Googolxx/STF

Using A100 GPU to train the model

Closed this issue · 1 comments

wangfeng9512 commented

When using A100 GPU to train the model "cnn", maybe we will get the following error:

NVIDIA A100-PCIE-40GB with CUDA capability sm_80 is not compatible with the current PyTorch installation.
The current PyTorch install supports CUDA capabilities sm_37 sm_50 sm_60 sm_70.
If you want to use the NVIDIA A100-PCIE-40GB GPU with PyTorch, please check the instructions at https://pytorch.org/get-started/locally/

wangfeng9512 commented

I find the following command can solve the problem:

pip3 install torch==1.8.1+cu111 torchvision==0.9.1+cu111 torchaudio==0.8.1 -f https://download.pytorch.org/whl/torch_stable.html

We can use the above command after ”pip install -e '.[dev]'“.

Reference link: https://discuss.pytorch.org/t/nvidia-a100-gpu-runtimeerror-cudnn-error-cudnn-status-mapping-error/121648