PyTorchSparseGAT icon indicating copy to clipboard operation
PyTorchSparseGAT copied to clipboard

Published 20 hours ago • verified publisher icon weiyangfb

AttributeError: module 'torch.sparse' has no attribute 'nn'

Open SwiftieH opened this issue 4 years ago • 4 comments

Hi, Thanks for share the implementation.

I have a question about how to call the function 'sparse.nn.functional.dropout' is correctly?

Error： AttributeError: module 'torch.sparse' has no attribute 'nn' is

Many thank!

SwiftieH avatar Apr 14 '20 00:04 SwiftieH

Hi, I also met this problem. Have you solved it? Thanks.

bbjy avatar Feb 24 '21 13:02 bbjy

It needs Pytorch Post release version

YayangLi avatar Mar 04 '21 03:03 YayangLi

It needs Pytorch Post release version

Hi, could you please tell me your program running environment? For example, the version of pytorch, the version of cuda... Thank you so much!

bbjy avatar Mar 04 '21 03:03 bbjy

It needs Pytorch Post release version

Hi, could you please tell me your program running environment? For example, the version of pytorch, the version of cuda... Thank you so much!

I didn't ran this code successfully, but you can achieve softmax and dropout for your own. I modified Diego's pyGAT with sparse.mm and it works. I modified his code because I have the same question with Spmm(#27).

YayangLi avatar Mar 04 '21 11:03 YayangLi