1 - 配置环境

1.0 测试显卡

!nvidia-smi -L

1.1 下载安装依赖

setup miniconda

import sys !wget https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86_64.sh !chmod +x Miniconda3-latest-Linux-x86_64.sh !bash ./Miniconda3-latest-Linux-x86_64.sh -b -f -p /usr/local sys.path.append('/usr/local/lib/python3.7/site-packages/') !rm Miniconda3-latest-Linux-x86_64.sh

1.2 设置环境

Setup environment, Gfpgan and Real-ESRGAN. Takes about 5-6 minutes

#@markdown ### Set up conda environment - Takes a while !conda env update -n base -f /content/stable-diffusion/environment.yaml

1.3 设置CFPGan和ESRGAN

#@markdown ### Build upscalers support #@markdown GFPGAN Automatically correct distorted faces with a built-in GFPGAN option, fixes them in less than half a second #@markdown ESRGAN Boosts the resolution of images with a built-in RealESRGAN option #@markdown LDSR and GoBig enable amazing upscale options in the new Image Lab

add_CFP = True #@param {type:"boolean"} add_ESR = True #@param {type:"boolean"} add_LDSR = False #@param {type:"boolean"} #@markdown ⚠️ LDSR is 1.9GB and make take time to download

if add_CFP: %cd /content/stable-diffusion/src/gfpgan/ !pip install basicsr facexlib yapf lmdb opencv-python pyyaml tb-nightly --no-deps !python setup.py develop !pip install realesrgan !wget https://github.com/TencentARC/GFPGAN/releases/download/v1.3.0/GFPGANv1.3.pth -P experiments/pretrained_models if add_ESR: %cd /content/stable-diffusion/src/realesrgan/ !wget https://github.com/xinntao/Real-ESRGAN/releases/download/v0.1.0/RealESRGAN_x4plus.pth -P experiments/pretrained_models !wget https://github.com/xinntao/Real-ESRGAN/releases/download/v0.2.2.4/RealESRGAN_x4plus_anime_6B.pth -P experiments/pretrained_models if add_LDSR: %cd /content/stable-diffusion/src !git clone https://github.com/devilismyfriend/latent-diffusion %cd latent-diffusion %mkdir -p experiments/ %cd experiments/ %mkdir -p pretrained_models %cd pretrained_models #project.yaml download !wget -O project.yaml https://heibox.uni-heidelberg.de/f/31a76b13ea27482981b4/?dl=1 #model.ckpt model download !wget -O model.ckpt https://heibox.uni-heidelberg.de/f/578df07c8fc04ffbadf3/?dl=1

%cd /content/stable-diffusion/ !wget https://github.com/matomo-org/travis-scripts/blob/master/fonts/Arial.ttf?raw=true -O arial.ttf

#2.配置NovelAI

可以展开配置密码，否则自动随机生成一个

每次更改需要运行一席下面单元格代码

下载复制文件

最快也得4分钟，稍等

如果执行失败，重新执行第二步和第三步即可

!sudo apt-get install aria2 !sudo apt-get install file

!mkdir /content/time !git clone https://github.com/pnpnpn/timeout-decorator.git /content/time %cd /content/time !pwd !ls -l

下载NA

%cd /content/time

import timeout_decorator

outTime=180

@timeout_decorator.timeout(outTime) def downNovelAI(): !rm -rf /content/n2 !mkdir /content/n2 %cd /content/n2 !aria2c "magnet:?xt=urn:btih:4a4b483d4a5840b6e1fee6b0ca1582c979434e4d&dn=naifu&tr=udp%3a%2f%2ftracker.opentrackr.org%3a1337%2fannounce"

def checkFile(): !file /content/n2/naifu/models/animefull-final-pruned/model.ckpt>fileinfo !file /content/n2/naifu/models/animevae.pt>fileinfo2 f1=open("fileinfo") res1=f1.read() f1.close f2=open("fileinfo2") res2=f2.read() f2.close return "Zip" in res1 and "Zip" in res2

while 1: try: downNovelAI() except: if checkFile(): print("下载完成") outTime+=60 break else: print("下载未完成，自动重试")

下载WebUI

!mkdir /content/novelai %cd /content/novelai !git clone https://github.com/RyensX/stable-diffusion-webui-zh /content/novelai %cd /content/novelai !git checkout -b master

复制模型

!cp /content/n2/naifu/models/animefull-final-pruned/model.ckpt /content/novelai/models/Stable-diffusion/ !cp /content/n2/naifu/models/animevae.pt /content/novelai/models/Stable-diffusion/model.pt

!mkdir -p /content/novelai/train_images/raw/ !mkdir -p /content/novelai/train_images/des/

设置密码

若不设置则随机生成一个

每次更改需要运行一下下面单元格代码

import random keys="abcdefghigklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789"

#@markdown # 服务账号 user="Iris" #@param {type:"string"} if len(user)==0: user="".join([random.choice(keys) for i in range(random.randint(4,6))]) #@markdown # 服务密码 pwd="212121" #@param {type:"string"} if len(pwd)==0: pwd="".join([random.choice(keys) for i in range(random.randint(6,8))])

#3.运行NovelAI

可主动停止和多次运行下列单元格代码控制NovelAI运行状态

%cd /content/novelai print("#####################################################################################################################") print(f"* 账号密码分别是{user}和{pwd}") print("#######################################") print("!!!运行成功时会显示两个蓝色的地址，点击下方类似 https://xxxx.gradio.app/ 的网址即可外部访问，支持分享给别人用") print("!!!注意看上面文本提示") print("#####################################################################################################################") !python launch.py --share --gradio-auth {user}:{pwd} --deepdanbooru