Expand description
Module for reinforcement learning.
Modules§
- environment
- Module for implementing an environment.
- policy
- Module for implementing a policy.
- transition_
buffer - Transition buffer.
Structs§
- Action
Context - An action along with additional context about the decision.
- Async
Policy - An asynchronous policy using an inference server with autobatching.
- RLTrain
Output - A training output.
- Step
Result - The result of taking a step in an environment.
- Transition
- A state transition in an environment.
- Transition
Batch - A batch of transitions.
- Transition
Buffer - A tensor-backed circular buffer for transitions.
Traits§
- Batchable
- Trait for a type that can be batched and unbatched (split).
- Environment
- Trait to be implemented for a RL environment.
- Environment
Init - Trait to define how to initialize an environment. By default, any function returning an environment implements it.
- Policy
- Trait for a RL policy.
- Policy
Learner - Learner for a policy.
- Policy
State - The state of a policy.
- Slice
Access - Trait for types that support tensor-like slice operations,
enabling storage in a
TransitionBuffer. - ToAction
- Defines how an environment’s action is converted to a policy’s action.
- ToObservation
- Defines how an environment’s state is converted to a policy’s observation.
Type Aliases§
- Learner
Transition Batch - Batched transitions for a PolicyLearner.