Discriminative learning rates

In many cases, it’s useful to train different layer groups with different learning rates. How do we achieve this with the stand keras.optimizers?

1 Like

I don’t think that we have an off the shelf solution.

Probably you could use a custom function with gradient_transformers