There is a newer version of the record available.

Published February 28, 2022 | Version v1.21.1
Software Open

JuliaDiff/ChainRules.jl: v1.21.1

Creators

  • 1. @invenia Labs
  • 2. Invenia Labs
  • 3. Beacon Biosignals
  • 4. @mlcolab
  • 5. @mit
  • 6. @JuliaComputing
  • 7. Artificial Intelligence Center, Czech Technical University
  • 8. Uppsala University
  • 9. @invenia
  • 10. Massachusetts Institute of Technology
  • 11. University of Basel
  • 12. Bocconi University
  • 13. Zuse Institute Berlin
  • 14. MIT
  • 15. @beacon-biosignals

Description

ChainRules v1.21.1

Diff since v1.21.0

Closed issues:

  • Rule for exp mutates its input (#512)
  • rrule for eachslice (#556)
  • Do ChainRules support differentiating aggregate types with respect to aggregate types? (#568)
  • ChainRules and ONNX(Runtime) (#580)
  • Rule for float(x), or convert(AbstractArray, x) (#581)
  • Can we define rules for NamedTuple{T}(...), Val{T}(...)? (#584)
  • using AD in rule definitions themselver for implicit function (#585)

Merged pull requests:

  • Use logten and logtwo from IrrationalConstants (#555) (@devmotion)
  • Rules for eachslice (#561) (@mcabbott)
  • Test for things adding rules to DataType/UnionAll (#563) (@oxinabox)
  • Add many frules (#565) (@mcabbott)
  • Some rules for UniformScaling (#571) (@mcabbott)
  • add sparse rrule (#579) (@CarloLucibello)
  • Disable ReversePropagation Integration test (#582) (@oxinabox)
  • Don't mutate cotangent in rule for exp (#588) (@sethaxen)
  • rrule for findnz (#590) (@CarloLucibello)
  • backport #571 (#593) (@mzgubic)

Files

JuliaDiff/ChainRules.jl-v1.21.1.zip

Files (104.2 kB)

Name Size Download all
md5:183fb28a42b3abcabd689c3391c44002
104.2 kB Preview Download

Additional details

Related works

Is supplement to
https://github.com/JuliaDiff/ChainRules.jl/tree/v1.21.1 (URL)