GraphStorm v0.0.1 release

v0.0.1 is the first release of GraphStorm which includes support for GNN models training on multi-GPU or multi-machine multi-GPU environments.

Major features

• Native support for multi-GPU and multi-machine multi-GPU GNN training and inference.
• GNN training and inference for node classification and regression tasks. (https://github.com/awslabs/graphstorm/tree/main/training_scripts/gsgnn_np, https://github.com/awslabs/graphstorm/tree/main/inference_scripts/np_infer)
• GNN training and inference for edge classification and regression tasks. (https://github.com/awslabs/graphstorm/tree/main/training_scripts/gsgnn_ep, https://github.com/awslabs/graphstorm/tree/main/inference_scripts/ep_infer)
• GNN training and inference for link prediction tasks. (https://github.com/awslabs/graphstorm/tree/main/training_scripts/gsgnn_lp, https://github.com/awslabs/graphstorm/tree/main/inference_scripts/lp_infer)
• Built-in model support for RGCN and RGAT.
• Support custom GNN model and provide an example of implementing HGT using GraphStorm framework. (https://github.com/awslabs/graphstorm/tree/main/examples/customized_models/HGT)
• Support Huggingface BERT-GNN co-training.
• Support Graph-aware Huggingface BERT fine-tuning.
• Support various evaluation metrics including Accuracy, F1, Roc-Auc, MSE, RMSE, MRR, etc for different graph ML tasks.
• AWS native support. We provide a guideline to build GraphStorm Docker images for AWS EC2.(https://github.com/awslabs/graphstorm/tree/main/docker)

Contributors

• Da Zheng from AWS
• Xiang Song from AWS
• Jian Zhang from AWS
• Theodore Vasiloudis from AWS
• Prateek M Desai from AWS
• Israt Nisa from AWS
• Vasileios Ioannidis from AWS