* fix_gradients

* fix_gradients, test=develop

* fix persistable;

* fix save load bugs; test=develop

* fix bug; test=develop

* add example for new io api; test=develop

* addd example; test=develop

* add Sequential api
test=develop

* fix unittest
test=develop

* refine code sample

* test=develop

support Tensor for split and concat, support -1 in num_or_sections, add check num_or_sections (#20780)

* improve split and concat op:
1. support Tensor for argument 'dim' in split op.
2. support Tensor for argument 'axis' in concat op.
test=develop

* redefine function GetDataFromTensor and set unknown output shape to - 1.
test=develop

* add check: Attr(sections) match Input(X). test=develop

* support Tensor for attr(sections) and attr(sections) can contain -1.
add check for attr(sections).
test=develop

* modify error message for concat and call Resize only when necessary. test=develop

* Refine the InferShape of ReadFrom and WriteTo op, and add comment to explain why not call ShareLoD for runtime.
test=develop

* Add comment for ReorderLoDTensorByRank op.

* Add comment for lod_tensor_to_tensor_array op to explain why only call DecreaseLoDLevel for compile time.
test=develop

* ShrinkRNNMemory op should call ShareLoD for compile time.
test=develop

* Add the implementation of IncreaseLoDLevel and add the compile-time check of lod_level in InferShape of sequence_pool.
test=develop

* Refine the unittest of DynamicRNN.
test=develop

* Change PADDLE_ENFORCE to PADDLE_ENFORCE_NE.
test=develop

* improve unsqueeze op to support int, Tensor and Tensor list for argument axes.
test=develop

* call Resize only when necessary. test=develop

*  Make shape tensor support int32

test=develop

* no longer need to define all embedding layers (no one less) of all slots in each program. make trainer_param repeated in ps.proto.
* add find_distributed_lookup_table_grads instead of hard code GRAD
* support embedding stop gradient. push sparse has error before fix this.* 
* fix fill sparse, skip slots which do not have embedding. each slot's embedding in a sparse table should be used in all training programs before fix this.
* fix pull sparse, skip slots which do not have embedding.
* fix collect feasign label info, skip slots which do not have embedding.
* support when there are multi sparse tables in one or multi training programs, each program can pull/push its own related sparse tables instead of all sparse tables.
* test=develop

test=develop

* All elements in attr(shape) of crop_tensor can be -1, test=develop, test=document_preview

* fix the bug that attr(offsets) should be initialized, test=develop

* test=develop

* test=develop

* test=develop

* test=develop

* make search_compute support avx only

* clean search_compute.h

* rename sse_axpy to avx_axpy

test=develop

* update CMakeLists.txt

test=develop

* jit/trace 1st version, test=develop

* add more unittests, test=develop

* refine Categorical and MultivariateNormalDiag en doc test=develop, test=document_fix

* refine Categorical and MultivariateNormalDiag en doc test=develop, test=document_fix

* mv sampcd_processor.py to tools

test=develop test=document_fix

* update example script

test=develop test=document_fix

* update the infer shape of matmul,  test=release/1.6

* add unittests of matmul, test=release/1.6

* change func names, test=develop

* add support for __get_item__ of Variable test=develop

* add cuda place and remove fluid.layers.data for test of python API, test=develop

* add cuda place and remove fluid.layers.data for test of python API, test=develop

* modified batch size for Python API test, test=develop

* test=develop, communicator_is_sgd_optimizer flags

* add check nan / inf in downpour worker during training
* test=develop