def play_round(self):
logger.info("Starting game round...")
if np.random.uniform(0, 1) <= 0.5:
white = self.current
black = self.best
w = "current"
b = "best"
else:
white = self.best
black = self.current
w = "best"
b = "current"
current_board = cboard()
checkmate = False
dataset = []
value = 0
t = 0.1
while checkmate == False and current_board.actions() != []:
dataset.append(copy.deepcopy(ed.encode_board(current_board)))
print("")
print(current_board.current_board)
if current_board.player == 0:
root = UCT_search(current_board, 777, white, t)
policy = get_policy(root, t)
print("Policy: ", policy, "white = %s" % (str(w)))
elif current_board.player == 1:
root = UCT_search(current_board, 777, black, t)
policy = get_policy(root, t)
print("Policy: ", policy, "black = %s" % (str(b)))
current_board = do_decode_n_move_pieces(current_board,\
np.random.choice(np.array([0,1,2,3,4,5,6,7,8]), \
p = policy)) # decode move and move piece(s)
if current_board.check_winner() == True: # someone wins
if current_board.player == 0: # black wins
value = -1
elif current_board.player == 1: # white wins
value = 1
checkmate = True
dataset.append(ed.encode_board(current_board))
if value == -1:
dataset.append(f"{b} as black wins")
return b, dataset
elif value == 1:
dataset.append(f"{w} as white wins")
return w, dataset
else:
dataset.append("Nobody wins")
return None, dataset
def play_ia(game, options):
made_moves, convertedMatrix = convert_matrix_to_alpha_zero(game)
print(convertedMatrix)
#########################################################################
# AlphaZero
#########################################################################
#best_net="c4_current_net_trained2_iter7.pth.tar"
best_net="c4_current_net_trained_iter8.pth.tar"
best_net_filename = os.path.join("C:\\Users\\Francesco\\Documents\\2_Schule\\HYU\\HYU-software-engineering-AI\\src_backend\\src\\ai_modules\\AlphaZero_Connect4\\src\\model_data\\", best_net)
best_cnet = ConnectNet()
cuda = torch.cuda.is_available()
if cuda:
best_cnet.cuda()
best_cnet.eval()
checkpoint = torch.load(best_net_filename)
best_cnet.load_state_dict(checkpoint['state_dict'])
net = best_cnet
white = None
black = net
current_board = cboard()
current_board.current_board = np.array(convertedMatrix)
checkmate = False
dataset = []
value = 0; t = 0.1; moves_count = made_moves
moves_count += 1
dataset.append(copy.deepcopy(ed.encode_board(current_board)))
print("AI is thinking.............")
root = UCT_search(current_board,777,black,t)
policy = get_policy(root, t)
current_board = do_decode_n_move_pieces(current_board, np.random.choice(np.array([0,1,2,3,4,5,6]), p = policy)) # decode move and move piece(s)
print(current_board.current_board); print(" ")
return find_which_move_ai_made(convertedMatrix, current_board.current_board)
def play_game(net):
# Asks human what he/she wanna play as
white = None
black = None
while (True):
play_as = input("What do you wanna play as? (\"O\"/\"X\")? Note: \"O\" starts first, \"X\" starts second\n")
if play_as == "O":
black = net; break
elif play_as == "X":
white = net; break
else:
print("I didn't get that.")
current_board = cboard()
checkmate = False
dataset = []
value = 0
t = 0.1
moves_count = 0
while not checkmate and current_board.actions() != []:
t = 1 if moves_count <= 5 else 0.1
moves_count += 1
dataset.append(copy.deepcopy(ed.encode_board(current_board)))
print(current_board.current_board)
print(" ")
if current_board.player == 0 and white != None:
print("AI is thinking........")
root = UCT_search(current_board,777,white,t)
policy = get_policy(root, t)
elif (
current_board.player == 0
or current_board.player == 1
and black is None
):
while(True):
col = input("Which column do you wanna drop your piece? (Enter 1-7)\n")
if int(col) in [1,2,3,4,5,6,7]:
policy = np.zeros([7], dtype=np.float32); policy[int(col)-1] += 1
break
elif current_board.player == 1:
print("AI is thinking.............")
root = UCT_search(current_board,777,black,t)
policy = get_policy(root, t)
current_board = do_decode_n_move_pieces(current_board,\
np.random.choice(np.array([0,1,2,3,4,5,6]), \
p = policy)) # decode move and move piece(s)
if current_board.check_winner() == True: # someone wins
if current_board.player == 0: # black wins
value = -1
elif current_board.player == 1: # white wins
value = 1
checkmate = True
dataset.append(ed.encode_board(current_board))
print(current_board.current_board)
print(" ")
if value == -1:
if play_as == "O":
dataset.append('AI as black wins')
print("YOU LOSE!!!!!!!")
else:
dataset.append('Human as black wins')
print("YOU WIN!!!!!!!")
return "black", dataset
elif value == 1:
if play_as == "O":
dataset.append('Human as white wins')
print("YOU WIN!!!!!!!!!!!")
else:
dataset.append('AI as white wins')
print("YOU LOSE!!!!!!!")
return "white", dataset
else:
dataset.append("Nobody wins")
print("DRAW!!!!!")
return None, dataset
