* fix qwen2vl float16 inference bug

* [run-slow] qwen2_vl

* update

* re-enable daily CI

---------

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

* fix

* fix

* fix

* fix

* skip

* skip more

---------

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

* fix: handle padding in contrastive search for decoder-only models

* fix: handle padding in contrastive search for encoder-decoder models

* tests: move padding contrastive test to test_util, add t5 test

* fix: handle if model_kwargs["decoder_attention_mask"] is None

* refactor: improve padding input contrastive search generation tests

* chore: _ranking_fast to use LongTensor for cosine_matrix_mask

* add check and prepare args for BC to ProcessorMixin, improve ProcessorTesterMixin

* change size and crop_size in processor kwargs tests to do_rescale and rescale_factor

* remove unnecessary llava processor kwargs test overwrite

* nit

* change data_arg_name to input_name

* Remove unnecessary test override

* Remove unnecessary tests Paligemma

* Move test_prepare_and_validate_optional_call_args to TesterMixin, add docstring

fix missing tests

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

* Add sdpa for BioGpt

* Updates

* Add the docs

* [run_slow] biogpt

* Use the copy mechanism to ensure consistency

* [run_slow] biogpt

Remove model tests

almost zero is not zero

* Update modeling_mamba2.py

Fix pad_size calculation to ensure it's less than self.chunk_size

* [run_slow] mamba2

* [run-slow] mamba2

* [run-slow] Add @require_read_token decorator to failing tests for token propagation

* [run_slow] mamba2

* enable

* fix

* add xpu skip

* add marker

* skip for xpu

* add more

* enable on accelerator

* add more cases

* add more tests

* add more

fix

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

fix

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

* Fix Llama 3 TikToken conversion

* No need to add tokens again

enable GemmaIntegrationTest

* enable

* fix

* add xpu skip

* add marker

* skip for xpu

* add more

* add one more

* Uniformize paligemma processor

* nit

* use kernel for dt calculations

* add small test

* [run-slow] mamba2

* change sequence_bias type of SequenceBiasLogitsProcessor tp list, add config tests for all processors

* fix format

* small fix for all_token_bias_pairs_are_valid internal func

* small typo fix in description

* improve test impl, some SequenceBiasLogitsProcessor refactoring

check attention mask in generate

* add initial design for uniform processors + align model

* add uniform processors for altclip + chinese_clip

* fix mutable default 👀

* add configuration test

* handle structured kwargs w defaults + add test

* protect torch-specific test

* fix style

* fix

* rebase

* update processor to generic kwargs + test

* fix style

* add sensible kwargs merge

* update test

* fix assertEqual

* move kwargs merging to processing common

* rework kwargs for type hinting

* just get Unpack from extensions

* run-slow[align]

* handle kwargs passed as nested dict

* add from_pretrained test for nested kwargs handling

* [run-slow]align

* update documentation + imports

* update audio inputs

* protect audio types, silly

* try removing imports

* make things simpler

* simplerer

* move out kwargs test to common mixin

* [run-slow]align

* skip tests for old processors

* [run-slow]align, clip

* !$#@!! protect imports, darn it

* [run-slow]align, clip

* [run-slow]align, clip

* update common processor testing

* add altclip

* add chinese_clip

* add pad_size

* [run-slow]align, clip, chinese_clip, altclip

* remove duplicated tests

* fix

* update doc

* improve documentation for default values

* add model_max_length testing

This parameter depends on tokenizers received.

* Raise if kwargs are specified in two places

* fix

* match defaults

* force padding

* fix tokenizer test

* clean defaults

* move tests to common

* remove try/catch block

* deprecate kwarg

* format

* add copyright + remove unused method

* [run-slow]altclip, chinese_clip

* clean imports

* fix version

* clean up deprecation

* fix style

* add corner case test on kwarg overlap

* resume processing - add Unpack as importable

* add tmpdirname

* fix altclip

* fix up

* add back crop_size to specific tests

* generalize tests to possible video_processor

* add back crop_size arg

* fixup overlapping kwargs test for qformer_tokenizer

* remove copied from

* fixup chinese_clip tests values

* fixup tests - qformer tokenizers

* [run-slow] altclip, chinese_clip

* remove prepare_image_inputs

* fix tests with main revision and read token

* [run-slow]mamba2

* test previously skipped tests

* [run-slow]mamba2

* skip some tests

* [run-slow]mamba2

* finalize tests

* [run-slow]mamba2

* add tests

* fix whisper

* update

* nit

* add qwen2-vl

* more updates!

* better this way

* fix this one

* fix more tests

* fix final tests, hope so

* fix led

* Update tests/generation/test_utils.py

Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>

* pr comments

* not pass pixels and extra for low-mem tests, very flaky because of visio tower

---------

Co-authored-by: Joao Gante <joaofranciscocardosogante@gmail.com>

* load and save from video-processor folder

* Update src/transformers/models/llava_onevision/processing_llava_onevision.py

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

---------

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* clean mimi commit

* some nits suggestions from Arthur

* make fixup

* rename repo id + change readme

* Update docs/source/en/model_doc/mimi.md

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* add flaky flag to batching equivalence due to audio_codes failing sometimes

---------

Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

* modify rt detr to improve inference times when compiled

* Remove redundant "to"

* Fix conditional lru_cache and missing shapes_list

* nit unnecessary list creation

* Fix compile error when ninja not available and custon kernel activated

* enforce original size to be a list

* formatting

* apply datatype change to unpad_image in llava_next

return attention mask in ASR pipeline

* Added support for bfloat16 to zero-shot classification pipeline

* Ensure support for TF.

Co-authored-by: Matt <Rocketknight1@users.noreply.github.com>

* Remove dependency on `torch`.

Co-authored-by: Matt <Rocketknight1@users.noreply.github.com>

---------

Co-authored-by: Matt <Rocketknight1@users.noreply.github.com>

* fix the wandb logging issue

* handle ConfigError in WandbCallback; move import to local scope

* update integration_utils.py; move import of ConfigError

* Update integration_utils.py: remove trailing whitespace

* Urdu docs added

* fixed the misaligned issue.