PyTorch Geometric SAGEConv - Expected scalar type Long, but found Float?

[BrandenKeck]

I am having tremendous trouble with my torch datatypes in the torch_geometric modules. It seems that no matter what I adjust in my code, SAGEConv( ) is expecting a Long datatype and is receiving a Float datatype in one of its linear layers. I also opened this issue to Stack Overflow here: PyTorch Geometric SAGEConv - Expected scalar type Long, but found Float?

I have trying to troubleshoot using the most basic example I could come up with:

import torch  
from torch_geometric.nn import SAGEConv  
sconv = SAGEConv((-1, -1), 64)  
x = torch.tensor([[1,0],[2,4],[5,7]])  
ei = torch.tensor([[1, 1],[0,2]])  
sconv(x, ei)  

Results in:

RuntimeError: expected m1 and m2 to have the same dtype, but got: __int64 != float

If I print my types, I get:

x.type()
'torch.LongTensor'
ei.type()
'torch.LongTensor'

Any help with this is extremely appreciated.

[rusty1s]

x should be a floating point tensor:

x = torch.tensor([[1,0],[2,4],[5,7]], dtype=torch.float)