GitLab

* Pin tensorflow-text to go with tensorflow

* Make it more convenient to pin TensorFlow

* setup don't like f-strings

* [deepspeed] offload + non-cpuadam optimizer exception doc

* deps

* time to say goodbye, torch 1.7 and 1.8

* clean up torch_int_div

* clean up is_torch_less_than_1_8-9

* update

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

* Add kernel size to NATTEN's QK arguments.

The new NATTEN 0.14.5 supports PyTorch 2.0, but also adds an additional
argument to the QK operation to allow optional RPBs.

This ends up failing NATTEN tests.

This commit adds NATTEN back to circleci and adds the arguments to get
it working again.

* Force NATTEN >= 0.14.5

* Use PyAV instead of Decord

* Get frame indices

* Fix number of frames

* Update src/transformers/models/videomae/image_processing_videomae.py

* Fix up

* Fix copies

* Update timesformer doctests

* Update docstrings

* Update setup.py

* suggestions

* Fix inclusion of non py files in package

* No need for the **

* Result of black 23.1

* Update target to Python 3.7

* Switch flake8 to ruff

* Configure isort

* Configure isort

* Apply isort with line limit

* Put the right black version

* adapt black in check copies

* Fix copies

* First draft

* Add initial draft of conversion script

* Convert all weights

* Fix config

* Add image processor

* Fix DetaImageProcessor

* Run make fix copies

* Remove timm dependency

* Fix dummy objects

* Improve loss function

* Remove conv_encoder attribute

* Update conversion scripts

* Improve postprocessing + docs

* Fix copied from statements

* Add tests

* Improve postprocessing

* Improve postprocessing

* Update READMEs

* More improvements

* Fix rebase

* Add is_torchvision_available

* Add torchvision dependency

* Fix typo and README

* Fix bug

* Add copied from

* Fix style

* Apply suggestions

* Fix thanks to @ydshieh

* Fix another dependency check

* Simplify image processor

* Add scipy

* Improve code

* Add threshold argument

* Fix bug

* Set default threshold

* Improve integration test

* Add another integration test

* Update setup.py

* Address review

* Improve deformable attention function

* Improve copied from

* Use relative imports

* Address review

* Replace assertions

* Address review

* Update dummies

* Remove dummies

* Address comments, update READMEs

* Remove custom kernel code

* Add image processor tests

* Add requires_backends

* Add minor comment

* Update scripts

* Update organization name

* Fix defaults, add doc tests

* Add id2label for object 365

* Fix tests

* Update task guide

* Adapt repository creation to latest hf_hub

* Update all examples

* Fix other tests, add Flax examples

* Address review comments

* update pyknp to rhoknp

* fix linter

* fix linter

* fix linter

* fix linter

* fix linter

* support rhoknp==1.1.0, fix testcase

* Install sentencepiece in DS CI image

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

* 🚧 wip video classification pipeline

* 🚧 wip - add is_decord_available check

* 🐛 add missing import

* ✅ add tests

* 🔧 add decord to setup extras

* 🚧 add is_decord_available

* ✨ add video-classification pipeline

* 📝 add video classification pipe to docs

* 🐛 add missing VideoClassificationPipeline import

* 📌 add decord install in test runner

* ✅ fix url inputs to video-classification pipeline

* ✨ updates from review

* 📝 add video cls pipeline to docs

* 📝 add docstring

* 🔥 remove unused import

* 🔥 remove some code

* 📝 docfix

* add minimal working gpt2 tokenizer

* graph mode and output equivalence tests working

* not today tensorflow. serialization test passing!

* fix style, documentation, docstrings and all that jazz

* passing consistency checks

* move keras nlp to tf dependencies

* fix tf modeling utils and gpt2 attention to enable compiling

* fix (I hope) keras nlp dependencies

* rever changes on generation

* remove debug prints

* remove redundant tf dummy objects

* add from config, get config and max length settings to address review

* let flake ignore the error on distillation you are welcome

* test from config

* add padding test

* address sgugger review

* Add DiNAT

* Adds DiNAT + tests

* Minor fixes

* Added HF model

* Add natten to dependencies.

* Cleanup

* Minor fixup

* Reformat

* Optional NATTEN import.

* Reformat & add doc to _toctree

* Reformat (finally)

* Dummy objects for DiNAT

* Add NAT + minor changes

Adds NAT as its own independent model + docs, tests
Adds NATTEN to ext deps to ensure ci picks it up.

* Remove natten from `all` and `dev-torch` deps, add manual pip install to ci tests

* Minor fixes.

* Fix READMEs.

* Requested changes to docs + minor fixes.

* Requested changes.

* Add NAT/DiNAT tests to layoutlm_job

* Correction to Dinat doc.

* Requested changes.

* Try PT1.13 by removing torch scatter

* Skip failing tests

* Style

* Remvoe testing extras for repo utils

* Try with all decorators

* Try to wipe the cache

* Fix all tests?

