def copy_move(src_path, dest_path):
'''
:param src_path 源文件
:param dest_path 需要剪切到的路径
将文件剪切到另一个文件夹
'''
try:
check_utils.check_path(dest_path)
shutil.move(src_path, dest_path)
except:
pass
def run(model_name):
'''
根据模型的不同选择使用的模型代码
:param model: 模型类型 bilstm_crf bert_bilstm_crf lstm_crf wordvec_bilstm_crf
:return:
'''
if model_name == 'bilstm_crf':
x_train, y_train, x_test, y_test, vocab_length, labels_to_ix_length = process_data_for_keras.process_data(
embeding="bilstm")
model = keras_BILSTM_CEF.build_embedding_bilstm2_crf_model(
vocab_length, labels_to_ix_length, 0)
save_path = normal_param.save_path_bilstm
elif model_name == "rnn_crf":
x_train, y_train, x_test, y_test, vocab_length, labels_to_ix_length = process_data_for_keras.process_data(
embeding="rnn")
model = keras_RNN_CRF.build_embedding_lstm2_crf_model(
vocab_length, labels_to_ix_length, 0)
save_path = normal_param.save_path_gru
elif model_name == 'bert_bilstm_crf':
x_train, y_train, x_test, y_test, vocab_length, labels_to_ix_length = process_data_for_keras.process_data(
embeding="bert")
model = keras_Bert_bilstm_crf.build_bilstm_crf_model(
labels_to_ix_length)
save_path = normal_param.save_path_bert_bilstm
elif model_name == 'lstm_crf':
x_train, y_train, x_test, y_test, vocab_length, labels_to_ix_length = process_data_for_keras.process_data(
embeding="lstm")
model = keras_LSTM_CRF.build_embedding_lstm2_crf_model(
vocab_length, labels_to_ix_length, 0)
save_path = normal_param.save_path_lstm
else:
embeddings_matrix, vocab = process_data_for_keras.txtpad_use_word2vec()
x_train, y_train, x_test, y_test, vocab_length, labels_to_ix_length = process_data_for_keras.process_data(
embeding="wordvec", vocab2=vocab)
model = keras_word2vec_bilstm_crf.build_embedding_bilstm2_crf_model(
labels_to_ix_length, embeddings_matrix, normal_param.max_length)
save_path = normal_param.save_path_wordVEC_bilstm
if check_utils.check_path(save_path):
model.load_weights(save_path)
model.fit(x_train,
y_train,
batch_size=18,
epochs=20,
validation_data=(x_test, y_test),
shuffle=False,
validation_split=0.2,
verbose=1)
keras_BILSTM_CEF.save_embedding_bilstm2_crf_model(model, save_path)
def read_path(head_path):
'''
根据头路径获取不同路径下对应的txt和label文件路径信息
:param head_path: 头路径的文件信息
:return: label_path, content_path
'''
content_head_path = os.path.join(head_path, "txt")
label_head_path = os.path.join(head_path, "label")
names = os.listdir(content_head_path)
content_paths = []
label_paths = []
for name in names:
content_path = os.path.join(content_head_path, name)
label_path = os.path.join(label_head_path, name)
if not check_utils.check_path(label_path):
continue
content_paths.append(content_path)
label_paths.append(label_path)
return label_paths, content_paths