CUDA enabled version installation problem #522
Replies: 3 comments 8 replies
-
It looks like that it cannot find a properly installed torch. Can you try to open python and try 'import torch'? Or better still try:
If that does not work, you might have two python installations and installed the torch under the wrong one. |
Beta Was this translation helpful? Give feedback.
-
@maljoras
It does not work. This is strange because I had already installed pytorch using conda install -c pytorch in miniconda. Could you please give me more suggestions? |
Beta Was this translation helpful? Give feedback.
-
You have |
Beta Was this translation helpful? Give feedback.
-
error 5.txt
When I tried to do CMAKE_PREFIX_PATH=$CONDA_PREFIX make build_cuda flags="-DRPU_CUDA_ARCHITECTURES='86'", I got the error message in the error5.txt
Pytorch version: 2.0.1 torchvision 0.15.2
Package version: aihwkit 0.7.1
OS: Ubuntu 22.04.2 LTS (GNU/Linux 5.15.90.1-microsoft-standard-WSL2 x86_64)
Python version: 3.8.17
NVIDIA GeForce RTX 3060 Laptop
Cuda compilation tools, release 11.5, V11.5.119
Beta Was this translation helpful? Give feedback.
All reactions