ServerlessPD: Fast RDMA-Codesigned Disaggregated Prefill-Decoding for Serverless Inference of Large Language Models.

Mingxuan Liu, Jianhua Gu, Tianhai Zhao

Browse the full ICWS paper archive.