smaller binary proto
avoid untrackable kernel

[Speed]Avoid init_nccl for every steps.

Add program cache for executor.py

update cluster_train page

* fix is_test issue

* add paddle enforce

* fix bug

* add new func

* small fix

* address comments

Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into add-program-cache-for-executor

* change learning_rate_decay to learning_rate_scheduler
* optimize code
* change nn.cast to tensor.cast

Create new structure for V2 and fluid doc

Implement detection mAP evaluator wrapper and unify label format between SSD loss and mAP evaluator  (#8736)

* Implement mAP evalutor Python interface.

* Fix unit testing and uniy label format between SSD loss and mAP evalutor.

* Update doc.

* add write_new_layers in rst tree

PaddlePaddle.org requires the nav tree to be generated in order to parse the site map.  Since we removed theme in https://github.com/PaddlePaddle/Paddle/pull/8752, this also removed the site navigation menu.  This changelist adds back the site navigation menu. (#8760)

remove doc_theme

disable nccl in memopt fit a line unittest

Merge branch 'develop' of https://github.com/PaddlePaddle/Paddle into add-program-cache-for-executor

Add layers for save/load op

fix warning: statement is unreachable