--- license: mit base_model: microsoft/Phi-3.5-mini-instruct --- # Phi-3.5-mini-instruct Sentis Conversion This repository contains a conversion of Microsoft's Phi-3.5 mini model into Unity's Sentis format. The model has been quantized to uint8 for optimized performance. ## Model Information - **Original Model**: [Microsoft Phi-3.5-mini-instruct](https://huggingface.co/microsoft/Phi-3.5-mini-instruct) - **Conversion**: Phi-3.5 mini to Unity Sentis format - **Quantization**: uint8 ## Installation To use this model in your Unity project: 1. Download the model JSON files and the `.sentis` model from this repository. 2. Place these files in your Unity project's `StreamingAssets` folder. 3. Download the Phi3Claude.cs script and put it into your Assets folder. ## Requirements - **Unity Sentis Version**: 1.6.0-pre.1 To run the LlamaTokenizer in Unity, you need to download and include the `Microsoft.ML.Tokenizers.dll` DLL from [the Microsoft's nuget servers](https://www.nuget.org/packages/Microsoft.ML.Tokenizers) and all its dependencies in your project: - `Google.Protobuf.dll` - `Microsoft.Bcl.AsyncInterfaces.dll` - `Microsoft.Bcl.HashCode.dll` - `Microsoft.ML.Tokenizers.dll` - `System.Runtime.CompilerServices.Unsafe.dll` - `System.Text.Encodings.Web.dll` - `System.Text.Json.dll` ## Usage To use the Phi-3.5-mini-instruct model in your Unity project: 1. Ensure you have completed the installation steps and added the required DLLs to your project. 2. In your Unity scene, add the `Phi3Claude` script to the Camera GameObject. 3. Click the Play button to run your scene. 4. Check the Console window to inspect the generated text output. The `Phi3Claude` script will run the text generation using the Sentis-converted Phi-3.5-mini-instruct model. You can modify this script to customize the input or adjust the generation as needed. ## Acknowledgements This model is a conversion of the Microsoft Phi-3.5 mini model. Please refer to the [original model repository](https://huggingface.co/microsoft/Phi-3.5-mini-instruct) for more information about the base model.