GitLab

Fix Loading of Flax(Speech)EncoderDecoderModel kwargs from PreTrained Encoder-Decoder Checkpoints (#16056)

* Fix Loading of Flax(Speech)EncoderDecoderModel kwargs from PreTrained Encoder-Decoder Checkpoints

* change wording

* Make Camembert great again!

* Add Camembert to TensorFlow ONNX tests

* [WIP] add support for bf16 mode

* prep for bf16

* prep for bf16

* fix; zero2/bf16 is ok

* check bf16 is available

* test fixes

* enable zero3_bf16

* config files

* docs

* split stage_dtype; merge back to non-dtype-specific config file

* fix doc

* cleanup

* cleanup

* bfloat16 => bf16 to match the PR changes

* s/zero_gather_fp16_weights_on_model_save/zero_gather_16bit_weights_on_model_save/; s/save_fp16_model/save_16bit_model/

* test fixes/skipping

* move

* fix

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

* backticks

* cleanup

* cleanup

* cleanup

* new version

* add note about grad accum in bf16
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* add possibility to softly regulate length when using sampling method in model.generate() function

* fix test config, fix formatting

* fix rag integration, fix docstyling

* fix wrong docstring

* change param to tuple, add test

* fix old param in rag_model, remove unused import

* change test according to new param

* fix formatting

* fix test case

* fix doc style

* move start_length calculation to Logitprocessor

* add possibility to softly regulate length when using sampling method in model.generate() function

* fix rag integration, fix docstyling

* fix test config, fix formatting

* change param to tuple, add test

* fix old param in rag_model, remove unused import

* add possibility to softly regulate length when using sampling method in model.generate() function

* change param to tuple, add test

* fix old param in rag_model, remove unused import

* remove unused import

* fix small errors

* fix test

* add possibility to softly regulate length when using sampling method in model.generate() function

* fix test config, fix formatting

* fix rag integration, fix docstyling

* change param to tuple, add test

* fix old param in rag_model, remove unused import

* change test according to new param

* fix test case

* move start_length calculation to Logitprocessor

* add possibility to softly regulate length when using sampling method in model.generate() function

* fix rag integration, fix docstyling

* fix test config, fix formatting

* change param to tuple, add test

* fix old param in rag_model, remove unused import

* add possibility to softly regulate length when using sampling method in model.generate() function

* fix test config, fix formatting

* fix rag integration, fix docstyling

* add possibility to softly regulate length when using sampling method in model.generate() function

* fix rag integration, fix docstyling

* change param to tuple, add test

* fix old param in rag_model, remove unused import

* fix small errors

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

* Update src/transformers/generation_utils.py

* Update src/transformers/generation_utils.py

* fix docstring, add type ind model rag

* fix docstrings

* introduce seq_length variable for cleaner code

* fix black formatting

* add input_ids_seq_length to modeling_rag

* add input_ids_seq_length to test

* retrigger checks

* retrigger checks
Co-authored-by: Kevin Bondzio <kev@AIM-LAP-02.local>
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>
Co-authored-by: Kevin Bondzio <kev@AIM-LAP-02.fritz.box>

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

* Fix duplicate arguments passed to dummy inputs in ONNX export

* Fix M2M100 ONNX config

* Ensure we check PreTrained model only if torch is available

* Remove TensorFlow tests for models without PyTorch parity

* support not sharing embeddings

* update modeling

* update tokenizer

* fix conversion script

* always use self.shared

* boom boom

* begin tests

* update tests

* fix resize_decoder_token_embeddings

* address Patrick's comments

* style

* update conversion script

* fix conversion script

* fix tokenizer

* better name target vocab

* add integration test for tokenizer with two vocabs

* style

* address Patrick's comments

* add integration test for model

* Fix Bug in Flax-Speech-Encoder-Decoder Test

* change thresholds for CPU precision

* Add attentions_option to common tester

* Fix tests, apply suggestion

* Apply suggestion from code review
Co-authored-by: Niels Rogge <nielsrogge@Nielss-MacBook-Pro.local>

* Add conversion script

* Improve script

* Fix bug

* Add option to push to hub

* Add support for classification models

* Update model name

* Upload feature extractor files first

* Remove hash checking

* Fix config

* Add id2label

* Add import

* Fix id2label file name

* Fix expected shape

* Add model to README

* Improve docs

* Add integration test and fix CI

* Fix code style

* Add missing init

* Add model to SPECIAL_MODULE_TO_TEST_MAP
Co-authored-by: Niels Rogge <nielsrogge@Nielss-MacBook-Pro.local>

* Freeze Feature Encoder in FlaxSpeechEncoderDecoder

* add backprop test

* add causal lm

* add CausalLM tests

* Add FlaxBartForCausalLM

* Add EncoderDecoder model tests

* change docstring

* make repo-consistency

* suggested changes

* remove jax ops

* correction

* rename pre-trained decoder model

* Add ONNX support for ViT

* Refactor to use generic preprocessor

* Add vision dep to tests

* Extend ONNX slow tests to ViT

* Add dummy image generator

* Use model_type to determine modality

* Add deprecation warnings for tokenizer argument

* Add warning when overwriting the preprocessor

* Add optional args to docstrings

* Add minimum PyTorch version to OnnxConfig

* Refactor OnnxConfig class variables from CONSTANT_NAME to snake_case

* Add reasonable value for default atol
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

* Use tiny model for TFRembertEncoderDecoderModelTest.get_pretrained_model()
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

* removed an outdated check about hdf5_version
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

* Adding ForInstanceSegmentation to pipelines.

