Features

This page gives a run-down of many features FluxTraining.jl brings to the table. Most features are implemented as callbacks and using them is as simple as passing the callback when constructing a Learner and training with fit!:

			
			
			
			cb
			 
			=
			 
			
			CoolFeature🕶️Callback
			(
			)
			

			
			learner
			 
			=
			 
			

	
			Learner
			(
			model
			,
			 
			lossfn
			
			;
			 
			
			callbacks
			=
			
			[
			cb
			]
			,
			 
			
			data
			=
			
			(
			trainiter
			,
			 
			validiter
			)
			)
			

			

	
			fit!
			(
			learner
			,
			 
			nepochs
			)

Metrics

By default, Learner will track only the loss function. You can track other metric with the Metrics callback. See also Metric, AbstractMetric.

Hyperparameter scheduling

The Scheduler callback takes care of hyperparameter scheduling. See the Hyperparameter scheduling tutorial and also Scheduler, Schedule , HyperParameter.

Logging

For logging, use the logging callbacks:

• LogMetrics

• LogHyperParams

• LogHistograms

They each can have multiple logging backends, but right now the only one implemented in FluxTraining.jl is TensorBoardBackend. See also LoggerBackend, log_to, and Loggables.Loggable.

There is also an external package Wandb.jl that implements a logging backend for Weights&Biases .

Checkpointing

Use the Checkpointer callback to create model checkpoints after every epoch.

Early Stopping

Use EarlyStopping to stop when a stopping criterion is met. Supports all criteria in EarlyStopping.jl .