reagent.reporting package

Submodules

reagent.reporting.actor_critic_reporter module

class reagent.reporting.actor_critic_reporter.ActorCriticReporter(report_interval: int = 100)

Bases: reagent.reporting.reporter_base.ReporterBase

property aggregating_observers

generate_training_report() → reagent.workflow.training_reports.ActorCriticTrainingReport

property value_list_observers

reagent.reporting.compound_reporter module

class reagent.reporting.compound_reporter.CompoundReporter(reporters: List[reagent.reporting.reporter_base.ReporterBase], merge_function: Callable[[List[reagent.reporting.reporter_base.ReporterBase]], reagent.core.result_registries.TrainingReport])

Bases: reagent.reporting.reporter_base.ReporterBase

flush(epoch: int)

generate_training_report() → reagent.core.result_registries.TrainingReport

log(**kwargs) → None

set_flush_function(flush_function)

reagent.reporting.discrete_crr_reporter module

class reagent.reporting.discrete_crr_reporter.DiscreteCRRReporter(actions: List[str], report_interval: int = 100, target_action_distribution: Optional[List[float]] = None, recent_window_size: int = 100)

Bases: reagent.reporting.reporter_base.ReporterBase

generate_training_report() → reagent.workflow.training_reports.DQNTrainingReport

reagent.reporting.discrete_dqn_reporter module

class reagent.reporting.discrete_dqn_reporter.DiscreteDQNReporter(actions: List[str], report_interval: int = 100, target_action_distribution: Optional[List[float]] = None, recent_window_size: int = 100)

Bases: reagent.reporting.reporter_base.ReporterBase

generate_training_report() → reagent.workflow.training_reports.DQNTrainingReport

reagent.reporting.parametric_dqn_reporter module

class reagent.reporting.parametric_dqn_reporter.ParametricDQNReporter(report_interval: int = 100)

Bases: reagent.reporting.reporter_base.ReporterBase

generate_training_report() → reagent.workflow.training_reports.ParametricDQNTrainingReport

reagent.reporting.reporter_base module

class reagent.reporting.reporter_base.ReporterBase(value_list_observers: Dict[str, reagent.core.observers.ValueListObserver], aggregating_observers: Dict[str, reagent.core.observers.IntervalAggregatingObserver])

Bases: reagent.core.observers.CompositeObserver

flush(epoch: int)

abstract generate_training_report() → reagent.core.result_registries.TrainingReport

log(**kwargs) → None

reagent.reporting.reward_network_reporter module

class reagent.reporting.reward_network_reporter.RewardNetworkReporter(loss_type: reagent.training.reward_network_trainer.LossFunction, model_description: str, report_interval: int = 100)

Bases: reagent.reporting.reporter_base.ReporterBase

property aggregating_observers

update_best_model(loss: float, model: reagent.models.base.ModelBase)

property value_list_observers

reagent.reporting.seq2reward_reporter module

class reagent.reporting.seq2reward_reporter.Seq2RewardCompressReporter(action_names: List[str], report_interval: int = 100)

Bases: reagent.reporting.seq2reward_reporter.Seq2RewardReporter

property aggregating_observers

class reagent.reporting.seq2reward_reporter.Seq2RewardReporter(action_names: List[str], report_interval: int = 100)

Bases: reagent.reporting.reporter_base.ReporterBase

property aggregating_observers

generate_training_report() → reagent.workflow.training_reports.Seq2RewardTrainingReport

property value_list_observers

reagent.reporting.slate_q_reporter module

class reagent.reporting.slate_q_reporter.SlateQReporter(report_interval: int = 100)

Bases: reagent.reporting.reporter_base.ReporterBase

property aggregating_observers

generate_training_report() → reagent.workflow.training_reports.SlateQTrainingReport

property value_list_observers

reagent.reporting.td3_reporter module

class reagent.reporting.td3_reporter.TD3Reporter(report_interval: int = 100)

Bases: reagent.reporting.actor_critic_reporter.ActorCriticReporter

property aggregating_observers

reagent.reporting.world_model_reporter module

class reagent.reporting.world_model_reporter.WorldModelReporter(report_interval: int = 100)

Bases: reagent.reporting.reporter_base.ReporterBase

property aggregating_observers

generate_training_report() → reagent.workflow.training_reports.WorldModelTrainingReport

property value_list_observers

Module contents

class reagent.reporting.CompoundReporter(reporters: List[reagent.reporting.reporter_base.ReporterBase], merge_function: Callable[[List[reagent.reporting.reporter_base.ReporterBase]], reagent.core.result_registries.TrainingReport])

Bases: reagent.reporting.reporter_base.ReporterBase

flush(epoch: int)

generate_training_report() → reagent.core.result_registries.TrainingReport

log(**kwargs) → None

set_flush_function(flush_function)

class reagent.reporting.ReporterBase(value_list_observers: Dict[str, reagent.core.observers.ValueListObserver], aggregating_observers: Dict[str, reagent.core.observers.IntervalAggregatingObserver])

Bases: reagent.core.observers.CompositeObserver

flush(epoch: int)

abstract generate_training_report() → reagent.core.result_registries.TrainingReport

log(**kwargs) → None