simulo.preview

Check a task’s robot the way training will run it, and return the report.

Builds the task’s environment with num_envs environments and runs the preview checks: Settle, Joint sweep, Random moves and Training loop. It records one MCAP recording with the findings on its timeline (and video when video is true), saves the report and the recording as job outputs, and returns the report as a plain dict; return it from the job. A preview makes no policy.

checks is "all" or a comma-separated list of settle, joint-sweep, random-moves, training-loop and action-map. seed seeds the resets and the random actions.

simulo.preview(
    task: TaskProtocol,
    *,
    num_envs: int = 16,
    checks: str = 'all',
    video: bool = True,
    seed: int = 0,
) -> Dict[str, Any]

---

Note

Call it from a @app.job(type="preview") job and return its report. simulo run preview.py then prints the report and exits 0 when every check passed (warnings allowed), 1 when a check failed. Like RLTrainer, simulo.preview resolves to a torch-free stand-in at submit time and to the runtime function on the GPU worker; the signature below is the same for both.