Implement skip connections using the JumpingKnowledge pyg module

ORNL / HydraGNN

Distributed PyTorch implementation of multi-headed graph convolutional neural networks

BSD 3-Clause "New" or "Revised" License

61 stars 27 forks source link

Open akhilpandey95 opened 5 months ago

akhilpandey95 commented 5 months ago

modify Base class to instantiate pyg.nn.models.JumpingKnowledge
modify the layer/dimensionality logic to account for concatenated aggregation scheme within JumpingKnowledge
within Base class include a reset_parameters() to reset skip connection choices
modify test cases and log the performance of the model on CI test case.