mandymo / darts Goto Github PK

Thank you for your sharing. Can you get the same result with author?

Hi, thanks for your updated version~

To clearly introduce it for us, can you describe your major changes in README or just under this question?

I read the code of the train_search.py carefully, and I found that when calculating Equation 6 in the paper, you used L_val calculated by the un-updated model parameters, L_val(w^', alpha) instead of w.

In the code, I think w_t should be used to calculate valid_loss instead of

valid_logits = self.model(valid_input)

Please advise me if my understanding is correct? Thank you.

I'm very interested in your code, because I don't know much about the main update history and differences between pytorch 0.3.1 and 1.0, so I would like to know what are the main modified code, Thanks