Add LRN efficient GPU implement

Add half precision float16 data type

* init nmt

* encoder ready

* only generation implementation

waiting for dynamic rnn ready to train

* init python

* remove decoder temporary

* clean

* clean

Fix nce op warning about comparison of integers of different signs

fix int overflow

Refine code（operator++）

* Add DataFeeder

A v2 API like data feeder for book demos.
We can feed data directly from reader.

* Fix CI

* Add an unittest for while/rnn op forward

* Add unittest for raw while op backward

* Fix CI

* Complete Dynamic RNN

This pull request adds "Build using Docker" documentation to the "Getting Started > Install and Build" menu on PaddlePaddle.org

* Using latex fraction syntax in sigmoid and logsigmoid op

* Fixing the rendering of the latex equations in adamax operator

* add crf_decoding layer

* fix some typo

* init trunk_evaluator

* add trunk_evaluator layer

* update chunk_eval_op and test, change int32 to int64

* fix a numeric problem

* change layers.trunk_evaluator to layers.trunk_eval

* fix typo

* add precision_val

Remove the CUDA stream synchronization between each operator.

* "add init seed"

* "fix compile error"

* "add program level seed setting"

* "fixed based on comments"

Fix 404 link in dist arch doc

* add crf_decoding layer

* fix some typo

* fix test_crf_decoding_op

Refine and speedup momentum operator.

[Profiling]  Update elementwise op

fix typo in pip_install_cn/en.rst

Update write docs cn