You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@mxnet.apache.org by la...@apache.org on 2020/04/04 21:49:07 UTC
[incubator-mxnet] branch leezu-patch-1 updated (369da39 -> 8a9599b)
This is an automated email from the ASF dual-hosted git repository.
lausen pushed a change to branch leezu-patch-1
in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git.
discard 369da39 Disable MKL_USE_STATIC_LIBS by default
add 517e3a1 Fix mxnet-native and Docker CD pipelines (#17784)
add 0aa2c78 Fix OpPerf in Master (#17735)
add b001006 add npx.broadcast_like (#17605)
add 2cf8e0f ffi invocation: expand_dims, tril, diff, broadcast_to (#17738)
add 4dddb08 Implement storage tagging, the first half of the memory profiler (#17656)
add afb8742 Add ffi benchmark (#17780)
add a5d55de [OpPerf] Consolidate array manipulation related operators (#17487)
add 713d962 [Numpy] FFI: Bincount, Percentile/Quantile, All/Any (#17717)
add bd6e917 [numpy] add op random.f (#17586)
add dfbcf6f fix np.clip scalar input case (#17788)
add 18c2a26 [numpy] add magic methods for symbol bitwise ops (#17807)
add 34010ea [CD] switch CD_RELEASE_JOB_NAME from global env var to job argument (#17775)
add c5989fa [CD] update pypi description, setup.py (#17681)
add bbba462 Update ccache on CI to fix cuda object caching (#17828)
add d495f22 fix chisquare (#17825)
add ea2daba [NumPy] add op random.laplace (#17316)
add bd6b80e CI: Disable AVX instructions in libzstd ccache dependency (#17842)
add 66b21b5 fixing batch_norm and layer_norm for large tensors (#17805)
add 1368a08 CI: switch to patched ccache 3 version (#17853)
add 796fa50 * impl - FFI for np_where_op (#17817)
add a7ecb35 ffi wrappers for polyval, ediff1d, nan_to_num (#17832)
add 1efb747 [Numpy] FFI for split and svd (#17816)
add ab4f7f6 CI: Test clang10 cpu & gpu builds with -WError (#17830)
add 5996544 Fix SoftReLU fused operator numerical stability (#17849)
add dfb1b88 HybridBlock.export() to return created filenames (#17758)
add 2fae7e4 add ffi for full_like, binary (#17811)
add 41d534b Skip test_multi_worker_dataloader_release_pool on OS X (#17797)
add 07d0b73 ffi for roll/rot90 (#17861)
add ea2320a * FFI for np.argmax and np.argmin (#17843)
add 04167df staticbuild: Fix potential user-assisted execution of arbitrary code (#17860)
add 4fa4e65 Refactor cpp-package CMakeLists.txt & add missing inference/imagenet_inference (#17835)
add f7c4323 Dynamic subgraph compile support (#17623)
add 3667e9a Update 3rdparty/mkldnn to v1.2.2 (#17313)
add f01dc80 Adding sparse support to MXTensor for custom operators (#17569)
add 2f358fd [Numpy] Add op fmax, fmin, fmod (#17567)
add b133899 Use multi-tensor sumSQ in clip_global_norm (#17652)
add 8e39518 Use FP32 copy of weights for norm (multitensor LAMB optimizer) (#17700)
add 83b5170 Add simplified HybridBlock.forward without F (#17530)
add 3840786 cmake: Set DMLC_LOG_FATAL_THROW only for building mxnet and not for tvm (#17878)
add 9a355eb [Numpy] Kron operator (#17323)
add d4052fd [Numpy] FFI for diag/diagonal/diag_indices_from (#17789)
add 56e7985 * impl - FFI for np einsum (#17869)
add 08528c5 [numpy] add op median (#17084)
add 66ee118 Fix Windows GPU CI (#17962)
add 792011e Omit kNullOp req when comparing changed NDArrays in static_shape=True (#17966)
add b6edefb CI: Re-enable code coverage for CPU builds (#17889)
add 8a9599b Disable MKL_USE_STATIC_LIBS by default
This update added new revisions after undoing existing revisions.
