def train_player():
# Treniraj # Zug
stdout.disable()
games = 10000 #10000, one game means up to 1001 Points!!!
last = 1000
wins = list()
#Training:
print("inside train player")
now = datetime.datetime.now()
for i in range(games):
game = Game(pairA, pairB)
pointsA, pointsB = game.play()
wins.append("A" if pointsA > pointsB else "B")
if i > 0 and i % last == 0:
winsA = wins[-last:].count("A")
winningPercentage = winsA / last * 100
if winningPercentage >= 90:#90 before
break
game.saveNetworks(str(i))
stdout.enable()
print("[RL] {} - win percentage (in last 100 games): {}% ({} / {}) at game:".format(pairA, last, winningPercentage, winsA, last), i, (datetime.datetime.now() - now))
stdout.disable()
# Igraj # spielen
#stdout.enable()
game.saveNetworks()
print(datetime.datetime.now() - now)
playerA1.eval()
playerA2.eval()
def notifyTrick(self, value):
# der schlimmste wert ist -17 (g10, g5, r1, r2)
# ausser wenn noch mal2 hinzukommt?! dann ist es wohl 21?!
value += 21
normalizedReward = value / 26 # 21 zuvor sonst 26
if abs(normalizedReward) > 1:
stdout.enable()
print(normalizedReward)
print(eeee)
#self.playingPolicy.feedback(normalizedReward)
self.witchesPolicy.feedback(normalizedReward)
def play(self):
number_of_won = [0, 0, 0, 0]
gameover_limit = -70
history = []
ai_player_index = 0
nuGames = 50
out_path = "models/rl_policy/"
stdout.disable()
stdout.write_file(out_path + "output.txt") # contains all logging!
start_time = datetime.datetime.now()
try:
for j in range(1, 100000000):
i = 0
while i < nuGames:
action = self.selectAction()
current_player = self.my_game.active_player
card = self.my_game.players[current_player].hand[action]
print("[{}] {} {}\t{}\tCard {}\tHand Index {}".format(
self.my_game.current_round, current_player,
self.my_game.names_player[current_player],
self.my_game.ai_player[current_player], card, action))
rewards, round_finished = self.my_game.step_idx(
action, auto_shift=False)
if round_finished:
# player idx of Reinforce
self.notifyTrick(rewards[ai_player_index])
print("Update rewards: ", rewards, "\n")
if len(self.my_game.players[current_player].hand
) == 0: # one game finished
print("update policy at end of one game!")
#self.playingPolicy.updatePolicy()
self.witchesPolicy.updatePolicy()
print(self.my_game.total_rewards)
if min(self.my_game.total_rewards
) <= gameover_limit:
winner_idx = np.where(
(self.my_game.total_rewards == max(
self.my_game.total_rewards)))
number_of_won[winner_idx[0][0]] += 1
self.my_game.total_rewards = [0, 0, 0, 0]
i += 1
if i == nuGames:
stdout.enable()
print("Win Stats:", number_of_won,
"at game", j * nuGames, "for:",
self.witchesPolicy.lr,
datetime.datetime.now() - start_time,
"\n")
active_player, state, options = self.my_game.getState(
)
state = torch.tensor(
state).float().resize_(180)
path = out_path + str(
self.witchesPolicy.lr) + "_" + str(
j * nuGames) + "__" + str(
number_of_won[ai_player_index])
self.exportONNX(self.witchesPolicy.network,
state, path)
history.append(number_of_won)
number_of_won = [0, 0, 0, 0]
stdout.disable()
self.my_game.reset_game()
self.plotHistory(history, ai_player_index, out_path)
except Exception as e:
stdout.enable()
print("ERROR!!!!", e)
print(traceback.format_exc())
print(number_of_won)
self.plotHistory(history, ai_player_index, out_path)
games = 10000
last = 100
wins = list()
for i in range(games):
game = Game(pairA, pairB)
pointsA, pointsB = game.play()
wins.append("A" if pointsA > pointsB else "B")
if i > 0 and i % last == 0:
winsA = wins[-last:].count("A")
winningPercentage = winsA / last * 100
# if winningPercentage >= 90:
# break
stdout.enable()
print("[RL] {} - postotak pobjeda (u zadnjih {} igara): {}% ({} / {})".format(pairA, last, winningPercentage, winsA, last))
stdout.disable()
# Igraj
stdout.enable()
playerA1.eval()
playerA2.eval()
playerMe = PlayerKeyboard("Ja")
playerFriend = PlayerRandom("On")
game = Game(
Pair(playerMe, playerFriend),
pairA
import stdout
# reinforcement learning pair
# TODO save & load policy networks
# TODO share policy networks across all RL players
playerA1 = PlayerRL("Borna")
playerA2 = PlayerRL("Mislav")
pairA = Pair(playerA1, playerA2)
# random pair
playerB1 = PlayerRandom("Luka")
playerB2 = PlayerRandom("Lovro")
pairB = Pair(playerB1, playerB2)
stdout.disable()
games = 10000
last=50
wins=list()
for i in range(games):
game = Game(pairA, pairB)
pointsA, pointsB = game.play()
wins.append("A" if pointsA>pointsB else "B")
if i>0 and i%last==0:
winningPercentage=wins[-last:].count("A")/last*100
stdout.enable()
print("[RL] {} - postotak pobjeda (u zadnjih {} igara): {}%".format(pairA, last, winningPercentage))
stdout.disable()