mrkwatkins.oakemu_rl.framework.StepPlan

Classes and functions

• ActionSourceKind
• LearningTargetKind
• StepPlan

Constants

POLICY_NO_LEARNING

POLICY_NO_LEARNING = StepPlan(ActionSourceKind.POLICY, LearningTargetKind.NONE)

Policy-driven step that should not be written into learner training buffers.

Source

POLICY_PPO

POLICY_PPO = StepPlan(ActionSourceKind.POLICY, LearningTargetKind.PPO)

Normal policy step recorded for PPO learning.

Source

RANDOM_NO_LEARNING

RANDOM_NO_LEARNING = StepPlan(ActionSourceKind.RANDOM, LearningTargetKind.NONE)

Random-controlled exploration step with no learner-side use.

Source