Enable QWen VL video preprocess

[xipingyan]

tickets: CVS-173219
1: Enable video preprocessing for Qwen VL model.
Add: ov::Property<std::vectorov::Tensor> video{"video"};

2: The main updates:
For video: For 2-in-1 merging, if 9 images are input, only 5 images are actually processed.
For image: For 2-in-1 merging, we only double each image, so if we input 9 images, we only actually process 9 images.
Introduce "If" node, merge video and image preprocess into one OV subgroup.

[rkazants]

please share CVS number in JIRA to access from arch perspective.

Thanks.

[Copilot]

Pull Request Overview

Enables video processing for QWen VL models by adding video input support throughout the VLM pipeline. The main change allows QWen VL models to handle video input through temporal patch processing, which groups video frames and merges them into combined patches for more efficient processing.

• Adds video parameter support to all VLM pipeline interfaces and binding functions
• Implements video encoding functionality specifically for QWen2VL models
• Updates the generation config to include is_video flag for video-specific processing

Reviewed Changes

Copilot reviewed 22 out of 22 changed files in this pull request and generated 2 comments.

Show a summary per file

File	Description
src/python/py_vlm_pipeline.cpp	Adds video parameter to Python VLM pipeline bindings
src/python/py_continuous_batching_pipeline.cpp	Updates continuous batching pipeline with video support
src/python/openvino_genai/py_openvino_genai.pyi	Adds is_video property and video parameter to type stubs
src/cpp/src/visual_language/vision_encoder.hpp	Adds virtual encode_video method to base VisionEncoder
src/cpp/src/visual_language/qwen2vl/classes.hpp	Declares video encoding implementation for QWen2VL
src/cpp/src/visual_language/qwen2vl/classes.cpp	Implements video preprocessing and encoding logic
src/cpp/src/visual_language/pipeline_base.hpp	Updates base pipeline interface to include video parameter
src/cpp/src/visual_language/pipeline.cpp	Updates main pipeline implementation with video support
src/cpp/src/visual_language/llava_next/classes.hpp	Updates LLaVANext interface with video parameter
src/cpp/src/visual_language/llava_next/classes.cpp	Adds video warning for unsupported models
src/cpp/src/visual_language/llava/classes.hpp	Updates LLaVA interface with video parameter
src/cpp/src/visual_language/llava/classes.cpp	Adds video warning for unsupported models
src/cpp/src/visual_language/inputs_embedder.hpp	Updates inputs embedder interface for video support
src/cpp/src/visual_language/inputs_embedder.cpp	Implements video encoding routing logic
src/cpp/src/visual_language/continuous_batching_adapter.hpp	Updates adapter interface with video parameter
src/cpp/src/visual_language/clip.cpp	Optimizes bicubic resize with early exit for same-size images
src/cpp/src/continuous_batching/pipeline_impl.cpp	Updates implementation to handle video parameters
src/cpp/src/continuous_batching/pipeline_base.hpp	Updates base interface with video support
src/cpp/src/continuous_batching/pipeline_base.cpp	Implements video parameter handling in pipeline
src/cpp/src/continuous_batching/pipeline.cpp	Updates main pipeline with video parameter support
src/cpp/include/openvino/genai/visual_language/pipeline.hpp	Adds video property and parameter to public interface
src/cpp/include/openvino/genai/continuous_batching_pipeline.hpp	Updates public interface with video support

---

_{Tip: Customize your code reviews with copilot-instructions.md. Create the file or learn how to get started.}