vLLM v0.31.0rc3 introduces Model Runner V2 with support for randomized dummy inputs

Core summary

One sentence to understand this update

vLLM's v0.31.0rc3 release candidate brings Model Runner V2, specifically adding support for randomized dummy inputs, which can be useful for testing and benchmarking.

Impact & opportunity

What this could mean

This update allows builders to perform more robust testing and performance evaluations of models in vLLM using varied dummy inputs.