Add Video Llava (#29733)

* add model draft

* update docstring

* add tests

* support image and video as input

* update for better handling of mixed input and clean-up a bit

* bug when mixed inputs & add tests

* Update README.md

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

* Merge remote-tracking branch 'upstream/main' into video_llava

* link to abstract of paper in README

* fix test

* fix-copies

* make tests happy

* skip docstest for now

* do not run doctest for now

* Update src/transformers/models/video_llava/processing_video_llava.py

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

* Update src/transformers/models/video_llava/image_processing_video_llava.py

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

* Update src/transformers/models/video_llava/image_processing_video_llava.py

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

* Update src/transformers/models/video_llava/image_processing_video_llava.py

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

* Update src/transformers/models/video_llava/image_processing_video_llava.py

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

* Update tests/models/video_llava/test_modeling_video_llava.py

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

* Update src/transformers/models/video_llava/image_processing_video_llava.py

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

* address review comments

* failing tests

* Fix vocab_size in common tests for VLMs

* codestyle

* Update src/transformers/models/video_llava/configuration_video_llava.py

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

* Update src/transformers/models/video_llava/configuration_video_llava.py

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

* Update src/transformers/models/video_llava/modeling_video_llava.py

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

* Update src/transformers/models/video_llava/modeling_video_llava.py

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

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

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

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

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

* Update src/transformers/models/video_llava/image_processing_video_llava.py

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

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

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

* Update src/transformers/models/video_llava/processing_video_llava.py

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

* Update tests/models/video_llava/test_modeling_video_llava.py

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

* Update tests/models/video_llava/test_modeling_video_llava.py

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

* Update tests/models/video_llava/test_modeling_video_llava.py

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

* PR suggestions

* fix-copies

* Update src/transformers/models/video_llava/configuration_video_llava.py

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

* Update src/transformers/models/video_llava/configuration_video_llava.py

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

* add full example in docs

* clean-up with new model-id

* [run-slow] video_llava

* update docstring

* [run-slow] video_llava

* remove all achive maps

* fix some tests

* test was supposed to be skipped for llava :)

---------

Co-authored-by: NielsRogge <48327001+NielsRogge@users.noreply.github.com>
Co-authored-by: amyeroberts <22614925+amyeroberts@users.noreply.github.com>

docs/source/en/perf_infer_gpu_one.md

@@ -56,6 +56,7 @@ FlashAttention-2 is currently supported for the following architectures:
 * [Llava](https://huggingface.co/docs/transformers/model_doc/llava)
 * [Llava-NeXT](https://huggingface.co/docs/transformers/model_doc/llava_next)
 * [VipLlava](https://huggingface.co/docs/transformers/model_doc/vipllava)
+* [VideoLlava](https://huggingface.co/docs/transformers/model_doc/video_llava)
 * [M2M100](https://huggingface.co/docs/transformers/model_doc/m2m_100)
 * [MBart](https://huggingface.co/docs/transformers/model_doc/mbart#transformers.MBartModel)
 * [Mistral](https://huggingface.co/docs/transformers/model_doc/mistral#transformers.MistralModel)