* MKL-DNN Data transform doc design

- Cosmetic

- Changes

- CPU -> Paddle conversion

- Finished data transform draft

- Cosmetic fixes

* - Fixes after internal review

* Add the description of the quantization-aware training model format.

* add the abstract.

* remove v2

* remove mobile

* fix_en

* Update distributed_architecture.md

* Start over

* Added basic docs
Added .travis.yml
Added scripts to build documentation on the Travis.

* Disable several deploy script commend for testing.

* fixed the deploy_docs.sh script.

* Update travis.yml

* Renamed docs to doc

* update .gitignore.

* Delete .gitignore

* Update .travis.yml

* Update .travis.yml

* Update deploy_docs.sh

* Update .travis.yml

* Develop doc (#1)

* Add paddle submodule

* Update the build script.

* Update script

* Use gen_doc_lib instead.

* Move files around

* cache external

* Update submodule

* try to cache batch 1

* add test code

* Update Paddle submodule

* Update submodule

* update script to print more

* update python path

* test

* test

* test

* clean up the code

* Update Script (#2)

* add new file

* Develop doc (#3)

* Add the rest of the submodules into the system

* Provide first symlink fit a line

* Update the rest of book to symlinks

* add models submodule

* Add link for models

* Update deploy_docs