When create one symbol, how to set multi-input data?

[tornadomeet]

Hello~
In some scenarios, one layer(the middle layer, not the input data layer) needs more than one input layer, and these inputs don't have the same formula mode, just like in caffe (one layer may have more bottom layers) , so how to set this multi-input data when buliding one layer in Symbolic?
assume that i have written a NewLayer in python side refer to new_op_howto, and the NewLayer has two inputs, how to set this in symbol? Does mxnet currently support this style ?or how i should do this?
ths.

[winstywang]

Check the implementation of Concat Op.

[tornadomeet]

@winstywang , the inputs don't have the same formula mode, so we cannot concat it. I'm writing the crop layer in python, just as(but not the same ) BVLC/caffe#1639

[piiswrong]

Just declare any input and weights in list_arguments ['input_1', 'input_2',...]
and when creating the symbol
new_layer = NewLayer()
new_layer(input_1=x1, input_2=x2...)

[tornadomeet]

@winstywang , Yeah, i got it, when more inputs, just as concat = mx.symbol.Concat(*[conv, pool])
thanks a lot!

[tornadomeet]

@piiswrong , thanks, i do it according to the your method, it works!
maybe you can consider put this in the guide new_op_howto~