You can specify the version of cudatoolkit you want, or use another channel with support.
Start Locally | PyTorch
docker or python venv are other options. python venv can use the system installed software also by using the ‘–system-site-packages’
Has options of what to install. On 3000 series (or any Ampere GPU) you need CUDA 11.1 or higher support.
conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch
or
pytorch.org states: NOTE: ‘conda-forge’ channel is required for cudatoolkit 11.6
conda install pytorch torchvision torchaudio cudatoolkit=11.6 -c pytorch -c conda-forge
NVIDIA also has a channel so this would work also:
conda install pytorch torchvision torchaudio cudatoolkit=11.6 -c pytorch -c nvidia