unilab.visualization.interactive_playback.load_play_actor¶
- unilab.visualization.interactive_playback.load_play_actor(algo_name, actor, normalizer, checkpoint, expected_actor_semantics=None)[source]¶
Restore an off-policy play actor and its optional observation normalizer.
For FlashSAC,
expected_actor_semanticsenables fail-closed validation of the checkpoint’s actor construction/observation/action semantics against the owner config before weights are loaded.