Breadth-First Pipeline Parallelism

We introduce Breadth-First Pipeline Parallelism, a novel training schedule which optimizes the combination of pipeline and data parallelism. Breadth-First Pipeline Parallelism lowers training time, cost and memory usage by combining a high GPU utilization with a small batch size per GPU, and by maki...

Full description

Saved in:
Bibliographic Details
Main Author Lamy-Poirier, Joel
Format Journal Article
LanguageEnglish
Published 10.11.2022
Subjects
Computer Science - Artificial Intelligence
Computer Science - Computation and Language
Computer Science - Distributed, Parallel, and Cluster Computing
Computer Science - Learning
Online AccessGet full text

Cover

More Information
Summary:We introduce Breadth-First Pipeline Parallelism, a novel training schedule which optimizes the combination of pipeline and data parallelism. Breadth-First Pipeline Parallelism lowers training time, cost and memory usage by combining a high GPU utilization with a small batch size per GPU, and by making use of fully sharded data parallelism. Experimentally, we observed an increase of up to 43% in training throughput for a 52 billion-parameter model using a small batch size per GPU compared to Megatron-LM, which would reduce the training time and cost by the same amount on a large GPU cluster.
DOI:10.48550/arxiv.2211.05953