LLama bindings for Mac OS not included in CPU

[juancc]

LLama bindings for Mac OS not included.
Requirement.txt path not in absolute.

As the folder "dependencies_dir" exists pip for installing llama_cpp_python is never called.
The problem is in operators.install_and_load_dependencies function
Modify this and show work:

def install_and_load_dependencies():
    install_pip()
    install_devel()

    dependencies_dir = absolute_path(".python_dependencies")
    # if os.path.exists(dependencies_dir) and len(os.listdir(dependencies_dir)) > 2:
    #     print("Requirements already installed, skipping")
    #     load_dependencies()
    #     return

    os.makedirs(dependencies_dir, exist_ok=True)

    if (sys.platform == "win32" or sys.platform == "linux") and check_cuda():
        requirements_file = "./requirements/cuda.txt"
    else:
        requirements_file = absolute_path("./requirements/cpu.txt")
    
    subprocess.run(
        [
            sys.executable,
            "-m",
            "pip",
            "install",
            "-r",
            requirements_file,
            "--upgrade",
            "--no-cache-dir",
            "--target",
            dependencies_dir,
        ],
        check=True,
        cwd=os.path.dirname(__file__),
    )

    load_dependencies()