|
Download README.md from KillzoneCCO/TEST: direct link, hf CLI and curl.
- Browser
- Download file 1.1 kB
-
https://huggingface.co/datasets/KillzoneCCO/TEST/resolve/main/README.md
- Command line
-
hf download hf://datasets/KillzoneCCO/TEST/README.md
-
curl -L -o README.md https://huggingface.co/datasets/KillzoneCCO/TEST/resolve/main/README.md
1.1 kB
| license: other | |
| task_categories: | |
| - robotics | |
| tags: | |
| - hbr | |
| - data-node-v2 | |
| - ainex | |
| # DATA_NODE HBR export (not LeRobot) | |
| This Hub repository received a **DATA_NODE v2 HBR v2** download artifact from | |
| the local dataset `dgdfr` (approved). It is **not** a LeRobot v3 / GR00T | |
| training layout. | |
| ## Why LeRobot / GR00T was not pushed | |
| DATA_NODE can commit to this repo only after a LeRobot-compatible payload | |
| builds. That builder requires one observation/action Parquet row per camera | |
| frame. This episode fails that gate: | |
| - `video/cam_main.mp4`: 200 frames | |
| - `data/observation_action.parquet`: 155 rows | |
| Quality review can still pass. Hugging Face delivery never starts until those | |
| counts match. The write token is a worker environment variable | |
| (`DATA_NODE_HF_TOKEN`), not a field in the Datasets UI. | |
| ## What is in `exports/dgdfr-hbr-v2.hbr` | |
| Gzip tar (HBR v2): | |
| - `manifest.json` | |
| - `checksums.sha256` | |
| - `meta/data-node-export.json` | |
| - episode parquet and `cam_main.mp4` | |
| Unpack with `tar -tzf exports/dgdfr-hbr-v2.hbr` (tar may warn about the | |
| `HBR.schema` pax header; that is expected). | |