ray.rllib.policy.policy.Policy.export_checkpoint#

Policy.export_checkpoint(export_dir: str, filename_prefix=-1, *, policy_state: Dict[str, numpy.array | jnp.ndarray | tf.Tensor | torch.Tensor | dict | tuple] | None = None, checkpoint_format: str = 'cloudpickle') None[source]#

Exports Policy checkpoint to a local directory and returns an AIR Checkpoint.

Parameters:
  • export_dir – Local writable directory to store the AIR Checkpoint information into.

  • policy_state – An optional PolicyState to write to disk. Used by Algorithm.save_checkpoint() to save on the additional self.get_state() calls of its different Policies.

  • checkpoint_format – Either one of ‘cloudpickle’ or ‘msgpack’.

from ray.rllib.algorithms.ppo import PPOTorchPolicy
policy = PPOTorchPolicy(...)
policy.export_checkpoint("/tmp/export_dir")