* license under bsd

* fix rat exclude

* add batch_axis in validation handler

* Add test case for batch axis support

* change test class name

…ion (#17388)

* fix build error due to  #17128

* whitelist all *MX* API

* Update libmxnet.sym

* Update libmxnet.ver

Co-authored-by: Sheng Zha <[email protected]>

- Add missing iot urlParam
- Fix use of potentially undefined variables. For example, if the &processor=
  url parameter is unspecified, this would previously cause use of undefined
  variable javascript error leading to an empty page.

* Update FindNCCL.cmake

If Cuda root dir is not specified, search for the symlinked unix default

* Update FindNCCL.cmake

* fix the cmake variable name

FindCUDAToolkit exposes result variables (cmake variables). Use those.

https://github.com/apache/incubator-mxnet/blob/28e053edb4f2079743458bf087557bcac7e58c62/cmake/Modules/FindCUDAToolkit.cmake#L427-L464

* search for library in /usr/local/cuda

* comment if

* indent, comment fix

* fix indent, add nccl_root alongside nccl_root_dir

Also add warning about deprecation

* fix indent, remove nccl_root checks

* Update FindNCCL.cmake

* add api

* make doc compatible with c_api.h INT64 docs

…her_nd error test (#17360)

* ndims and shape of dims

* add c_api definition and update invoking of capi in sparse

* minor fixes

* fix c_api for sparse64

* fix lint, add test

* change test, indext for thread

* fix test

* replace userdefined dtype with standard dtype

* revert dtype to int for ndim of 64bit SparseAPI, retain uint32 for 32bit API

* Sparse 32bit as well as 64bit API ndim dtype made int

* revert index_t to IType as it is int64 anyway no need for hardcode

* fix sparseEx API

* fix the api breaking change for 32bit API

* remove template parameter for ndim,aux_ndim since internally its always int

…ed memory footprint (#17410)

…17405)

https://cmake.org/cmake/help/latest/policy/CMP0074.html

* Add event handlers in validation handler

* update doc string

* fix template param as ndim is internally int

* Update c_api.cc

* use temp file

* fix dependency

* Update model_store.py

* Update test_gluon_model_zoo.py

* remove NamedTempFile

* norm

* full test

* add default behaviour

* add col norm forward

* add matrix col row norms forward

* row col norm backward

* improve tests

* beautify cpp

* update broadcast_op

* C1002

* billie holiday even told it even better

* probing for windows unittest numpy error

* update test

* update test

* fix style

* retrigger unix ci

* update according to reviews

* fix backward set_num_input

* fix CI

* fix seed for mkldnn test

* use fixed seed for mkldnn test

* fix cpp predcit license

* add white list (#210)

* fix white list (#211)

Co-authored-by: Lai Wei <[email protected]>

Includes ps-lite CMakeLists.txt refactor

* Add config.cmake

* Update guide

* Address comments

* Remove undocumented include(build/private/local_config.cmake)

* Fix typo

* safe load for yaml

* Trigger notification

)

* Add cmake/upstream folder

* Add upstream select_compute_arch.cmake

* Change CUDA_COMMON_GPU_ARCHITECTURES

* Include License

* Workaround rat-exclude problems

* Initial commit - added first batch of misc ops

* Initial commit - added first batch of misc ops

* Added remaining misc ops, including Custom op logic

* Added more test cases, fixed lint errors

* Update documentation

* Added run_backward=True for ops supporting backwards runs

* Added issue link for bilinear UpSampling

* Added remaining misc ops, including Custom op logic

* Update documentation

* Updated alias map

* Fixed missing and incorrect alias issues

* Added remaining missing aliases

* Fixed Custom profile dump parsing and alias

* Switched to using sets for O(1) op membership checks

* Added fix for dtype issue in master

* fix storage type of softmax backward

* remove unused variable

* Drop _cy2

* Drop Python 2 specific code in mxnet and tests

* Replace io.open with open

* Drop from __future__ imports

* Fix lint

* support broadcasting on the indexed axis

* support extra cases for broadcasting setitem, fixing a few other bugs in indexing

cublasGemmBatchedEx is only supported for GPU with architecture capabilities equal or greater than 5.0.

