pytorch-fid
pytorch-fid copied to clipboard
Feature: support multiple GPUS
Many GAN repos call calculate_fid_given_paths during training to monitor performance. We can train with n GPUS but have to pause all and run the validation with a single GPU, which is a speed bottleneck. Could it be distributed?
This sounds like a good idea, but I don't know when I will have time to implement it. I would welcome PRs for this.
OK, Ill work on a PR!