Language

unilab.visualization.playback.run_native_viewer_play_loop¶

unilab.visualization.playback.run_native_viewer_play_loop(env, *, initialize, step, num_steps=None, render_spacing=None, render_offset_mode=None)[source]¶

Run the backend-native viewer loop under torch.inference_mode().

Shared by the training entrypoints’ native-viewer playback paths; callers own the policy wiring through the initialize/step callbacks.

Parameters:

initialize (Callable[[], TypeVar(ObsT)])
step (Callable[[TypeVar(ObsT)], TypeVar(ObsT)])
num_steps (int | None)
render_spacing (float | None)
render_offset_mode (str | None)

Return type:

None