Crate burn_train 

Training and evaluation for Burn models.

Applications enable the train feature of burn and use this crate as burn::train.

• Learner bundles a model, optimizer and learning rate scheduler, and SupervisedTraining runs it over training and validation data loaders on one or several devices.
• metric: loss, accuracy, precision and recall, F-scores, AUROC, BLEU, CER, WER, perplexity, system usage and more; with the vision feature, image metrics such as PSNR, SSIM, LPIPS and FID.
• renderer: a terminal dashboard (tui feature) or plain CLI output for training progress.
• checkpoint: periodic and metric-based checkpointing, and resuming from a checkpoint.
• Early stopping, interruption and an Evaluator for test sets.

Failures on a device during training are returned as errors rather than panics.

Feature flags

• tui (default): terminal dashboard.
• sys-metrics (default): CPU, memory and GPU usage metrics.
• vision: image quality metrics.
• rl: reinforcement learning training through burn-rl.
• tracing: instrument operations with the tracing crate.

Modules

checkpoint

The checkpoint module.

logger

The logger module.

metric

The metric module.

renderer

Renderer modules to display metrics and training information.

train

The trainer module.

Structs

AgentEnvAsyncLooprl

An asynchronous agent/environement interface.

AgentEnvBaseLooprl

A simple, synchronized agent/environement interface.

AsyncAgentEnvLoopConfigrl

Configuration for an async agent/environment loop.

AsyncProcessorEvaluation

Event processor for the model evaluation.

AsyncProcessorTraining

Event processor for the training process.

ClassificationOutput

Simple classification output adapted for multiple metrics.

EpisodeSummaryrl

Summary of an episode.

EvaluationItem

An evaluation item.

EvaluationResult

The result of an evaluation.

Evaluator

Evaluates a model on a specific dataset.

EvaluatorBuilder

Struct to configure and create an evaluator.

EvaluatorComponentTypesMarker

A marker type used to provide evaluation components.

EventProcessorError

Every failure an event processor reported at once, oldest first.

FileApplicationLoggerInstaller

This struct is used to install a local file application logger to output logs to a given file path.

Interrupter

A handle that allows aborting the training/evaluation process early.

Interruption

A stop made through Interrupter::stop, without an error.

Learner

Learner struct encapsulating all components necessary to train a Neural Network model.

LearnerSummary

Detailed training summary.

LearnerSummaryConfig

Learning summary config.

LearningCheckpointer

Used to create, delete, or load checkpoints of the training process.

LearningResult

The result of a training, containing the model along with the renderer.

MetricEarlyStoppingStrategy

An early stopping strategy based on a metrics collected during training or validation.

MetricEntry

Contains the metric value at a given time.

MetricError

A metric that could not process an event.

MetricSummary

Contains the summary of recorded values for a given metric.

ModuleSharder

Describes how the module is distributed across multiple devices.

MultiAgentEnvLooprl

An asynchronous runner for multiple agent/environement interfaces.

MultiLabelClassificationOutput

Multi-label classification output adapted for multiple metrics.

OffPolicyConfigrl

Parameters of an on policy training with multi environments and double-batching.

OffPolicyStrategyrl

Off-policy reinforcement learning strategy with multi-env experience collection and double-batching.

RLCheckpointerrl

Used to create, delete, or load checkpoints of the training process.

RLComponentsrl

Struct to minimise parameters passed to RLStrategy::train.

RLComponentsMarkerrl

Concrete type that implements the RLComponentsTypes trait.

RLEventProcessorrl

An event processor that handles:

RLResultrl

The result of reinforcement learning, containing the final policy along with the renderer.

RLTrainingrl

Structure to configure and launch reinforcement learning trainings.

RegressionOutput

Regression output adapted for the loss metric.

SequenceOutput

Sequence prediction output adapted for multiple metrics.

SummaryMetrics

Contains the summary of recorded metrics for the training and validation steps.

SupervisedTraining

Structure to configure and launch supervised learning trainings.

TimeSteprl

A timestep debscribing an iteration of the state/decision process.

TrainOutput

A training output.

TrainingComponents

Struct to minimise parameters passed to SupervisedLearningStrategy::train. These components are used during training.

TrainingItem

A learner item.

Trajectoryrl

A trajectory, i.e. a list of ordered TimeStep.

Enums

AgentEvaluationEventrl

Event happening during evaluation of a reinforcement learning’s agent.

EvaluatorEvent

Event happening during the evaluation process.

EventProcessorFailure

A failure an event processor reported.

ExecutionStrategy

Describes where training runs.

LearnerEvent

Event happening during the training/validation process.

MultiDeviceOptim

Determine how the optimization is performed when training with multiple devices.

RLEventrl

Event happening during reinforcement learning.

RLStrategiesrl

The strategy for reinforcement learning.

StoppingCondition

The condition that early stopping strategies should follow.

TrainingError

An error that stopped training or evaluation.

TrainingStrategy

How should the learner run the learning for the model

Traits

AgentEnvLooprl

Trait for a structure that implements an agent/environement interface.

AgentMetricRegistrationrl

Trait to fake variadic generics for train step metrics.

AgentTextMetricRegistrationrl

Trait to fake variadic generics for train step text metrics.

ApplicationLoggerInstaller

This trait is used to install an application logger.

CloneEarlyStoppingStrategy

A helper trait to provide type-erased cloning.

EarlyStoppingStrategy

A strategy that checks if the training should be stopped.

EpisodeMetricRegistrationrl

Trait to fake variadic generics for episode metrics.

EpisodeTextMetricRegistrationrl

Trait to fake variadic generics for episode text metrics.

EvalMetricRegistration

Trait to fake variadic generics.

EvalTextMetricRegistration

Trait to fake variadic generics.

EvaluatorComponentTypes

All components necessary to evaluate a model grouped in one trait.

EventProcessorEvaluation

Process events happening during evaluation.

EventProcessorTraining

Process events happening during training and validation.

InferenceStep

Trait to be implemented for validating models.

ItemLazy

Items that are lazy are not ready to be processed by metrics.

LearnerModel

A single keyword trait for a Burn module used for learning.

MetricRegistration

Trait to fake variadic generics.

RLComponentsTypesrl

All components used by the reinforcement learning paradigm, grouped in one trait.

RLStrategyrl

Provides the fit function for any learning strategy

SupervisedLearningStrategy

Provides the fit function for any learning strategy

TextMetricRegistration

Trait to fake variadic generics.

TrainMetricRegistrationrl

Trait to fake variadic generics for env step metrics.

TrainStep

Trait to be implemented for models to be able to be trained.

TrainTextMetricRegistrationrl

Trait to fake variadic generics for env step text metrics.

Type Aliases

CustomLearningStrategy

A reference to an implementation of SupervisedLearningStrategy.

CustomRLStrategyrl

A reference to an implementation of RLStrategy.

RLAgentRecordrl

The record of the learning agent.

RLEventProcessorTyperl

The event processor type for reinforcement learning.

RLPolicyRecordrl

The record of the policy.

SupervisedTrainingEventProcessor

The event processor type for supervised learning.

TrainLoader

A reference to the training split DataLoader.

ValidLoader

A reference to the validation split DataLoader.