Fixes a bug in #16408

* add op insert

fix lint and compile error

fix pylint error

* fix slice bug

* modify code logic and patterns

fix sanity error

* reuse existing function / modify alignment

* support to insert a sigle number

* use range_fwd kernel

* reduce code

* fix a small mistake / remove redundant function

* reduce code again / resolove function conflict

* fix pylint error / decompose main function into small pieces

* test CI

* test CI 2

* test CI 3

* test CI 4

fix sanity error

* op : insert tensor

* op : insert scalar

fix sanity error

* op : insert by slice

fix compile error

fix cmopile error again

* fix code style problem

* enable mkldnn by default in pip packages

* add make file for native build

* rm *mkl build scripts

* remove *mkl variants

* Change readme/cd

* clear native build configurations

* fix static build test in ci

* git mv linux_cu102mkl.mk linux_cu102.mk

* fix merge conflict

* add gpu doc portion

* resolve sam comments

* add lib_api and resolve comments

* resolve sam comments

* try remove italic

* fix typo and add mutable example

* retrigger ci

* Add Bfloat16

* mshadow support bf16

* rebase bf16 mkldnn1.0

* support bf16 gemm

* resolve fp32 ip bwd bug

* add other bf16 ops

* change func name from fp16 to lp16 (low precision 16), to include bf16

* add amp_cast bf16 support for ndarray

* fix executor copy_params

* add test case for bf16

* remove numpy dtype hook for bf16

* add bf16 type support

* rebase to mxnet master

* add single conv test

* fix symbolic inference

* add dtype check when copy

* add single conv and bn test

* skip fp16 amp_cast test in cpu

* Fix resnet50 first convolution

* Skip first convolution for bfloat16

* support bf16 fallback compute

* recover origin test

* add some bf16 unittests

* fix bf16 bn test, enhance assert_almost_equal_with_err

* using assert_almost_equal_with_err for fallback bn test

* add relu6 bf16 support

* fix lint

* fix subgraph conv with data=0

* mkldnn doesn't support 0 dim tensor

* rm dtype check when copy

* using bf16 tvm

* rm bf16 mnist demo

* use official tvm

* change function name; fix lint error

* fix clang check error:conditional expression is ambiguous; 'float' can be converted to 'mshadow::bfloat::bf16_t' and vice versa

* nvcc compiler build pass

* fix gpu amp cast symbol error

* fix mnist training error

* fix cpp test: Engine.VarVersion error

* workaround cpp failed test mkldnn fc bwd

* to fix mkldnn test_mkldnn_ndarray_slice error

* 1. move some code from to np_broadcast_reduce_op_value.cc to np_broadcast_reduce_op_value_part2.cc to pass Win CPU/GPU build (fatal error C1002: compiler is out of heap space in pass 2)
2. rm debug code

* use official dlpack

* rename np_broadcast_reduce_op_value_part2.cc and add some description

* 1. update dlpack url in .gitmodule
2. disable mkldnn fc bwd

* fix remaining NodePtr due to tvm update

* mv some code from mxnet_op.h to mxnet_op_kernel_assign.h to avoid WIN compiler error 'fatal error C1002: compiler is out of heap space in pass 2'

* fix WIN CPU build fail:compiler is out of heap space in pass 2

* fix WIN build fail

* fix lint

* add print for test bf16_concat

* fix bf16 test fail

* disable bf16 concat test

* tmp skip to root cause edge test halt

* fix bf16_bn test error

* enable test_bulk

* tmp rm bf16 to locate edge error

* Revert "tmp rm bf16 to locate edge error"

This reverts commit 7360246.

* add Apache license header

* trigger CI

* add robust for test bf16 bn

Co-authored-by: Zhennan Qin <[email protected]>
Co-authored-by: YixinBao <[email protected]>
Co-authored-by: Xinyu Chen <[email protected]>
Co-authored-by: Wuxun Zhang <[email protected]>

* update symbol to json

add remove_amp_cast argument to keep same with symbol.save

* retrigger CI

Co-authored-by: JackieWu <[email protected]>

* Additional fix for vector access. See 9634786 for the original.

* CI

* ci

* ci

* retrigger CI

* ci

Co-authored-by: JackieWu <[email protected]>

* Fix OS X staticbuild and add tests

* Update OS X build from source documentation

* Fix

* Try to fix...

…17608)

Fixes `gzip: stdin: not in gzip format` errors due to CI trying to unzip HTML page containing the forward notice.

…d vsplit, add interoperability tests for h/v/dsplit (#17478)

* C++

* rayleigh

* exponential

* c++

* fix mkldnn fc bwd bug due to data inplace

* enable mkldnn fc bwd

* fix cpp tests

* try: fix random seed

* fix cpp test

* loose rtol for fc cpp test

* improve error message

* limit max value for mkldnn tensors

* limit the max value of test tensors

* fix lint

* remove fixed random seed

* address review comments

* Revert "address review comments"

This reverts commit 56d873f.

Co-authored-by: rongzha1 <[email protected]>

* Add fallback ops

* More fallback ops

* Add more fallback ops

* Fix astype

* Support mixed type of mx.np.ndarray and np.ndarray in binary ops

* More fix

* Add result_type

* Remove implemented ops

* Fix

* Add more tests

* fallback---------------

* 8

* all fallback

* ok

* delete repetition

Co-authored-by: reminisce <[email protected]>