THULab/so100_set_screw

SO100 Set Screw (TsFile)

Source: mtitg/so100_set_screw, revision d0c82c9b1ceb444baf17564d5455733e91ce01a8.

This is a numeric time-series conversion of the LeRobot dataset for task Grab the screw and set it in the hole.. The source records an SO100 robot with six-axis state and action vectors.

• Modalities: Time-series
• Split: train
• Sampling rate: 30 fps
• Scale: 52 episodes, 42,264 frames, 52 source episode Parquet files
• Converted layout: 1 TsFile with 42,264 rows

TsFile schema

The table is so100_set_screw_train.

Role	Columns	Representation
Time	Time	INT64 milliseconds; round(timestamp * 1000)
TAG	episode_index, task_index	Source episode and task dimensions
FIELD	frame_index, sample_index	Source scalar frame identifiers
FIELD	observation_state_0 ... _5	6 FLOAT robot-state values
FIELD	action_0 ... _5	6 FLOAT action values

Conversion notes

• The source timestamp is omitted because it is represented by Time in milliseconds.
• The source index is renamed to sample_index.
• Every numeric vector element and all 42,264 source rows are retained.
• Source metadata is mirrored under meta/; its data_path points to the TsFile and records the conversion mapping.
• Videos are not downloaded or uploaded. They remain in the original dataset videos, where they preserve frame alignment with the numeric rows.

Reading

from tsfile import TsFileReader

path = "data/so100_set_screw_train.tsfile"
reader = TsFileReader(path)
table = "so100_set_screw_train"
columns = ["episode_index", "task_index", "frame_index", "sample_index", "action_0"]
with reader.query_table(table, columns, batch_size=1024) as result:
    batch = result.read_arrow_batch()
    print(batch.to_pandas().head())

Source & license

The source card states that the dataset was created with LeRobot and is licensed under Apache-2.0. The source does not provide a paper or completed citation.