PR #1801 added attn_implementation for hf_base.py

added attn_implementation for hf_base.py #1801

ethantang-db merged 15 commits into main from ethantang-db/sdpa

ethantang-db requested a review 279 days ago

dakinggg commented on 2025-04-22

ethantang-db requested a review from

dakinggg 279 days ago

dakinggg commented on 2025-04-22

ethantang-db requested a review from

dakinggg 279 days ago

dakinggg approved these changes on 2025-04-22

ethantang-db changed the title ~~added attn_implementation for hf_base.py and deprecated use_flash_attention_v2~~ added attn_implementation for hf_base.py 278 days ago

added attn_implementation

192b3564

miving things

01501b3b

corrections wrt pr comments

de0a179c

fix tests

540f7a11

fix tests

47bc0fe8

fix other stuff

13cbbd6d

lint

59514497

ignore type

00019940

test fix

cd786c5f

right patch

e26f6f08

just make it gpu

efc7423f

fixes wrt comments

39143524

default value

2efece4a

fixes wrt PR comments

6016a884

ethantang-db force pushed from f1a97e77 to 6016a884 278 days ago

fix merge conflicts

5b418082

ethantang-db requested a review from

dakinggg 278 days ago

dakinggg approved these changes on 2025-04-23

ethantang-db merged 272dbd6c into main 278 days ago

Reviewers

dakinggg

Assignees

No one assigned

Labels

None yet

Milestone

No milestone

llm-foundry added attn_implementation for hf_base.py #1801 Merged

added attn_implementation for hf_base.py #1801

llm-foundry
added attn_implementation for hf_base.py
#1801

Merged