* 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

* 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

* fix multi-card multi-process bug

refine code

* Use iterable dataloader for image classification

* Support imagenet dataset with plain folder structure

* Add dali section to README.md

* Improve docs

* Remove some flags from docs

* Add dali reader for image classification

* Fix elapsed time calculation

* add profiler and max_iter for yolov3 model

* modify image_classification train

* update train.py

* modify args position

* move release 1.6 code to dev for tagging

* syn merged pr

* fix_infer_bug_release_some_model

* add_multiprocess_train_to_image_classification

Add EfficientNet

* remove use_inplace=True and add save_step params

* remove useless params when using global pooling

* Refine Image classification

This reverts commit 5e41760f.

* fix gpu detection issue on windows
1. use the fluid.core.get_cuda_device_count() to decide the GPU number
2. ignore the path missing issue

* Transfer resnext50 multi-process multi-gpu changes from benchmark to models
Signed-off-by: Nzhaoyuchen <zhaoyuchen01@baidu.com>

* refine code
Signed-off-by: Nzhaoyuchen <zhaoyuchen01@baidu.com>

* refine readme and fix train bug

* Add readme and disable memopt

memopt may affect GC result

test=develop
Signed-off-by: Nzhaoyuchen <zhaoyuchen01@baidu.com>

* Disable memopt as GC opend

test=develop
Signed-off-by: Nzhaoyuchen <zhaoyuchen01@baidu.com>

* set lr.persistable=True

* comment fuse flag

This changing can improve performance

test=develop

* fix padding bug and add resize_short_size args

* fix typo

* pass args:settings in eval and infer

* update image classification pretrained models

* update softmax

* Add mixup and distill and fix bugs

* recover test pyreader

* add error tip

* update README_CN

* fix distill bug

* add reader_cv2.py

* update run.sh

* fix docs

* fix typo

* polish train.py

* fix typo

* remove commented code

* fix typo

* fix_typo

* recover scale_loss

* fix typo

* fix bug

* fix test_program bug

* update ngraph

* fix typo