remove fluid.initializer.UniformInitializer, ConstantInitializer, NormalInitializer, TruncatedNormalInitializer, XavierInitializer, BilinearInitializer, MSRAInitializer, NumpyArrayInitializer and calculate_gain.. (#49498)

* move UniformInitializer and ConstantInitializer

* more modify

* circular import resolved

* another circular import resolved?

* more circular import 2

* circular import 3

* change import paddle in metric.py

* BuildStrategy import from fluid

* modify the framework import path in common.py

* change rnn.py import, from static to original framework

* change import static in the nn folder

* default_main_program should import from common_ops_import

* add import paddle in param_attr.py

* use core not paddle module for using VarDesc

* another old uniform

* mistake that use Uniform instead of UniformInitializer

* modify UniformInitializer doc

* move fluid.NormalInitializer to nn.initializer.NormalInitializer

* remove import of Normal in fluid.layers.nn.py

* remove more import of old Normal

* remove more import of old Normal

* sample code modify and tests modify import

* is_listen_failed passing arg should be log file

* problem solved

* a mistake solved

* comments resoleved and remove paddle.fluid.initializer.TruncatedNormalInitializer

* remove paddle.fluid.initializer.XavierInitializer and paddle.fluid.initializer.MSRAInitializer

* remove paddle.fluid.initializer.BilinearInitializer NumpyArrayInitializer and set_global_initializer

* change fluid to static

* change static to fluid to avoid circular import in distributed_strategy.py

* fix example code and test_initializer

* ValueType

* sample code fix

* change set_global_initializer back to fluid

* put paddle.static.BuildStrategy.ReduceStrategy into the fuction to avoid circular import

* remove calculate_gain, delete BilinearInitializer and revert set_global_initializer

* change the time of using UniformInitializer, ConstantInitializer, NormalInitializer, TruncatedNormalInitializer, XavierInitializer, MSRAInitializer, NumpyArrayInitializer as few as possible

* fix argument incampatible

* fix more arg incompatible

* fix test_prelu_op_xpu.py Constant

* fix inaccurate doc

* more doc fix: default value

* replace paddle.fluid.layers.data and remove io.data

* partial commit

* partial commit

* partial commit

* partial commit

* partial commit

* partial commit

* remove data in fluid.layers.io.__all__

* fix errors

* fix unitests

* fix unitest

* fix unitests

* fix unitest

* fix unitest

* fix unitests

* fix unitest

* fix test_layers unitests

* fix typro

* fix unitest

* fix unitest

* fix unitest

* fix typro

* fix unitest test_model_cast_to_bf16

* fix test_reducescatter

* fix collective unitest

* fix collective unitests

* fix collective unitests

* add coverage

* fix add layers.data

* re run ci

* fix some typro

* fix samplecode error

* fix samplecode error

* clean elem_arithmetic part6 unittest

* delete op_name_conflict unittest

* restore test_op_name_conflict

* fix codestyle test_op_function_generator

* update config

* re-blacken python code

* temporarily disable date and diff_py_file

* skip a format

* [CodeStyle] remove all future import

* revert test_error.py

* restore future import in example code

* use yapf to format all python file

* yapf exclude two unittests file for they rely on writing and reading file, and format will break them

* disable diff_py_file because too many diff files cause command following failed

* Create stateful OneDNNAXPYHandler object.

This makes it possible to call it multiple times without recreating the
oneDNN primitives every time.

* Prepare SGDOpKernel to reuse its implementation from OneDNN kernel.

* OneDNN SGD kernel.

* Update call to use new OneDNNAXPYHandler object api.

* Setup seed in proper place.

* Enable OneDNN kernel only for single case.

* For dense param and sparse grad.

* Small refactor.

* Enable oneDNN by op attr or by cmd line flag.

* Use int64_t type for number of elements.

* Support dense param and grad from OneDNN kernel.

* Enable SGD OneDNN kernel when use MP BF16 optimizer.

* Force non-copyable/movable OneDNNAXPYHandler.

* Reuse OneDNNAXPYHandler for spare tensors in SUM op.

* Fix SFINAE rules.

* Remove recording event inside AXPY.

* Get rid of internal primitive caching.

* Stop use PP cache mechanims to store mem and primitive obj.
* Handler obj store and reuse needed desc & prim

* Do not derive from MKLDNNHandlerT

* Enable BF16 for creating global tensor and reduce_mean.

* Functional test with small model.

* Add oneDNN AXPY handler.

* Add fallback for small tensors.

* Fix ifdefs

* Remove unnecessary namespace prefixes and add missing headers.

* Guard handler_axpy with proper ifdefs.

* Compilation of this function is possible only when Paddle is not build
with CUDA nor HIP.

* Move AXPY handler code to separate files.

* Use oneDNN AXPY handler in SGD op.

* Use axpy handler only when Paddle is built with oneDNN.

* Add test for SUM BF16 with big rows.

* Fix SFINAE rules for elementwise_add_to.

* Add test case for SGD with big rows.

* update

* update
Co-authored-by: NAdam Osewski <adam.osewski@intel.com>

* Initial draft for SGD BG16 kernel.

* Unit tests for SGD with BF16 data type.

* Add VLOG message to SGD BF16 op CPU kernel.

* Enhance error messages and error types.

* Refactor SGD op kernels to leverage some common code.

* Make easier to add new kerne invoke code.

* Fix SGD op kernel for sparse grad.

* Unify quotes style.

* Fix error for ROCM compilation.

* Use specialized PADDLE_ENFORCE_xx functions.