+[2026-01-22 17:30:37] (step=0000964) Train Loss mse: 0.0000, Train Loss ce: 0.0712, Train Steps/Sec: 0.14,
+[2026-01-22 17:30:44] (step=0000965) Train Loss mse: 0.0000, Train Loss ce: 0.0697, Train Steps/Sec: 0.14,
+[2026-01-22 17:30:51] (step=0000966) Train Loss mse: 0.0000, Train Loss ce: 0.0707, Train Steps/Sec: 0.15,
+[2026-01-22 17:30:57] (step=0000967) Train Loss mse: 0.0000, Train Loss ce: 0.0709, Train Steps/Sec: 0.15,
+[2026-01-22 17:31:04] (step=0000968) Train Loss mse: 0.0000, Train Loss ce: 0.0701, Train Steps/Sec: 0.15,
+[2026-01-22 17:31:11] (step=0000969) Train Loss mse: 0.0000, Train Loss ce: 0.0712, Train Steps/Sec: 0.14,
+[2026-01-22 17:31:18] (step=0000970) Train Loss mse: 0.0000, Train Loss ce: 0.0717, Train Steps/Sec: 0.14,
+[2026-01-22 17:31:25] (step=0000971) Train Loss mse: 0.0000, Train Loss ce: 0.0715, Train Steps/Sec: 0.15,
+[2026-01-22 17:31:32] (step=0000972) Train Loss mse: 0.0000, Train Loss ce: 0.0696, Train Steps/Sec: 0.15,
+[2026-01-22 17:31:39] (step=0000973) Train Loss mse: 0.0000, Train Loss ce: 0.0715, Train Steps/Sec: 0.14,
+[2026-01-22 17:31:46] (step=0000974) Train Loss mse: 0.0000, Train Loss ce: 0.0721, Train Steps/Sec: 0.14,
+[2026-01-22 17:31:52] (step=0000975) Train Loss mse: 0.0000, Train Loss ce: 0.0721, Train Steps/Sec: 0.15,
+[2026-01-22 17:32:00] (step=0000976) Train Loss mse: 0.0000, Train Loss ce: 0.0719, Train Steps/Sec: 0.14,
+[2026-01-22 17:32:07] (step=0000977) Train Loss mse: 0.0000, Train Loss ce: 0.0726, Train Steps/Sec: 0.14,
+[2026-01-22 17:32:13] (step=0000978) Train Loss mse: 0.0000, Train Loss ce: 0.0716, Train Steps/Sec: 0.15,
+[2026-01-22 17:32:20] (step=0000979) Train Loss mse: 0.0000, Train Loss ce: 0.0713, Train Steps/Sec: 0.15,
+[2026-01-22 17:32:27] (step=0000980) Train Loss mse: 0.0000, Train Loss ce: 0.0738, Train Steps/Sec: 0.14,
+[2026-01-22 17:32:34] (step=0000981) Train Loss mse: 0.0000, Train Loss ce: 0.0728, Train Steps/Sec: 0.14,
+[2026-01-22 17:32:42] (step=0000982) Train Loss mse: 0.0000, Train Loss ce: 0.0739, Train Steps/Sec: 0.14,
+[2026-01-22 17:32:49] (step=0000983) Train Loss mse: 0.0000, Train Loss ce: 0.0705, Train Steps/Sec: 0.14,
+[2026-01-22 17:32:55] (step=0000984) Train Loss mse: 0.0000, Train Loss ce: 0.0706, Train Steps/Sec: 0.15,
+[2026-01-22 17:33:02] (step=0000985) Train Loss mse: 0.0000, Train Loss ce: 0.0720, Train Steps/Sec: 0.14,
 FullyShardedDataParallel(
   (_fsdp_wrapped_module): Bagel(
     (language_model): Qwen2ForCausalLM(
@@ -1055,129 +1178,6 @@ Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap
   fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
   fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
 ce_avg: 0.14597271382808685, mse_avg: 0.0
-[2026-01-22 17:18:51] (step=0000863) Train Loss mse: 0.0000, Train Loss ce: 0.0719, Train Steps/Sec: 0.14,
