【AI作画】使用stable-diffusion-webui搭建AI作画平台

一、安装配置Anaconda

进入官网下载安装包https://www.anaconda.com/并安装，然后将Anaconda配置到环境变量中。

打开命令行，依次通过如下命令创建Python运行虚拟环境。

conda env create novelai python==3.10.6

E:\workspace\02_Python\novalai>conda info --envs
# conda environments:
#
base                  *  D:\anaconda3
novelai                  D:\anaconda3\envs\novelai

conda activate novelai

二、安装CUDA

笔者的显卡为NVIDIA，需安装NVIDIA的开发者工具进入官网https://developer.nvidia.com/，根据自己计算机的系统情况，选择合适的安装包下载安装。

打开安装程序后，依照提示完成安装。

安装完成后，在命令窗口输入如下命令，输出CUDA版本即安装成功。

C:\Users\yefuf>nvcc -V
nvcc: NVIDIA (R) Cuda compiler driver
Copyright (c) 2005-2022 NVIDIA Corporation
Built on Wed_Sep_21_10:41:10_Pacific_Daylight_Time_2022
Cuda compilation tools, release 11.8, V11.8.89
Build cuda_11.8.r11.8/compiler.31833905_0

三、安装pytorch

进入官网https://pytorch.org/，根据计算机配置选择合适的版本进行安装。这里需要注意的是CUDA的平台选择，先打开NVIDIA控制面板-帮助-系统信息-组件查看CUDA版本，官网上选择的计算平台需要低于计算机的NVIDIA版本。

配置选择完成后，官网会生成相应的安装命令。

将安装命令复制出，命令窗口执行安装即可。

conda install pytorch torchvision torchaudio pytorch-cuda=11.6 -c pytorch -c nvidia

当查到Pytorch官网推荐的CUDA版本跟你的显卡版本不匹配时，就需要根据官网的CUDA版本找到对应的显卡驱动版本并升级显卡驱动，对应关系可通过https://docs.nvidia.com/cuda/cuda-toolkit-release-notes/index.html查看

四、安装git

进入git官网https://git-scm.com/，下载安装即可。

五、搭建stable-diffusion-webui

进入项目地址https://github.com/AUTOMATIC1111/stable-diffusion-webui，通过git将项目克隆下来。

git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui.git
Cloning into 'stable-diffusion-webui'...
remote: Enumerating objects: 10475, done.
remote: Counting objects: 100% (299/299), done.
remote: Compressing objects: 100% (199/199), done.
remote: Total 10475 (delta 178), reused 199 (delta 100), pack-reused 10176
Receiving objects: 100% (10475/10475), 23.48 MiB | 195.00 KiB/s, done.
Resolving deltas: 100% (7312/7312), done.

克隆下载扩展库。

git clone https://github.com/AUTOMATIC1111/stable-diffusion-webui-aesthetic-gradients “extensions/aesthetic-gradients”
Cloning into 'extensions/aesthetic-gradients'...
remote: Enumerating objects: 21, done.
remote: Counting objects: 100% (21/21), done.
remote: Compressing objects: 100% (12/12), done.
remote: Total 21 (delta 3), reused 18 (delta 3), pack-reused 0
Receiving objects: 100% (21/21), 1.09 MiB | 1.34 MiB/s, done.
Resolving deltas: 100% (3/3), done.

git clone https://github.com/yfszzx/stable-diffusion-webui-images-browser “extensions/images-browser”
Cloning into 'extensions/images-browser'...
remote: Enumerating objects: 118, done.
remote: Counting objects: 100% (118/118), done.
remote: Compressing objects: 100% (70/70), done.
remote: Total 118 (delta 42), reused 65 (delta 24), pack-reused 0
Receiving objects: 100% (118/118), 33.01 KiB | 476.00 KiB/s, done.
Resolving deltas: 100% (42/42), done.

克隆完成后，extensions目录会多如下文件夹：

下载模型库https://github.com/AUTOMATIC1111/stable-diffusion-webui/wiki/Dependencies，并将下载的.ckpt

