GitLab

Co-authored-by: Lysandre Debut <hi@lysand.re>
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

[Wav2vec2 + LM Test] Improve wav2vec2 with lm tests and make torch version dependent for now (#18749)

* add first generation tutorial

* remove generation

* make version dependent expected values

* Apply suggestions from code review

* Update tests/models/wav2vec2_with_lm/test_processor_wav2vec2_with_lm.py

* fix typo

* add first generation tutorial

* VisionEnocderDecoder gradient checkpointing

* remove generation

* add tests

* Automatic detection for framework to use when exporting to ONNX

* Log message change

* Incorporating PR comments, adding unit test

* Adding tf for pip install for run_tests_onnxruntime CI

* Restoring past changes to circleci yaml and test_onnx_v2.py, tests moved to tests/onnx/test_features.py

* Fixup

* Adding test to fetcher

* Updating circleci config to log more

* Changing test class name

* Comment typo fix in tests/onnx/test_features.py
Co-authored-by: lewtun <lewis.c.tunstall@gmail.com>

* Moving torch_str/tf_str to self.framework_pt/tf

* Remove -rA flag in circleci config
Co-authored-by: lewtun <lewis.c.tunstall@gmail.com>

* Implement ONNX support for Longformer

Fix repo consistency check complaints

Fix value mismatches

Add pooler output for default model

Increase validation atol to accommodate multiple-choice error

Fix copies

Fix chunking for longer sequence lengths

Add future comment

* Fix issue in mask_invalid_locations

* Remove torch imports in configuration_longformer

* Change config access to fix LED

* Push opset version to support tril

* Work in review comments (mostly style)

* Add Longformer to ONNX tests

add warning to let the user know that the `__call__` method is faster than `encode` + `pad` for a fast tokenizer (#18693)

* add warning to let the user know that the  method is slower that  for a fast tokenizer

* user warnings

* fix layoutlmv2

* fix layout*

* change warnings into logger.warning

* Add TFXGLM models 

* Add todo: self.supports_xla_generation = False
Co-authored-by: Daniel Stancl <stancld@Daniels-MacBook-Pro.local>
Co-authored-by: Daniel Stancl <stancld@daniels-mbp.home>
Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>
Co-authored-by: Daniel <daniel.stancl@rossum.ai>
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* Update methods to optionally rescale
This is necessary to allow for casting our images / videos to numpy arrays within the feature extractors' call. We want to do this to make sure the behaviour is as expected when flags like  are False. If some transformations aren't applied, then the output type can't be unexpected e.g. a list of PIL images instead of numpy arrays.

* Cast images to numpy arrays in call to enable consistent behaviour with different configs

* Remove accidental clip changes

* Update tests to reflect the scaling logic
We write a generic  function to handle rescaling of our arrays. In order for the API to be intuitive, we take some factor c and rescale the image values by that. This means, the rescaling done in normalize and to_numpy_array are now done with array * (1/255) instead of array / 255. This leads to small differences in the resulting image. When testing, this was in the order of 1e-8, and so deemed OK

Co-authored-by: lewtun <lewis.c.tunstall@gmail.com>
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

* bnb minor modifications

- refactor documentation
- add troubleshooting README
- add PyPi library on DockerFile

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

* Apply suggestions from code review

* Apply suggestions from code review

* Apply suggestions from code review

* put in one block

- put bash instructions in one block

* update readme

- refactor a bit hardware requirements

* change text a bit

* Apply suggestions from code review
Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com>

* apply suggestions
Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com>

* add link to paper

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

* Update tests/mixed_int8/README.md

* Apply suggestions from code review

* refactor a bit

* add instructions Turing & Amperer
Co-authored-by: Stas Bekman <stas00@users.noreply.github.com>

* add A6000

* clarify a bit

* remove small part

* Update tests/mixed_int8/README.md
Co-authored-by: Stas Bekman <stas00@users.noreply.github.com>
Co-authored-by: Yih-Dar <2521628+ydshieh@users.noreply.github.com>

