Metrics

			Metrics(metrics...) <: Callback

Callback that tracks metrics during training.

You can pass any number of metrics with every argument being

• an AbstractMetric like Metric; or

• a function f(ŷs, ys) -> val

This callback is added by default to every Learner unless you pass in usedefaultcallbacks = false. A metric tracking learner.lossfn Loss is included by default.

The computed metrics can be access in learner.cbstate.metricsstep and learner.cbstate.metricsepoch for steps and epochs, respectively.

Examples

Track accuracy:

			
			
			
			cb
			 
			=
			 
			

			Metrics
			(

	
			accuracy
			)

Pass in [ Metric]s:

			
			
			
			cb
			 
			=
			 
			

			Metrics
			(
			
    
			

	
			Metric
			(
			

	
			Flux
			.
			
			mse
			,
			 
			
			device
			 
			=
			 

	
			gpu
			)
			,
			
    
			

	
			Metric
			(
			

	
			Flux
			.
			
			mae
			,
			 
			
			device
			 
			=
			 

	
			gpu
			)
			

			)