Custom Defined classification Layer

Hi, I have defined a classification layer using cross entropy loss function and replaced the last classification layer in 'layers' with this one. Though this layer has cleared the validity test for CNN layer but when I try to train the RCNN network then it gives me error of 'no classification layer' in the network. Layers also get defined but it is not training the network. Following are the layers of simplest network I am trying to use right now:

layers =

6x1 Layer array with layers:

1 '' Image Input 32x32x3 images with 'zerocenter' normalization

2 '' Convolution 96 11x11 convolutions with stride [1 1] and padding [0 0 0 0]

3 '' ReLU ReLU

4 '' Fully Connected 2 fully connected layer

5 '' Softmax softmax

6 '' Classification Output cross entropy based classification

It is showing following error when it tries to train the network. "The network must contain one classification layer."