GPipe: Efficient Training of Giant Neural Networks using Pipeline Parallelism
GPipe: Efficient Training of Giant Neural Networks using Pipeline Parallelism
Yanping Huang,Yonglong Cheng,4 Authors,Z. Chen
2018 · DBLP: conf/nips/HuangCBFCCLNLWC19
Neural Information Processing Systems · 2,223 Citations
TLDR
GPipe is introduced, a pipeline parallelism library that allows scaling any network that can be expressed as a sequence of layers by pipelining different sub-sequences of layers on separate accelerators, resulting in almost linear speedup when a model is partitioned across multiple accelerators.
