【AIGC】如何在Windows/Linux上部署stable diffusion

文章目录

整体安装步骤

windows10安装stable diffusion

环境要求

安装步骤

注意事项

参考博客

其他事项

安装显卡驱动

https://www.volcengine.com/docs/6419/69858

安装cuda

cuda版本与显卡驱动对照表
https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html

要安装的显卡驱动是，所以对应的cuda版本应该是 CUDA 12.0.x 、12.1.x等

安装成功后可以在cmd使用nvcc -V查看cuda版本，使用nsmi

好像需要安装cuda11.8版本？
下载链接：https://developer.nvidia.com/cuda-12-1-0-download-archive?target_os=Windows&target_arch=x86_64&target_version=10&target_type=exe_local

检查环境

python -m torch.utils.collect_env

卸载cuda

cuda 和 cudnn 库的卸载与安装：https://zhuanlan.zhihu.com/p/102966512

安装对应版本pytorch

在以下网址搜索：2.0.1+cu118
https://download.pytorch.org/whl/torch/
其中cu118对应刚刚安装的cuda11.8版本

安装git上的python包

cd path to CLIP
# python setup.py install --user
python setup.py install

Q&A

解决方法：运行命令修改为./webui.sh --precision full --no-half

在webui-user.bat中添加命令行参数：
set COMMANDLINE_ARGS=–skip-torch-cuda-test --precision full --no-half

What’s happening is that, since you’re not running on localhost and everyone in the whole wide world can in theory access your Web UI, you need to explicitly enable extensions.

解决方法：adding the --enable-insecure-extension-access flag

linux安装stable diffusion

简单版可查看之前的博客：【AIGC】阿里云ecs部署stable diffusion

安装anaconda

安装Python、wget、git
sudo apt install python3 python3-pip python3-virtualenv wget git

安装前置依赖
sudo apt-get install libgl1-mesa-glx libegl1-mesa libxrandr2 libxrandr2 libxss1 libxcursor1 libxcomposite1 libasound2 libxi6 libxtst6

下载anaconda
wget https://repo.anaconda.com/archive/Anaconda3-2023.03-1-Linux-x86_64.sh

创建Python 3.10.6的虚拟环境
conda create --name sdwebui python=3.10.6

安装cuda

os=ubuntu2004
arch=x86_64
wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2004/x86_64/cuda-keyring_1.0-1_all.deb
sudo dpkg -i cuda-keyring_1.0-1_all.deb
sudo apt update
sudo apt install cuda-11-8

git 加速

开启加速
export http_proxy=http://172.181.217.43:12798 && export https_proxy=http://172.181.217.43:12798
关闭加速
unset http_proxy && unset https_proxy

【todo：注意关闭配置】https://blog.csdn.net/zhiboqingyun/article/details/123912058

git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git

配置虚拟环境

启动虚拟环境
conda activate sdwebui

如果想要于bash script自动以conda启动虚拟环境，可以把这二行加入至webui-user.sh顶端：
eval “$(conda shell.bash hook)”
conda activate sdwebui

下载模型
wget -nc -P ~/stable-diffusion-webui/models/Stable-diffusion https://huggingface.co/andite/anything-v4.0/resolve/main/anything-v4.5-pruned.safetensors -O anything-v4.5-pruned.safetensors

首次启动
./webui.sh

./webui.sh --xformers --skip-install --enable-insecure-extension-access

挂载oss（optional）

https://help.aliyun.com/document_detail/153892.html?spm=a2c4g.153893.0.0.1d877f1fHkYVYt

端口开放

https://www.bilibili.com/read/cv23060248

启动服务时需要添加 --listen参数
./webui.sh --listen

sed -i ‘s/can_run_as_root=0/can_run_as_root=1/g’ webui.sh &&
./webui.sh --no-download-sd-model --xformers

启动后的一些权限问题

sudo chown -R root stable-diffusion-webui
sudo chgrp -R root stable-diffusion-webui

服务地址：http://xxx.xxx.xxx.xxx:7860/

启动服务

sudo ./webui.sh --listen --skip-install --enable-insecure-extension-access

controlnet

https://github.com/Mikubill/sd-webui-controlnet/issues/1399

Package cairo was not found in the pkg-config search path.
Perhaps you should add the directory containing `cairo.pc’
to the PKG_CONFIG_PATH environment variable
No package ‘cairo’ found
ERROR: Could not build wheels for pycairo, which is required to install pyproject.toml-based projects

sudo apt install libcairo2-dev

I fixed itby this way:

run command below on linux
sudo apt install libcairo2-dev

install svglib package manually or relaunch webui
pip install svglib

显卡使用监控

watch -n 1 nvidia-smi

插件安装

抠图插件安装：https://zhuanlan.zhihu.com/p/632888046

参考文章

查看Linux服务器配置|是否有GPU以及GPU信息：https://blog.csdn.net/qq_41725313/article/details/123784429

安装至Linux：https://ivonblog.com/posts/stable-diffusion-webui-manuals/zh-cn/installation/linux-installation/

