def convert_to_training_data(data):
"""
:param data: format is SelfPlayWorker.buffer
:return:
"""
state_list = []
policy_list = []
z_list = []
for state, policy, z in data:
env = ChessEnv().update(state)
black_ary, white_ary = env.black_and_white_plane()
state = [black_ary, white_ary] if env.board.turn == chess.BLACK else [white_ary, black_ary]
state_list.append(state)
policy_list.append(policy)
z_list.append(z)
return np.array(state_list), np.array(policy_list), np.array(z_list)
def convert_to_training_data(data):
"""
:param data: format is SelfPlayWorker.buffer
:return:
"""
state_list = []
policy_list = []
z_list = []
aux_move_number = 1
movements = []
for state, policy, z in data:
move_number = int(
(ChessEnv().update(state,
movements)).board.fen().split(" ")[5])
if aux_move_number < move_number:
if len(movements) > 8:
movements.pop(0)
movements.append(env.observation)
aux_move_number = move_number
else:
aux_move_number = 1
movements = []
env = ChessEnv().update(state, movements)
black_ary, white_ary, current_player, move_number = env.black_and_white_plane(
)
state = [black_ary, white_ary
] if env.board.fen().split(" ")[1] == 'b' else [
white_ary, black_ary
]
state = np.reshape(np.reshape(np.array(state), (18, 6, 8, 8)),
(108, 8, 8))
state = np.vstack((state, np.reshape(current_player, (1, 8, 8)),
np.reshape(move_number, (1, 8, 8))))
state_list.append(state)
policy_list.append(policy)
z_list.append(z)
return np.array(state_list), np.array(policy_list), np.array(z_list)
