HetSeq: Distributed GPU Training on Heterogeneous Infrastructure
This addresses the challenge for organizations like universities that have heterogeneous computing resources and cannot train large models, though it is incremental as it adapts existing frameworks.
The authors tackled the problem of training large neural network models on heterogeneous GPU infrastructure, which existing systems cannot handle, and developed HetSeq, a PyTorch-based software package that enables such training, demonstrating scalability in experiments with transformer translation and BERT language models.
Modern deep learning systems like PyTorch and Tensorflow are able to train enormous models with billions (or trillions) of parameters on a distributed infrastructure. These systems require that the internal nodes have the same memory capacity and compute performance. Unfortunately, most organizations, especially universities, have a piecemeal approach to purchasing computer systems resulting in a heterogeneous infrastructure, which cannot be used to compute large models. The present work describes HetSeq, a software package adapted from the popular PyTorch package that provides the capability to train large neural network models on heterogeneous infrastructure. Experiments with transformer translation and BERT language model shows that HetSeq scales over heterogeneous systems. HetSeq can be easily extended to other models like image classification. Package with supported document is publicly available at https://github.com/yifding/hetseq.