def paintEvent(self, e):
#data = self.itemData(self.currentIndex())
data = NWM.intToStyle(NWM.LINE_STYLE_LIST[self.currentIndex()])
line_color = self.network_index.model().createIndex(self.network_index.row(),
NWM.NET_COLOR).data()
#line_color = self.network_index.model().data(self.network_index.row(),
# WLM.NET_COLOR)
p = QStylePainter(self)
p.setPen(self.palette().color(QPalette.Text))
opt = QStyleOptionComboBox()
self.initStyleOption(opt)
p.drawComplexControl(QStyle.CC_ComboBox, opt)
painter = QPainter(self)
painter.save()
rect = p.style().subElementRect(QStyle.SE_ComboBoxFocusRect, opt, self)
rect.adjust(+5, 0, -5, 0)
pen = QPen()
pen.setColor(line_color)
pen.setWidth(3)
pen.setStyle(data)
painter.setPen(pen)
middle = (rect.bottom() + rect.top()) / 2
painter.drawLine(rect.left(), middle, rect.right(), middle)
painter.restore()
def paint(self, painter, option, index):
data = NWM.intToStyle(int(index.data()))
line_color = self.network_index.model().createIndex(self.network_index.row(),
NWM.NET_COLOR).data()
#line_color = self.network_model.data(index.row(), WLM.NET_COLOR)
painter.save()
rect = option.rect
rect.adjust(+5, 0, -5, 0)
pen = QPen()
pen.setColor(line_color)
pen.setWidth(3)
pen.setStyle(data)
painter.setPen(pen)
middle = (rect.bottom() + rect.top()) / 2
painter.drawLine(rect.left(), middle, rect.right(), middle)
painter.restore()
import sys
from digits import DigitData
from model import NetworkModel
if __name__ == '__main__':
model_file = sys.argv[1]
test_data = DigitData.from_json(sys.argv[2])
model = NetworkModel.load(model_file)
print('Accuracy: %s' % model.evaluate(test_data, progress=True))
import activation
from digits import DigitData
from model import NetworkModel
EPOCHS = 60
def train(model, train_data, test_data, model_file, epochs=EPOCHS):
try:
for i in range(epochs):
print('Epoch %s' % i)
train_data.shuffle()
for datum in tqdm(train_data.data[:10000]):
model.train(datum.features(), datum.label_vec)
print('Evaluating...')
print('Accuracy: %s' % model.evaluate(test_data, progress=True))
except KeyboardInterrupt:
pass
finally:
model.save(model_file)
# python train.py data/train.json data/weights.json data/model.json
if __name__ == '__main__':
train_data = DigitData.from_json(sys.argv[1])
test_data = DigitData.from_json(sys.argv[2])
model_file = sys.argv[3]
model = NetworkModel(train_data.num_features(), 30, train_data.num_labels(),
activation_fns=[activation.Sigmoid, activation.Sigmoid])
train(model, train_data, test_data, model_file)
from osero import Osero
env = Osero(board_size=6)
exit()
from logger import EpisodeLogger
from model import NetworkModel
from keras.optimizers import Adam
from rl.agents.dqn import DQNAgent
from rl.policy import EpsGreedyQPolicy
from rl.memory import SequentialMemory
env = Osero(board_size=4)
nb_actions = env.action_space.n
model = NetworkModel.create_simple_nn(env)
print(model.summary())
memory = SequentialMemory(limit=10000, window_length=1)
policy = EpsGreedyQPolicy(eps=0.1)
dqn = DQNAgent(model=model,
nb_actions=nb_actions,
memory=memory,
nb_steps_warmup=100,
target_model_update=1e-2,
policy=policy)
dqn.compile(Adam(lr=1e-3), metrics=['mae'])
# dqn.load_weights('results/190621/10000.h5')
history = dqn.fit(env,
nb_steps=10000,
