|
Download README.md from HeliosRobotic/SCVLA_dts: direct link, hf CLI and curl.
- Browser
- Download file 970 Bytes
-
https://huggingface.co/datasets/HeliosRobotic/SCVLA_dts/resolve/main/README.md
- Command line
-
hf download hf://datasets/HeliosRobotic/SCVLA_dts/README.md
-
curl -L -o README.md https://huggingface.co/datasets/HeliosRobotic/SCVLA_dts/resolve/main/README.md
970 Bytes
| license: cc-by-4.0 | |
| task_categories: | |
| - robotics | |
| tags: | |
| - libero | |
| - rgb-d | |
| - robot-manipulation | |
| - historical-commitments | |
| # SC-VLA datasets | |
| This private dataset registry stores SC-VLA-derived LIBERO commitment, | |
| predicate-validity, perturbation, consumer-deadline, and executed-subtask | |
| datasets used by the registered paper experiments. | |
| The original LIBERO dataset is not mirrored. LIBERO releases its datasets | |
| under CC BY 4.0; SC-VLA derived artifacts preserve that attribution and must be | |
| used with the upstream LIBERO citation. RGB-D memmap caches and frozen OpenVLA | |
| feature caches are reproducible intermediates and are intentionally excluded. | |
| Every published release has a manifest containing file sizes, SHA-256 digests, | |
| source Git revision, and immutable Hub paths. In-progress collections are not | |
| eligible for publication. | |
| Source and protocols: https://github.com/heliosink/SCVLA | |
| Upstream LIBERO project: https://github.com/Lifelong-Robot-Learning/LIBERO | |