chore: enable simplify in ruff lint (#2085)

src/lerobot/datasets/video_utils.py

@@ -437,7 +437,9 @@ def concatenate_video_files(
         tmp_concatenate_path, mode="r", format="concat", options={"safe": "0"}
     )  # safe = 0 allows absolute paths as well as relative paths
 
-    tmp_output_video_path = tempfile.NamedTemporaryFile(suffix=".mp4", delete=False).name
+    with tempfile.NamedTemporaryFile(suffix=".mp4", delete=False) as tmp_named_file:
+        tmp_output_video_path = tmp_named_file.name
+
     output_container = av.open(
         tmp_output_video_path, mode="w", options={"movflags": "faststart"}
     )  # faststart is to move the metadata to the beginning of the file to speed up loading

src/lerobot/policies/act/modeling_act.py

@@ -398,10 +398,7 @@ class ACT(nn.Module):
                 "actions must be provided when using the variational objective in training mode."
             )
 
-        if OBS_IMAGES in batch:
-            batch_size = batch[OBS_IMAGES][0].shape[0]
-        else:
-            batch_size = batch[OBS_ENV_STATE].shape[0]
+        batch_size = batch[OBS_IMAGES][0].shape[0] if OBS_IMAGES in batch else batch[OBS_ENV_STATE].shape[0]
 
         # Prepare the latent for input to the transformer encoder.
         if self.config.use_vae and ACTION in batch and self.training:

src/lerobot/processor/hil_processor.py

@@ -340,7 +340,7 @@ class GripperPenaltyProcessorStep(ComplementaryDataProcessorStep):
         """
         action = self.transition.get(TransitionKey.ACTION)
 
-        raw_joint_positions = complementary_data.get("raw_joint_positions", None)
+        raw_joint_positions = complementary_data.get("raw_joint_positions")
         if raw_joint_positions is None:
             return complementary_data
 

src/lerobot/processor/normalize_processor.py

@@ -119,13 +119,12 @@ class _NormalizationMixin:
                     )
                 self.features = reconstructed
 
-        if self.norm_map:
-            # if keys are strings (JSON), rebuild enum map
-            if all(isinstance(k, str) for k in self.norm_map.keys()):
-                reconstructed = {}
-                for ft_type_str, norm_mode_str in self.norm_map.items():
-                    reconstructed[FeatureType(ft_type_str)] = NormalizationMode(norm_mode_str)
-                self.norm_map = reconstructed
+        # if keys are strings (JSON), rebuild enum map
+        if self.norm_map and all(isinstance(k, str) for k in self.norm_map):
+            reconstructed = {}
+            for ft_type_str, norm_mode_str in self.norm_map.items():
+                reconstructed[FeatureType(ft_type_str)] = NormalizationMode(norm_mode_str)
+            self.norm_map = reconstructed
 
         # Convert stats to tensors and move to the target device once during initialization.
         self.stats = self.stats or {}

src/lerobot/processor/observation_processor.py

@@ -152,7 +152,7 @@ class VanillaObservationProcessorStep(ObservationProcessorStep):
         """
         # Build a new features mapping keyed by the same FeatureType buckets
         # We assume callers already placed features in the correct FeatureType.
-        new_features: dict[PipelineFeatureType, dict[str, PolicyFeature]] = {ft: {} for ft in features.keys()}
+        new_features: dict[PipelineFeatureType, dict[str, PolicyFeature]] = {ft: {} for ft in features}
 
         exact_pairs = {
             "pixels": OBS_IMAGE,

src/lerobot/utils/visualization_utils.py

@@ -32,11 +32,8 @@ def init_rerun(session_name: str = "lerobot_control_loop") -> None:
 
 
 def _is_scalar(x):
-    return (
-        isinstance(x, float)
-        or isinstance(x, numbers.Real)
-        or isinstance(x, (np.integer | np.floating))
-        or (isinstance(x, np.ndarray) and x.ndim == 0)
+    return isinstance(x, (float | numbers.Real | np.integer | np.floating)) or (
+        isinstance(x, np.ndarray) and x.ndim == 0
     )