Upload checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins

checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/wandb/offline-run-20260129_221514-checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins-run0/files/output.log

@@ -1,180 +1,3 @@
-FullyShardedDataParallel(
-  (_fsdp_wrapped_module): Bagel(
-    (language_model): Qwen2ForCausalLM(
-      (model): Qwen2Model(
-        (embed_tokens): Embedding(152064, 3584)
-        (layers): ModuleList(
-          (0-27): 28 x FullyShardedDataParallel(
-            (_fsdp_wrapped_module): CheckpointWrapper(
-              (_checkpoint_wrapped_module): Qwen2MoTDecoderLayer(
-                (self_attn): PackedAttentionMoT(
-                  (q_proj): Linear(in_features=3584, out_features=3584, bias=True)
-                  (k_proj): Linear(in_features=3584, out_features=512, bias=True)
-                  (v_proj): Linear(in_features=3584, out_features=512, bias=True)
-                  (o_proj): Linear(in_features=3584, out_features=3584, bias=False)
-                  (q_norm): Qwen2RMSNorm((128,), eps=1e-06)
-                  (k_norm): Qwen2RMSNorm((128,), eps=1e-06)
-                  (q_norm_moe_gen): Qwen2RMSNorm((128,), eps=1e-06)
-                  (k_norm_moe_gen): Qwen2RMSNorm((128,), eps=1e-06)
-                  (q_proj_moe_gen): Linear(in_features=3584, out_features=3584, bias=True)
-                  (k_proj_moe_gen): Linear(in_features=3584, out_features=512, bias=True)
-                  (v_proj_moe_gen): Linear(in_features=3584, out_features=512, bias=True)
-                  (o_proj_moe_gen): Linear(in_features=3584, out_features=3584, bias=False)
-                )
-                (mlp): Qwen2MLP(
-                  (gate_proj): Linear(in_features=3584, out_features=18944, bias=False)
-                  (up_proj): Linear(in_features=3584, out_features=18944, bias=False)
-                  (down_proj): Linear(in_features=18944, out_features=3584, bias=False)
-                  (act_fn): SiLU()
-                )
-                (mlp_moe_gen): Qwen2MLP(
-                  (gate_proj): Linear(in_features=3584, out_features=18944, bias=False)
-                  (up_proj): Linear(in_features=3584, out_features=18944, bias=False)
-                  (down_proj): Linear(in_features=18944, out_features=3584, bias=False)
-                  (act_fn): SiLU()
-                )
-                (input_layernorm): Qwen2RMSNorm((3584,), eps=1e-06)
-                (input_layernorm_moe_gen): Qwen2RMSNorm((3584,), eps=1e-06)
-                (post_attention_layernorm): Qwen2RMSNorm((3584,), eps=1e-06)
-                (post_attention_layernorm_moe_gen): Qwen2RMSNorm((3584,), eps=1e-06)
-              )
-            )
-          )
-        )
-        (norm): Qwen2RMSNorm((3584,), eps=1e-06)
-        (norm_moe_gen): Qwen2RMSNorm((3584,), eps=1e-06)
-        (rotary_emb): Qwen2RotaryEmbedding()
-      )
-      (lm_head): Linear(in_features=3584, out_features=152064, bias=False)
-    )
-    (vit_model): SiglipVisionModel(
-      (vision_model): FullyShardedDataParallel(
-        (_fsdp_wrapped_module): SiglipVisionTransformer(
-          (embeddings): SiglipVisionEmbeddings(
-            (position_embedding): Embedding(4900, 1152)
-            (patch_embedding): Linear(in_features=588, out_features=1152, bias=True)
-          )
-          (encoder): SiglipEncoder(
-            (layers): ModuleList(
-              (0-25): 26 x FullyShardedDataParallel(
-                (_fsdp_wrapped_module): CheckpointWrapper(
-                  (_checkpoint_wrapped_module): SiglipEncoderLayer(
-                    (self_attn): SiglipFlashAttention2(
-                      (k_proj): Linear(in_features=1152, out_features=1152, bias=True)
-                      (v_proj): Linear(in_features=1152, out_features=1152, bias=True)
-                      (q_proj): Linear(in_features=1152, out_features=1152, bias=True)
-                      (out_proj): Linear(in_features=1152, out_features=1152, bias=True)
-                    )
-                    (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True)
-                    (mlp): SiglipMLP(
-                      (activation_fn): PytorchGELUTanh()
-                      (fc1): Linear(in_features=1152, out_features=4304, bias=True)
-                      (fc2): Linear(in_features=4304, out_features=1152, bias=True)
-                    )
-                    (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True)
-                  )
-                )
-              )
-            )
-          )
-          (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True)
-        )
-      )
-    )
