Releases: SciML/SciMLSensitivity.jl
Releases · SciML/SciMLSensitivity.jl
v7.0.1
SciMLSensitivity v7.0.1
v7.0.0
SciMLSensitivity v7.0.0
Closed issues:
- Minimisation of a mixed-type loss function (#544)
- Still problems with the docu (#641)
- UndefVarError: solve_up not defined (#628)
- Stable documentation links return 404 (#642)
- "Neural ODEs on GPUs" example not working (#646)
- tutorials are not following the SciML style (#658)
Merged pull requests:
- swap signs adjoint (#545) (@ArnoStrouwen)
- Don't allow Zygote to return nothing by default (#633) (@ChrisRackauckas)
- Update the tutorial of Neural Graph ODE (#635) (@MilkshakeForReal)
- more docpages converted to examples (#636) (@ChrisRackauckas)
- Mutation robust VJPs for SimpleChains (#638) (@ChrisRackauckas)
- Update Downstream.yml (#639) (@MartinuzziFrancesco)
- try a few more docs example building (#643) (@ChrisRackauckas)
- Fix tests for out-of-place constant requirement (#644) (@ChrisRackauckas)
- fix gpu tests for type constancy (#645) (@ChrisRackauckas)
- more doctests (#647) (@ChrisRackauckas)
- Doctests for training tips (#648) (@Abhishek-1Bhatt)
- Even more docs clean up (#649) (@ChrisRackauckas)
- make PEM example run in docs (#650) (@baggepinnen)
- [ci-skip] Redo badges (#651) (@ChrisRackauckas)
- remove some hard-coded images (#652) (@baggepinnen)
- Change to GPUArraysCore (#654) (@ChrisRackauckas)
- Enable coverage statistics on buildkite runs (#656) (@ChrisRackauckas)
- Allow for mixed-type cost contributions in continuous adjoints (#657) (@frankschae)
- some more doctests (#659) (@ChrisRackauckas)
- add doctests for Bayesian (#660) (@ChrisRackauckas)
- More doctests: SDE control tutorials (#662) (@frankschae)
- Simplify Lux multiple NN example (#664) (@ChrisRackauckas)
- Rename and set as a breaking change (#665) (@ChrisRackauckas)
v6.79.0
DiffEqSensitivity v6.79.0
Closed issues:
- EnsembleProblems and equivalent for loops lead to different Zygote gradients (#611)
- Better error message for non-mutating ODEForwardSensitivityProblem (#624)
Merged pull requests:
- consolidate automated vjp choice and throw more warnings (#553) (@ChrisRackauckas)
- remove
noise=..
in solve (#621) (@frankschae) - enable strict doctesting (#622) (@ChrisRackauckas)
- more doctests (#623) (@ChrisRackauckas)
- Better error message for out of place ODEForwardSensitivityProblem (#625) (@ChrisRackauckas)
- some fixes (#626) (@Abhishek-1Bhatt)
- Allow for alternative AD front-ends (#632) (@ChrisRackauckas)
v6.78.5
DiffEqSensitivity v6.78.5
Merged pull requests:
- Improve SteadyStateAdjoint linear solver handling (#617) (@ChrisRackauckas)
- start slapping some doctests around (#619) (@ChrisRackauckas)
- Fix SDE reversediff regression (#620) (@ChrisRackauckas)
v6.78.4
DiffEqSensitivity v6.78.4
Merged pull requests:
- Handle all of the probs in TrackerAdjoint (#616) (@ChrisRackauckas)
v6.78.3
DiffEqSensitivity v6.78.3
v6.78.2
DiffEqSensitivity v6.78.2
Closed issues:
- Neural ODEs + Implicit methods + TrackerAdjoint Fail (#608)
- Hybrid Differential Equations (example) not working for implicit solvers (at the example of Rosenbrock23) (#579)
- Precompilation warning: method definition setindex() overwritten (#603)
UndefVarError: VecJacOperator not defined
thrown while working withSteadyStateAdjointProblem
(#605)
Merged pull requests:
- Test and fix
VectorContinuousCallback
(#602) (@frankschae) - Test for using callbacks within
ForwardDiffSensitivity
(#606) (@frankschae) - Document TrackerAdjoints issues with ForwardDiff (#614) (@ChrisRackauckas)
- Direct definition of isinplace in TrackerAdjoint (#615) (@ChrisRackauckas)
v6.78.1
DiffEqSensitivity v6.78.1
Merged pull requests:
- Partially revert removal of DiffEqSensitivity from deps (#604) (@DhairyaLGandhi)
v6.78.0
DiffEqSensitivity v6.78.0
Closed issues:
- Callbacks with initial values as a matrix (#555)
Merged pull requests:
- Reshape state of BacksolveAdjoint to match shape of forward pass (#565) (@frankschae)
- replaced deprecated functions (#592) (@Abhishek-1Bhatt)
- split callback tests (#600) (@frankschae)
- No arch on GPU (#601) (@ChrisRackauckas)
v6.77.0
DiffEqSensitivity v6.77.0
Merged pull requests:
- Fix BacksolveAdjoint() for implicit solvers (#580) (@frankschae)
- NILSS fix to reduce memory allocations and remove initial time step estimate (#593) (@vleon1234)
- add tests from DiffEqFlux (#594) (@ChrisRackauckas)
- Add high level interpretable error messages for parameter compatability (#596) (@ChrisRackauckas)
- GPU test fixes (#597) (@ChrisRackauckas)
- Remove old cruft and reduce dependences (#598) (@ChrisRackauckas)