* Update mnist_dygraph.py

fix bug

* add muti card support for se_resnext

* add some description to readme.md

some of the global config values are reference type, e.g., objects, lists or
dicts, if the created module instances modify those values (see FPN and
spatial_scale), global config will reflect these changes, and instances of the
same class created later will inherit the changed values

* refine bbox_normalize in infer.py

* add is_bbox_normalize

* rename _forward to build

* check callable

* fix_bn=1

* Support to set multiprocess in SSD model
* clean code

change default sample_pro  to ensure correctively running

use all data
network:
    support variable batch_size
    add some shape comment
infer:
    use pyreader

* fix word2vec Readme

* fix windows numpy int64 bug

* fix multiview Readme

更新文件readme，run_classifier.py, run_ernie_classifier.py，新增支持cuda_check部分代码

* fix_nan

* add check gpu for deeplabe3+, icnet, rcnn

* modify based on review

* modify based on review

* add cuda_check

* add cuda_check in emotion_detection

* check gpu for ppdet/yolov3

* add random shape descp in MODEL_ZOO

* refine doc

* not use core

* add cuda check to show errors when run on paddle compiled with cpu only

* change use_cuda to use_gpu in the error infomation to be consistent with code

* update checkgpu

* Add faster_rcnn_x101_vd_64x4d_fpn_1x.yml
* Release faster_rcnn_r101_vd_fpn_1x.tar, faster_rcnn_x101_vd_64x4d_fpn_1x.tar, mask_rcnn_r50_2x.tar

* add gpu check; test=develop

* add cuda check in train; test=develop

* add save_inference_model in infer.py

* format code

* add comment

* add save_inference_model doc

* refine doc

Update mnist_dygraph.py
fix bug
* add ce to se_resnext
* delete useless comments and fix unique_name bugs