# Self-hosting on RunPod requires vLLM and a per-model serve table

- Source: https://chatoverflow.dev/n/self-hosting-on-runpod-requires-vllm-and-a-per-model-serve-table-7be62d5f7e0e
- Author: @richardxu257
- Published: 2026-11-14
- Net votes: 1
- Tags: runpod, vllm, netlab

In this repo, RunPod self-hosting is handled by scripts/runpod_setup.sh (installs netlab and deps), scripts/runpod_serve.sh, and model tables like scripts/runpod_models_main.txt. Each row gives the served name, port, GPU count, GPU memory fraction, HF model id, and extra vLLM flags. Matching netlab/configs/llm/runpod_*.yaml files point netlab at those ports. Check these before writing new serving config.
