1. 安装 Nvidia 显卡驱动
安装 Nvidia 显卡驱动
$ sudo ubuntu-drivers autoinstall
查看驱动的安装情况：
$ nvidia-smi
Thu Aug 26 17:20:29 2021
+-----------------------------------------------------------------------------+
| NVIDIA-SMI 470.57.02 Driver Version: 470.57.02 CUDA Version: 11.4 |
|-------------------------------+----------------------+----------------------+
| GPU Name Persistence-M| Bus-Id Disp.A | Volatile Uncorr. ECC |
| Fan Temp Perf Pwr:Usage/Cap| Memory-Usage | GPU-Util Compute M. |
| | | MIG M. |
|===============================+======================+======================|
| 0 NVIDIA GeForce ... Off | 00000000:01:00.0 On | N/A |
| 32% 43C P8 11W / 180W | 674MiB / 8117MiB | 0% Default |
| | | N/A |
+-------------------------------+----------------------+----------------------+
+-----------------------------------------------------------------------------+
| Processes: |
| GPU GI CI PID Type Process name GPU Memory |
| ID ID Usage |
|=============================================================================|
| 0 N/A N/A 1293 G /usr/lib/xorg/Xorg 96MiB |
| 0 N/A N/A 1985 G /usr/lib/xorg/Xorg 384MiB |
| 0 N/A N/A 2132 G /usr/bin/gnome-shell 45MiB |
| 0 N/A N/A 2502 G ...AAAAAAAAA= --shared-files 95MiB |
| 0 N/A N/A 3281 G ...AAAAAAAA== --shared-files 39MiB |
+-----------------------------------------------------------------------------+
2. 安装 CUDA ToolKit
从这里下载 CUDA ToolKit, 页面上需要根据你自己的系统情况选择合适的版本，最好选择 run file 的版本，会下载所有的文件，包括驱动。
$ wget https://developer.download.nvidia.com/compute/cuda/11.4.1/local_installers/cuda_11.4.1_470.57.02_linux.run
$ sudo sh cuda_11.4.1_470.57.02_linux.run
如果你已经安装了驱动，注意不要再选择安装 cuda 的驱动。
安装完成后, 在家目录的 .bashrc 中加入
PATH=$PATH:/usr/local/cuda-11.4/bin
LD_LIBRARY_PATH=$LD_LIBRARY_PATH:/usr/local/cuda-11.4/lib64
安装 cupy
pip3 install cupy-cuda114
python3 -m cupyx.tools.install_library --cuda 11.4 --library cutensor