English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

With the increasing adoption of graph neural networks (GNNs) in the machinelearning community, GPUs have become an essential tool to accelerate GNNtraining. However, training GNNs on very large graphs that do not fit in GPUmemory is still a challenging task. Unlike conventional neural networks,mini-batching input samples in GNNs requires complicated tasks such astraversing neighboring nodes and gathering their feature values. While thisprocess accounts for a significant portion of the training time, we findexisting GNN implementations using popular deep neural network (DNN) librariessuch as PyTorch are limited to a CPU-centric approach for the entire datapreparation step. This "all-in-CPU" approach has negative impact on the overallGNN training performance as it over-utilizes CPU resources and hinders GPUacceleration of GNN training. To overcome such limitations, we introducePyTorch-Direct, which enables a GPU-centric data accessing paradigm for GNNtraining. In PyTorch-Direct, GPUs are capable of efficiently accessingcomplicated data structures in host memory directly without CPU intervention.Our microbenchmark and end-to-end GNN training results show that PyTorch-Directreduces data transfer time by 47.1% on average and speeds up GNN training by upto 1.6x. Furthermore, by reducing CPU utilization, PyTorch-Direct also savessystem power by 12.4% to 17.5% during training. To minimize programmer effort,we introduce a new "unified tensor" type along with necessary changes to thePyTorch memory allocator, dispatch logic, and placement rules. As a result,users need to change at most two lines of their PyTorch GNN training code foreach tensor object to take advantage of PyTorch-Direct.

PyTorch-Direct: Enabling GPU Centric Data Access for Very Large Graph Neural Network Training with Irregular Accesses