transformers
Add ALL_ATTENTION_FUNCTIONS compatibility for Pixtral model
#37960

Merged

Add ALL_ATTENTION_FUNCTIONS compatibility for Pixtral model #37960

ArthurZucker merged 7 commits into huggingface:main from uminaty:pixtral-all-attn

Add ALL_ATTENTION_FUNCTIONS compatibility for Pixtral model

c67c173c

github-actions marked this pull request as draft 275 days ago

uminaty marked this pull request as ready for review 275 days ago

github-actions requested a review from

ArthurZucker 275 days ago

github-actions requested a review from

zucchini-nlp 275 days ago

Fix invalid operand type

54b71c24

uminaty force pushed from 7c43f751 to 54b71c24 275 days ago

Allow image_sizes to be optional in forward pass to fit tests

0d7a1b7e

Disallow using sdpa with output_attentions

71827ac0

uminaty force pushed from 20f777d4 to 71827ac0 275 days ago

qubvel commented on 2025-05-05

Delete useless comments, use eager attention from smolvlm, use patter…

9e78ceef

ArthurZucker approved these changes on 2025-05-06

uminaty force pushed from 9e78ceef to 50cc674b 274 days ago

zucchini-nlp commented on 2025-05-06

add _supports_attention_backend

9503c773

uminaty force pushed from 50cc674b to 9503c773 274 days ago

use kwargs instead of position_ids

7457672b

ArthurZucker merged f6664ee7 into main 272 days ago

uminaty deleted the pixtral-all-attn branch 272 days ago

Reviewers

ArthurZucker

zucchini-nlp

qubvel

Assignees

No one assigned

Labels

None yet

Milestone

No milestone