放到models/Stable-diffusion文件夹中。模型很大，推荐使用下载器。

安装项目所需的Python依赖库。

pip install -r requirements.txt

安装完成之后，运行如下命令，顺利的话，当程序加载完成模型之后，会自动打开http://127.0.0.1:7860/显示平台主页。

python launch.py --autolaunch

进入平台的设置页面，选择语言为中文，重启程序之后，即可看到页面显示为中文。

在界面中输入作画内容的正向提示词（画想要什么特征）和反向提示词（画不想要什么特征），点击生成即可开始自动作画。

如上述的提示词作出的画如图（由于随机种子不同，生成的画会有差异）。

六、如何设置提示词

这里建议使用元素法典https://docs.qq.com/doc/DWHl3am5Zb05QbGVs，上面有前人整理好的提示词及效果，以供参考。

七、可能遇到的问题

1、GitHub访问不了或访问慢

一般为DNS解析问题，需要修改本地host文件，增加配置内容，绕过域名解析，达到加速访问的目的。

访问https://www.ipaddress.com/，分别输入github.com和github.global.ssl.fastly.net，获取域名对应的IP地址。

打开系统的Host文件，将IP和域名的对应关系配置到Host文件中。

配置文件内容如下：

140.82.114.4   github.com
199.232.5.194 github.global.ssl.fastly.net

执行命令ipconfig /flushdns刷新DNS即可。

2、pip安装依赖库慢或常下载失败

pip安装依赖库时默认选择国外的源，安装速度会非常慢，可以考虑切换为国内源，常用的国内源如下：

阿里云 https://mirrors.aliyun.com/pypi/simple/
中国科技大学 https://pypi.mirrors.ustc.edu.cn/simple/
豆瓣(douban) https://pypi.douban.com/simple/
清华大学 https://pypi.tuna.tsinghua.edu.cn/simple/
中国科学技术大学 https://pypi.mirrors.ustc.edu.cn/simple/

在安装依赖库时，可使用pip install -i 源空格安装包名称进行源的选择，如pip install -i https://mirrors.aliyun.com/pypi/simple numpy。

也可以通过增加配置文件，使安装依赖库时默认选择国内的源，在用户目录下增加pip.ini文件。

在文件中写入如下内容。

[global]
timeout = 60000
index-url = https://pypi.tuna.tsinghua.edu.cn/simple[install]
use-mirrors = true
mirrors = https://pypi.tuna.tsinghua.edu.cn

3、安装CLIP时提示Connection was aborted, errno 10053

出错时的错误打印如下：

(novelai) E:\workspace\02_Python\novalai\stable-diffusion-webui>python launch.py
Python 3.10.6 | packaged by conda-forge | (main, Oct 24 2022, 16:02:16) [MSC v.1916 64 bit (AMD64)]
Commit hash: b8f2dfed3c0085f1df359b9dc5b3841ddc2196f0
Installing clip
Traceback (most recent call last):File "E:\workspace\02_Python\novalai\stable-diffusion-webui\launch.py", line 251, in <module>prepare_enviroment()File "E:\workspace\02_Python\novalai\stable-diffusion-webui\launch.py", line 178, in prepare_enviromentrun_pip(f"install {clip_package}", "clip")File "E:\workspace\02_Python\novalai\stable-diffusion-webui\launch.py", line 63, in run_pipreturn run(f'"{python}" -m pip {args} --prefer-binary{index_url_line}', desc=f"Installing {desc}", errdesc=f"Couldn't install {desc}")File "E:\workspace\02_Python\novalai\stable-diffusion-webui\launch.py", line 34, in runraise RuntimeError(message)
RuntimeError: Couldn't install clip.
Command: "D:\anaconda3\envs\novelai\python.exe" -m pip install git+https://github.com/openai/CLIP.git@d50d76daa670286dd6cacf3bcd80b5e4823fc8e1 --prefer-binary
Error code: 1
stdout: Looking in indexes: https://pypi.tuna.tsinghua.edu.cn/simple
Collecting git+https://github.com/openai/CLIP.git@d50d76daa670286dd6cacf3bcd80b5e4823fc8e1Cloning https://github.com/openai/CLIP.git (to revision d50d76daa670286dd6cacf3bcd80b5e4823fc8e1) to c:\users\yefuf\appdata\local\temp\pip-req-build-f8w7kbzgstderr:   Running command git clone --filter=blob:none --quiet https://github.com/openai/CLIP.git 'C:\Users\yefuf\AppData\Local\Temp\pip-req-build-f8w7kbzg'fatal: unable to access 'https://github.com/openai/CLIP.git/': OpenSSL SSL_read: Connection was aborted, errno 10053error: subprocess-exited-with-errorgit clone --filter=blob:none --quiet https://github.com/openai/CLIP.git 'C:\Users\yefuf\AppData\Local\Temp\pip-req-build-f8w7kbzg' did not run successfully.exit code: 128See above for output.note: This error originates from a subprocess, and is likely not a problem with pip.
error: subprocess-exited-with-errorgit clone --filter=blob:none --quiet https://github.com/openai/CLIP.git 'C:\Users\yefuf\AppData\Local\Temp\pip-req-build-f8w7kbzg' did not run successfully.
exit code: 128See above for output.note: This error originates from a subprocess, and is likely not a problem with pip.

