Whale: Efficient Giant Model Training over Heterogeneous GPUs.
Xianyan JiaLe JiangAng WangWencong XiaoZiji ShiJie ZhangXinyuan LiLangshi ChenYong LiZhen ZhengXiaoyong LiuWei LinPublished in: USENIX Annual Technical Conference (2022)
Keyphrases
- real time
- formal model
- probabilistic model
- probability distribution
- computational model
- neural network
- genetic algorithm
- feature selection
- objective function
- markov chain
- experimental data
- graphics hardware
- neural network model
- conceptual model
- statistical model
- mathematical model
- em algorithm
- hidden markov models
- prior knowledge
- training data
- data sets