Upload checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins

checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/wandb/offline-run-20260126_192812-checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins-run0/files/output.log

@@ -1077,20 +1077,6 @@ wandb: For more information, check out the docs at: https://weave-docs.wandb.ai/
 [2026-01-26 19:50:13] (step=0000896) Train Loss mse: 0.0000, Train Loss ce: 0.1327, Train Steps/Sec: 1.02,
 [2026-01-26 19:50:14] (step=0000897) Train Loss mse: 0.0000, Train Loss ce: 0.0637, Train Steps/Sec: 1.02,
 [2026-01-26 19:50:15] (step=0000898) Train Loss mse: 0.0000, Train Loss ce: 0.0733, Train Steps/Sec: 1.02,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins_step1500
-Preparing Dataset vlm_gym_match_equation_sos_celoss_no_mse_evalonce/vlm_gym_match_equation_sos_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-ce_avg: 0.05330345034599304, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins_step2000
-Preparing Dataset vlm_gym_match_equation_sos_celoss_no_mse_evalonce/vlm_gym_match_equation_sos_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-ce_avg: 0.05127852410078049, mse_avg: 0.0
 [2026-01-26 19:50:16] (step=0000899) Train Loss mse: 0.0000, Train Loss ce: 0.0557, Train Steps/Sec: 1.02,
 [2026-01-26 19:50:17] (step=0000900) Train Loss mse: 0.0000, Train Loss ce: 0.0701, Train Steps/Sec: 0.80,
 [2026-01-26 19:50:18] (step=0000901) Train Loss mse: 0.0000, Train Loss ce: 0.0699, Train Steps/Sec: 1.01,
@@ -1303,6 +1289,27 @@ ce_avg: 0.05127852410078049, mse_avg: 0.0
 [2026-01-26 19:54:02] (step=0001108) Train Loss mse: 0.0000, Train Loss ce: 0.0604, Train Steps/Sec: 1.01,
 [2026-01-26 19:54:03] (step=0001109) Train Loss mse: 0.0000, Train Loss ce: 0.0570, Train Steps/Sec: 0.81,
 [2026-01-26 19:54:04] (step=0001110) Train Loss mse: 0.0000, Train Loss ce: 0.0559, Train Steps/Sec: 1.01,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins_step1500
+Preparing Dataset vlm_gym_match_equation_sos_celoss_no_mse_evalonce/vlm_gym_match_equation_sos_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+ce_avg: 0.05330345034599304, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins_step2000
+Preparing Dataset vlm_gym_match_equation_sos_celoss_no_mse_evalonce/vlm_gym_match_equation_sos_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+ce_avg: 0.05127852410078049, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins_step2500
+Preparing Dataset vlm_gym_match_equation_sos_celoss_no_mse_evalonce/vlm_gym_match_equation_sos_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+ce_avg: 0.05304804816842079, mse_avg: 0.0
 [2026-01-26 19:54:05] (step=0001111) Train Loss mse: 0.0000, Train Loss ce: 0.0425, Train Steps/Sec: 1.01,
 [2026-01-26 19:54:06] (step=0001112) Train Loss mse: 0.0000, Train Loss ce: 0.0792, Train Steps/Sec: 1.01,
 [2026-01-26 19:54:08] (step=0001113) Train Loss mse: 0.0000, Train Loss ce: 0.0488, Train Steps/Sec: 0.76,
@@ -2474,20 +2481,6 @@ ce_avg: 0.05127852410078049, mse_avg: 0.0
 [2026-01-26 20:14:52] (step=0002279) Train Loss mse: 0.0000, Train Loss ce: 0.0346, Train Steps/Sec: 1.02,
 [2026-01-26 20:14:53] (step=0002280) Train Loss mse: 0.0000, Train Loss ce: 0.0447, Train Steps/Sec: 1.02,
 [2026-01-26 20:14:54] (step=0002281) Train Loss mse: 0.0000, Train Loss ce: 0.0545, Train Steps/Sec: 0.75,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins_step2500
-Preparing Dataset vlm_gym_match_equation_sos_celoss_no_mse_evalonce/vlm_gym_match_equation_sos_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-ce_avg: 0.05304804816842079, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins_step3000
-Preparing Dataset vlm_gym_match_equation_sos_celoss_no_mse_evalonce/vlm_gym_match_equation_sos_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-ce_avg: 0.058707185089588165, mse_avg: 0.0
