keras 解决加载lstm+crf模型出错的问题
时间：2021-05-22 08:43:20|栏目：Python代码|点击： 次
错误展示
new_model = load_model(“model.h5”)
报错：
1、keras load_model valueError: Unknown Layer :CRF
2、keras load_model valueError: Unknown loss function:crf_loss
错误修改
1、load_model修改源码：custom_objects = None 改为 def load_model(filepath, custom_objects, compile=True)：
2、new_model = load_model(“model.h5”,custom_objects={‘CRF': CRF,‘crf_loss': crf_loss,‘crf_viterbi_accuracy': crf_viterbi_accuracy}
以上修改后，即可运行。
补充知识：用keras搭建bilstm crf
使用 https://github.com/keras-team/keras-contrib实现的crf layer，
安装 keras-contrib
pip install git+https://www.github.com/keras-team/keras-contrib.git
Code Example：
# coding: utf-8
from keras.models import Sequential
from keras.layers import Embedding
from keras.layers import LSTM
from keras.layers import Bidirectional
from keras.layers import Dense
from keras.layers import TimeDistributed
from keras.layers import Dropout
from keras_contrib.layers.crf import CRF
from keras_contrib.utils import save_load_utils
VOCAB_SIZE = 2500
EMBEDDING_OUT_DIM = 128
TIME_STAMPS = 100
HIDDEN_UNITS = 200
DROPOUT_RATE = 0.3
NUM_CLASS = 5
def build_embedding_bilstm2_crf_model():
"""
带embedding的双向LSTM + crf
"""
model = Sequential()
model.add(Embedding(VOCAB_SIZE, output_dim=EMBEDDING_OUT_DIM, input_length=TIME_STAMPS))
model.add(Bidirectional(LSTM(HIDDEN_UNITS, return_sequences=True)))
model.add(Dropout(DROPOUT_RATE))
model.add(Bidirectional(LSTM(HIDDEN_UNITS, return_sequences=True)))
model.add(Dropout(DROPOUT_RATE))
model.add(TimeDistributed(Dense(NUM_CLASS)))
crf_layer = CRF(NUM_CLASS)
model.add(crf_layer)
model.compile('rmsprop', loss=crf_layer.loss_function, metrics=[crf_layer.accuracy])
return model
def save_embedding_bilstm2_crf_model(model, filename):
save_load_utils.save_all_weights(model,filename)
def load_embedding_bilstm2_crf_model(filename):
model = build_embedding_bilstm2_crf_model()
save_load_utils.load_all_weights(model, filename)
return model
if __name__ == '__main__':
model = build_embedding_bilstm2_crf_model()
注意：
如果执行build模型报错，则很可能是keras版本的问题。在keras-contrib==2.0.8且keras==2.0.8时，上面代码不会报错。
