Microsoft Phi 4 Mini Reasoning GGUF

🚀 Llamacpp imatrix Quantizations of Phi-4-mini-reasoning by microsoft

This project offers quantized versions of the microsoft/Phi-4-mini-reasoning model. It uses the llama.cpp library for quantization, providing various quantization types to meet different performance and quality requirements.

📦 Installation

• Using huggingface-cli: First, make sure you have huggingface-cli installed:

  pip install -U "huggingface_hub[cli]"
  

  Then, you can target the specific file you want:

  huggingface-cli download bartowski/microsoft_Phi-4-mini-reasoning-GGUF --include "microsoft_Phi-4-mini-reasoning-Q4_K_M.gguf" --local-dir ./
  

  If the model is bigger than 50GB, it will have been split into multiple files. In order to download them all to a local folder, run:

  huggingface-cli download bartowski/microsoft_Phi-4-mini-reasoning-GGUF --include "microsoft_Phi-4-mini-reasoning-Q8_0/*" --local-dir ./
  

✨ Features

• Multiple Quantization Types: Offers a wide range of quantization types, such as bf16, Q8_0, Q6_K_L, etc., to balance between model quality and file size.
• Online Repacking: Some quantization types support "online repacking" for weights, which can automatically optimize performance on ARM and AVX machines.
• Embed/Output Weights Optimization: Some quants use Q8_0 for embeddings and output weights, improving model quality.

💻 Usage Examples

• Run in LM Studio: You can run the quantized models in LM Studio.
• Run with llama.cpp: Run the models directly with llama.cpp, or any other llama.cpp based project.

📚 Documentation

Prompt format

<|system|>Your name is Phi, an AI math expert developed by Microsoft. {system_prompt}<|end|><|user|>{prompt}<|end|><|assistant|>

Download a file

Embed/output weights

Some of these quants (Q3_K_XL, Q4_K_L etc) are the standard quantization method with the embeddings and output weights quantized to Q8_0 instead of what they would normally default to.

ARM/AVX information

Previously, you would download Q4_0_4_4/4_8/8_8, and these would have their weights interleaved in memory in order to improve performance on ARM and AVX machines by loading up more data in one pass.

Now, however, there is something called "online repacking" for weights. details in this PR. If you use Q4_0 and your hardware would benefit from repacking weights, it will do it automatically on the fly.

As of llama.cpp build b4282 you will not be able to run the Q4_0_X_X files and will instead need to use Q4_0.

Additionally, if you want to get slightly better quality for , you can use IQ4_NL thanks to this PR which will also repack the weights for ARM, though only the 4_4 for now. The loading time may be slower but it will result in an overall speed incrase.

Which file should I choose?

🔧 Technical Details

• Quantization Method: Uses the llama.cpp release b5228 for quantization.
• Calibration Dataset: All quants are made using the imatrix option with a dataset from here.

📄 License

This project is licensed under the MIT License.

Credits

• Thank you kalomaze and Dampf for assistance in creating the imatrix calibration dataset.
• Thank you ZeroWw for the inspiration to experiment with embed/output.
• Thank you to LM Studio for sponsoring my work.

Want to support my work? Visit my ko-fi page here: https://ko-fi.com/bartowski