GitLab

* [WIP] Proposal for model outputs

* All Bert models

* Make CI green maybe?

* Fix ONNX test

* Isolate ModelOutput from pt and tf

* Formatting

* Add Electra models

* Auto-generate docstrings from outputs

* Add TF outputs

* Add some BERT models

* Revert TF side

* Remove last traces of TF changes

* Fail with a clear error message

* Add Albert and work through Bart

* Add CTRL and DistilBert

* Formatting

* Progress on Bart

* Renames and finish Bart

* Formatting

* Fix last test

* Add DPR

* Finish Electra and add FlauBERT

* Add GPT2

* Add Longformer

* Add MMBT

* Add MobileBert

* Add GPT

* Formatting

* Add Reformer

* Add Roberta

* Add T5

* Add Transformer XL

* Fix test

* Add XLM + fix XLMForTokenClassification

* Style + XLMRoberta

* Add XLNet

* Formatting

* Add doc of return_tuple arg

* Add B I handling to grouping

* Add fix to include separate entity as last token

* move last_idx definition outside loop

* Use first entity in entity group as reference for entity type

* Add test cases

* Take out extra class accidentally added

* Return tf ner grouped test to original

* Take out redundant last entity

* Get last_idx safely
Co-authored-by: ColleterVi <36503688+ColleterVi@users.noreply.github.com>

* Fix first entity comment

* Create separate functions for group_sub_entities and group_entities (splitting call method to testable functions)

* Take out unnecessary last_idx

* Remove additional forward pass test

* Move token classification basic tests to separate class

* Move token classification basic tests back to monocolumninputtestcase

* Move base ner tests to nerpipelinetests

* Take out unused kwargs

* Add back mandatory_keys argument

* Add unitary tests for group_entities in _test_ner_pipeline

* Fix last entity handling

* Fix grouping fucntion used

* Add typing to group_sub_entities and group_entities
Co-authored-by: ColleterVi <36503688+ColleterVi@users.noreply.github.com>

* tf_train

* adapt timing for tpu

* fix timing

* fix timing

* fix timing

* fix timing

* update notebook

* add tests

improve unittests for finetuning, especially w.r.t testing frozen parameters
fix freeze_embeds for T5
add streamlit setup.cfg

[Almost all TF models] TF clean up: add missing CLM / MLM loss; fix T5 naming and keras compile (#5395)

* add first version of clm tf

* make style

* add more tests for bert

* update tf clm loss

* fix tests

* correct tf ner script

* add mlm loss

* delete bogus file

* clean tf auto model + add tests

* finish adding clm loss everywhere

* fix training in distilbert

* fix flake8

* save intermediate

* fix tf t5 naming

* remove prints

* finish up

* up

* fix tf gpt2

* fix new test utils import

* fix flake8

* keep backward compatibility

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

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

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

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

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

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

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

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

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

* fix test imports

* fix max_length

* style

* fix tests

* Passing all but one torchscript test

* Style

* move comment

* remove unneeded assert

* beginning of dpr modeling

* wip

* implement forward

* remove biencoder + better init weights

* export dpr model to embed model for nlp lib

* add new api

* remove old code

* make style

* fix dumb typo

* don't load bert weights

* docs

* docs

* style

* move the `k` parameter

* fix init_weights

* add pretrained configs

* minor

* update config names

* style

* better config

* style

* clean code based on PR comments

* change Dpr to DPR

* fix config

* switch encoder config to a dict

* style

* inheritance -> composition

* add messages in assert startements

* add dpr reader tokenizer

* one tokenizer per model

* fix base_model_prefix

* fix imports

* typo

* add convert script

* docs

* change tokenizers conf names

* style

* change tokenizers conf names

* minor

* minor

* fix wrong names

* minor

* remove unused convert functions

* rename convert script

* use return_tensors in tokenizers

* remove n_questions dim

* move generate logic to tokenizer

* style

* add docs

* docs

* quality

* docs

* add tests

* style

* add tokenization tests

* DPR full tests

* Stay true to the attention mask building

* update docs

* missing param in bert input docs

* docs

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

* Default decoder inputs to encoder ones for T5 if neither are specified.

* Fixing typo, now all tests are passing.

* Changing einsum to operations supported by onnx

* Adding a test to ensure T5 can be exported to onnx op>9

* Modified test for onnx export to make it faster

* Styling changes.

* Styling changes.

* Changing notation for matrix multiplication
Co-authored-by: Abel Riboulot <tkai@protomail.com>

* fix attention mask

* fix slow test

* refactor attn masks

* fix fp16 generate test

* Added data collator for XLNet language modeling and related calls

Added DataCollatorForXLNetLanguageModeling in data/data_collator.py
to generate necessary inputs for language modeling training with
XLNetLMHeadModel. Also added related arguments, logic and calls in
examples/language-modeling/run_language_modeling.py.

Resolves: #4739, #2008 (partially)

* Changed name to `DataCollatorForPermutationLanguageModeling`

Changed the name of `DataCollatorForXLNetLanguageModeling` to the more general `DataCollatorForPermutationLanguageModelling`.
Removed the `--mlm` flag requirement for the new collator and defined a separate `--plm_probability` flag for its use.
CTRL uses a CLM loss just like GPT and GPT-2, so should work out of the box with this script (provided `past` is taken care of
similar to `mems` for XLNet).
Changed calls and imports appropriately.

