https://www.youngju.dev/blog 천천히 올바르게. AI Researcher & DevOps Engineer Youngju's blog. GPU/CUDA, LLM, MLOps, Kubernetes AI workloads, and data engineering — plus mindset essays on confidence, routines, health, and sport psychology. ko fjvbn2003@gmail.com (Youngju Kim) fjvbn2003@gmail.com (Youngju Kim) Wed, 12 Aug 2026 00:00:00 GMT https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-tuning-and-pitfalls.en https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-tuning-and-pitfalls.en A practical order for tuning a vLLM deployment. Covers what gpu_memory_utilization actually sets, when to use tensor parallelism versus pipeline parallelism, how to choose quantization and a KV cache data type, and a diagnostic order for OOM that separates the startup stage from the operating stage, all checked against the official documentation. The final part of the Inside vLLM series. Wed, 12 Aug 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) vllmgpuquantizationtensor-parallelllm https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-tuning-and-pitfalls.ja https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-tuning-and-pitfalls.ja vLLMのデプロイを実際にチューニングする手順を整理する。gpu_memory_utilizationが何を決めるのか、テンソル並列とパイプライン並列をいつ使うのか、量子化とKVキャッシュのデータ型をどう選ぶのか、そしてOOMが起きたときに起動段階と運用段階を分けて診断する手順まで、公式ドキュメントを基準に確認して整理した。vLLM内部構造シリーズ最終回。 Wed, 12 Aug 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) vllmgpuquantizationtensor-parallelllm https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-tuning-and-pitfalls https://www.youngju.dev/blog/ai-platform/2026-08-12-vllm-tuning-and-pitfalls vLLM 배포를 실제로 튜닝하는 순서를 정리했습니다. gpu_memory_utilization이 무엇을 정하는지, 텐서 병렬과 파이프라인 병렬을 언제 쓰는지, 양자화와 KV 캐시 자료형을 어떻게 고르는지, 그리고 OOM이 났을 때 기동 단계와 운영 단계를 나눠 진단하는 순서까지 공식 문서 기준으로 확인해 정리했습니다. vLLM 내부 구조 시리즈 마지막 편. Wed, 12 Aug 2026 00:00:00 GMT fjvbn2003@gmail.com (Youngju Kim) vllmgpuquantizationtensor-parallelllm