* Update BLOOM parameter counts

* Update BLOOM parameter counts

* First draft

* Improve script

* Update script

* Make conversion work

* Add final_layer_norm attribute to Swin's config

* Add DonutProcessor

* Convert more models

* Improve feature extractor and convert base models

* Fix bug

* Improve integration tests

* Improve integration tests and add model to README

* Add doc test

* Add feature extractor to docs

* Fix integration tests

* Remove register_buffer

* Fix toctree and add missing attribute

* Add DonutSwin

* Make conversion script work

* Improve conversion script

* Address comment

* Fix bug

* Fix another bug

* Remove deprecated method from docs

* Make Swin and Swinv2 untouched

* Fix code examples

* Fix processor

* Update model_type to donut-swin

* Add feature extractor tests, add token2json method, improve feature extractor

* Fix failing tests, remove integration test

* Add do_thumbnail for consistency

* Improve code examples

* Add code example for document parsing

* Add DonutSwin to MODEL_NAMES_MAPPING

* Add model to appropriate place in toctree

* Update namespace to appropriate organization
Co-authored-by: Niels Rogge <nielsrogge@Nielss-MacBook-Pro.local>

* Supporting seq2seq models for `bitsandbytes` integration

- `bitsandbytes` integration supports now seq2seq models
- check if a model has tied weights as an additional check

* small modification

- tie the weights before looking at tied weights!

* validate generate model_kwargs

* generate tests -- not all models have an attn mask

* initial commit

* add small test

* add cross pt tf flag to test

* fix quality

* style

* update test with new repo

* fix failing test

* update

* fix wrong param ordering

* style

* update based on review

* update related to recent new caching mechanism

* quality

* Update based on review
Co-authored-by: sgugger <sylvain.gugger@gmail.com>

* quality and style

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

* onnx config for clip

* default opset as 14

* changes from the original repo

* input values order fix

* outputs fix

* remove unused import

* ran make fix-copies

* black format

* review comments: forward ref, import fix, model change revert, .to cleanup

* make style

* formatting fixes

* revert groupvit

* comment for cast to int32

* comment fix

* make .T as .t() for onnx conversion

* ran make fix-copies

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

* fix copies

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

* Use commit hash to look in cache instead of calling head

* Add tests

* Add attr for local configs too

* Stupid typos

* Fix tests

* Update src/transformers/utils/hub.py
Co-authored-by: Julien Chaumond <julien@huggingface.co>

* Address Julien's comments
Co-authored-by: Julien Chaumond <julien@huggingface.co>

* first commit

* correct replace function

* add final changes

- works like charm!
- cannot implement tests yet
- tested

* clean up a bit

* add bitsandbytes dependencies

* working version

- added import function
- added bitsandbytes utils file

* small fix

* small fix

- fix import issue

* fix import issues

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

* refactor a bit

- move bitsandbytes utils to utils
- change comments on functions

* reformat docstring

- reformat docstring on init_empty_weights_8bit

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

* revert bad formatting

* change to bitsandbytes

* refactor a bit

- remove init8bit since it is useless

* more refactoring

- fixed init empty weights issue
- added threshold param

* small hack to make it work

* Update src/transformers/modeling_utils.py

* Update src/transformers/modeling_utils.py

* revmoe the small hack

* modify utils file

* make style + refactor a bit

* create correctly device map

* add correct dtype for device map creation

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

* apply suggestions

- remove with torch.grad
- do not rely on Python bool magic!

* add docstring

 - add docstring for new kwargs

* add docstring

- comment `replace_8bit_linear` function
- fix weird formatting

* - added more documentation
- added new utility function for memory footprint tracking
- colab demo to add

* few modifs

- typo doc
- force cast into float16 when load_in_8bit is enabled

* added colab link

* add test architecture + docstring a bit

* refactor a bit testing class

* make style + refactor a bit

* enhance checks

- add more checks
- start writing saving test

* clean up a bit

* male style

* add more details on doc

* add more tests

- still needs to fix 2 tests

* replace by "or"

- could not fix it from GitHub GUI
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* refactor a bit testing code + add readme

* make style

* fix import issue

* Update src/transformers/modeling_utils.py
Co-authored-by: Michael Benayoun <mickbenayoun@gmail.com>

* add few comments

* add more doctring + make style

* more docstring

* raise error when loaded in 8bit

* make style

* add warning if loaded on CPU

* add small sanity check

* fix small comment

* add bitsandbytes on dockerfile

* Improve documentation

- improve documentation from comments

* add few comments

* slow tests pass on the VM but not on the CI VM

* Fix merge conflict

* make style

* another test should pass on a multi gpu setup

* fix bad import in testing file

* Fix slow tests

- remove dummy batches
- no more CUDA illegal memory errors

* odify dockerfile

* Update docs/source/en/main_classes/model.mdx

* Update Dockerfile

* Update model.mdx

* Update Dockerfile

* Apply suggestions from code review

* few modifications

- lm head can stay on disk/cpu
- change model name so that test pass

* change test value

- change test value to the correct output
- torch bmm changed to baddmm in bloom modeling when merging

