admin/lerobot
1
0
Fork 0
mirror of https://github.com/huggingface/lerobot.git synced 2026-07-24 18:26:11 +00:00
Code Issues Packages Projects Releases Wiki Activity

fix(update video info): ditching the differentiated approahces for video info update - video info are always updated unless for preserved keys.

Browse Source
This commit is contained in:
CarolinePascal
2026-06-16 16:09:20 +02:00
parent 2c96703505
commit 95c01a6b12
1 changed files with 6 additions and 16 deletions
Download Patch File Download Diff File Expand all files Collapse all files
src/lerobot/datasets/dataset_metadata.py
+6 -16
View File
@@ -608,13 +608,10 @@ class LeRobotDatasetMetadata:
Warning: this function writes info from first episode videos, implicitly assuming that all videos have Warning: this function writes info from first episode videos, implicitly assuming that all videos have
been encoded the same way. Also, this means it assumes the first episode exists. been encoded the same way. Also, this means it assumes the first episode exists.
Two modes, selected by ``preserve_keys``: Always re-probes the videos and overwrites existing info for every recomputed
key. ``preserve_keys`` lists keys whose existing values must be kept (e.g.
- **Populate** (``None``, default): write info for video keys that lack it, data-intrinsic entries like ``is_depth_map`` and depth quantization params)
skip the rest. Used when first encoding a dataset. instead of being recomputed.
- **Refresh** (any iterable): re-probe and overwrite existing info, keeping
the listed keys. Used after re-encoding to preserve data-intrinsic entries
(``is_depth_map``, depth quantization params) while codec params change.
Args: Args:
video_key: If provided, only update this video key. Otherwise update video_key: If provided, only update this video key. Otherwise update
@@ -623,23 +620,16 @@ class LeRobotDatasetMetadata:
videos. When provided, its fields are recorded as videos. When provided, its fields are recorded as
``video.<field>`` entries alongside the stream-derived ``video.<field>`` entries alongside the stream-derived
``video.*`` entries (see :func:`get_video_info`). ``video.*`` entries (see :func:`get_video_info`).
preserve_keys: ``None`` (default) for populate-once mode. An iterable preserve_keys: Keys whose existing values are kept instead of being
(possibly empty) switches to refresh mode, keeping these keys' recomputed. ``None`` (default) recomputes every key.
existing values while recomputing the rest.
""" """
if video_key is not None and video_key not in self.video_keys: if video_key is not None and video_key not in self.video_keys:
raise ValueError(f"Video key {video_key} not found in dataset") raise ValueError(f"Video key {video_key} not found in dataset")
video_keys = [video_key] if video_key is not None else self.video_keys video_keys = [video_key] if video_key is not None else self.video_keys
refresh = preserve_keys is not None
preserve_set = set(preserve_keys or ()) preserve_set = set(preserve_keys or ())
for key in video_keys: for key in video_keys:
existing = self.features[key].get("info") or {} existing = self.features[key].get("info") or {}
# ``is_depth_map`` is written at feature creation and does not count as real video info here.
already_populated = bool(set(existing.keys()) - {"is_depth_map"})
# Populate-once: never clobber info that has already been written unless a refresh is requested.
if already_populated and not refresh:
continue
video_path = self.root / self.video_path.format(video_key=key, chunk_index=0, file_index=0) video_path = self.root / self.video_path.format(video_key=key, chunk_index=0, file_index=0)
new_info = get_video_info(video_path, video_encoder=video_encoder) new_info = get_video_info(video_path, video_encoder=video_encoder)
# Drop preserved keys so the existing values win on merge. # Drop preserved keys so the existing values win on merge.