* [NPU] add tril_triu

* [NPU] delete debug codes

* [NPU] add more test cases, and api test

* [NPU] optimize codes style

* fix the docs bug in collective file

* psgpu:edit cuda remote_streams; test=develop

* Support getitem by ellipsis index in dynamic mode

* change some code style

* Support getitem by none index in dynamic mode

* modify a comments style and remove useless code

* [NPU] add cumsum

* [NPU] delete debug codes

* [NPU] add attr flatten and unittests, and api tests

* [NPU] delete comment codes

* [NPU] add attr flatten and axis exclusive check

* [NPU] delete skipIf

Fix the bug in paddle.distributed.split demo. The paddle.distributed.split op just can be used in static mode. (#34306)

* fix the bug in paddle.distributed.split demo

* [NPU] add hard_sigmoid

* [NPU] delete check_dygraph=False and max_relative_error

* [NPU] delete debug codes

* [NPU] add more test cases

* [NPU] add api test TestHardsigmoidAPI

* [NPU] temp delete hard_sigmoid for resovle conficts

* [NPU] resolve conflicts

* fixed CI failing

* removed unnecessary imports

Revert "[Dy2Stat] Refactor ExecutorCache logic and pre-support BuildStrategy for pass (#34181)" (#34348)

This reverts commit 609f8225.

* logical ops support int8, int16, int32, int64, float, double

* update docs of logical ops

* fix npu and xpu logical ops

* fix npu and xpu logical ops

* fix bug in xpu logical op code

* update test_logical_op_npu and test_logical_op_xpu

* correct error type

* add index_sample_op_npu and tests

* update

* fix variable name

* fix variable name

* copy found_inf to cpu in advance to improve performance

* add npu test

* add npu test

* refine code

* refine memcpy op

* fix adam

* modify into program_id

* fix cache_info declare problem

* fix python int to C long problem

* modify point to reference

* add ENVS

* [NPU] update NPU ci tests, test=npu_aarch64

* [NPU] fix x86 build and add disable_ut for NPU, test=npu_aarch64

* [NPU] address review comments, test=develop

* added sigmoid BF16 FWD/BWD and gelu BF16 BWD

* added newline at EOF

* switched from lambdas to local functions

* changed function names

* add int16 kernel for lookup_talbe and dequantize_abs_max op

* Support getitem by ellipsis index in dynamic mode

* change some code style

* upgrade test_tensorrt to trt8

* format