def load_data_val_6(testList, W, index, batch):
x_train_1 = []
x_train_2 = []
x_train_3 = []
for i in range(0, batch):
true_index = index + i
if (true_index >= len(testList)):
true_index = len(testList) - 1
items = testList[true_index].split(' ')
q_words = items[2].split('_')
a_words = items[3].split('_')
x_train_1_words = []
x_train_2_words = []
x_train_3_words = []
for i in range(50):
x_train_1_words.append(
ModelUtils.build_text_image(W, q_words[i], padding=1))
x_train_2_words.append(
ModelUtils.build_text_image(W, a_words[i], padding=1))
x_train_3_words.append(
ModelUtils.build_text_image(W, a_words[i], padding=1))
x_train_1.append(np.array(x_train_1_words).reshape((50, 300)))
x_train_2.append(np.array(x_train_2_words).reshape((50, 300)))
x_train_3.append(np.array(x_train_3_words).reshape((50, 300)))
return np.array(x_train_1), np.array(x_train_2), np.array(x_train_3)
def load_data_6(W, alist, raw, size):
x_train_1 = []
x_train_2 = []
x_train_3 = []
for i in range(0, size):
items = raw[random.randint(0, len(raw) - 1)]
nega = rand_qa(alist)
q_words = items[2].split('_')
a_words = items[3].split('_')
neg_words = nega.split('_')
x_train_1_words = []
x_train_2_words = []
x_train_3_words = []
for i in range(50):
x_train_1_words.append(
ModelUtils.build_text_image(W, q_words[i], padding=1))
x_train_2_words.append(
ModelUtils.build_text_image(W, a_words[i], padding=1))
x_train_3_words.append(
ModelUtils.build_text_image(W, neg_words[i], padding=1))
x_train_1.append(np.array(x_train_1_words).reshape((50, 300)))
x_train_2.append(np.array(x_train_2_words).reshape((50, 300)))
x_train_3.append(np.array(x_train_3_words).reshape((50, 300)))
return np.array(x_train_1), np.array(x_train_2), np.array(x_train_3)