'numIterations': 9,
'numEpisodes': 25,
'tempThreshold': 15,
'updateThreshold': 0.6,
'maxlenOfQueue': 200000,
'numMCTSSims': 25,
'arenaCompare': 40,
'cpuct': 1,
}
loadSaveArgs = {'checkpoint': './checkpoints/tictactoe/keras/' + args_to_filename(model_params),
'load_model': False,
'load_folder_file': ('models/tictactoe/keras/' + args_to_filename(model_params), 'best.pth.tar'),
'numItersForTrainExamplesHistory': 10}
args = dotdict({**model_params,**loadSaveArgs})
if __name__ == "__main__":
game = TicTacToeGame(3)
nnet = keras_tictactoe_neuralnet(game)
c = Coach(game, nnet, args)
pathlib.Path(args.checkpoint).mkdir(parents=True, exist_ok=True)
if args.load_model:
pathlib.Path(args.load_folder_file[0]).mkdir(parents=True, exist_ok=True)
nnet.load_checkpoint(args.load_folder_file[0], args.load_folder_file[1])
print("Load trainExamples from file")
c.loadTrainExamples()
from tictactoe.TicTacToeGame import TicTacToeGame
from tictactoe.TicTacToePlayers import HumanPlayer, RandomPlayer, KerasNeuralNetPlayer
from alpha_zero.Arena import Arena
from alpha_zero.utils import dotdict
"""
use this script to play any two agents against each other, or play manually with
any agent.
"""
game = TicTacToeGame(3)
# all players
random_player1 = RandomPlayer(game, "Random1")
random_player2 = RandomPlayer(game, "Random1")
human_player = HumanPlayer(game, "Human")
args1 = dotdict({'numMCTSSims': 25, 'cpuct': 1.0})
# nnet players
checkpoints_dir = "./checkpoints/tictactoe/keras/3numIterations-25numEpisodes-15tempThreshold-0.6updateThreshold-200000maxlenOfQueue-25numMCTSSims-40arenaCompare-1cpuct-"
neural_net_player1 = KerasNeuralNetPlayer(game, args1, "Neural_1")
neural_net_player1.load_brain(checkpoints_dir, 'best.pth.tar')
if __name__ == '__main__':
# arena = Arena(neural_net_player1, neural_net_player2, game, display=display)
arena = Arena(human_player, neural_net_player1, game, display=display)
results = arena.play_games(2, verbose=True)
results_format = 'Results {0}'.format(results)
print(results_format)
print("")
from .TicTacToeNNet import TicTacToeNNet as onnet
"""
NeuralNet wrapper class for the TicTacToeNNet.
Author: Evgeny Tyurin, github.com/evg-tyurin
Date: Jan 5, 2018.
Based on (copy-pasted from) the NNet by SourKream and Surag Nair.
"""
args = dotdict({
'learningRate': 0.001,
'dropout': 0.3,
'epochs': 10,
'batch_size': 64,
'cuda': True,
'num_channels': 512,
})
class NNetWrapper(NeuralNet):
def __init__(self, game):
self.nnet = onnet(game, args)
self.board_x, self.board_y = game.getBoardSize()
self.action_size = game.getActionSize()
def train(self, examples:List[Tuple[Board,List[float],List[int]]]):
"""
examples: list of examples, each example is of form (board, pi, v)
"""
input_boards, target_pis, target_vs = list(zip(*examples))
import math
import sys
from alpha_zero.NeuralNet import NeuralNet
from alpha_zero.utils import dotdict
sys.path.append('..')
import argparse
from .OthelloNNet import OthelloNNet as onnet
args = dotdict({
'lr': 0.001,
'dropout': 0.3,
'epochs': 10,
'batch_size': 64,
'cuda': False,
'num_channels': 512,
})
class NNetWrapper(NeuralNet):
def __init__(self, game):
self.nnet = onnet(game, args)
self.board_x, self.board_y = game.getBoardSize()
self.action_size = game.getActionSize()
def train(self, examples):
"""
examples: list of examples, each example is of form (board, pi, v)
"""
input_boards, target_pis, target_vs = list(zip(*examples))