* Add bf16 pool2d and unify bf16 unit tests

* Add change default ops test

* default not show cpp statck & add hint

* fix failed unittest

* fix failed unittests

* change load dict file one by one to warn

* add unittests for coverage

* polish error message

* fix cond error

* Rename variables when use 'jit.load'

* Check whether the original graph contains the variable with the same name

* add comment

* rename output/input of op and edit unittest

* modify the code according to CI

* edit code according to CI

* edit code according to CI

* edit code according to CI

* edit code according to CI

* edit code according to CI

* edit code according to CI

* fix syncbn

* add unittest

add uint8 for reshape operator

* Support pop for dict in dy2stat

* Move convert_pop to convert_operators.py and polish convert_pop

* fix tensor detach to zero copy

* fix tensor detach to zero copy

* polish parallel api impl & doc details

* add unittest for coverage

* remove spawn test in py2.7

* add parallel api into white list

As the title

* add soft_label and axis for CrossEntropyLoss and improve performance,test=develop

* fix conflict in nn/functional/loss.py, test=develop

Add support for using tuple as tensor.shape (For example: a, b, c, d = x.shape)

* add lamb optimizer and unittest

* fix lamb

* fix lamb v2 op

* fix sampling id

* fix lamb sample code

* Update lamb.py

* fix doc

* fix doc

* Update lamb.py

* upgrade comment string to raw string

* fix string in

* fix string with ' '

* revert update on comments

* upgrade only necessary

* fix sample code checker

* fix comments with '''

[Dynamic-to-Static] Fix bug of convert_logical_and/convert_logical_or: the operands are executed sequentially(#28993)

  
  1) The operands are executed sequentially according to the running logic of Python.
  
  2) If the left hand operand is True(for convert_logical_or)/False(for convert_logical_and), the right hand operand should be executed.

The PR description is long. See details in the PR link.

* update, test=develop

* refactor momentum op to combine weight_decay (scale op and sum op)

fix error for Can not find test to add properties to:test_parallel_dygraph_transformer,test_parallel_dygraph_sparse_embedding (#28968)

* fix conv_transpose unittest. test=develop

* fix dataloader. test=develop

* fix param base trainable set failed

* add unittest

* fix typo

* polish comment

Adding uint8 support for squeeze operator.

* Add bf16 matmul, fc, elementwise add and mul

* Correct unit test

* Fix wrong out dtype from helper.input_dtype

* add unittest

* remove disable_static in op_test

* fix param name typo

* 1.add xpu slice op 2. add xpu top_k op 3.modify xpu cast to new api

* 1.add xpu slice op 2. add xpu top_k op 3.modify xpu cast to new api

* add lstm, simple rnn op kernel

* fix the test_lstm for the rnn op

* change func name

* fix forward postprocess bug

* add gru forward, backward code

* remove unittest.skipIf; use a big rnn op instead of combination op

* fix input doesn't have gradient bug

* add eigen lstm forward, backward
Co-authored-by: Nwawltor <fangzeyang0904@hotmail.com>

* adjust kunlun header file
*test=kunlun

* update kunlun unittest
*test=kunlun

* update xpu unitest
* test = kunlun

* update xpu unittest
* test=kunlun

* update xpu unitest
* test=kunlun

[Dynamic-to-Static] Fix bug in loop_transformer: loop vars should contain the var from ancestor-for-node (#28735)