ConnectSix/loop.py at master · StepNeverStop/ConnectSix · GitHub

1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
import sys

import numpy as np

from player.policy import MyPolicy
from utils.GymRender import GymRender


def _choose_valid_action(player, env_or_state, use_env=True):
    """Retry until the player returns a legal move."""
    while True:
        try:
            x, y = player.choose_action(env_or_state)
            if use_env:
                is_ok, _msg = env_or_state.can_place(x, y)
                if not is_ok:
                    continue
            return x, y
        except KeyboardInterrupt:
            print('游戏中断')
            sys.exit()


def _apply_turn(env, player, now_player, move_step):
    """Apply one move and return updated turn state."""
    x, y = _choose_valid_action(player, env, use_env=True)
    print(player.name, x, y)
    env.step(x, y)
    move_step += 1
    end, winner = env.is_over()
    if end:
        return move_step, now_player, True, winner
    if move_step == 2:
        return 0, (now_player + 1) % 2, False, None
    return move_step, now_player, False, None


def battle(env, players):
    """Headless battle loop for model evaluation."""
    env.register(players[0].name, players[-1].name)
    env.reset()
    now_player = 0
    move_step = 1

    while True:
        if now_player == 0:
            print('黑子')
        else:
            print('白子')
        move_step, now_player, finished, winner = _apply_turn(
            env, players[now_player], now_player, move_step
        )
        if finished:
            print(env.last_move)
            print(f'游戏结束，{players[now_player].name}获胜')
            break


def test_loop(env, players):
    render = GymRender()
    env.register(players[0].name, players[-1].name)
    env.reset()
    now_player = 0
    move_step = 1

    while True:
        env.render()
        if now_player == 0:
            print('黑子')
        else:
            print('白子')
        move_step, now_player, finished, winner = _apply_turn(
            env, players[now_player], now_player, move_step
        )
        if finished:
            env.render()
            render.render(env)
            print(env.last_move)
            print(f'游戏结束，{players[now_player].name}获胜')
            input()
            break


def train_loop(env, players):
    info = ['黑子', '白子']
    offset = [2, 1]
    wins = [0, 0]

    for episode in range(10000):
        if np.random.rand() > 0.5:
            players.reverse()
            wins.reverse()
        for i in range(2):
            if isinstance(players[i], MyPolicy):
                players[i].update_offset(offset[i])

        env.register(players[0].name, players[-1].name)

        state = env.reset()
        now_player = 0
        move_step = 1
        total_step = 0
        states = [
            [state, state],
            [state, state],
        ]

        while True:
            total_step += 1
            while True:
                try:
                    x, y = players[now_player].choose_action(state)
                    is_ok, _msg = env.can_place(x, y)
                    if not is_ok:
                        continue
                    break
                except KeyboardInterrupt:
                    print('游戏中断')
                    sys.exit()

            state = env.step(x, y)
            states[now_player][-1] = state
            move_step += 1
            if wins[now_player] - wins[(now_player + 1) % 2] < 6:
                players[now_player].learn()

            end, winner = env.is_over()
            if end:
                if winner == -1:
                    r0, r1 = 0.1, 0.1
                    print(f'episode: {episode}, step: {total_step:>3d}, 平局')
                else:
                    r0, r1 = 1, -move_step
                    print(
                        f'episode: {episode:>4d}, step: {total_step:>3d}, '
                        f'{info[now_player]}{players[now_player].name}获胜'
                    )
                    wins[now_player] += 1

                players[now_player].store(
                    s=states[now_player][0], r=r0,
                    s_=states[now_player][1], done=True,
                )
                players[now_player].writer_loop_summary(episode, reward=r0, step=total_step)
                next_player = (now_player + 1) % 2
                players[next_player].store(
                    s=states[next_player][0], r=r1,
                    s_=states[next_player][1], done=True,
                )
                players[next_player].writer_loop_summary(
                    episode, reward=r1, step=total_step
                )
                break

            if move_step == 2:
                move_step = 0
                now_player = (now_player + 1) % 2
                players[now_player].store(
                    s=states[now_player][0], r=0,
                    s_=states[now_player][1], done=False,
                )
                states[now_player][0] = states[(now_player + 1) % 2][1]