Diffusion Policy: Configured policy training launcher

Pinned upstream Hydra training entrypoint that resolves a selected workspace configuration and invokes its operational training run. Statically reviewed source only; not run and no Hub preparation adapter.

diffusion-policy-train-py | Record version: source-4dc8d6e876db25a62bf6c647444d7ee88f6754cda992c75728419d47f26374e5

Record JSON | Markdown | Pinned upstream source

Access

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

Hub preparation adapter: false. Follow the upstream setup and invocation contract.

Source and license

Repository: https://github.com/real-stanford/diffusion_policy
Revision: 5ba07ac6661db573af695b419a7947ecb704690f
Path: train.py
Source SHA256: 4dc8d6e876db25a62bf6c647444d7ee88f6754cda992c75728419d47f26374e5
Original authors: Cheng Chi, Siyuan Feng, Yilun Du, Zhenjia Xu, Eric Cousineau, Benjamin Burchfiel, Shuran Song
License: MIT
License scope: Selected first-party Python workflow source only. Preserve upstream copyright and per-file notices. Model weights, datasets, third-party code, robot drivers/firmware and services retain separate terms; no rights for those assets are inferred.
License SHA256: 304e99a48271e10820f71a1b86ef395c06c100d7f68dd1d308790d1064fc4d41
Hub commit: 1d6d9eb134e397d910265a8d6e7d8f257f9e8d7f

Artifact

{
  "kind": "workflow",
  "format": "python",
  "workflow_role": "policy-training"
}

Requirements

{
  "python": "Python 3.9 in upstream simulation/real conda declarations.",
  "packages": [
    "Pinned diffusion_policy checkout",
    "Workflow-specific conda environment with PyTorch1.12.1 and CUDA toolkit11.6",
    "Configured simulator or real-robot dependencies"
  ],
  "display": "Hardware: README benchmark-reproduction setup specifies Linux with an NVIDIA GPU. Exact device and memory depend on chosen workspace/data; macOS is documented for development only. Weights: Checkpoint initialization/resume behavior depends on the explicitly chosen workspace configuration; no blanket external pretrained-weight requirement is declared by this dispatcher. Datasets: Requires a selected demonstration dataset/task configuration; README's Push-T example downloads a separate training dataset. Services/auth: README's documented training path explicitly logs in to W&B. The dispatcher itself delegates logging to its chosen workspace; no particular API-key variable is named.",
  "execution_domain": "computation",
  "source_imports": [
    "diffusion_policy.workspace.base_workspace",
    "hydra",
    "omegaconf",
    "pathlib",
    "sys"
  ],
  "notes": "Included because README names it as the operational training entrypoint and it calls workspace.run, not as a config-only expansion. Source registers an eval resolver, so only reviewed upstream configurations could be considered for later manual execution. Code licensing does not establish model, checkpoint, dataset or service permission.",
  "declared_requirements": {
    "python": "Python 3.9 in upstream simulation/real conda declarations.",
    "weights": "Checkpoint initialization/resume behavior depends on the explicitly chosen workspace configuration; no blanket external pretrained-weight requirement is declared by this dispatcher.",
    "datasets": "Requires a selected demonstration dataset/task configuration; README's Push-T example downloads a separate training dataset.",
    "hardware": "README benchmark-reproduction setup specifies Linux with an NVIDIA GPU. Exact device and memory depend on chosen workspace/data; macOS is documented for development only.",
    "external_services_and_auth": "README's documented training path explicitly logs in to W&B. The dispatcher itself delegates logging to its chosen workspace; no particular API-key variable is named.",
    "evidence": [
      {
        "anchor": "OmegaConf.register_new_resolver(\"eval\", eval",
        "line": 18
      },
      {
        "anchor": "workspace.run()",
        "line": 32
      }
    ],
    "license_scopes": {
      "code": "MIT for selected first-party workflow source.",
      "weights": "README links separately downloadable trained checkpoints but the reviewed repository text does not separately establish their license. Do not extend first-party-code MIT to weights.",
      "datasets": "README links separate task demonstration datasets and dependencies such as robomimic. Individual dataset licenses were not reviewed; code MIT does not grant dataset rights.",
      "third_party_code": "External dependencies retain their own licenses."
    },
    "interpretation": "Pinned upstream declarations only; no dependency resolution, model/data access, compatibility inference or runtime validation was performed."
  }
}

Invocation

{
  "entrypoint": "python train.py --config-name=<reviewed-training-workspace>",
  "notes": "Included because README names it as the operational training entrypoint and it calls workspace.run, not as a config-only expansion. Source registers an eval resolver, so only reviewed upstream configurations could be considered for later manual execution."
}

Explicit compatibility

[]

Missing environment fields remain unknown. 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.