Django+Keras开发中遇到的问题

1.django多次加载keras_model导致错误

           将加载过程放到{path_to}\__init__.py(自己创建的专门处理keras相关事务的包)中，在web启动时加载model

print('>>>loading model ...')

with open('model.json', 'r')as f:

    json_string = json.load(f)

model = model_from_json(json_string)

model.load_weights('model.h5')

print('>>>load done')

model.summary()

2.多次执行predict报错，Tensor Tensor(“activation_5/Softmax:0”, shape=(?, 4), dtype=float32) is not an element of this graph

           由于django的线程机制，导致多次predict时tf错误，解决graph = tf.get_default_graph()

参考：https://github.com/keras-team/keras/issues/2397

参考：https://qiita.com/itisyuu/items/7c9d7ff43b3936704918

#加载model之后

graph = tf.get_default_graph()

def category_class(x):

    global model

    global graph

    with graph.as_default():

        prediction = np.argmax(

            model.predict(x)

        )

    return prediction

3.print（）时报错，OSError: raw write() returned invalid length 134 (should have been between 0 and 67)

这个错误不时出现，出现的原因可能和win10系统有关

import win_unicode_console

#for raw write()

win_unicode_console.enable()

参考：https://www.cnblogs.com/yanjj/p/8275995.html