[Guide?] How to run v4.5 on google colab

I hate anything which come along with a fat framework, especially when those extra ~600mb garbage is just for an unnecessary purpose. So I made some tweaks to make v4.5 run on google colab. If you also want to avoid this totally unoptimized thing cooking your pc, you might be interested.

Unfortunately, I can’t provide you guys full code cuz the author of this is clearly willing to keep something in “secret” by obfuscating his codes. ( scriptkiddie-tier dumb obfuscation, tho )

Here are the steps:

• 1. Fork the v4.5 gdrive release to your own gdrive.
• 1. Make a colab notebook, mount gdrive, insert 1st code block like this:

import os
!unzip '/content/drive/My Drive/<your path>/Style2PaintsV45配布/style2paints45beta1214B.zip' > 00.txt
os.rename('/content/style2paints45beta1214B/assets/nets','/content/nets')
os.rename('/content/style2paints45beta1214B/assets/game','/content/game')
os.rename('/content/style2paints45beta1214B/assets/gpu.conf','/content/gpu.conf')

run it.

• 1. Modify /content/game/src/project.js, remove all ‘window.server_url+’ string in it, or just replace it with the attachment. project.txt
• 1. Get the decompiled v4.5 code by some mean. As you see a decompiled script starts with ‘def main_exec():’, and full of sh*t like OO0O000OO000OOO00(op), that’s it. Paste the decompiled code to 2nd code block.
• 1. Erase all the import xxx, from xxx import xxxx in the 2nd code, replace all ‘8233’ with ‘viewppp’, paste

%tensorflow_version 1.x
!pip install bottle
import sys
import socket, os
import tensorflow as tf
import numpy as np, datetime, pickle, base64, json, gzip, re, os, shutil
import cv2
from tensorflow.python.client import device_lib
from threading import Thread
from google.colab.output import eval_js
from tqdm import tqdm
from cv2.ximgproc import l0Smooth, createGuidedFilter, guidedFilter
from bottle import route, run, static_file, request, BaseRequest

BaseRequest.MEMFILE_MAX = 10000000


def replaceservname(servaddr):
  data=[]
  with open('/content/game/src/project.js','rt') as fjs:
    data=fjs.read()
    data=data.replace('http://127.0.0.1:8233',servaddr)
  with open('/content/game/src/project.js','wt') as fjs:
    fjs.write(data)
   
def main_exec(viewppp):
    ruul = eval_js( 'google.colab.kernel.proxyPort(' + str(viewppp) +')')[:-1]
    print(ruul+'/index.html')
    replaceservname(ruul)

to 2nd code’s head (overwrite first line of def main_exec()😃

• 1. Comment out these lines:

os.system('start ./dotnet/Style2Paints.exe')

Thread(target=O0000O0OO0O0000O0, args=('Now Loading ...', )).start()

        if os.path.exists('_pywrap_tensorflow_internal_gpu.pyd'):
            if os.path.exists('_pywrap_tensorflow_internal.pyd'):
                shutil.move('_pywrap_tensorflow_internal.pyd', '_pywrap_tensorflow_internal_cpu.pyd')
            shutil.move('_pywrap_tensorflow_internal_gpu.pyd', '_pywrap_tensorflow_internal.pyd')
        else:
            print('Style2Paints Customed Tensorflow Shifting to CPU.')
            if os.path.exists('_pywrap_tensorflow_internal_cpu.pyd'):
                if os.path.exists('_pywrap_tensorflow_internal.pyd'):
                    shutil.move('_pywrap_tensorflow_internal.pyd', '_pywrap_tensorflow_internal_gpu.pyd')
                shutil.move('_pywrap_tensorflow_internal_cpu.pyd', '_pywrap_tensorflow_internal.pyd')
                

• 1. Run 2nd code block, make 3rd code block with

main_exec(8233)

1. Run 3rd code block, after message like

Listening on http://0.0.0.0:8233/
Hit Ctrl-C to quit.

shows, click the first url in output which looks like: https://xxxxxxxxxxxxxxxxxxxxx-8233-colab.googleusercontent.com/index.html than that’s it.

[SylveonBottle]

How does this compare to just running the orriginal on your own machine, in terms of performance / time required to color?

[basemdabbour]

Hello, can someone share google Colab for Style2Paints V4.5?.