Can you confirm that was for inference? I thought that was only for training 55min on 8x v100
You are right, inference only uses one single v100 according to the paper.
I missed the bit about using 8 of them to run it! Wow that’s a lot of GPU horsepower to do this. More efficient to just use a vtuber style pipeline using unreal engine and the metahumans or other avatars… only need one good GOU for that.
Follow up,glad to know I read it right the first time.