GitLab

* Add min and max question length option to the tokenizer

* Add corresponding test

* Fix the loss calculation of ProphetNet

* Fix the loss calculation of ProphetNet

Fix the loss calculation of ProphetNet and remove warning

* Update torch.utils.data namespaces to the latest.

* Format

* Update Dataloader.

* Style

* up

* up

* splinter template

* initialize splinter classes

* Splinter Tokenizer

* splinter.rst

* tokenization fixes

* Documentation & some minor variable name changes

* bug fix (added back question_token_id to config) + variable names

* Minor bug fixes + variable name changes

* Fix Splinter references after merge with new transformers

* changes after running make style & quality

* Fix documentation unindent

* Fix doc indentation in tokenization_splinter

* Fix also SplinterTokenizerFast

* Add Splinter to index.rst and README

* Fixdouble whitespace from index.rst

* Fixed index.rst with 'make fix-copies'

* Update docs/source/model_doc/splinter.rst
Co-authored-by: Suraj Patil <surajp815@gmail.com>

* Update docs/source/model_doc/splinter.rst
Co-authored-by: Suraj Patil <surajp815@gmail.com>

* Update docs/source/model_doc/splinter.rst
Co-authored-by: Suraj Patil <surajp815@gmail.com>

* Update docs/source/model_doc/splinter.rst
Co-authored-by: Suraj Patil <surajp815@gmail.com>

* Update src/transformers/models/splinter/__init__.py
Co-authored-by: Suraj Patil <surajp815@gmail.com>

* Added "copied from BERT" comments

* Removing unnexessary code from modeling_splinter

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

* Update src/transformers/models/splinter/configuration_splinter.py
Co-authored-by: Suraj Patil <surajp815@gmail.com>

* Remove references to TF modeling from splinter

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

* Remove unnecessary check

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

* Add differences between Splinter and Bert tokenizers

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

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

* Remove unnecessary check

* Doc formatting

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

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

* bug fix: remove load_tf_weights attribute

* Some minor quality changes

* Update docs/source/model_doc/splinter.rst
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

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

* Change FullyConnectedLayer to SplinterFullyConnectedLayer

* Variable naming

* Reove gather_positions function

* Remove ClassificationHead as it's outdated

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

* Remove hardcoded 102 token id

* Minor style change

* Added "tau" organization to all model identifiers & URLS

* Added tau to the tests as well

* Copy-from comments

* Removed all unnecessary classes (e.g. SplinterForMaskedLM)

* Running make fix-copies

* Bug fix: Further removed unnecessary classes

* Add Splinter to AutoTokenization

* Add an integration test for Splinter

* Removed initialize_new_qass from config - It will be done through different checkpoints

* Removed `initialize_new_qass` from documentation as well

* Added new checkpoint names (`tau/splinter-base-qass` and same for large) in the code

* Minor change to test

* SplinterTokenizer now doesn't abstract from BertTokenizer

* SplinterTokenizerFast also dosn't abstract from Bert

* style and quality

* bug fix: import ing torch in tests only if it's available

* Auto mappings

* Changed copyrights in Splinter's files

* Update src/transformers/models/splinter/configuration_splinter.py
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
Co-authored-by: yuvalkirstain <kirstain.yuval@gmail.com>
Co-authored-by: Suraj Patil <surajp815@gmail.com>
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
Co-authored-by: Lysandre <lysandre.debut@reseau.eseo.fr>
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>

* Starting to optimize ByT5.

* Making ByT5Tokenizer faster.

* Even faster.

* Cleaning up.

* Continue on error

* Specific

* Temporary patch

* Fix frameworks table so it's alphabetical

* Update index.rst

* Don't differentiate when sorting between upper and lower case

* Fix omitted lazy import for xlm-prophetnet

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

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

* Fill mask pipelines test updates.

* Model eval !!

* Adding slow test with actual values.

* Making all tests pass (skipping quite a bit.)