-[2026-01-22 17:18:58] (step=0000864) Train Loss mse: 0.0000, Train Loss ce: 0.0714, Train Steps/Sec: 0.15,
-[2026-01-22 17:19:04] (step=0000865) Train Loss mse: 0.0000, Train Loss ce: 0.0717, Train Steps/Sec: 0.15,
-[2026-01-22 17:19:12] (step=0000866) Train Loss mse: 0.0000, Train Loss ce: 0.0730, Train Steps/Sec: 0.13,
-[2026-01-22 17:19:19] (step=0000867) Train Loss mse: 0.0000, Train Loss ce: 0.0715, Train Steps/Sec: 0.15,
-[2026-01-22 17:19:25] (step=0000868) Train Loss mse: 0.0000, Train Loss ce: 0.0706, Train Steps/Sec: 0.15,
-[2026-01-22 17:19:32] (step=0000869) Train Loss mse: 0.0000, Train Loss ce: 0.0715, Train Steps/Sec: 0.15,
-[2026-01-22 17:19:39] (step=0000870) Train Loss mse: 0.0000, Train Loss ce: 0.0710, Train Steps/Sec: 0.15,
-[2026-01-22 17:19:46] (step=0000871) Train Loss mse: 0.0000, Train Loss ce: 0.0730, Train Steps/Sec: 0.15,
-[2026-01-22 17:19:54] (step=0000872) Train Loss mse: 0.0000, Train Loss ce: 0.0704, Train Steps/Sec: 0.13,
-[2026-01-22 17:20:01] (step=0000873) Train Loss mse: 0.0000, Train Loss ce: 0.0713, Train Steps/Sec: 0.14,
-[2026-01-22 17:20:07] (step=0000874) Train Loss mse: 0.0000, Train Loss ce: 0.0723, Train Steps/Sec: 0.15,
-[2026-01-22 17:20:14] (step=0000875) Train Loss mse: 0.0000, Train Loss ce: 0.0707, Train Steps/Sec: 0.15,
-[2026-01-22 17:20:20] (step=0000876) Train Loss mse: 0.0000, Train Loss ce: 0.0735, Train Steps/Sec: 0.15,
-[2026-01-22 17:20:27] (step=0000877) Train Loss mse: 0.0000, Train Loss ce: 0.0719, Train Steps/Sec: 0.15,
-[2026-01-22 17:20:34] (step=0000878) Train Loss mse: 0.0000, Train Loss ce: 0.0719, Train Steps/Sec: 0.14,
-[2026-01-22 17:20:41] (step=0000879) Train Loss mse: 0.0000, Train Loss ce: 0.0729, Train Steps/Sec: 0.13,
-[2026-01-22 17:20:49] (step=0000880) Train Loss mse: 0.0000, Train Loss ce: 0.0727, Train Steps/Sec: 0.13,
-[2026-01-22 17:20:56] (step=0000881) Train Loss mse: 0.0000, Train Loss ce: 0.0714, Train Steps/Sec: 0.14,
-[2026-01-22 17:21:03] (step=0000882) Train Loss mse: 0.0000, Train Loss ce: 0.0722, Train Steps/Sec: 0.14,
-[2026-01-22 17:21:10] (step=0000883) Train Loss mse: 0.0000, Train Loss ce: 0.0699, Train Steps/Sec: 0.14,
-[2026-01-22 17:21:17] (step=0000884) Train Loss mse: 0.0000, Train Loss ce: 0.0729, Train Steps/Sec: 0.14,
-[2026-01-22 17:21:24] (step=0000885) Train Loss mse: 0.0000, Train Loss ce: 0.0717, Train Steps/Sec: 0.14,
-[2026-01-22 17:21:31] (step=0000886) Train Loss mse: 0.0000, Train Loss ce: 0.0698, Train Steps/Sec: 0.15,
-[2026-01-22 17:21:38] (step=0000887) Train Loss mse: 0.0000, Train Loss ce: 0.0709, Train Steps/Sec: 0.15,
-[2026-01-22 17:21:44] (step=0000888) Train Loss mse: 0.0000, Train Loss ce: 0.0713, Train Steps/Sec: 0.15,
-[2026-01-22 17:21:51] (step=0000889) Train Loss mse: 0.0000, Train Loss ce: 0.0704, Train Steps/Sec: 0.15,
