def test_priornet_tictactoe_self_play(self):
ttt = TicTacToe()
nn = NeuralNetwork(ttt, PriorNet)
t = Trainer(ttt, nn, num_simulations=2, num_games=1, num_updates=0, buffer_size_limit=None, cpuct=1, num_threads=4)
data = t.self_play(temperature=0)
np.testing.assert_equal(data[:,-1], np.array([1, -1, 1, -1, 1, -1, 1]))
s = ttt.get_initial_state()
np.testing.assert_equal(data[0,0], s)
np.testing.assert_equal(data[0,1], np.array([0, 1, 0, 0, 0, 0, 0, 0, 0]))
s = ttt.take_action(s, np.array([[0,1,0],[0,0,0],[0,0,0]])) # Top-middle X
np.testing.assert_equal(data[1,0], s)
np.testing.assert_equal(data[1,1], np.array([1, 0, 0, 0, 0, 0, 0, 0]))
s = ttt.take_action(s, np.array([[1,0,0],[0,0,0],[0,0,0]])) # Top-left O
np.testing.assert_equal(data[2,0], s)
np.testing.assert_equal(data[2,1], np.array([1, 0, 0, 0, 0, 0, 0]))
s = ttt.take_action(s, np.array([[0,0,1],[0,0,0],[0,0,0]])) # Top-right X
np.testing.assert_equal(data[3,0], s)
np.testing.assert_equal(data[3,1], np.array([1, 0, 0, 0, 0, 0]))
s = ttt.take_action(s, np.array([[0,0,0],[1,0,0],[0,0,0]])) # Mid-left O
np.testing.assert_equal(data[4,0], s)
np.testing.assert_equal(data[4,1], np.array([1, 0, 0, 0, 0]))
s = ttt.take_action(s, np.array([[0,0,0],[0,1,0],[0,0,0]])) # Mid-mid X
np.testing.assert_equal(data[5,0], s)
np.testing.assert_equal(data[5,1], np.array([1, 0, 0, 0]))
s = ttt.take_action(s, np.array([[0,0,0],[0,0,1],[0,0,0]])) # Mid-right O
np.testing.assert_equal(data[6,0], s)
np.testing.assert_equal(data[6,1], np.array([1, 0, 0]))
def test_policy_iteration(self):
ttt = TicTacToe()
nn = NeuralNetwork(ttt, PriorNet)
t = Trainer(ttt, nn, num_simulations=2, num_games=100, num_updates=0, buffer_size_limit=None, cpuct=1, num_threads=4)
t.policy_iteration()
states = t.training_data[:,0]
inits = 0
for s in states:
if (s.astype(np.float32) == ttt.get_initial_state()).all():
inits += 1
self.assertEqual(inits, 100)
def test_8(self):
game = TicTacToe(['X', 1, 'X', 3, 'O', 5, 6, 'O', 'X'])
self.assertEqual({
3: {
'X': 0,
'O': 0
},
2: {
'X': 2,
'O': 1
}
}, game.count_lines(game.state))
def test_5(self):
game = TicTacToe(['X', 'X', 'X', 'O', 'O', 'X', 'X', 'O', 'O'])
self.assertEqual({
3: {
'X': 1,
'O': 0
},
2: {
'X': 0,
'O': 0
}
}, game.count_lines(game.state))
def test_1(self):
game = TicTacToe(['O', 1, 2, 3, 4, 5, 6, 7, 8])
self.assertEqual({
3: {
'X': 0,
'O': 0
},
2: {
'X': 0,
'O': 0
}
}, game.count_lines(game.state))
def get_game_from_name(game_name):
if game_name.endswith("tic_tac_toe"):
return TicTacToe()
elif game_name.endswith("connect_four"):
return ConnectFour()
elif game_name.endswith("reversi"):
return Reversi()
def __init__(self):
self.db = MongoClient('mongodb://localhost:27017/').ai_data
self.process_strings = []
self.emails = []
count = 2
for person in self.db.posts.find({'game':'tic tac toe'}):
if not count:
break
if person['lang'] == 'java':
with open('bridges/java/src/com/ai/api/MyAI.java', 'w') as f:
f.write(person['code'])
os.chdir('bridges/java')
p = Popen(shlex.split('mvn install'))
p.communicate()
os.chdir('../..')
