How to use NeighborLoader to train a GraphSAGE on multiple graphs

[harirRzi]

Hello. I have multiple graphs and I want to train a GraphSAGE model on them and then test and evaluate on some other graphs. here is my code to make a graphSAGE:

class GraphSAGE(torch.nn.Module):
  """GraphSAGE"""
  def __init__(self, in_channels, hidden_channels, out_channels):
    
    super().__init__()
    self.sage1 = SAGEConv(in_channels, hidden_channels)
    self.linear = torch.nn.Linear(hidden_channels, out_channels)

def forward(self, x, edge_index,edge_weight):  
        x = self.sage1(x, edge_index,edge_weight)
        x = F.relu(x)
        x = F.dropout(x, training=self.training)
        x = self.linear(x)
        out=x

        out = F.softmax(out, dim=1)
        label = out.argmax(dim=-1)
        return label,out

val_dataset = data_list[:22]
test_dataset = data_list[22:46]
train_dataset = data_list[46:]

train_loader = DataLoader(train_dataset, batch_size=1, shuffle=True)
val_loader = DataLoader(val_dataset, batch_size=2)
test_loader = DataLoader(test_dataset, batch_size=2)        
SAGE_model = GraphSAGE(in_channels, hidden_channels, max_num_classes).to(device)

def SAGE_train():
    SAGE_model.train()

    total_loss = 0
    for data in train_loader:
        # Create NeighborLoader instance
        # data= NeighborLoader(data, num_neighbors=num_neighbors,batch_size=1, shuffle=True)
        data = data.to(device)
        data.requires_grad_(True)
        optimizer.zero_grad()
        ....    

Now my question is how should I use NeighborLoader to train my model?

[rusty1s]

I don't think there is a need to use NeighborLoader here at all. If you want to imitate sampling, you can also just use the dropout_edge functionality to drop random edges. Combining DataLoader and NeighborLoader should definitely be avoided.

[harirRzi]

Is GraphSAGE, sensitive on setting a specific parameter so its performance gets better then? I wanted to use weights for the edges of my Graphs but I was not able to use them for GraphSAGE. It seems it doesn't support edge weights.