Push-based KV transfer to improve TTFT and TPOT for disaggregated inference in vLLM
vLLM's prefill-decode disaggregation splits LLM inference across engine instances but must transfer the KV cache between them. This post introduces a push-based NIXL connector that moves transfer coordination off the critical path, delivering up to 3x better TTFT and 30% better TPOT in benchmarks on Amazon EC2 P5en instances with AWS EFA.
阅读原文 · AWS 机器学习博客本文由 SoccerDream AI 自动聚合索引，内容版权归原作者与 AWS 机器学习博客 所有。
本站仅作技术动态聚合，不构成任何形式的建议。
本站仅作技术动态聚合，不构成任何形式的建议。