gem-llm-vllm-debug

Diagnose vLLM startup and runtime failures with validated dependency and launch configurations.

Updated Apr 4, 2026
One-click install
npx skills add https://github.com/saintgo7/claude-skills --skill gem-llm-vllm-debug
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: gem-llm-vllm-debug
Source: https://github.com/saintgo7/claude-skills/tree/main/gem-llm-vllm-debug
Command: npx skills add https://github.com/saintgo7/claude-skills --skill gem-llm-vllm-debug

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This Skill helps you quickly identify and resolve vLLM model-server startup and runtime failures, including dependency mismatches and GPU-related crashes, so your Claude Code workflow can resume without long downtime.

Core Features & Use Cases

  • Compatibility & dependency matrix: Provides a validated version set (Python, vLLM, transformers, flashinfer, etc.) tailored to common failure patterns.
  • Dual-model vLLM launch templates: Supports operational debugging for Qwen2.5-Coder-32B and Qwen3-Coder-30B-A3B with correct tensor-parallel settings and tool-calling parser configuration.
  • Failure-mode playbook: Covers frequent issues like option incompatibilities, flashinfer version mismatches, DeepGEMM backend unavailability, transformers/mistral_common API breaks, model architecture recognition errors, port conflicts, and leftover GPU memory after crashes.

Quick Start

Run the installer hook for gem-llm-vllm-debug to use the validated debugging setup for your current environment.

Frequently Asked Questions about gem-llm-vllm-debug

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
Why does my vLLM server fail to start with a flashinfer version mismatch?

vLLM startup failures from flashinfer mismatches require strict version alignment across vLLM, transformers, and flashinfer, resolved by applying a validated compatibility matrix to your environment.

How do I configure tool calling when serving Qwen2.5-Coder-32B in vLLM?

To configure tool calling for Qwen2.5-Coder-32B in vLLM, apply correct launch templates specifying tensor-parallel settings and tool-calling parser configuration for your OpenAI compatible API server.

What is the best way to fix GPU OOM errors and leftover memory remnants after a vLLM crash?

Fix GPU OOM errors and leftover memory remnants after a vLLM crash by executing actionable recovery steps that clear residual GPU memory before relaunching the model server.

Does vLLM support DeepGEMM backend unavailability errors during model serving?

DeepGEMM backend unavailability errors during vLLM serving indicate dependency misconfigurations, requiring strict version alignment and correct launch parameters to restore backend functionality.

How to resolve port conflicts when launching a vLLM OpenAI compatible API server?

Resolve port conflicts when launching a vLLM OpenAI compatible API server by applying correct launch parameters that specify alternative ports and clear processes occupying the default server configuration.

Can I use vLLM to serve Qwen3-Coder-30B-A3B without encountering model architecture recognition errors?

Serving Qwen3-Coder-30B-A3B in vLLM without model architecture recognition errors requires strict version alignment of transformers and applying correct dual-model launch templates for configuration.