-    (connector): FullyShardedDataParallel(
-      (_fsdp_wrapped_module): CheckpointWrapper(
-        (_checkpoint_wrapped_module): MLPconnector(
-          (activation_fn): PytorchGELUTanh()
-          (fc1): Linear(in_features=1152, out_features=3584, bias=True)
-          (fc2): Linear(in_features=3584, out_features=3584, bias=True)
-        )
-      )
-    )
-    (vit_pos_embed): FullyShardedDataParallel(
-      (_fsdp_wrapped_module): PositionEmbedding()
-    )
-  )
-)
-_flat_param True
-language_model.model.layers.0._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.1._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.2._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.3._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.4._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.5._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.6._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.7._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.8._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.9._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.10._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.11._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.12._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.13._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.14._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.15._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.16._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.17._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.18._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.19._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.20._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.21._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.22._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.23._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.24._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.25._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.26._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-language_model.model.layers.27._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.0._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.1._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.2._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.3._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.4._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.5._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.6._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.7._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.8._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.9._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.10._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.11._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.12._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.13._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.14._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.15._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.16._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.17._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.18._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.19._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.20._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.21._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.22._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.23._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.24._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_model.vision_model._fsdp_wrapped_module.encoder.layers.25._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-connector._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
-vit_pos_embed._fsdp_wrapped_module._flat_param False
-Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse/vlm_gym_mujoco_pick_and_place_train
-base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step0
-Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-ce_avg: 0.2096945196390152, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step500
-Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-ce_avg: 0.01328575424849987, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step1000
-Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-ce_avg: 0.03305660933256149, mse_avg: 0.0
 wandb: Detected [huggingface_hub.inference] in use.
 wandb: Use W&B Weave for improved LLM call tracing. Install Weave with `pip install weave` then add `import weave` to the top of your script.
 wandb: For more information, check out the docs at: https://weave-docs.wandb.ai/
