AssertionError: Torch not compiled with CUDA enabled despite CUDA installation

[Btr4k]

Hello,

I'm encountering an issue where PyTorch is not recognizing CUDA, even though I have CUDA 12.5 installed on my system. Despite following the installation instructions for a CUDA-compatible version of PyTorch, torch.cuda.is_available() returns False, and attempting to utilize CUDA results in the error: "AssertionError: Torch not compiled with CUDA enabled."

Environment:

OS: Windows 11
PyTorch Version: (specify the version you tried installing)
Python Version: 3.8
CUDA Version: 12.5
GPU Model: NVIDIA GeForce RTX 3070
Command Used for Installation: conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch (adjust this according to the actual command you used)

Steps to Reproduce:

Install PyTorch with the specified CUDA toolkit version.
Run python -c "import torch; print(torch.cuda.is_available())".

The expected outcome is True, but it results in False, indicating that CUDA is not enabled in the installed PyTorch build.

Could you please help me understand if I am missing a step, or if there is an issue with compatibility between the specified versions of PyTorch and CUDA? Any guidance would be appreciated.

Thank you!

[Asma-Alkhaldi]

Maybe your installation defaults to a CPU-only version. Make sure you installed the correct version of PyTorch that is compatible with your machine. Check this link: https://pytorch.org/