def main():
cur_state = SquadroState()
agents = [
getattr(__import__(agent_0), 'MyAgent')(),
getattr(__import__(agent_1), 'MyAgent')()
]
agents[0].set_id(0)
agents[1].set_id(1)
cur_state.cur_pos[0][0] = (100, 400)
cur_state.cur_pos[0][1] = (200, 100)
cur_state.returning[0][1] = True
cur_state.finished[0][2] = True
cur_state.finished[0][3] = True
cur_state.cur_player = 0
l1 = [
cur_state.get_pawn_advancement(cur_state.cur_player, pawn)
for pawn in [0, 1, 2, 3, 4]
]
l2 = [
cur_state.get_pawn_advancement(1 - cur_state.cur_player, pawn)
for pawn in [0, 1, 2, 3, 4]
]
#print(l1, l2)
for i in range(2, 50):
agents[0].MC_steps = i
action = agents[0].get_action(cur_state.copy(), 0, 10)
cur_state.apply_action(action)
def main(agent_0, agent_1, first):
# Initialisation
cur_state = SquadroState()
if first != -1:
cur_state.cur_player = first
agents = [
getattr(__import__(agent_0), 'MyAgent')(),
getattr(__import__(agent_1), 'MyAgent')()
]
agents[0].set_id(0)
agents[1].set_id(1)
last_action = None
while not cur_state.game_over():
# Make move
cur_player = cur_state.get_cur_player()
action = get_action_timed(agents[cur_player], cur_state.copy(),
last_action)
if cur_state.is_action_valid(action):
cur_state.apply_action(action)
last_action = action
else:
cur_state.set_invalid_action(cur_player)
print(cur_player)
def game(agent, validation, i, new_model_path):
results = 0
init = 1
# Initialisation
cur_state = SquadroState()
cur_state.cur_player = 0 if i % 2 == 0 else 1
agents = [getattr(__import__(agent), 'MyAgent')(), getattr(__import__(agent), 'MyAgent')()]
agents[0].set_id(0)
agents[1].set_id(1)
#print_network(agents[0].deepnetwork)
if validation: # Use different models during validation phase
ind = 0 if i % 4 < 2 else 1
agents[ind].set_model_path(new_model_path)
print('Main model:', ind, ', First player:', cur_state.cur_player)
# Remove stochastic actions
#agents[0].epsilonMCTS = 0
#agents[1].epsilonMCTS = 0
agents[0].epsilonMove = 0
agents[1].epsilonMove = 0
'''
print('Current network model...............................................')
print_network(agents[0].deepnetwork)
print('New network model....................................................')
print_network(agents[1].deepnetwork)
'''
last_action = None
while not cur_state.game_over():
# Make move
cur_player = cur_state.get_cur_player()
action = get_action_timed(agents[cur_player], cur_state.copy(), last_action)
if cur_state.is_action_valid(action):
cur_state.apply_action(action)
last_action = action
else:
cur_state.set_invalid_action(cur_player)
if init:
results = agents[cur_player].results
init = 0
else:
results = np.append(results, agents[cur_player].results, 0)
#if validation:
# print(results)
# print(cur_player)
return (results, cur_player)
def game(agent_0, agent_1, i):
model_path = 'model/{}.pt'.format(agent_0)
results = 0
init = 1
# Initialisation
cur_state = SquadroState()
cur_state.cur_player = 0 if i % 4 < 2 else 1
if i % 2 == 0:
agents = [getattr(__import__(agent_0), 'MyAgent')(), getattr(__import__(agent_1), 'MyAgent')()]
DNN_player = 0
else:
agents = [getattr(__import__(agent_1), 'MyAgent')(), getattr(__import__(agent_0), 'MyAgent')()]
DNN_player = 1
print(DNN_player, cur_state.cur_player)
agents[0].set_id(0)
agents[1].set_id(1)
agents[DNN_player].epsilonMove = 0
#if i == 0:
#print('Network 0 (main) -------------------------------------------------------')
#print_network(agents[0].deepnetwork)
#print('Network 1 (other) -------------------------------------------------------')
#print_network(agents[1].deepnetwork)
last_action = None
while not cur_state.game_over():
# Make move
cur_player = cur_state.get_cur_player()
action = get_action_timed(agents[cur_player], cur_state.copy(), last_action)
if cur_state.is_action_valid(action):
cur_state.apply_action(action)
last_action = action
else:
cur_state.set_invalid_action(cur_player)
#print('Last state:', agents[DNN_player].results)
return (0, abs(cur_player - DNN_player))
