Commit Graph

3 Commits

Author SHA1 Message Date
Raushan Turganbay
8c102e2eb1 Rename _supports_flash_attn_2 in examples and tests (#39471)
* delete `_supports_flash_attn_2` from examples and tests

* simplify docs
2025-07-21 14:02:57 +02:00
Yaswanth Gali
0e1c281745 [Tests] Update model_id in AIMv2 Tests (#39281)
* Update model_id in tests

* fix

---------

Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
2025-07-08 21:46:32 +02:00
Yaswanth Gali
fbdaa7b099 Add Aimv2 model (#36625)
* Model skelton

* changes

* temp push

* changes

* Added support for aimv2-native

* More changes

* More changes

* Stupid mistake correction

* Added config and refactor

* Added vison model

* update

* Refactor for lit variant

* Added Text Model

* Minor fixes

* nits

* update

* Preliminary tests

* More fixes

* Updated tests 🤗

* Refactor

* Updated testcase

* Updated config

* make fixup

* more fixes

* Bug fix and updates

* deadcode

* Fixes

* nit

* up

* Happy CI 

* Reduce LOC

* nit

* nit

* make style

* return_dict refactor

* bug fix

* fix

* doc update

* nit

* make fixup

* Minor update

* _init_weigths modifcation

* update tests

* Minor fixes post review

* Update w.r.t GradientCheckpointingLayer

* docs update

* update

* nit

* Use more Modular 😉

* Change name from AIMv2 to Aimv2

* Nit

* make style

* Add model doc pointer

* make style

* Update model doc section

* updates

* Modify attn mask and interface

* update test

* Final change

* Utilize flash and flex attn

* keep attn mask

* camelcase model name in test file

* Fix docstring

* Fix config warning finally and create_causal_mask

* disable torchscript

* remove unused arg

* remove from tests

* balance model size for tests

* fix device

* tests

* tests

* flaky test

* fix import

---------

Co-authored-by: Cyril Vallez <cyril.vallez@huggingface.co>
Co-authored-by: Cyril Vallez <cyril.vallez@gmail.com>
2025-07-08 11:53:21 +02:00