用keras调用load_model时报错ValueError: Unknown Layer:LayerName_load_metrics(rouge

出现该错误是因为要保存的model中包含了自定义的层（Custom Layer），导致加载模型的时候无法解析该Layer
解决该问题的方法是在load_model函数中添加custom_objects参数，该参数接受一个字典，键值为自定义的层：

ner_model = load_model(model_path + 'ner_crf_bilstm_model.h5', custom_objects={"CRF": CRF,"crf_loss": crf_loss, "crf_accuracy": crf_accuracy})
1

或者你可以使用对象的方法

from keras.utils import CustomObjectScope
with CustomObjectScope({'AttentionLayer': AttentionLayer}):
    model = load_model('my_model.h5')
1
2
3

Custom objects handling works the same way for load_model, model_from_json, model_from_yaml:

from keras.models import model_from_json
model = model_from_json(json_string, custom_objects={'AttentionLayer': AttentionLayer})
1
2

如果有两个以上的自定义网络层的解决方法如下：

custom_ob = {'AttLayer1': custom_layer.Attention,'AttLayer2': custom_layer.Attention}
model = load_model('lstm.h5', custom_objects=custom_ob)
或者
with CustomObjectScope(custom_ob):
model = load_model('lstm.h5')
1
2
3
4
5