notesum.ai

Published at November 27

JPPO: Joint Power and Prompt Optimization for Accelerated Large Language Model Services

eess.AS

cs.CL

cs.SD

Released Date: November 27, 2024

Authors: Feiran You¹, Hongyang Du¹, Kaibin Huang¹, Abbas Jamalipour²

Aff.: ¹Department of Electrical and Electronic Engineering, University of Hong Kong, Hong Kong SAR, China; ²School of Electrical and Computer Engineering, University of Sydney, Sydney, Australia

Arxiv: http://arxiv.org/abs/2411.18010v1

Parameter	Value
Learning rate	$10^{-3}$
$\alpha_{1}$ , $\alpha_{2}$ , and $\alpha_{3}$	$0.4$ , $0.3$ and $0.3$
Total test runs range	$[1,10]$
Episodes per test run	$10,000$