Alex
Alex
Hi! With 16 gigs I've managed to get as high as 400x400 with batch size 2-4. Possible tips: if you run out of memory, it's best to restart the session,...
Hi! It's briefly described at the end of this repo's readme. Here's the paired image2image - https://github.com/aarcosg/fastai-course-v3-notes/blob/master/refactored_by_topics/CNN_L7_gan_feature-loss.md
Hi! What dataset are you training on? And what perceptual loss you are using? (and what model) Usually l1 should be a lot weaker (like 0.1 or less) than perceptual...
Hi, You've closed the issue, so I've thought you figured this out :D Which notebook are you talking about?
Just replace all the `.cuda()` with `.cpu()` in the code. I guess I should add dynamic selection based on the environment.
Ah, probably half-precision isn't supported on CPU, so try replacing `.half()` with `.float()` as well. This might still not work because of hardcoded .jit datatypes inside the model, but worth...
Hi! Well, this is usually used to train Deep Learning models. The files are images, you can view them with any suitable software.
It's been a while (almost 3 years), so I can't really tell which one It was. I'm trying to do another run with stylegan3, this time I will try to...
Looks awesome! Have you been using vanilla AnimeGANv2 code for training? What dataset did you use for face2portrait model?
Looks like you are trying to use two different models, and their layer names don't match. You should try printing all the models and check their layer names (and overall...