@@ -1235,20 +1058,6 @@ wandb: For more information, check out the docs at: https://weave-docs.wandb.ai/
 [2026-01-29 23:22:44] (step=0001047) Train Loss mse: 0.0000, Train Loss ce: 0.0137, Train Steps/Sec: 0.32,
 [2026-01-29 23:22:47] (step=0001048) Train Loss mse: 0.0000, Train Loss ce: 0.0137, Train Steps/Sec: 0.30,
 [2026-01-29 23:22:50] (step=0001049) Train Loss mse: 0.0000, Train Loss ce: 0.0067, Train Steps/Sec: 0.29,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step1500
-Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-ce_avg: 0.11604302376508713, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step2000
-Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-ce_avg: 0.14839798212051392, mse_avg: 0.0
 [2026-01-29 23:22:54] (step=0001050) Train Loss mse: 0.0000, Train Loss ce: 0.0090, Train Steps/Sec: 0.27,
 [2026-01-29 23:22:57] (step=0001051) Train Loss mse: 0.0000, Train Loss ce: 0.0099, Train Steps/Sec: 0.30,
 [2026-01-29 23:23:01] (step=0001052) Train Loss mse: 0.0000, Train Loss ce: 0.0126, Train Steps/Sec: 0.29,
@@ -1331,6 +1140,204 @@ ce_avg: 0.14839798212051392, mse_avg: 0.0
 [2026-01-29 23:27:22] (step=0001129) Train Loss mse: 0.0000, Train Loss ce: 0.0117, Train Steps/Sec: 0.31,
 [2026-01-29 23:27:25] (step=0001130) Train Loss mse: 0.0000, Train Loss ce: 0.0151, Train Steps/Sec: 0.30,
 [2026-01-29 23:27:28] (step=0001131) Train Loss mse: 0.0000, Train Loss ce: 0.0081, Train Steps/Sec: 0.29,
+FullyShardedDataParallel(
+  (_fsdp_wrapped_module): Bagel(
+    (language_model): Qwen2ForCausalLM(
+      (model): Qwen2Model(
+        (embed_tokens): Embedding(152064, 3584)
+        (layers): ModuleList(
+          (0-27): 28 x FullyShardedDataParallel(
+            (_fsdp_wrapped_module): CheckpointWrapper(
+              (_checkpoint_wrapped_module): Qwen2MoTDecoderLayer(
+                (self_attn): PackedAttentionMoT(
+                  (q_proj): Linear(in_features=3584, out_features=3584, bias=True)
+                  (k_proj): Linear(in_features=3584, out_features=512, bias=True)
+                  (v_proj): Linear(in_features=3584, out_features=512, bias=True)
+                  (o_proj): Linear(in_features=3584, out_features=3584, bias=False)
+                  (q_norm): Qwen2RMSNorm((128,), eps=1e-06)
+                  (k_norm): Qwen2RMSNorm((128,), eps=1e-06)
+                  (q_norm_moe_gen): Qwen2RMSNorm((128,), eps=1e-06)
+                  (k_norm_moe_gen): Qwen2RMSNorm((128,), eps=1e-06)
+                  (q_proj_moe_gen): Linear(in_features=3584, out_features=3584, bias=True)
+                  (k_proj_moe_gen): Linear(in_features=3584, out_features=512, bias=True)
+                  (v_proj_moe_gen): Linear(in_features=3584, out_features=512, bias=True)
+                  (o_proj_moe_gen): Linear(in_features=3584, out_features=3584, bias=False)
+                )
+                (mlp): Qwen2MLP(
+                  (gate_proj): Linear(in_features=3584, out_features=18944, bias=False)
+                  (up_proj): Linear(in_features=3584, out_features=18944, bias=False)
+                  (down_proj): Linear(in_features=18944, out_features=3584, bias=False)
+                  (act_fn): SiLU()
+                )
+                (mlp_moe_gen): Qwen2MLP(
+                  (gate_proj): Linear(in_features=3584, out_features=18944, bias=False)
+                  (up_proj): Linear(in_features=3584, out_features=18944, bias=False)
+                  (down_proj): Linear(in_features=18944, out_features=3584, bias=False)
+                  (act_fn): SiLU()
+                )
+                (input_layernorm): Qwen2RMSNorm((3584,), eps=1e-06)
+                (input_layernorm_moe_gen): Qwen2RMSNorm((3584,), eps=1e-06)
+                (post_attention_layernorm): Qwen2RMSNorm((3584,), eps=1e-06)
+                (post_attention_layernorm_moe_gen): Qwen2RMSNorm((3584,), eps=1e-06)
+              )
+            )
+          )
+        )
+        (norm): Qwen2RMSNorm((3584,), eps=1e-06)
+        (norm_moe_gen): Qwen2RMSNorm((3584,), eps=1e-06)
+        (rotary_emb): Qwen2RotaryEmbedding()
+      )
+      (lm_head): Linear(in_features=3584, out_features=152064, bias=False)
+    )
+    (vit_model): SiglipVisionModel(
+      (vision_model): FullyShardedDataParallel(
+        (_fsdp_wrapped_module): SiglipVisionTransformer(
+          (embeddings): SiglipVisionEmbeddings(
+            (position_embedding): Embedding(4900, 1152)
+            (patch_embedding): Linear(in_features=588, out_features=1152, bias=True)
+          )
+          (encoder): SiglipEncoder(
+            (layers): ModuleList(
+              (0-25): 26 x FullyShardedDataParallel(
+                (_fsdp_wrapped_module): CheckpointWrapper(
