GitLab

* fix inits

* fix embed dtype

* fix embed dtype

* add test to check default dtype

* quality

* add type conversion methods for flax models

* more robust casting

* cast sinusoidal positions

* update pegasus

* update albert

* update test

* make sure dtype is passed to every module

* style

* fix electra dense

* fix t5

* quality

* add more tests

* better name

* use the dtype for lm head computation

* fix albert

* style

* fix albert embed dtype

* more tests

* fix vision enc-dec

* cleanup

* fix embed dtype pegasus

* fix default param test

* doc

* update template

* fix final_logits_bias dtype

* Apply suggestions from code review
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* fix doc

* fix doc

* add detailed docstring for dtype parameter

* remove un-necessary import
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* Update postporcessing accordingly to use SQuAD metric.

* Update assets accordingly based on SQuAD metrics.

* Fix function naming error.

* Update Transformers to huggingface_hub >= 0.1.0

* Forgot to save...

* Style

* Fix test

* Remove n_ctx from configs

* Fix GPTJ and OpenAIGPT, both are acceptable breaking changes as there are no configs such that it breaks

* Remove unecessary n_positions from TFOpenAIGPT

* up

* up

* fix

* up

* Update examples/pytorch/test_xla_examples.py

* correct labels

* up

* up

* up

* up

* up

* up

Updated masked-language modeling examples in pytorch
with convention defined by #12789

* specify the text column name in the error message

* pluralize the word fields

[Speech Recognition] - Distributed training: Make sure vocab file removal and creation don't interfer  (#14161)

* up

* better

* Add seq2seq example for QnA on SQuAD Dataset.

* Changes from review - Fixing styling mistakes.

* Added how to example in README, simplified the access to dataset's preprocess function.

* Added tests for the seq2seq QA example.

* Change dataset column name to fix tests.

* Fix test command mistake.

* Add missing argument 'ignore_pad_token_for_loss' from DataTrainingArguments.

* Add missing argument 'num_beams' from DataTrainingArguments.

* Fix processing of output predicted token ids so that tokenizer decode gets appropriate input. Updated assertion conditions on the tests.

last line: "# limitations under the License." is missing

* Update SEW integration test tolerance

* Add audio classification notebooks

* up

* Update notebooks/README.md

* Update SEW integration test tolerance

* Update audio classification

* Update test

* Remove torchaudio

* Add dataset revision

* Hub branch naming

* Revert dataset revisions

* Update datasets

* up

* up

* up

* finish

* finish

* up

* finish all

* up

* fixed horizon_length

* fixed horizon_length

* fix style

This reverts commit 5b6bd4e7.