LearnerThrottler

LearnerThrottler(config: DQNExperimentConfiguration) -> None

Throttler for the learner to ensure training doesn't outpace data collection.

Initialize the LearnerThrottler.

Parameters:

• config (DQNExperimentConfiguration) – The experiment configuration.

Source

Methods

should_step

should_step(env_steps: int, current_iterations: int, replay_buffer: ReplayBuffer) -> bool

Determine if the learner should perform a training step.

Parameters:

• env_steps (int) – Total environment steps.
• current_iterations (int) – Total learner iterations.
• replay_buffer (ReplayBuffer) – The experience replay buffer.

Returns:

• bool – True if the learner should step, False otherwise.

Source

Values

config

config = config

Source