+                  (_checkpoint_wrapped_module): SiglipEncoderLayer(
+                    (self_attn): SiglipFlashAttention2(
+                      (k_proj): Linear(in_features=1152, out_features=1152, bias=True)
+                      (v_proj): Linear(in_features=1152, out_features=1152, bias=True)
+                      (q_proj): Linear(in_features=1152, out_features=1152, bias=True)
+                      (out_proj): Linear(in_features=1152, out_features=1152, bias=True)
+                    )
+                    (layer_norm1): LayerNorm((1152,), eps=1e-06, elementwise_affine=True)
+                    (mlp): SiglipMLP(
+                      (activation_fn): PytorchGELUTanh()
+                      (fc1): Linear(in_features=1152, out_features=4304, bias=True)
+                      (fc2): Linear(in_features=4304, out_features=1152, bias=True)
+                    )
+                    (layer_norm2): LayerNorm((1152,), eps=1e-06, elementwise_affine=True)
+                  )
+                )
+              )
+            )
+          )
+          (post_layernorm): LayerNorm((1152,), eps=1e-06, elementwise_affine=True)
+        )
+      )
+    )
+    (connector): FullyShardedDataParallel(
+      (_fsdp_wrapped_module): CheckpointWrapper(
+        (_checkpoint_wrapped_module): MLPconnector(
+          (activation_fn): PytorchGELUTanh()
+          (fc1): Linear(in_features=1152, out_features=3584, bias=True)
+          (fc2): Linear(in_features=3584, out_features=3584, bias=True)
+        )
+      )
+    )
+    (vit_pos_embed): FullyShardedDataParallel(
+      (_fsdp_wrapped_module): PositionEmbedding()
+    )
+  )
+)
+_flat_param True
+language_model.model.layers.0._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.1._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.2._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.3._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.4._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.5._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.6._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.7._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.8._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.9._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.10._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.11._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.12._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.13._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.14._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.15._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.16._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.17._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.18._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.19._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.20._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.21._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.22._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.23._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.24._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.25._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.26._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+language_model.model.layers.27._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.0._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.1._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.2._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.3._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.4._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.5._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.6._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.7._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.8._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.9._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.10._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.11._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.12._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.13._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.14._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.15._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.16._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.17._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.18._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.19._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.20._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.21._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.22._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.23._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.24._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_model.vision_model._fsdp_wrapped_module.encoder.layers.25._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+connector._fsdp_wrapped_module._checkpoint_wrapped_module._flat_param True