* Try this way

* Fix comma

* Update to main

* Try with less deps

* Quality

* pin torch to < 1.13

* pin torch to < 1.13
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

* Change the import of kenlm from github to pypi

* Change the import of kenlm from github to pypi in circleci config

* Fix code quality issues

* Fix isort issue, add kenlm in extras for audio

* Add kenlm to deps

* Add kenlm to deps

* Commit 'make fixup' changes

* Remove version from kenlm deps

* commit make fixup changes

* Remove manual installation of kenlm

* Remove manual installation of kenlm

* Remove manual installation of kenlm

* [Wav2Vec2] Allow user-managed Pool in Wav2Vec2ProcessorWithLM.batch_decode

* [Wav2Vec2] Add user-managed LM's pool tests and usage examples

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

* [Wav2Vec2] Fix hyperlink references
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* add sudachipy and jumanpp tokenizers for bert_japanese

* use ImportError instead of ModuleNotFoundError in SudachiTokenizer and JumanppTokenizer

* put test cases of test_tokenization_bert_japanese in one line

* add require_sudachi and require_jumanpp decorator for testing

* add sudachi and pyknp(jumanpp) to dependencies

* remove sudachi_dict_small and sudachi_dict_full from dependencies

* empty commit for ci

* Update protobuf dependency to fix vulnerability

* Update `dependency_versions_table.py` to include updated protobuf.

* initial files

* initial model via cli

* typos

* make a start on the model config

* ready with configuation

* remove tokenizer ref.

* init the transformer

* added initial model forward to return dec_output

* require gluonts

* update dep. ver table and add as extra

* fixed typo

* add type for prediction_length

* use num_time_features

* use config

* more config

* typos

* opps another typo

* freq can be none

* default via transformation is 1

* initial transformations

* fix imports

* added transform_start_field

* add helper to create pytorch dataloader

* added inital val and test data loader

* added initial distr head and loss

* training working

* remove TimeSeriesTransformerTokenizer
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/__init__.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/time_series_transformer/__init__.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* fixed copyright

* removed docs

* remove time series tokenizer

* fixed docs

* fix text

* fix second

* fix default

* fix order

* use config directly

* undo change

* fix comment

* fix year

* fix import

* add additional arguments for training vs. test

* initial greedy inference loop

* fix inference

* comment out token inputs to enc dec

* Use HF encoder/decoder

* fix inference

* Use Seq2SeqTSModelOutput output

* return Seq2SeqTSPredictionOutput

* added default arguments

* fix return_dict true

* scale is a tensor

* output static_features for inference

* clean up some unused bits

* fixed typo

* set return_dict if none

* call model once for both train/predict

* use cache if future_target is none

* initial generate func

* generate arguments

* future_time_feat is required

* return SampleTSPredictionOutput

* removed unneeded classes

* fix when params is none

* fix return dict

* fix num_attention_heads

* fix arguments

* remove unused shift_tokens_right

* add different dropout configs

* implement FeatureEmbedder, Scaler and weighted_average

* remove gluonts dependency

* fix class names

* avoid _variable names

* remove gluonts dependency

* fix imports

* remove gluonts from configuration

* fix docs

* fixed typo

* move utils to examples

* add example requirements

* config has no freq

* initial run_ts_no_trainer

* remove from ignore

* fix output_attentions and removed unsued getters/setters

* removed unsed tests

* add dec seq len

* add test_attention_outputs

* set has_text_modality=False

* add config attribute_map

* make style

* make fix-copies

* add encoder_outputs to TimeSeriesTransformerForPrediction forward

* Improve docs, add model to README

* added test_forward_signature

* More improvements

* Add more copied from

* Fix README

* Fix remaining quality issues

* updated encoder and decoder

* fix generate

* output_hidden_states and use_cache are optional

* past key_values returned too

* initialize weights of distribution_output module

* fixed more tests

* update test_forward_signature

* fix return_dict outputs

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

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

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

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

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

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

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

* removed commented out tests

* added neg. bin and normal output

* Update src/transformers/models/time_series_transformer/configuration_time_series_transformer.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* move to one line

* Add docstrings

* Update src/transformers/models/time_series_transformer/configuration_time_series_transformer.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* add try except for assert and raise

* try and raise exception

* fix the documentation formatting

* fix assert call

* fix docstring formatting

* removed input_ids from DOCSTRING

* Update input docstring

* Improve variable names

* Update order of inputs

* Improve configuration

* Improve variable names

* Improve docs

* Remove key_length from tests

* Add extra docs

* initial unittests

* added test_inference_no_head test

* added test_inference_head

* add test_seq_to_seq_generation

* make style

* one line

* assert mean prediction

* removed comments

* Update src/transformers/models/time_series_transformer/modeling_time_series_transformer.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* Update src/transformers/models/time_series_transformer/modeling_time_series_transformer.py
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>

* fix order of args

* make past_observed_mask optional as well

* added Amazon license header

* updated utils with new fieldnames

* make style

* cleanup

* undo position of past_observed_mask

* fix import

* typo

* more typo

* rename example files

* remove example for now

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

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

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

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

* Update modeling_time_series_transformer.py

fix style

* fixed typo

* fix typo and grammer

* fix style
Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>
Co-authored-by: NielsRogge <niels.rogge1@gmail.com>
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* Add `beautifulsoup4` to extras["testing"]
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

* Poc to use safetensors

* Typo

* Final version

* Add tests

* Save with the right name!

* Update tests/test_modeling_common.py
Co-authored-by: Julien Chaumond <julien@huggingface.co>

* Support for sharded checkpoints

* Test from Hub part 1

* Test from hub part 2

* Fix regular checkpoint sharding

* Bump for fixes
Co-authored-by: Julien Chaumond <julien@huggingface.co>

* Use  instead of  from huggingface_hub

* bump huggingface_hub to 0.10.0 + make deps_table_update

* Allowing users to use the latest `tokenizers` release !

* Upgrading the versions table too.