Torch does not have access to GPU after Aydin installation

[leopold-franz]

Describe the bug
After following the conda installation instructions from (https://github.com/royerlab/aydin) pytorch does not have access to the GPU.

To Reproduce
Steps to reproduce the behavior:

1. Run conda installation instructions from (https://github.com/royerlab/aydin).
2. Run python in the bash.
3. Run import torch.
4. Run torch.cuda.is_available().

Expected behavior
Expected True but gives False

Screenshots

Desktop (please complete the following information):

• Windows 10 Enterprise LTSC
• GPU: NVIDIA Quadro P1000
• Cuda compilation tools, release 11.6, V11.6.124
  Build cuda_11.6.r11.6/compiler.31057947_0

Additional context
Interestingly after also installing cudnn with conda install cudnn, tensorflow successfully can find the GPU using:
import tensorflow as tf
print("Num GPUs Available: ", len(tf.config.list_physical_devices('GPU')))

[AhmetCanSolak]

Hi @leopold-franz , thank you for the issue. Our migration from TensorFlow to PyTorch has not completed yet but is in the progress. I will make sure to update installation guides accordingly with this migration and make sure we can use torch with GPU.

[AhmetCanSolak]

hello @leopold-franz , given the response here: https://discuss.pytorch.org/t/torch-cuda-installation-issues-and-best-practices-when-torch-is-a-dependency/161992 , soon there will be less compatibility issues between pytorch and cuda drivers/cudatoolkit. We depend on latest version of pytorch and will get updates as they make a new release. Meanwhile have you tried installing torch with this: pip install torch --extra-index-url https://download.pytorch.org/whl/cu116? It might help given your cudatoolkit version is also 11.6