通过访CLIP项目GitHub主页，发现该项目可以通过如下命令进行安装解决。

pip install ftfy regex tqdm
pip install git+https://github.com/openai/CLIP.git

4、项目启动中提示Connection was reset in connection to github.com

出错时的错误打印如下：

(novelai) E:\workspace\02_Python\novalai\stable-diffusion-webui>python launch.py
Python 3.10.6 | packaged by conda-forge | (main, Oct 24 2022, 16:02:16) [MSC v.1916 64 bit (AMD64)]
Commit hash: b8f2dfed3c0085f1df359b9dc5b3841ddc2196f0
Cloning Stable Diffusion into repositories\stable-diffusion...
Cloning Taming Transformers into repositories\taming-transformers...
Traceback (most recent call last):File "E:\workspace\02_Python\novalai\stable-diffusion-webui\launch.py", line 251, in <module>prepare_enviroment()File "E:\workspace\02_Python\novalai\stable-diffusion-webui\launch.py", line 201, in prepare_enviromentgit_clone(taming_transformers_repo, repo_dir('taming-transformers'), "Taming Transformers", taming_transformers_commit_hash)File "E:\workspace\02_Python\novalai\stable-diffusion-webui\launch.py", line 85, in git_clonerun(f'"{git}" clone "{url}" "{dir}"', f"Cloning {name} into {dir}...", f"Couldn't clone {name}")File "E:\workspace\02_Python\novalai\stable-diffusion-webui\launch.py", line 34, in runraise RuntimeError(message)
RuntimeError: Couldn't clone Taming Transformers.
Command: "git" clone "https://github.com/CompVis/taming-transformers.git" "repositories\taming-transformers"
Error code: 128
stdout: <empty>
stderr: Cloning into 'repositories\taming-transformers'...
fatal: unable to access 'https://github.com/CompVis/taming-transformers.git/': OpenSSL SSL_connect: Connection was reset in connection to github.com:443

在命令窗口中输入如下命令，然后重新运行程序，但实际操作下来，仍有较大概率在克隆项目的过程中失败。

git config --global http.postBuffer 524288000
git config --global http.sslVerify "false"

查看lauch.py中的代码可以发现，程序在启动时有对依赖项目进行检查，如项目不存在，则克隆下来。

def prepare_enviroment():torch_command = os.environ.get('TORCH_COMMAND', "pip install torch==1.12.1+cu113 torchvision==0.13.1+cu113 --extra-index-url https://download.pytorch.org/whl/cu113")requirements_file = os.environ.get('REQS_FILE', "requirements_versions.txt")commandline_args = os.environ.get('COMMANDLINE_ARGS', "")gfpgan_package = os.environ.get('GFPGAN_PACKAGE', "git+https://github.com/TencentARC/GFPGAN.git@8d2447a2d918f8eba5a4a01463fd48e45126a379")clip_package = os.environ.get('CLIP_PACKAGE', "git+https://github.com/openai/CLIP.git@d50d76daa670286dd6cacf3bcd80b5e4823fc8e1")deepdanbooru_package = os.environ.get('DEEPDANBOORU_PACKAGE', "git+https://github.com/KichangKim/DeepDanbooru.git@d91a2963bf87c6a770d74894667e9ffa9f6de7ff")xformers_windows_package = os.environ.get('XFORMERS_WINDOWS_PACKAGE', 'https://github.com/C43H66N12O12S2/stable-diffusion-webui/releases/download/f/xformers-0.0.14.dev0-cp310-cp310-win_amd64.whl')stable_diffusion_repo = os.environ.get('STABLE_DIFFUSION_REPO', "https://github.com/CompVis/stable-diffusion.git")taming_transformers_repo = os.environ.get('TAMING_REANSFORMERS_REPO', "https://github.com/CompVis/taming-transformers.git")k_diffusion_repo = os.environ.get('K_DIFFUSION_REPO', 'https://github.com/crowsonkb/k-diffusion.git')codeformer_repo = os.environ.get('CODEFORMET_REPO', 'https://github.com/sczhou/CodeFormer.git')blip_repo = os.environ.get('BLIP_REPO', 'https://github.com/salesforce/BLIP.git')stable_diffusion_commit_hash = os.environ.get('STABLE_DIFFUSION_COMMIT_HASH', "69ae4b35e0a0f6ee1af8bb9a5d0016ccb27e36dc")taming_transformers_commit_hash = os.environ.get('TAMING_TRANSFORMERS_COMMIT_HASH', "24268930bf1dce879235a7fddd0b2355b84d7ea6")k_diffusion_commit_hash = os.environ.get('K_DIFFUSION_COMMIT_HASH', "f4e99857772fc3a126ba886aadf795a332774878")codeformer_commit_hash = os.environ.get('CODEFORMER_COMMIT_HASH', "c5b4593074ba6214284d6acd5f1719b6c5d739af")blip_commit_hash = os.environ.get('BLIP_COMMIT_HASH', "48211a1594f1321b00f14c9f7a5b4813144b2fb9")

