fix

src/lerobot/policies/rlearn/configuration_rlearn.py

@@ -34,7 +34,7 @@ class RLearNConfig(PreTrainedConfig):
 
     Notes:
       - This follows the ReWiND paper architecture. It uses frozen vision/text encoders
-        (DINO v3 for vision, sentence-transformers for language) and trains a 
+        (DINO v3 for vision, sentence-transformers for language) and trains a
         lightweight temporal aggregator + head.
     """
 
@@ -61,15 +61,17 @@ class RLearNConfig(PreTrainedConfig):
     use_tanh_head: bool = False  # when True, bound outputs in [-1, 1]
 
     # Training
-    learning_rate: float = 1e-4
+    learning_rate: float = 3e-5
     weight_decay: float = 0.01
 
     # ReWiND-specific parameters
-    use_video_rewind: bool = True  # Enable video rewinding augmentation
+    use_video_rewind: bool = False  # Enable video rewinding augmentation
     rewind_prob: float = 0.8  # Probability of applying rewind to each sample (paper: ~80%)
     rewind_last3_prob: float = 0.1  # Of the rewinds, 10% only rewind the last 3 frames
-    use_mismatch_loss: bool = True  # Enable mismatched language-video loss
-    mismatch_prob: float = 0.2  # Probability to include a mismatched video-language forward pass (paper: ~20%)
+    use_mismatch_loss: bool = False  # Enable mismatched language-video loss
+    mismatch_prob: float = (
+        0.2  # Probability to include a mismatched video-language forward pass (paper: ~20%)
+    )
 
     # Loss hyperparameters (simplified for ReWiND)
     # The main loss is just MSE between predicted and target progress
@@ -85,7 +87,7 @@ class RLearNConfig(PreTrainedConfig):
     # Architectural knobs to better mirror ReWiND
     num_register_tokens: int = 4
     mlp_predictor_depth: int = 3  # depth of the per-frame MLP head
-    
+
     # HLGauss loss parameters
     use_hl_gauss_loss: bool = True
     reward_min_value: float = 0.0