You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@mxnet.apache.org by bg...@apache.org on 2022/03/31 07:45:07 UTC
[incubator-mxnet] branch master updated (4672ca1 -> e8ff13c)
This is an automated email from the ASF dual-hosted git repository.
bgawrych pushed a change to branch master
in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git.
from 4672ca1 [FEATURE] Add oneDNN support for npi: exp, square and sqrt (#20981)
add e8ff13c Improve AMP, bf16 support. Support oneDNN ops in AMP (#20753)
No new revisions were added by this update.
Summary of changes:
.github/workflows/os_x_mklbuild.yml | 2 +-
include/mxnet/c_api.h | 70 +--
include/mxnet/runtime/packed_func.h | 4 +
python/mxnet/amp/amp.py | 369 +++++-------
python/mxnet/amp/lists/symbol_bf16.py | 223 +------
python/mxnet/executor.py | 12 +-
python/mxnet/gluon/block.py | 13 +-
python/mxnet/gluon/nn/basic_layers.py | 4 +-
python/mxnet/gluon/parameter.py | 19 +-
python/mxnet/ndarray/__init__.py | 2 +-
python/mxnet/ndarray/ndarray.py | 51 +-
python/mxnet/ndarray/numpy/_op.py | 31 +-
python/mxnet/ndarray/numpy/random.py | 11 +-
python/mxnet/ndarray/numpy_extension/_op.py | 7 +-
python/mxnet/ndarray/register.py | 15 +-
python/mxnet/ndarray/sparse.py | 10 +-
python/mxnet/numpy/multiarray.py | 10 +-
python/mxnet/operator.py | 6 +-
python/mxnet/rtc.py | 8 +-
python/mxnet/symbol/register.py | 16 +-
python/mxnet/symbol/symbol.py | 37 +-
python/mxnet/test_utils.py | 6 +-
src/c_api/c_api_symbolic.cc | 222 ++-----
src/nnvm/amp_infer_unknown.cc | 151 -----
src/nnvm/low_precision_pass.cc | 647 ++++++++++-----------
src/operator/nn/dnnl/dnnl_convolution-inl.h | 4 +
src/operator/nn/dnnl/dnnl_fully_connected-inl.h | 4 +
src/operator/subgraph/dnnl/dnnl_conv.cc | 10 +-
src/operator/subgraph/dnnl/dnnl_fc.cc | 10 +-
.../subgraph/dnnl/dnnl_post_amp_property.h | 152 +++++
.../subgraph/dnnl/dnnl_subgraph_property.cc | 4 +
src/operator/subgraph/dnnl/dnnl_transformer-inl.h | 5 +
src/operator/subgraph/dnnl/dnnl_transformer.cc | 71 ++-
src/operator/tensor/amp_cast.h | 17 +-
src/serialization/cnpy.cc | 4 +-
tests/python/dnnl/subgraphs/test_amp_subgraph.py | 243 ++++++++
.../python/dnnl/subgraphs/test_matmul_subgraph.py | 53 +-
tests/python/dnnl/test_amp.py | 217 ++++---
tests/python/dnnl/test_bf16_operator.py | 64 +-
tests/python/gpu/test_amp.py | 139 ++---
tests/python/unittest/test_ndarray.py | 3 +-
41 files changed, 1398 insertions(+), 1548 deletions(-)
delete mode 100644 src/nnvm/amp_infer_unknown.cc
create mode 100644 src/operator/subgraph/dnnl/dnnl_post_amp_property.h
create mode 100644 tests/python/dnnl/subgraphs/test_amp_subgraph.py