so101_teleop_vials_rack_left

This dataset is a TsFile conversion of the Hugging Face dataset sreetz-nv/so101_teleop_vials_rack_left. It contains the numeric LeRobot frame data for an SO-101 teleoperation task, stored as one Apache TsFile table.

Source Dataset

• Original dataset: sreetz-nv/so101_teleop_vials_rack_left
• LeRobot codebase version: v3.0
• Robot type: so101_follower
• Split: train (0:75)
• Scale: 75 episodes, 18,250 frames, 1 task
• Sampling rate: 30 fps
• Source modalities: tabular time-series data and video

Converted Files

• TsFile: data/so101_teleop_vials_rack_left.tsfile
• Table name: so101_teleop_vials_rack_left
• Rows: 18,250
• Time precision: milliseconds
• Metadata: meta/ mirrors the source metadata, with meta/info.json updated to describe the TsFile conversion.

Schema

Column	Role	Type	Notes
Time	TIME	INT64	round(timestamp * 1000) in milliseconds; restarts per episode.
episode_index	TAG	INT64	Original source episode index.
task_index	TAG	INT64	Original source task index.
frame_index	FIELD	INT64	Original source frame index.
sample_index	FIELD	INT64	Renamed from source column index.
observation_state_0 ... observation_state_5	FIELD	FLOAT	Flattened from source vector column observation.state.
action_0 ... action_5	FIELD	FLOAT	Flattened from source vector column action.

Conversion Notes

• The source timestamp column is used to create Time and is not retained as a separate field, because it equals Time / 1000 seconds.
• Vector columns are flattened by preserving the source column prefix, replacing . with _, and appending element indexes.
• Source video features observation.images.ego and observation.images.external_D455 are not included in this converted repository. Videos remain in the original dataset under videos/.
• The original source data path is data/chunk-{chunk_index:03d}/file-{file_index:03d}.parquet; the converted data path is data/so101_teleop_vials_rack_left.tsfile.

Usage

from tsfile import TsFileReader

path = "data/so101_teleop_vials_rack_left.tsfile"
reader = TsFileReader(path)
print(reader.get_all_table_schemas())