review.tizen.org Git - platform/upstream/pytorch.git/commit

projects / platform / upstream / pytorch.git / commit

author	Jongsoo Park <jongsoo@fb.com>
	Mon, 10 Dec 2018 09:06:17 +0000 (01:06 -0800)
committer	Facebook Github Bot <facebook-github-bot@users.noreply.github.com>
	Mon, 10 Dec 2018 09:08:21 +0000 (01:08 -0800)
commit	b039a715ce4e9cca82ae3bf72cb84652957b2844
tree	c7807d0219b5b67bd49b8a3e60073f8421ce9207	tree \| snapshot
parent	e747acbebbae2de381ccda8b70010953046c397d	commit \| diff

pre-pack operation of dnnlowp conv with 16-bit accumulation (#14881)

Summary:
Pull Request resolved: https://github.com/pytorch/pytorch/pull/14881

This diff allows us to pre-quantize and pre-pack weight matrix used in DNNLOWP_ACC16 .
The intended use pattern is run Int8ConvPackWeight in init_net that generates a packed weight and Int8Conv with DNNLOWP_ACC16 engine uses the the packed weight.

Reviewed By: csummersea

Differential Revision: D13374662

fbshipit-source-id: dd02b9a4eb7af1fe208aa857fcd0b445e6e395af

26 files changed:

caffe2/quantization/server/CMakeLists.txt		diff \| blob \| history
caffe2/quantization/server/conv_depthwise_dnnlowp_op_test.py		diff \| blob \| history
caffe2/quantization/server/conv_dnnlowp_acc16_op.cc		diff \| blob \| history
caffe2/quantization/server/conv_dnnlowp_acc16_op.h		diff \| blob \| history
caffe2/quantization/server/conv_dnnlowp_acc16_op_test.py		diff \| blob \| history
caffe2/quantization/server/conv_dnnlowp_op.cc		diff \| blob \| history
caffe2/quantization/server/conv_dnnlowp_op.h		diff \| blob \| history
caffe2/quantization/server/conv_dnnlowp_op_test.py		diff \| blob \| history
caffe2/quantization/server/conv_groupwise_dnnlowp_acc16_op_test.py		diff \| blob \| history
caffe2/quantization/server/conv_groupwise_dnnlowp_op_test.py		diff \| blob \| history
caffe2/quantization/server/conv_pool_dnnlowp_op_base.h		diff \| blob \| history
caffe2/quantization/server/dnnlowp_op.h		diff \| blob \| history
caffe2/quantization/server/fbgemm_pack_blob.h	[new file with mode: 0644]	blob
caffe2/quantization/server/fbgemm_pack_op.cc	[new file with mode: 0644]	blob
caffe2/quantization/server/fbgemm_pack_op.h	[new file with mode: 0644]	blob
caffe2/quantization/server/fully_connected_dnnlowp_acc16_op.cc		diff \| blob \| history
caffe2/quantization/server/fully_connected_dnnlowp_acc16_op.h		diff \| blob \| history
caffe2/quantization/server/fully_connected_dnnlowp_acc16_op_test.py		diff \| blob \| history
caffe2/quantization/server/fully_connected_dnnlowp_op.cc		diff \| blob \| history
caffe2/quantization/server/fully_connected_dnnlowp_op.h		diff \| blob \| history
caffe2/quantization/server/fully_connected_dnnlowp_op_test.py		diff \| blob \| history
caffe2/quantization/server/fully_connected_rowwise_dnnlowp_op.cc		diff \| blob \| history
caffe2/quantization/server/fully_connected_rowwise_dnnlowp_op.h		diff \| blob \| history
caffe2/quantization/server/fully_connected_rowwise_dnnlowp_op_test.py		diff \| blob \| history
caffe2/quantization/server/group_norm_dnnlowp_op_test.py		diff \| blob \| history
caffe2/quantization/server/utils.py		diff \| blob \| history

Domain: Machine Learning / ML Framework;

RSS Atom