ray.rllib.callbacks.callbacks.RLlibCallback

class ray.rllib.callbacks.callbacks.RLlibCallback

Abstract base class for RLlib callbacks (similar to Keras callbacks).

These callbacks can be used for custom metrics and custom postprocessing.

By default, all of these callbacks are no-ops. To configure custom training callbacks, subclass RLlibCallback and then set {“callbacks”: YourCallbacksClass} in the algo config.

Methods

__init__
on_algorithm_init	Callback run when a new Algorithm instance has finished setup.
on_checkpoint_loaded	Callback run when an Algorithm has loaded a new state from a checkpoint.
on_create_policy	Callback run whenever a new policy is added to an algorithm.
on_env_runners_recreated	Callback run after one or more EnvRunner actors have been recreated.
on_environment_created	Callback run when a new environment object has been created.
on_episode_created	Callback run when a new episode is created (but has not started yet!).
on_episode_end	Called when an episode is done (after terminated/truncated have been logged).
on_episode_start	Callback run right after an Episode has been started.
on_episode_step	Called on each episode step (after the action(s) has/have been logged).
on_evaluate_end	Runs when the evaluation is done.
on_evaluate_offline_end	Runs when the offline evaluation is done.
on_evaluate_offline_start	Callback before offline evaluation starts.
on_evaluate_start	Callback before evaluation starts.
on_learn_on_batch	Called at the beginning of Policy.learn_on_batch().
on_offline_eval_runners_recreated	Callback run after one or more OfflineEvaluationRunner actors have been recreated.
on_postprocess_trajectory	Called immediately after a policy's postprocess_fn is called.
on_sample_end	Called at the end of EnvRunner.sample().
on_sub_environment_created	Callback run when a new sub-environment has been created.
on_train_result	Called at the end of Algorithm.train().