-[2026-01-22 17:21:58] (step=0000890) Train Loss mse: 0.0000, Train Loss ce: 0.0721, Train Steps/Sec: 0.14,
-[2026-01-22 17:22:06] (step=0000891) Train Loss mse: 0.0000, Train Loss ce: 0.0711, Train Steps/Sec: 0.13,
-[2026-01-22 17:22:12] (step=0000892) Train Loss mse: 0.0000, Train Loss ce: 0.0741, Train Steps/Sec: 0.15,
-[2026-01-22 17:22:19] (step=0000893) Train Loss mse: 0.0000, Train Loss ce: 0.0709, Train Steps/Sec: 0.15,
-[2026-01-22 17:22:26] (step=0000894) Train Loss mse: 0.0000, Train Loss ce: 0.0720, Train Steps/Sec: 0.14,
-[2026-01-22 17:22:33] (step=0000895) Train Loss mse: 0.0000, Train Loss ce: 0.0705, Train Steps/Sec: 0.15,
-[2026-01-22 17:22:40] (step=0000896) Train Loss mse: 0.0000, Train Loss ce: 0.0720, Train Steps/Sec: 0.13,
-[2026-01-22 17:22:47] (step=0000897) Train Loss mse: 0.0000, Train Loss ce: 0.0715, Train Steps/Sec: 0.15,
-[2026-01-22 17:22:54] (step=0000898) Train Loss mse: 0.0000, Train Loss ce: 0.0719, Train Steps/Sec: 0.14,
-[2026-01-22 17:23:01] (step=0000899) Train Loss mse: 0.0000, Train Loss ce: 0.0713, Train Steps/Sec: 0.15,
-[2026-01-22 17:23:08] (step=0000900) Train Loss mse: 0.0000, Train Loss ce: 0.0717, Train Steps/Sec: 0.15,
-[2026-01-22 17:23:14] (step=0000901) Train Loss mse: 0.0000, Train Loss ce: 0.0731, Train Steps/Sec: 0.15,
-[2026-01-22 17:23:22] (step=0000902) Train Loss mse: 0.0000, Train Loss ce: 0.0701, Train Steps/Sec: 0.14,
-[2026-01-22 17:23:29] (step=0000903) Train Loss mse: 0.0000, Train Loss ce: 0.0726, Train Steps/Sec: 0.14,
-[2026-01-22 17:23:36] (step=0000904) Train Loss mse: 0.0000, Train Loss ce: 0.0711, Train Steps/Sec: 0.15,
-[2026-01-22 17:23:42] (step=0000905) Train Loss mse: 0.0000, Train Loss ce: 0.0736, Train Steps/Sec: 0.15,
-[2026-01-22 17:23:49] (step=0000906) Train Loss mse: 0.0000, Train Loss ce: 0.0708, Train Steps/Sec: 0.14,
-[2026-01-22 17:23:57] (step=0000907) Train Loss mse: 0.0000, Train Loss ce: 0.0712, Train Steps/Sec: 0.14,
-[2026-01-22 17:24:04] (step=0000908) Train Loss mse: 0.0000, Train Loss ce: 0.0736, Train Steps/Sec: 0.14,
-[2026-01-22 17:24:11] (step=0000909) Train Loss mse: 0.0000, Train Loss ce: 0.0709, Train Steps/Sec: 0.14,
-[2026-01-22 17:24:17] (step=0000910) Train Loss mse: 0.0000, Train Loss ce: 0.0711, Train Steps/Sec: 0.15,
-[2026-01-22 17:24:25] (step=0000911) Train Loss mse: 0.0000, Train Loss ce: 0.0730, Train Steps/Sec: 0.13,
-[2026-01-22 17:24:32] (step=0000912) Train Loss mse: 0.0000, Train Loss ce: 0.0745, Train Steps/Sec: 0.15,
-[2026-01-22 17:24:38] (step=0000913) Train Loss mse: 0.0000, Train Loss ce: 0.0722, Train Steps/Sec: 0.15,
-[2026-01-22 17:24:45] (step=0000914) Train Loss mse: 0.0000, Train Loss ce: 0.0705, Train Steps/Sec: 0.14,
-[2026-01-22 17:24:53] (step=0000915) Train Loss mse: 0.0000, Train Loss ce: 0.0704, Train Steps/Sec: 0.14,