因此，我们打开git bash重新执行上述两条git命令，预先将项目克隆下来。

git clone https://github.com/CompVis/taming-transformers.git "repositories\taming-transformers"git clone https://github.com/crowsonkb/k-diffusion.git "repositories\k-diffusion"git clone https://github.com/sczhou/CodeFormer.git "repositories\CodeFormer"git clone https://github.com/salesforce/BLIP.git "repositories\BLIP"

克隆完成之后如图：

5、项目启动中提示CUDA out of memory

出错时的错误打印如下：

(novelai) E:\workspace\02_Python\novalai\stable-diffusion-webui>python launch.py
Python 3.10.6 | packaged by conda-forge | (main, Oct 24 2022, 16:02:16) [MSC v.1916 64 bit (AMD64)]
Commit hash: b8f2dfed3c0085f1df359b9dc5b3841ddc2196f0
Fetching updates for BLIP...
Checking out commit for BLIP with hash: 48211a1594f1321b00f14c9f7a5b4813144b2fb9...
Installing requirements for CodeFormer
Installing requirements for Web UI
Launching Web UI with arguments:
Moving sd-v1-4.ckpt from E:\workspace\02_Python\novalai\stable-diffusion-webui\models to E:\workspace\02_Python\novalai\stable-diffusion-webui\models\Stable-diffusion.
LatentDiffusion: Running in eps-prediction mode
DiffusionWrapper has 859.52 M params.
making attention of type 'vanilla' with 512 in_channels
Working with z of shape (1, 4, 32, 32) = 4096 dimensions.
making attention of type 'vanilla' with 512 in_channels
Downloading: 100%|██████████████████████████████████████████████████████████████████| 939k/939k [00:00<00:00, 1.26MB/s]
Downloading: 100%|███████████████████████████████████████████████████████████████████| 512k/512k [00:01<00:00, 344kB/s]
Downloading: 100%|████████████████████████████████████████████████████████████████████████████| 389/389 [00:00<?, ?B/s]
Downloading: 100%|████████████████████████████████████████████████████████████████████████████| 905/905 [00:00<?, ?B/s]
Downloading: 100%|████████████████████████████████████████████████████████████████████████| 4.41k/4.41k [00:00<?, ?B/s]
Downloading: 100%|████████████████████████████████████████████████████████████████| 1.59G/1.59G [03:56<00:00, 7.23MB/s]
Loading weights [7460a6fa] from E:\workspace\02_Python\novalai\stable-diffusion-webui\models\Stable-diffusion\sd-v1-4.ckpt
Global Step: 470000
Traceback (most recent call last):File "E:\workspace\02_Python\novalai\stable-diffusion-webui\launch.py", line 252, in <module>start()File "E:\workspace\02_Python\novalai\stable-diffusion-webui\launch.py", line 247, in startwebui.webui()File "E:\workspace\02_Python\novalai\stable-diffusion-webui\webui.py", line 148, in webuiinitialize()File "E:\workspace\02_Python\novalai\stable-diffusion-webui\webui.py", line 83, in initializemodules.sd_models.load_model()File "E:\workspace\02_Python\novalai\stable-diffusion-webui\modules\sd_models.py", line 252, in load_modelsd_model.to(shared.device)File "D:\anaconda3\envs\novelai\lib\site-packages\pytorch_lightning\core\mixins\device_dtype_mixin.py", line 113, in toreturn super().to(*args, **kwargs)File "D:\anaconda3\envs\novelai\lib\site-packages\torch\nn\modules\module.py", line 987, in toreturn self._apply(convert)File "D:\anaconda3\envs\novelai\lib\site-packages\torch\nn\modules\module.py", line 639, in _applymodule._apply(fn)File "D:\anaconda3\envs\novelai\lib\site-packages\torch\nn\modules\module.py", line 639, in _applymodule._apply(fn)File "D:\anaconda3\envs\novelai\lib\site-packages\torch\nn\modules\module.py", line 639, in _applymodule._apply(fn)[Previous line repeated 2 more times]File "D:\anaconda3\envs\novelai\lib\site-packages\torch\nn\modules\module.py", line 662, in _applyparam_applied = fn(param)File "D:\anaconda3\envs\novelai\lib\site-packages\torch\nn\modules\module.py", line 985, in convertreturn t.to(device, dtype if t.is_floating_point() or t.is_complex() else None, non_blocking)
torch.cuda.OutOfMemoryError: CUDA out of memory. Tried to allocate 20.00 MiB (GPU 0; 2.00 GiB total capacity; 1.68 GiB already allocated; 0 bytes free; 1.72 GiB reserved in total by PyTorch) If reserved memory is >> allocated memory try setting max_split_size_mb to avoid fragmentation.  See documentation for Memory Management and PYTORCH_CUDA_ALLOC_CONF

