GPipe: Efficient Training of Giant Neural Networks using Pipeline Parallelism

Explore this paper's citation graph

Summary

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.

Type
preprint
Published
2018-11-16
Cited by
2,248
References
68
Access
Open access

Keywords

Computer science, Speedup, Pipeline (software), Artificial neural network, Machine translation

References

Cited by

Related papers