HUGGING FACE DATASET

so101_pick_cube_chunked

Name: so101_pick_cube_chunked
Creator: gpudad

gpudad/so101_pick_cube_chunked — source facts cached from the per-dataset Hugging Face API.

DIRECT ANSWER

License: Apache-2.0; size: 1M<n<10M; creator: gpudad; paper: No arXiv link in cached metadata.

Commercial use: Review the linked license and upstream rights.

API snapshot · Jul 14, 2026

Downloads

6,209

Hugging Face metadata

Size category

1M<n<10M

Hugging Face metadata

Last modified

Jan 14, 2026

Hugging Face metadata

Access

Not marked as gated

Per-dataset API

Dataset-card excerpt

SO101 Pick Cube Dataset (Chunked) This is a restructured version of the gpudad/so101_pick_cube dataset with episode-level video files for faster data loading during training. Why Chunked? The original dataset has 3 monolithic video files (one per camera, 13+ hours each). Random access during training is slow because the decoder must seek through huge files. This version splits videos into 1000 episodes per chunk, making data loading ~50x faster.…

Hugging Face metadata

Captured tags

Formats: Parquet
Modalities: Tabular, Video

Hugging Face

Open the source dataset

Review the current card, files, access terms, and metadata.

One methodology for every dataset

How this profile is built

See how license, size, creator, downloads, paper links, missing fields, and fetch failures are handled.

Robotics datasets