Skip to content

Releases: calad0i/HGQ

v0.2.3

07 Aug 23:55
36d1f2d
Compare
Choose a tag to compare
  • Add PPermute layer.
  • Update install instruction
    • Official hls4ml support since 1.0.

v0.2.2

27 Jun 16:17
e2657ca
Compare
Choose a tag to compare

Change logs:

  • Allow QKeras layers with linear activations / Qkeras QDense w/o bias
  • bump max allowed tensorflow version to 2.16
  • compatibility with np_config.enable_numpy_behavior() is set

v0.2.1

26 Apr 21:53
599a5e4
Compare
Choose a tag to compare

Change logs:

  • Serialize FixedPointQuantizer and UnaryLUT classes, say goodbye to custom_objects when loading.
  • Fix pooling request bw issue that cause overflow in corner cases.

v0.2.0.post1

06 Apr 01:21
e6dc16e
Compare
Choose a tag to compare

Minor dependency requirement change.

As tf 2.15 passes all tests, allow tf >=2.13, <2.16. tf 2.16 is not tested, but may work.

v0.2.0

21 Feb 21:10
253e051
Compare
Choose a tag to compare

Change logs:

  • proper hls4ml integration through proxy-model.
    • bit-accurate emulation of hls4ml firmware up to fp32 precision.
    • zero-hassle conversion. You get what you see in the HGQ model, upon fp32 precision and overflowing.
  • (experimental) supports nested models
  • (limited) support of arbitrary unary activation functions
  • many bug fixes here and there

Full Changelog: 0.1.3...v0.2.0

v0.2.0rc1

02 Dec 02:51
bba76ac
Compare
Choose a tag to compare

Bug fixes and doc update:

Issues with converting Signature layer into proxy-model, PConcatenate layer init check, fixing var name typo, bit-mismatch in edge some cases, and other minor changes.

Back support python 3.10, better dependency requirement.

Full Changelog: v0.2.0b2...v0.2.0rc1

v0.2.0b2

27 Nov 22:22
f042973
Compare
Choose a tag to compare
v0.2.0b2 Pre-release
Pre-release

Fix Qkeras -> proxy-model conversion when QActivation is defined with string activation.

0.2.0b1 - overhaul

27 Nov 06:42
80cf874
Compare
Choose a tag to compare
0.2.0b1 - overhaul Pre-release
Pre-release
  • Use proxy-model for hls4ml conversion, no more monkey and file patches
    • bit-accurate emulation to hls4ml, with or without overflow
  • many bug fixes regarding inaccurate synthesis results
  • Documentation
  • bump to python==3.11 and tensorflow==2.13
  • QKeras -> proxy-model support for bit-accurate synthesis
  • Much more...

v0.1.3

07 Oct 22:17
db46d31
Compare
Choose a tag to compare
  • Fix an issue causing all io_stream models to fail to convert

Minor fixes

07 Oct 19:59
368d053
Compare
Choose a tag to compare
  1. Avoid overflow on some corner cases
  2. Avoid unnecessarily large cover range for bounded activation functions when cover_factor is set
  3. Support json serialization of model architecture