PR #2158 Transformers 4.48

Transformers 4.48 #2158

IlyasMoutawwakil merged 26 commits into main from transformers-4.48

test

51902804

testing tensor cache x)

6a03d765

fix logger

7207215f

condition cache class usage

6261094b

update opset for beit and data2vec vision and skip flattened/fused pk…

822066d3

style

3ab38fdb

fix args patcher

d713e5a4

fix modernbert testing

bf4d1f3a

adaot to new whisper returned generation length

230c3a00

fix is_causal in transformers

3d5d9c96

fix modernbert failures

96e27141

IlyasMoutawwakil commented on 2025-01-20

style

78a2dba7

traceable cache

967c6e2c

IlyasMoutawwakil requested a review from

echarlaix 338 days ago

IlyasMoutawwakil requested a review from

xenova 338 days ago

echarlaix approved these changes on 2025-01-20

use pkv index

1d743882

add version gard and clean up other model patcher version gards

d452c464

patch sdpa attention in optimum for now

5dcab7f1

remove modernbert condition

656941a4

style

1bcb38f3

fix MistralModelPatcher

23fa20eb

correctly patch gpt2 in vision encoder decoder

24c8f4b8

patch sdpa attention forward everywhere

3694ea4e

fix gpt2 cross attention in seq2seq as well

3d7d5869

IlyasMoutawwakil commented on 2025-01-26

IlyasMoutawwakil requested a review from

echarlaix 331 days ago

echarlaix approved these changes on 2025-01-27

moved traceable cache to a file for simplicity of model patcher

10833d8f

IlyasMoutawwakil commented on 2025-01-29

Apply suggestions from code review

9491d17f

style

2b731297

fix

dea98a04

IlyasMoutawwakil merged d1bcdf76 into main 330 days ago

IlyasMoutawwakil deleted the transformers-4.48 branch 330 days ago

Reviewers

echarlaix

xenova

Assignees

No one assigned

Labels

None yet

Milestone

No milestone

optimum Transformers 4.48 #2158 Merged

Transformers 4.48 #2158

optimum
Transformers 4.48
#2158

Merged