BLIP
BLIP copied to clipboard
Extracting attention value
Hi, I want to extract decoder's cross attention value, but there is no option or way to return attention layers. Can you tell how should I do? Thanks