Skip to content

Releases: FluxML/Zygote.jl

v0.4.4

08 Jan 15:10
v0.4.4
d144c1b
Compare
Choose a tag to compare

v0.4.4 (2020-01-08)

Diff since v0.4.3

Merged pull requests:

v0.4.3

20 Dec 16:33
v0.4.3
727ea3c
Compare
Choose a tag to compare

v0.4.3 (2019-12-20)

Diff since v0.4.2

Merged pull requests:

v0.4.2

20 Dec 12:45
v0.4.2
b954f84
Compare
Choose a tag to compare

v0.4.2 (2019-12-20)

Diff since v0.4.1

Closed issues:

  • I’m sure you know this, but since I wanted to see what goes wrong: Comprehensions with Iterators.ProductIterator don’t seem to work right now, but surely they could be made to. (#420)
  • Remove Flux page? (#416)
  • setindex_shape_check method error for custom adjoint (#403)
  • Warnings on fresh Zygote install (#397)
  • Precompile error: Non-differentiable function mul_int (breaks Turing.jl + others?) (#396)
  • Error when trying to precompile Zygote (#395)
  • ZygoteRules compat (#392)
  • Facilitate array construction without mutation (#377)

Merged pull requests:

v0.4.1

06 Nov 16:18
v0.4.1
6ac5b5b
Compare
Choose a tag to compare

v0.4.1 (2019-11-06)

Diff since v0.4.0

Closed issues:

  • differentiate with π failed (#309)
  • Blocks are not in domtree order (#227)
  • Wrong gradient when using backslash (#222)

Merged pull requests:

v0.4.0

04 Nov 14:18
v0.4.0
b2c3bc4
Compare
Choose a tag to compare

v0.4.0 (2019-11-04)

Diff since v0.3.4

Closed issues:

  • Unbroadcasting is not supported when scalars are involved (#382)
  • Missing adjoint for [ ] in broadcasting (#381)
  • Missing adjoint for Ref (#380)
  • Error in pullback construction (#368)
  • Behavior of real changes when broadcasted (#356)
  • Adjoint of exp inaccurate for complex matrices (#348)
  • Splatting and mutation? (#347)
  • Separate gradcheck (#346)
  • Need convenient utilities for looking inside Grads(...) (#345)
  • Getting rid of type-hacks mode (#341)
  • Type instability on tuple since d74f3c (#322)
  • DimensionMismatch on nested for loop (#320)
  • Zygote.jl tries to back-propagate nothings when branching happens inside broadcasting (#317)
  • Diagonal * Vector adjoint can be faster (#316)
  • Customizing broadcasted yields MethodError on unbroadcast (#313)
  • scalar getindex on gpu in the backwards pass (#261)
  • Can't differentiate cat (#198)
  • ERROR: Can't differentiate simdloop expression (#157)
  • Complex Number Interfaces (#142)
  • Error if a vector or range is indexed (#116)
  • Hessian-vector products (#115)

Merged pull requests:

v0.3.4

20 Aug 13:04
v0.3.4
Compare
Choose a tag to compare

v0.3.4 (2019-08-20)

Diff since v0.3.3

Closed issues:

  • Error loading Zygote 0.3.3 (#306)
  • New release (#302)
  • Poor gradient performance on map reduce (#292)

Merged pull requests:

v0.3.3

19 Aug 14:19
v0.3.3
a207f56
Compare
Choose a tag to compare

v0.3.3 (2019-08-19)

Diff since v0.3.2

Closed issues:

  • Custom type getproperty adjoint (#298)
  • fma not supported (#296)
  • precompile error in Julia 1.0.1 (#287)
  • Double derivatives breaking the compiler (#278)
  • does Zygote support functions with variant input sizes? (#274)
  • DimensionMismatch on branch (#271)
  • Internal error: encountered unexpected error in runtime in toy example (#267)
  • Gradient error when trying to use getindex on size of array (#263)
  • gradient not working with div (#258)
  • Nested differentiation works but the LLVM code looks... really bad (#253)
  • Mixing ForwardDiff and Zygote (#250)

Merged pull requests:

v0.3.2

25 Jun 18:41
v0.3.2
e842cb8
Compare
Choose a tag to compare

v0.3.2 (2019-06-25)

Diff since v0.3.1

Closed issues:

  • Zygote does not go through DiffEq all the way (#243)
  • Failing simple use case (#242)
  • Forward-Over-Reverse HesVec Error (#241)
  • derivative not defined (#239)
  • 31638 bus error julia (#237)
  • Error: Variable not defined (#236)
  • Type Inference for backward pass of broadcasted division fails (#233)
  • Fail to build, expand! required (#224)
  • custom adjoint fails for functions that take tuple as additional argument (#214)
  • A hvcat related issue (#211)
  • Wrong gradient for x -\> x / maximum\(\[x\]\) (#208)
  • LoadError: LoadError: UndefVarError: expand! not defined (#206)
  • Variable %8 in block 7 is not defined. (#201)
  • Incorrect Gradient for a normalization function (#197)
  • complex value gradient is not correct (#196)
  • ERROR: Variable %20 in block 4 is not defined. (#192)
  • relation between builtin FillArray and FillArrays (#190)
  • Failure in Color Optimisation example (#179)
  • Dev Docs (#99)
  • Function with views error when taking gradient (#52)

Merged pull requests:

v0.3.1

09 May 11:03
fd36e6e
Compare
Choose a tag to compare
Update Project.toml

v0.3.0

30 Apr 12:44
Compare
Choose a tag to compare
version, compat, irtools stable