Request for qwen_2.5-omni-3B deploying with xnnpack backend

[hayyaw]

🚀 The feature, motivation and pitch

If I want to deploy the qwen_2.5-omni-3B model （https://huggingface.co/Qwen/Qwen2.5-Omni-3B）on the Android side, like xnnpack or qnn backend, what should I do？ Is there a development manual to guide development or some advice？ thanks very much.

Alternatives

No response

Additional context

No response

RFC (Optional)

No response

cc @mergennachin @iseeyuan @lucylq @helunwencser @tarun292 @kimishpatel @jackzhxng

[cccclai]

@jackzhxng have you tried this model in ExecuTorch?

[jackzhxng]

@hayyaw not enabled natively in ExecuTorch at the moment. We'll keep this ticket open since it might be a feature down the road, but in the meantime, Optimum ExecuTorch supports more models out of the box (currently not as performant, but it's getting much better recently) that you can explore