Trouble with CUDA capability sm_86

Hi, all
I am trying to train a network on my NVIDIA RTX A4000. I receive the following error:
NVIDIA RTX A4000 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_70.
pytorch-lightning 1.5.0
torch 1.11.0
torchaudio 0.11.0
torchmetrics 0.6.0
torchvision 0.12.0
CUDA Version : 11.4
Python 3.8.10

You’ve most likely install the pip wheel or conda binary with the CUDA 10.2 runtime, which doesn’t support your Ampere GPU. Select CUDA 11.3 in the install setup here and it should work.

I did it. but after I received the following error:
UserWarning: CUDA initialization: CUDA unknown error - this may be due to an incorrectly set up environment, e.g. changing env variable CUDA_VISIBLE_DEVICES after program start. Setting the available devices to be zero. (Triggered internally at …/c10/cuda/CUDAFunctions.cpp:112.)
return torch._C._cuda_getDeviceCount() > 0
No CUDA runtime is found, using CUDA_HOME=’/usr’

CUDA unknown error could point to anything broken in your setup.
If you get stuck, reinstall the NVIDIA drivers, then the conda binaries again and recheck the functionality.