tensorplay.utils.checkpoint.checkpoint

tensorplay.utils.checkpoint.checkpoint(function: Callable[[...], Any], *args: Any, use_reentrant: bool = True, context_fn: Callable[[], Any] | None = None, determinism_check: str = 'default', debug: bool = False, **kwargs: Any) → Any[source]

Runs function(*args, **kwargs) eagerly (no recomputation).

Mirrors torch.utils.checkpoint.checkpoint’s signature; emits a warning once when grad is enabled because backward will re-run the graph normally.