def run_epoch(self, sess, verbose=10):
total_loss = []
total_acc_num = []
total_num = []
for step, indices in enumerate(
batch_index(len(self.train_doc_y), self.config.batch_size, 1),
1):
feed_dict = self.create_feed_dict(self.train_x[indices],
self.train_sen_len[indices],
self.train_doc_len[indices],
self.train_doc_y[indices],
self.config.keep_prob1,
self.config.keep_prob2)
_, loss, acc_num, lr = sess.run(
[self.train_op, self.doc_loss, self.accuracy_num, self.lr],
feed_dict=feed_dict)
total_loss.append(loss)
total_acc_num.append(acc_num)
total_num.append(len(indices))
if verbose and step % verbose == 0:
print '\n[INFO] {} : loss = {}, acc = {}, lr = {}'.format(
step, np.mean(total_loss[-verbose:]),
sum(total_acc_num[-verbose:]) * 1.0 /
sum(total_num[-verbose:]), lr)
return np.mean(total_loss), sum(total_acc_num) * 1.0 / sum(total_num)
def run_op(self,
sess,
op,
data_x,
sen_len,
doc_len,
doc_y=None,
kp1=1.0,
kp2=1.0):
res_list = []
len_list = []
for indices in batch_index(len(data_x), self.config.batch_size, 1,
False, False):
if doc_y is not None:
feed_dict = self.create_feed_dict(data_x[indices],
sen_len[indices],
doc_len[indices],
doc_y[indices], kp1, kp2)
else:
feed_dict = self.create_feed_dict(data_x[indices],
sen_len[indices],
doc_len[indices], None, kp1,
kp2)
res = sess.run(op, feed_dict=feed_dict)
res_list.append(res.tolist())
len_list.append(len(indices))
if type(res_list[0]) is list:
res = np.concatenate(res_list, axis=1)
elif op is self.accuracy_num:
res = sum(res_list)
elif op is self.doc_logits:
res = np.asarray(res_list)
else:
res = sum(res_list) * 1.0 / len(len_list)
return res
def get_batch_data(xo, slo, yy, batch_size, kp1, kp2, is_shuffle=True):
for index in batch_index(len(yy), batch_size, 1, is_shuffle):
feed_dict = {
x: xo[index],
y: yy[index],
sen_len: slo[index],
keep_prob1: kp1,
keep_prob2: kp2,
}
yield feed_dict, len(index)
def get_batch_data(xo, slo, dlo, xr, slr, dlr, yy, batch_size, kp1, kp2, is_shuffle=True):
for index in batch_index(len(yy), batch_size, 1, is_shuffle):
feed_dict = {
x_o: xo[index],
x_r: xr[index],
y: yy[index],
sen_len_o: slo[index],
sen_len_r: slr[index],
doc_len_o: dlo[index],
doc_len_r: dlr[index],
keep_prob1: kp1,
keep_prob2: kp2,
}
yield feed_dict, len(index)
def get_batch_data(x_in,
y_in,
sen_len_in,
doc_len_in,
batch_size,
kp1,
kp2,
is_shuffle=True):
for index in batch_index(len(y_in), batch_size, 1, is_shuffle):
feed_dict = {
x: x_in[index],
y: y_in[index],
sen_len: sen_len_in[index],
doc_len: doc_len_in[index],
keep_prob1: kp1,
keep_prob2: kp2,
}
yield feed_dict, len(index)
def get_batch_data(x1,
x2,
len1,
len2,
yy,
batch_size,
kp1,
kp2,
is_shuffle=True):
for index in batch_index(len(yy), batch_size, 1, is_shuffle):
feed_dict = {
x_l: x1[index],
x_r: x2[index],
y: yy[index],
len_l: len1[index],
len_r: len2[index],
keep_prob1: kp1,
keep_prob2: kp2,
}
yield feed_dict, len(index)