-[2026-01-22 17:24:59] (step=0000916) Train Loss mse: 0.0000, Train Loss ce: 0.0720, Train Steps/Sec: 0.15,
-[2026-01-22 17:25:06] (step=0000917) Train Loss mse: 0.0000, Train Loss ce: 0.0714, Train Steps/Sec: 0.15,
-[2026-01-22 17:25:13] (step=0000918) Train Loss mse: 0.0000, Train Loss ce: 0.0727, Train Steps/Sec: 0.15,
-[2026-01-22 17:25:20] (step=0000919) Train Loss mse: 0.0000, Train Loss ce: 0.0725, Train Steps/Sec: 0.14,
-[2026-01-22 17:25:27] (step=0000920) Train Loss mse: 0.0000, Train Loss ce: 0.0723, Train Steps/Sec: 0.14,
-[2026-01-22 17:25:35] (step=0000921) Train Loss mse: 0.0000, Train Loss ce: 0.0734, Train Steps/Sec: 0.14,
-[2026-01-22 17:25:41] (step=0000922) Train Loss mse: 0.0000, Train Loss ce: 0.0729, Train Steps/Sec: 0.15,
-[2026-01-22 17:25:48] (step=0000923) Train Loss mse: 0.0000, Train Loss ce: 0.0705, Train Steps/Sec: 0.15,
-[2026-01-22 17:25:55] (step=0000924) Train Loss mse: 0.0000, Train Loss ce: 0.0713, Train Steps/Sec: 0.14,
-[2026-01-22 17:26:02] (step=0000925) Train Loss mse: 0.0000, Train Loss ce: 0.0730, Train Steps/Sec: 0.15,
-[2026-01-22 17:26:09] (step=0000926) Train Loss mse: 0.0000, Train Loss ce: 0.0706, Train Steps/Sec: 0.15,
-[2026-01-22 17:26:16] (step=0000927) Train Loss mse: 0.0000, Train Loss ce: 0.0732, Train Steps/Sec: 0.14,
-[2026-01-22 17:26:23] (step=0000928) Train Loss mse: 0.0000, Train Loss ce: 0.0733, Train Steps/Sec: 0.15,
-[2026-01-22 17:26:29] (step=0000929) Train Loss mse: 0.0000, Train Loss ce: 0.0700, Train Steps/Sec: 0.15,
-[2026-01-22 17:26:36] (step=0000930) Train Loss mse: 0.0000, Train Loss ce: 0.0698, Train Steps/Sec: 0.16,
-[2026-01-22 17:26:43] (step=0000931) Train Loss mse: 0.0000, Train Loss ce: 0.0708, Train Steps/Sec: 0.15,
-[2026-01-22 17:26:50] (step=0000932) Train Loss mse: 0.0000, Train Loss ce: 0.0699, Train Steps/Sec: 0.13,
-[2026-01-22 17:26:58] (step=0000933) Train Loss mse: 0.0000, Train Loss ce: 0.0707, Train Steps/Sec: 0.13,
-[2026-01-22 17:27:05] (step=0000934) Train Loss mse: 0.0000, Train Loss ce: 0.0708, Train Steps/Sec: 0.14,
-[2026-01-22 17:27:12] (step=0000935) Train Loss mse: 0.0000, Train Loss ce: 0.0715, Train Steps/Sec: 0.14,
-[2026-01-22 17:27:19] (step=0000936) Train Loss mse: 0.0000, Train Loss ce: 0.0724, Train Steps/Sec: 0.13,
-[2026-01-22 17:27:26] (step=0000937) Train Loss mse: 0.0000, Train Loss ce: 0.0721, Train Steps/Sec: 0.15,
-[2026-01-22 17:27:32] (step=0000938) Train Loss mse: 0.0000, Train Loss ce: 0.0724, Train Steps/Sec: 0.15,
-[2026-01-22 17:27:40] (step=0000939) Train Loss mse: 0.0000, Train Loss ce: 0.0710, Train Steps/Sec: 0.14,
-[2026-01-22 17:27:46] (step=0000940) Train Loss mse: 0.0000, Train Loss ce: 0.0695, Train Steps/Sec: 0.15,
-[2026-01-22 17:27:53] (step=0000941) Train Loss mse: 0.0000, Train Loss ce: 0.0715, Train Steps/Sec: 0.14,