根据提示，先尝试用如下命令改变pytorch配置，仍旧报错!

set PYTORCH_CUDA_ALLOC_CONF=garbage_collection_threshold:0.6,max_split_size_mb:128

尝试增加代码with torch.no_grad()，使内存就不会分配参数梯度的空间，仍旧报错！

由于提示内存溢出，先通过控制面板->所有控制面板项->管理工具->系统信息，查看显卡内存大小。

官方推荐的显卡内存大小为4GB以上，而笔者的显卡内存只有2GB，显然GPU不符合要求。查看项目的命令选项，发现项目支持CPU计算--use-cpu。

(novelai) E:\workspace\02_Python\novalai\stable-diffusion-webui>python launch.py -h
Python 3.10.6 | packaged by conda-forge | (main, Oct 24 2022, 16:02:16) [MSC v.1916 64 bit (AMD64)]
Commit hash: b8f2dfed3c0085f1df359b9dc5b3841ddc2196f0
Installing requirements for Web UI
Launching Web UI with arguments: -h
usage: launch.py [-h] [--config CONFIG] [--ckpt CKPT] [--ckpt-dir CKPT_DIR] [--gfpgan-dir GFPGAN_DIR][--gfpgan-model GFPGAN_MODEL] [--no-half] [--no-half-vae] [--no-progressbar-hiding][--max-batch-count MAX_BATCH_COUNT] [--embeddings-dir EMBEDDINGS_DIR][--hypernetwork-dir HYPERNETWORK_DIR] [--localizations-dir LOCALIZATIONS_DIR] [--allow-code][--medvram] [--lowvram] [--lowram] [--always-batch-cond-uncond] [--unload-gfpgan][--precision {full,autocast}] [--share] [--ngrok NGROK] [--ngrok-region NGROK_REGION][--enable-insecure-extension-access] [--codeformer-models-path CODEFORMER_MODELS_PATH][--gfpgan-models-path GFPGAN_MODELS_PATH] [--esrgan-models-path ESRGAN_MODELS_PATH][--bsrgan-models-path BSRGAN_MODELS_PATH] [--realesrgan-models-path REALESRGAN_MODELS_PATH][--scunet-models-path SCUNET_MODELS_PATH] [--swinir-models-path SWINIR_MODELS_PATH][--ldsr-models-path LDSR_MODELS_PATH] [--clip-models-path CLIP_MODELS_PATH] [--xformers][--force-enable-xformers] [--deepdanbooru] [--opt-split-attention] [--opt-split-attention-invokeai][--opt-split-attention-v1] [--disable-opt-split-attention][--use-cpu {all,sd,interrogate,gfpgan,swinir,esrgan,scunet,codeformer} [{all,sd,interrogate,gfpgan,swinir,esrgan,scunet,codeformer} ...]][--listen] [--port PORT] [--show-negative-prompt] [--ui-config-file UI_CONFIG_FILE][--hide-ui-dir-config] [--freeze-settings] [--ui-settings-file UI_SETTINGS_FILE] [--gradio-debug][--gradio-auth GRADIO_AUTH] [--gradio-img2img-tool {color-sketch,editor}] [--opt-channelslast][--styles-file STYLES_FILE] [--autolaunch] [--theme THEME] [--use-textbox-seed][--disable-console-progressbars] [--enable-console-prompts] [--vae-path VAE_PATH][--disable-safe-unpickle] [--api] [--nowebui] [--ui-debug-mode] [--device-id DEVICE_ID][--administrator] [--cors-allow-origins CORS_ALLOW_ORIGINS] [--tls-keyfile TLS_KEYFILE][--tls-certfile TLS_CERTFILE] [--server-name SERVER_NAME]options:-h, --help            show this help message and exit--config CONFIG       path to config which constructs model--ckpt CKPT           path to checkpoint of stable diffusion model; if specified, this checkpoint will be added tothe list of checkpoints and loaded--ckpt-dir CKPT_DIR   Path to directory with stable diffusion checkpoints--gfpgan-dir GFPGAN_DIRGFPGAN directory--gfpgan-model GFPGAN_MODELGFPGAN model file name--no-half             do not switch the model to 16-bit floats--no-half-vae         do not switch the VAE model to 16-bit floats--no-progressbar-hidingdo not hide progressbar in gradio UI (we hide it because it slows down ML if you have hardwareacceleration in browser)--max-batch-count MAX_BATCH_COUNTmaximum batch count value for the UI--embeddings-dir EMBEDDINGS_DIRembeddings directory for textual inversion (default: embeddings)--hypernetwork-dir HYPERNETWORK_DIRhypernetwork directory--localizations-dir LOCALIZATIONS_DIRlocalizations directory--allow-code          allow custom script execution from webui--medvram             enable stable diffusion model optimizations for sacrificing a little speed for low VRM usage--lowvram             enable stable diffusion model optimizations for sacrificing a lot of speed for very low VRMusage--lowram              load stable diffusion checkpoint weights to VRAM instead of RAM--always-batch-cond-unconddisables cond/uncond batching that is enabled to save memory with --medvram or --lowvram--unload-gfpgan       does not do anything.