Hacker News new | ask | show | jobs
by godmode2019 1252 days ago
Can you confirm that was for inference? I thought that was only for training 55min on 8x v100
1 comments

You are right, inference only uses one single v100 according to the paper.