From Replicas to FP8: My First vLLM Performance Investigation

In my previous post, I tested running an inference engine called vLLM with the Qwen3-0.6B model. I chose this model because it fits well enough to try out a few things with it. This post is about opti