simulo.PolicyCheckpoint

A resolved policy checkpoint supplied to a play or evaluation job.

The contract contains only standard-library values. Loading the policy is deliberately deferred until execution, where simulo-backend provides the concrete policy-network loader.

@dataclass(frozen=True)
class PolicyCheckpoint

Field	Type	Default	Description
checkpoint	str	required
policy_id	str	required
kind	str	required
path	Path	required
meta	Mapping[str, Any]	required
iteration	int	required	Training iteration stored in the checkpoint sidecar.
max_iterations	int	required	Configured training limit stored in the checkpoint sidecar.
reward	`float	None`	required
obs_dim	int	required	Observation width used to train the checkpoint.
act_dim	int	required	Action width produced by the checkpoint.
task_name	str	required	Name of the task class used for training, if the sidecar has it.
task_digest	str	required	Digest of the file that defined the training task, if available.