Add layers for save/load op

fix warning: statement is unreachable

Refine document of build_and_install 

* Refine the doc in detection_output API.

* Refine the doc.

Add guideline for the doc of cmd parameters

v2 API overview refactor doc

fix only shared variables could be declared as static in the device code

Fix vgg16 fc layer size: 4096 rather than 512

Fix fluid distribute build

Add CPU time and MemCopy to the timeline.

make memory optimization module compatible with parallel_do

Add sequence error output to edit distance evaluator

Python Unit Tests for CSP

* Simple Channel Send and Receive test
* Daisy Chain test with 100 channels/Go ops

Fix rename de-ref bug

First timeline version

* add the type of tuple

* add lod_tensor to tuple

Add global_step in nn.py