def selfPlayOnly(args): g = nimGame(config) nnet = nn(g) coach_0 = Coach(g, nnet, args) for i in range(args.numIters): print("Self-play iteration: " + str(i)) nnet.load_checkpoint(args.load_folder_file[0], args.load_folder_file[1]) coach_0.selfPlay()
def main(): log.info('Loading %s...', Game.__name__) g = Game(6) 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()
nimConfig = {'maxPileSize': 10, 'maxNumPile': 3, 'initialState': None} if __name__ == "__main__": print("Serial Flag: " + str(serialFlag)) if serialFlag: if gameChoice == 0: g = Game(6) elif gameChoice == 1: g = TicTacToeGame() elif gameChoice == 2: g = nimGame(nimConfig) 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() else: def selfPlayOnly(args):
from Coach import Coach from othello.OthelloGame import OthelloGame as Game from othello.pytorch.NNet import NNetWrapper as nn from utils import * args = dotdict({ 'numIters': 1000, 'numEps': 100, 'tempThreshold': 15, 'updateThreshold': 0.6, 'maxlenOfQueue': 200000, 'numMCTSSims': 25, 'arenaCompare': 40, 'cpuct': 1, 'checkpoint': './temp/', 'load_model': False, 'load_folder_file': ('/dev/models/8x100x50','best.pth.tar'), }) if __name__=="__main__": g = Game(6) #game env nnet = nn(g) #network if args.load_model: nnet.load_checkpoint(args.load_folder_file[0], args.load_folder_file[1]) c = Coach(g, nnet, args)#set train para c.learn()#train