当前位置：AIGC资讯 > AIGC > 正文

手把手教你从0开始在服务器上部署stable diffusion

StableDiffusion 服务器部署

0. 服务器

GPU A5000-24G 数量： 1 显存： 24 GB
CPU AMD EPYC 7551P 核心： 8 核
实例内存： 63G
系统 Ubuntu20.04

验证是否有nvidia驱动

nvidia-smi

如果没有显示出显卡信息（如下）

+---------------------------------------------------------------------------------------+
| NVIDIA-SMI 530.30.02              Driver Version: 530.30.02    CUDA Version: 12.1     |
|-----------------------------------------+----------------------+----------------------+
| 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 RTX A5000                On | 00000000:06:00.0 Off |                  Off |
| 30%   33C    P8               15W / 230W|      1MiB / 24564MiB |      0%      Default |
|                                         |                      |                  N/A |
+-----------------------------------------+----------------------+----------------------+

则需要参考ubuntu安装nvidia驱动

https://blog.csdn.net/Perfect886/article/details/119109380

1. 使用工具

远程连接服务器工具：VS Code

https://code.visualstudio.com/Download

VS Code 插件：Remote

文件传输工具 FileZilla

https://www.filezilla.cn/download

2. 装python环境

下载地址：https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh

./Miniconda3-latest-Linux-x86_64.sh

# 一路回车+yes

在.bashrc中添加以下内容

# >>> conda initialize >>>
# !! Contents within this block are managed by 'conda init' !!
__conda_setup="$('/usr/local/miniconda3/bin/conda' 'shell.bash' 'hook' 2> /dev/null)"
if [ $? -eq 0 ]; then
    eval "$__conda_setup"
else
    if [ -f "/usr/local/miniconda3/etc/profile.d/conda.sh" ]; then
        . "/usr/local/miniconda3/etc/profile.d/conda.sh"
    else
        export PATH="/usr/local/miniconda3/bin:$PATH"
    fi
fi
unset __conda_setup
# <<< conda initialize <<<
# 根据自己的安装路径去修改

3. 设置Conda、PIP国内源

codna 
touch .condarc # 如果有了就不用了

写入以下内容

channels:
  - defaults
show_channel_urls: true
auto_activate_base: true
channel_alias: https://mirrors.tuna.tsinghua.edu.cn/anaconda
default_channels:
  - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/main
  - https://mirrors.tuna.tsinghua.edu.cn/anaconda/pkgs/r
custom_channels:
  conda-forge: https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud
  msys2: https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud
  bioconda: https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud
  menpo: https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud
  pytorch: https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud
  simpleitk: https://mirrors.tuna.tsinghua.edu.cn/anaconda/cloud
pip 
mkdir .pip
cd .pip
touch pip.conf

写入以下内容

[global]
index-url = https://pypi.tuna.tsinghua.edu.cn/simple

4. 设置git加速

该方案为使用镜像网站加速git代码的下载

git config --global url."https://hub.njuu.cf/".insteadOf https://github.com

http://lib.zuotiyi.cn/tool/github.html # 从这里可以找其他镜像网站

注意，镜像网站可能存在滞后

5. 拉取代码

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

6. 创建python虚拟环境并下载依赖包

conda create -n sd python=3.11 -y
conda activate sd
# 优先安装torch会避免很多麻烦
pip install torch torchvision torchaudio
pip install -r requirements.txt 
pip install xformers

可能遇到的问题

在执行的时候可能会卡在安装 basicsr 这里,可以尝试进行以下操作

# 1.下载  tb_nightly
https://files.pythonhosted.org/packages/c0/1c/4d95416f1d9c8bab3d2c3271642084a9a93324a85b1b897ef44fb05ab734/tb_nightly-2.15.0a20230816-py3-none-any.whl
# 2. 安装tb_nightly
pip install tb_nightly-2.15.0a20230816-py3-none-any.whl

# 3.拉取basicsr源码
git clone https://github.com/XPixelGroup/BasicSR.git
# 4. 安装basicsr
cd BasicSR
pip install -r requirements.txt 
python setup.py install

单独安装完basicsr后再执行

pip install -r requirements.txt

出现 ImportError: libGL.so.1: cannot open shared object file: No such file or directory

apt install libgl1-mesa-glx

7. 修改webui.sh

使用现有conda环境而不是新建一个环境

在webui.sh中搜索 use_venv
use_venv=1 (默认) -> use_venv=0

允许root用户运行

在webui.sh中搜索 can_run_as_root
can_run_as_root=0 (默认) -> can_run_as_root=1

8. 修改 paths_internal.py

# stable-diffusion-webui/modules/paths_internal.py
# 搜索 commandline_args
# 将 
commandline_args = os.environ.get('COMMANDLINE_ARGS', "")
# 改为
commandline_args = os.environ.get('COMMANDLINE_ARGS', "--xformers")

