GitLab

* fix the deepspeed test failures

* apex fix

* FSDP save ckpt fix

* Update src/transformers/trainer.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

---------
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* fix bnb config json serialization

* forward contrib credits from discussions

---------
Co-authored-by: Andrechang <Andrechang@users.noreply.github.com>

* preventllama fast from returning token type ids

* remove type hints

* normalised False

* fix

* fix

---------
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

* Fix typo in Llama docstrings
Signed-off-by: Serge Panev <spanev@nvidia.com>

* Update
Signed-off-by: Serge Panev <spanev@nvidia.com>

* make style
Signed-off-by: Serge Panev <spanev@nvidia.com>

---------
Signed-off-by: Serge Panev <spanev@nvidia.com>

* add trust_remote_code option

* require_torch

[`GPT2`] Add correct keys on `_keys_to_ignore_on_load_unexpected` on all child classes of `GPT2PreTrainedModel` (#24113)

* add correct keys on `_keys_to_ignore_on_load_unexpected`

* oops

* fix get_keys_to_not_convert funct

* Fix style

* v1

* some refactor

- add ST format as well

* fix

* add `ADAPTER_WEIGHTS_NAME` & `ADAPTER_SAFE_WEIGHTS_NAME`

* Add AzureOpenAiAgent

* quality

* Update src/transformers/tools/agents.py
Co-authored-by: Lysandre Debut <lysandre.debut@reseau.eseo.fr>

---------
Co-authored-by: Lysandre Debut <lysandre.debut@reseau.eseo.fr>

* Min accelerate

* Also min version

* Min accelerate

* Also min version

* To different minor version

* Empty

* fix mixed precision prep during eval only mode

* update to address comments

* update to reflect the changes in accelerate

* Do not prepare lr scheduler as it as the right number of steps

* Trigger CI

* Trigger CI

* Trigger CI

* Add fake comment

* Remove fake comment

* Trigger CI please!

* fix executable batch size issue

* fix

* undo

fix base workflow name

* Fix expected value in tests of the test fetcher

* Fix trigger for repo util tests

* Let's see if we can use the smallest possible dummies

* Make GPT-2's dummies a little longer

* Just use (1,2) as the default shape

* Update other dummies in sync

* Correct imports for Keras 2.13

* Shrink the Wav2Vec2 dummies

* fix

* fix

---------
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

* fix skip modules test

* oops

* address comments

Fix is_optimum_neuron_available

add `safe_serialization` in push_to_hub

* support PEFT models when saving the model using trainer

* fixup

* Add support for non-rust implemented tokenization for `__getitem__` method.

* Update for error message on adding new sub-branch for `__item__` method.

---------
Co-authored-by: liuyang17 <liuyang17@zhihu.com>

* [Wav2Vec2] Fix torch srcipt

* fix more

increase atol

* Fix model load when it has both code on the Hub and locally

* Add input check with timeout

* Add tests

* Apply suggestions from code review
Co-authored-by: Lysandre Debut <lysandre.debut@reseau.eseo.fr>

* Some non-saved stuff

* Add feature extractors

* Add image processor

* Add model

* Add processor and tokenizer

* Reduce timeout

---------
Co-authored-by: Lysandre Debut <lysandre.debut@reseau.eseo.fr>

* Fix device placement for model-parallelism in generate for encoder/decoders

* Remove debug statements

fix
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

* Use new parametrization based weight norm if available

See https://github.com/pytorch/pytorch/pull/103001

Signed-off-by: Edward Z. Yang <ezyang@meta.com>

* handle copies
Signed-off-by: Edward Z. Yang <ezyang@meta.com>

* black
Signed-off-by: Edward Z. Yang <ezyang@meta.com>

---------
Signed-off-by: Edward Z. Yang <ezyang@meta.com>

* A fun new PR where I break the entire codebase again

* A fun new PR where I break the entire codebase again

* Handle cross-attention

* Move calls to model(model.dummy_inputs) to the new build() method

* Seeing what fails with the build context thing

* make fix-copies

* Let's see what fails with new build methods

* Fix the pytorch crossload build calls

* Fix the overridden build methods in vision_text_dual_encoder

* Make sure all our build methods set self.built or call super().build(), which also sets it

* make fix-copies

* Remove finished TODO

* Tentatively remove unneeded (?) line

* Transpose b in deberta correctly and remove unused threading local

* Get rid of build_with_dummies and all it stands for

* Rollback some changes to TF-PT crossloading

* Correctly call super().build()

Oops

* Make the default dummies (2, 2) instead of (3, 3)

* Fix for Funnel

* Actually fix Funnel