--precision {full,autocast}evaluate at this precision--share               use share=True for gradio and make the UI accessible through their site--ngrok NGROK         ngrok authtoken, alternative to gradio --share--ngrok-region NGROK_REGIONThe region in which ngrok should start.--enable-insecure-extension-accessenable extensions tab regardless of other options--codeformer-models-path CODEFORMER_MODELS_PATHPath to directory with codeformer model file(s).--gfpgan-models-path GFPGAN_MODELS_PATHPath to directory with GFPGAN model file(s).--esrgan-models-path ESRGAN_MODELS_PATHPath to directory with ESRGAN model file(s).--bsrgan-models-path BSRGAN_MODELS_PATHPath to directory with BSRGAN model file(s).--realesrgan-models-path REALESRGAN_MODELS_PATHPath to directory with RealESRGAN model file(s).--scunet-models-path SCUNET_MODELS_PATHPath to directory with ScuNET model file(s).--swinir-models-path SWINIR_MODELS_PATHPath to directory with SwinIR model file(s).--ldsr-models-path LDSR_MODELS_PATHPath to directory with LDSR model file(s).--clip-models-path CLIP_MODELS_PATHPath to directory with CLIP model file(s).--xformers            enable xformers for cross attention layers--force-enable-xformersenable xformers for cross attention layers regardless of whether the checking code thinks youcan run it; do not make bug reports if this fails to work--deepdanbooru        enable deepdanbooru interrogator--opt-split-attentionforce-enables Doggettx's cross-attention layer optimization. By default, it's on for torchcuda.--opt-split-attention-invokeaiforce-enables InvokeAI's cross-attention layer optimization. By default, it's on when cuda isunavailable.--opt-split-attention-v1enable older version of split attention optimization that does not consume all the VRAM it canfind--disable-opt-split-attentionforce-disables cross-attention layer optimization--use-cpu {all,sd,interrogate,gfpgan,swinir,esrgan,scunet,codeformer} [{all,sd,interrogate,gfpgan,swinir,esrgan,scunet,codeformer} ...]use CPU as torch device for specified modules--listen              launch gradio with 0.0.0.0 as server name, allowing to respond to network requests--port PORT           launch gradio with given server port, you need root/admin rights for ports < 1024, defaults to7860 if available--show-negative-promptdoes not do anything--ui-config-file UI_CONFIG_FILEfilename to use for ui configuration--hide-ui-dir-config  hide directory configuration from webui--freeze-settings     disable editing settings--ui-settings-file UI_SETTINGS_FILEfilename to use for ui settings--gradio-debug        launch gradio with --debug option--gradio-auth GRADIO_AUTHset gradio authentication like "username:password"; or comma-delimit multiple like"u1:p1,u2:p2,u3:p3"--gradio-img2img-tool {color-sketch,editor}gradio image uploader tool: can be either editor for ctopping, or color-sketch for drawing--opt-channelslast    change memory type for stable diffusion to channels last--styles-file STYLES_FILEfilename to use for styles--autolaunch          open the webui URL in the system's default browser upon launch--theme THEME         launches the UI with light or dark theme--use-textbox-seed    use textbox for seeds in UI (no up/down, but possible to input long seeds)--disable-console-progressbarsdo not output progressbars to console--enable-console-promptsprint prompts to console when generating with txt2img and img2img--vae-path VAE_PATH   Path to Variational Autoencoders model--disable-safe-unpickledisable checking pytorch models for malicious code--api                 use api=True to launch the api with the webui--nowebui             use api=True to launch the api instead of the webui--ui-debug-mode       Don't load model to quickly launch UI--device-id DEVICE_IDSelect the default CUDA device to use (export CUDA_VISIBLE_DEVICES=0,1,etc might be neededbefore)--administrator       Administrator rights--cors-allow-origins CORS_ALLOW_ORIGINSAllowed CORS origins--tls-keyfile TLS_KEYFILEPartially enables TLS, requires --tls-certfile to fully function--tls-certfile TLS_CERTFILEPartially enables TLS, requires --tls-keyfile to fully function--server-name SERVER_NAMESets hostname of server

