Group-based Interleaved Pipeline Parallelism for Large-scale DNN Training.
PengCheng YangXiaoming ZhangWenpeng ZhangMing YangHong WeiPublished in: ICLR (2022)
Keyphrases
- training process
- real life
- small scale
- training phase
- training algorithm
- real world
- parallel processing
- web scale
- test set
- feedforward neural networks
- genetic algorithm
- computer vision
- training set
- online learning
- computational power
- shared memory
- training samples
- image classification
- object detection
- case study
- social networks
- machine learning
- data mining