# NVIDIA Isaac GR00T: GR00T policy server

Serve a configured GR00T checkpoint or the upstream dataset-replay alternative through the policy server protocol with an explicit embodiment tag.

ID: `isaac-groot-gr00t-eval-run-gr00t-server-py`

Record version: `source-b94d8f879bc83b97120cfcf3e3c9d7a506d93f633a036474c7f60db81792705d`

Hub commit: `66a13cd6ae8ab828219a4916587fb83d6dc71809`

## Access

Fetch the pinned repository, verify this source checksum, and use upstream installation and invocation. This record does not install or execute code.

Optional inert file retrieval: `python /PATH_TO_SKILL/scripts/rsh.py fetch-source isaac-groot-gr00t-eval-run-gr00t-server-py --expected-source-revision dac54b313160c201c450f97df0fd74ebd379c892 --expected-source-sha256 b94d8f879bc83b97120cfcf3e3c9d7a506d93f633a036474c7f60db81792705d --project ./experiment --json`. This verifies one file, preserves provenance and never installs or runs it. See [source-file retrieval](https://hub.superfastrobotics.com/source-files.md).

Hub preparation adapter: **false**. Runtime verification: **not_run**.

Upstream setup: https://github.com/NVIDIA/Isaac-GR00T/blob/dac54b313160c201c450f97df0fd74ebd379c892/README.md

## Source and license

- Repository: https://github.com/NVIDIA/Isaac-GR00T
- Pinned source: https://github.com/NVIDIA/Isaac-GR00T/blob/dac54b313160c201c450f97df0fd74ebd379c892/gr00t/eval/run_gr00t_server.py
- Revision: `dac54b313160c201c450f97df0fd74ebd379c892`
- Path: `gr00t/eval/run_gr00t_server.py`
- Source SHA256: `b94d8f879bc83b97120cfcf3e3c9d7a506d93f633a036474c7f60db81792705d`
- Source bytes: 6882
- Original authors: NVIDIA CORPORATION & AFFILIATES
- License: [Apache-2.0](https://github.com/NVIDIA/Isaac-GR00T/blob/dac54b313160c201c450f97df0fd74ebd379c892/LICENSE)
- License scope: Only the 12 selected Python source files, each with explicit matching Apache-2.0 header. Preserve copyright and attribution, including source adaptation notices. Third-party code, datasets, simulators, robot firmware, services, model architecture components and weights retain separate terms.

License SHA256: `93363510f62ce25ad1a47605b0b06ab335edd82f1a444531c58dd30b6e5a5a0c`

## Artifact

```json
{
  "kind": "workflow",
  "format": "python",
  "workflow_role": "policy-serving"
}
```

## Requirements

```json
{
  "python": "Root environment declares >=3.12,<3.13; per-workflow or robot-client environments may differ and must follow their pinned setup documents.",
  "packages": [
    "Pinned gr00t source checkout",
    "Upstream PyTorch/CUDA or TensorRT stack as selected",
    "Matching simulator or robot connector only for selected workflow"
  ],
  "display": "No environment has been installed. Training, inference, platform architecture and robot-client environments differ; setup text is a declaration.",
  "execution_domain": "computation",
  "source_imports": [
    "dataclasses",
    "gr00t.configs.data.embodiment_configs",
    "gr00t.data.embodiment_tags",
    "gr00t.data.types",
    "gr00t.policy.gr00t_policy",
    "gr00t.policy.replay_policy",
    "gr00t.policy.server_client",
    "importlib",
    "json",
    "os",
    "pathlib",
    "sys",
    "tyro"
  ],
  "notes": "These are upstream source/document declarations, not verified installed dependencies or operating capabilities.",
  "declared_requirements": {
    "gpu_notes": "Model-serving branch defaults to CUDA; upstream learned-policy inference guide declares >=16 GB VRAM and CUDA 12.6+ for the dGPU profile. The dataset-replay branch does not establish a GPU or learned-model requirement.",
    "checkpoint_or_model_access": "Either a configured checkpoint or the source's dataset-replay branch is selected. Checkpoint licensing remains separate from the source-code license; no model or dataset was acquired.",
    "python": "Root environment declares >=3.12,<3.13; per-workflow or robot-client environments may differ and must follow their pinned setup documents.",
    "external_service_or_network": "Upstream model/data retrieval, configured policy server and optional logging/robot connectivity depend on selection. No network service was invoked by this entrypoint during review.",
    "interpretation": "These are upstream source/document declarations, not verified installed dependencies or operating capabilities.",
    "model_license_scope": {
      "upstream_readme_license_section": "Code: Apache-2.0; model weights: NVIDIA Open Model License",
      "evidence_path": "README.md",
      "checkpoint_license_applicability": "not_reviewed",
      "weights_acquired": false,
      "scope": "Source declarations only; no checkpoint access or downstream licensing clearance is inferred."
    }
  }
}
```

## Invocation

```json
{
  "entrypoint": "python gr00t/eval/run_gr00t_server.py",
  "notes": "Read the pinned source and workflow setup. Model-serving branch defaults to CUDA; upstream learned-policy inference guide declares >=16 GB VRAM and CUDA 12.6+ for the dGPU profile. The dataset-replay branch does not establish a GPU or learned-model requirement. Either a configured checkpoint or the source's dataset-replay branch is selected. Checkpoint licensing remains separate from the source-code license; no model or dataset was acquired. No environment was resolved and no source was executed."
}
```

## Compatibility

```json
[]
```

Missing environment fields remain unknown. Source discovery labels do not establish robot compatibility.

## Evidence

Level: **source_reviewed**. Runtime verification: **not_run**.

Verified exact source, applicable license and author/setup document bytes at a full Git commit; parsed the Python AST and reviewed this guarded operational entrypoint and its explicit license notices. No upstream code, installation, robot command, dataset, mesh or checkpoint was executed or loaded.

Source review and contributor reports do not establish task completion. Preserve upstream attribution and licenses.