I'm writing a test model using keras, where I want do some mathematics depends on numerical values of the output of a layer and its the derivatives.

I'm using tensorflow backend.
I use K.function in order to get the values of the outputs of the Lambda layer and derivative layers. However I got some weird err if I choose the function in the Lambda layer as power function, e.g. x**2. If I change x**2 to sin(x), it works fine.

import numpy as np
from keras.models import Model
from keras.layers import Input, Layer, Lambda
from keras import backend as K

x = Input(shape=(1,))

# the Lambda layer 
c = Lambda(lambda x: x**2)(x) # this will causs err
#c = Lambda(lambda x: K.sin(x))(x) # but this works fine


class dc_layer(Layer):

 def __init__(self,*args,**kwargs):
 self.is_placeholder = True
 super(dc_layer, self).__init__(*args,**kwargs)

 def call(self,inputs):
 x = inputs[0]
 c0 = inputs[1]
 c1 = K.gradients(c0,x)
 return c1

# the derivatives of the lambda layer 
c1 = dc_layer()([x,c])
c2 = dc_layer()([x,c1])

Then I use backend.function to define a function in order to get layer outputs

# define a function to get the derivatives
get_layer_outputs = K.function([x],[c2])

x_data = np.linspace(0,1,6)
val = get_layer_outputs([x_data])[0]
print(val)

I got the following err message in jupyter notebook

InvalidArgumentError: data[0].shape = [1] does not start with indices[0].shape = [2]

which tracback to

---> 36 val = get_layer_outputs([x_data])[0]

but if I look at the c1 layer

# define a function to get the derivatives
get_layer_outputs = K.function([x],[c1])

x_data = np.linspace(0,1,6)
val = get_layer_outputs([x_data])[0]
print(val)

it works fine.

I guess it is some thing wrong when I use K.function. Any solutions/suggestions would be appreciated.

======================================================

Additional question:

Even if I try a very simple code, I got err when use K.function, as follows

x = Input(shape=(1,))
h = Dense(10,activation='sigmoid')(x)
c = Dense(1)(h)

get_layer_outputs = K.function([x],[c])

x_data = np.linspace(0,1,6)
val = get_layer_outputs([x_data])[0]
print(val)

I got

InvalidArgumentError: In[0] is not a matrix
 [[Node: dense_24/MatMul = MatMul[T=DT_FLOAT, transpose_a=false, transpose_b=false, _device="/job:localhost/replica:0/task:0/device:CPU:0"](_arg_input_19_0_0, dense_24/kernel/read)]]

Now I'm really confused on how to use K.function properly. Please help if you have any idea. Thanks!

For me this works - your x_data vector was 0-Dimensional:

import numpy as np

from keras.models import Model
from keras.layers import Input, Layer, Lambda, Dense
from keras import backend as K

x = Input(shape=(1,))

# the Lambda layer 
c = Lambda(lambda x: x**2)(x) # this will causs err
#c = Lambda(lambda x: K.sin(x))(x) # but this works fine


class dc_layer(Layer):

 def __init__(self,*args,**kwargs):
 self.is_placeholder = True
 super(dc_layer, self).__init__(*args,**kwargs)

 def call(self,inputs):
 x = inputs[0]
 c0 = inputs[1]
 c1 = K.gradients(c0,x)
 return c1

# the derivatives of the lambda layer 
c1 = dc_layer()([x,c]) # in Keras 2.0.2 need to unpack results, Keras 2.2.4 seems fine.
c2 = dc_layer()([x,c1])

# define a function to get the derivatives
get_layer_outputs = K.function([x],[c2])

x_data = np.linspace(0,1,6)[:,None] # ensure vector is 1D, not 0D
val = get_layer_outputs([x_data])[0]
print(val)

output:

[[2.]
 [2.]
 [2.]
 [2.]
 [2.]
 [2.]]