9. 安装TCMalloc

sudo apt-get install google-perftools

10. 下载一个模型到指定路径

https://civitai.com/api/download/models/130312

将下载到的模型放入stable-diffusion-webui/models/Stable-diffusion

# 模型网站
https://civitai.com/
https://huggingface.co/

11. 运行webui.sh

./webui.sh

正常启动后，你就能看到vs code贴心的帮你转发了127.0.0.1:7860，通过vs code点击这个链接，你就能看到SD的页面了

完成以上操作就可以给自己用了

好了，可以Ctrl+c关掉程序了

下面的操作是用来给别人用的

12. 修改 webui.py

# webui.py中搜索 api.launch
# 将
    api.launch(
        server_name="0.0.0.0" if cmd_opts.listen else "127.0.0.1",
        port=cmd_opts.port if cmd_opts.port else 7861,
        root_path=f"/{cmd_opts.subpath}" if cmd_opts.subpath else ""
    )
# 修改为
    api.launch(
        server_name="127.0.0.1",
        port=18101, # 自己选一个端口
        root_path="/StableDiffusion/Web"
    )
    
# webui.py中搜索 shared.demo.launch
# 将
        app, local_url, share_url = shared.demo.launch(
            share=cmd_opts.share,
            server_name=server_name,
            server_port=cmd_opts.port,
            ssl_keyfile=cmd_opts.tls_keyfile,
            ssl_certfile=cmd_opts.tls_certfile,
            ssl_verify=cmd_opts.disable_tls_verify,
            debug=cmd_opts.gradio_debug,
            auth=gradio_auth_creds,
            inbrowser=cmd_opts.autolaunch and os.getenv('SD_WEBUI_RESTARTING') != '1',
            prevent_thread_lock=True,
            allowed_paths=cmd_opts.gradio_allowed_path,
            app_kwargs={
                "docs_url": "/docs",
                "redoc_url": "/redoc",
            },
            root_path=f"/{cmd_opts.subpath}" if cmd_opts.subpath else "",
        )
# 修改为
        app, local_url, share_url = shared.demo.launch(
            share=cmd_opts.share,
            server_name="127.0.0.1",
            server_port=18101,
            ssl_keyfile=cmd_opts.tls_keyfile,
            ssl_certfile=cmd_opts.tls_certfile,
            ssl_verify=cmd_opts.disable_tls_verify,
            debug=cmd_opts.gradio_debug,
            auth=gradio_auth_creds,
            inbrowser=cmd_opts.autolaunch and os.getenv('SD_WEBUI_RESTARTING') != '1',
            prevent_thread_lock=True,
            allowed_paths=cmd_opts.gradio_allowed_path,
            app_kwargs={
                "docs_url": "/docs",
                "redoc_url": "/redoc",
            },
            root_path="/StableDiffusion/Web",
        )

13. 安装Nginx

sudo apt install nginx -y
nginx

14. 设置Nginx转发

把配置文件软链到自己目录下好修改

cd ~
ln -s /etc/nginx/ Nginx

# 打开Nginx下的nginx.conf
# 搜索 include /etc/nginx/sites-enabled/*;
# 在这行下添加
include /etc/nginx/myHost/*.conf;

mkdir myHost
cd myHost
touch StableDiffusion.conf

# 在StableDiffusion.conf中写入如下内容
server {
    listen 8080;
    server_name example.com www.example.com;  # Change this to your domain name

    location /StableDiffusion/Web/ {  # Change this if you'd like to server your Gradio app on a different path
        proxy_pass http://127.0.0.1:18101/; # Change this if your Gradio app will be running on a different port
        proxy_redirect off;
        proxy_http_version 1.1;
        proxy_set_header Upgrade $http_upgrade;
        proxy_set_header Connection "upgrade";
        proxy_set_header Host $host;
    }
}

# 验证配置是否正确
nginx -t
# 重启nginx
nginx -s reload

15. 后台启动webui

nohup webui.sh &

codewebwebuiurlgitdiffusionsharedocgradioiconproxyappstablediffusionpythondebugapivs codegithub服务器gpubashivalinuxclonectoctrpytorchelocpushellhuggingfacegoogle虚拟环境网站加速htmlcreate

更新时间 2023-11-25