How does chainer operater other image in one batch?

[machanic]

Because the chainer's Variable has omit Linker's first batch dimension.
According to https://docs.chainer.org/en/stable/tutorial/basic.html#forward-backward-computation
The document said

Most functions and links only accept mini-batch input, where the first dimension of the input array is considered as the batch dimension. In the above Linear link case, input must have shape of (N, 3), where N is the mini-batch size.

If I want to get/operate the other image in this batch, How to write the code?

[amasky]

In this repository, I used concatenations for "multi-scale glimpse" (see ram.py). Each batch has multiple glimpses of an input image and I used chainer.functions.concat in this line
rho = F.concat((rho, patch), axis=1)
"axis=1" means the 2nd axis for channel dimension