[OPT] Low-bit Quantization (#2116)
authorziheng <ziheng@apache.org>
Thu, 31 Jan 2019 21:53:44 +0000 (13:53 -0800)
committerGitHub <noreply@github.com>
Thu, 31 Jan 2019 21:53:44 +0000 (13:53 -0800)
commit741b6bbe2973e45a537a1c8770220b53487c8fd4
treeddbd554541083600bd4905f690fe906164b2897e
parentda972bdfc783f1931306bb8c327c4e59c7cf0176
[OPT] Low-bit Quantization (#2116)

* [QUANTIZE] Quantization implementation.

* Update.

* Update.

* Update.

* Update.
17 files changed:
include/tvm/relay/attrs/transform.h
include/tvm/relay/op.h
python/tvm/relay/__init__.py
python/tvm/relay/build_module.py
python/tvm/relay/quantize/__init__.py [new file with mode: 0644]
python/tvm/relay/quantize/_annotate.py [new file with mode: 0644]
python/tvm/relay/quantize/_quantize.py [new file with mode: 0644]
python/tvm/relay/quantize/quantize.py [new file with mode: 0644]
src/relay/ir/expr_functor.cc
src/relay/op/tensor/transform.cc
src/relay/op/tensor/unary.cc
src/relay/pass/pattern_util.h
src/relay/pass/quantize.cc [new file with mode: 0644]
src/relay/pass/quantize.h [new file with mode: 0644]
tests/python/relay/test_ad.py [deleted file]
tests/python/relay/test_pass_gradient.py [new file with mode: 0644]
tests/python/relay/test_pass_quantize.py [new file with mode: 0644]