Module entity_gym_rs::agent
source · [−]Expand description
High level API for interacting with neural network agents.
Structs
A channel for receiving an agent action returned by AgentOps::act_async
or Agent::act_async_dyn
.
An observation that defines what an agent can see.
Agent that samples all actions uniformly at random.
Agent that implements the RogueNet entity neural network. Can be loaded from checkpoints produced by enn-trainer.
bevy
Bevy asset containing a RogueNetAgent
.
Bevy asset loader for loading .roguenet checkpoint archives as RogueNetAsset
s.
python
Used during training to interface with an external agent implementation.
python
An Environment
implementation that is paired with one or more TrainAgent
.
python
Traits
Defines a categorical action. Can be derived for enums.
Agents are given observations and return actions.
Augments the Agent
trait with more ergonomic typed versions of the Agent::act_dyn
and Agent::act_async_dyn
methods.
A data structure that can be serialized into a data format that can be processed by a neural network.
Functions
Loads an agent from a checkpoint directory.
Loads an agent from an archive of a checkpoint directory.
Returns a boxed RandomAgent
.
Returns a boxed RandomAgent
with the given seed.