foundry local and window ml

[linxiaochou]

Hello everyone, I'm a newcomer who has only recently come into contact with foundry local. I also read the introduction of Windows ai foundry before. There is an architecture diagram in it showing that Windows ml provides services from the bottom layer to the upper layer, while foundry local is in the upper layer. However, when I was using foundry local, I found that the directory contained onnx_runtime.lib, and the architecture diagram of foundry local also showed that its runtime was onnx runtime.

The question I want to raise is: Why doesn't foundry local directly use window ml?

I hope what I have said can be made clear. I am looking forward to anyone's answer. Thank you very much.

architecture diagram of window ai foundry: https://learn.microsoft.com/zh-cn/windows/ai/images/ai-foundry-overview.png

architecture diagram of foundry local: https://learn.microsoft.com/zh-cn/azure/ai-foundry/foundry-local/media/architecture/foundry-local-arch.png

[bmehta001]

Hi linxiaochou , Thanks for your question. Foundry Local uses WinML to download EPs, but uses a newer version of ONNX Runtime, as well as uses ONNX Runtime GenAI -- a library built on ONNX Runtime focused on GenAI models -- for the latest features and best performance.