vllm-project/vllm · error · ValueError

Could not determine Python executable. Please provide it man

Error message

Could not determine Python executable. Please provide it manually.

What it means

tools/generate_cmake_presets.py needs a Python interpreter path to embed in CMakePresets.json. It first tries sys.executable; if that is empty (can happen under embedded/restricted launchers) it prompts interactively. When stdin yields an empty answer, this ValueError aborts generation.

Source

Thrown at tools/generate_cmake_presets.py:68

            "path to nvcc (e.g., /usr/local/cuda/bin/nvcc): "
        )
        nvcc_path = nvcc_path_input.strip()
    print(f"Using NVCC path: {nvcc_path}")

    # Detect Python executable
    python_executable = get_python_executable()
    if python_executable:
        print(f"Found Python via sys.executable: {python_executable}")
    else:
        python_executable_prompt = (
            "Could not automatically find Python executable. Please provide "
            "the full path to your Python executable for vLLM development "
            "(typically from your virtual environment, e.g., "
            "/home/user/venvs/vllm/bin/python): "
        )
        python_executable = input(python_executable_prompt).strip()
        if not python_executable:
            raise ValueError(
                "Could not determine Python executable. Please provide it manually."
            )

    print(f"Using Python executable: {python_executable}")

    # Get CPU cores
    cpu_cores = get_cpu_cores()
    nvcc_threads = min(4, cpu_cores)
    cmake_jobs = max(1, cpu_cores // nvcc_threads)
    print(
        f"Detected {cpu_cores} CPU cores. "
        f"Setting NVCC_THREADS={nvcc_threads} and CMake jobs={cmake_jobs}."
    )

    # Get vLLM project root (assuming this script is in vllm/tools/)
    project_root = os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))
    print(f"VLLM project root detected as: {project_root}")

View on GitHub (pinned to c794754062)

Solutions

Invoke the script with your venv interpreter directly so sys.executable is populated: `/path/to/vllm-venv/bin/python tools/generate_cmake_presets.py`.
Run it in an interactive terminal and paste the full venv python path (e.g. /home/user/venvs/vllm/bin/python) when prompted.
Ensure the virtual environment is activated before running the tool.

Example fix

# before
python tools/generate_cmake_presets.py < /dev/null
# -> ValueError: Could not determine Python executable.

# after
source .venv/bin/activate && .venv/bin/python tools/generate_cmake_presets.py

Defensive patterns

Strategy: validation

Validate before calling

import sys
if not sys.executable:
    raise SystemExit("Run via an interpreter with a populated sys.executable (e.g. .venv/bin/python)")

Prevention

Invoke the script with the venv's python binary directly rather than through wrappers.
Never run generate_cmake_presets.py with stdin closed unless sys.executable is populated.
Activate the venv before any CMake preset generation step.

When it happens

Trigger: Running generate_cmake_presets.py non-interactively (CI, piped stdin) where sys.executable is empty and the prompt receives EOF/empty input.

Common situations: Scripting CMake preset generation in CI with stdin closed; invoking the tool through an embedded interpreter that reports an empty sys.executable.

Related errors

AI-assisted analysis of vllm-project/vllm@c794754062 (2026-08-14). Data as JSON: /api/errors/a22ca57ebcb43836. Report an issue: GitHub.