search

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

Corrected implementation of tgcn example #209

Closed JoelMathewC closed 1 year ago

JoelMathewC commented 1 year ago

Corrected the tgcn_example code to allow previous hidden states to passed for every timestamp. Changes were made in the RecurrentGCN model definition to allow prev_hidden_state to be passed as argument. The RecurrentGCN forward function was updated to return both the hidden state and the output. Changes were made in the training and evaluating portions to pass the previous hidden state as input when the model is called.

JoelMathewC commented 1 year ago

Was taking a look at the issues and realized that an issue was filed for the same at #202

benedekrozemberczki commented 1 year ago

Thank you @JoelMathewC.