what outputs must be for second stage model in predict function?

[deadfrog333]

Im a newbie in the theano framework and in machine learning. so Im not fully understand the problem Im facing. I hope someone will be able to point me to the right direction.

the demo file has this line, which creates predict function: predict = theano.function(inputs=[x], outputs=net.y) where outputs are net.y from model stage 1.

I trained model stage 2 on a smaller dataset. Model loads to the memory:

Number of parameters is 17926152
Number of parameters is 62019
Number of parameters with stage1 params is 17988171

However when I try to created predict function it gives the following error:

theano.gof.fg.MissingInputError: Input 0 of the graph (indices start from 0), used to compute Subtensor{int64::}(p, Constant{0}), was not provided and not given a value. Use the Theano flag exception_verbosity='high', for more information on this error. tracing to variable: p = T.matrix('p')

So obviously the empty tensor I created before p = T.matrix('p') (by looking at main2.py script) is not working. And it is not very clear what p should be in that case. and what outputs for second stage model should be? The softmax layer maybe? I will really appreciate any help.