NOTE: The implementation of Phi Silica has changed and you must complete these additional steps to make the application run correctly: https://learn.microsoft.com/en-us/windows/ai/apis/get-started?tabs=winget%2Cwinui%2Cwinui2 (look for the steps starting with "Edit the Package.appxmanifest file"). You may also need to change the implementation in PhiSilicaClient.cs.
👉 The PowerPoint presentation, the Jupyter notebook, and code snippets are in the _GOODIES INSIDE folder.
👉 You must make sure that your device is capable of running Phi Silica (Copilot+ PC with NPU, Windows Insider beta or higher). Be sure you can run a Phi Silica sample from the AI Dev Gallery first. https://learn.microsoft.com/en-us/windows/ai/ai-dev-gallery/
NOTE: This repo does NOT include the ONNX model because it's too big.
- Download
model.onnxfrom https://huggingface.co/protectai/deberta-v3-large-zeroshot-v1-onnx/tree/main.- Rename that file to
zero_shot_classifier.onnxfile, drop it in the AIModels folder (alongsideLabels.csandvocab.txt)- If required, set its Build Action to
Contentand Copy to Output Directory toCopy if newer
The Azure resource URLs/keys in App.xaml will need to be replaced, as they are no longer functional after the lab.
This project is licensed under the MIT License. Please provide attribution.