Fix optimizer state restoration and multi-host sharding in WAN trainer
- Fix no-op state.replace(...) when restoring opt_state and step by reassigning state. - Add restore_optimizer_state_by_path to normalize and align Orbax checkpoint paths with live Optax PyTree structure. - Update max_utils.device_put_replicated to convert to NumPy host arrays before callback slicing to avoid multi-host dispatch failures, and fast-path matching shardings.
N
Nina Shvetsova committed
b8a3fba7e3ccac4293c355fcb1168c84bbc04d8a
Parent: a4ff3ae