* skip ReplaceAllReduceOp in GraphtoBlock when nccl_ctxs_ is nullptr

* update ut

* test_dist_allreduce_op failed

* fix test_dist_allreduce_op

* add ut

* fix nccl cpu compile

* fix

* [CodeStyle][py2] remove `compat` module (to_bytes)

* remove some unused imports

* clean up to_bytes definition and unittests

* Revert "clean up to_bytes definition and unittests"

This reverts commit e726539e1768172a411ff60e63fab82f164343cf.

* use `b` prefix instead of `encode()`

test=kunlun

* add singleton to custom device

* Update custom_device.cc

Init device_init_flag_ in default

* simplify conv_mkldnn op registration

* remove custom type value in conv grad op

* update test_image.py

* update test_image.py

* for gpt-gen

* fix reshard

* adapt assign and shape op

* add dist_assign & unittest

* add conditional block unittest

* rename unittest

* Fix hAPI bug of not compatible with LayerHook

* draft with debug print

* Add more record event in run program op

* Refine code

* Restore code

* Rename event

* refind logic of choose phi kernrel

* fix complie budg

* fix immutable op quantize bugs

* fix

* fix build bug

* fix test

* notest,test=inference

* fix ppyoloe acc drop bugs

* fix test

* fix test

* add test

* fix

* fix

* fix test

* fix refined name bug

* fix test

* bias fix

* fix matmul weight dequant bug

* re-ci

* fix tester

* fix test

* fix tester

* update weight dequantize func

* update code

* update test for converage

* update test

* update cmake

* update cmakelist

* update code

* rerun ci

* remove useless code

Add symbolic shape deduction function for unfold, scatter_nd_add, p_norm, grid_sampler, pad3d, etc (#46291)

* test=infer-coverage

* [BUG]Fix expand_as_v2 bug while X and Y with different dtype

* fix commit

* combine dp and stage2 hybrid parallel.

* update condition.