tkhkaeio/AlphaZero

I researched and explained AlphaGo/AlphaGo Zero papers, which had beaten the world the game of Go champion in 2016, 2017. Especially, I applied Alpha Zero algorithm to Othello to catch the whole idea.

Alpha"Othello" Zero

Alpha Zero explanation: here

Alpha Go, DeepMind: here

please run this scripts in Google Colaboratory

1. Select GPU.

Edit -> Notebook settings -> GPU

2. Mount your Google Drive

#to access your google drive folder
from google.colab import auth, drive, files, output
drive.mount('/content/drive')

3. Import the folder from github & setup

#to clone codes
!git clone https://github.com/takehiko-ohkawa/AlphaZero.git
import os
os.chdir('./AlphaZero')
!pwd
!chmod ugo+x ./setup_colab.sh #give a permission
! ./setup_colab.sh #install torch 

4. Train

!chmod ugo+x ./train_colab.sh
!./train_colab.sh

• make a folder or copy a folder

#make
!mkdir <folder>
#copy
!cp -r <folderA> <folderB> #copy A->B

• Overwrite your code

%%writefile oo.py
<contents of oo.py file>

• Check GPU status

!df -h #desk status
!free -h #memory status
!ps aux #process status
!nvidia-smi #GPU status
!cat /proc/uptime |awk '{print $1/86400"days"}' #remaining time - upto 0.5days

• Download files

files.download(<file>)

Reference: here

Result

Alpha Zero explanation & result: here