Episodes Preview SO-100 Visualizer
100 episodes · 30 fps

0605 Pi0 100 TsFile

Apache TsFile conversion of weiye11/0605_pi0_100, a LeRobot v2.1 SO-100 manipulation dataset.

Source Dataset

  • Original dataset: weiye11/0605_pi0_100
  • Original author and repository uploader: weiye11
  • License: Apache-2.0
  • Robot type: so100
  • Task: Grasp a lego block and put it in the bin.
  • Split: train; sampling rate: 30 fps
  • Scale: 100 episodes, 24,804 frame rows, 1 task, 100 source Parquet shards
  • Source frame layout: data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet

Data Layout

  • TsFile: data/weiye11_0605_pi0_100.tsfile
  • Table: weiye11_0605_pi0_100
  • Rows: 24,804; episodes/devices: 100
  • Time precision: milliseconds
  • meta/ retains source metadata, with meta/info.json describing the TsFile data path and schema.

Schema

Column TsFile role Type Meaning
Time TIME INT64 round(timestamp * 1000) milliseconds
episode_index TAG STRING tag (source INT64) LeRobot episode index
task_index TAG STRING tag (source INT64) LeRobot task index
frame_index FIELD INT64 Frame number within an episode
sample_index FIELD INT64 Source index, renamed
action_0 ... action_5 FIELD FLOAT Flattened action[6], source order
observation_state_0 ... observation_state_5 FIELD FLOAT Flattened observation.state[6], source order

Time restarts for each episode and is monotonic under its (episode_index, task_index) TAG device. Source timestamp is omitted because it equals Time / 1000. No source rows or numeric action/state dimensions are omitted; index is preserved as sample_index.

Storage

  • Time: TS_2DIFF + LZ4
  • FLOAT/DOUBLE fields: GORILLA + LZ4
  • INT32/INT64 fields: TS_2DIFF + LZ4
  • BOOLEAN fields: RLE + LZ4 when present; this source has none
  • Episode and task identifiers: TsFile table-model TAG/device mechanism

The TsFile is 36.36% of the combined source Parquet size.

Videos

The original dataset has 200 episode-aligned MP4 files in two camera streams, 100 per stream:

Source video template: videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4. Videos are not included here; retrieve them from the original dataset. Numeric rows remain frame-aligned through episode_index and frame_index.

Minimal Read Example

from tsfile import TsFileReader

reader = TsFileReader("data/weiye11_0605_pi0_100.tsfile")
with reader.query_table("weiye11_0605_pi0_100", ["episode_index", "task_index", "frame_index", "action_0", "observation_state_0"], batch_size=65536) as result:
    print(result.read_arrow_batch().to_pandas().head())
reader.close()

Citation

The source card provides no paper or BibTeX citation. Please cite weiye11/0605_pi0_100 and weiye11.

Downloads last month
83