You are viewing a plain text version of this content. The canonical link for it is here.
Posted to commits@mxnet.apache.org by ta...@apache.org on 2020/01/03 04:27:53 UTC
[incubator-mxnet] branch master updated (2a9ec0e -> 89fe1f6)
This is an automated email from the ASF dual-hosted git repository.
taolv pushed a change to branch master
in repository https://gitbox.apache.org/repos/asf/incubator-mxnet.git.
from 2a9ec0e Softmax primitive cache and in-place computation (#17152)
add 89fe1f6 [MKLDNN] Support channel wise quantization for FullyConnected (#17187)
No new revisions were added by this update.
Summary of changes:
include/mxnet/c_api.h | 5 +-
include/mxnet/op_attr_types.h | 4 +-
python/mxnet/contrib/quantization.py | 52 ++-
src/c_api/c_api_symbolic.cc | 3 +
.../nn/mkldnn/mkldnn_fully_connected-inl.h | 7 +-
src/operator/nn/mkldnn/mkldnn_fully_connected.cc | 25 +-
src/operator/quantization/quantize_graph_pass.cc | 22 +-
src/operator/quantization/quantized_batch_norm.cc | 8 +-
src/operator/quantization/quantized_indexing_op.cc | 8 +-
src/operator/subgraph/mkldnn/mkldnn_common.h | 138 ++++++++
src/operator/subgraph/mkldnn/mkldnn_conv.cc | 110 +------
src/operator/subgraph/mkldnn/mkldnn_fc.cc | 358 +++++++++++++++------
tests/python/mkl/test_subgraph.py | 69 ++--
13 files changed, 536 insertions(+), 273 deletions(-)
create mode 100644 src/operator/subgraph/mkldnn/mkldnn_common.h