loss.jl

			
			
			
			
			
			"""
			

			    blocklossfn(predblock, yblock)

			

			Construct a loss function that compares a batch of model outputs

			`ŷs` and encoded targets `ys` and returns a scalar loss.

			

			For example for `block = OneHotLabel(classes)` (i.e. an encoded

			`Label(classes)`), we have

			`blocklossfn(block, block) == Flux.Losses.logitcrossentropy`.

			"""
			

			
			function
			 

	
			blocklossfn
			 
			end
			

			

			
			

	
			blocklossfn
			(
			predblock
			)
			 
			=
			 
			

	
			blocklossfn
			(
			predblock
			,
			 
			predblock
			)