尝试构造如下运行参数，--use-cpu all使所有模块均使用CPU计算，--lowram --always-batch-cond-uncond使用低内存配置选项，程序可以成功运行。

(novelai) E:\workspace\02_Python\novalai\stable-diffusion-webui>python launch.py --lowram --always-batch-cond-uncond --use-cpu all
Python 3.10.6 | packaged by conda-forge | (main, Oct 24 2022, 16:02:16) [MSC v.1916 64 bit (AMD64)]
Commit hash: b8f2dfed3c0085f1df359b9dc5b3841ddc2196f0
Installing requirements for Web UI
Launching Web UI with arguments: --lowram --always-batch-cond-uncond --use-cpu all
Warning: caught exception 'Expected a cuda device, but got: cpu', memory monitor disabled
LatentDiffusion: Running in eps-prediction mode
DiffusionWrapper has 859.52 M params.
making attention of type 'vanilla' with 512 in_channels
Working with z of shape (1, 4, 32, 32) = 4096 dimensions.
making attention of type 'vanilla' with 512 in_channels
Loading weights [7460a6fa] from E:\workspace\02_Python\novalai\stable-diffusion-webui\models\Stable-diffusion\sd-v1-4.ckpt
Global Step: 470000
Applying cross attention optimization (Doggettx).
Model loaded.
Loaded a total of 0 textual inversion embeddings.
Embeddings:
Running on local URL:  http://127.0.0.1:7860To create a public link, set `share=True` in `launch()`.

然而，开始作画时提示RuntimeError: "LayerNormKernelImpl" not implemented for 'Half'错误！如果安装网上的处理方法，将half函数在工程中替换为float函数，则会出现device不匹配问题。

