Merge pull request !4308 from yuchaojie/fix_resnet_gpu_distribute_init

Merge pull request !4303 from liuxiao78/master

Merge pull request !4307 from YuJianfeng/internal_output

Merge pull request !4311 from anthonyaje/monitor_fix

Merge pull request !3771 from cathwong/ckw_c_api_textfile_nosampler

Merge pull request !4081 from ava/visionOps

Merge pull request !4312 from nhussain/fix_lite_builds

Merge pull request !4245 from nhussain/solarize

Merge pull request !3978 from Peilin/master

Merge pull request !4301 from liuzhongkai/leaky_relu

Merge pull request !4297 from guozhijian/code_sync_incubator_f3c32baf_to_master_fcfc75a3_0811

Merge pull request !4061 from LiHongzhang/fix_summary

Merge pull request !4292 from liubuyu/bug_fix2

Merge pull request !4306 from mengchunyang/master

Merge pull request !4276 from hangq/master

Merge pull request !4287 from yankai10/merge

add opencl prelu kernel

Merge pull request !4261 from yangruoqi713/test_dw

- handle collection for multiple trains
- how many tensors to collect when sunk
- change loglevel for get_learning_rate
- update calculation of `max_file_size`
- fix how collect_tensor_freq counting

Merge pull request !4289 from xiefangqi/fix_pad_validate

Merge pull request !4258 from leopz/fix_doc

Merge pull request !4285 from yanghaitao/yht_fix_doc

Merge pull request !4268 from liubuyu/master

Merge pull request !4270 from songhonglei413/roi

Merge pull request !4296 from chenjianping/lite_dev2

Merge pull request !4275 from yuchaojie/add_allreduce_group_for_resnet_gpu

Merge pull request !4263 from chenjianping/lite_dev