+vit_pos_embed._fsdp_wrapped_module._flat_param False
+Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse/vlm_gym_mujoco_pick_and_place_train
+base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step0
+Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+ce_avg: 0.2096945196390152, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step500
+Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+ce_avg: 0.01328575424849987, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step1000
+Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+ce_avg: 0.03305660933256149, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step1500
+Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+ce_avg: 0.11604302376508713, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step2000
+Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+ce_avg: 0.14839798212051392, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step2500
+Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+ce_avg: 0.18175341188907623, mse_avg: 0.0
 [2026-01-29 23:27:32] (step=0001132) Train Loss mse: 0.0000, Train Loss ce: 0.0176, Train Steps/Sec: 0.28,
 [2026-01-29 23:27:35] (step=0001133) Train Loss mse: 0.0000, Train Loss ce: 0.0101, Train Steps/Sec: 0.29,
 [2026-01-29 23:27:39] (step=0001134) Train Loss mse: 0.0000, Train Loss ce: 0.0164, Train Steps/Sec: 0.30,
@@ -2540,27 +2547,6 @@ ce_avg: 0.14839798212051392, mse_avg: 0.0
 [2026-01-30 00:36:39] (step=0002338) Train Loss mse: 0.0000, Train Loss ce: 0.0100, Train Steps/Sec: 0.29,
 [2026-01-30 00:36:42] (step=0002339) Train Loss mse: 0.0000, Train Loss ce: 0.0061, Train Steps/Sec: 0.31,
 [2026-01-30 00:36:46] (step=0002340) Train Loss mse: 0.0000, Train Loss ce: 0.0084, Train Steps/Sec: 0.26,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step2500
-Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-ce_avg: 0.18175341188907623, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step3000
-Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-ce_avg: 0.3127627372741699, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step3500
-Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-ce_avg: 0.4038556218147278, mse_avg: 0.0
 [2026-01-30 00:36:50] (step=0002341) Train Loss mse: 0.0000, Train Loss ce: 0.0155, Train Steps/Sec: 0.28,
 [2026-01-30 00:36:53] (step=0002342) Train Loss mse: 0.0000, Train Loss ce: 0.0095, Train Steps/Sec: 0.29,
 [2026-01-30 00:36:56] (step=0002343) Train Loss mse: 0.0000, Train Loss ce: 0.0115, Train Steps/Sec: 0.31,
@@ -2919,6 +2905,20 @@ ce_avg: 0.4038556218147278, mse_avg: 0.0
 [2026-01-30 00:57:16] (step=0002696) Train Loss mse: 0.0000, Train Loss ce: 0.0089, Train Steps/Sec: 0.29,
 [2026-01-30 00:57:19] (step=0002697) Train Loss mse: 0.0000, Train Loss ce: 0.0051, Train Steps/Sec: 0.29,
 [2026-01-30 00:57:23] (step=0002698) Train Loss mse: 0.0000, Train Loss ce: 0.0030, Train Steps/Sec: 0.30,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step3000
+Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+ce_avg: 0.3127627372741699, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step3500
+Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+ce_avg: 0.4038556218147278, mse_avg: 0.0
 [2026-01-30 00:57:26] (step=0002699) Train Loss mse: 0.0000, Train Loss ce: 0.0062, Train Steps/Sec: 0.30,
 [2026-01-30 00:57:29] (step=0002700) Train Loss mse: 0.0000, Train Loss ce: 0.0085, Train Steps/Sec: 0.30,
 [2026-01-30 00:57:32] (step=0002701) Train Loss mse: 0.0000, Train Loss ce: 0.0060, Train Steps/Sec: 0.30,
@@ -3606,20 +3606,6 @@ ce_avg: 0.4038556218147278, mse_avg: 0.0
 [2026-01-30 01:36:44] (step=0003383) Train Loss mse: 0.0000, Train Loss ce: 0.0041, Train Steps/Sec: 0.30,
 [2026-01-30 01:36:48] (step=0003384) Train Loss mse: 0.0000, Train Loss ce: 0.0091, Train Steps/Sec: 0.30,
 [2026-01-30 01:36:51] (step=0003385) Train Loss mse: 0.0000, Train Loss ce: 0.0047, Train Steps/Sec: 0.27,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step4000
-Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-ce_avg: 0.4453171491622925, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step4500
-Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-ce_avg: 0.45896753668785095, mse_avg: 0.0
 [2026-01-30 01:36:55] (step=0003386) Train Loss mse: 0.0000, Train Loss ce: 0.0072, Train Steps/Sec: 0.29,
 [2026-01-30 01:36:58] (step=0003387) Train Loss mse: 0.0000, Train Loss ce: 0.0067, Train Steps/Sec: 0.28,
