Efficient Large-Scale Language Model Training on GPU Clusters

Deepak Narayanan, Mohammad Shoeybi, Jared Casper, Patrick LeGresley, Mostofa Patwary, Vijay Korthikanti, Dmitri Vainbrand, Prethvi Kashinkunti, Julie Bernauer, Bryan Catanzaro, Amar Phanishayee, Matei Zaharia
arXiv:2104.04473 [cs.CL], (9 Apr 2021)


   title={Efficient Large-Scale Language Model Training on GPU Clusters},

   author={Deepak Narayanan and Mohammad Shoeybi and Jared Casper and Patrick LeGresley and Mostofa Patwary and Vijay Korthikanti and Dmitri Vainbrand and Prethvi Kashinkunti and Julie Bernauer and Bryan Catanzaro and Amar Phanishayee and Matei Zaharia},






Download Download (PDF)   View View   Source Source   Source codes Source codes




Large language models have led to state-of-the-art accuracies across a range of tasks. However, training these large models efficiently is challenging for two reasons: a) GPU memory capacity is limited, making it impossible to fit large models on a single GPU or even on a multi-GPU server; and b) the number of compute operations required to train these models can result in unrealistically long training times. New methods of model parallelism such as tensor and pipeline parallelism have been proposed to address these challenges; unfortunately, naive usage leads to fundamental scaling issues at thousands of GPUs due to various reasons, e.g., expensive cross-node communication or idle periods waiting on other devices. In this work, we show how to compose different types of parallelism methods (tensor, pipeline, and data paralleism) to scale to thousands of GPUs, achieving a two-order-of-magnitude increase in the sizes of models we can efficiently train compared to existing systems. We discuss various implementations of pipeline parallelism and propose a novel schedule that can improve throughput by more than 10% with comparable memory footprint compared to previously-proposed approaches. We quantitatively study the trade-offs between tensor, pipeline, and data parallelism, and provide intuition as to how to configure distributed training of a large model. The composition of these techniques allows us to perform training iterations on a model with 1 trillion parameters at 502 petaFLOP/s on 3072 GPUs with achieved per-GPU throughput of 52% of peak; previous efforts to train similar-sized models achieve much lower throughput (36% of theoretical peak). Our code has been open-sourced.
No votes yet.
Please wait...

* * *

* * *

HGPU group © 2010-2021 hgpu.org

All rights belong to the respective authors

Contact us: