* add FP16 dtype for CastNumpy2Scalar

* fix throw message

* add test

* fix SyntaxWarning

* test skip for float16

* fix dtype mistakes

* [CustomDevice] add custom device api

* update

* update

* test=document_fix

* update

* update

* add  examples

* support inplaced variable in cinn_launch

* fix error hint when compiling

* fix inplaced output variable of the subgraph

* skip CinnCompiler check

* using existed definition

* fix namespace reference error

* modify error message

* update cinn tage

* fix namespace

* skip enforce check

* fix unittest attribute throw

* decouple phi custom_op

* decouple phi custom_op, remove codes

* delete custom symbol of inference

* rm flags_retain grad in pybind

* retain grads for xpu test

* set retain grad for xpu

* rm flag

* lint

---------
Co-authored-by: Nwanghuancoder <wanghuan29@baidu.com>

replace all TensorFromVector & TensorToVector

AssignKernel async copy

* Support stream priority for standalone executor

* Fix compile error

* Fix compile error

* Fix compile error

* Fix compile error

* Fix compile error

* add set slot_num for psgpuwraper (#177)

* add set slot_num_for_pull_feature for psgpuwarper

* Add get_epoch_finish python interface (#182)

* add get_epoch_finish interface

* add return

* delete return

* add unzip op (#183)

* fix miss key for error dataset (#186)

* fix miss key for error dataset

* fix miss key for error dataset
Co-authored-by: Nyangjunchao <yangjunchao@baidu.com>

* add excluded_train_pair and infer_node_type (#187)

* support return of degree (#188)

* fix task stuck in barrier (#189)
Co-authored-by: Nyangjunchao <yangjunchao@baidu.com>

* check node/feature format when loading (#190)

* check node&feature format when loading

* check node&feature format when loading (2£ (2)

* degrade log (#191)

* [PGLBOX]fix conflict

* [PGLBOX]fix conflict

* [PGLBOX]replace LodTensor with phi::DenseTensor

* [PGLBOX]fix gpu_primitives.h include path

* [PGLBOX]from platform::PADDLE_CUDA_NUM_THREADS to phi::PADDLE_CUDA_NUM_THREADS

* [PGLBOX]fix unzip example code

* [PGLBOX]fix unzip example code

* [PGLBOX]fix unzip example code

* [PGLBOX]fix unzip example code

* [PGLBOX]fix unzip ut

* [PGLBOX]fix unzip ut

* [PGLBOX]fix code style

* [PGLBOX]fix code style

* [PGLBOX]fix code style

* fix code style

* fix code style

* fix unzip ut

* fix unzip ut

* fix unzip ut

* fix unzip

* fix code stype

* add ut

* add c++ ut & fix train_mode_ set

* fix load into memory

* fix c++ ut

* fix c++ ut

* fix c++ ut

* fix c++ ut

* fix code style

* fix collective

* fix unzip_op.cc

* fix barrier

* fix code style

* fix barrier

* fix barrier

* fix code styple

* fix unzip

* add unzip.py

* add unzip.py

* fix unzip.py

---------
Co-authored-by: Nchao9527 <33347532+chao9527@users.noreply.github.com>
Co-authored-by: NSiming Dai <908660116@qq.com>
Co-authored-by: Nhuwei02 <53012141+huwei02@users.noreply.github.com>
Co-authored-by: Nyangjunchao <yangjunchao@baidu.com>

* add missing proto file

* fix windows ci

* fix ci compile error

[CustomDevice] registering feed_dense_tensor, feed_sparse_coo_tensor, feed_strings kernels for custom device (#50042)

* [CustomDevice] registering feed_dense_tensor, feed_sparse_coo_tensor, feed_strings kernels for custom device

* update

* update

* update

* remove max_slot_num

* fix test case

* [CINN] collect inplace var into cinn op desc's kInplaceVarNames attribute

* attr move from op desc to subgraph

* GetFetchIds from var_map instead of var_model_to_program_map_

* update xccl lib & use native Reduce in dygraph

* minor

* modify name

* merge develop

* fix param

* fix exp gen bug

* fix sum_grad

* comment

* [KUNLUN] add op: maxpool_with_index

* use DeviceContext::Alloc() instead of DenseTensor::mutable_data()

* fix file format

* solve clip unittest failure

* minor fix

* Revert "solve clip unittest failure" since the issue is fixed
in #49535

This reverts commit 1127adc66e79afe35ac3c00bb34e6aaa7cd7d78b.

* align with xdnn on the definition of mask in max_pool_with_index

* minor

* support PaddlePaddle Backend on Triton

* fix test cases

* fix Codestyle

* add test case

* add test case

* extract fuse pass logic to header file

* adjust namespaces

* Update paddle/fluid/framework/ir/mkldnn/activation_onednn_fuse_pass.h

update date
Co-authored-by: NTomasz Socha <tomasz.socha@intel.com>

* add inline remove static
Co-authored-by: NTomasz Socha <tomasz.socha@intel.com>

* fix cast issue

* add ut

* Add the cumsum 0d tensor

* xpu and cpu judge the 0d  tensor

* change to 2022 to 2023 in new commit

* fix the reverse logic

* revert to use default XPU stream for computing

XPUContext now has a null stream by default. If you want to use a separate stream
 (e.g. in async collective communication), you should create a dedicated XPUContext
and invoke its XPUContext::CreateStream()

* minor

* add test for composite with dy2st

* add more log

* refine munmap freq for ref_cnt_mmap_allocator

* add shm reuse logic

* fix compile bug

* fix compile bug

* fix bug of file refcount

* fix compile bug

* fix compile bug

* refine code for delete shm case

* polish code

* refine shm cache pool size setting logic

* set buffer is 2

* refine shm cache size logic

* refine max shm cache

* refine shm cache size

* reshape_transpose_matmul_pass_tester rewritten

* matmul_transpose_reshape_pass_tester rewritten

* mkldnn to onednn

* new exe supports CUDA Graph

* fix

* fix

* fix

* fix FLAGS_use_stream_safe_cuda_allocator in unittest

* insert output of coalesce_tensor op to skip_gc_var

* fix