 [2026-01-30 01:37:02] (step=0003388) Train Loss mse: 0.0000, Train Loss ce: 0.0043, Train Steps/Sec: 0.30,
@@ -3995,6 +3981,27 @@ ce_avg: 0.45896753668785095, mse_avg: 0.0
 [2026-01-30 01:58:17] (step=0003758) Train Loss mse: 0.0000, Train Loss ce: 0.0057, Train Steps/Sec: 0.27,
 [2026-01-30 01:58:20] (step=0003759) Train Loss mse: 0.0000, Train Loss ce: 0.0044, Train Steps/Sec: 0.31,
 [2026-01-30 01:58:23] (step=0003760) Train Loss mse: 0.0000, Train Loss ce: 0.0036, Train Steps/Sec: 0.31,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step4000
+Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+ce_avg: 0.4453171491622925, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step4500
+Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+ce_avg: 0.45896753668785095, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step5000
+Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+ce_avg: 0.46307557821273804, mse_avg: 0.0
 [2026-01-30 01:58:27] (step=0003761) Train Loss mse: 0.0000, Train Loss ce: 0.0039, Train Steps/Sec: 0.30,
 [2026-01-30 01:58:30] (step=0003762) Train Loss mse: 0.0000, Train Loss ce: 0.0064, Train Steps/Sec: 0.34,
 [2026-01-30 01:58:33] (step=0003763) Train Loss mse: 0.0000, Train Loss ce: 0.0077, Train Steps/Sec: 0.28,
@@ -5134,13 +5141,6 @@ ce_avg: 0.45896753668785095, mse_avg: 0.0
 [2026-01-30 03:04:00] (step=0004897) Train Loss mse: 0.0000, Train Loss ce: 0.0064, Train Steps/Sec: 0.29,
 [2026-01-30 03:04:03] (step=0004898) Train Loss mse: 0.0000, Train Loss ce: 0.0113, Train Steps/Sec: 0.30,
 [2026-01-30 03:04:07] (step=0004899) Train Loss mse: 0.0000, Train Loss ce: 0.0041, Train Steps/Sec: 0.29,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step5000
-Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-ce_avg: 0.46307557821273804, mse_avg: 0.0
 [2026-01-30 03:04:10] (step=0004900) Train Loss mse: 0.0000, Train Loss ce: 0.0053, Train Steps/Sec: 0.30,
 [2026-01-30 03:04:13] (step=0004901) Train Loss mse: 0.0000, Train Loss ce: 0.0057, Train Steps/Sec: 0.29,
 [2026-01-30 03:04:17] (step=0004902) Train Loss mse: 0.0000, Train Loss ce: 0.0051, Train Steps/Sec: 0.28,

checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/wandb/offline-run-20260130_175015-checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins-run0/files/output.log

@@ -1148,6 +1148,7 @@ wandb: For more information, check out the docs at: https://weave-docs.wandb.ai/
 [2026-01-30 18:52:55] (step=0000967) Train Loss mse: 0.0000, Train Loss ce: 0.0138, Train Steps/Sec: 0.28,
 [2026-01-30 18:52:59] (step=0000968) Train Loss mse: 0.0000, Train Loss ce: 0.0096, Train Steps/Sec: 0.27,
 [2026-01-30 18:53:02] (step=0000969) Train Loss mse: 0.0000, Train Loss ce: 0.0131, Train Steps/Sec: 0.31,
+[2026-01-30 18:53:05] (step=0000970) Train Loss mse: 0.0000, Train Loss ce: 0.0099, Train Steps/Sec: 0.30,
 base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step1000
 Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
 [eval debug] first 3 batch fingerprints:
@@ -1169,7 +1170,6 @@ Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_m
   fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
   fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
 ce_avg: 0.16064727306365967, mse_avg: 0.0
-[2026-01-30 18:53:05] (step=0000970) Train Loss mse: 0.0000, Train Loss ce: 0.0099, Train Steps/Sec: 0.30,
 [2026-01-30 18:53:09] (step=0000971) Train Loss mse: 0.0000, Train Loss ce: 0.0102, Train Steps/Sec: 0.28,
 [2026-01-30 18:53:12] (step=0000972) Train Loss mse: 0.0000, Train Loss ce: 0.0103, Train Steps/Sec: 0.30,
 [2026-01-30 18:53:16] (step=0000973) Train Loss mse: 0.0000, Train Loss ce: 0.0145, Train Steps/Sec: 0.29,
@@ -2600,27 +2600,6 @@ ce_avg: 0.16064727306365967, mse_avg: 0.0
 [2026-01-30 20:15:49] (step=0002398) Train Loss mse: 0.0000, Train Loss ce: 0.0071, Train Steps/Sec: 0.29,
 [2026-01-30 20:15:52] (step=0002399) Train Loss mse: 0.0000, Train Loss ce: 0.0055, Train Steps/Sec: 0.31,
 [2026-01-30 20:15:56] (step=0002400) Train Loss mse: 0.0000, Train Loss ce: 0.0069, Train Steps/Sec: 0.28,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step2500
-Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-ce_avg: 0.38690388202667236, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step3000
-Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-ce_avg: 0.47913289070129395, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step3500
-Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-ce_avg: 0.5199082493782043, mse_avg: 0.0
 [2026-01-30 20:15:59] (step=0002401) Train Loss mse: 0.0000, Train Loss ce: 0.0036, Train Steps/Sec: 0.29,
