Using validation accuracy vs validation loss to determine best model

vision
1

Hi,

I notice that in the tutorials for fine tuning torchvision models. The code uses the best validation accuracy to determine the best model. Although other materials I about fine tuning models tend to use the lowest validation loss to decide which is the best model. Is there a reason that validation accuracy is being used here. Thanks for any help in advice.

https://pytorch.org/tutorials/beginner/finetuning_torchvision_models_tutorial.html

Thanks

1 Like