* fix py3

To avoid the error when use_program_cache is True in Transformer infe…rence.

To avoid the data decode error.

* Unified object detection framework based on PaddlePaddle.
* Include algo: Faster, Mask, FPN, Cascade, RetinaNet, Yolo v3, SSD.

* fix grad and scipy version

* fix some bug of gan

fix op init

* fix batch generate bug in eval.py.

add AttGAN and STGAN

* 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

* move benchmark changing of cycle gan to models/develop, test=develop

* add flag exist check, test=develop

This changing can improve performance

test=develop

* add ce for dygraph mnist

* add ce for dygraph mnist

* del mnist_dygraph.py

* change mnist_dygraph to train

* fix print style

* add multi process reader

* use distributed_batch_reader

* update

* fix

* Enable the use of python multi-processes to read data.

* Add the setting of default value of some paddle flags, such as GC.

* update transformer

* Update the optimization of PaddingRNN model in benchmark repo.

* Move the configuration of rnn model to a serapate file.

* Fix the error of forgetting feeding learning_rate when use py_reader.

* Add a argument and some check.

* Fix the feeding problem and remove some build options.

* Add some comments and refine the format.

* Fix an error when define the inference program.

* Remove the setting of use_experimental_executor, and simplfy some codes.

* Refine the setting of device_count.

load_persistables function returns a tuple which include parameters and optimizer

* fix padding bug and add resize_short_size args

* fix typo

* pass args:settings in eval and infer

* Fix SSD file path.

* Move object_detection to ssd.

1.Modify the checkpoint save path
2. Modifiy the img save paht
3. Synchronous modification for Readme
4. update generated result img