policy

Policy components and utilities.

A policy is a function or a model that maps observations from the environment to actions. It can be deterministic or stochastic, and it is typically used in reinforcement learning to decide what action to take based on the current state of the environment.

Policy is a specialized form of inference pipeline, which includes additional components for interacting with the environment, such as action sampling and state management.

Submodules

• base
• evaluator
• remote