ray.rllib.algorithms.algorithm.Algorithm.save

Algorithm.save(checkpoint_dir: Optional[str] = None) → ray.train._internal.session._TrainingResult

Saves the current model state to a checkpoint.

Subclasses should override save_checkpoint() instead to save state.

Parameters

checkpoint_dir – Optional dir to place the checkpoint.

Returns

The given or created checkpoint directory.

Note the return value matches up with what is expected of restore().

DeveloperAPI: This API may change across minor Ray releases.