mirror of
https://github.com/lucidrains/DALLE2-pytorch.git
synced 2026-02-23 08:15:18 +01:00
from my vision transformer experience, dimension of attention head of 32 is sufficient for image feature maps
This commit is contained in: