A step-by-step guide to deploying open-source LLMs using vLLM on Hugging Face Inference Endpoints. Covers creating a custom container image, programmatically setting up endpoints with the huggingface_hub Python library, deploying Meta-Llama-3-8B-Instruct on an A10G GPU instance, and testing the deployment via the OpenAI SDK. Also notes current UI limitations with custom containers.

4m read timeFrom philschmid.de
Post cover image
Table of contents
Create Inference EndpointsTest Endpoint using OpenAI SDKConclusion
3 Impressions