feat: add support for HyperHessian backend (#940)

* WIP: Add support for HyperHessian backend

* only run HyperHessians tests

* less tests

* also have source for HyperHessians

* package is now registerd

* address review comments

* updates

* fixups and use ADTypes HyperHessians

* bump compat for ADTypes

* Add JET to test env

* Docs fixes and version bump

---------

Co-authored-by: KristofferC <kristoffer.carlsson@juliacomputing.com>
Co-authored-by: Guillaume Dalle <22795598+gdalle@users.noreply.github.com>

Author: 3 people

.github/workflows/Test.yml

@@ -100,6 +100,7 @@ jobs:
           - FiniteDifferences
           - ForwardDiff
           - GTPSA
+          - HyperHessians
           - Mooncake
           - Mooncake-old
           - PolyesterForwardDiff

DifferentiationInterface/CHANGELOG.md

@@ -5,7 +5,13 @@ All notable changes to this project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.1.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
-## [Unreleased](https://github.com/JuliaDiff/DifferentiationInterface.jl/compare/DifferentiationInterface-v0.7.17...main)
+## [Unreleased](https://github.com/JuliaDiff/DifferentiationInterface.jl/compare/DifferentiationInterface-v0.7.18...main)
+
+## [0.7.18](https://github.com/JuliaDiff/DifferentiationInterface.jl/compare/DifferentiationInterface-v0.7.17...DifferentiationInterface-v0.7.18)
+
+### Added
+
+- Add support for HyperHessians backend ([#940](https://github.com/JuliaDiff/DifferentiationInterface.jl/pull/940))
 
 ## [0.7.17](https://github.com/JuliaDiff/DifferentiationInterface.jl/compare/DifferentiationInterface-v0.7.16...DifferentiationInterface-v0.7.17)
 

DifferentiationInterface/Project.toml

@@ -1,6 +1,6 @@
 name = "DifferentiationInterface"
 uuid = "a0c0ee7d-e4b9-4e03-894e-1c5f64a51d63"
-version = "0.7.17"
+version = "0.7.18"
 authors = ["Guillaume Dalle", "Adrian Hill"]
 
 [deps]
@@ -15,6 +15,7 @@ Diffractor = "9f5e2b26-1114-432f-b630-d3fe2085c51c"
 Enzyme = "7da242da-08ed-463a-9acd-ee780be4f1d9"
 EnzymeCore = "f151be2c-9106-41f4-ab19-57ee4f262869"
 FastDifferentiation = "eb9bf01b-bf85-4b60-bf87-ee5de06c00be"
+HyperHessians = "06b494a0-c8e0-40cc-ad32-d99506a00a6c"
 FiniteDiff = "6a86dc24-6348-571c-b903-95158fe2bd41"
 FiniteDifferences = "26cc04aa-876d-5657-8c51-4c34ba976000"
 ForwardDiff = "f6369f11-7733-5829-9624-2563aa707210"
@@ -40,6 +41,7 @@ DifferentiationInterfaceFiniteDiffExt = "FiniteDiff"
 DifferentiationInterfaceFiniteDifferencesExt = "FiniteDifferences"
 DifferentiationInterfaceForwardDiffExt = ["ForwardDiff", "DiffResults"]
 DifferentiationInterfaceGPUArraysCoreExt = ["GPUArraysCore", "Adapt"]
+DifferentiationInterfaceHyperHessiansExt = "HyperHessians"
 DifferentiationInterfaceGTPSAExt = "GTPSA"
 DifferentiationInterfaceMooncakeExt = "Mooncake"
 DifferentiationInterfacePolyesterForwardDiffExt = [
@@ -58,7 +60,7 @@ DifferentiationInterfaceZygoteExt = ["Zygote", "ForwardDiff"]
 
