THULab/anvilbot-patrickhhh_SO101_record_test

SO101 Record Test TsFile

Apache TsFile edition of anvilbot-patrickhhh/SO101_record_test, a LeRobot v2.1 SO101 dataset for putting a green cube in the area marked with black tape. The numeric trajectories are stored in one table-model TsFile.

Source and attribution

• Original repository owner and uploader: anvilbot-patrickhhh. The source does not provide a personal author name.
• License: Apache-2.0.
• The source card does not provide a homepage, paper, or completed BibTeX citation.
• Split: train; 100 episodes; 30,169 frame rows; one task; 30 FPS; 100 source Parquet shards.
• Task 0: Put the green cube in the area with the black tape.

Data layout

The table is anvilbot_patrickhhh_so101_record_test and contains 30,169 rows across 100 TAG devices. The source Parquet shards total 1,663,535 bytes; the TsFile is 770,267 bytes (46.3% of the source Parquet size).

Column	TsFile role	Type	Meaning
Time	TIME	INT64 milliseconds	round(timestamp * 1000), restarting at zero per episode
episode_index	TAG	STRING from source INT64	Source episode identity
task_index	TAG	STRING from source INT64	Source task identity
frame_index	FIELD	INT64	Frame position within the episode
sample_index	FIELD	INT64	Source index, renamed for clarity
action_0 ... action_5	FIELD	FLOAT	Flattened action[6]
observation_state_0 ... observation_state_5	FIELD	FLOAT	Flattened observation.state[6]

timestamp is not retained as a separate FIELD because it is represented by Time / 1000 seconds. The vector column names preserve their source prefixes, with dots changed to underscores. No trajectory row, episode, task, action dimension, or state dimension is removed.

Videos and alignment

The 100 source AV1 videos are not included here. They remain under videos/chunk-000 in the original repository:

• observation.images.front: 100 per-episode MP4 files, 1280x720 at 30 FPS. Hugging Face displays the directory size as about 587 MB.

The source template is videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4. Use episode_index and frame_index to align each numeric row with the 30 FPS video stream.

Read example

from tsfile import TsFileReader

reader = TsFileReader("data/anvilbot_patrickhhh_so101_record_test.tsfile")
with reader.query_table(
    "anvilbot_patrickhhh_so101_record_test",
    ["episode_index", "task_index", "frame_index", "sample_index", "action_0", "observation_state_0"],
    batch_size=1024,
) as result:
    batch = result.read_arrow_batch()
    print(batch.to_pandas().head())
reader.close()