cturan/MiniMax-M2-GGUF

Building and Running the Experimental minimax Branch of llama.cpp

Note:
This setup is experimental. The minimax branch will not work with the standard llama.cpp. Use it only for testing GGUF models with experimental features.

---

System Requirements (you can use any supported this is for ubuntu build commands)

• Ubuntu 22.04
• NVIDIA GPU with CUDA support
• CUDA Toolkit 12.8 or later
• CMake

---

Installation Steps

1. Install CUDA Toolkit 12.8

wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.1-1_all.deb
sudo dpkg -i cuda-keyring_1.1-1_all.deb
sudo apt-get update
sudo apt-get -y install cuda-toolkit-12-8

2. Set Environment Variables

export CUDA_HOME=/usr/local/cuda
export LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda/lib64:/usr/local/cuda/extras/CUPTI/lib64
export PATH=$PATH:$CUDA_HOME/bin

3. Install Build Tools

sudo apt install cmake

4. Clone the Experimental Branch

git clone --branch minimax --single-branch https://github.com/cturan/llama.cpp.git
cd llama.cpp

5. Build the Project

mkdir build
cd build
cmake .. -DLLAMA_CUDA=ON -DLLAMA_CURL=OFF
cmake --build . --config Release --parallel $(nproc --all)

---

Build Output

After the build is complete, the binaries will be located in:

llama.cpp/build/bin

---

Running the Model

Example command:

./llama-server -m minimax-m2-Q4_K.gguf -ngl 999 --cpu-moe --jinja -fa on -c 32000 --reasoning-format auto

This configuration offloads the experts to the CPU, so approximately 16 GB of VRAM is sufficient.

---

Notes

• --cpu-moe enables CPU offloading for mixture-of-experts layers.
• --jinja activates the Jinja templating engine.
• Adjust -c (context length) and -ngl (GPU layers) according to your hardware.
• Ensure the model file (minimax-m2-Q4_K.gguf) is available in the working directory.

---

All steps complete. The experimental CUDA-enabled build of llama.cpp is ready to use.