Fix ElementwiseOpInferVarType in elementwise_op

Disable and fix tests on multi devices.

Add `op_role` into OpDesc.

* Expose bilinear operator into Python API

* delete unused lines

* delete unused lines

* fix typos

* fix name arg and typos

fix typo in workflow_of_capi.md

Fix cuda9 prod image

* add lod_tensor utility python module

* add lod_tensor test code

* add more lod tensor tests

* modify word2vec example code using new api

* add comment

multi-thread handlerequest

add checkpoint util class and implement

Add groups for conv transpose ops

* wip integrate benchmark scripts

* testing nlp models

* k8s script to start dist benchmark job

* update script

* done support all models

* add README.md

* update by comment

* clean up

* follow comments

expose random_data_generator

Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into feature/refine_parallel_executor

Add Inference doc for fluid

Add api docs for RMSProp and modelAverage

send use pinned memory

Add the demo of subgraph splitter

fix dist test

* Enhance reduce op for multi dims.

* Uncomment some unitest.

* Uncomment unitest.

* Remove unused code.

* Fix infershape and python wrapper.

* Add more examples.

* Fix l2_normalize.

* Fix normalization_wrapper.

* Polish code.
1. Rename unitest function.
2. Rename const variable.