Implementing Spatio-Temporal Graph Convolutional Networks on Graphcore IPUs

Moe, Johannes; Pogorelov, Konstantin; Schroeder, Daniel Thilo; Langguth, Johannes

doi:10.1109/ipdpsw55747.2022.00016

Cited by 4 publications

(1 citation statement)

References 9 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Accelerators designed specifically for AI/ML applications show improved processing speed, scalability, and energy efficiency, allowing faster training on larger datasets. In particular, Graphcore's IPU accelerators show a 4× speedup over NVIDIA V100 GPUs for training of GNNs [20]. Graphcore's high-level development framework PopTorch was used to implement the Pytorch Geometric (PyG) library [21], which includes the SchNet framework.…”

Section: Methodsmentioning

confidence: 99%

Reducing Down(stream)time: Pretraining Molecular GNNs using Heterogeneous AI Accelerators

Bilbrey¹,

Herman²,

Sprueill³

et al. 2022

Preprint

View full text Add to dashboard Cite

The demonstrated success of transfer learning has popularized approaches that involve pretraining models from massive data sources and subsequent finetuning towards a specific task. While such approaches have become the norm in fields such as natural language processing, implementation and evaluation of transfer learning approaches for chemistry are in the early stages. In this work, we demonstrate finetuning for downstream tasks on a graph neural network (GNN) trained over a molecular database containing 2.7 million water clusters. The use of Graphcore IPUs as an AI accelerator for training molecular GNNs reduces training time from a reported 2.7 days on 0.5M clusters to 1.2 hours on 2.7M clusters. Finetuning the pretrained model for downstream tasks of molecular dynamics and transfer to a different potential energy surface took only 8.3 hours and 28 minutes, respectively, on a single GPU.Preprint. Under review.

show abstract

Section: Methodsmentioning

confidence: 99%