caffe - How to deploy the model in Deep Lab V2 -

i trying train , test deep cnn model segmentation. using model deep lab v2 based on caffe. have gone through tutorial of script of run_pascal.sh train model. however, if want train own segmentation model , deploy it, how should write deploy.prototxt file? can provide example of deploy file according latest v2 version?

i used train model using cmd:

caffe.bin train --solver=solver.prototxt

my solver.prototxt follows:

train_net: "train.prototxt"

lr_policy: "poly"

power: 0.9

base_lr: 5e-2

average_loss: 1

display: 20

max_iter: 10000

weight_decay: 0.0005

snapshot: 10000

snapshot_prefix: "/path/to/save/snapshots"

type: "adadelta"

solver_mode: gpu

train.prototxt model used train

make sure include paths wherever necessary