NVIDIA NVIDIA RTX A5000 with CUDA capability sm_86 is not compatible with the current PyTorch installation

I set up my PyTorch environment following the ’ [environment_gpu.yml’ file from this Github Project https://github.com/graphdeeplearning/graphtransformer, which goes like this:
name: graph_transformer
channels:

  • pytorch
  • dglteam
  • conda-forge
  • fragcolor
  • anaconda
  • defaults
    dependencies:
  • cudatoolkit=10.2
  • cudnn=7.6.5
  • python=3.7.4
  • python-dateutil=2.8.0
  • pip=19.2.3
  • pytorch=1.6.0
  • torchvision==0.7.0
  • pillow==6.1
  • dgl-cuda10.2=0.5.2
  • numpy=1.16.4
  • matplotlib=3.1.0
  • tensorboard=1.14.0
  • tensorboardx=1.8
  • future=0.18.2
  • absl-py
  • networkx=2.3
  • scikit-learn=0.21.2
  • scipy=1.3.0
  • notebook=6.0.0
  • h5py=2.9.0
  • mkl=2019.4
  • ipykernel=5.1.2
  • ipython=7.7.0
  • ipython_genutils=0.2.0
  • ipywidgets=7.5.1
  • jupyter=1.0.0
  • jupyter_client=5.3.1
  • jupyter_console=6.0.0
  • jupyter_core=4.5.0
  • plotly=4.1.1
  • scikit-image=0.15.0
  • requests==2.22.0
  • tqdm==4.43.0
  • pip:
    • tensorflow-gpu==2.1.0
    • tensorflow-estimator==2.1.0
    • tensorboard==2.1.1

However, it raised warning ‘NVIDIA RTX A5000 with CUDA capability sm_86 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA capabilities sm_37 sm_50 sm_60 sm_61 sm_70 sm_75 compute_37.’

It seems that my environment is unable to work, and the command ‘model.to(device)’ just froze and nothing shows up.

I have read other related posts that suggests I could solve the problem with:

conda install pytorch torchvision torchaudio cudatoolkit=11.3 -c pytorch

But sadly it shows conflicts and does not work as well, can anybody help me with this problem?

Your Ampere GPU needs CUDA11+ while your current installation uses cudatoolkit=10.2 and will thus not work. The CUDA10.2 binaries are also not shipping sm_86 for the same reason.
Install the CUDA11.3 binary and it should work. If you are hitting conflicts, create a new conda environment and install it again.

1 Like