As part of the Phi-3 launch Microsoft has released optimized ONNX models as detailed in ONNX Runtime supports Phi-3 mini models across platforms and devices and published the models on HuggingFace 🤗 at Phi-3 Mini-4K-Instruct ONNX models for consumption in for example the ONNX Runtime GenAI. This makes it very easy to run this model locally in just a few lines of C# as I’ll show in this blog post.
