HetPipe: Enabling Large DNN Training on (Whimpy) Heterogeneous GPU Clusters through Integration of Pipelined Model Parallelism and Data Parallelism

Abstract

Deep Neural Network (DNN) models have continuously been growing in size in order to improve the accuracy and quality of the models. Moreover, for training of large DNN models, the use of heterogeneous GPUs is inevitable due to the short release cycle of new GPU architectures. In this paper, we investigate how to enable training of large DNN models on a heterogeneous GPU cluster that possibly includes whimpy GPUs that, as a standalone, could not be used for training. We present a DNN training system, HetPipe (Heterogeneous Pipeline), that integrates pipelined model parallelism (PMP) with data parallelism (DP). In HetPipe, a group of multiple GPUs, called a virtual worker, processes minibatches in a pipelined manner, and multiple such virtual workers employ data parallelism for higher performance. We also propose a novel parameter synchronization model, which we refer to as Wave Synchronous Parallel (WSP) to accommodate both PMP and DP for virtual workers, and provide convergence proof of WSP. Our experimental results on a given heterogeneous setting show that with HetPipe, DNN models converge up to 49% faster compared to the state-of-the-art DP technique.

Cited by in corpus (6)

• PanGu-$\alpha$: Large-scale Autoregressive Pretrained Chinese Language Models with Auto-parallel Computation
• Towards Efficient Post-training Quantization of Pre-trained Language Models
• PipeTransformer: Automated Elastic Pipelining for Distributed Training of Transformers
• Pipeline Parallelism for Inference on Heterogeneous Edge Computing
• Automatic Graph Partitioning for Very Large-scale Deep Learning
• MXDAG: A Hybrid Abstraction for Cluster Applications