Compute validation loss while training?

[doantientai]

I would like to separate my dataset into 3 sets: train, val and test, so that I can use the "val" dir for validating during the training phase and "test" dir for the final test.

From the Readme.md I think here "val" is used as a test set instead of a validation set. If I would like to make val and test separated as I said above, what can I do?

Thank you very much!

[heidotrofimov]

I tried to this, take a look:
#200