What the connection between pool 1 and conv2 in caffe implementation

[Xiong Yunyang]

Hi,

I tried to do the inference by the trained model using my own network and I used the same structure as Lenet.
But when I use the same input and the same model, the difference comes at the conv2.
I used all the input in pool 1, so I was wondering if the connection between pool 1 and conv 2 is specified in caffe implementation?
If it is specified, then where can I get the parameter to specify the partial connection between pool 1 and conv 2?
I looked the caffe.pro and I did not find it.
Can someone help me out?