* Doc styling.

* Better doc cleanup.

* Making an explicit test with no pad token tokenizer.

* Typo.

* Fix inconsistency of the last element in hidden_states between PyTorch/Flax GPT2(Neo) (#13102)

* Fix missing elements in outputs tuple

* Apply suggestions from code review
Co-authored-by: Suraj Patil <surajp815@gmail.com>

* Fix local variable 'all_hidden_states' referenced before assignment

* Fix by returning tuple containing None values

* Fix quality
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>
Co-authored-by: Suraj Patil <surajp815@gmail.com>

* Create py.typed

This creates a [py.typed as per PEP 561](https://www.python.org/dev/peps/pep-0561/#packaging-type-information) that should be distributed to mark that the package includes (inline) type annotations.

* Update setup.py

Include py.typed as package data

* Update setup.py

Call `setup(...)` with `zip_safe=False`.

* Fix VisualBERT docs

* Show example notebooks as lists

* Fix style

* conditional declare `TOKENIZER_MAPPING_NAMES` within a `if TYPE_CHECKING` block so that type checkers dont need to evaluate the RHS of the assignment.

this improves performance of the pylance/pyright type checkers

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

* adding missing import

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

* Only report failures on failures

* Fix typo

* Put it everywhere

* allow passing params to image and text feature method

* ifx for hybrid clip as well

* Remove hf_api module and use hugginface_hub

* Style

* Fix to test_fetcher

* Quality

* up

* up

* up

* Change FlaxBartForConditionalGeneration.decode() argument: deterministic -> train

* Also change the parameter name to train for flax marian and mbart
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

* Reactive test fecthers on scheduled test with proper git install

* Proper fetch-depth

* TFDeberta

moved weights to build and fixed name scope

added missing ,

bug fixes to enable graph mode execution

updated setup.py

fixing typo

fix imports

embedding mask fix

added layer names avoid autmatic incremental names

+XSoftmax

cleanup

added names to layer

disable keras_serializable
Distangled attention output shape hidden_size==None
using symbolic inputs

test for Deberta tf

make style

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

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

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

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

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

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

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

removed tensorflow-probability

removed blank line

* removed tf experimental api
+torch_gather tf implementation from @Rocketknight1

* layername DeBERTa --> deberta

* copyright fix

* added docs for TFDeberta & make style

* layer_name change to fix load from pt model

* layer_name change as pt model

* SequenceClassification layername change,
to same as pt model

* switched to keras built-in LayerNormalization

* added `TFDeberta` prefix most layer classes

* updated to tf.Tensor in the docstring

* Doctests

* Limit to 4 decimals

* Try with separate PT/TF tests

* Remove test for TF

* Ellips the predictions

* Doctest continue on failure
Co-authored-by: Sylvain Gugger <sylvain.gugger@gmail.com>

Classification head of AlbertForMultipleChoice uses `hidden_dropout_prob` instead of `classifier_dropout_prob`.  This
is not desirable as we cannot change classifer head dropout probability without changing the dropout probabilities of
the whole model.

* Install git

* Add TF tests

* And last TF test

* Add in commented code too
Co-authored-by: Sylvain Gugger <sylvain.gugger@gmail.com>

* Initialize VisualBERT demo

* Update demo

* Add commented URL

* Update README

* Update README

* Fix doctests for quicktour

* Adapt causal LM exemple

* Remove space

* Fix until summarization

* End of task summary

* Style

* With last changes in quicktour

* Use original key for label in DataCollatorForTokenClassification

DataCollatorForTokenClassification accepts either `label` or `labels` as key for label in it's input. However after padding the label it assigns the padded labels to key `labels`. If originally `label` was used as key than the original upadded labels still remains in the batch. Then at line 192 when we try to convert the batch elements to torch tensor than these original unpadded labels cannot be converted as the labels for different samples have different lengths.

* Fixed style.