self.process_strings.append('java -jar bridges/java/target/JavaAPI-0.0.1-SNAPSHOT-jar-with-dependencies.jar')
self.emails.append(person['email'])
elif person['lang'] == 'javascript':
with open('bridges/javascript/myAI.js', 'w') as f:
f.write(person['code'])
self.process_strings.append('node bridges/javascript/bridge.js')
self.emails.append(person['email'])
else:
pass
count -= 1
while len(self.process_strings) < 2:
self.process_strings.append(('java -jar bridges/java/target/JavaAPI-0.0.1-SNAPSHOT-jar-with-dependencies.jar', 'node bridges/javascript/bridge.js')[random.randint(0,1)])
self.game = TicTacToe()
self.processes = []
def move(self, game: TicTacToe, possible_steps=None):
available = game.get_possible_next_steps()
while True:
choice = int(input('Your move? [0-8]'))
if choice in available:
return choice
print('Not available')
def test_resets_work(self):
game = TicTacToe()
p1, p2 = UninformedMCTSPlayer(game, 1), UninformedMCTSPlayer(game, 1)
for _ in range(100):
scores, outcomes = play_match(game, [p1, p2])
self.assertEqual(scores, {p1: 1.0, p2: 0.0})
self.assertEqual(outcomes, {p1: "Win", p2: "Lose"})
for _ in range(100):
scores, outcomes = play_match(game, [p1, p2], permute=True)
self.assertEqual(scores, {p1: 0.5, p2: 0.5})
self.assertEqual(outcomes, {p1: "Tie", p2: "Tie"})
def __init__(self):
self.db = MongoClient('mongodb://localhost:27017/').ai_data
self.process_strings = []
self.emails = []
count = 2
for person in self.db.posts.find({'game': 'tic tac toe'}):
if not count:
break
if person['lang'] == 'java':
with open('bridges/java/src/com/ai/api/MyAI.java', 'w') as f:
f.write(person['code'])
os.chdir('bridges/java')
p = Popen(shlex.split('mvn install'))
p.communicate()
os.chdir('../..')
self.process_strings.append(
'java -jar bridges/java/target/JavaAPI-0.0.1-SNAPSHOT-jar-with-dependencies.jar'
)
self.emails.append(person['email'])
elif person['lang'] == 'javascript':
with open('bridges/javascript/myAI.js', 'w') as f:
f.write(person['code'])
self.process_strings.append(
'node bridges/javascript/bridge.js')
self.emails.append(person['email'])
else:
pass
count -= 1
while len(self.process_strings) < 2:
self.process_strings.append((
'java -jar bridges/java/target/JavaAPI-0.0.1-SNAPSHOT-jar-with-dependencies.jar',
'node bridges/javascript/bridge.js')[random.randint(0, 1)])
self.game = TicTacToe()
self.processes = []
def main(board_obj: TicTacToe):
"""
The main function to play a game of Tic Tac Toe.
:param board_obj: The Board object to run the game on.
:return:
"""
print("Welcome to Tic Tac Toe.")
print("Possible moves are:")
board_obj.print_move_grid()
while board_obj.active:
# Print a newline.
print()
# Get the move the player made.
move = input(
f"{board_obj.get_player_icon(board_obj.current_player)} > ")
# Check the validity of the move.
while not board_obj.check_move_validity(move):
print("Invalid move.")
move = input(
f"{board_obj.get_player_icon(board_obj.current_player)} > ")
# Convert the move to an int.
move = int(move)
# Make the move.
board_obj.make_move(move)
# Notify the player of the new board.
board_obj.print_grid()
# Print the winner.
print("\nThe game ended!")