 [2026-01-26 20:14:55] (step=0002282) Train Loss mse: 0.0000, Train Loss ce: 0.0328, Train Steps/Sec: 0.81,
 [2026-01-26 20:14:56] (step=0002283) Train Loss mse: 0.0000, Train Loss ce: 0.0494, Train Steps/Sec: 1.02,
 [2026-01-26 20:14:57] (step=0002284) Train Loss mse: 0.0000, Train Loss ce: 0.0751, Train Steps/Sec: 1.02,
@@ -2820,6 +2813,20 @@ ce_avg: 0.058707185089588165, mse_avg: 0.0
 [2026-01-26 20:20:47] (step=0002611) Train Loss mse: 0.0000, Train Loss ce: 0.0525, Train Steps/Sec: 1.01,
 [2026-01-26 20:20:48] (step=0002612) Train Loss mse: 0.0000, Train Loss ce: 0.0433, Train Steps/Sec: 0.99,
 [2026-01-26 20:20:49] (step=0002613) Train Loss mse: 0.0000, Train Loss ce: 0.0471, Train Steps/Sec: 0.76,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins_step3000
+Preparing Dataset vlm_gym_match_equation_sos_celoss_no_mse_evalonce/vlm_gym_match_equation_sos_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+ce_avg: 0.058707185089588165, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins_step3500
+Preparing Dataset vlm_gym_match_equation_sos_celoss_no_mse_evalonce/vlm_gym_match_equation_sos_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+ce_avg: 0.10416685044765472, mse_avg: 0.0
 [2026-01-26 20:20:50] (step=0002614) Train Loss mse: 0.0000, Train Loss ce: 0.0586, Train Steps/Sec: 1.01,
 [2026-01-26 20:20:51] (step=0002615) Train Loss mse: 0.0000, Train Loss ce: 0.0357, Train Steps/Sec: 1.01,
 [2026-01-26 20:20:52] (step=0002616) Train Loss mse: 0.0000, Train Loss ce: 0.0313, Train Steps/Sec: 1.00,
@@ -3496,27 +3503,6 @@ ce_avg: 0.058707185089588165, mse_avg: 0.0
 [2026-01-26 20:32:48] (step=0003287) Train Loss mse: 0.0000, Train Loss ce: 0.0356, Train Steps/Sec: 1.01,
 [2026-01-26 20:32:49] (step=0003288) Train Loss mse: 0.0000, Train Loss ce: 0.0347, Train Steps/Sec: 1.01,
 [2026-01-26 20:32:51] (step=0003289) Train Loss mse: 0.0000, Train Loss ce: 0.0473, Train Steps/Sec: 0.75,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins_step3500
-Preparing Dataset vlm_gym_match_equation_sos_celoss_no_mse_evalonce/vlm_gym_match_equation_sos_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-ce_avg: 0.10416685044765472, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins_step4000
-Preparing Dataset vlm_gym_match_equation_sos_celoss_no_mse_evalonce/vlm_gym_match_equation_sos_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-ce_avg: 0.09295430034399033, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins_step4500
-Preparing Dataset vlm_gym_match_equation_sos_celoss_no_mse_evalonce/vlm_gym_match_equation_sos_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-ce_avg: 0.09719827771186829, mse_avg: 0.0
 [2026-01-26 20:32:52] (step=0003290) Train Loss mse: 0.0000, Train Loss ce: 0.0485, Train Steps/Sec: 0.81,
 [2026-01-26 20:32:53] (step=0003291) Train Loss mse: 0.0000, Train Loss ce: 0.0363, Train Steps/Sec: 1.02,
 [2026-01-26 20:32:54] (step=0003292) Train Loss mse: 0.0000, Train Loss ce: 0.0417, Train Steps/Sec: 1.01,
@@ -3921,27 +3907,27 @@ ce_avg: 0.09719827771186829, mse_avg: 0.0
 [2026-01-26 20:40:04] (step=0003691) Train Loss mse: 0.0000, Train Loss ce: 0.0398, Train Steps/Sec: 1.00,
 [2026-01-26 20:40:05] (step=0003692) Train Loss mse: 0.0000, Train Loss ce: 0.0272, Train Steps/Sec: 1.02,
 [2026-01-26 20:40:06] (step=0003693) Train Loss mse: 0.0000, Train Loss ce: 0.0313, Train Steps/Sec: 1.02,
-[2026-01-26 20:40:07] (step=0003694) Train Loss mse: 0.0000, Train Loss ce: 0.0528, Train Steps/Sec: 1.02,
-[2026-01-26 20:40:08] (step=0003695) Train Loss mse: 0.0000, Train Loss ce: 0.0489, Train Steps/Sec: 1.02,
-[2026-01-26 20:40:09] (step=0003696) Train Loss mse: 0.0000, Train Loss ce: 0.0267, Train Steps/Sec: 0.79,
-[2026-01-26 20:40:10] (step=0003697) Train Loss mse: 0.0000, Train Loss ce: 0.0365, Train Steps/Sec: 0.74,
-[2026-01-26 20:40:11] (step=0003698) Train Loss mse: 0.0000, Train Loss ce: 0.0334, Train Steps/Sec: 1.01,
