1 paper
Rui Bao, Yaping Sun, Zhiyong Chen +4
Edge deployments of generative inference increasingly face two practical realities: per-device per-model performance is often unknown at deployment time, and it is non-stationary d…