Upload checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed

checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/wandb/offline-run-20260111_233308-checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed-run0/files/output.log

@@ -819,35 +819,6 @@ wandb: For more information, check out the docs at: https://weave-docs.wandb.ai/
 [2026-01-12 02:47:03] (step=0000808) Train Loss mse: 0.0447, Train Loss ce: 0.0600, Train Steps/Sec: 0.09,
 [2026-01-12 02:47:19] (step=0000809) Train Loss mse: 0.0325, Train Loss ce: 0.0641, Train Steps/Sec: 0.06,
 [2026-01-12 02:47:31] (step=0000810) Train Loss mse: 0.0435, Train Loss ce: 0.0588, Train Steps/Sec: 0.08,
-[2026-01-12 02:47:43] (step=0000811) Train Loss mse: 0.0598, Train Loss ce: 0.0609, Train Steps/Sec: 0.08,
-[2026-01-12 02:47:55] (step=0000812) Train Loss mse: 0.0389, Train Loss ce: 0.0585, Train Steps/Sec: 0.09,
-[2026-01-12 02:48:08] (step=0000813) Train Loss mse: 0.0492, Train Loss ce: 0.0590, Train Steps/Sec: 0.07,
-[2026-01-12 02:48:21] (step=0000814) Train Loss mse: 0.0543, Train Loss ce: 0.0658, Train Steps/Sec: 0.08,
-[2026-01-12 02:48:38] (step=0000815) Train Loss mse: 0.0495, Train Loss ce: 0.0667, Train Steps/Sec: 0.06,
-[2026-01-12 02:48:53] (step=0000816) Train Loss mse: 0.0343, Train Loss ce: 0.0635, Train Steps/Sec: 0.06,
-[2026-01-12 02:49:09] (step=0000817) Train Loss mse: 0.0452, Train Loss ce: 0.0661, Train Steps/Sec: 0.06,
-[2026-01-12 02:49:23] (step=0000818) Train Loss mse: 0.0347, Train Loss ce: 0.0633, Train Steps/Sec: 0.07,
-[2026-01-12 02:49:34] (step=0000819) Train Loss mse: 0.0487, Train Loss ce: 0.0630, Train Steps/Sec: 0.09,
-[2026-01-12 02:49:45] (step=0000820) Train Loss mse: 0.0425, Train Loss ce: 0.0600, Train Steps/Sec: 0.09,
-[2026-01-12 02:49:58] (step=0000821) Train Loss mse: 0.0287, Train Loss ce: 0.0599, Train Steps/Sec: 0.07,
-[2026-01-12 02:50:09] (step=0000822) Train Loss mse: 0.0400, Train Loss ce: 0.0655, Train Steps/Sec: 0.09,
-[2026-01-12 02:50:22] (step=0000823) Train Loss mse: 0.0512, Train Loss ce: 0.0626, Train Steps/Sec: 0.08,
-[2026-01-12 02:50:38] (step=0000824) Train Loss mse: 0.0549, Train Loss ce: 0.0732, Train Steps/Sec: 0.06,
-[2026-01-12 02:50:51] (step=0000825) Train Loss mse: 0.0471, Train Loss ce: 0.0684, Train Steps/Sec: 0.07,
-[2026-01-12 02:51:03] (step=0000826) Train Loss mse: 0.0601, Train Loss ce: 0.0645, Train Steps/Sec: 0.08,
-[2026-01-12 02:51:14] (step=0000827) Train Loss mse: 0.0475, Train Loss ce: 0.0641, Train Steps/Sec: 0.09,
-[2026-01-12 02:51:28] (step=0000828) Train Loss mse: 0.0399, Train Loss ce: 0.0642, Train Steps/Sec: 0.07,
-[2026-01-12 02:51:44] (step=0000829) Train Loss mse: 0.0344, Train Loss ce: 0.0644, Train Steps/Sec: 0.06,
-[2026-01-12 02:51:55] (step=0000830) Train Loss mse: 0.0483, Train Loss ce: 0.0616, Train Steps/Sec: 0.09,
-[2026-01-12 02:52:07] (step=0000831) Train Loss mse: 0.0494, Train Loss ce: 0.0635, Train Steps/Sec: 0.08,
-[2026-01-12 02:52:20] (step=0000832) Train Loss mse: 0.0520, Train Loss ce: 0.0662, Train Steps/Sec: 0.08,
