一、安装pip和vim

sudo apt-get -y install python3-pip vim git

二、配置pip源

mkdir ~/.pip && vim ~/.pip/pip.conf

pip.conf

[global]
index-url = https://mirrors.cloud.tencent.com/pypi/simple
[install]
trusted-host = mirrors.cloud.tencent.com

三、安装显卡驱动和pytorch 

# rocm5.4.2需要Ubuntu内核5.15+
cd ~ && uname -srmv

• Ubuntu 22.04 + rocm 5.4.2

$ wget https://repo.radeon.com/amdgpu-install/5.4.2/ubuntu/jammy/amdgpu-install_5.4.50402-1_all.deb 
$ sudo apt-get install ./amdgpu-install_5.4.50402-1_all.deb
$ sudo apt-get update
$ amdgpu-install -y --usecase=graphics,rocm

# 设置运行权限
$ ls -l /dev/dri/render*
$ sudo usermod -a -G render $LOGNAME
$ sudo usermod -a -G video $LOGNAME

# 重启
$ reboot

$ pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm5.4.2

•  Ubuntu 20.04 + rocm 5.4.2

$ wget https://repo.radeon.com/amdgpu-install/5.4.2/ubuntu/focal/amdgpu-install_5.4.50402-1_all.deb
$ sudo apt-get install ./amdgpu-install_5.4.50402-1_all.deb
$ sudo apt-get update
$ amdgpu-install -y --usecase=graphics,rocm

# 设置运行权限
$ ls -l /dev/dri/render*
$ sudo usermod -a -G render $LOGNAME
$ sudo usermod -a -G video $LOGNAME

# 重启
$ reboot

$ pip3 install torch torchvision torchaudio --index-url https://download.pytorch.org/whl/rocm5.4.2

•  Ubuntu 20.04 + rocm 5.2

$ wget https://repo.radeon.com/amdgpu-install/22.40.3/ubuntu/focal/amdgpu-install_5.4.50403-1_all.deb
$ sudo apt-get install ./amdgpu-install_5.4.50403-1_all.deb
$ sudo apt-get update
$ amdgpu-install -y --usecase=graphics

$ sudo apt-get purge amdgpu-install

$ wget https://repo.radeon.com/amdgpu-install/22.20/ubuntu/focal/amdgpu-install_22.20.50200-1_all.deb 
$ sudo apt-get install ./amdgpu-install_22.20.50200-1_all.deb
$ sudo apt-get update
$ amdgpu-install -y --usecase=rocm

# 设置运行权限
$ ls -l /dev/dri/render*
$ sudo usermod -a -G render $LOGNAME
$ sudo usermod -a -G video $LOGNAME

# 重启
$ reboot

$ pip3 install torch==1.13.1+rocm5.2 torchvision==0.14.1+rocm5.2 torchaudio==0.13.1 --extra-index-url https://download.pytorch.org/whl/rocm5.2

四、检测显卡驱动是否安装成功

# 显示GPU活动信息
$ rocm-smi
# 检查内核模式驱动程序的安装
$ dkms status

# 如果两个命令都列出了GPU，则认为安装成功
$ /opt/rocm-5.4.2/bin/rocminfo
# 或者
$ /opt/rocm-5.4.2/opencl/bin/clinfo

五、检测pytorch是否安装成功

$ vim ~/check.py
$ python3 ~/check.py

 check.py

import torch

print(torch.cuda.is_available())

六、错误解决

• "hipErrorNoBinaryForGpu: Unable to find code object for all current devices!" 已放弃 (核心已转储)

check.py 

import torch
import os

os.environ["HSA_OVERRIDE_GFX_VERSION"] = "10.3.0"

print(torch.cuda.is_available())

 加入环境变量

$ vim ~/.bash_profile
$ source ~/.bash_profile

 .bash_profile末尾加入

export HSA_OVERRIDE_GFX_VERSION=10.3.0

参考资料

pytorch官网
AMD显卡驱动下载

AMD显卡驱动安装文档

ROCm安装文档