def play():
env = OmokEnv(BOARD_SIZE, HISTORY)
mcts = MCTS(BOARD_SIZE, HISTORY, N_SIMUL)
result = {'Black': 0, 'White': 0, 'Draw': 0}
for g in range(GAME):
print('#' * (BOARD_SIZE - 4),
' GAME: {} '.format(g + 1),
'#' * (BOARD_SIZE - 4))
# reset state
state, board = env.reset()
done = False
while not done:
env.render()
# start simulations
action = mcts.get_action(state, board)
state, board, z, done = env.step(action)
if done:
if z == 1:
result['Black'] += 1
elif z == -1:
result['White'] += 1
else:
result['Draw'] += 1
# render & reset tree
env.render()
mcts.reset_tree()
# result
blw, whw, drw = result['Black'], result['White'], result['Draw']
print('')
print('=' * 20, " {} Game End ".format(blw + whw + drw), '=' * 20)
stats = (
'Black Win: {} White Win: {} Draw: {} Winrate: {:.2f}%'.format(
blw, whw, drw, (blw + 0.5 * drw) / (blw + whw + drw) * 100))
print(stats, '\n')
def main():
env = OmokEnv(BOARD_SIZE, HISTORY)
manager = HumanUI()
result = {'Black': 0, 'White': 0, 'Draw': 0}
for g in range(GAME):
print('########## Game: {} ##########'.format(g + 1))
state, board = env.reset()
done = False
idx = 0
while not done:
env.render()
# start simulations
action = manager.get_action(state, board, idx)
state, board, z, done = env.step(action)
idx += 1
if done:
if z == 1:
result['Black'] += 1
elif z == -1:
result['White'] += 1
else:
result['Draw'] += 1
# render & reset tree
env.render()
manager.ai.reset_tree()
# result
print('')
print('=' * 20, " {} Game End ".format(g + 1), '=' * 20)
blw, whw, drw = result['Black'], result['White'], result['Draw']
stat = (
'Black Win: {} White Win: {} Draw: {} Winrate: {:0.1f}%'.format(
blw, whw, drw,
1 / (1 + np.exp(whw / (g + 1)) / np.exp(blw / (g + 1))) * 100))
print(stat, '\n')
def play():
USE_CUDA = torch.cuda.is_available()
env = OmokEnv(BOARD_SIZE, HISTORY)
manager = HumanUI()
if USE_CUDA:
manager.ai.model.cuda()
model_path = None
if model_path:
print('load model: {}\n'.format(model_path))
manager.ai.model.load_state_dict(torch.load(model_path))
result = {'Black': 0, 'White': 0, 'Draw': 0}
for g in range(N_GAME):
print('##### Game: {} #####'.format(g + 1))
state, board = env.reset()
done = False
idx = 0
while not done:
env.render()
# start simulations
action = manager.get_action(state, board, idx)
state, board, z, done = env.step(action)
idx += 1
if done:
if z == 1:
result['Black'] += 1
elif z == -1:
result['White'] += 1
else:
result['Draw'] += 1
# render & reset tree
env.render()
manager.ai.reset_tree()
# result
blw, whw, drw = result['Black'], result['White'], result['Draw']
print('')
print('=' * 20, " {} Game End ".format(blw + whw + drw), '=' * 20)
stats = (
'Black Win: {} White Win: {} Draw: {} Winrate: {:.2f}%'.format(
blw, whw, drw, (blw + 0.5 * drw) / (blw + whw + drw) * 100))
print(stats, '\n')