the p is not updated?

[ry-jojo]

Hi there, thanks for sharing the codebase.
I am confused about the implementation of AuxiNash. Where can I find the task preference vector p update after the initialization?

auxinash/methods/weight_methods.py

Line 321 in f97bbd1

def set_p(self, p):

[AvivSham]

Hi @ry-jojo
Thank you for your interest in our work. The parameters p are updated using the meta optimizer after hyper grads calculation, please check the following code snip:
https://github.com/AvivSham/auxinash/blob/f97bbd1817659ff58cf85c9d79ed3f60b3a460ce/implicit_diff/optim.py