Project Name: *
PyTorch Geometric Temporal
Email Address: *
Project Summary and Goals: *
PyTorch Geometric Temporal provides spatiotemporal deep learning layers, spatiotemporal data iteratiors and benchmark dataset loaders. It covers nearly 20 deep learning models from major conferences (AAAI, IJCAI, KDD, WWW, CIKM and so on):
GConvLSTM, GConvGRU, MTGNN, ASTGCN, MSTGCN, STGCN, DCRNN, GCLSTM, TGCN, A3TGCN, GMAN, MPNNLSTM, DyGRAE, LRGCN
Provides easy to use data loaders for these data types:
- Temporal signal on static graph
- Temporal signal on dynamic graph
- Static signal on dynamic graph
- Chickenpox forecasting
- Denmark windmill output
- Twitter tennis mentions
- PedalMe deliveries
- Wikipedia traffic
- COVID forecasting
Are there currently other projects similar to yours? If yes, what are they?
PyTorch Geometric is similar, but it has no support for temporal datasets and it does not cover spatiotemporal deep learning models.
Other similar projects on static graphs:
Repo URL: *
Project license: *
GitHub handles of the project maintainers: *
benedekrozemberczki, SherylHYX , paulmorio
Is there a corporate or academic entity backing the project? If yes, which one?
The University of Edinburgh, University of Cambridge, University of Oxford
Do you have a set of getting started tutorials and complete documentation? *
Documentation is available under read the docs.
We provide a getting started tutorial in the introduction section with explaining cumulative and incremental batching.
How is continuous integration provided today? *
Github Actions deployed on Ubuntu and Windows machines, the CI is integrated with the read the docs based auto documentation generation and the automated code coverage reports are provided by CodeCov.
Does the project support TorchScript today for production usage? *
Not yet. We did not try it.
How long do you expect to maintain the project? *
As long as people do research in this area.