def main():
log.info('Loading %s...', Game.__name__)
g = Game()
log.info('Loading %s...', nn.__name__)
nnet = nn(g)
if args.load_model:
log.info('Loading checkpoint "%s/%s"...', 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()
import numpy as np
from utils import *
"""
use this script to play any two agents against each other, or play manually with
any agent.
"""
g = Connect4Game()
# all players
rp = RandomPlayer(g).play
gp = OneStepLookaheadConnect4Player(g).play
hp = HumanConnect4Player(g).play
# nnet players
n1 = nn(g)
#n1.load_checkpoint('./checkpoint-connect4','best.pth.tar')
n1.load_checkpoint('./temp', 'best.pth.tar')
args1 = dotdict({'numMCTSSims': 50, 'cpuct': 1.0})
mcts1 = MCTS(g, n1, args1)
def n1p(x):
probs = mcts1.getActionProb(x, temp=0)
return np.argmax(probs)
#n2 = NNet(g)
#n2.load_checkpoint('/dev/8x50x25/','best.pth.tar')
#args2 = dotdict({'numMCTSSims': 25, 'cpuct':1.0})
#mcts2 = MCTS(g, n2, args2)
args = dotdict({
'numIters': 1000,
'numEps': 100,
'tempThreshold': 15,
'updateThreshold': 0.6,
'maxlenOfQueue': 200000,
'numMCTSSims': 25,
'arenaCompare': 40,
'cpuct': 1,
'checkpoint': './temp/',
'load_model': True,
'load_folder_file': ('temp','best.pth.tar'),
'numItersForTrainExamplesHistory': 5,
})
if __name__=="__main__":
g = Connect4Game()
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)
if args.load_model:
print("Load trainExamples from file")
c.loadTrainExamples()
c.learn()
