def test_loaded_dataset_white_move(self):
s_idcs_val, x_val, yv_val, yp_val, pgn_datasets_val = load_pgn_dataset(
dataset_type='test',
part_id=0,
print_statistics=True,
print_parameters=True,
normalize=True)
board = chess.variant.CrazyhouseBoard()
mv_converted = policy_to_move(yp_val[0], is_white_to_move=True)
mv_converted_is_legal = False
# check if the move is legal in the starting position
for mv in board.legal_moves:
if mv == mv_converted:
mv_converted_is_legal = True
self.assertTrue(
mv_converted_is_legal,
msg=
'Convert move %s is not a legal move in the starting position for WHITE'
% mv_converted.uci())
def moves_single_game(params_inp):
"""
Iterates over all moves of a given pgn file and comparse the reconverted policy representation with
the move stored in the pgn file
:param params_inp: pgn file, policy vector, starting index of the game
:return: Boolean indicating if the all moves have been equal in uci notation and the start_idx of the pgn
"""
pgn, yp_test, start_idx = params_inp
all_ok = True
pgn = deepcopy(pgn)
cur_game = chess.pgn.read_game(pgn)
# Iterate through all moves and play them on a board.
board = cur_game.board()
for i, move in enumerate(cur_game.main_line()):
# get the move in python chess format based on the policy representation
converted_move = policy_to_move(yp_test[i], is_white_to_move=board.turn)
cur_ok = converted_move == move
all_ok = all_ok and cur_ok
if not cur_ok:
logging.error('mat_move != move: %s - %s', converted_move, move)
board.push(move)
return all_ok, start_idx
def test_move_roundtrip_black(self):
mv = chess.Move.from_uci('e7e5')
policy_vec = move_to_policy(mv, is_white_to_move=False)
mv_conv = policy_to_move(policy_vec, is_white_to_move=False)
self.assertTrue(LABELS_MIRRORED[policy_vec.argmax()] == mv.uci())
self.assertTrue(mv == mv_conv)
def test_move_roundtrip_white(self):
move = chess.Move.from_uci('e2e4')
policy_vec = move_to_policy(move, is_white_to_move=True)
mv_converted = policy_to_move(policy_vec, is_white_to_move=True)
self.assertTrue(LABELS[policy_vec.argmax()] == move.uci())
self.assertTrue(move == mv_converted)
def test_move_roundtrip_black(self):
""" TODO: docstring"""
move = chess.Move.from_uci("e7e5")
policy_vec = move_to_policy(move, is_white_to_move=False)
mv_conv = policy_to_move(policy_vec, is_white_to_move=False)
self.assertTrue(LABELS_MIRRORED[policy_vec.argmax()] == move.uci())
self.assertTrue(move == mv_conv)
def test_loaded_dataset_black_move(self):
"""
Loads the dataset file and checks the first move policy vector for black for correctness
:return:
"""
s_idcs_val, x_val, yv_val, yp_val, pgn_datasets_val = load_pgn_dataset(
dataset_type='test',
part_id=0,
print_statistics=True,
print_parameters=True,
normalize=True)
board = chess.variant.CrazyhouseBoard()
# push a dummy move
board.push_uci('e2e4')
mv_conv0 = policy_to_move(yp_val[1], is_white_to_move=False)
mv_conv1, prob = policy_to_best_move(board, yp_val[1])
self.assertEqual(prob,
1,
msg='The policy vector has to be one hot encoded.')
selected_moves, move_probabilites = policy_to_moves(board, yp_val[1])
mv_conv2 = selected_moves[0]
self.assertGreater(move_probabilites[0],
0,
msg='The move probability must be greater 0')
self.assertEqual(move_probabilites[0],
1,
msg='The policy vector has to be one hot encoded.')
converted_moves = [mv_conv0, mv_conv1, mv_conv2]
for mv_converted in converted_moves:
mv_converted_is_legal = False
# check if the move is legal in the starting position
for mv in board.legal_moves:
if mv == mv_converted:
mv_converted_is_legal = True
self.assertTrue(
mv_converted_is_legal,
msg=
'Convert move %s is not a legal move in the starting position for BLACK'
% mv_converted.uci())
def test_loaded_dataset_white_move(self):
"""
Loads the dataset file and checks the first move policy vector for white for correctness
:return:
"""
_, _, _, yp_val, _ = load_pgn_dataset(dataset_type="test",
part_id=0,
print_statistics=True,
print_parameters=True,
normalize=True)
board = chess.variant.CrazyhouseBoard()
mv_conv0 = policy_to_move(yp_val[1], is_white_to_move=True)
mv_conv1, prob = policy_to_best_move(board, yp_val[1])
self.assertEqual(prob,
1,
msg="The policy vector has to be one hot encoded.")
selected_moves, move_probabilities = policy_to_moves(board, yp_val[1])
mv_conv2 = selected_moves[0]
self.assertGreater(move_probabilities[0],
0,
msg="The move probability must be greater 0")
self.assertEqual(move_probabilities[0],
1,
msg="The policy vector has to be one hot encoded.")
converted_moves = [mv_conv0, mv_conv1, mv_conv2]
for mv_converted in converted_moves:
mv_converted_is_legal = False
# check if the move is legal in the starting position
for move in board.legal_moves:
if move == mv_converted:
mv_converted_is_legal = True
self.assertTrue(
mv_converted_is_legal,
msg=
"Convert move %s is not a legal move in the starting position for BLACK"
% mv_converted.uci(),
)