从python3.6转换keras模型。到3.5

时间:2018-05-27 10:13:27

标签: python tensorflow keras raspberry-pi3 raspbian

我有一个使用python 3.6训练的keras模型,并且使用带有python 3.5的raspbian。

当您将使用python 3.6训练的模型(或至少我的模型)加载到python 3.6中时,您会遇到异常:

IndexError: tuple index out of range

问题是由于不同的原因我不能将训练平台改为3.5或RPi改为3.6,所以我必须转换model.h5。

有没有办法将h5转换为中间值,然后在其他平台中从中间转换为h5?

调用load_module时错误上升

问题是由于不同的原因我不能将训练平台改为3.5或RPi改为3.6,所以我必须转换de model.h5。

有没有办法将h5转换为中间值,然后在其他平台中从中间转换为h5?

load_model("model1527371035.h5")    
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "/usr/local/lib/python3.5/dist-packages/keras/models.py", line 270, in load_model
    model = model_from_config(model_config, custom_objects=custom_objects)
  File "/usr/local/lib/python3.5/dist-packages/keras/models.py", line 347, in model_from_config
    return layer_module.deserialize(config, custom_objects=custom_objects)
  File "/usr/local/lib/python3.5/dist-packages/keras/layers/__init__.py", line 55, in deserialize
    printable_module_name='layer')
  File "/usr/local/lib/python3.5/dist-packages/keras/utils/generic_utils.py", line 144, in deserialize_keras_object
    list(custom_objects.items())))
  File "/usr/local/lib/python3.5/dist-packages/keras/models.py", line 1412, in from_config
    model.add(layer)
  File "/usr/local/lib/python3.5/dist-packages/keras/models.py", line 497, in add
    layer(x)
  File "/usr/local/lib/python3.5/dist-packages/keras/engine/topology.py", line 619, in __call__
    output = self.call(inputs, **kwargs)
  File "/usr/local/lib/python3.5/dist-packages/keras/layers/core.py", line 685, in call
    return self.function(inputs, **arguments)
  File "<ipython-input-11-b85ceb3c6761>", line 64, in <lambda>
IndexError: tuple index out of range

模型如下:

model = Sequential()
model.add(Lambda(lambda x: x / 255.0 - 0.5, input_shape=(84, 84, 3)))
model.add(BatchNormalization())
model.add(Conv2D(36,(5,5), strides=(2,2), activation='relu'))
model.add(Dropout(dropout))
model.add(Conv2D(64,(3,3), activation='relu'))
model.add(Dropout(dropout))
model.add(Flatten())
model.add(Dropout(dropout))
model.add(Dense(40))
model.add(Dropout(dropout))
model.add(Dense(10))
model.add(Dense(6, activation='softmax'))
model.compile(loss='categorical_crossentropy', optimizer="adam",metrics=['mae', 'acc'])

1 个答案:

答案 0 :(得分:2)

是的,通过正在进行的issue查看评论,当前的解决方法看起来就像保存和加载权重一样:

model.save_weights(filename)
# you have to rebuild model again
model.load_weights(filename)

这种情况下保存的文件不包含架构,您每次都必须重建它。这不是昂贵的,所以它应该不是问题。

编辑:这可能只影响Lambda图层,可能是一个简单的custom layer可以避免此问题:

class MyLayer(Layer):
   def call(self, x):
     return x / 255.0 - 0.5