start.py

import subprocess

# Uninstall the existing llama-cpp-python package
subprocess.run("pip uninstall -y llama-cpp-python", shell=True)

# Install llama-cpp-python with CUDA support
install_command = "CMAKE_ARGS='-DLLAMA_CUDA=on' pip install llama-cpp-python"
subprocess.run(install_command, shell=True)

# Start the Hugging Face Space
subprocess.run("uvicorn app:app --host 0.0.0.0 --port 7860", shell=True)