-[2026-01-12 02:52:35] (step=0000833) Train Loss mse: 0.0426, Train Loss ce: 0.0628, Train Steps/Sec: 0.06,
-[2026-01-12 02:52:47] (step=0000834) Train Loss mse: 0.0460, Train Loss ce: 0.0740, Train Steps/Sec: 0.08,
-[2026-01-12 02:53:03] (step=0000835) Train Loss mse: 0.0393, Train Loss ce: 0.0605, Train Steps/Sec: 0.06,
-[2026-01-12 02:53:17] (step=0000836) Train Loss mse: 0.0367, Train Loss ce: 0.0604, Train Steps/Sec: 0.07,
-[2026-01-12 02:53:29] (step=0000837) Train Loss mse: 0.0563, Train Loss ce: 0.0676, Train Steps/Sec: 0.08,
-[2026-01-12 02:53:45] (step=0000838) Train Loss mse: 0.0371, Train Loss ce: 0.0716, Train Steps/Sec: 0.06,
-[2026-01-12 02:53:59] (step=0000839) Train Loss mse: 0.0382, Train Loss ce: 0.0621, Train Steps/Sec: 0.07,
 FullyShardedDataParallel(
   (_fsdp_wrapped_module): Bagel(
     (language_model): Qwen2ForCausalLM(
@@ -1034,27 +1005,35 @@ Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
   fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
   fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
 ce_avg: 0.389182448387146, mse_avg: 0.06303369253873825
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed_step1000
-Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-ce_avg: 0.28593534231185913, mse_avg: 0.058155789971351624
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed_step1500
-Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-ce_avg: 0.13890723884105682, mse_avg: 0.054675184190273285
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed_step2000
-Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-ce_avg: 0.09731192141771317, mse_avg: 0.049818094819784164
+[2026-01-12 02:47:43] (step=0000811) Train Loss mse: 0.0598, Train Loss ce: 0.0609, Train Steps/Sec: 0.08,
+[2026-01-12 02:47:55] (step=0000812) Train Loss mse: 0.0389, Train Loss ce: 0.0585, Train Steps/Sec: 0.09,
+[2026-01-12 02:48:08] (step=0000813) Train Loss mse: 0.0492, Train Loss ce: 0.0590, Train Steps/Sec: 0.07,
+[2026-01-12 02:48:21] (step=0000814) Train Loss mse: 0.0543, Train Loss ce: 0.0658, Train Steps/Sec: 0.08,
+[2026-01-12 02:48:38] (step=0000815) Train Loss mse: 0.0495, Train Loss ce: 0.0667, Train Steps/Sec: 0.06,
+[2026-01-12 02:48:53] (step=0000816) Train Loss mse: 0.0343, Train Loss ce: 0.0635, Train Steps/Sec: 0.06,
+[2026-01-12 02:49:09] (step=0000817) Train Loss mse: 0.0452, Train Loss ce: 0.0661, Train Steps/Sec: 0.06,
+[2026-01-12 02:49:23] (step=0000818) Train Loss mse: 0.0347, Train Loss ce: 0.0633, Train Steps/Sec: 0.07,
+[2026-01-12 02:49:34] (step=0000819) Train Loss mse: 0.0487, Train Loss ce: 0.0630, Train Steps/Sec: 0.09,
+[2026-01-12 02:49:45] (step=0000820) Train Loss mse: 0.0425, Train Loss ce: 0.0600, Train Steps/Sec: 0.09,
+[2026-01-12 02:49:58] (step=0000821) Train Loss mse: 0.0287, Train Loss ce: 0.0599, Train Steps/Sec: 0.07,
+[2026-01-12 02:50:09] (step=0000822) Train Loss mse: 0.0400, Train Loss ce: 0.0655, Train Steps/Sec: 0.09,
+[2026-01-12 02:50:22] (step=0000823) Train Loss mse: 0.0512, Train Loss ce: 0.0626, Train Steps/Sec: 0.08,
+[2026-01-12 02:50:38] (step=0000824) Train Loss mse: 0.0549, Train Loss ce: 0.0732, Train Steps/Sec: 0.06,
