admin/lerobot
1
0
Fork 0
mirror of https://github.com/huggingface/lerobot.git synced 2026-07-24 10:16:09 +00:00
Code Issues Packages Projects Releases Wiki Activity

feat(torchcodec): setting torchcodec as default as the new official release supports audio decoding

Browse Source
This commit is contained in:
CarolinePascal
2025-04-24 11:52:53 +02:00
parent 6d726266fd
commit bff91f9927
2 changed files with 6 additions and 7 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/lerobot/datasets/audio_utils.py
+3 -4
View File
@@ -41,7 +41,7 @@ def decode_audio(
audio_path: Path | str, audio_path: Path | str,
timestamps: list[float], timestamps: list[float],
duration: float, duration: float,
backend: str | None = "torchaudio", backend: str | None = "torchcodec",
) -> torch.Tensor: ) -> torch.Tensor:
""" """
Decodes audio using the specified backend. Decodes audio using the specified backend.
@@ -49,7 +49,7 @@ def decode_audio(
audio_path (Path): Path to the audio file. audio_path (Path): Path to the audio file.
timestamps (list[float]): List of (starting) timestamps to extract audio chunks. timestamps (list[float]): List of (starting) timestamps to extract audio chunks.
duration (float): Duration of the audio chunks in seconds. duration (float): Duration of the audio chunks in seconds.
backend (str, optional): Backend to use for decoding. Defaults to "torchaudio". backend (str, optional): Backend to use for decoding. Defaults to "torchcodec".
Returns: Returns:
torch.Tensor: Decoded audio chunks. torch.Tensor: Decoded audio chunks.
@@ -57,8 +57,7 @@ def decode_audio(
Currently supports torchaudio. Currently supports torchaudio.
""" """
if backend == "torchcodec": if backend == "torchcodec":
# return decode_audio_torchcodec(audio_path, timestamps, duration) #TODO(CarolinePascal): uncomment this line at next torchcodec release return decode_audio_torchcodec(audio_path, timestamps, duration)
raise ValueError("torchcodec backend is not available yet.")
elif backend == "torchaudio": elif backend == "torchaudio":
return decode_audio_torchaudio(audio_path, timestamps, duration) return decode_audio_torchaudio(audio_path, timestamps, duration)
else: else:
src/lerobot/datasets/lerobot_dataset.py
+3 -3
View File
@@ -757,7 +757,7 @@ class LeRobotDataset(torch.utils.data.Dataset):
download_audio (bool, optional): Flag to download the audio. Defaults to True. download_audio (bool, optional): Flag to download the audio. Defaults to True.
video_backend (str | None, optional): Video backend to use for decoding videos. Defaults to torchcodec when available int the platform; otherwise, defaults to 'pyav'. video_backend (str | None, optional): Video backend to use for decoding videos. Defaults to torchcodec when available int the platform; otherwise, defaults to 'pyav'.
You can also use the 'pyav' decoder used by Torchvision, which used to be the default option, or 'video_reader' which is another decoder of Torchvision. You can also use the 'pyav' decoder used by Torchvision, which used to be the default option, or 'video_reader' which is another decoder of Torchvision.
audio_backend (str | None, optional): Audio backend to use for decoding audio. Defaults to 'torchaudio'. audio_backend (str | None, optional): Audio backend to use for decoding audio. Defaults to 'torchcodec'.
batch_encoding_size (int, optional): Number of episodes to accumulate before batch encoding videos. batch_encoding_size (int, optional): Number of episodes to accumulate before batch encoding videos.
Set to 1 for immediate encoding (default), or higher for batched encoding. Defaults to 1. Set to 1 for immediate encoding (default), or higher for batched encoding. Defaults to 1.
vcodec (str, optional): Video codec for encoding videos during recording. Options: 'h264', 'hevc', vcodec (str, optional): Video codec for encoding videos during recording. Options: 'h264', 'hevc',
@@ -776,7 +776,7 @@ class LeRobotDataset(torch.utils.data.Dataset):
self.revision = revision if revision else CODEBASE_VERSION self.revision = revision if revision else CODEBASE_VERSION
self.video_backend = video_backend if video_backend else get_safe_default_codec() self.video_backend = video_backend if video_backend else get_safe_default_codec()
self.audio_backend = ( self.audio_backend = (
audio_backend if audio_backend else "torchaudio" audio_backend if audio_backend else "torchcodec"
) # Waiting for torchcodec release #TODO(CarolinePascal) ) # Waiting for torchcodec release #TODO(CarolinePascal)
self.delta_indices = None self.delta_indices = None
self.batch_encoding_size = batch_encoding_size self.batch_encoding_size = batch_encoding_size
@@ -1950,7 +1950,7 @@ class LeRobotDataset(torch.utils.data.Dataset):
obj._recorded_frames = 0 obj._recorded_frames = 0
obj._writer_closed_for_reading = False obj._writer_closed_for_reading = False
obj.audio_backend = ( obj.audio_backend = (
audio_backend if audio_backend is not None else "torchaudio" audio_backend if audio_backend is not None else "torchcodec"
) # Waiting for torchcodec release #TODO(CarolinePascal) ) # Waiting for torchcodec release #TODO(CarolinePascal)
return obj return obj