我需要在Keras（1.1）中创建具有可训练权重（与输入相同的形状）的自定义图层。我尝试通过随机值初始化权重。 有我的'mylayer.py'文件：
from keras import backend as K
from keras.engine.topology import Layer
import numpy as np
from numpy import random
class MyLayer(Layer):
def __init__(self,**kwargs):
super(MyLayer, self).__init__(**kwargs)
def build(self, input_shape):
# Create a trainable weight variable for this layer.
self.W_init = np.random(input_shape)
self.W = K.variable(self.W_init, name="W")
self.trainable_weights = [self.W]
super(MyLayer, self).build(input_shape) # Be sure to call this somewhere!
def call(self, x):
num, n, m = x.shape
res=np.empty(num,1,1)
for i in range(num):
res[i,0,0]=K.dot(x[i,:,:], self.W[i,:,:])
return res
def compute_output_shape(self, input_shape):
return (input_shape[0], 1,1)
但是当我尝试使用它时：
from mylayer import *
def get_my_model_2():
model = Sequential([
Lambda(norm_input, input_shape=(1,28,28)),
Convolution2D(32,3,3, activation='relu'),
BatchNormalization(axis=1),
Convolution2D(32,3,3, activation='relu'),
MaxPooling2D(),
BatchNormalization(axis=1),
Convolution2D(64,3,3, activation='relu'),
BatchNormalization(axis=1),
Convolution2D(64,3,3, activation='relu'),
MaxPooling2D(),
MyLayer(input_shape=(64,4,4)), # MY LAYER
Flatten(),
BatchNormalization(),
Dense(512, activation='relu'),
BatchNormalization(),
Dropout(0.5),
Dense(10, activation='softmax')
])
model.compile(Adam(), loss='categorical_crossentropy', metrics=['accuracy'])
return model
model=get_my_model_2()
我有错误“'模块'对象不可调用”：
>
> /home/universal/anaconda3/envs/practicecourse2/mylayer.py in
> build(self, input_shape)
> 15 # initializer='uniform',
> 16 # trainable=True)
> ---> 17 self.W_init = np.random(input_shape)
> 18 self.W = K.variable(self.W_init, name="W")
> 19 self.trainable_weights = [self.W]
>
> TypeError: 'module' object is not callable
怎么了？
提前谢谢
添加了：
解决“随机”问题后，我又出现了另一个错误：
---------------------------------------------------------------------------
TypeError Traceback (most recent call last)
<ipython-input-31-8e5ded840273> in <module>()
----> 1 model=get_my_model_2()
2 model.summary()
<ipython-input-30-09ee1207017c> in get_my_model_2()
20 BatchNormalization(),
21 Dropout(0.5),
---> 22 Dense(10, activation='softmax')
23 ])
24 model.compile(Adam(), loss='categorical_crossentropy', metrics=['accuracy'])
/home/universal/anaconda3/envs/practicecourse2/lib/python2.7/site-packages/keras/models.pyc in __init__(self, layers, name)
253
254 for layer in layers:
--> 255 self.add(layer)
256
257 def add(self, layer):
/home/universal/anaconda3/envs/practicecourse2/lib/python2.7/site-packages/keras/models.pyc in add(self, layer)
310 output_shapes=[self.outputs[0]._keras_shape])
311 else:
--> 312 output_tensor = layer(self.outputs[0])
313 if type(output_tensor) is list:
314 raise Exception('All layers in a Sequential model '
/home/universal/anaconda3/envs/practicecourse2/lib/python2.7/site-packages/keras/engine/topology.pyc in __call__(self, x, mask)
512 if inbound_layers:
513 # this will call layer.build() if necessary
--> 514 self.add_inbound_node(inbound_layers, node_indices, tensor_indices)
515 input_added = True
516
/home/universal/anaconda3/envs/practicecourse2/lib/python2.7/site-packages/keras/engine/topology.pyc in add_inbound_node(self, inbound_layers, node_indices, tensor_indices)
570 # creating the node automatically updates self.inbound_nodes
571 # as well as outbound_nodes on inbound layers.
--> 572 Node.create_node(self, inbound_layers, node_indices, tensor_indices)
573
574 def get_output_shape_for(self, input_shape):
/home/universal/anaconda3/envs/practicecourse2/lib/python2.7/site-packages/keras/engine/topology.pyc in create_node(cls, outbound_layer, inbound_layers, node_indices, tensor_indices)
147
148 if len(input_tensors) == 1:
--> 149 output_tensors = to_list(outbound_layer.call(input_tensors[0], mask=input_masks[0]))
150 output_masks = to_list(outbound_layer.compute_mask(input_tensors[0], input_masks[0]))
151 # TODO: try to auto-infer shape if exception is raised by get_output_shape_for
TypeError: call() got an unexpected keyword argument 'mask'
在我看来，我的自定义图层仍然存在错误