GitLab

* add post_process_semantic_segmentation method to BeiTFeatureExtractor

* Add documentation of Trainer.create_model_card

* Expand to TF version

need to find out solution for following cases
     *if we need to use trial in model_init, how to do it for non-main rank, sync the model with rank0 in app?
     *how to use optuna prune feature for DDP, if we do it in rank0, how does other rank know it.
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>
Signed-off-by: Wang, Yi A <yi.a.wang@intel.com>

* Add tips

* Add BEiT figure

* Fix URL

* Move tip to start

* Add tip to TF model as well
Co-authored-by: Niels Rogge <nielsrogge@Nielss-MacBook-Pro.local>

* Added type hints to ResNetForImageClassification

* Resolved check_repository_consistency failure issue

Running fix-copies changed the type hints for RegNetForImageClassification in modeling_regnet.py file

* Added type hints for TFMPNetModel

* Added type hints for TFMPNetForMaskedLM

* Added type hints for TFMPNetForSequenceClassification

* Added type hints for TFMPNetForMultipleChoice

* Added type hints for TFMPNetForTokenClassification

* Added Type hints for TFMPNetForQuestionAnswering

* Added Type hints for VIT MAE

* Ran make fixup

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

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

* german autoclass

* Update _toctree.yml

Bumps [mako](https://github.com/sqlalchemy/mako) from 1.2.0 to 1.2.2.
- [Release notes](https://github.com/sqlalchemy/mako/releases)
- [Changelog](https://github.com/sqlalchemy/mako/blob/main/CHANGES)
- [Commits](https://github.com/sqlalchemy/mako/commits

)

---
updated-dependencies:
- dependency-name: mako
  dependency-type: direct:production
...
Signed-off-by: dependabot[bot] <support@github.com>
Signed-off-by: dependabot[bot] <support@github.com>
Co-authored-by: dependabot[bot] <49699333+dependabot[bot]@users.noreply.github.com>

* Fix tokenizer load from one file

* Add a test

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

* added type hints pytorch unispeech

* added type hints pytorch  MPNet

* added type hints nystromformer

* resolved copy inconsistencies

* make fix-copies
Co-authored-by: matt <rocketknight1@gmail.com>

* resized models that we can actually load

* separate embeddings check

* add test for embeddings out of bounds

* add fake slows

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

* Add CLIP to zero-shot-image-classification

* Make mapping private as it's not used for AutoClassing

* Tests conditional run

* Syntax

* Deps

* Try early exit

* Another way

* Test with no tests to run

* Test all

* Typo

* Try this way

* With tests to run

* Mostly finished

* Typo

* With a modification in one file only

* No change, no tests

* Final cleanup

* Address review comments

* Adding cast to fp32 in convnext layernorm to prevent rounding errors in the case of fp16 input

* Trigger CI

* Support for ConvNext

* Support for Wav2Vec2

* Support for Resnet

* Fix small issue in test_modeling_convnext

- set `use_cache` to `True` for consistency with other `transformers` models

* Adds package and requirement spec output to version check exception

It's difficult to understand what package is affected when `got_ver`
here comes back None, so output the requirement and the package. The
requirement probably contains the package but let's output both for good
measure.

Non-exhaustive references for this problem aside from my own encounter:

* https://stackoverflow.com/questions/70151167/valueerror-got-ver-is-none-when-importing-tensorflow
* https://discuss.huggingface.co/t/valueerror-got-ver-is-none/17465
* https://github.com/UKPLab/sentence-transformers/issues/1186
* https://github.com/huggingface/transformers/issues/13356

I speculate that the root of the error comes from a conflict of
conda-managed and pip-managed Python packages but I've not yet proven
this.

* Combines version presence check and streamlines exception message

See also: https://github.com/huggingface/transformers/pull/18702#discussion_r953223275

Co-authored-by: Stas Bekman <stas@stason.org>

* Enable torchdynamo tests

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

* Fix CI for custom tokenizers

* Add nightly tests

* Run CI, run!

* Fix paths

* Typos

* Fix test

* Fixing OPT fast tokenizer option.

* Remove dependency on `pt`.

* Move it to GPT2 tokenization tests.

* Added a few tests.

* init PR

* optimize top p and add edge case

* styling

* style

* revert tf and flax test

* add edge case test for FLAX and TF

* update doc with smallest set sampling for top p

* make style

* Override save() to use the serving signature as the default

* Replace int32 with int64 in all our serving signatures

* Remember one very important line so as not to break every test at once

* Dtype fix for TFLED

* dtype fix for shift_tokens_right in general

* Dtype fixes in mBART and RAG

* Fix dtypes for test_unpack_inputs

* More dtype fixes

* Yet more mBART + RAG dtype fixes

* Yet more mBART + RAG dtype fixes

* Add a check that the model actually has a serving method

* Fix test_save_load for TFViTMAEModelTest
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

* Updated test values

The image segmentation pipeline tests - tests/pipelines/test_pipelines_image_segmentation.py - were failing after the merging of #1849  (49e44b21). This was due to the difference in rescaling. Previously the images were rescaled by `image = image / 255`. In the new commit, a `rescale` method was added, and images rescaled using `image = image * scale`. This was known to cause small differences in the processed images (see
[PR comment](https://github.com/huggingface/transformers/pull/18499#discussion_r940347575)).

Testing locally, changing the `rescale` method to divide by a scale factor (255) resulted in the tests passing. It was therefore decided the test values could be updated, as there was no logic difference between the commits.

* Use double quotes, like previous example

* Fix up