Feature Request: Qwen-Image-Edit OpenVINO Support

Feature Request: Qwen-Image-Edit OpenVINO Support

Could Optimum Intel add official OpenVINO support for Qwen-Image-Edit based on the current Diffusers architecture?

There was an earlier experimental implementation in "[PR #1727](https://github.com/huggingface/optimum-intel/pull/1727)" (https://github.com/huggingface/optimum-intel/pull/1727), but it was closed as WIP.

It would be very useful to have:

- OpenVINO export/inference for Qwen-Image-Edit
- Qwen2.5-VL + Qwen Image Transformer + VAE support
- Intel CPU and GPU support
- INT8 quantization
- INT4/weight-only quantization if feasible

With the model's high memory requirements, INT8/INT4 OpenVINO support would make Qwen-Image-Edit much more practical on Intel integrated GPUs and other memory-constrained systems.

Is there a plan to implement or revisit Qwen-Image-Edit support in Optimum Intel?

Thank you!

Start by reviewing PR #1727 and the current Diffusers architecture referenced in the issue. Compare the existing experimental work with the requested Qwen2.5-VL, Qwen Image Transformer, VAE, CPU/GPU, and quantization support. Done would mean an agreed implementation covering the requested OpenVINO export and inference scope.

Written by the indexing model from the issue text.