The module object torch._C has no attribute

vision
1

I try to run open source code on GPU where I got an error like this:

File “/usr/local/lib/python2.7/dist-packages/excitationbp-0.1-py2.7.egg/excitationbp/functions/eb_convNd.py”, line 123, in _update_output self._cudnn_info = torch._C._cudnn_convolution_full_forward( AttributeError: ‘module’ object has no attribute ‘_cudnn_convolution_full_forward’

Settings:
Python 2.7.12
[GCC 5.4.0 20160609] on linux2
print(torch.version) 0.4.1

Thanks!

2

It looks like cuDNN is not available.
What does

torch.backends.cudnn.is_available()

return?

3

Thanks for your response!

I do have a GPU.
It return True.

4

I guess you are using this repo.
The error might be due to a mismatch of your PyTorch version. Although they state PyTorch 0.3+ should work, the last update was 6 months ago.
I’m not familiar with the repo, but you could try to create a conda environment with PyTorch 0.3 just to try, if it runs. Have a look here for more information on conda envs.