-[2026-01-22 17:28:00] (step=0000942) Train Loss mse: 0.0000, Train Loss ce: 0.0715, Train Steps/Sec: 0.15,
-[2026-01-22 17:28:07] (step=0000943) Train Loss mse: 0.0000, Train Loss ce: 0.0744, Train Steps/Sec: 0.15,
-[2026-01-22 17:28:14] (step=0000944) Train Loss mse: 0.0000, Train Loss ce: 0.0714, Train Steps/Sec: 0.15,
-[2026-01-22 17:28:21] (step=0000945) Train Loss mse: 0.0000, Train Loss ce: 0.0712, Train Steps/Sec: 0.14,
-[2026-01-22 17:28:29] (step=0000946) Train Loss mse: 0.0000, Train Loss ce: 0.0706, Train Steps/Sec: 0.13,
-[2026-01-22 17:28:35] (step=0000947) Train Loss mse: 0.0000, Train Loss ce: 0.0723, Train Steps/Sec: 0.15,
-[2026-01-22 17:28:42] (step=0000948) Train Loss mse: 0.0000, Train Loss ce: 0.0707, Train Steps/Sec: 0.15,
-[2026-01-22 17:28:49] (step=0000949) Train Loss mse: 0.0000, Train Loss ce: 0.0706, Train Steps/Sec: 0.14,
-[2026-01-22 17:28:56] (step=0000950) Train Loss mse: 0.0000, Train Loss ce: 0.0711, Train Steps/Sec: 0.15,
-[2026-01-22 17:29:03] (step=0000951) Train Loss mse: 0.0000, Train Loss ce: 0.0697, Train Steps/Sec: 0.14,
-[2026-01-22 17:29:11] (step=0000952) Train Loss mse: 0.0000, Train Loss ce: 0.0703, Train Steps/Sec: 0.13,
-[2026-01-22 17:29:18] (step=0000953) Train Loss mse: 0.0000, Train Loss ce: 0.0714, Train Steps/Sec: 0.13,
-[2026-01-22 17:29:26] (step=0000954) Train Loss mse: 0.0000, Train Loss ce: 0.0713, Train Steps/Sec: 0.14,
-[2026-01-22 17:29:32] (step=0000955) Train Loss mse: 0.0000, Train Loss ce: 0.0714, Train Steps/Sec: 0.15,
-[2026-01-22 17:29:39] (step=0000956) Train Loss mse: 0.0000, Train Loss ce: 0.0696, Train Steps/Sec: 0.15,
-[2026-01-22 17:29:46] (step=0000957) Train Loss mse: 0.0000, Train Loss ce: 0.0720, Train Steps/Sec: 0.14,
-[2026-01-22 17:29:53] (step=0000958) Train Loss mse: 0.0000, Train Loss ce: 0.0712, Train Steps/Sec: 0.14,
-[2026-01-22 17:30:01] (step=0000959) Train Loss mse: 0.0000, Train Loss ce: 0.0727, Train Steps/Sec: 0.13,
-[2026-01-22 17:30:08] (step=0000960) Train Loss mse: 0.0000, Train Loss ce: 0.0722, Train Steps/Sec: 0.13,
-[2026-01-22 17:30:15] (step=0000961) Train Loss mse: 0.0000, Train Loss ce: 0.0740, Train Steps/Sec: 0.15,
-[2026-01-22 17:30:22] (step=0000962) Train Loss mse: 0.0000, Train Loss ce: 0.0714, Train Steps/Sec: 0.14,
-[2026-01-22 17:30:30] (step=0000963) Train Loss mse: 0.0000, Train Loss ce: 0.0717, Train Steps/Sec: 0.13,
-[2026-01-22 17:30:37] (step=0000964) Train Loss mse: 0.0000, Train Loss ce: 0.0712, Train Steps/Sec: 0.14,
-[2026-01-22 17:30:44] (step=0000965) Train Loss mse: 0.0000, Train Loss ce: 0.0697, Train Steps/Sec: 0.14,
-[2026-01-22 17:30:51] (step=0000966) Train Loss mse: 0.0000, Train Loss ce: 0.0707, Train Steps/Sec: 0.15,
-[2026-01-22 17:30:57] (step=0000967) Train Loss mse: 0.0000, Train Loss ce: 0.0709, Train Steps/Sec: 0.15,
