SonyCSLParis / pesto

Self-supervised learning for fast pitch estimation

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

ALWA reduction seems to use too much GPU resources

aRI0U opened this issue · comments

maybe gather is involved?

Cannot manage to reproduce the issue. On a GTX 1080 CUDA OOM happens when trying to pitch estimate more than 9min of audio in parallel with both argmax and alwa reduction methods