Closed sudhirkumardubey closed 5 months ago
I found solution to this this,
in gnn_pool.py replacing
x = self.convs(x, edge_index) # applying con5v
resolves the error as in pyg_nn.GCN, does not need edge_attr.
Doing this changes, the code runs and gives the output as mentioned in the paper. Please let me if thats correct way to rectify the changes.
I found solution to this this,
in gnn_pool.py replacing
x = self.convs(x, edge_index, edge_atrr) # applying con5v
x = self.convs(x, edge_index) # applying con5v
resolves the error as in pyg_nn.GCN, does not need edge_attr.
Doing this changes, the code runs and gives the output as mentioned in the paper. Please let me if thats correct way to rectify the changes.
I use your solution, but it doesn't work, and still the same error
I also use your solution, but it doesn't work, and still the same error
Hello,
I am trying to replicate your results running the example.ipynb file, but the code shows following error, I have not made any changes to the code still getting this error;
1 stride = 8 2 device = 'cuda' if torch.cuda.is_available() else 'cpu' ----> 4 segment.GNN_seg(mode, epochs, K, pretrained_weights, in_dir, out_dir, save, cc, bs, log_bin, res, facet, layer, stride, 5 device)
TypeError: GNN_seg() missing 2 required positional arguments: 'stride' and 'device'