安装驱动：https://ivonblog.com/posts/ubuntu-install-nvidia-drivers/

安装驱动（阿里云官方文档）：https://help.aliyun.com/document_detail/163824.htm?spm=a2c4g.298150.0.0.1e0b5355t3jXPh#concept-ecy-qrz-wgb

脚本安装stable diffusion

# git配置
git config --global http.postBuffer 524288000
git config --global https.postBuffer 524288000
git config --global core.compression -1
git config --global http.sslVerify "false"

# 拉取stable diffusion
cd ~
git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git
cd ~/stable-diffusion-webui
git checkout 89f9faa63388756314e8a1d96cf86bf5e0663045

# 配置python环境
conda create -y --name sdwebui python=3.10.6
source activate sdwebui
python -m venv venv
source venv/bin/activate            # conda activate sdwebui 可能会失败
# source venv/Scripts/activate      # for win10 git bash

# 安装GPU版torch
pip install torch==2.0.0+cu117 torchvision==0.15.0+cu117 --extra-index-url https://download.pytorch.org/whl/cu117
pip install xformers==0.0.17

# python github环境包
mkdir tmp_envs
cd tmp_envs
git config --global http.sslVerify "false"      # 加上，以防断连
git clone https://github.com/TencentARC/GFPGAN.git
cd GFPGAN && git checkout 8d2447a2d918f8eba5a4a01463fd48e45126a379 
pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple
python setup.py install
cd ..

git config --global http.sslVerify "false"
git clone https://github.com/openai/CLIP.git
cd CLIP && git checkout d50d76daa670286dd6cacf3bcd80b5e4823fc8e1
pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple
python setup.py install
cd ..

git config --global http.sslVerify "false"
git clone https://github.com/mlfoundations/open_clip.git
cd open_clip && git checkout bb6e834e9c70d9c27d0dc3ecedeebeaeb1ffad6b
pip install -r requirements.txt -i https://pypi.tuna.tsinghua.edu.cn/simple
python setup.py install
cd ..


# 下载模型
wget -nc -P ~/stable-diffusion-webui/models/Stable-diffusion https://huggingface.co/andite/anything-v4.0/resolve/main/anything-v4.5-pruned.safetensors -O anything-v4.5-pruned.safetensors

./webui.sh --xformers --skip-install


# 面部修复功能报错
# 下载模型文件：
cd ~/stable-diffusion-webui
wget https://github.com/sczhou/CodeFormer/releases/download/v0.1.0/codeformer.pth
mv codeformer.pth ~/stable-diffusion-webui/models/Codeformer/codeformer-v0.1.0.pt
# 下载权重文件
wget https://github.com/xinntao/facexlib/releases/download/v0.1.0/detection_Resnet50_Final.pth
mv detection_Resnet50_Final.pth ~/stable-diffusion-webui/repositories/CodeFormer/weights/facelib/detection_Resnet50_Final.pth
wget https://github.com/sczhou/CodeFormer/releases/download/v0.1.0/parsing_parsenet.pth
mv parsing_parsenet.pth ~/stable-diffusion-webui/repositories/CodeFormer/weights/facelib/parsing_parsenet.pth

stable diffusion教程

models

stable-diffusion-v1-5: https://huggingface.co/runwayml/stable-diffusion-v1-5/tree/main

The 10 Best Stable Diffusion Models by Popularity (SD Models Explained)：https://aituts.com/models/

AI 绘画与作画 stable diffusion webui 常见模型汇总及简介：https://www.tjsky.net/tutorial/583

C站下载模型：https://civitai.com/

hugging face下载：https://huggingface.co/models

ControlNet下载： https://huggingface.co/lllyasviel/ControlNet-v1-1/tree/main

lora模型下载：https://openai.wiki/lora-model-part-1.html

tutorial

AIGC入门教程：Stable Diffusion，万字保姆篇：https://www.woshipm.com/ai/5813208.html

超详细！外婆都能看懂的Stable Diffusion入门教程：https://www.uisdc.com/stable-diffusion-3

How to use Stable Diffusion v2.1 (AUTOMATIC1111 Guide)：https://aituts.com/install-stable-diffusion-v2-1/

stable-diffusion-webui使用手册：https://ivonblog.com/posts/stable-diffusion-webui-manuals/zh-cn/installation/windows-installation/

Stable Diffusion 速览: https://www.zhihu.com/question/550101073/answer/2931261853

AI绘画指南 stable diffusion webui （SD webui）如何设置与使用：https://www.tjsky.net/tutorial/488

How to use models（LoRA）：https://github.com/civitai/civitai/wiki/How-to-use-models#lora

How to generate realistic people in Stable Diffusion：https://stable-diffusion-art.com/realistic-people/

How to make a video with Stable Diffusion (Deforum)：https://stable-diffusion-art.com/deforum/

AI绘画教程：从入门到放弃(xiaobai)：https://zhuanlan.zhihu.com/p/607892849

LoRA模型的训练脚本：https://github.com/kohya-ss/sd-scripts

prompt reference

风格提示词：https://lexica.art/

image on playgroundai: https://playgroundai.com/c/landscapes