def main(agent_0, agent_1, time_out, first):
# Initialisation
pygame.init()
n_tiles = 7
n_pawns = 5
board = Board(n_tiles, n_pawns)
cur_state = SquadroState()
if first != -1:
cur_state.cur_player = first
agents = [
getattr(__import__(agent_0), 'MyAgent')(),
getattr(__import__(agent_1), 'MyAgent')()
]
agents[0].set_id(0)
agents[1].set_id(1)
times_left = [time_out, time_out]
last_action = None
while not cur_state.game_over():
# Draw board
board.screen.fill(0)
board.draw_board(cur_state)
board.show_turn(cur_state)
# Update screen
pygame.display.flip()
# Make move
cur_player = cur_state.get_cur_player()
timer_stop = [False]
timer = TimerDisplay(board, cur_player, times_left.copy(), timer_stop)
timer.start()
try:
action, exe_time = get_action_timed(agents[cur_player],
cur_state.copy(), last_action,
times_left[cur_player])
timer_stop[0] = True
times_left[cur_player] -= exe_time
if cur_state.is_action_valid(action):
cur_state.apply_action(action)
last_action = action
else:
cur_state.set_invalid_action(cur_player)
except TimeoutError:
timer_stop[0] = True
cur_state.set_timed_out(cur_player)
timer.join()
# Events
for event in pygame.event.get():
# Quit when pressing the X button
if event.type == pygame.QUIT:
pygame.quit()
exit(0)
# Game finished: display the winner
while True:
# Draw board
board.screen.fill(0)
board.draw_board(cur_state)
# Print the winner
font = pygame.font.Font("freesansbold.ttf", 48)
if cur_state.get_winner() == 0:
text = font.render(" Yellow wins! ", True, (255, 164, 0),
(34, 34, 34))
else:
text = font.render(" Red wins! ", True, (150, 0, 0), (34, 34, 34))
textRect = text.get_rect()
textRect.center = (n_tiles * 100 // 2, n_tiles * 100 // 2)
board.screen.blit(text, textRect)
# Print if time-out or invalid action
if cur_state.timeout_player != None:
font2 = pygame.font.Font("freesansbold.ttf", 18)
text2 = font2.render("The opponent timed out", True,
(255, 255, 255), (34, 34, 34))
textRect2 = text2.get_rect()
textRect2.center = (n_tiles * 100 // 2, n_tiles * 100 // 2 + 34)
board.screen.blit(text2, textRect2)
elif cur_state.invalid_player != None:
font2 = pygame.font.Font("freesansbold.ttf", 18)
text2 = font2.render("The opponent made an invalid move", True,
(255, 255, 255), (34, 34, 34))
textRect2 = text2.get_rect()
textRect2.center = (n_tiles * 100 // 2, n_tiles * 100 // 2 + 34)
board.screen.blit(text2, textRect2)
# Update screen
pygame.display.flip()
for event in pygame.event.get():
if event.type == pygame.QUIT:
pygame.quit()
exit(0)
sleep(1)
def game(agent_0, agent_1, i):
model_path = 'model/{}.pt'.format(agent_0)
other_model_path = 'model/{}.pt'.format(agent_1)
results = 0
init = 1
# Initialisation
cur_state = SquadroState()
agents = [
getattr(__import__(agent_0), 'MyAgent')(),
getattr(__import__(agent_0), 'MyAgent')()
]
DNN_main_player = 0 if i % 2 == 0 else 1
cur_state.cur_player = 0 if (2 * i) % 2 == 0 else 1
print(DNN_main_player, cur_state.cur_player)
agents[0].set_id(0)
agents[1].set_id(1)
agents[1 - DNN_main_player].set_model_path(other_model_path)
'''
agents[0].epsilonMCTS = 0
agents[1].epsilonMCTS = 0
'''
agents[0].epsilonMove = 0
agents[1].epsilonMove = 0
if i == 0:
print(
'Network 0 (main) -------------------------------------------------------'
)
print_network(agents[DNN_main_player].deepnetwork)
print(
'Network 1 (other) -------------------------------------------------------'
)
print_network(agents[1 - DNN_main_player].deepnetwork)
last_action = None
while not cur_state.game_over():
# Make move
cur_player = cur_state.get_cur_player()
action = get_action_timed(agents[cur_player], cur_state.copy(),
last_action)
if cur_state.is_action_valid(action):
cur_state.apply_action(action)
last_action = action
else:
cur_state.set_invalid_action(cur_player)
if init:
results = agents[cur_player].results
init = 0
else:
results = np.append(results, agents[cur_player].results, 0)
return (results, abs(DNN_main_player - cur_player))