导入 torch.fx ModuleNotFoundError：没有名为“torch.fx”的模块

Question

After enabling torch and Cuda for my system according to my system GPU compatibility, whenever I am trying to run any program which needs to be run on GPU to enable the system, this error is coming. I could not able to find any solution for this. though I read about this that create another environment and this error will be solved but did not work for me. Please find the details of my system. for reference, I am using Python 3.7.2.

image include torch and Cuda version and device name

Details of NVIDIA Cuda compiler driver

nvidia-smi output

Can anyone help to solve this problem? Thank you

Answer 1

Instead of using

import torch.fx

You have to import from torch.fx as below:

from torch.fx import symbolic_trace

You can view more in the official documentations .

Answer 2

torch.fx was added in PyTorch 1.8.0. Check release post . You're probably using an older version. Upgrade pytorch from website .