+[2026-01-12 02:50:51] (step=0000825) Train Loss mse: 0.0471, Train Loss ce: 0.0684, Train Steps/Sec: 0.07,
+[2026-01-12 02:51:03] (step=0000826) Train Loss mse: 0.0601, Train Loss ce: 0.0645, Train Steps/Sec: 0.08,
+[2026-01-12 02:51:14] (step=0000827) Train Loss mse: 0.0475, Train Loss ce: 0.0641, Train Steps/Sec: 0.09,
+[2026-01-12 02:51:28] (step=0000828) Train Loss mse: 0.0399, Train Loss ce: 0.0642, Train Steps/Sec: 0.07,
+[2026-01-12 02:51:44] (step=0000829) Train Loss mse: 0.0344, Train Loss ce: 0.0644, Train Steps/Sec: 0.06,
+[2026-01-12 02:51:55] (step=0000830) Train Loss mse: 0.0483, Train Loss ce: 0.0616, Train Steps/Sec: 0.09,
+[2026-01-12 02:52:07] (step=0000831) Train Loss mse: 0.0494, Train Loss ce: 0.0635, Train Steps/Sec: 0.08,
+[2026-01-12 02:52:20] (step=0000832) Train Loss mse: 0.0520, Train Loss ce: 0.0662, Train Steps/Sec: 0.08,
+[2026-01-12 02:52:35] (step=0000833) Train Loss mse: 0.0426, Train Loss ce: 0.0628, Train Steps/Sec: 0.06,
+[2026-01-12 02:52:47] (step=0000834) Train Loss mse: 0.0460, Train Loss ce: 0.0740, Train Steps/Sec: 0.08,
+[2026-01-12 02:53:03] (step=0000835) Train Loss mse: 0.0393, Train Loss ce: 0.0605, Train Steps/Sec: 0.06,
+[2026-01-12 02:53:17] (step=0000836) Train Loss mse: 0.0367, Train Loss ce: 0.0604, Train Steps/Sec: 0.07,
+[2026-01-12 02:53:29] (step=0000837) Train Loss mse: 0.0563, Train Loss ce: 0.0676, Train Steps/Sec: 0.08,
+[2026-01-12 02:53:45] (step=0000838) Train Loss mse: 0.0371, Train Loss ce: 0.0716, Train Steps/Sec: 0.06,
+[2026-01-12 02:53:59] (step=0000839) Train Loss mse: 0.0382, Train Loss ce: 0.0621, Train Steps/Sec: 0.07,
 [2026-01-12 02:54:15] (step=0000840) Train Loss mse: 0.0466, Train Loss ce: 0.0669, Train Steps/Sec: 0.06,
 [2026-01-12 02:54:28] (step=0000841) Train Loss mse: 0.0454, Train Loss ce: 0.0709, Train Steps/Sec: 0.08,
 [2026-01-12 02:54:40] (step=0000842) Train Loss mse: 0.0398, Train Loss ce: 0.0702, Train Steps/Sec: 0.08,
@@ -2093,6 +2072,20 @@ ce_avg: 0.09731192141771317, mse_avg: 0.049818094819784164
 [2026-01-12 06:52:48] (step=0001875) Train Loss mse: 0.0338, Train Loss ce: 0.0594, Train Steps/Sec: 0.08,
 [2026-01-12 06:53:05] (step=0001876) Train Loss mse: 0.0441, Train Loss ce: 0.0604, Train Steps/Sec: 0.06,
 [2026-01-12 06:53:18] (step=0001877) Train Loss mse: 0.0328, Train Loss ce: 0.0611, Train Steps/Sec: 0.07,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed_step1000
+Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+ce_avg: 0.28593534231185913, mse_avg: 0.058155789971351624
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed_step1500
+Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+ce_avg: 0.13890723884105682, mse_avg: 0.054675184190273285
 [2026-01-12 06:53:31] (step=0001878) Train Loss mse: 0.0325, Train Loss ce: 0.0575, Train Steps/Sec: 0.07,
 [2026-01-12 06:53:45] (step=0001879) Train Loss mse: 0.0372, Train Loss ce: 0.0609, Train Steps/Sec: 0.08,
 [2026-01-12 06:53:58] (step=0001880) Train Loss mse: 0.0411, Train Loss ce: 0.0609, Train Steps/Sec: 0.07,
