rskill-gr00t-n17-libero

OpenRAL rSkill — NVIDIA Isaac GR00T N1.7 (3B) finetuned on the LIBERO benchmark, packaged for the OpenRAL robot agent framework.

This package bundles the inference checkpoint of nvidia/GR00T-N1.7-LIBERO at the repo root (config + experiment_cfg + statistics + safetensors; no DeepSpeed optimizer shards), plus a rskill.yaml manifest that adds capability checking, license surfacing, latency budgets, and local registry integration. The upstream repo keeps its weights in a libero_spatial/ subfolder with no root model.safetensors, so this repackage is what makes the GR00T sidecar's from_pretrained load it directly. Weights are redistributed under the NVIDIA Open Model License (see source_repo).

Runtime status. Runtime-validated (ADR-0046). The out-of-process adapter (openral_sim.policies.gr00t) + tools/gr00t_sidecar.py boot helper auto-spawn a Python-3.10 GR00T sidecar that NF4-quantizes the Cosmos backbone on load (~3 GB VRAM). Verified end-to-end on an RTX 4070 (8 GB): episode_done on scenes/sim/libero_spatial, ~100 ms/step, with weights fetched from this repo via the manifest's weights_uri.

Upstream model, architecture & training

Field	Value
Source repo	nvidia/GR00T-N1.7-LIBERO
Base model	nvidia/GR00T-N1.7-3B
VLM backbone	Cosmos-Reason2-2B (SigLip2 vision encoder)
Paper	arXiv:2503.14734 — GR00T N1: An Open Foundation Model for Generalist Humanoid Robots
Parameters	~3.1 B
License	NVIDIA Open Model License Agreement (commercial use permitted)
Pretraining	20K hours EgoScale human video + diverse robot demonstrations
Finetune	LIBERO task suite (Franka Panda, MuJoCo via robosuite)

GR00T is a cross-embodiment foundation model with variable-dimension proprioception and per-embodiment action heads. This checkpoint specializes the N1.7 base on the LIBERO Franka embodiment.

Supported robots / embodiments

Robot	Embodiment tag	Status	Notes
Franka Panda (LIBERO sim)	franka_panda	packaged	Native finetune embodiment; live eval in ADR-0046 PR2

GR00T exposes a LIBERO_PANDA embodiment tag internally; OpenRAL maps it to the canonical franka_panda embodiment from robots/.

Sensors / observation contract

Key	Type	Min resolution	Description
observation.images.camera1	RGB camera	224 × 224	Agentview / overhead
observation.images.camera2	RGB camera	224 × 224	Wrist / end-effector
state	Proprioception	(8,)	End-effector pose + gripper, LIBERO layout

The policy emits a 16-step action chunk; each action is 7-D (delta_ee_6d_plus_gripper). Images and state are normalized inside the GR00T checkpoint's own experiment_cfg metadata rather than a lerobot processor pipeline — hence no processors block in the manifest.

Manifest summary

Field	Value
name	OpenRAL/rskill-gr00t-n17-libero
version	0.1.0
license	nvidia_open_model (commercial OK)
role	s1
model_family	gr00t
embodiment_tags	franka_panda
runtime	pytorch (out-of-process sidecar, ADR-0046)
quantization.dtype	bf16
weights_uri	hf://OpenRAL/rskill-gr00t-n17-libero (this repo; root-level checkpoint)
chunk_size	16
state_contract.dim / action_contract.dim	8 / 7
latency_budget.per_chunk_ms	1500 ms (sidecar round-trip + 3B inference)

Full schema: openral_core.schemas.RSkillManifest.

Hardware

GR00T N1.7-3B (bf16, ~6 GB weights) plus the Cosmos-Reason VLM does not fit on an 8 GB GPU without NF4 quantization; the sidecar (ADR-0046 PR2) follows the NF4 isolated-venv recipe used by the RLDX and detector sidecars. A ≥ 16 GB GPU runs bf16 directly.

License

This rSkill package (rskill.yaml, README.md) is Apache-2.0.

The wrapped model weights (nvidia/GR00T-N1.7-LIBERO) are governed by the NVIDIA Open Model License Agreement, which permits commercial use. This is the key distinction from GR00T N1 / N1.5 / N1.6, which ship under the NVIDIA OneWay Noncommercial License and are blocked in commercial deployments by the OpenRAL loader unless OPENRAL_ALLOW_NONCOMMERCIAL=1 is set (CLAUDE.md §3, ADR-0046).