Upload checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins

checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/wandb/offline-run-20260125_150523-checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins-run0/files/output.log

@@ -1,3 +1,173 @@
+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_colorization_celoss_no_mse/vlm_gym_colorization_train
+base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step0
+Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+ce_avg: 0.8449353575706482, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step500
+Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+ce_avg: 0.2859387993812561, 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/
@@ -1094,197 +1264,6 @@ wandb: For more information, check out the docs at: https://weave-docs.wandb.ai/
 [2026-01-25 16:03:40] (step=0001083) Train Loss mse: 0.0000, Train Loss ce: 0.2555, Train Steps/Sec: 0.39,
 [2026-01-25 16:03:42] (step=0001084) Train Loss mse: 0.0000, Train Loss ce: 0.2940, Train Steps/Sec: 0.42,
 [2026-01-25 16:03:45] (step=0001085) Train Loss mse: 0.0000, Train Loss ce: 0.2703, Train Steps/Sec: 0.33,
-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_colorization_celoss_no_mse/vlm_gym_colorization_train
-base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step0
-Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-ce_avg: 0.8449353575706482, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step500
-Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-ce_avg: 0.2859387993812561, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step1000
-Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-ce_avg: 0.46788886189460754, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step1500
-Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-ce_avg: 0.6231057643890381, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step2000
-Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-ce_avg: 0.6934272646903992, mse_avg: 0.0
 [2026-01-25 16:03:48] (step=0001086) Train Loss mse: 0.0000, Train Loss ce: 0.2461, Train Steps/Sec: 0.38,
 [2026-01-25 16:03:51] (step=0001087) Train Loss mse: 0.0000, Train Loss ce: 0.2672, Train Steps/Sec: 0.40,
 [2026-01-25 16:03:53] (step=0001088) Train Loss mse: 0.0000, Train Loss ce: 0.2795, Train Steps/Sec: 0.34,
@@ -1316,6 +1295,27 @@ ce_avg: 0.6934272646903992, mse_avg: 0.0
 [2026-01-25 16:05:05] (step=0001114) Train Loss mse: 0.0000, Train Loss ce: 0.2708, Train Steps/Sec: 0.40,
 [2026-01-25 16:05:08] (step=0001115) Train Loss mse: 0.0000, Train Loss ce: 0.2779, Train Steps/Sec: 0.37,
 [2026-01-25 16:05:11] (step=0001116) Train Loss mse: 0.0000, Train Loss ce: 0.2583, Train Steps/Sec: 0.39,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step1000
+Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+ce_avg: 0.46788886189460754, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step1500
+Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+ce_avg: 0.6231057643890381, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step2000
+Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+ce_avg: 0.6934272646903992, mse_avg: 0.0
 [2026-01-25 16:05:14] (step=0001117) Train Loss mse: 0.0000, Train Loss ce: 0.2621, Train Steps/Sec: 0.33,
 [2026-01-25 16:05:16] (step=0001118) Train Loss mse: 0.0000, Train Loss ce: 0.2592, Train Steps/Sec: 0.35,
 [2026-01-25 16:05:19] (step=0001119) Train Loss mse: 0.0000, Train Loss ce: 0.2544, Train Steps/Sec: 0.35,
@@ -2782,27 +2782,6 @@ ce_avg: 0.6934272646903992, mse_avg: 0.0
 [2026-01-25 17:13:45] (step=0002580) Train Loss mse: 0.0000, Train Loss ce: 0.2436, Train Steps/Sec: 0.31,
 [2026-01-25 17:13:48] (step=0002581) Train Loss mse: 0.0000, Train Loss ce: 0.2514, Train Steps/Sec: 0.42,
 [2026-01-25 17:13:51] (step=0002582) Train Loss mse: 0.0000, Train Loss ce: 0.2485, Train Steps/Sec: 0.29,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step2500
-Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-ce_avg: 0.719638466835022, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step3000
-Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-ce_avg: 0.7213400602340698, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step3500
-Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-ce_avg: 0.6960676312446594, mse_avg: 0.0
 [2026-01-25 17:13:54] (step=0002583) Train Loss mse: 0.0000, Train Loss ce: 0.2590, Train Steps/Sec: 0.37,
 [2026-01-25 17:13:56] (step=0002584) Train Loss mse: 0.0000, Train Loss ce: 0.2530, Train Steps/Sec: 0.40,
 [2026-01-25 17:14:00] (step=0002585) Train Loss mse: 0.0000, Train Loss ce: 0.2502, Train Steps/Sec: 0.32,
@@ -2842,6 +2821,27 @@ ce_avg: 0.6960676312446594, mse_avg: 0.0
 [2026-01-25 17:15:34] (step=0002619) Train Loss mse: 0.0000, Train Loss ce: 0.2615, Train Steps/Sec: 0.39,
 [2026-01-25 17:15:37] (step=0002620) Train Loss mse: 0.0000, Train Loss ce: 0.2592, Train Steps/Sec: 0.34,
 [2026-01-25 17:15:39] (step=0002621) Train Loss mse: 0.0000, Train Loss ce: 0.2630, Train Steps/Sec: 0.38,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step2500
+Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+ce_avg: 0.719638466835022, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step3000
+Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+ce_avg: 0.7213400602340698, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step3500
+Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+ce_avg: 0.6960676312446594, mse_avg: 0.0
 [2026-01-25 17:15:42] (step=0002622) Train Loss mse: 0.0000, Train Loss ce: 0.2424, Train Steps/Sec: 0.38,
 [2026-01-25 17:15:45] (step=0002623) Train Loss mse: 0.0000, Train Loss ce: 0.2457, Train Steps/Sec: 0.37,
 [2026-01-25 17:15:48] (step=0002624) Train Loss mse: 0.0000, Train Loss ce: 0.2660, Train Steps/Sec: 0.36,
@@ -3800,20 +3800,6 @@ ce_avg: 0.6960676312446594, mse_avg: 0.0
 [2026-01-25 18:00:35] (step=0003577) Train Loss mse: 0.0000, Train Loss ce: 0.2664, Train Steps/Sec: 0.29,
 [2026-01-25 18:00:38] (step=0003578) Train Loss mse: 0.0000, Train Loss ce: 0.2234, Train Steps/Sec: 0.35,
 [2026-01-25 18:00:41] (step=0003579) Train Loss mse: 0.0000, Train Loss ce: 0.2551, Train Steps/Sec: 0.36,
-base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step4000
-Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-ce_avg: 0.7003546953201294, mse_avg: 0.0
-base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step4500
-Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
-[eval debug] first 3 batch fingerprints:
-  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
-ce_avg: 0.6831608414649963, mse_avg: 0.0
 [2026-01-25 18:00:44] (step=0003580) Train Loss mse: 0.0000, Train Loss ce: 0.2514, Train Steps/Sec: 0.37,
 [2026-01-25 18:00:46] (step=0003581) Train Loss mse: 0.0000, Train Loss ce: 0.2396, Train Steps/Sec: 0.39,
 [2026-01-25 18:00:49] (step=0003582) Train Loss mse: 0.0000, Train Loss ce: 0.2210, Train Steps/Sec: 0.40,
@@ -3880,6 +3866,20 @@ ce_avg: 0.6831608414649963, mse_avg: 0.0
 [2026-01-25 18:03:40] (step=0003643) Train Loss mse: 0.0000, Train Loss ce: 0.2529, Train Steps/Sec: 0.37,
 [2026-01-25 18:03:43] (step=0003644) Train Loss mse: 0.0000, Train Loss ce: 0.2446, Train Steps/Sec: 0.35,
 [2026-01-25 18:03:46] (step=0003645) Train Loss mse: 0.0000, Train Loss ce: 0.2420, Train Steps/Sec: 0.35,
+base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step4000
+Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+ce_avg: 0.7003546953201294, mse_avg: 0.0
+base_dir is /dev/shm/models/checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins/eval_used_rows, step_tag is checkpoints_vlm_gym_colorization_one_image_lr2e_5_ce_no_mse_ins_step4500
+Preparing Dataset vlm_gym_colorization_celoss_no_mse_evalonce/vlm_gym_colorization_val
+[eval debug] first 3 batch fingerprints:
+  fp[0]: [{'data_indexes': [0], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[1]: [{'data_indexes': [8], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+  fp[2]: [{'data_indexes': [16], 'worker_id': 0, 'dataset_name': 'vlm_gym_colorization_celoss_no_mse_evalonce'}]
+ce_avg: 0.6831608414649963, mse_avg: 0.0
 [2026-01-25 18:03:48] (step=0003646) Train Loss mse: 0.0000, Train Loss ce: 0.2638, Train Steps/Sec: 0.40,
 [2026-01-25 18:03:51] (step=0003647) Train Loss mse: 0.0000, Train Loss ce: 0.2499, Train Steps/Sec: 0.38,
 [2026-01-25 18:03:54] (step=0003648) Train Loss mse: 0.0000, Train Loss ce: 0.2567, Train Steps/Sec: 0.36,