Change keras regularizer during training / dynamic regularization

Question

I am looking for a proper way to implement a dynamic regularization of weights for a layer during the training. As an example after 10 calls I want to replace the L2 regular