Skip to content
(ง ˙o˙)ว

Cuda-relative

Check if you have cuda

nvidia-smi: if the output showing cuda version, then you have cuda (I currently using 12.4)

Check if you have cuda-toolkit

nvcc --version : check the cuda-toolkit version (not the cuda itself version, the toolkit version can be different from cuda version)

Download the cuda-toolkit version you want

Be aware that when go through some posted repository/paper, they are using different toolkit. Cuda can back tolerate, but the cuda-toolkit must match the author’s version

Terminal window
sudo dnf install -y cuda-toolkit-11-8

Set/switch cuda-toolkit version

Terminal window
export CUDA_HOME=/usr/local/cuda-11.8
export PATH=$CUDA_HOME/bin:$PATH
export LD_LIBRARY_PATH=$CUDA_HOME/lib64:$LD_LIBRARY_PATH
source ~/.bashrc
  • after executing PATH, then can check the output of echo $PATH
  • LD_LIBRARY_PATH environment path, tells the system where to look for shared libraries (.so files) when running programs
  • 这是一种临时办法， 并不是永久的，如果是永久就打开~/.bashrc 或 ~/.bash_profile加入这些命令

CUDA_HOME set the path for CUDA TOOLKIT!!

Set/switch cuda-toolkit version in Conda environment

  1. 首先要明白自己现在用的是什么版本的cuda toolkit
    ls -ld /usr/local/cuda 看返回的结果指向哪里
Terminal window
readlink -f /usr/local/cuda
Terminal window
which nvcc
readlink -f $(which nvcc)

Check if you have available cuda (for specific env/proj)

# 检查 CUDA 是否可用
print(f"CUDA 是否可用: {torch.cuda.is_available()}")
if torch.cuda.is_available():
    print(f"CUDA 设备数量: {torch.cuda.device_count()}")
    print(f"当前 CUDA 设备: {torch.cuda.current_device()}")
    print(f"CUDA 设备名称: {torch.cuda.get_device_name(0)}")