Traceback (most recent call last):File "E:\workspace\02_Python\novalai\stable-diffusion-webui\modules\ui.py", line 185, in fres = list(func(*args, **kwargs))File "E:\workspace\02_Python\novalai\stable-diffusion-webui\webui.py", line 57, in fres = func(*args, **kwargs)File "E:\workspace\02_Python\novalai\stable-diffusion-webui\modules\txt2img.py", line 48, in txt2imgprocessed = process_images(p)File "E:\workspace\02_Python\novalai\stable-diffusion-webui\modules\processing.py", line 423, in process_imagesres = process_images_inner(p)File "E:\workspace\02_Python\novalai\stable-diffusion-webui\modules\processing.py", line 508, in process_images_inneruc = prompt_parser.get_learned_conditioning(shared.sd_model, len(prompts) * [p.negative_prompt], p.steps)File "E:\workspace\02_Python\novalai\stable-diffusion-webui\modules\prompt_parser.py", line 138, in get_learned_conditioningconds = model.get_learned_conditioning(texts)File "E:\workspace\02_Python\novalai\stable-diffusion-webui\repositories\stable-diffusion\ldm\models\diffusion\ddpm.py", line 558, in get_learned_conditioningc = self.cond_stage_model(c)File "D:\anaconda3\envs\novelai\lib\site-packages\torch\nn\modules\module.py", line 1190, in _call_implreturn forward_call(*input, **kwargs)File "E:\workspace\02_Python\novalai\stable-diffusion-webui\modules\sd_hijack.py", line 338, in forwardz1 = self.process_tokens(tokens, multipliers)File "E:\workspace\02_Python\novalai\stable-diffusion-webui\extensions\aesthetic-gradients\aesthetic_clip.py", line 202, in __call__z = self.process_tokens(remade_batch_tokens, multipliers)File "E:\workspace\02_Python\novalai\stable-diffusion-webui\modules\sd_hijack.py", line 353, in process_tokensoutputs = self.wrapped.transformer(input_ids=tokens, output_hidden_states=-opts.CLIP_stop_at_last_layers)File "D:\anaconda3\envs\novelai\lib\site-packages\torch\nn\modules\module.py", line 1190, in _call_implreturn forward_call(*input, **kwargs)File "D:\anaconda3\envs\novelai\lib\site-packages\transformers\models\clip\modeling_clip.py", line 722, in forwardreturn self.text_model(File "D:\anaconda3\envs\novelai\lib\site-packages\torch\nn\modules\module.py", line 1190, in _call_implreturn forward_call(*input, **kwargs)File "D:\anaconda3\envs\novelai\lib\site-packages\transformers\models\clip\modeling_clip.py", line 643, in forwardencoder_outputs = self.encoder(File "D:\anaconda3\envs\novelai\lib\site-packages\torch\nn\modules\module.py", line 1190, in _call_implreturn forward_call(*input, **kwargs)File "D:\anaconda3\envs\novelai\lib\site-packages\transformers\models\clip\modeling_clip.py", line 574, in forwardlayer_outputs = encoder_layer(File "D:\anaconda3\envs\novelai\lib\site-packages\torch\nn\modules\module.py", line 1190, in _call_implreturn forward_call(*input, **kwargs)File "D:\anaconda3\envs\novelai\lib\site-packages\transformers\models\clip\modeling_clip.py", line 316, in forwardhidden_states = self.layer_norm1(hidden_states)File "D:\anaconda3\envs\novelai\lib\site-packages\torch\nn\modules\module.py", line 1190, in _call_implreturn forward_call(*input, **kwargs)File "D:\anaconda3\envs\novelai\lib\site-packages\torch\nn\modules\normalization.py", line 190, in forwardreturn F.layer_norm(File "D:\anaconda3\envs\novelai\lib\site-packages\torch\nn\functional.py", line 2515, in layer_normreturn torch.layer_norm(input, normalized_shape, weight, bias, eps, torch.backends.cudnn.enabled)
RuntimeError: "LayerNormKernelImpl" not implemented for 'Half'

考虑到--use-cpu参数可以指定模块，则尝试使工程中的部分模块用CPU计算，其余在可用内存方位内用GPU计算，最终构造参数如下，项目可成功作画。

然而，此方式作画效率非常低，一般每张图片约5-6分钟。当参数设置较大时，会达到数小时。因此如果有条件可以升级计算机的显卡配置，或租赁云服务器效果会更好。


(novelai) E:\workspace\02_Python\novalai\stable-diffusion-webui>python launch.py --lowram --always-batch-cond-uncond  --precision full --no-half --opt-split-attention-v1 --use-cpu sd --autolaunch
Python 3.10.6 | packaged by conda-forge | (main, Oct 24 2022, 16:02:16) [MSC v.1916 64 bit (AMD64)]
Commit hash: b8f2dfed3c0085f1df359b9dc5b3841ddc2196f0
Installing requirements for Web UI
Launching Web UI with arguments: --lowram --always-batch-cond-uncond --precision full --no-half --opt-split-attention-v1 --use-cpu sd
Warning: caught exception 'Expected a cuda device, but got: cpu', memory monitor disabled
LatentDiffusion: Running in eps-prediction mode
DiffusionWrapper has 859.52 M params.
making attention of type 'vanilla' with 512 in_channels
Working with z of shape (1, 4, 32, 32) = 4096 dimensions.
making attention of type 'vanilla' with 512 in_channels
Loading weights [7460a6fa] from E:\workspace\02_Python\novalai\stable-diffusion-webui\models\Stable-diffusion\sd-v1-4.ckpt
Global Step: 470000
Applying v1 cross attention optimization.
Model loaded.
Loaded a total of 0 textual inversion embeddings.
Embeddings:
Running on local URL:  http://127.0.0.1:7860To create a public link, set `share=True` in `launch()`.
100%|███████████████████████████████████████████████████████████████████████████████████████████████████████████████████████| 20/20 [06:30<00:00, 19.50s/it]
Total progress: 100%|███████████████████████████████████████████████████████████████████████████████████████████████████████| 20/20 [06:10<00:00, 18.51s/it]

参考文献：

AI作画保姆级教程来了！逆天，太强了！

【作者：墨叶扶风http://blog.csdn.net/yefufeng】