* add amp for ResNet200_vd

* change use_fp16 to use_amp

* fix script format

* change batch_size

* add amp support for model SE_ResNet50_vd

* optimize code, add model ResNet50 and SE_ResNet50_vd to list

* merge develop

* add pure fp16 training.

* update the usage of cast_parameters_to_fp16.

* use fp32 to initilize parameter values and then cast to fp16.

* enable multi_precision when training with fp16.

* add static loss scaling.

* add the rescale_grad function in the pure fp16 training.

* use cast to complete the pure fp16 training.

* use sum loss for fp16 trainging and mean loss for fp32 training.

* refine codes according reviewers' suggestions

* Create inception_v3.py

* Update __init__.py

* Create InceptionV3.sh

* Revert "add fuse_bn_add_act_ops args (#4864)"

* add fuse_bn_add_act_ops args

* fix ptb_dy time print for benchmark, test=develop

* use yiqun(Xreki)'s PR, test=develop

* add empty line, test=develop

* add empty line, test=develop

* fix ptb_dy time print for benchmark, test=develop

* refine resnet benchmard print, test=develop

test=develop

* add new cls link
* fix content
* fix cv readme
* fix moodels readme

* Update README.md (#4267)

* test=develop (#4269)

* 3d use new api (#4275)

* PointNet++ and PointRCNN use new API

* Update Readme of Dygraph BERT (#4277)

Fix some typos.

* Update run_classifier_multi_gpu.sh (#4279)

remove the CUDA_VISIBLE_DEVICES

* Update README.md (#4280)

* 17 update api (#4294)

* update1.7 save/load & fluid.data

* update datafeed to dataloader

* Update resnet_acnet.py (#4297)

Bias attr of square conv should be "False" rather than None during training mode.

* test=develop
Co-authored-by: NKaipeng Deng <dengkaipeng@baidu.com>
Co-authored-by: Nzhang wenhui <frankwhzhang@126.com>
Co-authored-by: Nparap1uie-s <parap1uie-s@users.noreply.github.com>

* fix multi-card multi-process bug

* add the fp16 support for ResNet50

* add docs for AMP.

* support out json in eval and add exclude fc params in finetune

* fix cpu bug