undefined | Better HN

0 pointsgodmode20193y ago0 comments

Can you confirm that was for inference? I thought that was only for training 55min on 8x v100

0 comments

1 comments · 1 top-level

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

j / k navigate · click thread line to collapse