def main():
log.info('Loading %s...', TicTacToeGame.__name__)
g = TicTacToeGame()
log.info('Loading %s...', nn.__name__)
nnet = nn(g)
if args.load_model:
log.info(f'Loading checkpoint "{args.load_folder_file}" ...')
nnet.load_checkpoint(args.load_folder_file[0],
args.load_folder_file[1])
else:
log.warning('Not loading a checkpoint!')
log.info('Loading the Coach...')
c = Coach(g, nnet, args)
if args.load_model:
log.info("Loading 'trainExamples' from file...")
c.loadTrainExamples()
log.info('Starting the learning process 🎉')
c.learn()
# all players
#rp = RandomPlayer(g).play
if gameChoice == 0:
gp = GreedyOthelloPlayer(g).play
hp = HumanOthelloPlayer(g).play
elif gameChoice == 1:
hp = HumanTicTacToePlayer(g).play
elif gameChoice == 2:
hp = HumanNimPlayer(g).play
# nnet players
if gameChoice == 0:
n1 = NNet(g)
else:
n1 = nn(g)
if gameChoice == 0:
if mini_othello:
n1.load_checkpoint('./pretrained_models/othello/pytorch/',
'6x100x25_best.pth.tar')
else:
n1.load_checkpoint('./pretrained_models/othello/pytorch/',
'8x8_100checkpoints_best.pth.tar')
elif gameChoice == 1:
n1.load_checkpoint(
'/Users/mettinger/github/alpha-zero-general/pretrained_models/tictactoe/keras',
'best-25eps-25sim-10epch.pth.tar')
elif gameChoice == 2:
n1.load_checkpoint('''/Users/mettinger/Google Drive/models''',
'best.pth.tar')
25,
'tempThreshold':
15,
'updateThreshold':
0.55,
'maxlenOfQueue':
200000,
'numMCTSSims':
25,
'arenaCompare':
50,
'cpuct':
1,
'checkpoint':
'./temp/tictactoe/',
'load_model':
False,
'load_folder_file': ('./models/tictactoe/10x25x25', 'best.pth.tar'),
})
if __name__ == "__main__":
g = TicTacToeGame()
nnet = nn(g)
if args.load_model:
nnet.load_checkpoint(args.load_folder_file[0],
args.load_folder_file[1])
c = Coach(g, nnet, args)
c.learn()
from utils import *
"""
use this script to play any two agents against each other, or play manually with
any agent.
"""
human_vs_cpu = True
g = TicTacToeGame()
# all players
rp = RandomPlayer(g).play
hp = HumanTicTacToePlayer(g).play
# nnet players
n1 = nn(g)
n1.load_checkpoint('./pretrained_models/tictactoe/keras/',
'best-25eps-25sim-10epch.pth.tar')
args1 = dotdict({'numMCTSSims': 50, 'cpuct': 1.0})
mcts1 = MCTS(g, n1, args1)
n1p = lambda x: np.argmax(mcts1.getActionProb(x, temp=0))
if human_vs_cpu:
player2 = hp
else:
n2 = nn(g)
n2.load_checkpoint('./pretrained_models/othello/pytorch/',
'8x8_100checkpoints_best.pth.tar')
args2 = dotdict({'numMCTSSims': 50, 'cpuct': 1.0})
mcts2 = MCTS(g, n2, args2)
n2p = lambda x: np.argmax(mcts2.getActionProb(x, temp=0))