ml-research / feathers Goto Github PK

Dear @J0nasSeng,

Thanks a lot for sharing the code. I wanted to bring to your notice a possible bug in the code.

When I print the grad of self.alphas_normal at the end of step function of architect, I get the gradient as None.

To reproduce

Please add the line after L39 in code:

print(self.model.alphas_normal.grad)

it displays None to me and because of that, self.alphas_normal and self.alphas_reduce are not updated at all during the search phase. Please check.