@@ -2303,20 +2296,6 @@ ce_avg: 0.09731192141771317, mse_avg: 0.049818094819784164
 [2026-01-12 07:41:52] (step=0002085) Train Loss mse: 0.0384, Train Loss ce: 0.0615, Train Steps/Sec: 0.08,
 [2026-01-12 07:42:04] (step=0002086) Train Loss mse: 0.0376, Train Loss ce: 0.0600, Train Steps/Sec: 0.09,
 [2026-01-12 07:42:19] (step=0002087) Train Loss mse: 0.0531, Train Loss ce: 0.0583, Train Steps/Sec: 0.07,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed_step2500
-Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-ce_avg: 0.0866585522890091, mse_avg: 0.04947679862380028
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed_step3000
-Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-ce_avg: 0.07865892350673676, mse_avg: 0.049521658569574356
 [2026-01-12 07:42:35] (step=0002088) Train Loss mse: 0.0440, Train Loss ce: 0.0601, Train Steps/Sec: 0.06,
 [2026-01-12 07:42:48] (step=0002089) Train Loss mse: 0.0460, Train Loss ce: 0.0582, Train Steps/Sec: 0.08,
 [2026-01-12 07:43:04] (step=0002090) Train Loss mse: 0.0397, Train Loss ce: 0.0569, Train Steps/Sec: 0.06,
@@ -3104,6 +3083,20 @@ ce_avg: 0.07865892350673676, mse_avg: 0.049521658569574356
 [2026-01-12 10:44:53] (step=0002869) Train Loss mse: 0.0422, Train Loss ce: 0.0587, Train Steps/Sec: 0.06,
 [2026-01-12 10:45:05] (step=0002870) Train Loss mse: 0.0456, Train Loss ce: 0.0590, Train Steps/Sec: 0.08,
 [2026-01-12 10:45:19] (step=0002871) Train Loss mse: 0.0379, Train Loss ce: 0.0585, Train Steps/Sec: 0.07,
+[2026-01-12 10:45:32
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed_step2000
+Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+ce_avg: 0.09731192141771317, mse_avg: 0.049818094819784164
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed_step2500
+Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
 [2026-01-12 10:45:32] (step=0002872) Train Loss mse: 0.0395, Train Loss ce: 0.0561, Train Steps/Sec: 0.08,
 [2026-01-12 10:45:46] (step=0002873) Train Loss mse: 0.0464, Train Loss ce: 0.0642, Train Steps/Sec: 0.07,
 [2026-01-12 10:46:00] (step=0002874) Train Loss mse: 0.0353, Train Loss ce: 0.0605, Train Steps/Sec: 0.07,
@@ -3211,20 +3204,6 @@ ce_avg: 0.07865892350673676, mse_avg: 0.049521658569574356
 [2026-01-12 11:08:58] (step=0002976) Train Loss mse: 0.0449, Train Loss ce: 0.0565, Train Steps/Sec: 0.07,
 [2026-01-12 11:09:14] (step=0002977) Train Loss mse: 0.0384, Train Loss ce: 0.0630, Train Steps/Sec: 0.06,
 [2026-01-12 11:09:28] (step=0002978) Train Loss mse: 0.0437, Train Loss ce: 0.0567, Train Steps/Sec: 0.07,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed_step3500
-Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-ce_avg: 0.07448671013116837, mse_avg: 0.045566461980342865
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed_step4000
-Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-ce_avg: 0.07164748013019562, mse_avg: 0.045982010662555695
 [2026-01-12 11:09:39] (step=0002979) Train Loss mse: 0.0554, Train Loss ce: 0.0585, Train Steps/Sec: 0.09,
 [2026-01-12 11:09:55] (step=0002980) Train Loss mse: 0.0392, Train Loss ce: 0.0592, Train Steps/Sec: 0.06,
