allenact.algorithms.onpolicy_sync.runner

[view_source]

Defines the reinforcement learning OnPolicyRunner.

SaveDirFormat

class SaveDirFormat(enum.Enum)

[view_source]

Directory formats that can be used when saving tensorboard logs, checkpoints, etc. during training/evaluation. FLAT: the first-level directories are logs, checkpoints, metrics, etc; the second-level are time strings of each experiment NESTED: the opposite to FLAT.