* Last fix `category_id` renamed to `label_id`.

* Can't be none no more.

* No `is_thing_map` anymore.

* Seed get_train_sampler's generator with arg seed to improve reproducibility

and make the world_size<=1 code path more similar to the others

* move test file into trainer test explicitly

* dumb typo

* make style lint happy

* per discussion, switch to data_seed

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

* Remove packed past from generation_tf_utils

* update models with the new past format

* update template accordingly

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

* Fix test across both cpu and gpu

* Fix typo

* Fix LayoutLMv2 test

* Update black

* set scale_embedding to False to avoid large (> 1e-5) output differences between PT/TF
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>

* Backprop Test for Freeze FlaxWav2Vec2 Feature Encoder

* remove jnp.ndarray type suggestion

* assert frozen grads are precisely zero

[Bug Fix] Beam search example in docs fails & a fix (integrating `max_length` in `BeamScorer.finalize()`) (#15555)

* added the test and fix

* had left out a comment

* added classes to get started with constrained beam search

* in progress, think i can directly force tokens now but not yet with the round robin

* think now i have total control, now need to code the bank selection

* technically works as desired, need to optimize and fix design choices leading to undersirable outputs

* complete PR #1 without disjunctive decoding

* removed incorrect tests

* Delete k.txt

* Delete test.py

* Delete test.sh

* revert changes to test scripts

* genutils

* full implementation with testing, no disjunctive yet

* shifted docs

* passing all tests realistically ran locally

* removing accidentally included print statements

* fixed source of error in initial PR test

* fixing the get_device() vs device trap

* fixed documentation docstrings about constrained_beam_search

* fixed tests having failing for Speech2TextModel's floating point inputs

* fix cuda long tensor

* added examples and testing for them and founx & fixed a bug in beam_search and constrained_beam_search

* deleted accidentally added test halting code with assert False

* code reformat

* Update tests/test_generation_utils.py
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* Update tests/test_generation_utils.py
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* Update tests/test_generation_utils.py
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* Update tests/test_generation_utils.py
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* Update tests/test_generation_utils.py

* fixing based on comments on PR

* took out the testing code that should but work fails without the beam search moditification ; style changes

* fixing comments issues

* docstrings for ConstraintListState

* typo in PhrsalConstraint docstring

* docstrings improvements

* finished adding what is sort of an opinionated implementation of disjunctive generation, but it revealed errors in inner beam search logic during testing.

* fixed bug found in constrained beam search that used beam_idx that were not global across all the batches

* disjunctive constraint working 100% correctly

* passing all tests

* Accidentally included mlruns

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

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

* complete overhaul of type complexities and other nits

* strict type checks in generate()

* fixing second round of feedback by narsil

* fixed failing generation test because of type check overhaul

* generation test fail fix

* fixing test fails
Co-authored-by: Patrick von Platen <patrick.v.platen@gmail.com>

* proper tests for post_process*** methods in feature extractor

* mask th == 0

* Update tests/maskformer/test_feature_extraction_maskformer.py
Co-authored-by: Sylvain Gugger <35901082+sgugger@users.noreply.github.com>

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

* Adding Flax XLM-RoBERTa

* Add Flax to __init__

* Adding doc and dummy objects

* Add tests

* Add Flax XLM-R models autodoc

* Fix tests

* Add Flask XLM-RoBERTa to TEST_FILES_WITH_NO_COMMON_TESTS

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

* Update tests/xlm_roberta/test_modeling_flax_xlm_roberta.py
Co-authored-by: Suraj Patil <surajp815@gmail.com>

* Update tests/xlm_roberta/test_modeling_flax_xlm_roberta.py
Co-authored-by: Suraj Patil <surajp815@gmail.com>

* Remove test on large Flask XLM-RoBERTa

* Add tokenizer to the test
Co-authored-by: Suraj Patil <surajp815@gmail.com>

Linked to https://github.com/huggingface/transformers/pull/15826

* Fix to support fast tokenizer with `CLIPProcessor`

* Update CLIPProcessor test for fast tokenizer

* Fix Docstring Style

* Rename into meaningful Variable name in test code

* Enabling MaskFormer in ppipelines

No AutoModel though :(

* Ooops local file.

* finish speech doc tests

* finish

* boom

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

in the scoring (which is more correct)

* Freeze FlaxWav2Vec2 Feature Encoder

* add to all module apply

* add backprop test

* Draft

* Add test

* Update src/transformers/models/realm/modeling_realm.py

* Apply suggestion

* Add block_mask

* Update

* Update

* Add block_embedding_to

* Remove no_grad

* Use AutoTokenizer

* Remove model.to overridding

* send PyTorch inputs to the correct device

* Fix: TypeError: can't convert cuda:0 device type tensor to numpy. Use Tensor.cpu() to copy the tensor to host memory first.
Co-authored-by: ydshieh <ydshieh@users.noreply.github.com>