 [2026-01-30 20:16:03] (step=0002402) Train Loss mse: 0.0000, Train Loss ce: 0.0051, Train Steps/Sec: 0.27,
 [2026-01-30 20:16:06] (step=0002403) Train Loss mse: 0.0000, Train Loss ce: 0.0061, Train Steps/Sec: 0.30,
@@ -2681,6 +2660,27 @@ ce_avg: 0.5199082493782043, mse_avg: 0.0
 [2026-01-30 20:19:17] (step=0002458) Train Loss mse: 0.0000, Train Loss ce: 0.0087, Train Steps/Sec: 0.29,
 [2026-01-30 20:19:20] (step=0002459) Train Loss mse: 0.0000, Train Loss ce: 0.0069, Train Steps/Sec: 0.28,
 [2026-01-30 20:19:24] (step=0002460) Train Loss mse: 0.0000, Train Loss ce: 0.0052, Train Steps/Sec: 0.28,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step2500
+Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+ce_avg: 0.38690388202667236, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step3000
+Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+ce_avg: 0.47913289070129395, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step3500
+Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+ce_avg: 0.5199082493782043, mse_avg: 0.0
 [2026-01-30 20:19:27] (step=0002461) Train Loss mse: 0.0000, Train Loss ce: 0.0033, Train Steps/Sec: 0.30,
 [2026-01-30 20:19:31] (step=0002462) Train Loss mse: 0.0000, Train Loss ce: 0.0102, Train Steps/Sec: 0.28,
 [2026-01-30 20:19:34] (step=0002463) Train Loss mse: 0.0000, Train Loss ce: 0.0104, Train Steps/Sec: 0.29,
@@ -3679,27 +3679,6 @@ ce_avg: 0.5199082493782043, mse_avg: 0.0
 [2026-01-30 21:17:00] (step=0003456) Train Loss mse: 0.0000, Train Loss ce: 0.0121, Train Steps/Sec: 0.30,
 [2026-01-30 21:17:04] (step=0003457) Train Loss mse: 0.0000, Train Loss ce: 0.0080, Train Steps/Sec: 0.29,
 [2026-01-30 21:17:07] (step=0003458) Train Loss mse: 0.0000, Train Loss ce: 0.0110, Train Steps/Sec: 0.30,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step4000
-Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-ce_avg: 0.5366948843002319, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step4500
-Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-ce_avg: 0.5437333583831787, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step5000
-Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
-ce_avg: 0.5457080006599426, mse_avg: 0.0
 [2026-01-30 21:17:10] (step=0003459) Train Loss mse: 0.0000, Train Loss ce: 0.0090, Train Steps/Sec: 0.30,
 [2026-01-30 21:17:14] (step=0003460) Train Loss mse: 0.0000, Train Loss ce: 0.0042, Train Steps/Sec: 0.28,
 [2026-01-30 21:17:17] (step=0003461) Train Loss mse: 0.0000, Train Loss ce: 0.0094, Train Steps/Sec: 0.27,
@@ -3788,6 +3767,20 @@ ce_avg: 0.5457080006599426, mse_avg: 0.0
 [2026-01-30 21:22:15] (step=0003544) Train Loss mse: 0.0000, Train Loss ce: 0.0024, Train Steps/Sec: 0.31,
 [2026-01-30 21:22:18] (step=0003545) Train Loss mse: 0.0000, Train Loss ce: 0.0104, Train Steps/Sec: 0.30,
 [2026-01-30 21:22:22] (step=0003546) Train Loss mse: 0.0000, Train Loss ce: 0.0060, Train Steps/Sec: 0.29,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step4000
+Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+ce_avg: 0.5366948843002319, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step4500
+Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+ce_avg: 0.5437333583831787, mse_avg: 0.0
 [2026-01-30 21:22:25] (step=0003547) Train Loss mse: 0.0000, Train Loss ce: 0.0049, Train Steps/Sec: 0.30,
 [2026-01-30 21:22:28] (step=0003548) Train Loss mse: 0.0000, Train Loss ce: 0.0052, Train Steps/Sec: 0.30,
 [2026-01-30 21:22:32] (step=0003549) Train Loss mse: 0.0000, Train Loss ce: 0.0053, Train Steps/Sec: 0.26,
@@ -5245,4 +5238,11 @@ ce_avg: 0.5457080006599426, mse_avg: 0.0
 [2026-01-30 22:46:46] Saving checkpoint to /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/0005000.
 /opt/conda/lib/python3.11/site-packages/torch/distributed/fsdp/fully_sharded_data_parallel.py:690: FutureWarning: FSDP.state_dict_type() and FSDP.set_state_dict_type() are being deprecated. Please use APIs, get_state_dict() and set_state_dict(), which can support different parallelisms, FSDP1, FSDP2, DDP. API doc: https://pytorch.org/docs/stable/distributed.checkpoint.html#torch.distributed.checkpoint.state_dict.get_state_dict .Tutorial: https://pytorch.org/tutorials/recipes/distributed_checkpoint_recipe.html .
   warnings.warn(
-[2026-01-30 22:49:19] Done!
+[2026-01-30 22:49:19] Done!
+base_dir is /dev/shm/models/checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_mujoco_pick_and_place_one_image_lr2e_5_ce_no_mse_ins_step5000
+Preparing Dataset vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce/vlm_gym_mujoco_pick_and_place_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_mujoco_pick_and_place_celoss_no_mse_evalonce'}]
+ce_avg: 0.5457080006599426, mse_avg: 0.0