* check md5 before load weights.

* add TX2 cpp build document and benchmark

* turn WITH_MKL=ON

* add tensorrt version

* set WITH_GPU=OFF for CI

* keep same as TX2_build

* enhance voc to coco docs

* fix voc annotation float type

* add solov2_mobilenetv3

* update link on README

* use relative path

* add benchmark on win10 cpp finer, fix bug of _mkdir on windows VS2019 Release bulig

* win cpp use Predictor to inference

* 测试环境描述更加详细

* add enable_static()

* follow comments

* remove blank line

* modify cpp output path

* fix compile error

* modify docs for windows

* add android demo

* support infer for SOLOv2

* add ppyolo mobilenetv3

* unify log_smooth_windows and log_iter

test=develop

* remove duplicate log_iter

* remain mobile yml, since they are soft-link

test=develop

* share train_batch_size from reader to model

* update floor_divide

* add makedirs for dumping config

test=develop

* add solov2 model

* fix train batch size

* add solov2_r101_vd_fpn_3x

* fix batch size and update modelzoo

* refactor code of solov2

* fix deploy/python

* clean useless log in DataLoader

* fix infer set_sample_list_generator

* check md5 after weights download

* move train_batch_size sync from load_config to merge_config

test=develop

* warning when YOLOv3Loss.batch_size setting

* expand ratio=2.0 & add ppyolo_2x_bs12.yml

* add batch_size in YOLOv3Loss for backward compatable.