-[2026-01-22 17:31:04] (step=0000968) Train Loss mse: 0.0000, Train Loss ce: 0.0701, Train Steps/Sec: 0.15,
-[2026-01-22 17:31:11] (step=0000969) Train Loss mse: 0.0000, Train Loss ce: 0.0712, Train Steps/Sec: 0.14,
-[2026-01-22 17:31:18] (step=0000970) Train Loss mse: 0.0000, Train Loss ce: 0.0717, Train Steps/Sec: 0.14,
-[2026-01-22 17:31:25] (step=0000971) Train Loss mse: 0.0000, Train Loss ce: 0.0715, Train Steps/Sec: 0.15,
-[2026-01-22 17:31:32] (step=0000972) Train Loss mse: 0.0000, Train Loss ce: 0.0696, Train Steps/Sec: 0.15,
-[2026-01-22 17:31:39] (step=0000973) Train Loss mse: 0.0000, Train Loss ce: 0.0715, Train Steps/Sec: 0.14,
-[2026-01-22 17:31:46] (step=0000974) Train Loss mse: 0.0000, Train Loss ce: 0.0721, Train Steps/Sec: 0.14,
-[2026-01-22 17:31:52] (step=0000975) Train Loss mse: 0.0000, Train Loss ce: 0.0721, Train Steps/Sec: 0.15,
-[2026-01-22 17:32:00] (step=0000976) Train Loss mse: 0.0000, Train Loss ce: 0.0719, Train Steps/Sec: 0.14,
-[2026-01-22 17:32:07] (step=0000977) Train Loss mse: 0.0000, Train Loss ce: 0.0726, Train Steps/Sec: 0.14,
-[2026-01-22 17:32:13] (step=0000978) Train Loss mse: 0.0000, Train Loss ce: 0.0716, Train Steps/Sec: 0.15,
-[2026-01-22 17:32:20] (step=0000979) Train Loss mse: 0.0000, Train Loss ce: 0.0713, Train Steps/Sec: 0.15,
-[2026-01-22 17:32:27] (step=0000980) Train Loss mse: 0.0000, Train Loss ce: 0.0738, Train Steps/Sec: 0.14,
-[2026-01-22 17:32:34] (step=0000981) Train Loss mse: 0.0000, Train Loss ce: 0.0728, Train Steps/Sec: 0.14,
-[2026-01-22 17:32:42] (step=0000982) Train Loss mse: 0.0000, Train Loss ce: 0.0739, Train Steps/Sec: 0.14,
-[2026-01-22 17:32:49] (step=0000983) Train Loss mse: 0.0000, Train Loss ce: 0.0705, Train Steps/Sec: 0.14,
-[2026-01-22 17:32:55] (step=0000984) Train Loss mse: 0.0000, Train Loss ce: 0.0706, Train Steps/Sec: 0.15,
-[2026-01-22 17:33:02] (step=0000985) Train Loss mse: 0.0000, Train Loss ce: 0.0720, Train Steps/Sec: 0.14,
 [2026-01-22 17:33:10] (step=0000986) Train Loss mse: 0.0000, Train Loss ce: 0.0711, Train Steps/Sec: 0.14,
 [2026-01-22 17:33:16] (step=0000987) Train Loss mse: 0.0000, Train Loss ce: 0.0720, Train Steps/Sec: 0.16,
 [2026-01-22 17:33:23] (step=0000988) Train Loss mse: 0.0000, Train Loss ce: 0.0711, Train Steps/Sec: 0.15,
@@ -2426,20 +2426,6 @@ ce_avg: 0.14597271382808685, mse_avg: 0.0
 [2026-01-22 19:57:55] (step=0002231) Train Loss mse: 0.0000, Train Loss ce: 0.0706, Train Steps/Sec: 0.14,
 [2026-01-22 19:58:01] (step=0002232) Train Loss mse: 0.0000, Train Loss ce: 0.0705, Train Steps/Sec: 0.15,
 [2026-01-22 19:58:09] (step=0002233) Train Loss mse: 0.0000, Train Loss ce: 0.0713, Train Steps/Sec: 0.13,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step2500
-Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-ce_avg: 0.1571701020002365, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step3000
-Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-ce_avg: 0.0698586106300354, mse_avg: 0.0
 [2026-01-22 19:58:15] (step=0002234) Train Loss mse: 0.0000, Train Loss ce: 0.0691, Train Steps/Sec: 0.15,
 [2026-01-22 19:58:23] (step=0002235) Train Loss mse: 0.0000, Train Loss ce: 0.0701, Train Steps/Sec: 0.14,
 [2026-01-22 19:58:29] (step=0002236) Train Loss mse: 0.0000, Train Loss ce: 0.0688, Train Steps/Sec: 0.15,
@@ -2567,6 +2553,27 @@ ce_avg: 0.0698586106300354, mse_avg: 0.0
 [2026-01-22 20:12:43] (step=0002358) Train Loss mse: 0.0000, Train Loss ce: 0.0688, Train Steps/Sec: 0.14,
 [2026-01-22 20:12:50] (step=0002359) Train Loss mse: 0.0000, Train Loss ce: 0.0684, Train Steps/Sec: 0.14,
 [2026-01-22 20:12:57] (step=0002360) Train Loss mse: 0.0000, Train Loss ce: 0.0691, Train Steps/Sec: 0.14,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step2500
+Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+ce_avg: 0.1571701020002365, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step3000
+Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+ce_avg: 0.0698586106300354, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step3500
+Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+ce_avg: 0.07012835144996643, mse_avg: 0.0
 [2026-01-22 20:13:03] (step=0002361) Train Loss mse: 0.0000, Train Loss ce: 0.0687, Train Steps/Sec: 0.15,
 [2026-01-22 20:13:10] (step=0002362) Train Loss mse: 0.0000, Train Loss ce: 0.0672, Train Steps/Sec: 0.15,
 [2026-01-22 20:13:17] (step=0002363) Train Loss mse: 0.0000, Train Loss ce: 0.0698, Train Steps/Sec: 0.14,
@@ -3521,6 +3528,20 @@ ce_avg: 0.0698586106300354, mse_avg: 0.0
 [2026-01-22 22:06:04] (step=0003309) Train Loss mse: 0.0000, Train Loss ce: 0.0683, Train Steps/Sec: 0.15,
 [2026-01-22 22:06:11] (step=0003310) Train Loss mse: 0.0000, Train Loss ce: 0.0683, Train Steps/Sec: 0.14,
 [2026-01-22 22:06:18] (step=0003311) Train Loss mse: 0.0000, Train Loss ce: 0.0683, Train Steps/Sec: 0.14,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step4000
+Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+ce_avg: 0.06947071850299835, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step4500
+Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+ce_avg: 0.069743812084198, mse_avg: 0.0
 [2026-01-22 22:06:26] (step=0003312) Train Loss mse: 0.0000, Train Loss ce: 0.0709, Train Steps/Sec: 0.13,
 [2026-01-22 22:06:33] (step=0003313) Train Loss mse: 0.0000, Train Loss ce: 0.0682, Train Steps/Sec: 0.15,
 [2026-01-22 22:06:39] (step=0003314) Train Loss mse: 0.0000, Train Loss ce: 0.0677, Train Steps/Sec: 0.14,
@@ -3597,27 +3618,6 @@ ce_avg: 0.0698586106300354, mse_avg: 0.0
 [2026-01-22 22:14:53] (step=0003385) Train Loss mse: 0.0000, Train Loss ce: 0.0689, Train Steps/Sec: 0.15,
