Deep Learning with Shogi AI on Mac and Google Colab

Book "Deep Learning Learned with Shogi AI" by Tadao Yamaoka Original source code https://github.com/TadaoYamaoka/python-dlshogi.git

Purpose

Run "Deep Learning with Shogi AI" on iMac and Google Colab. Build the basic environment on the iMac and use the GPU of Google Colab only when learning.

environment

iMac iMac 2020 27inch default model 3.1GHz 6-core 10th generation Intel Core i5 processor Up to 4.5GHz when using Turbo Boost 8GB 2,666MHz DDR4 memory 256GB SSD storage Radeon Pro 5300 (with 4GB GDDR6 memory) OS : Catalina Python: 3.8.2 and 2.7 pre-installed. This time I used 3.8.2. Google Colab Python 3.6.9

Summary of results

Prepare files on the iMac side and sync with Google Drive. image.png Learning execution on the Google Colab side. image.png Play on iMac. image.png Game video https://youtu.be/vPCsmi3_Zu8

table of contents

Use Google Colab </ b> [Getting Google Colab ready](https://qiita.com/kazunoriri/items/ef116e1cf88b4649a9b7#google-colab%E3%82%92%E4%BD%BF%E3%81%88%E3%82% 8B% E7% 8A% B6% E6% 85% 8B% E3% 81% AB% E3% 81% 99% E3% 82% 8B) Local side Google Colab side [Mount the drive](https://qiita.com/kazunoriri/items/ef116e1cf88b4649a9b7#%E3%83%89%E3%83%A9%E3%82%A4%E3%83%96%E3%82% 92% E3% 83% 9E% E3% 82% A6% E3% 83% B3% E3% 83% 88% E3% 81% 99% E3% 82% 8B) [Enable GPU](https://qiita.com/kazunoriri/items/ef116e1cf88b4649a9b7#gpu%E3%82%92%E6%9C%89%E5%8A%B9%E3%81%AB%E3% 81% 99% E3% 82% 8B) [Install chainer](https://qiita.com/kazunoriri/items/ef116e1cf88b4649a9b7#chainer%E3%82%92%E3%82%A4%E3%83%B3%E3%82%B9%E3%83% 88% E3% 83% BC% E3% 83% AB) [Install python-shogi and pydlshogi](https://qiita.com/kazunoriri/items/ef116e1cf88b4649a9b7#python-shogi%E3%81%A8pydlshogi%E3%82%92%E3%82%A4%E3%83% B3% E3% 82% B9% E3% 83% 88% E3% 83% BC% E3% 83% AB) Use Google Colab Local side Colab side Learning execution Experiment 　　CPU vs GPU

Chapter 1-6 </ b> Chapter 1-5 [Chapter 6 Deep Learning Framework](https://qiita.com/kazunoriri/items/32c45e46bb122ae1ef7c#%E7%AC%AC6%E7%AB%A0-%E3%83%87%E3%82%A3% E3% 83% BC% E3% 83% 97% E3% 83% A9% E3% 83% BC% E3% 83% 8B% E3% 83% B3% E3% 82% B0% E3% 83% 95% E3% 83% AC% E3% 83% BC% E3% 83% A0% E3% 83% AF% E3% 83% BC% E3% 82% AF) Chainer and numpy Warning

Chapter 7 1-4 </ b> [Chapter 7 Policy Network](https://qiita.com/kazunoriri/items/f91e39e36cf179750e5e#%E7%AC%AC7%E7%AB%A0-%E6%96%B9%E7%AD%96%E3% 83% 8D% E3% 83% 83% E3% 83% 88% E3% 83% AF% E3% 83% BC% E3% 82% AF) 　　7.1～7.4 　　　policy.py [Meaning of 194 filters](https://qiita.com/kazunoriri/items/f91e39e36cf179750e5e#%E3%83%95%E3%82%A3%E3%83%AB%E3%82%BF%E3% 83% BC% E6% 95% B0194% E5% 80% 8B% E3% 81% AE% E6% 84% 8F% E5% 91% B3) [Meaning of 1x1 filter](https://qiita.com/kazunoriri/items/f91e39e36cf179750e5e#1x1%E3%83%95%E3%82%A3%E3%83%AB%E3%82%BF%E3%83 % BC% E3% 81% AE% E6% 84% 8F% E5% 91% B3) 　　　common.py 　　　　bb_rotate180() 　　　features.py 　　　　make_input_features() 　　　　make_input_features_from_board() 　　　　make_output_label() 　　　　make_features()

