SD安装笔记

./webui.sh --server-name=0.0.0.0 --listen --device-id 1

set COMMANDLINE_ARGS=–share

/etc/apt/sources.list.d/cuda-ubuntu2204-12-2-local.list
如果要删除nividia驱动，这个源必须删除。

wget https://developer.download.nvidia.com/compute/cuda/12.2.0/local_installers/cuda_12.2.0_535.54.03_linux.run
sudo sh cuda_12.2.0_535.54.03_linux.run

run文件删除，是成功率比较大的

https://developer.nvidia.com/cuda-12-2-0-download-archive?target_os=Linux&target_arch=x86_64&Distribution=Ubuntu&target_version=22.04&target_type=runfile_local

CUDA的下载页面

sudo vim /etc/modprobe.d/blacklist-nouveau.conf

blacklist nouveau
options nouveau modeset=0

sudo update-initramfs -u

lsmod | grep nouveau

禁用nouveau

sudo ubuntu-drivers devices

查看已经安装的驱动

 import torch

 torch.version

‘2.1.0+cu121’

 torch.cuda.is_available()

查看Torch版本

 import torch

 torch.version

‘2.1.0+cu121’

 torch.cuda.is_available()

查看Torch版本

https://blog.csdn.net/zxdd2018/article/details/127705627

cuDNN的安装

sudo apt-get -f install
修复依赖问题，但是大多数的时候不太好用， 还是得删除了，重新装最干净

sudo apt-get upgrade

cannot import name ‘_compare_version‘ from ‘torchmetrics.utilities.imports‘

https://zhuanlan.zhihu.com/p/619901627
比较全的安装CUDA的方法。

export CUDA_VISIBLE_DEVICES=0,1,2,3

pip config set global.index-url https://mirrors.aliyun.com/pypi/simple

还是不行，然后换成下面的命令：

pip install open-clip-torch
open-clip-torch 使用 阿里的镜像好用

–xformers
–reinstall-torch

sudo .run --silent --driver

CUDA_VISIBLE_DEVICES=0,1,2,3 python launch.py --share

No module ‘xformers’. Proceeding without it.

pip install xformers==0.0.16
pip install xformers-0.0.23.post1

装0.0.16版本，装高版本的没法用。

Installing collected packages: torch, xformers

Attempting uninstall: torch

Found existing installation: torch 1.13.1+cu117

Uninstalling torch-1.13.1+cu117:

Successfully uninstalled torch-1.13.1+cu117

Attempting uninstall: xformers

Found existing installation: xformers 0.0.16

Uninstalling xformers-0.0.16:

Successfully uninstalled xformers-0.0.16

set COMMANDLINE_ARGS=--xformers

webui.sh --xformers