Views
No views yet
allros/lerobot_grab_cube1_trimmed
(SO101 single arm, front + wrist cameras). Final checkpoint only (step 10000, merged weights, bfloat16, ~6.3GB).| Item | Value |
|---|---|
| Base model | nvidia/GR00T-N1.7-3B |
| Embodiment | NEW_EMBODIMENT |
| Modality config | SO100 dual-cam (examples/SO100/so100_config.py) |
| Steps | 10000 |
| VLM | frozen (tune_llm=false, tune_visual=false) |
| DiT | frozen (tune_diffusion_model=false) |
| Action horizon | 16 |
1python gr00t/eval/run_gr00t_server.py \
2 --model-path <path-to-this-repo> \
3 --embodiment-tag NEW_EMBODIMENT \
4 --device cuda:0 --port 55551python scripts/evaluation/policy_inference.py \
2 --task=LeIsaac-SO101-GrabCube-Real-v0 \
3 --policy_type=gr00tn1.6 \
4 --policy_host=127.0.0.1 --policy_port=5555 \
5 --policy_action_horizon=16 \
6 --policy_language_instruction="Grab Cube" \
7 --enable_cameras --device=cudamodel-*.safetensors, config.json).processor/, experiment_cfg/: required by Gr00tPolicy.meta/modality.json: LeRobot modality mapping (SO100 layout).