Early stopping in deep learning studio?


#1

Hi,

Is there any way to specify that I want to perform early stopping based on the validation loss? Moreover, I’d like to specify the “patience” parameter that specifies how long to keep the training after the validation loss starts to increase to see if eventually starts decreasing again.

Thanks


#2

Yes, you can specify saving model at every epoch, best validation accuracy, and lowest validation loss.

While the training will keep running for the configured number of epochs, your model is only being saved based on the above condition.

There is no way to specify the patience parameter.