Conv GRU

[carter Frank]

Hi,

    I want to add a convolutional GRU layer following FCN (first part of my network) as a recurrent refinement step, but I only find GRU layer in lasagne, in which dot product is used for input/hidden state transformation rather than convolution.  Is there any ConvGRU layer code exsited? Can you give me some help on this if I need to rewrite GRU layer? Thanks!