Preprint
Jul 2026
Attention to Detail: Evaluating Energy, Performance, and Accuracy Trade-offs Across vLLM Configurations
The results show that the studied configuration options significantly impact energy and performance, mainly driven by attention type and prefix caching, while chunked prefill has a limited effect under the default vLLM serving configuration and evaluated workloads.
Nada Zine, Tristan Coignion, V. Stoico et al.
· 1 citation