projects / platform / upstream / tvm.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 5357f49) | patch
[Relay][Quantization] KL-divergence-based per-layer calibration (#3538)
authorWuwei Lin <vincentl13x@gmail.com>
Fri, 2 Aug 2019 03:55:27 +0000 (20:55 -0700)
committerziheng <ziheng@apache.org>
Fri, 2 Aug 2019 03:55:27 +0000 (20:55 -0700)
commit33ab3c602874d128d600f9cfa5c23da37a9f1ab2
treefb53a8102521a4166fba8291793be8b4872bafe6tree | snapshot
parent5357f49b0ed02e33faa590aa104760ee370a00eccommit | diff
[Relay][Quantization] KL-divergence-based per-layer calibration (#3538)

* [Relay][Quantization] Support floating-point scale

* [Relay][Quantization] KL-divergence calibration on dataset

* Fix unhandled LeftShift case in QuantizeRealize

* Fix lint

* drop QBias

* fix lint

* address comments

* address comments

* Update comments

* address comments

* lint

* kQIdentity = 0
python/tvm/relay/quantize/__init__.py diff | blob | history
python/tvm/relay/quantize/_annotate.py diff | blob | history
python/tvm/relay/quantize/kl_divergence.py [new file with mode: 0644] blob
python/tvm/relay/quantize/quantize.py diff | blob | history
src/relay/pass/quantize/calibrate.cc [new file with mode: 0644] blob
src/relay/pass/quantize/quantize.cc [moved from src/relay/pass/quantize.cc with 93% similarity] diff | blob | history
src/relay/pass/quantize/quantize.h [moved from src/relay/pass/quantize.h with 89% similarity] diff | blob | history
Domain: Machine Learning / ML Framework;