That is to say, some revisions that were in the old version of the
branch are not in the new version. This situation occurs
when a user --force pushes a change and generates a repository
containing something like this:
* -- * -- B -- O -- O -- O (369da39)
\
N -- N -- N refs/heads/leezu-patch-1 (8a9599b)
You should already have received notification emails for all of the O
revisions, and so the following emails describe only the N revisions
from the common base, B.
Any revisions marked "omit" are not gone; other references still
refer to them. Any revisions marked "discard" are gone forever.
No new revisions were added by this update.
Summary of changes:
.gitmodules | 1 -
3rdparty/dmlc-core | 2 +-
3rdparty/mshadow/CMakeLists.txt | 9 +-
CMakeLists.txt | 147 +++--
Makefile | 7 +
benchmark/opperf/README.md | 8 +-
benchmark/opperf/nd_operations/README.md | 107 +---
.../nd_operations/array_manipulation_operators.py | 264 +++++++++
benchmark/opperf/opperf.py | 28 +-
benchmark/opperf/rules/default_params.py | 48 +-
benchmark/opperf/utils/op_registry_utils.py | 136 ++++-
benchmark/opperf/utils/profiler_utils.py | 4 +-
benchmark/python/ffi/benchmark_ffi.py | 170 ++++++
cd/Jenkinsfile_cd_pipeline | 15 +-
cd/Jenkinsfile_release_job | 2 +-
cd/Jenkinsfile_utils.groovy | 20 +-
cd/python/docker/python_images.sh | 2 +-
ci/build_windows.py | 68 ++-
ci/docker/Dockerfile.build.ubuntu_gpu_cu100 | 3 +
ci/docker/Dockerfile.build.ubuntu_gpu_cu101 | 3 +
ci/docker/Dockerfile.build.ubuntu_gpu_cu102 | 3 +
ci/docker/Dockerfile.build.ubuntu_gpu_cu80 | 3 +
ci/docker/Dockerfile.build.ubuntu_gpu_cu90 | 3 +
ci/docker/Dockerfile.build.ubuntu_gpu_cu92 | 3 +
ci/docker/install/centos7_ccache.sh | 12 +-
ci/docker/install/deb_ubuntu_ccache.sh | 41 +-
ci/docker/install/ubuntu_binutils.sh | 12 +-
ci/docker/install/ubuntu_clang.sh | 10 +-
ci/docker/runtime_functions.sh | 127 +++--
ci/jenkins/Jenkins_steps.groovy | 86 ++-
ci/jenkins/Jenkinsfile_clang | 4 +-
ci/jenkins/Jenkinsfile_miscellaneous | 2 +
cmake/BuildTVM.cmake | 8 +-
cmake/Modules/FindNVML.cmake | 84 +++
config/distribution/linux_native.cmake | 31 +
cpp-package/CMakeLists.txt | 63 ++-
cpp-package/example/CMakeLists.txt | 103 ++--
cpp-package/example/charRNN.cpp | 2 +
.../example/inference/CMakeLists.txt | 13 +-
cpp-package/scripts/OpWrapperGenerator.py | 15 +-
example/extensions/lib_custom_op/Makefile | 10 +-
.../extensions/lib_custom_op/test_transposecsr.py | 78 +++
.../lib_custom_op/test_transposerowsp.py | 73 +++
.../extensions/lib_custom_op/transposecsr_lib.cc | 197 +++++++
.../extensions/lib_custom_op/transposerowsp_lib.cc | 199 +++++++
example/extensions/lib_subgraph/README.md | 69 ++-
example/extensions/lib_subgraph/subgraph_lib.cc | 45 +-
example/extensions/lib_subgraph/test_subgraph.py | 62 +-
.../predict-cpp/image-classification-predict.cc | 6 +-
include/mxnet/c_api.h | 53 +-
include/mxnet/imperative.h | 90 ++-
include/mxnet/lib_api.h | 387 +++++++++++--
include/mxnet/libinfo.h | 4 +
include/mxnet/ndarray.h | 72 +--
include/mxnet/resource.h | 70 ++-
include/mxnet/runtime/c_runtime_api.h | 18 +
include/mxnet/runtime/container.h | 2 +-
.../mxnet/runtime/ndarray_handle.h | 51 +-
include/mxnet/runtime/packed_func.h | 11 +-
include/mxnet/storage.h | 11 +-
mkldnn.mk | 1 +
perl-package/AI-MXNet/lib/AI/MXNet/Gluon/Block.pm | 7 +-
perl-package/AI-MXNetCAPI/mxnet.i | 2 +
python/mxnet/__init__.py | 3 +
python/mxnet/_deferred_compute.py | 106 ++++
python/mxnet/_ffi/_ctypes/function.py | 23 +-
python/mxnet/_ffi/_ctypes/object.py | 47 +-
python/mxnet/_ffi/_cython/base.pxi | 2 +
python/mxnet/_ffi/_cython/core.pyx | 1 +
python/mxnet/_ffi/_cython/function.pxi | 26 +
python/mxnet/_ffi/_cython/object.pxi | 98 ++++
python/mxnet/_ffi/object.py | 101 +++-
python/mxnet/_numpy_op_doc.py | 165 ------
python/mxnet/container.py | 54 ++
python/mxnet/gluon/block.py | 212 ++++++-
python/mxnet/gluon/parameter.py | 8 +-
python/mxnet/gluon/utils.py | 24 +-
python/mxnet/ndarray/ndarray.py | 11 +-
python/mxnet/ndarray/numpy/_op.py | 626 +++++++++++++++++----
python/mxnet/ndarray/numpy/linalg.py | 3 +-
python/mxnet/ndarray/numpy/random.py | 110 +++-
python/mxnet/ndarray/sparse.py | 1 +
python/mxnet/numpy/multiarray.py | 496 +++++++++++++++-
python/mxnet/numpy/random.py | 95 +++-
python/mxnet/numpy_dispatch_protocol.py | 5 +
python/mxnet/operator.py | 8 +-
python/mxnet/optimizer/updater.py | 4 +-
python/mxnet/profiler.py | 52 ++
python/mxnet/symbol/numpy/_symbol.py | 301 +++++++++-
python/mxnet/symbol/numpy/random.py | 93 ++-
python/mxnet/symbol/register.py | 11 +
python/mxnet/symbol/symbol.py | 34 +-
.../np_gesvd.cc} | 24 +-
.../numpy/{np_init_op.cc => np_bincount_op.cc} | 51 +-
.../numpy/np_broadcast_reduce_op_boolean.cc | 93 +++
.../operator/numpy/np_broadcast_reduce_op_index.cc | 98 ++++
..._init_op.cc => np_broadcast_reduce_op_value.cc} | 34 +-
.../{np_elemwise_broadcast_op.cc => np_diff_op.cc} | 28 +-
src/api/operator/numpy/np_ediff1d_op.cc | 75 +++
.../numpy/{np_init_op.cc => np_einsum_op.cc} | 57 +-
src/api/operator/numpy/np_elemwise_broadcast_op.cc | 52 ++
...cc => np_elemwise_broadcast_op_extended_sec.cc} | 27 +-
src/api/operator/numpy/np_init_op.cc | 65 +++
.../{np_elemwise_broadcast_op.cc => np_kron.cc} | 19 +-
src/api/operator/numpy/np_matrix_op.cc | 220 ++++++++
...np_elemwise_broadcast_op.cc => np_memory_op.cc} | 18 +-
src/api/operator/numpy/np_nan_to_num_op.cc | 72 +++
...p_elemwise_broadcast_op.cc => np_nonzero_op.cc} | 20 +-
src/api/operator/numpy/np_percentile_op.cc | 98 ++++
...lemwise_broadcast_op.cc => np_polynomial_op.cc} | 21 +-
src/api/operator/numpy/np_tensordot_op.cc | 8 +-
.../{np_elemwise_broadcast_op.cc => np_tril_op.cc} | 27 +-
.../numpy/{np_init_op.cc => np_unique_op.cc} | 46 +-
src/api/operator/numpy/np_where_op.cc | 101 ++++
