Trained weight about pwcnettensorflow

[djl11]

Unfortunately I no longer have the trained weights saved! If you repeat the training process though I would be happy to add your trained weights to the repo.

As for the error, it appears as though the custom ops have not compiled correctly.
There can often be g++ issue for different versions/operating systems etc. with tf custom ops.

Can you tell me what the output for the nvcc and g++ commands are when run in the terminal as opposed to through python subprocess?
These are at lines 48 and 55 in compiled.py respecitvely.
You need to have nvcc runnable from the command line, so the cuda/bin directory should be in your PATH variable.

If you are not using tf 1.8, cuda 9.0, and ubuntu 16.04 though, this is a gray area for me, and with the known difficulties with tf custom ops across different platforms, I will unfortunately have to close the issue.

[saxenarohan97]

@Choneke I have the same error, did you manage to solve it?