我的測試環境
Ubuntu 20.04
NVIDIA-SMI 535.183.01 Driver Version: 535.183.01 CUDA Version: 12.2
用以下的方法，可以確保版本正確的問題
1. **Add NVIDIA package repositories**:
Ensure you have the CUDA repository added to your system. Follow these steps:
```sh
sudo apt-key adv --fetch-keys https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/3bf863cc.pub
sudo sh -c 'echo "deb https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64 /" > /etc/apt/sources.list.d/cuda.list'
sudo sh -c 'echo "deb https://developer.download.nvidia.com/compute/machine-learning/repos/ubuntu2004/x86_64 /" > /etc/apt/sources.list.d/cuda_learn.list'
```
2. **Update the package list**:
```sh
sudo apt update
```
3. **Install CUDA**:
Ensure you have CUDA installed. You can install a specific version if needed.
```sh
sudo apt install cuda-12-0
```
4. **Install cuDNN**:
Install the cuDNN package for CUDA 12.
```sh
sudo apt install libcudnn8=8.9.7.29-1+cuda12.2
sudo apt install libcudnn8-dev=8.9.7.29-1+cuda12.2
sudo apt install libcudnn8-samples=8.9.7.29-1+cuda12.2
```
Replace `8.x.x.x-1+cuda12.0` with the actual version available in the repository. You can check for the available versions using:
```sh
apt-cache policy libcudnn8
```
5. **Verify the installation**:
Ensure CUDA and cuDNN are installed correctly.
```sh
nvcc --version
```
For cuDNN, you can check the version as follows:
```sh
cat /usr/include/cudnn_version.h | grep CUDNN_MAJOR -A 2
```
參考：
https://chatgpt.com/share/cbe6f18b-d8af-4621-8355-874f205f6a86
