dev.to
Serverless Gemma 4: E2B on a Lambda GPU Function, with Google's QAT Weights and Drafter
Gemma 4 E2B with Google's QAT 4-bit weights and its assistant drafter, served by vLLM inside an AWS Lambda function on an NVIDIA L4. How to build it on the day Lambda gained GPUs, a local chat page, and what the drafter does to speed from 1 to 48 requests at once.
Original reporting from dev.to.
Read original at dev.to