* Added detailed comments, changed variable names

Added more detailed comments to `DataCollatorForPermutationLanguageModeling` in `data/data_collator.py` to explain working. Also cleaned up variable names and made them more informative.

* Added tests for new data collator

Added tests in `tests/test_trainer.py` for DataCollatorForPermutationLanguageModeling based on those in DataCollatorForLanguageModeling. A specific test has been added to check for odd-length sequences.

* Fixed styling issues

* BertTokenizerFast - Do not specify strip_accents by default

* Bump tokenizers to new version

* Add test for AddedToken serialization

* Exposing prepare_for_model for both slow & fast tokenizers

* Update method signature

* The traditional style commit

* Hide the warnings behind the verbose flag

* update default truncation strategy and prepare_for_model

* fix tests and prepare_for_models methods
Co-authored-by: Thomas Wolf <thomwolf@users.noreply.github.com>

* Changed expected_output_ids in TransfoXL generation test to match #4826 generation PR.

* making black happy

* making isort happy

* fix conflicts

* fix

* happy rebasing

* allow tensor label inputs to default collator

* replace try/except with type check

* refactor naming

* add small slow test

* refactor

* refactor naming

* rename selected to extra

* big global attention refactor

* make style

* refactor naming

* save intermed

* refactor functions

* finish function refactor

* fix tests

* fix longformer

* fix longformer

* fix longformer

* fix all tests but one

* finish longformer

* address sams and izs comments

* fix transpose

* Added PipelineException
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>

* fill-mask pipeline raises exception when more than one mask_token detected.
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>

* Put everything in a function.
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>

* Added tests on pipeline fill-mask when input has != 1 mask_token
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>

* Fix numel() computation for TF
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>

* Addressing PR comments.
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>

* Remove function typing to avoid import on specific framework.
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>

* Quality.
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>

* Retry typing with @julien-c tip.
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>

* Quality².
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>

* Simplify fill-mask mask_token checking.
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>

* Trigger CI

* first doc version

* add benchmark docs

* fix typos

* improve README

* Update docs/source/benchmarks.rst
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>

* fix naming and docs
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>

* remove references to old API in docstring - update data processors

* style

* fix tests - better type checking error messages

* better type checking

* include awesome fix by @LysandreJik for #5310

* updated doc and examples

* Add new parameter `pad_to_multiple_of` on tokenizers.

* unittest for pad_to_multiple_of

* Add .name when logging enum.

* Fix missing .items() on dict in tests.

* Add special check + warning if the tokenizer doesn't have proper pad_token.

* Use the correct logger format specifier.

* Ensure tokenizer with no pad_token do not modify the underlying padding strategy.

* Skip test if tokenizer doesn't have pad_token

* Fix RobertaTokenizer on empty input

* Format.
Signed-off-by: Morgan Funtowicz <funtowiczmo@gmail.com>

* fix and updating to simpler API
Co-authored-by: Thomas Wolf <thomwolf@users.noreply.github.com>

* Refactor code samples

* Test docstrings

* Style

* Tokenization examples

* Run rust of tests

* First step to testing source docs

* Style and BART comment

* Test the remainder of the code samples

* Style

* let to const

* Formatting fixes

* Ready for merge

* Fix fixture + Style

* Fix last tests

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

* Addressing @sgugger's comments + Fix MobileBERT in TF
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* avoid recursion in id checks for fast tokenizers

* better typings and fix #5232

* align slow and fast tokenizers behaviors for Roberta and GPT2

* style and quality

* fix tests - improve typings

[Tokenization] Fix #5181 - make #5155 more explicit - move back the default logging level in tests to WARNING (#5252)

* fix-5181

Padding to max sequence length while truncation to another length was wrong on slow tokenizers

* clean up and fix #5155

* fix XLM test

* Fix tests for Transfo-XL

* logging only above WARNING in tests

* switch slow tokenizers tests in @slow

* fix Marian truncation tokenization test

* style and quality

* make the test a lot faster by limiting the sequence length used in tests

* update tests for fast tokenizers + fix small bug in saving/loading

* better tests on serialization

* fixing serialization

* comment cleanup

* Cleaning TensorFlow models

Update all classes


stylr

* Don't average loss

* fix use cache

* add bart use cache

* fix bart

* finish bart

* add benchmark for all kinds of models

* improved import

* delete bogus files

* make style

* Add return lengths

* make pad a bit more flexible so it can be used as collate_fn

* check all kwargs sent to encoding method are known

* fixing kwargs in encodings

* New AddedToken class in python

This class let you specify specifique tokenization behaviors for some special tokens. Used in particular for GPT2 and Roberta, to control how white spaces are stripped around special tokens.

* style and quality

* switched to hugginface tokenizers library for AddedTokens

* up to tokenizer 0.8.0-rc3 - update API to use AddedToken state

* style and quality

* do not raise an error on additional or unused kwargs for tokenize() but only a warning

* transfo-xl pretrained model requires torch

* Update src/transformers/tokenization_utils.py
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>
Co-authored-by: Lysandre Debut <lysandre@huggingface.co>