Created by: pengwangucla

I tested using cost layer, it seems the network parsing is stopped after meeting the cost layer ?

First I test copy the and add them together:

    cost_rs = pd.mse_cost(input=input_rs, label=label_rs)
    cost = pd.addto(input=[cost_rs, cost_rs])

cost_rs is [2] and cost is also [2], there is no effect of paddle addto layer.

I also try to use mixed_layer add a activation later, for example:

    cost_rs = pd.mse_cost(input=input_rs, label=label_rs)
   sqrt_l2_cost = math_op(input=cost_rs, act=pd.activation.Sqrt())

sqrt_l2_cost is still the same as cost, no operation added later.

Is there anyway to fix this ? @reyoung