 [2026-01-12 11:10:09] (step=0002981) Train Loss mse: 0.0305, Train Loss ce: 0.0559, Train Steps/Sec: 0.07,
@@ -4194,6 +4173,41 @@ ce_avg: 0.07164748013019562, mse_avg: 0.045982010662555695
 [2026-01-12 14:52:27] (step=0003945) Train Loss mse: 0.0261, Train Loss ce: 0.0557, Train Steps/Sec: 0.09,
 [2026-01-12 14:52:38] (step=0003946) Train Loss mse: 0.0454, Train Loss ce: 0.0571, Train Steps/Sec: 0.09,
 [2026-01-12 14:52:50] (step=0003947) Train Loss mse: 0.0509, Train Loss ce: 0.0586, Train Steps/Sec: 0.08,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed_step3000
+Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+ce_avg: 0.07865892350673676, mse_avg: 0.049521658569574356
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed_step3500
+Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+ce_avg: 0.07448671013116837, mse_avg: 0.045566461980342865
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed_step4000
+Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+ce_avg: 0.07164748013019562, mse_avg: 0.045982010662555695
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed_step4500
+Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+ce_avg: 0.06939579546451569, mse_avg: 0.04529677331447601
+base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed_step5000
+Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
+ce_avg: 0.06746137142181396, mse_avg: 0.044675178825855255
 [2026-01-12 14:53:03] (step=0003948) Train Loss mse: 0.0378, Train Loss ce: 0.0602, Train Steps/Sec: 0.08,
 [2026-01-12 14:53:17] (step=0003949) Train Loss mse: 0.0307, Train Loss ce: 0.0570, Train Steps/Sec: 0.07,
 [2026-01-12 14:53:29] (step=0003950) Train Loss mse: 0.0482, Train Loss ce: 0.0563, Train Steps/Sec: 0.08,
@@ -4506,20 +4520,6 @@ ce_avg: 0.07164748013019562, mse_avg: 0.045982010662555695
 [2026-01-12 16:05:02] (step=0004257) Train Loss mse: 0.0372, Train Loss ce: 0.0558, Train Steps/Sec: 0.09,
 [2026-01-12 16:05:15] (step=0004258) Train Loss mse: 0.0310, Train Loss ce: 0.0591, Train Steps/Sec: 0.07,
 [2026-01-12 16:05:31] (step=0004259) Train Loss mse: 0.0422, Train Loss ce: 0.0573, Train Steps/Sec: 0.06,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed_step4500
-Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-ce_avg: 0.06939579546451569, mse_avg: 0.04529677331447601
-base_dir is /dev/shm/models/checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed/eval_used_rows, step_tag is checkpoints_vlm_gym_jigsaw_one_image_lr2e_5_ce_ema9999_hashed_step5000
-Preparing Dataset vlm_gym_jigsaw_celoss_evalonce/vlm_gym_jigsaw_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_jigsaw_celoss_evalonce'}]
-ce_avg: 0.06746137142181396, mse_avg: 0.044675178825855255
 [2026-01-12 16:05:44] (step=0004260) Train Loss mse: 0.0486, Train Loss ce: 0.0621, Train Steps/Sec: 0.08,
 [2026-01-12 16:05:55] (step=0004261) Train Loss mse: 0.0576, Train Loss ce: 0.0519, Train Steps/Sec: 0.09,
 [2026-01-12 16:06:11] (step=0004262) Train Loss mse: 0.0290, Train Loss ce: 0.0579, Train Steps/Sec: 0.06,