* [Demo] custom kernel based on pten kernel

* merge and npu custom work well

* del comments

* delete other code

* fix CUDAContext

* fix not found small_vector.h

* support NPU

* fix NPUContext

* fix DeviceContext support

* add UT

* fix call

* add UT

* fix

* fix for comments and ut

* add MACRO control

* fix multi input output

* support env CUSTOM_DEVICE_ROOT

* deal with special cases

* fix for Windows

* try coverage with test_custom_kernel_dot.py

* fix test_custom_kernel_dot

* fix test_custom_kernel_dot

* fix merge

* fix merge

* fix CI

* update

* merge and fix

* remove WITH_CUSTOM_KERNEL

* fix merge

* merge and fix

* fix ut

* fix ut for mac

* add more UT

* add more UT

* fix

* Upadate pass in quant2_int8_mkldnn_pass

* Back to the previous scale_matmul order

* Change place of cpu_quantize_placement_pass

* shuffle channel pass

* add ut

* timeout fix

* makefile fix

* update planner

* update unitest

* update dist matmul

* update auto converter

* optimize kunlun/xpu softmax_with_cross_entropy add add unitest
*test=kunlun

* minor
*test=kunlun

* minor
*test=kunlun

* minor
*test=kunlun

* minor
*test=kunlun

* fix bug:
1. atten: set the default value of attn_dropout_rate to None
2. ffn: add activation parameter

* for pure fp16

* Add a SparseCsrTensor

* remove unused functional

* remove const

* remove SetMemoberTensor

* remove non_zero_nums_, the number of non zero elements of each batch can be obtained from the crows

* SparseCooTensor

* add SetMember

* merge upstream; add SetMember

* merge upstream

* merge upstream; add newline at end of file

* add newline at end of file

* remove newline at end of file

* remove newline at end of file

* stash

* user pten::framework::make_ddim

* user pten::framework::make_ddim

* merge upstream; use the latest mutable_data

* merge upstream; use the latest mutable_data

* return mutable dense tensor

* update dist param grad for pass

* update unitest

* update unitests

* fix conflict

* Paddle-Inference:fix_concat_slice

* Paddle-Inference:fix_concat_slice

* Paddle-Inference:fix_concat_slice

* Paddle-Inference:fix_concat_slice

* [Paddle-Inference]: fix concat slice

* [Paddle-Inference]: fix concat slice

* [Paddle-Inference]: fix concat slice

Support the cases that the indices shape size is larger than the arr shape size

* add master weight for opt state_dict

* check empty of master weight

* strict gpu test

* refine unittest

* add fuse_relu_depthwise_conv_pass unittest

* fix atol and rtol

* fix according to review

* add FuseBatchNormAddActPass and unittest

* Update test_dist_fuse_bn_add_act_pass.py

* solve conflict

* Added selected_rows and rw_lock to pten

* Renamed the unit test target to fix CI

* Removed Class SelectedRows in Fluid, changed include/cmake relationship, use pten::SelectedRows in Fluid

* Remove rw_lock.h,rw_lock_test.cc in fluid

* Use pten::RWLock and pten::AutoRDLock, fix CI

* Use pten::SelectedRows

* Use pten::SelectedRows

* Fix to pass NPU CI

* Selected_Rows inherits from TensorBase

* Use pten::SelectedRows, to pass NPU CI

* To fix NPU CI

* To fix NPU CI again

* Use paddle/pten/core/enforce and polish code

* Support imperative selected_rows_to_lod_tensor

* Polish code

* [MLU]Add conv2d op

* [MLU]fix comment

* [MLU]adapt NCHW of conv2d op

* sync misc changes

* apply comments 01

* fix compile error

* remove is_ipu_place check

* add authors
Co-authored-by: NXiaobing Wang <xiaobingw@graphcore.ai>
Co-authored-by: NAllen Guo <alleng@graphcore.ai>
Co-authored-by: NZhixin Yao <zhixiny@graphcore.ai>
Co-authored-by: NHaicheng Jiang <haichengj@graphcore.ai>
Co-authored-by: NHan Zhao <hanzhao@graphcore.ai>

* sync changes

* restore cmake

* update ir cmake and setup.py

* update inference_lib cmake

* restore for split PR
Co-authored-by: NXiaobing Wang <xiaobingw@graphcore.ai>
Co-authored-by: NZhixin Yao <zhixiny@graphcore.ai>
Co-authored-by: NHaicheng Jiang <haichengj@graphcore.ai>
Co-authored-by: NHan Zhao <hanzhao@graphcore.ai>

* Optimize layer_norm fwd when cols is 1024.

* add sigmoid cross entropy with logits to kl2. test=kunlun

* add sigmoid cross entropy with logits to kl2. test=kunlun

* follow comments. test=kunlun

* add fuse_relu_depthwise_conv_pass unittest

* fix atol and rtol

* fix according to review

* Add fuse_bn_act_pass unittest

* rm others

* add fuse_bn_act_pass

* [inference] update convert reduce op&ut,test=develop

* update

* update

* update

* add int32 support

* add int32 support

* add comments

* trt < 7.0 do not support int32

* test=develop

* update

* test=develop

* [MLU]add mlu kernel for fill_constant op

* delete device_context DEPS

* assert _compile_dir include file existence

* polish

* [MLU]add mlu kernel for concat and split op

* delete device_context DEPS

* support param groups in grad_clip

* update

* modify for review

* update reshard for newest completion

* update unitest

* merge newest

* [autograd] static Jacobian pass tests.

* [autograd] apply CR suggested changes.

* [autograd] more tests.

* [autograd] add CPUPlace in tests.

* [autograd] bug fixes.

* [autograd] reformatted.