PR #37363 Fix SDPA attention precision issue in Qwen2.5-VL

Fix SDPA attention precision issue in Qwen2.5-VL #37363

zucchini-nlp merged 3 commits into huggingface:main from JJJYmmm:main

github-actions marked this pull request as draft 1 year ago

JJJYmmm marked this pull request as ready for review 1 year ago

github-actions requested a review from

ArthurZucker 1 year ago

github-actions requested a review from

Rocketknight1 1 year ago

JJJYmmm closed this 1 year ago

JJJYmmm reopened this 1 year ago

github-actions marked this pull request as draft 1 year ago

JJJYmmm force pushed from 3ca2045e to 483409e0 1 year ago

JJJYmmm marked this pull request as ready for review 1 year ago

ArthurZucker commented on 2025-04-09

JJJYmmm force pushed from 483409e0 to 026c55d4 1 year ago

JJJYmmm force pushed from 50a81dcd to 397511e0 1 year ago

JJJYmmm requested a review from

ArthurZucker 1 year ago

ArthurZucker approved these changes on 2025-04-11

JJJYmmm force pushed from 6b9e0c66 to fa6bb3f1 1 year ago

JJJYmmm changed the title ~~Implement improved window attention in eager/sdpa version for Qwen2.5VL~~ Fix SDPA attention precision issue in Qwen2.5-VL 1 year ago

JJJYmmm force pushed from 65f1395c to 5c821843 1 year ago

JJJYmmm force pushed from 5c821843 to 21e6406b 1 year ago

zucchini-nlp approved these changes on 2025-07-02

solve conflicts and remove redundant attention_mask in qwenvit

bddd2c9d

JJJYmmm force pushed from 95cfafe3 to bddd2c9d 1 year ago

update decoded text check

d2f83f42

remove trailing whitespace

970eecdf

ArthurZucker approved these changes on 2025-07-08

zucchini-nlp merged 25343aaf into main 1 year ago

Reviewers

ArthurZucker

zucchini-nlp

Rocketknight1

Assignees

No one assigned

Labels

None yet

Milestone

No milestone

transformers Fix SDPA attention precision issue in Qwen2.5-VL #37363 Merged

Fix SDPA attention precision issue in Qwen2.5-VL #37363

transformers
Fix SDPA attention precision issue in Qwen2.5-VL
#37363

Merged