def test_bare_model_player_from_model():
game = OthelloGame(6)
some_net = OthelloNNet(game)
player = BareModelPlayer(game, some_net)
board = game.get_init_board()
action = player.play(board)
assert action
def test_alpha_zero_player_from_model():
game = OthelloGame(6)
some_net = OthelloNNet(game)
player = AlphaZeroPlayer(game, some_net)
board = game.get_init_board()
action = player.play(board)
assert action
def test_human_player_valid_action():
action_valid = "1,2"
game = OthelloGame(6)
player = HumanPlayer(game)
board = game.get_init_board()
with patch("builtins.input", side_effect=[action_valid]):
action = player.play(board)
assert action == 8
def test_alpha_zero_player_reset():
game = OthelloGame(6)
player = AlphaZeroPlayer(game, OthelloNNet, num_mcts_sims=4)
assert not player.mcts.Qsa
board = game.get_init_board()
player.play(board)
assert player.mcts.Qsa
player.reset()
assert not player.mcts.Qsa
def test_human_player_invalid_action(capsys):
action_valid = "1,2"
action_invalid = "0,0"
game = OthelloGame(6)
player = HumanPlayer(game)
board = game.get_init_board()
with patch("builtins.input", side_effect=[action_invalid, action_valid]):
player.play(board)
out, _err = capsys.readouterr()
assert out == "Invalid action.\n"
def test_bare_model_player_from_checkpoint():
game = OthelloGame(6)
some_net = OthelloNNet(game)
folder, filename = "/tmp/", "checkpoint_bare_model_player"
some_net.save_checkpoint(folder, filename)
del some_net
player = BareModelPlayer(game,
OthelloNNet,
folder=folder,
filename=filename)
board = game.get_init_board()
action = player.play(board)
assert action
def test_alpha_zero_player_from_checkpoint():
game = OthelloGame(6)
some_net = OthelloNNet(game)
folder, filename = "/tmp/", "checkpoint_alpha_zero_player"
some_net.save_checkpoint(folder, filename)
del some_net
player = AlphaZeroPlayer(game,
OthelloNNet,
folder=folder,
filename=filename,
num_mcts_sims=4)
board = game.get_init_board()
action = player.play(board)
assert action
def test_self_play(local_ray, tmpdir):
game = OthelloGame(6)
nnet = OthelloNNet(game)
s = SharedStorage.remote(nnet.get_weights())
r = ReplayBuffer.remote(games_to_play=1, games_to_use=1, folder=tmpdir)
assert ray.get(r.get_number_of_games_played.remote()) == 0
self_play = SelfPlay.remote(r, s, game, nnet.__class__, dict(args))
ray.get(self_play.start.remote())
assert ray.get(r.get_number_of_games_played.remote()) == 1
assert ray.get(r.played_enough.remote()) is True
games = ray.get(ray.get(r.get_examples.remote()))
assert len(games) == 1
examples = games[0]
assert len(examples) > 2
board, policy, winner = examples[0]
assert isinstance(board, type(game.get_init_board()))
assert len(policy) == game.get_action_size()
assert all(0 <= value <= 1 for value in policy)
assert winner in [1, -1]
def test_bare_model_player():
game = OthelloGame(6)
player = BareModelPlayer(game, OthelloNNet)
board = game.get_init_board()
action = player.play(board)
assert action
def test_alpha_zero_player():
game = OthelloGame(6)
player = AlphaZeroPlayer(game, OthelloNNet, num_mcts_sims=4)
board = game.get_init_board()
action = player.play(board)
assert action
def test_random_player():
game = OthelloGame(6)
player = RandomPlayer(game)
board = game.get_init_board()
action = player.play(board)
assert action
def test_greedy_player():
game = OthelloGame(6)
player = GreedyPlayer(game)
board = game.get_init_board()
action = player.play(board)
assert action