print(f"Player {board_obj.get_player_icon(board_obj.winner)} won!")
class Runner:
def __init__(self):
self.db = MongoClient('mongodb://*****:*****@osai.com", "AI Results", "Your AI code has finished running:<br>", 'http://127.0.0.1:5000/static/replay.html?id='+str(ident))
for email in self.emails:
message.add_to(email)
s.smtp.send(message)
class Runner:
def __init__(self):
self.db = MongoClient('mongodb://*****:*****@osai.com", "AI Results",
"Your AI code has finished running:<br>",
'http://127.0.0.1:5000/static/replay.html?id=' + str(ident))
for email in self.emails:
message.add_to(email)
s.smtp.send(message)
def test_permutation(self):
game = TicTacToe()
p1, p2 = UninformedMCTSPlayer(game, 1), UninformedMCTSPlayer(game, 1)
scores, outcomes = play_match(game, [p1, p2], permute=True)
self.assertEqual(scores, {p1: 0.5, p2: 0.5})
self.assertEqual(outcomes, {p1: "Tie", p2: "Tie"})
# Print the winner.
print("\nThe game ended!")
print(f"Player {board_obj.get_player_icon(board_obj.winner)} won!")
if __name__ == '__main__':
try:
# Ask the user for parameters.
grid_size = int(
validate_input("Size of the grid: ", lambda x: x.isnumeric()))
amount_of_players = int(
validate_input("Amount of players: ", lambda x: x.isnumeric()))
player_symbols = validate_input(
"Enter the symbols for the players, separated by space.\n",
lambda x: len(set(x.split())) >= amount_of_players).split(
)[:amount_of_players]
player_symbols = [x[:1] for x in player_symbols
] # Only use the first character of each symbol.
# Create the Board class
board = TicTacToe(grid_size=grid_size,
players=amount_of_players,
player_symbols=player_symbols)
# Run the main function.
main(board)
except KeyboardInterrupt:
print("\n\nThanks for playing.")
exit(0)
window = 1
error = np.convolve(error, np.ones(window), mode="valid") / window
min_len = len(error) if min_len is None else min(min_len, len(error))
plt.plot(error, label=model_class.__name__)
plt.title("Training loss for {}".format(game.__class__.__name__))
ax.set_xlim(left=0, right=min_len)
ax.set_ylabel("Error")
ax.set_xlabel("Iteration")
plt.legend()
plt.show()
if __name__ == "__main__":
checkpoint = 13
game = TicTacToe()
model_class = MiniVGG
sims = 50
cuda = False
print("*** Rank Checkpoints ***")
rank_checkpoints(game, model_class, sims, cuda)
print("*** One vs All ***")
one_vs_all(checkpoint, game, model_class, sims, cuda)
print("*** Effective Model Power ***")
effective_model_power(checkpoint, game, model_class, sims, cuda)
plot_train_loss(game, [model_class], [cuda])
'''
x = [10,20,40,80,160,320,640,1280,2560,5120,10240,20480]
control = [1., .5, .5, .25, .25, .25, .25, .25, .25, .25, .5, .5]
senet = [1., .75, .75, .75, .75, .75, .75, .5, .5, .5, .5, .5]
def test_buffer_size_limit_100(self):
ttt = TicTacToe()
nn = NeuralNetwork(ttt, PriorNet)
t = Trainer(ttt, nn, num_simulations=2, num_games=100, num_updates=0, buffer_size_limit=100, cpuct=1, num_threads=4)
t.policy_iteration()
self.assertEqual(len(t.training_data), 100)
def test_simple(self):
game = TicTacToe()
p1, p2 = UninformedMCTSPlayer(game, 1), UninformedMCTSPlayer(game, 1)
scores, outcomes = play_match(game, [p1, p2])
self.assertEqual(scores, {p1: 1.0, p2: 0.0})
self.assertEqual(outcomes, {p1: "Win", p2: "Lose"})
