Is there Hyperparameter project for graphConv same as SAGEConv

[harirRzi]

Hello. I used a SAGEConv layer with hyper parameter project=True for my model and the prediction is very well. Now I want to improve the prediction of it by using GraphConv using aggr='mean' and edge_weight that predicts similar to SAGEConv without project=True. I checked PyTorch website and I didn't see any hyper parameter 'project' for GraphConv. Is there any alternative for this hyper parameter?

[rusty1s]

project is specific to SAGEConv. You can achieve a similar thing via

self.conv(self.lin(x), edge_index)

but IMO it is not a crucial parameter.