cglemon / pydlgo Goto Github PK

View Code? Open in Web Editor NEW

53.0 2.0 8.0 12.41 MB

基於深度學習的 GTP 圍棋（围棋）引擎，KGS 指引文件以及演算法教學。

License: MIT License

Python 100.00%

deep-learning mcts weiqi goban baduk alphago game-of-go

pydlgo's Issues

Complete GTP with ids

pyDLGO/gtp.py

Lines 149 to 154 in d330691

    
           def process(self, cmd): 
        
               # TODO: Support analyze and genmove_analyze commands. 
        
               main = cmd[0] 
        
               if main == "name": 
        
                   self.success_print("dlgo")

It is not a complete GTP like this:

1 showboard
=1 
    A B C ...
19 
18
...
1

2 play b k10
=2

0 final_score
=0 7.5

5 quit
=5

Here. numbers before commands are ids.

Why are the features designed this way?

pyDLGO/board.py

Line 440 in c237ab6

def get_features(self):

Is the 18-layer feature design following some paper or trying to fit another library? The most direct thought would perhaps be just to output the past 16 boards, and flip the color depending on who is playing (so that the states are symmetric with respect to black/white). Wonder if there is any reason why the current one is used instead.

Can't run it on Sabaki in windows

The situation is below

你好，我设法让他在windows上和sabaki连接，但是他自己断开了

state size does not change dynamically with custom board size

pyDLGO/board.py

Line 89 in c237ab6

self.state = np.full(NUM_VERTICES, INVLD) # positions state

This line seems to use the global constant for 19x19 board. But if the board is initialized for smaller sizes, there seems to be a redundancy in storing the states (a bunch of unnecessary "3"'s). Should it be changed?

cglemon / pydlgo Goto Github PK

pydlgo's Issues

Complete GTP with ids

Why are the features designed this way?

Can't run it on Sabaki in windows

你好，我设法让他在windows上和sabaki连接，但是他自己断开了

state size does not change dynamically with custom board size

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent

	def process(self, cmd):
	# TODO: Support analyze and genmove_analyze commands.
	main = cmd[0]

	if main == "name":
	self.success_print("dlgo")