AttributeError: module 'torch._C' has no attribute '_cuda_setDevice' about ganomaly

[rick-lzr]

i get the same problem. did you slove it?

[HaraldusBluumTheSecond]

I come from here pytorch/pytorch#33655

Where it firstly says to comment out that line, which gets us a NEW ERROR, praise the sun..

AssertionError: Torch not compiled with CUDA enabled

[HaraldusBluumTheSecond]

It is mentioned here:
pytorch/pytorch#37022

To download cuda dlls and place them in system32

Like @peterjc123 said, I downloaded additional .dll files from https://drive.google.com/u/0/uc?id=1injUyo3lnarMgWyRcXqKg4UGnN0ysmuq&export=download and copied them to C:\Windows\System32 and it worked.

[rick-lzr]

I solved this problem by a linux sys and gpu device . colab is a good choice. if you are in china you can use many gpu lease platform. and i spend 1 hour to run this code by V100 (ps: i modified batchsize=2048) . hope thoes informations can help others.