azithromycin/Haribot099_so101_60

Haribot099 SO101 60 TsFile

This repository contains an Apache TsFile representation of Haribot099/so101_60, a LeRobot v2.1 SO101 dataset for the task Grasp a Lego block and put it in the bin.

It contains numeric robot states, actions, frame timing, and episode/task tags. The camera streams remain in the original Hugging Face dataset.

Source Dataset and Attribution

• Original dataset: Haribot099/so101_60
• Original repository creator and uploader: Perla Harish (Haribot099)
• License: Apache-2.0
• Robot type: so101
• LeRobot codebase version: v2.1
• Task: Grasp a lego block and put it in the bin. (task_index = 0)
• Split: train (0:60)
• Sampling rate: 30 fps
• Scale: 60 episodes, 27,418 frame rows, 1 task, 60 source Parquet shards, 120 source videos
• Source frame layout: data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet
• Source video layout: videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4

The source dataset card provides no paper or completed BibTeX citation.

TsFile Contents

• Path: data/haribot099_so101_60.tsfile
• Table: haribot099_so101_60
• Rows: 27,418
• Episode/task devices: 60
• Time precision: milliseconds
• Numeric source Parquet size: 1,449,872 bytes
• TsFile size: 463,488 bytes
• TsFile/Parquet size ratio: 0.3197

Schema

Time is an INT64 millisecond timestamp computed as round(timestamp * 1000). It restarts at zero for every episode.

TAG columns use the TsFile table-model device/tag mechanism:

• episode_index
• task_index

FIELD columns:

• frame_index
• sample_index
• action_0
• action_1
• action_2
• action_3
• action_4
• action_5
• observation_state_0
• observation_state_1
• observation_state_2
• observation_state_3
• observation_state_4
• observation_state_5

Flattened vector groups:

• action -> action_0 ... action_5 (6 FLOAT fields)
• observation.state -> observation_state_0 ... observation_state_5 (6 FLOAT fields)

Transformation Notes

• All 60 train episodes are merged into one table-model TsFile. Filter by episode_index and task_index to select an episode or task.
• Time uses TS_2DIFF + LZ4; FLOAT/DOUBLE use GORILLA + LZ4; INT32/INT64 use TS_2DIFF + LZ4; BOOLEAN uses RLE + LZ4 when present.
• action[6] and observation.state[6] become scalar FLOAT fields while preserving source prefixes; . is replaced with _.
• The source timestamp field is omitted after Time synthesis because it is exactly represented by Time / 1000 seconds.
• The source index field is retained as sample_index; frame_index is retained unchanged.
• No numeric rows, action dimensions, or state dimensions are omitted.

Videos

Videos are not duplicated here. The original repository contains two frame-aligned streams at the source video layout shown above:

• observation.images.phone - 60 per-episode MP4 files, 497,918,244 bytes
• observation.images.wrist - 60 per-episode MP4 files, 97,883,599 bytes

The 120 MP4 files total 595,801,843 bytes. Use episode_index and frame_index to align numeric rows with the original per-episode videos.

Minimal Read Example

from tsfile import TsFileReader

reader = TsFileReader("data/haribot099_so101_60.tsfile")
columns = [
    "episode_index",
    "task_index",
    "frame_index",
    "sample_index",
    "action_0",
    "observation_state_0",
]

with reader.query_table("haribot099_so101_60", columns, batch_size=65536) as result:
    batch = result.read_arrow_batch()
    print(batch.to_pandas().head())
reader.close()

Citation

The source dataset card provides no paper or completed BibTeX entry. Cite the original Hugging Face dataset and Perla Harish (Haribot099) when using this TsFile representation.