How can i decrease the learning rate when the validation loss stop decreasing?

1

i know how to make scheduler and decrease the learning rate after few steps,
i want something like ReduceLROnPlateau() in Kears.

2

Hi @Bassel,

I’d recommend looking at the following for the current state (“not included”) and the various ways to do it with current torch (be sure to read the later posts to match current pytorch versions):

Best regards

Thomas

1 Like