Optimize code to prevent repeated calls to convolution operator

benedekrozemberczki / pytorch_geometric_temporal

PyTorch Geometric Temporal: Spatiotemporal Signal Processing with Neural Machine Learning Models (CIKM 2021)

MIT License

2.61k stars 367 forks source link

Closed xunil17 closed 2 years ago

xunil17 commented 2 years ago

Currently the convolution operator in each gate is called n times for every node type in the graph.