OakEmu-RL v0.1.0

Appearance

StepPlan ​

python
StepPlan(action_source: ActionSourceKind, learning_target: LearningTargetKind) -> None

Flyweight step-plan describing action selection and learning behaviour.

Source

Properties ​

is_policy_trainable ​

python
is_policy_trainable: bool

Return whether PPO should optimise this step on-policy.

Returns:

  • bool – True when the learning target is PPO.

Source

requires_observation ​

python
requires_observation: bool

Return whether this step needs materialized image observations.

Returns:

  • bool – True when observations are needed for policy inference or BC storage.

Source

requires_policy_inference ​

python
requires_policy_inference: bool

Return whether this step needs a policy forward pass.

Returns:

  • bool – True when actions come from the policy.

Source

stores_rollout_step ​

python
stores_rollout_step: bool

Return whether this step should be written into rollout storage.

Returns:

  • bool – True when PPO or BC consumes the step later.

Source

Values ​

action_source ​

python
action_source: ActionSourceKind

Source

learning_target ​

python
learning_target: LearningTargetKind

Source