 [2026-01-22 22:15:00] (step=0003386) Train Loss mse: 0.0000, Train Loss ce: 0.0678, Train Steps/Sec: 0.15,
 [2026-01-22 22:15:07] (step=0003387) Train Loss mse: 0.0000, Train Loss ce: 0.0687, Train Steps/Sec: 0.15,
-[
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step3500
-Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-ce_avg: 0.07012835144996643, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step4000
-Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-ce_avg: 0.06947071850299835, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step4500
-Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
 [2026-01-22 22:15:14] (step=0003388) Train Loss mse: 0.0000, Train Loss ce: 0.0697, Train Steps/Sec: 0.13,
 [2026-01-22 22:15:22] (step=0003389) Train Loss mse: 0.0000, Train Loss ce: 0.0691, Train Steps/Sec: 0.13,
 [2026-01-22 22:15:29] (step=0003390) Train Loss mse: 0.0000, Train Loss ce: 0.0710, Train Steps/Sec: 0.14,
@@ -4930,13 +4930,6 @@ Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap
 [2026-01-23 00:48:12] (step=0004697) Train Loss mse: 0.0000, Train Loss ce: 0.0683, Train Steps/Sec: 0.13,
 [2026-01-23 00:48:19] (step=0004698) Train Loss mse: 0.0000, Train Loss ce: 0.0700, Train Steps/Sec: 0.15,
 [2026-01-23 00:48:25] (step=0004699) Train Loss mse: 0.0000, Train Loss ce: 0.0676, Train Steps/Sec: 0.15,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step5000
-Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
-ce_avg: 0.07005834579467773, mse_avg: 0.0
 [2026-01-23 00:48:32] (step=0004700) Train Loss mse: 0.0000, Train Loss ce: 0.0675, Train Steps/Sec: 0.14,
 [2026-01-23 00:48:40] (step=0004701) Train Loss mse: 0.0000, Train Loss ce: 0.0679, Train Steps/Sec: 0.14,
 [2026-01-23 00:48:47] (step=0004702) Train Loss mse: 0.0000, Train Loss ce: 0.0690, Train Steps/Sec: 0.14,
@@ -4965,6 +4958,13 @@ ce_avg: 0.07005834579467773, mse_avg: 0.0
 [2026-01-23 00:51:29] (step=0004725) Train Loss mse: 0.0000, Train Loss ce: 0.0672, Train Steps/Sec: 0.14,
 [2026-01-23 00:51:37] (step=0004726) Train Loss mse: 0.0000, Train Loss ce: 0.0686, Train Steps/Sec: 0.14,
 [2026-01-23 00:51:44] (step=0004727) Train Loss mse: 0.0000, Train Loss ce: 0.0688, Train Steps/Sec: 0.14,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_no_mse_ins_step5000
+Preparing Dataset vlm_gym_jigsaw_swap_celoss_no_mse_evalonce/vlm_gym_jigsaw_swap_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_swap_celoss_no_mse_evalonce'}]
+ce_avg: 0.07005834579467773, mse_avg: 0.0
 [2026-01-23 00:51:51] (step=0004728) Train Loss mse: 0.0000, Train Loss ce: 0.0680, Train Steps/Sec: 0.14,
 [2026-01-23 00:51:57] (step=0004729) Train Loss mse: 0.0000, Train Loss ce: 0.0686, Train Steps/Sec: 0.15,
 [2026-01-23 00:52:04] (step=0004730) Train Loss mse: 0.0000, Train Loss ce: 0.0677, Train Steps/Sec: 0.14,