transformers
🚨 Modeling changes for export, compile, and hybrid-attention standardization
#46738

Merged

🚨 Modeling changes for export, compile, and hybrid-attention standardization #46738

IlyasMoutawwakil merged 48 commits into main from hf-exporters-models

model fixes for export compatibility

5b6993a1

Merge branch 'main' into hf-exporters-models

43326dff

add minicpmv vision utils

9e9da667

IlyasMoutawwakil requested a review from

vasqu 17 days ago

IlyasMoutawwakil assigned

ArthurZucker 17 days ago

can compile

6e6e3ac6

IlyasMoutawwakil unassigned

ArthurZucker 17 days ago

IlyasMoutawwakil unassigned

ArthurZucker 17 days ago

make ssms fully compilable

1fce6c3d

IlyasMoutawwakil commented on 2026-06-18

more ssms

7ba1546e

vasqu commented on 2026-06-18

address new comments

ff33d8b3

fast ci fixes

a2b3a366

IlyasMoutawwakil marked this pull request as ready for review 16 days ago

fix minimax

07bc5c7b

fix the last falcon fast ci failure

fadcdf8e

IlyasMoutawwakil requested a review from

vasqu 16 days ago

cleanup

1ea08b68

use can_return_tuple in patchmixer

21a8fd1c

avoid 5D attention

a9eef0a0

granite export fix

7587ae69

Merge branch 'main' into hf-exporters-models

3ef899e3

vasqu commented on 2026-06-19

address modeling comments

7caddb13

update

6fddf68a

address more comments, unify linear attention and make more models su…

60908510

style

a1be3efb

style

bf4a9c75

Merge branch 'main' into hf-exporters-models

95f5ae15

modeling changes for attn impl compatibility

77d95f3f

fix

17e3d60d

fixes

9e738821

Merge branch 'main' into hf-exporters-models

1ecddd30

fix lfm2 conv errors

b2f9a31e

IlyasMoutawwakil changed the title ~~Model fixes for export compatibility~~ 🚨 Modeling changes for export, compile, and hybrid-attention standardization 13 days ago

vasqu commented on 2026-06-22

Cyrilvallez commented on 2026-06-22

address most of anton's review comments

f520f198

address cyril's comments

5e7bdeee

vasqu commented on 2026-06-22

revert conv rename

707df61d

simpler modular

32929f67

keep max_batch_size check

d8f85a6d

vasqu commented on 2026-06-22

fixes

25873ea2

fix moshi cuda graphs

b7d3137a

IlyasMoutawwakil requested a review from

vasqu 12 days ago

IlyasMoutawwakil requested a review from

Cyrilvallez 12 days ago

added layer type class attr in mixers

5f7f241b

fix repo

a14a41c4

vasqu commented on 2026-06-23

address comments

753b0861

IlyasMoutawwakil requested a review from

vasqu 12 days ago

vasqu commented on 2026-06-24

address comments

5e5c8928

Merge branch 'main' into hf-exporters-models

cca58651

fix merge

2f4e2634

Merge branch 'main' into hf-exporters-models

6070f0ed

claude review findings

1faf03eb

IlyasMoutawwakil requested a review from

vasqu 6 days ago

vasqu approved these changes on 2026-06-29

docs comments

e6883b5b

address comments

5b6d9f92

Merge branch 'main' into hf-exporters-models

49b70375

style

90adb3eb

claude review

5d7a9184

Merge branch 'main' into hf-exporters-models

b52f9820

IlyasMoutawwakil requested a review from

copilot-pull-request-reviewer 5 days ago

copilot-pull-request-reviewer commented on 2026-06-30

small fix caught by copilot

d4e05ddc

IlyasMoutawwakil merged 181beb3b into main 5 days ago

IlyasMoutawwakil deleted the hf-exporters-models branch 5 days ago

Reviewers

vasqu

Cyrilvallez

copilot-pull-request-reviewer

Assignees

No one assigned

Labels

None yet

Milestone

No milestone

transformers 🚨 Modeling changes for export, compile, and hybrid-attention standardization #46738 Merged

🚨 Modeling changes for export, compile, and hybrid-attention standardization #46738

transformers
🚨 Modeling changes for export, compile, and hybrid-attention standardization
#46738

Merged