First, thank you very much for sharing the code with the community.
I am curious about why and how the gradient penalty works.
As you said in the Table5 (on page 13 of the arxiv version paper), the gradient penalty loss is very import for the results, but I didn't find the specific description for it in the paper.
So can you point out how it works and where the code is?
First, thank you very much for sharing the code with the community.
I am curious about why and how the gradient penalty works.
As you said in the Table5 (on page 13 of the arxiv version paper), the gradient penalty loss is very import for the results, but I didn't find the specific description for it in the paper.
So can you point out how it works and where the code is?
Thanks again for the attractive work.