How to search NAS for convolution classification task with one-shot strategy and lightning module?

[mashu]

Hi,

I managed to run NAS using DARTs strategy and one-shot. As far as I can tell, I had to modify examples to exclude logging of intermediate results because these are not use, which wasn't clear from the beginning. So it is optimizing using gradients. But than final model when I try to train it, performs terribly. Can I make it such that it optimizes the model on the accuracy using the same approach? What alternatives do I have ?

Thanks