lr = 0.001 #learning rate
num_epochs = 5 #epochs
batch_size = 64 #batch_size
keep_p = 0.5 #keep_prob
train_filename = './data/cnews.train.txt' #train data
test_filename = './data/cnews.test.txt' #test data
val_filename = './data/cnews.val.txt' #validation data
vocab_filename = './data/vocab_word.txt' #vocabulary
vector_word_filename = './data/vector_word.txt' #vector_word trained by word2vec
vector_word_npz = './data/vector_word.npz' # save vector_word to numpy file
categories, cat_to_id = read_category()
wordid = get_wordid(vocab_filename)
embeddings = get_word2vec(vector_word_npz)
############################################################
class data_loader():
def __init__(self):
self.train_x, self.train_y = process(train_filename,
wordid,
cat_to_id,
max_length=300)
self.test_x, self.test_y = process(test_filename,
wordid,
cat_to_id,
val_x, val_y = process_predict(pm.val_filename, wordid, cat_to_id, max_length=pm.seq_length)
batch_val = batch_iter_predict(val_x, val_y, batch_size=64)
for x_batch, y_batch in batch_val:
real_seq_len = seq_length(x_batch)
feed_dict = model.feed_data(x_batch, y_batch, real_seq_len, 1.0)
pre_lab = session.run(model.predict, feed_dict=feed_dict)
pre_label.extend(pre_lab)
return pre_label
if __name__ == '__main__':
start = time.time()
pm = pm
categories, cat_to_id = read_category()
wordid = get_wordid(pm.vocab_filename)
pm.vocab_size = len(wordid)
pm.pre_training = get_word2vec(pm.vector_word_npz)
model = Lstm_CNN()
pre_label = val()
contents_list = list(zip(pre_label, [num for item in pm.contents_finally for num in item]))
# 在当前目录下创建文件
i, j = 1, 1
finally_data_dir = './data/Finally_Data'
if os.path.exists(os.path.join(finally_data_dir, 'No_Value.txt')):
os.remove(os.path.join(finally_data_dir, 'No_Value.txt'))
if os.path.exists(os.path.join(finally_data_dir, 'Value.txt')):
os.remove(os.path.join(finally_data_dir, 'Value.txt'))
if not os.path.exists(finally_data_dir):
test_loss, test_accuracy = model.test(session, x_test, y_test)
print('global_step:', global_step, 'train_loss:', train_loss, 'train_accuracy:', train_accuracy,
'test_loss:', test_loss, 'test_accuracy:', test_accuracy)
summary = tf.Summary(value=[tf.Summary.Value(tag="accuracy", simple_value=test_accuracy)])
writer_test.add_summary(summary, global_step)
summary = tf.Summary(value=[tf.Summary.Value(tag="loss", simple_value=test_loss)])
writer_test.add_summary(summary, global_step)
if global_step % num_batchs == 0:
print('Saving Model...')
saver.save(session, save_path, global_step=global_step)
pm.learning_rate *= pm.lr_decay
if __name__ == '__main__':
# pm = pm
start = time.time()
filenames = [pm.train_filename, pm.test_filename, pm.val_filename]
categories, cat_to_id = read_category() # 给10个类别编号
wordid = get_wordid(pm.vocab_filename) # 给10000个常用词词典单词编号
pm.vocab_size = len(wordid)
pm.pre_training = get_word2vec(pm.vector_word_npz)
model = Lstm_CNN()
train()
stop = time.time()
print("程序运行时间： ", (stop - start))