Chapter 7 5-7 </ b> 　7.5～7.7 　　read_kifu.py 　　　read_kifu()

Chapter 7 8 </ b> 　7.8 [Automatic use of GPU and CPU](https://qiita.com/kazunoriri/items/e8541358bb030742cb19#gpu%E3%81%A8cpu%E3%81%AE%E8%87%AA%E5%8B%95%E4 % BD% BF% E3% 81% 84% E5% 88% 86% E3% 81% 91) [Pickle Protocol](https://qiita.com/kazunoriri/items/e8541358bb030742cb19#pickle%E3%81%AE%E3%83%97%E3%83%AD%E3%83%88%E3%82% B3% E3% 83% AB) 　　train_policy.py

Chapter 7 9 </ b> 　7.9 Learning execution

Chapter 8 1-4 </ b> [Implementation of USI engine](https://qiita.com/kazunoriri/items/e313efd0026d56e65442#usi%E3%82%A8%E3%83%B3%E3%82%B8%E3%83%B3%E3%81 % AE% E5% AE% 9F% E8% A3% 85) 　　policy_player.py 　　　y = self.model(x) 　　　logits = y.data[0] 　　　probabilities = F.softmax(y).data[0] [Automatic switching between GPU / CPU and PC](https://qiita.com/kazunoriri/items/e313efd0026d56e65442#gpucpu%E3%81%A8pc%E3%81%AE%E8%87%AA%E5%8B%95 % E5% 88% 87% E3% 82% 8A% E6% 9B% BF% E3% 81% 88) Strategy setting All codes Test [Test from command line](https://qiita.com/kazunoriri/items/e313efd0026d56e65442#%E3%82%B3%E3%83%9E%E3%83%B3%E3%83%89%E3%83% A9% E3% 82% A4% E3% 83% B3% E3% 81% 8B% E3% 82% 89% E3% 83% 86% E3% 82% B9% E3% 83% 88) [Test from Google Colab](https://qiita.com/kazunoriri/items/e313efd0026d56e65442#google-colab%E3%81%8B%E3%82%89%E3%83%86%E3%82%B9%E3 % 83% 88) Coordinates

Chapter 8 5-9 </ b> [Register with GUI software](https://qiita.com/kazunoriri/items/e0791fb8a975b58db275#gui%E3%82%BD%E3%83%95%E3%83%88%E3%81%AB%E7%99 % BB% E9% 8C% B2) [Problem that engine registration is not completed](https://qiita.com/kazunoriri/items/e0791fb8a975b58db275#%E3%82%A8%E3%83%B3%E3%82%B8%E3%83%B3%E7% 99% BB% E9% 8C% B2% E3% 81% 8C% E7% B5% 82% E3% 82% 8F% E3% 82% 89% E3% 81% AA% E3% 81% 84% E5% 95% 8F% E9% A1% 8C) Game

Chapter 9 </ b> [Learning Technique](https://qiita.com/kazunoriri/items/74c800bfa48cba34bb1c#%E5%AD%A6%E7%BF%92%E3%83%86%E3%82%AF%E3%83%8B% E3% 83% 83% E3% 82% AF) 　　SGD 　　Momentum SGD 　　SGD , Batch Normalization

Word </ b> 　shogi.BB_SQUARES 　shogi.COLORS 　shogi.CSA.Parser.parse_file(filepath) 　shogi.PIECE_TYPES_WITH_NONE 　shogi.MAX_PIECES_IN_HAND 　shogi.SQUARES Move class 　　from_square 　　to_square Board class 　　piece_bb 　　occupied 　　pieces_in_hand

Recommended Posts