---
license: apache-2.0
authors:
- Hugging Face user sattgle
task_categories:
- robotics
tags:
- tsfile
- timeseries
- tabular
modality:
- timeseries
- tabular
pretty_name: Clean Test TsFile
configs:
- config_name: default
  data_files:
  - split: train
    path: data/sattgle_clean_test.tsfile
size_categories:
- 10K<n<100K
---

# Clean Test TsFile

This dataset is an Apache TsFile conversion of
[`sattgle/clean-test`](https://huggingface.co/datasets/sattgle/clean-test), a LeRobot v2.1 SO101 follower
robot-manipulation dataset for the clean task.
It contains numeric trajectories, timing, and episode/task tags. Videos remain
in the original Hugging Face repository.

## Source Dataset and Attribution

- Original dataset: [`sattgle/clean-test`](https://huggingface.co/datasets/sattgle/clean-test)
- Repository owner, uploader, and sole contributor:
  [Hugging Face user sattgle (`sattgle`)](https://huggingface.co/sattgle)
- License: Apache-2.0
- Task: "clean"
- Robot: `so101_follower`; LeRobot version: `v2.1`
- Split: `train`; sampling rate: 30 fps
- Scale: 50 episodes, 32,220 frames, 1 task
- Source shards: 50 Parquet files totaling 1,680,276 bytes
- Source frame layout:
  `data/chunk-{episode_chunk:03d}/episode_{episode_index:06d}.parquet`
- Paper, external homepage, and completed citation: not provided by the source card.

The source metadata, all 50 Parquet shards, and episode metadata consistently describe this 50-episode dataset.

## TsFile Dataset

- TsFile: `data/sattgle_clean_test.tsfile` (718,233 bytes)
- Table: `sattgle_clean_test`
- Rows: 32,220; episodes/devices: 50; tasks: 1
- Time precision: milliseconds
- TsFile/source-Parquet size ratio: 0.427
- `meta/` is preserved from the source, with only `meta/info.json` rewritten to
  describe the TsFile schema and source-video alignment.

## Schema and Mapping

`Time = round(timestamp * 1000)` milliseconds. Time starts at zero and is
strictly increasing inside every episode. The source `timestamp` is dropped
afterward because it is represented by `Time / 1000` seconds at the selected
precision.

| TsFile column | Role | Type | Source mapping |
|---|---|---|---|
| `Time` | TIME | TIMESTAMP | `round(timestamp * 1000)` ms |
| `episode_index` | TAG | STRING | Original INT64 episode index |
| `task_index` | TAG | STRING | Original INT64 task index |
| `frame_index` | FIELD | INT64 | Preserved |
| `sample_index` | FIELD | INT64 | Renamed from `index` |
| `action_0` ... `action_5` | FIELD | FLOAT | Flattened from `action[6]` |
| `observation_state_0` ... `observation_state_5` | FIELD | FLOAT | Flattened from `observation.state[6]` |

The six vector dimensions are `shoulder_pan.pos`, `shoulder_lift.pos`,
`elbow_flex.pos`, `wrist_flex.pos`, `wrist_roll.pos`, and `gripper.pos`.
Dots in source names are replaced by underscores. No numeric row, episode,
task, state dimension, or action dimension is dropped.

## Encodings and Compression

- FLOAT/DOUBLE: GORILLA + LZ4
- INT32/INT64: TS_2DIFF + LZ4
- Time: TS_2DIFF + LZ4
- BOOLEAN: RLE + LZ4 (the source contains no BOOLEAN field)
- TAG: TsFile table/device TAG storage

The physical table schema, every field codec, TAG roles, and all 32,220 rows
were read back with the Apache TsFile Java API.

## Videos

Videos are not included in this TsFile repository. The source contains
100 frame-aligned AV1 MP4 files (597,891,922 bytes), 640x360 at 30 fps with
no audio, in two streams with 50 episode files each:

- [`observation.images.front`](https://huggingface.co/datasets/sattgle/clean-test/tree/main/videos/chunk-000/observation.images.front)
- [`observation.images.rear`](https://huggingface.co/datasets/sattgle/clean-test/tree/main/videos/chunk-000/observation.images.rear)

The source layout is
`videos/chunk-{episode_chunk:03d}/{video_key}/episode_{episode_index:06d}.mp4`.
Use `episode_index`, `frame_index`, and `meta/episodes.jsonl` to align numeric
rows with the original video frames.

## Quality Checks

Source Parquet, staged Parquet, and complete Java TsFile readback all contain
32,220 rows. Scalar values, all flattened vector values, TAGs, episode indexes,
Time mapping and monotonicity, physical codecs, size, and SHA-256 were checked
locally. Conversion scripts and local quality-check reports are intentionally excluded
from this upload-ready directory.

## Minimal Read Example

```python
from tsfile import TsFileReader

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

## Citation

The source card provides no personal name, paper, external homepage, or completed citation. Cite the original
Hugging Face dataset and Hugging Face user sattgle (`sattgle`) when using this conversion.