GitLab

* fix_torch_device_generate_test

* remove @

* start adding tests

* correct wav2vec2 pretraining

* up

* up
Co-authored-by: Patrick von Platen <patrick@huggingface.co>

* fix_torch_device_generate_test

* remove @

* push

* fix marian

* fix

* up

* [trainer] release tmp memory in checkpoint load

* 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>

* Deprecate TFTrainer

* Style pass

* Base test

* More test

* Fix mistake

* Add a docstring change

* Add doc ignore

* Add changes

* Add recursive dep search

* Add recursive dep search

* save

* Finalize test mapping

* Fix bug

* Print prettier

* Ignore comments and empty lines

* Make script runnable from anywhere

* Need dev install

* Like that

* Adapt

* Add as artifact

* Try on torch tests

* Fix yaml error

* Install GitPython

* Apply everywhere

* Be more defensive

* Revert to all tests if something is wrong

* Install GitPython

* Test if there are tests before launching.

* Fixes

* Fixes

* Fixes

* Fixes

* Bash syntax is horrible

* Be less stupid

* Try differently

* Typo

* Typo

* Typo

* Style

* Better name

* Escape quotes

* Ignore black unhelpful re-formatting

* Not a docstring

* Deal with inits in dependency map

* Run all tests once PR is merged.

* Add last job

* Apply suggestions from code review
Co-authored-by: Stas Bekman <stas00@users.noreply.github.com>

* Stronger dependencies gather

* Ignore empty lines too!

* Clean up

* Fix quality
Co-authored-by: Stas Bekman <stas00@users.noreply.github.com>

* We need to provide mask_time_indices to `_mask_hidden_states` to avoid applying the mask two times

* apply the same to wav2vec2

* Uniformize the style between hubert and wav2vec2

* fix tf as well
Co-authored-by: patrickvonplaten <patrick.v.platen@gmail.com>

* Global 60-300 seconds timeout

* Add verbose option

* [skip ci] typo

* zero_to_fp32 tests

* args change

* remove unnecessary work

* use transformers.trainer_utils.get_last_checkpoint

* document the new features

* cleanup

* wip

* fix fsmt

* add bert

* cleanup

* add xlm-roberta

* electra works

* cleanup

* sync

* split off the model zoo tests

* cleanup

* cleanup

* cleanup

* cleanup

* reformat

* cleanup

* casing

* deepspeed>=0.4.3

* adjust distilbert

* Update docs/source/main_classes/deepspeed.rst
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

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

* fix_torch_device_generate_test

* remove @

* correct greedy search

* save intertmed

* add final logits bias

* correct

* up

* add more tests

* fix another bug

* finish tests

* finish marian tests

* up
Co-authored-by: Patrick von Platen <patrick@huggingface.co>

* change deprecation to be easily actionable

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

* rework as suggested

* one warning together

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

* Add option to load a pretrained model with mismatched shapes

* Fail at loading when mismatched shapes in Flax

* Fix tests

* Update src/transformers/modeling_flax_utils.py
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* Address review comments
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* fix_torch_device_generate_test

* remove @

* fix docs

* PoC, it pickles!

* Remove old method.

* Apply to every auto object

Co-authored-by: Lewis Bails <Lewis.Bails@infomedia.dk>

* TFHubert

* Update with TFWav2Vec Bug Fixes

* Add OOV Error

* Feedback changes

* Fix kwargs call

* fix_torch_device_generate_test

* remove @

* up

* fix_torch_device_generate_test

* remove @

* finish better examples for marian flax

* fix_torch_device_generate_test

* remove @

* fix mt5 auto

* Pass model_kwargs when loading a model in pipeline

* Add test for model_kwargs parameter of pipeline()

* Rewrite test to not download model

* Fix failing style checks

* Base test

* More test

* Fix mistake

* Add a docstring change

* Add doc ignore

* Simplify logic for unk token in Unigram tokenizers

* Remove changes from otehr branch

* fix_torch_device_generate_test

* remove @

* add marian

* finish make style

* add model

* add docs

* add test

* add integration tests

* up

* solve bug

* correct tests

* correct some tests

* Apply suggestions from code review
Co-authored-by: Suraj Patil <surajp815@gmail.com>
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* correct adapt marian

* finish
Co-authored-by: Patrick von Platen <patrick@huggingface.co>
Co-authored-by: Suraj Patil <surajp815@gmail.com>
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* This will reduce "Already borrowed error":

Original issue https://github.com/huggingface/tokenizers/issues/537



The original issue is caused by transformers calling many times
mutable functions on the rust tokenizers.
Rust needs to guarantee that only 1 agent has a mutable reference
to memory at a given time (for many reasons which don't need explaining
here). Usually, the rust compiler can guarantee that this property is
true at compile time.

Unfortunately, this is impossible for Python to do that, so PyO3, the
bridge between rust and python used by `tokenizers`, will change the
compile guarantee for a dynamic guarantee, so if multiple agents try
to have multiple mutable borrows at the same time, then the runtime will
yell with "Already borrowed".

The proposed fix here in transformers, is simply to reduce the actual
number of calls that really need mutable borrows. By reducing them,
we reduce the risk of running into "Already borrowed" error.
The caveat is now we add a call to read the current configuration of the
`_tokenizer`, so worst case we have 2 calls instead of 1, and best case
we simply have 1 + a Python comparison of a dict (should be negligible).

* Adding a test.

* trivial error :(.

* Update tests/test_tokenization_fast.py
Co-authored-by: SaulLu <55560583+SaulLu@users.noreply.github.com>

* Adding reference to original issues in the tests.

* Update the tests with fast tokenizer.
Co-authored-by: SaulLu <55560583+SaulLu@users.noreply.github.com>

* [model.from_pretrained] raise exception early on failed load

Currently if `load` pretrained weights fails in `from_pretrained`, we first print a whole bunch of successful messages and then fail - this PR puts the exception first to avoid all the misleading messages.

* style
Co-authored-by: Suraj Patil <surajp815@gmail.com>

* Fixing the pipeline optimization by rescaling the logits first.

* Add test for target equivalence
Co-authored-by: Lysandre <lysandre.debut@reseau.eseo.fr>