modal/vllm_service.py

import modal

vllm_image = (
    modal.Image.debian_slim(python_version="3.12")
    .pip_install(
        "vllm==0.7.2",
        "transformers==4.51.0",
        "huggingface_hub[hf_transfer]",
        "flashinfer-python==0.2.0.post2",
        extra_index_url="https://flashinfer.ai/whl/cu124/torch2.5",
    )
    .env({"HF_HUB_ENABLE_HF_TRANSFER": "1"})  # faster model transfers
)
vllm_image = vllm_image.env({"VLLM_USE_V1": "1"})

hf_cache_vol = modal.Volume.from_name("mcp-datascientist-model-weights-vol")
vllm_cache_vol = modal.Volume.from_name("vllm-cache", create_if_missing=True)

app = modal.App("example-vllm-openai-compatible")

N_GPU = 1  # tip: for best results, first upgrade to more powerful GPUs, and only then increase GPU count
API_KEY = "super-secret-key-mcp-hackathon"  # api key, for auth. for production use, replace with a modal.Secret

MINUTES = 60  # seconds
VLLM_PORT = 8000

MODEL_NAME = "Qwen/Qwen3-14B"


@app.function(
    image=vllm_image,
    gpu=f"A100-40GB",
    scaledown_window=15 * MINUTES,  # how long should we stay up with no requests?
    timeout=10 * MINUTES,  # how long should we wait for container start?
    volumes={
        "/root/.cache/huggingface": hf_cache_vol,
        "/root/.cache/vllm": vllm_cache_vol,
    },
)
@modal.concurrent(
    max_inputs=10
)  # how many requests can one replica handle? tune carefully!
@modal.web_server(port=VLLM_PORT, startup_timeout=5 * MINUTES)
def serve():
    import subprocess

    cmd = [
        "vllm",
        "serve",
        "--uvicorn-log-level=info",
        MODEL_NAME,
        "--host",
        "0.0.0.0",
        "--port",
        str(VLLM_PORT),
        "--api-key",
        API_KEY,
    ]

    subprocess.Popen(" ".join(cmd), shell=True)