collector#

Source code: tianshou/highlevel/params/collector.py

class CollectorFactory[source]#

Bases: ABC

abstract create_collector(algorithm: Algorithm, vector_env: BaseVectorEnv, buffer: ReplayBuffer | None = None, exploration_noise: bool = False) → BaseCollector[source]#

Creates a collector for the given algorithm and vectorized environment.

Parameters:

  • algorithm – the algorithm

  • vector_env – the vectorized environment

  • buffer – the replay buffer to be used by the collector; if None, a new buffer will be created with default parameters

  • exploration_noise – whether action shall be modified using the policy’s exploration noise

Returns:

the collector

class CollectorFactoryDefault[source]#

Bases: CollectorFactory

create_collector(algorithm: Algorithm, vector_env: BaseVectorEnv, buffer: ReplayBuffer | None = None, exploration_noise: bool = False) → BaseCollector[source]#

Creates a collector for the given algorithm and vectorized environment.

Parameters:

  • algorithm – the algorithm

  • vector_env – the vectorized environment

  • buffer – the replay buffer to be used by the collector; if None, a new buffer will be created with default parameters

  • exploration_noise – whether action shall be modified using the policy’s exploration noise

Returns:

the collector