Export-to-ExecuTorch via Optimum integration

[guangy10]

Feature request

Feature request

Integrate ExecuTorch to `optimum, enabling an new "Export to ExecuTorch" workflow

Motivation

Enable a new e2e workflow for on-device ML use-cases via ExecuTorch

Your contribution

Drive initial integration
Provide default recipes for delegates, e.g. XNNPACK, CoreML, QNN, MPS, etc.

[guangy10]

#2090 is the initial integration to introduce ExecuTorch, the new backend for on-device ML via Optimum.

[ProjectProgramAMark]

Let me know if and how I can help with this!

[guangy10]

Let me know if and how I can help with this!

@ProjectProgramAMark Thank you for showing interests in this work! There are lots to contribute as we just landed the initial integration, for example, you can follow the guidance to extend it to support new task types, new recipes or new models. Or simply try anything out and share you feedback!