Datasets:
Download scripts/setup_lerobot_groot.sh from Nabidnur/RoboRACE-Bench: direct link, hf CLI and curl.
- Browser
- Download file 816 Bytes
-
https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/scripts/setup_lerobot_groot.sh
- Command line
-
hf download hf://datasets/Nabidnur/RoboRACE-Bench/scripts/setup_lerobot_groot.sh
-
curl -L -o setup_lerobot_groot.sh https://huggingface.co/datasets/Nabidnur/RoboRACE-Bench/resolve/main/scripts/setup_lerobot_groot.sh
816 Bytes
| # LeRobot + GR00T N1.7 environment (per the HF/NVIDIA blog, Jul 2026). | |
| # Needs: Linux, NVIDIA GPU (16 GB+ inference, 40 GB+ full fine-tune), HF_TOKEN with | |
| # access granted to nvidia/Cosmos-Reason2-2B and nvidia/GR00T-N1.7-3B. | |
| set -euo pipefail | |
| : "${HF_TOKEN:?set HF_TOKEN (see .env.example)}" | |
| command -v uv >/dev/null || curl -LsSf https://astral.sh/uv/install.sh | sh | |
| command -v ffmpeg >/dev/null || sudo apt-get install -y ffmpeg | |
| mkdir -p third_party && cd third_party | |
| [ -d lerobot ] || git clone https://github.com/huggingface/lerobot.git | |
| cd lerobot | |
| uv venv --python 3.12 | |
| # groot = policy, training = trainer deps, feetech = SO-101 motors, libero = sim benchmark | |
| uv pip install -e ".[groot,training,feetech,viz,libero]" | |
| uv run python -c "import lerobot; print('lerobot', lerobot.__version__)" | |