diff --git a/src/lerobot/scripts/lerobot_train.py b/src/lerobot/scripts/lerobot_train.py
index 08d2a6ac8..23bd822e2 100644
--- a/src/lerobot/scripts/lerobot_train.py
+++ b/src/lerobot/scripts/lerobot_train.py
@@ -254,16 +254,20 @@ def train(cfg: TrainPipelineConfig, accelerator: Accelerator | None = None):
         from lerobot.processor.delta_action_processor import to_delta_actions
 
         max_samples = min(100_000, len(dataset))
-        indices = np.random.choice(len(dataset), max_samples, replace=False)
+        indices = np.random.choice(len(dataset), max_samples, replace=False).tolist()
         logging.info(
             f"use_delta_actions is enabled — computing delta action stats from {max_samples} dataset chunks"
         )
 
+        # Read only action and state from parquet (no video decoding)
+        hf = dataset.hf_dataset
+        actions_raw = hf.select(indices)["action"]
+        states_raw = hf.select(indices)["observation.state"]
+
         all_delta_actions = []
-        for i in indices:
-            item = dataset[int(i)]
-            action = item["action"]
-            state = item["observation.state"]
+        for action, state in zip(actions_raw, states_raw):
+            action = torch.as_tensor(action).float()
+            state = torch.as_tensor(state).float()
             if action.ndim == 1:
                 action = action.unsqueeze(0)
             mask = [True] * action.shape[-1]