* Feed/Fetch op just plain operator, not a OpWithKernel
* Do not register OpInfoMaker since Feed/Fetch will never be
  configured by users
* Feed/Fetch op has empty gradient
* Feed/Fetch op do not hard code `feed_variable`, `fetch_variable` as
  its input and output, make it as a plain Operator input/output

* fix gometalinter versioning

* stop gometalinter

* Expose Executor to Python

* Follow comments

* Final step of backward, return a map from param_name to grad

* Complete the final step of backward

Return the param_name to grad_info

* add target to Backward, generate var in block when call backward

* modify backward_test

* fix executor_test

* set var desc default type to LOD_TENSOR

* update backward_test

* insert loss in the top level of backward

* create grad vars for all blocks in current program

* optimize code

* update test_program.py

* only create var for newly create blocks when backward

* Add debug string for Python ProtoBuf

and Rename `Sync` to `Flush`

* Add check of ProtoBuf initialization

* Add grad_name_map

* Fix bug

* Fix bug

* Follow comments

* Update VarDesc from design doc

* Fix GCC compile

* Fix unittest

and add branch net comparing with cpu result

* Adding Hard Sigmoid Activation

* Adding a comment for slope to be only positive

* Fixing grammatical mistake in comment

* Add no_grad_vars for grad_op_maker

* Add unittest

* Fix unittest

* Fix unittest

* Follow comment

Use less GPU memory