src/api/operator/numpy/random/np_laplace_op.cc | 96 ++++
src/api/operator/op_utils.cc | 23 +-
src/api/operator/op_utils.h | 3 +-
src/api/operator/utils.cc | 21 +-
src/c_api/c_api.cc | 227 ++++++--
src/c_api/c_api_ndarray.cc | 60 +-
src/c_api/c_api_profile.cc | 19 +
src/c_api/c_api_symbolic.cc | 84 ++-
src/common/cuda_utils.h | 19 +
src/common/utils.h | 20 +-
src/executor/eliminate_common_expr_pass.cc | 2 +-
src/executor/graph_executor.cc | 36 +-
src/executor/infer_graph_attr_pass.cc | 4 +-
src/imperative/cached_op.cc | 43 +-
src/imperative/cached_op.h | 12 +-
src/imperative/cached_op_threadsafe.cc | 3 -
src/imperative/imperative.cc | 199 ++++++-
src/imperative/imperative_utils.h | 70 ++-
src/io/iter_image_recordio_2.cc | 7 +
src/kvstore/comm.h | 19 +-
src/kvstore/comm_tree.h | 12 +-
src/kvstore/kvstore_local.h | 22 +
src/ndarray/ndarray.cc | 148 ++++-
src/nnvm/legacy_json_util.cc | 2 +-
src/operator/contrib/amp_graph_pass.cc | 2 +-
src/operator/contrib/fft-inl.h | 4 -
src/operator/contrib/ifft-inl.h | 4 -
src/operator/contrib/multi_lamb-inl.h | 12 +-
src/operator/cross_device_copy.cc | 2 +-
src/operator/fusion/fused_op-inl.h | 5 +-
src/operator/linalg_impl.h | 47 +-
src/operator/mshadow_op.h | 50 ++
src/operator/nn/batch_norm.cc | 5 +-
src/operator/nn/concat-inl.h | 7 +
src/operator/nn/cudnn/cudnn_convolution-inl.h | 5 +-
src/operator/nn/cudnn/cudnn_deconvolution-inl.h | 5 +-
src/operator/nn/layer_norm.cc | 2 +-
src/operator/nn/mkldnn/mkldnn_base.cc | 6 +-
src/operator/nn/mkldnn/mkldnn_rnn-inl.h | 1 -
src/operator/numpy/np_bincount_op-inl.h | 8 +
src/operator/numpy/np_broadcast_reduce_op.h | 7 +
.../numpy/np_broadcast_reduce_op_boolean.cc | 4 +-
.../numpy/np_broadcast_reduce_op_boolean.cu | 4 +-
src/operator/numpy/np_cumsum-inl.h | 2 +-
src/operator/numpy/np_cumsum.cc | 1 +
src/operator/numpy/np_diff-inl.h | 10 +-
src/operator/numpy/np_ediff1d_op-inl.h | 13 +
src/operator/numpy/np_einsum_op-inl.h | 9 +
.../numpy/np_elemwise_broadcast_logic_op.cc | 6 +
.../numpy/np_elemwise_broadcast_op_extended_sec.cc | 142 +++++
.../numpy/np_elemwise_broadcast_op_extended_sec.cu | 77 +++
src/operator/numpy/np_init_op.h | 8 +
src/operator/numpy/np_kron-inl.h | 322 +++++++++++
src/operator/numpy/np_kron.cc | 94 ++++
.../op_utils.h => operator/numpy/np_kron.cu} | 22 +-
src/operator/numpy/np_matrix_op-inl.h | 62 +-
src/operator/numpy/np_matrix_op.cc | 16 +-
src/operator/numpy/np_matrix_op.cu | 10 +-
src/operator/numpy/np_percentile_op-inl.h | 12 +
src/operator/numpy/np_tril_op-inl.h | 6 +
src/operator/numpy/np_unique_op.h | 11 +
src/operator/numpy/np_where_op-inl.h | 12 +
src/operator/numpy/random/np_laplace_op.cc | 69 +++
.../numpy/random/np_laplace_op.cu} | 20 +-
src/operator/numpy/random/np_laplace_op.h | 231 ++++++++
src/operator/operator_tune-inl.h | 4 +
src/operator/operator_tune.cc | 4 +
src/operator/operator_tune.h | 4 +
