train_runners.rst

Trainers and Training Runners

This page contains the reference documentation for trainers and training runners:

Overview

General

These are general interfaces, classes and utility functions for trainers and training runners:

maze.train.trainers

~common.trainer.Trainer ~common.training_runner.TrainingRunner ~common.config_classes.TrainConfig ~common.config_classes.ModelConfig ~common.config_classes.AlgorithmConfig

~common.model_selection.model_selection_base.ModelSelectionBase ~common.model_selection.best_model_selection.BestModelSelection

~common.evaluators.evaluator.Evaluator ~common.evaluators.multi_evaluator.MultiEvaluator ~common.evaluators.rollout_evaluator.RolloutEvaluator

~common.value_transform.ValueTransform ~common.value_transform.ReduceScaleValueTransform ~common.value_transform.support_to_scalar ~common.value_transform.scalar_to_support

~common.replay_buffer.replay_buffer.BaseReplayBuffer ~common.replay_buffer.uniform_replay_buffer.UniformReplayBuffer

Trainers

These are interfaces, classes and utility functions for built-in trainers:

Actor-Critics (AC)

maze.train.trainers

~common.actor_critic.actor_critic_runners.ACRunner ~common.actor_critic.actor_critic_runners.ACDevRunner ~common.actor_critic.actor_critic_runners.ACLocalRunner ~common.actor_critic.actor_critic_trainer.ActorCritic ~common.actor_critic.actor_critic_events.ActorCriticEvents

~a2c.a2c_trainer.A2C ~a2c.a2c_algorithm_config.A2CAlgorithmConfig

~ppo.ppo_trainer.PPO ~ppo.ppo_algorithm_config.PPOAlgorithmConfig

~impala.impala_trainer.IMPALA ~impala.impala_algorithm_config.ImpalaAlgorithmConfig ~impala.impala_events.ImpalaEvents ~impala.impala_runners.ImpalaRunner ~impala.impala_runners.ImpalaDevRunner ~impala.impala_runners.ImpalaLocalRunner ~impala.impala_vtrace.log_probs_from_logits_and_actions_and_spaces ~impala.impala_vtrace.from_logits ~impala.impala_vtrace.from_importance_weights ~impala.impala_vtrace.get_log_rhos

~sac.sac_trainer.SAC ~sac.sac_algorithm_config.SACAlgorithmConfig ~sac.sac_events.SACEvents ~sac.sac_runners.SACRunner ~sac.sac_runners.SACDevRunner

Evolutionary Strategies (ES)

maze.train.trainers

~es.es_trainer.ESTrainer ~es.es_algorithm_config.ESAlgorithmConfig ~es.es_events.ESEvents ~es.es_runners.ESMasterRunner ~es.es_runners.ESDevRunner ~es.es_shared_noise_table.SharedNoiseTable ~es.optimizers.base_optimizer.Optimizer ~es.optimizers.sgd.SGD ~es.optimizers.adam.Adam ~es.distributed.es_distributed_rollouts.ESRolloutResult ~es.distributed.es_dummy_distributed_rollouts.ESDummyDistributedRollouts ~es.distributed.es_distributed_rollouts.ESDistributedRollouts ~es.distributed.es_rollout_wrapper.ESAbortException ~es.distributed.es_rollout_wrapper.ESRolloutWorkerWrapper ~es.es_utils.get_flat_parameters ~es.es_utils.set_flat_parameters

Imitation Learning (IL) and Learning from Demonstrations (LfD)

maze.train.trainers

~imitation.imitation_events.ImitationEvents ~imitation.bc_runners.BCRunner ~imitation.bc_trainer.BCTrainer ~imitation.bc_algorithm_config.BCAlgorithmConfig ~imitation.bc_validation_evaluator.BCValidationEvaluator ~imitation.bc_loss.BCLoss

Utilities

maze.train

~utils.train_utils.stack_numpy_dict_list ~utils.train_utils.unstack_numpy_list_dict ~utils.train_utils.compute_gradient_norm ~utils.train_utils.stack_torch_dict_list