tester.py

import subprocess
import re
import sys

def install_llama_cpp_python(cuda=None):
    try:
        # Detect CUDA version
        result = subprocess.run(['nvcc', '--version'], stdout=subprocess.PIPE, stderr=subprocess.PIPE, text=True)
        version_output = result.stdout
        version_match = re.search(r"release (\d+\.\d+),", version_output)
        if not version_match:
            raise RuntimeError("CUDA version not found in nvcc output.")
        
        cuda_version = version_match.group(1)
        version_map = {
            "12.1": "cu121",
            "12.2": "cu122",
            "12.3": "cu123",
            "12.4": "cu124",
        }
        major_minor = '.'.join(cuda_version.split('.')[:2])
        if cuda is None:
            cuda_suffix = version_map.get(major_minor)
        else:
            cuda_suffix = cuda

        if not cuda_suffix:
            raise ValueError(f"No suitable wheel found for CUDA version {cuda_version}.")

        extra_index_url = f"https://abetlen.github.io/llama-cpp-python/whl/{cuda_suffix}"
        subprocess.run([sys.executable, '-m', 'pip', 'install', 'llama-cpp-python', '--extra-index-url', extra_index_url])
        print(f"Successfully installed llama-cpp-python with CUDA {cuda_suffix}")

    except FileNotFoundError:
        print("Error: nvcc (CUDA) is not installed or not in PATH.")
    except Exception as e:
        print(f"An error occurred: {e}")

# Run the function
install_llama_cpp_python('cu124')