agent

Source code: tianshou/highlevel/agent.py

---

class A2CAgentFactory(params: TActorCriticParams, sampling_config: SamplingConfig, actor_factory: ActorFactory, critic_factory: CriticFactory, optimizer_factory: OptimizerFactory)

class ActorCriticAgentFactory(params: TActorCriticParams, sampling_config: SamplingConfig, actor_factory: ActorFactory, critic_factory: CriticFactory, optimizer_factory: OptimizerFactory)

create_actor_critic_module_opt(envs: Environments, device: str | device, lr: float) → ActorCriticOpt

class ActorDualCriticsAgentFactory(params: TActorDualCriticsParams, sampling_config: SamplingConfig, actor_factory: ActorFactory, critic1_factory: CriticFactory, critic2_factory: CriticFactory, optim_factory: OptimizerFactory)

class AgentFactory(sampling_config: SamplingConfig, optim_factory: OptimizerFactory)

Factory for the creation of an agent’s policy, its trainer as well as collectors.

create_policy(envs: Environments, device: str | device) → BasePolicy

create_train_test_collector(policy: BasePolicy, envs: Environments) → tuple[Collector, Collector]

abstract create_trainer(world: World, policy_persistence: PolicyPersistence) → BaseTrainer

set_policy_wrapper_factory(policy_wrapper_factory: PolicyWrapperFactory | None) → None

set_trainer_callbacks(callbacks: TrainerCallbacks) → None

class DDPGAgentFactory(params: DDPGParams, sampling_config: SamplingConfig, actor_factory: ActorFactory, critic_factory: CriticFactory, optim_factory: OptimizerFactory)

class DQNAgentFactory(params: TDiscreteCriticOnlyParams, sampling_config: SamplingConfig, model_factory: ModuleFactory, optim_factory: OptimizerFactory)

class DiscreteCriticOnlyAgentFactory(params: TDiscreteCriticOnlyParams, sampling_config: SamplingConfig, model_factory: ModuleFactory, optim_factory: OptimizerFactory)

class DiscreteSACAgentFactory(params: TActorDualCriticsParams, sampling_config: SamplingConfig, actor_factory: ActorFactory, critic1_factory: CriticFactory, critic2_factory: CriticFactory, optim_factory: OptimizerFactory)

class IQNAgentFactory(params: TDiscreteCriticOnlyParams, sampling_config: SamplingConfig, model_factory: ModuleFactory, optim_factory: OptimizerFactory)

class NPGAgentFactory(params: TActorCriticParams, sampling_config: SamplingConfig, actor_factory: ActorFactory, critic_factory: CriticFactory, optimizer_factory: OptimizerFactory)

class OffPolicyAgentFactory(sampling_config: SamplingConfig, optim_factory: OptimizerFactory)

create_trainer(world: World, policy_persistence: PolicyPersistence) → OffpolicyTrainer

class OnPolicyAgentFactory(sampling_config: SamplingConfig, optim_factory: OptimizerFactory)

create_trainer(world: World, policy_persistence: PolicyPersistence) → OnpolicyTrainer

class PGAgentFactory(params: PGParams, sampling_config: SamplingConfig, actor_factory: ActorFactory, optim_factory: OptimizerFactory)

class PPOAgentFactory(params: TActorCriticParams, sampling_config: SamplingConfig, actor_factory: ActorFactory, critic_factory: CriticFactory, optimizer_factory: OptimizerFactory)

class REDQAgentFactory(params: REDQParams, sampling_config: SamplingConfig, actor_factory: ActorFactory, critic_ensemble_factory: CriticEnsembleFactory, optim_factory: OptimizerFactory)

class SACAgentFactory(params: TActorDualCriticsParams, sampling_config: SamplingConfig, actor_factory: ActorFactory, critic1_factory: CriticFactory, critic2_factory: CriticFactory, optim_factory: OptimizerFactory)

class TD3AgentFactory(params: TActorDualCriticsParams, sampling_config: SamplingConfig, actor_factory: ActorFactory, critic1_factory: CriticFactory, critic2_factory: CriticFactory, optim_factory: OptimizerFactory)

class TRPOAgentFactory(params: TActorCriticParams, sampling_config: SamplingConfig, actor_factory: ActorFactory, critic_factory: CriticFactory, optimizer_factory: OptimizerFactory)