[static cache] fix device map per layer in VLMs #38488
zucchini-nlp
merged
2 commits into
huggingface:main
from
zucchini-nlp:multi-gpu-vlms
return lm as decoder
a1aacfb2
gante
approved these changes
on 2025-06-20
Merge branch 'main' into multi-gpu-vlms
5c9d7643
Login to write a write a comment.
Login via GitHub