def uct_playerlist():
return [
UCTPlayer(ucb_const=100, num_samples=10, name="A"),
UCTPlayer(ucb_const=100, num_samples=10, name="A"),
UCTPlayer(ucb_const=100, num_samples=10, name="A"),
UCTPlayer(ucb_const=100, num_samples=10, name="A")
]
def main():
game_state = sample_game_states[18]
num_simulations = 100
ucb_constant = 0.1
player = UCTPlayer(ucb_const=ucb_constant, num_samples=1, num_simulations=num_simulations)
root_node = MCNode(game_state=game_state)
mc_tree = MCTree(root_node=root_node)
for sim_num in range(1, player.num_simulations + 1):
selected_node = player.selection(mc_tree)
rewards = player.simulation(selected_node)
mc_tree.backup_rewards(leaf_node=selected_node, rewards=rewards)
mc_tree.visualize_tree(ucb=player.ucb_const,
filename="img_{}".format(sim_num))
filenames = ["img_{}.png".format(i) for i in range(player.num_simulations, 0, -1)]
gif_name = "Gif_test.gif"
duration = 0.2
make_gif(filenames=filenames, outputname=gif_name, duration=duration)
im = Image.open(gif_name)
frames = [frame.copy() for frame in ImageSequence.Iterator(im)]
frames.reverse()
frames[0].save('reversed.gif', save_all=True, append_images=frames[1:])
def main():
sim_player_list = [
NNPlayer(
game_mode_nn='../players/models/bigger_classifier200.hdf5',
partner_nn='../players/models/partner_model_wider_data_2.hdf5',
solo_nn='../players/models/solo_model_wider_data_10.hdf5',
wenz_nn='../players/models/wenz_model_wider_data_10.hdf5'),
NNPlayer(
game_mode_nn='../players/models/bigger_classifier200.hdf5',
partner_nn='../players/models/partner_model_wider_data_2.hdf5',
solo_nn='../players/models/solo_model_wider_data_10.hdf5',
wenz_nn='../players/models/wenz_model_wider_data_10.hdf5'),
NNPlayer(
game_mode_nn='../players/models/bigger_classifier200.hdf5',
partner_nn='../players/models/partner_model_wider_data_2.hdf5',
solo_nn='../players/models/solo_model_wider_data_10.hdf5',
wenz_nn='../players/models/wenz_model_wider_data_10.hdf5'),
NNPlayer(
game_mode_nn='../players/models/bigger_classifier200.hdf5',
partner_nn='../players/models/partner_model_wider_data_2.hdf5',
solo_nn='../players/models/solo_model_wider_data_10.hdf5',
wenz_nn='../players/models/wenz_model_wider_data_10.hdf5')
]
game_state = sample_game_states[15]
num_simulations = 100
ucb_constant = 0.1
player = UCTPlayer(ucb_const=ucb_constant,
num_samples=1,
num_simulations=num_simulations,
simulation_player_list=None)
root_node = MCNode(game_state=game_state)
mc_tree = MCTree(root_node=root_node)
for sim_num in range(1, player.num_simulations + 1):
selected_node = player.selection(mc_tree)
rewards = player.simulation(selected_node)
mc_tree.backup_rewards(leaf_node=selected_node, rewards=rewards)
mc_tree.visualize_tree(
ucb=player.ucb_const,
filename="Tree_{}nodes{}ucb_const{}game_mode".format(
num_simulations, ucb_constant, game_state["game_mode"]))
def main():
sim_player_list = [NNPlayer(game_mode_nn='../players/models/bigger_classifier200.hdf5',
partner_nn='../players/models/partner_model_wider_data_2.hdf5',
solo_nn='../players/models/solo_model_wider_data_10.hdf5',
wenz_nn='../players/models/wenz_model_wider_data_10.hdf5'),
NNPlayer(game_mode_nn='../players/models/bigger_classifier200.hdf5',
partner_nn='../players/models/partner_model_wider_data_2.hdf5',
solo_nn='../players/models/solo_model_wider_data_10.hdf5',
wenz_nn='../players/models/wenz_model_wider_data_10.hdf5'),
NNPlayer(game_mode_nn='../players/models/bigger_classifier200.hdf5',
partner_nn='../players/models/partner_model_wider_data_2.hdf5',
solo_nn='../players/models/solo_model_wider_data_10.hdf5',
wenz_nn='../players/models/wenz_model_wider_data_10.hdf5'),
NNPlayer(game_mode_nn='../players/models/bigger_classifier200.hdf5',
partner_nn='../players/models/partner_model_wider_data_2.hdf5',
solo_nn='../players/models/solo_model_wider_data_10.hdf5',