.../mkldnn/mkldnn_quantized_elemwise_add.cc | 2 +-
src/operator/quantization/quantize_graph_pass.cc | 4 +-
.../quantization/quantized_elemwise_mul.cc | 2 +-
src/operator/random/shuffle_op.cc | 6 +-
src/operator/rnn-inl.h | 4 +
src/operator/subgraph/build_subgraph.cc | 13 +-
.../mkldnn_post_quantize_align_scale_property.h | 3 +-
.../partitioner/custom_subgraph_property.h | 247 +++++++-
src/operator/subgraph/subgraph_property.h | 9 +
src/operator/tensor/broadcast_reduce_op.h | 12 +
src/operator/tensor/broadcast_reduce_op_value.cc | 1 +
src/operator/tensor/elemwise_unary_op.h | 12 +
src/operator/tensor/init_op.h | 17 +-
src/operator/tensor/matrix_op-inl.h | 18 +
src/profiler/profiler.cc | 18 +
src/profiler/profiler.h | 14 +-
src/profiler/storage_profiler.cc | 130 +++++
src/profiler/storage_profiler.h | 95 +++-
src/resource.cc | 30 +-
src/runtime/container.cc | 72 +++
.../ndarray_handle.cc} | 29 +-
src/runtime/object.cc | 14 +
src/runtime/object_internal.h | 9 +
src/storage/gpu_device_storage.h | 8 +-
src/storage/pinned_memory_storage.h | 5 +
src/storage/pooled_storage_manager.h | 13 +-
src/storage/storage.cc | 2 +-
tests/cpp/include/test_core_op.h | 1 +
tests/cpp/include/test_op.h | 1 +
tests/cpp/include/test_util.h | 2 +-
tests/cpp/misc/libinfo_test.cc | 2 +-
tests/cpp/operator/mkldnn_test.cc | 2 +-
tests/cpp/storage/storage_test.cc | 4 +-
tests/nightly/test_np_random.py | 19 +
.../python/gpu/test_deferred_compute_gpu.py | 36 +-
tests/python/gpu/test_fusion.py | 3 +
tests/python/gpu/test_gluon_gpu.py | 14 +-
tests/python/unittest/test_deferred_compute.py | 536 ++++++++++++++++++
tests/python/unittest/test_extensions.py | 14 +
tests/python/unittest/test_gluon.py | 4 +-
tests/python/unittest/test_gluon_data.py | 6 +
.../python/unittest/test_numpy_interoperability.py | 41 ++
tests/python/unittest/test_numpy_op.py | 198 ++++++-
tests/python/unittest/test_profiler.py | 156 ++++-
tests/python/unittest/test_subgraph_op.py | 4 +-
tests/python/unittest/test_symbol.py | 3 +-
tests/python/unittest/test_thread_local.py | 1 +
tools/dependencies/libturbojpeg.sh | 2 +-
tools/pip/doc/CPU_ADDITIONAL.md | 7 +
tools/pip/doc/CU100_ADDITIONAL.md | 7 +
tools/pip/doc/CU101_ADDITIONAL.md | 7 +
tools/pip/doc/CU102_ADDITIONAL.md | 7 +
tools/pip/doc/CU90_ADDITIONAL.md | 9 +
tools/pip/doc/CU92_ADDITIONAL.md | 7 +
tools/pip/doc/NATIVE_ADDITIONAL.md | 7 +
tools/pip/doc/PYPI_README.md | 1 -
tools/pip/setup.py | 21 +-
239 files changed, 10014 insertions(+), 1556 deletions(-)
create mode 100644 benchmark/opperf/nd_operations/array_manipulation_operators.py
create mode 100644 benchmark/python/ffi/benchmark_ffi.py
create mode 100644 cmake/Modules/FindNVML.cmake
create mode 100644 config/distribution/linux_native.cmake
copy example/extensions/lib_custom_op/Makefile => cpp-package/example/inference/CMakeLists.txt (72%)
create mode 100644 example/extensions/lib_custom_op/test_transposecsr.py
