* optimize: vit 384

* fix:bug

* fix:bug

* fix:supoort rocm complie

* refactor:name

* fix:support rocm

* fix:__HIP_NO_HALF_CONVERSIONS__

* optimize: delete scalar

* fix:rocm can't support

* fix:ernie error

* fix typo `Fasle`/`Flase` -> `Flase`

* fix typo `Ture` -> `True`

* fix fused_attention fused_feedforward

* fix ci

* fix ci

* fix ci PADDLE_GET_CONST

* fix ci ut

* namespace modify

* update by comment

* remove PADDLE_WITH_MKLDNN, test white_list=abs

* fix unique_ptr

* fix op.Type()

* remove TODO in kernel_dispatch.h

* remove IndicateVarDataType function, update white_list

* remove mkldnn hard code

* add comments

* fix ==

* update mkldnn_op_list

* delete hard code of OPs

* update mkldnn_op_list

* update mkldnn_op_list, remove interp

* add error check for ExecutionContext

* update mkldnn_op_list, remove transpose2_grad

* remove interpolate mkldnn

* remove fill_constant mkldnn

* opt HasAttr in DygraphExecutionContext

* deprecated commit, test mkldnn_white_list

* deprecated commit, test mkldnn_white_list

* deprecated commit, test mkldnn_black_list

* update mkldnn_op_list, add assert error op

* solve cudnn related op

* fix error

* add mkldnn fallback in phi_utils.cc

* remove mkldnn fallback in phi_utils.cc

* opt code implementation

* polish Copyright License

* remove using lodtensor part1

* polish history code format

make fused_multi_transformer support dynamically set the cache_kvs' shape and support input prefix_caches. (#46777)

* make fused_multi_transformer support dynamically set the cache_kvs' shape and support input prefix_caches.

* support pure bfloat16

* support bf16 linear

* update PR to pass CI

* tiny fix where_grad_kernel.cu

* add bfloat16 to selu_grad to pass CI

* fix selu grad compilation error

* remove needless using tensor

* remove needless using tensor

* resolve conflict

* replace tensor using

* fix format error

* revert needless changing

* fix rocm and npu compile error

* fix cinn compile error

* fix format error

* fix mkldnn format error

* fix mkldnn format error

* fix cinn compile error

* fix cinn compile error

* fix cinn compile error

* resolve conflict

* add gemm_epilogue

* xpu-paddlepaddle-40 [任务] fused_gemm_epilogue 支持 test=kunlun

* fix multi stream error.

* convert multihead to oss

* fix:bug

* fix:delete const cast

* fix:don't support bias_qk

* add vit pass

* fix:convert bug and add preln_residual_bias

* support length=-1

* add UT for convert

* add no_bias_qk support for gpu_multihead_op

* delete infer_shape depends on bias_qk

* oss just can be used in T4 and A*

* fix:change api for ROCM CI

* convert_fp16 support multi block

* update

* update

* add fused_multi_transformer post_layer_norm

* add test post_layer_norm

* multihead matmul add fp16

* fix windows error

* fix rocm error

* fix rocm error

* remove cudaDeviceContext

* remove more template

* fix rocm compile

* remove alias name CUDADeviceContext

* fix compile

* fix tests

* revert changes

* remove cudaDeviceContext

* remove more template

* fix rocm compile

* add some fp16 op for kunlun resnet50 model
*test=kunlun

* tmp
*test=kunlun

* fused_fc_elementwise_layernorm support fp16

* fused_fc_elementwise_layernorm support double

* multi stream support handle lazy init.

* support eigen lazy init

* update

* fix ci problem

* Rename BOOST_GET macros

* Fix conflicts