wenz_nn='../players/models/wenz_model_wider_data_10.hdf5')]
cum_depth = 0
for game_state in sample_game_states:
num_simulations = 100
ucb_constant = 0.1
player = UCTPlayer(ucb_const=ucb_constant,
num_samples=1,
num_simulations=num_simulations,
simulation_player_list=None)
root_node = MCNode(game_state=game_state)
mc_tree = MCTree(root_node=root_node)
for sim_num in range(1, player.num_simulations + 1):
selected_node = player.selection(mc_tree)
rewards = player.simulation(selected_node)
mc_tree.backup_rewards(leaf_node=selected_node, rewards=rewards)
depth = mc_tree.max_depth()
cum_depth += depth
print('Average tree depth without NN:', cum_depth / len(sample_game_states))
cum_depth_with_nn = 0
for game_state in sample_game_states:
num_simulations = 100
ucb_constant = 0.1
player = UCTPlayer(ucb_const=ucb_constant,
num_samples=1,
num_simulations=num_simulations,
simulation_player_list=sim_player_list)
root_node = MCNode(game_state=game_state)
mc_tree = MCTree(root_node=root_node)
for sim_num in range(1, player.num_simulations + 1):
selected_node = player.selection(mc_tree)
rewards = player.simulation(selected_node)
mc_tree.backup_rewards(leaf_node=selected_node, rewards=rewards)
depth = mc_tree.max_depth()
cum_depth_with_nn += depth
print('Average tree depth with NN:', cum_depth_with_nn / len(sample_game_states))
partner_nn='../players/models/partner_model_bigger_1.hdf5',
solo_nn='../players/models/solo_model_bigger_1.hdf5',
wenz_nn='../players/models/wenz_model_bigger_1.hdf5'),
NNPlayer(game_mode_nn='../players/models/bigger_classifier200.hdf5',
partner_nn='../players/models/partner_model_bigger_1.hdf5',
solo_nn='../players/models/solo_model_bigger_1.hdf5',
wenz_nn='../players/models/wenz_model_bigger_1.hdf5'),
NNPlayer(game_mode_nn='../players/models/bigger_classifier200.hdf5',
partner_nn='../players/models/partner_model_bigger_1.hdf5',
solo_nn='../players/models/solo_model_bigger_1.hdf5',
wenz_nn='../players/models/wenz_model_bigger_1.hdf5')
]
playerlist = [
UCTPlayer(name="A",
num_samples=10,
num_simulations=1000,
simulation_player_list=None),
RandomPlayer(name="B"),
RandomPlayer(name="C"),
RandomPlayer(name="D")
]
player_hands_partner = [[(OBER, ACORNS), (OBER, BELLS), (UNTER, BELLS),
(ACE, BELLS), (KING, LEAVES), (TEN, ACORNS),
(SEVEN, ACORNS), (NINE, ACORNS)],
[(OBER, LEAVES), (OBER, HEARTS), (UNTER, ACORNS),
(ACE, HEARTS), (SEVEN, HEARTS), (ACE, ACORNS),
(KING, BELLS), (SEVEN, BELLS)],
[(UNTER, LEAVES), (TEN, LEAVES), (KING, HEARTS),
(KING, ACORNS), (TEN, HEARTS), (SEVEN, LEAVES),
(EIGHT, ACORNS), (NINE, BELLS)],
wenz_nn='../players/models/trickplay_model_wenz_extended.hdf5'),
NNPlayer(
game_mode_nn='../players/models/bigger_classifier200.hdf5',
partner_nn='../players/models/trickplay_model_partner_extended.hdf5',
solo_nn='../players/models/trickplay_model_solo_extended.hdf5',
wenz_nn='../players/models/trickplay_model_wenz_extended.hdf5'),
NNPlayer(
game_mode_nn='../players/models/bigger_classifier200.hdf5',
partner_nn='../players/models/trickplay_model_partner_extended.hdf5',
solo_nn='../players/models/trickplay_model_solo_extended.hdf5',
wenz_nn='../players/models/trickplay_model_wenz_extended.hdf5')
]
playerlist = [
UCTPlayer(num_samples=10,
num_simulations=20,
ucb_const=1,
simulation_player_list=sim_player_list),
# UCTPlayer(name="A", num_samples=10, num_simulations=100, ucb_const=1),
# NNPlayer(game_mode_nn='../players/models/bigger_classifier200.hdf5',
# partner_nn='../players/models/trickplay_model_partner_extended.hdf5',
# solo_nn='../players/models/trickplay_model_solo_extended.hdf5',
# wenz_nn='../players/models/trickplay_model_wenz_extended.hdf5',
# use_extended_models=True),
# HeuristicsPlayer(),
# RandomPlayer(),
# ISUCTPlayer(num_simulations=1000, ucb_const=2, simulation_player_list=None),
HeuristicsPlayer(name="B"),
HeuristicsPlayer(name="C"),
HeuristicsPlayer(name="D")
]