Adapt the Pytorch Lightning example to a GConvLSTM

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

Adapt the Pytorch Lightning example to a GConvLSTM #185

Closed AndreaPi closed 1 year ago

AndreaPi commented 2 years ago

The PyTorch Lightning example uses a DCRNN, but I want to use a GConvLSTM. However, this layer doesn't return a single tensor, but three. How should I modify the training_step and the validation_step in order to be able to use a GConvLSTM?