Support for OneDNN

[david20571015]

Hi developers.

I'd like to request support for the OneDNN Execution Provider (EP) in ONNX Runtime GenAI.

Integrating OneDNN could significantly enhance performance, especially on Intel CPUs, by leveraging its highly optimized deep learning primitives. This would be incredibly beneficial for users deploying GenAI models on Intel-based hardware.

Thanks for considering this request!