create mode 100644 example/extensions/lib_custom_op/test_transposerowsp.py
create mode 100644 example/extensions/lib_custom_op/transposecsr_lib.cc
create mode 100644 example/extensions/lib_custom_op/transposerowsp_lib.cc
copy src/runtime/object_internal.h => include/mxnet/runtime/ndarray_handle.h (53%)
create mode 100644 python/mxnet/_deferred_compute.py
create mode 100644 python/mxnet/_ffi/_cython/object.pxi
create mode 100644 python/mxnet/container.py
copy src/api/operator/numpy/{np_elemwise_broadcast_op.cc => linalg/np_gesvd.cc} (67%)
copy src/api/operator/numpy/{np_init_op.cc => np_bincount_op.cc} (52%)
create mode 100644 src/api/operator/numpy/np_broadcast_reduce_op_boolean.cc
create mode 100644 src/api/operator/numpy/np_broadcast_reduce_op_index.cc
copy src/api/operator/numpy/{np_init_op.cc => np_broadcast_reduce_op_value.cc} (60%)
copy src/api/operator/numpy/{np_elemwise_broadcast_op.cc => np_diff_op.cc} (62%)
create mode 100644 src/api/operator/numpy/np_ediff1d_op.cc
copy src/api/operator/numpy/{np_init_op.cc => np_einsum_op.cc} (51%)
copy src/api/operator/numpy/{np_elemwise_broadcast_op.cc => np_elemwise_broadcast_op_extended_sec.cc} (57%)
copy src/api/operator/numpy/{np_elemwise_broadcast_op.cc => np_kron.cc} (70%)
create mode 100644 src/api/operator/numpy/np_matrix_op.cc
copy src/api/operator/numpy/{np_elemwise_broadcast_op.cc => np_memory_op.cc} (70%)
create mode 100644 src/api/operator/numpy/np_nan_to_num_op.cc
copy src/api/operator/numpy/{np_elemwise_broadcast_op.cc => np_nonzero_op.cc} (74%)
create mode 100644 src/api/operator/numpy/np_percentile_op.cc
copy src/api/operator/numpy/{np_elemwise_broadcast_op.cc => np_polynomial_op.cc} (68%)
copy src/api/operator/numpy/{np_elemwise_broadcast_op.cc => np_tril_op.cc} (64%)
copy src/api/operator/numpy/{np_init_op.cc => np_unique_op.cc} (56%)
create mode 100644 src/api/operator/numpy/np_where_op.cc
create mode 100644 src/api/operator/numpy/random/np_laplace_op.cc
create mode 100644 src/operator/numpy/np_elemwise_broadcast_op_extended_sec.cc
create mode 100644 src/operator/numpy/np_elemwise_broadcast_op_extended_sec.cu
create mode 100644 src/operator/numpy/np_kron-inl.h
create mode 100644 src/operator/numpy/np_kron.cc
copy src/{api/operator/op_utils.h => operator/numpy/np_kron.cu} (71%)
create mode 100644 src/operator/numpy/random/np_laplace_op.cc
copy src/{api/operator/op_utils.h => operator/numpy/random/np_laplace_op.cu} (74%)
create mode 100644 src/operator/numpy/random/np_laplace_op.h
create mode 100644 src/profiler/storage_profiler.cc
create mode 100644 src/runtime/container.cc
copy src/{api/operator/numpy/np_elemwise_broadcast_op.cc => runtime/ndarray_handle.cc} (63%)
copy ci/docker/install/centos7_ccache.sh => tests/python/gpu/test_deferred_compute_gpu.py (62%)
mode change 100755 => 100644
create mode 100644 tests/python/unittest/test_deferred_compute.py