* modify installation guidelines

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

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

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

* replace `n`by `name`

* merge `load_in_8bit` and `low_cpu_mem_usage`

* first try - keep the lm head in full precision

* better check

- check the attribute `base_model_prefix` instead of computing the number of parameters

* added more tests

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

* Merge branch 'integration-8bit' of https://github.com/younesbelkada/transformers

 into integration-8bit

* improve documentation

- fix typos for installation
- change title in the documentation
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
Co-authored-by: Michael Benayoun <mickbenayoun@gmail.com>

* Adding a new `align_to_words` param to qa pipeline.

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

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

* update features

* MT5OnnxConfig added with updated with tests and docs

* fix imports

* fix onnc_config_cls for mt5

Co-authored-by: Thomas Chaigneau <thomas.deeptools.ai>

* feat: add the data2vec conf that are missing https://huggingface.co/docs/transformers/serialization

* fix: wrong config

* Clean up utils.hub

* Remove imports

* More fixes

* Last fix

* [DX fix] Fixing QA pipeline streaming a dataset.

QuestionAnsweringArgumentHandler would iterate over the whole dataset
effectively killing all properties of the pipeline.
This restores nice properties when using `Dataset` or `Generator` since
those are meant to be consumed lazily.

* Handling TF better.

* zero chance anyone's using that constant no?

* `transformers-cli login` => `huggingface-cli login`

* `transformers-cli repo create` => `huggingface-cli repo create`

* `make style`

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

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

* Draft new cached_file

* Initial draft for config and model

* Small fixes

* Fix first batch of tests

* Look in cache when internet is down

* Fix last tests

* Bad black, not fixing all quality errors

* Make diff less

* Implement change for TF and Flax models

* Add tokenizer and feature extractor

* For compatibility with main

* Add utils to move the cache and auto-do it at first use.

* Quality

* Deal with empty commit shas

* Deal with empty etag

* Address review comments

* Fix pipeline tests

* Make sure all pipelines tests run with init changes

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

* Adding a better error message when the model is improperly configured

within transformers.

* Update src/transformers/pipelines/__init__.py

* Black version.

* Overriding task aliases so that tokenizer+feature_extractor

values are correct.

* Fixing task aliases by overriding their names early

* X.

* Fixing feature-extraction.

* black again.

* Normalizing `translation` too.

* Fixing last few corner cases.

translation need to use its non normalized name (translation_XX_to_YY,
so that the task_specific_params are correctly overloaded).
This can be removed and cleaned up in a later PR.

`speech-encode-decoder` actually REQUIRES to pass a `tokenizer` manually
so the error needs to be discarded when the `tokenizer` is already
there.

* doc-builder fix.

* Fixing the real issue.

* Removing dead code.

* Do not import the actual config classes.

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

* First draft

* Add VideoMAEForVideoClassification

* Improve conversion script

* Add VideoMAEForPreTraining

* Add VideoMAEFeatureExtractor

* Improve VideoMAEFeatureExtractor

* Improve docs

* Add first draft of model tests

* Improve VideoMAEForPreTraining

* Fix base_model_prefix

* Make model take pixel_values of shape (B, T, C, H, W)

* Add loss computation of VideoMAEForPreTraining

* Improve tests

* Improve model testsé

* Make all tests pass

* Add VideoMAE to main README

* Add tests for VideoMAEFeatureExtractor

* Add integration test

* Improve conversion script

* Rename patch embedding class

* Remove VideoMAELayer from init

* Update design of patch embeddings

* Improve comments

* Improve conversion script

* Improve conversion script

* Add conversion of pretrained model

* Add loss verification of pretrained model

* Add loss verification of unnormalized targets

* Add integration test for pretraining model

* Apply suggestions from code review

* Fix bug to make feature extractor resize only shorter edge

* Address more comments

* Improve normalization of videos

* Add doc examples

* Move constants to dedicated script

* Remove scripts

* Transfer checkpoints, fix docs

* Update script

* Update image mean and std

* Fix doc tests

* Set return_tensors to NumPy by default

* Revert the previous change
Co-authored-by: Niels Rogge <nielsrogge@Nielss-MacBook-Pro.local>