TENSILE: A Tensor granularity dynamic GPU memory scheduler method towards multiple dynamic workloads system
TENSILE: A Tensor granularity dynamic GPU memory scheduler method towards multiple dynamic workloads system
Kaixin Zhang,Hongzhi Wang,3 Authors,Songling Zou
2021 · DBLP: journals/corr/abs-2105-13336
arXiv.org · 0 Citations
TLDR
Although there are some extensive works proposed for dynamic GPU memory management, they are hard to be applied to systems with multiple dynamic workloads, such as in-database machine learning system.
