THULab/RickRain_so101_test_single_lego

SO101 Test Single LEGO TsFile

Apache TsFile edition of RickRain/so101_test_single_lego, a LeRobot v2.1 SO101 robot dataset for "Grasp a lego block and put it in the bin." Numeric trajectories are stored in one table-model TsFile.

Source And Attribution

• Original author and repository owner: RickRain
• License: Apache-2.0
• Task: Grasp a lego block and put it in the bin.
• Split: train; 95 episodes, 25,330 frames, 1 task, sampled at 30 FPS, from 95 source Parquet shards.
• The source card does not provide a homepage, paper, or completed BibTeX citation.

TsFile Output

The file data/rickrain_so101_test_single_lego.tsfile contains 25,330 rows across 95 episode/task devices. Time is round(timestamp * 1000) in milliseconds and restarts at zero for each episode. The source timestamp field is omitted because it is equivalent to Time / 1000 seconds.

Column	Role	Type	Source mapping
Time	TIME	INT64 milliseconds	round(timestamp * 1000)
episode_index, task_index	TAG	STRING	Original INT64 identifiers
frame_index	FIELD	INT64	Preserved frame position
sample_index	FIELD	INT64	Renamed from source index
action_0 ... action_5	FIELD	FLOAT	Flattened from action[6]
observation_state_0 ... observation_state_5	FIELD	FLOAT	Flattened from observation.state[6]

Vector prefixes are preserved and dots are replaced with underscores. The six dimensions retain the source order: main_shoulder_pan, main_shoulder_lift, main_elbow_flex, main_wrist_flex, main_wrist_roll, main_gripper. No trajectory rows, episodes, tasks, action dimensions, or state dimensions are removed.

Storage Profile

• FLOAT and DOUBLE: GORILLA + LZ4
• INT32 and INT64: TS_2DIFF + LZ4
• Time: TS_2DIFF + LZ4
• BOOLEAN, when present: RLE + LZ4
• TAG values: TsFile table/device TAG mechanism

Source Videos

Videos are not included in this repository. They remain in the original Hugging Face dataset under videos/chunk-000. The two source streams are observation.images.wrist and observation.images.webcam, with 95 MP4 files per stream. 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 corresponding 30 FPS frames.

Read Example

from tsfile import TsFileReader

reader = TsFileReader("data/rickrain_so101_test_single_lego.tsfile")
with reader.query_table(
    "rickrain_so101_test_single_lego",
    ["episode_index", "task_index", "frame_index", "sample_index", "action_0", "observation_state_0"],
    batch_size=1024,
) as result:
    print(result.read_arrow_batch().to_pandas().head())
reader.close()