diff --git a/src/lerobot/datasets/utils.py b/src/lerobot/datasets/utils.py index c75aef4ef..34197fa6e 100644 --- a/src/lerobot/datasets/utils.py +++ b/src/lerobot/datasets/utils.py @@ -417,6 +417,11 @@ def load_image_as_numpy( def load_audio_from_path(fpath: str | Path) -> np.ndarray: audio_data, _ = read(fpath, dtype="float32") + + # Fill missing channel dimension when loading mono audio data + if audio_data.ndim == 1: + audio_data = np.expand_dims(audio_data, axis=1) + return audio_data