def __init__(self, sequence_len, batch_size, vocab_size, embedding_size, filter_sizes, num_filters, visible_size, hidden_size, dropout=1.0, l2_reg=0.0, learning_rate=1e-2, params=None,embeddings=None,loss='svm',trainable=True,score_type='nn_output'):
core.__init__(self,sequence_len,batch_size,vocab_size,embedding_size,filter_sizes,num_filters, visible_size, hidden_size, dropout,l2_reg,params,learning_rate,embeddings,loss,trainable,score_type)
self.model_type = "Gen"
self.reward = tf.placeholder(tf.float32, shape=[None], name='reward')
self.neg_index = tf.placeholder(tf.int32, shape=[None], name='neg_index')
self.gan_score = -tf.abs(self.neg_score - self.pos_score)
#self.gan_score = self.neg_score - self.pos_score
self.batch_scores =tf.nn.softmax(self.gan_score)
self.prob = tf.gather(self.batch_scores,self.neg_index)
self.gan_loss = -tf.reduce_mean(tf.log(self.prob) *self.reward) +l2_reg* self.l2_loss
#self.gan_loss = -tf.reduce_sum(tf.log(tf.clip_by_value(self.prob,1e-12,tf.reduce_max(self.prob))) *self.reward)
self.global_step = tf.Variable(0, name="global_step", trainable=False)
#optimizer = tf.train.AdamOptimizer(self.learning_rate)
#grads_and_vars = optimizer.compute_gradients(self.gan_loss)
#self.gan_updates = optimizer.apply_gradients(grads_and_vars, global_step=self.global_step)
optimizer = tf.train.GradientDescentOptimizer(self.learning_rate)
self.gan_updates = optimizer.minimize(self.gan_loss, global_step=self.global_step)
# minize attention
self.gans=-tf.reduce_mean(self.gan_score)
self.dns_score=self.neg_score
self.positive= tf.reduce_mean(self.pos_score)
self.negative= tf.reduce_mean(self.neg_score)
def __init__(self,_dataFilePath, _templateFilePath):
core.__init__(self, _dataFilePath, _templateFilePath)
self.memorizedData = {}
self.dataFilePath = _dataFilePath
self.templateFilePath = _templateFilePath
self.shellCommand=self.properties.get('shellCommand') or None
if not self.shellCommand:
self.error('Parameter shellCommand is REQUIRED in the properties file !!!')
raise
def __init__(self, _dataFilePath, _templateFilePath):
core.__init__(self, _dataFilePath, _templateFilePath)
self.min_random = properties.getInt('dummy.sleep.min', 0)
self.max_random = properties.getInt('dummy.sleep.max', 0)
if self.min_random > self.max_random:
self.error(
'Parameter dummy.sleep.min [%d] and dummy.sleep.max [%d] are not consistent ! '
% (self.min_random, self.max_random))
raise
self.sleep = True if self.min_random > 0 and self.max_random > 0 else False
def __init__(self,
sequence_len,
batch_size,
vocab_size,
embedding_size,
filter_sizes,
num_filters,
visible_size,
hidden_size,
dropout=1.0,
l2_reg=0.0,
learning_rate=1e-2,
params=None,
embeddings=None,
loss='svm',
trainable=True,
score_type='nn_output'):
core.__init__(self, sequence_len, batch_size, vocab_size,
embedding_size, filter_sizes, num_filters, visible_size,
hidden_size, dropout, l2_reg, params, learning_rate,
embeddings, loss, trainable, score_type)
self.model_type = 'Dis'
with tf.name_scope('output'):
if loss == 'svm':
self.losses = tf.maximum(
0.0, 0.05 - (self.pos_score - self.neg_score))
self.loss = tf.reduce_sum(
self.losses) + self.l2_reg * self.l2_loss
self.reward = 2 * (
tf.sigmoid(0.05 - (self.pos_score - self.neg_score)))
self.correct = tf.equal(0.0, self.losses)
elif loss == 'log':
self.losses = tf.log(
tf.sigmoid(self.pos_score - self.neg_score))
self.loss = -tf.reduce_mean(
self.losses) + self.l2_reg * self.l2_loss
self.reward = tf.reshape(
tf.log(tf.sigmoid(self.neg_score - self.pos_score) + 0.5),
[-1])
self.correct = tf.greater(0.0, self.losses)
self.positive = tf.reduce_mean(self.pos_score)
self.negative = tf.reduce_mean(self.neg_score)
#self.correct = tf.equal(0.0, self.losses)
self.accuracy = tf.reduce_mean(tf.cast(self.correct, 'float'),
name='accuracy')
self.global_step = tf.Variable(0, name='global_step', trainable=False)
optimizer = tf.train.GradientDescentOptimizer(self.learning_rate)
#optimizer = tf.train.AdamOptimizer(self.learning_rate)
self.updates = optimizer.minimize(self.loss,
global_step=self.global_step)
def __init__(self, _dataFilePath, _templateFilePath):
core.__init__(self, _dataFilePath, _templateFilePath)
self.min_random = grinder.getProperties().getInt('dummy.sleep.min', 0)
self.max_random = grinder.getProperties().getInt('dummy.sleep.max', 0)
if self.min_random > self.max_random:
self.error(
'Parameter dummy.sleep.min [%d] and dummy.sleep.max [%d] are not consistent ! '
% (self.min_random, self.max_random))
raise
self.sleep = True if self.min_random > 0 and self.max_random > 0 else False
properties = grinder.getProperties()
self.displayReadResponse = properties.getBoolean(
'displayReadResponse', False)
def __init__(self, _dataFilePath, _templateFilePath):
core.__init__(self, _dataFilePath, _templateFilePath)
self.lastStmt = 'OTHER'
self.db_type = None
self.connections = {}
str_conn = self.properties.get('db_connection') or None
logger.info('db_connection=%s' % (str_conn))
self.connection = None
if str_conn:
self.connection = self.getConnection(str_conn)
if self.connection:
self.connections[str_conn] = self.connection
self.cursor = self.connection.cursor()
self.alias = None
self.dictBind = {}
:param path: str, путь к файлу (без цифры)
:return: str
"""
i = 1
while os.path.isfile(path):
if i == 1:
path += '.1'
else:
path = path[:-2] + '.' + str(i)
i += 1
return path
if __name__ == '__main__':
core.__init__() # инициализация модуля
str_time = time.strftime('%b %Y %H:%M:%S', time.localtime())
profile = cProfile.Profile()
if core.config['MAIN']['cgitb'] == 'yes': # включить вывод ошибок
if core.config['MAIN']['tolog'] == 'yes': # вывод в лог
cgitb.enable(0, get_file(os.path.join(core.PATH_CGITB, str_time
+ 'log')))
else:
cgitb.enable()
if core.config['MAIN']['profile'] == 'yes': # включить профилирование
profile.enable()
header, content, cookies = core.run(cgi.FieldStorage()) # получение данных
# вывод заголовка
for line in header:
print(line)
print(get_all(cookies)) # вывод куков (относятся к заголовку)
def __init__(self, _dataFilePath, _templateFilePath):
core.__init__(self, _dataFilePath, _templateFilePath)