vllm-installer

What problem does it solve?

This skill provides a turnkey guide to install, configure, and diagnose vLLM on NVIDIA GPUs, removing guesswork and reducing setup time.

Core Features & Use Cases

• End-to-end installation: CUDA, PyTorch, vLLM, FlashInfer, and KV transfer components.
• Environment validation and debugging: LSSD mount checks, LD_LIBRARY_PATH setup, and DeepEP readiness for MoE models.
• Server deployment and testing: Launch an OpenAI-compatible vLLM API server and verify model endpoints.
• Troubleshooting and maintenance: Diagnose common issues and re-run the diagnostic workflow.

Quick Start

source ./scripts/setup_env.sh pip install vllm==0.14.1 flashinfer-python==0.5.3 flashinfer-cubin==0.5.3 pip install nvidia-nccl-cu12==2.28.3 nvidia-cudnn-cu12==9.16.0.29 vllm serve Qwen/Qwen2.5-7B-Instruct --tensor-parallel-size 4 --port 8000 --host 0.0.0.0