-[2026-01-26 20:40:13] (step=0003699) Train Loss mse: 0.0000, Train Loss ce: 0.0528, Train Steps/Sec: 0.97,
-[2026-01-26 20:40:13] (step=0003700) Train Loss mse: 0.0000, Train Loss ce: 0.0260, Train Steps/Sec: 1.01,
-[2026-01-26 20:40:14] (step=0003701) Train Loss mse: 0.0000, Train Loss ce: 0.0514, Train Steps/Sec: 1.01,
-[2026-01-26 20:40:15] (step=0003702) Train Loss mse: 0.0000, Train Loss ce: 0.0590, Train Steps/Sec: 1.01,
-[2026-01-26 20:40:16] (step=0003703) Train Loss mse: 0.0000, Train Loss ce: 0.0275, Train Steps/Sec: 1.01,
-[2026-01-26 20:40:18] (step=0003704) Train Loss mse: 0.0000, Train Loss ce: 0.0255, Train Steps/Sec: 0.80,
-[2026-01-26 20:40:19] (step=0003705) Train Loss mse: 0.0000, Train Loss ce: 0.0308, Train Steps/Sec: 0.74,
-[2026-01-26 20:40:20] (step=0003706) Train Loss mse: 0.0000, Train Loss ce: 0.0407, Train Steps/Sec: 1.00,
-[2026-01-26 20:40:21] (step=0003707) Train Loss mse: 0.0000, Train Loss ce: 0.0359, Train Steps/Sec: 1.00,
-[2026-01-26 20:40:22] (step=0003708) Train Loss mse: 0.0000, Train Loss ce: 0.0362, Train Steps/Sec: 1.01,
-[2026-01-26 20:40:23] (step=0003709) Train Loss mse: 0.0000, Train Loss ce: 0.0446, Train Steps/Sec: 1.02,
-[2026-01-26 20:40:24] (step=0003710) Train Loss mse: 0.0000, Train Loss ce: 0.0333, Train Steps/Sec: 1.01,
-[2026-01-26 20:40:25] (step=0003711) Train Loss mse: 0.0000, Train Loss ce: 0.0461, Train Steps/Sec: 0.80,
-[2026-01-26 20:40:27] (step=0003712) Train Loss mse: 0.0000, Train Loss ce: 0.0473, Train Steps/Sec: 0.75,
-[2026-01-26 20:40:28] (step=0003713) Train Loss mse: 0.0000, Train Loss ce: 0.0338, Train Steps/Sec: 1.00,
-[2026-01-26 20:40:29] (step=0003714) Train Loss mse: 0.0000, Train Loss ce: 0.0298, Train Steps/Sec: 1.00,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins_step4000
+Preparing Dataset vlm_gym_match_equation_sos_celoss_no_mse_evalonce/vlm_gym_match_equation_sos_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+ce_avg: 0.09295430034399033, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins_step4500
+Preparing Dataset vlm_gym_match_equation_sos_celoss_no_mse_evalonce/vlm_gym_match_equation_sos_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+ce_avg: 0.09719827771186829, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins_step5000
+Preparing Dataset vlm_gym_match_equation_sos_celoss_no_mse_evalonce/vlm_gym_match_equation_sos_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
+ce_avg: 0.10010946542024612, mse_avg: 0.0
 [2026-01-26 20:40:30] (step=0003715) Train Loss mse: 0.0000, Train Loss ce: 0.0235, Train Steps/Sec: 1.01,
 [2026-01-26 20:40:31] (step=0003716) Train Loss mse: 0.0000, Train Loss ce: 0.0562, Train Steps/Sec: 1.01,
 [2026-01-26 20:40:32] (step=0003717) Train Loss mse: 0.0000, Train Loss ce: 0.0530, Train Steps/Sec: 0.99,
@@ -4909,13 +4895,6 @@ ce_avg: 0.09719827771186829, mse_avg: 0.0
 [2026-01-26 20:57:42] (step=0004679) Train Loss mse: 0.0000, Train Loss ce: 0.0178, Train Steps/Sec: 0.99,
 [2026-01-26 20:57:43] (step=0004680) Train Loss mse: 0.0000, Train Loss ce: 0.0502, Train Steps/Sec: 0.79,
 [2026-01-26 20:57:44] (step=0004681) Train Loss mse: 0.0000, Train Loss ce: 0.0485, Train Steps/Sec: 1.01,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_match_equation_sos_one_image_lr2e_5_ce_no_mse_ins_step5000
-Preparing Dataset vlm_gym_match_equation_sos_celoss_no_mse_evalonce/vlm_gym_match_equation_sos_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_match_equation_sos_celoss_no_mse_evalonce'}]
-ce_avg: 0.10010946542024612, mse_avg: 0.0
 [2026-01-26 20:57:46] (step=0004682) Train Loss mse: 0.0000, Train Loss ce: 0.0333, Train Steps/Sec: 0.74,
 [2026-01-26 20:57:47] (step=0004683) Train Loss mse: 0.0000, Train Loss ce: 0.0376, Train Steps/Sec: 1.01,
 [2026-01-26 20:57:48] (step=0004684) Train Loss mse: 0.0000, Train Loss ce: 0.0151, Train Steps/Sec: 1.02,