if __name__ == "__main__":
parser = argparse.ArgumentParser()
parser.add_argument("--batch_size", type=int, default=256)
parser.add_argument("--memory_limit", type=float, default=0.2)
parser.add_argument("--learn_rate", type=float, default=1e-3)
parser.add_argument("--eplen_bonus", type=float,
default=0) # reward on the episode length
parser.add_argument("--discount_rate", type=float, default=0.99)
parser.add_argument("--replay_size", type=int, default=10000)
parser.add_argument("--exploration", type=float, default=0.2)
parser.add_argument("--save_freq", type=int, default=100)
parser.add_argument("--save_folder", type=str, default="model")
parser.add_argument("--reload", type=str, default=None)
args = parser.parse_args()
gmm = gameMgr.tetris(20, 10)
gmm.setScore(drop=1, eplen_bonus=args.eplen_bonus, terminal=-1.)
epoch = 0
write_epoch = 100
reward_history = collections.deque(maxlen=1000)
rewardDrop_history = collections.deque(maxlen=1000)
agt = agent.agent(gmm.getActionList(),
gmm.getScreenSize(),
gmm.getNextBlockSize(),
n_batch=args.batch_size,
learning_rate=args.learn_rate,
discountRate=args.discount_rate,
saveFreq=args.save_freq,
saveFolder=args.save_folder,
memoryLimit=args.memory_limit)
import numpy as np
import gameMgr
import dqn_agent_tetris
import collections
import matplotlib.pyplot as plt
import csv
def mean(x):
return float(sum(x)) / len(x)
showInterval = -1
if __name__ == "__main__":
gmm = gameMgr.tetris(12, 6)
epoch = 0
reward_history = collections.deque(maxlen=1000)
loss_history = collections.deque(maxlen=1000)
n_batch = 256
agt = dqn_agent_tetris.DQNAgent(gmm.getActionList(), "tetris",
gmm.getScreenSize()[0],
gmm.getScreenSize()[1])
#agt.load("model/model.ckpt")
fig = plt.figure(figsize=(gmm.getScreenSize()[0], gmm.getScreenSize()[1]))
fig.canvas.set_window_title("TeTris")
logFile = file("log", "w")
logCSV = csv.writer(logFile)
logCSV.writerow([
"epoch", "last_loss", "loss_mean", "last_reward", "mean_reward",
"max_reward"