 [compat]
 Adapt = "4.5.0"
-ADTypes = "1.18.0"
+ADTypes = "1.22.0"
 ChainRulesCore = "1.23.0"
 DiffResults = "1.1.0"
 Diffractor = "=0.2.6"
@@ -70,6 +72,7 @@ FiniteDifferences = "0.12.31"
 ForwardDiff = "0.10.36,1"
 GPUArraysCore = "0.2"
 GTPSA = "1.4.0"
+HyperHessians = "0.2"
 LinearAlgebra = "1"
 Mooncake = "0.5.1"
 PolyesterForwardDiff = "0.1.2"

DifferentiationInterface/README.md

@@ -34,6 +34,7 @@ We support the following backends defined by [ADTypes.jl](https://github.com/Sci
 - [FiniteDiff.jl](https://github.com/JuliaDiff/FiniteDiff.jl)
 - [FiniteDifferences.jl](https://github.com/JuliaDiff/FiniteDifferences.jl)
 - [ForwardDiff.jl](https://github.com/JuliaDiff/ForwardDiff.jl)
+- [HyperHessians.jl](https://github.com/KristofferC/HyperHessians.jl)
 - [GTPSA.jl](https://github.com/bmad-sim/GTPSA.jl)
 - [Mooncake.jl](https://github.com/chalk-lab/Mooncake.jl)
 - [PolyesterForwardDiff.jl](https://github.com/JuliaDiff/PolyesterForwardDiff.jl)

DifferentiationInterface/docs/Project.toml

@@ -6,6 +6,7 @@ Documenter = "e30172f5-a6a5-5a46-863b-614d45cd2de4"
 DocumenterInterLinks = "d12716ef-a0f6-4df4-a9f1-a5a34e75c656"
 FiniteDiff = "6a86dc24-6348-571c-b903-95158fe2bd41"
 ForwardDiff = "f6369f11-7733-5829-9624-2563aa707210"
+HyperHessians = "06b494a0-c8e0-40cc-ad32-d99506a00a6c"
 Markdown = "d6f4376e-aef5-505a-96c1-9c027394607a"
 PrettyTables = "08abe8d2-0d0c-5749-adfa-8a2ac140af0d"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
@@ -21,6 +22,7 @@ Documenter = "1"
 DocumenterInterLinks = "1.1"
 FiniteDiff = "2.29"
 ForwardDiff = "1.2.2"
+HyperHessians = "0.2"
 PrettyTables = "3.1"
 SparseConnectivityTracer = "1.1.2"
 SparseMatrixColorings = "0.4.23"

DifferentiationInterface/docs/src/explanation/backends.md

@@ -12,6 +12,7 @@ We support the following dense backend choices from [ADTypes.jl](https://github.
 - [`AutoFiniteDifferences`](@extref ADTypes.AutoFiniteDifferences)
 - [`AutoForwardDiff`](@extref ADTypes.AutoForwardDiff)
 - [`AutoGTPSA`](@extref ADTypes.AutoGTPSA)
+- [`AutoHyperHessians`](@extref ADTypes.AutoHyperHessians)
 - [`AutoMooncake`](@extref ADTypes.AutoMooncake) and [`AutoMooncakeForward`](@extref ADTypes.AutoMooncake) (the latter is experimental)
 - [`AutoPolyesterForwardDiff`](@extref ADTypes.AutoPolyesterForwardDiff)
 - [`AutoReverseDiff`](@extref ADTypes.AutoReverseDiff)
@@ -32,11 +33,11 @@ In practice, many AD backends have custom implementations for high-level operato
 !!! details
 
     In the rough summary table below,
-    
+
       - ✅ means that we reuse the custom implementation from the backend;
       - ❌ means that a custom implementation doesn't exist, so we use our default fallbacks;
       - 🔀 means it's complicated or not done yet.
-    
+
     |                            | `pf` | `pb` | `der` | `grad` | `jac` | `hess` | `hvp` | `der2` |
     |:-------------------------- |:---- |:---- |:----- |:------ |:----- |:------ |:----- |:------ |
     | `AutoChainRules`           | ❌    | ✅    | ❌     | ❌      | ❌     | ❌      | ❌     | ❌      |
@@ -48,6 +49,7 @@ In practice, many AD backends have custom implementations for high-level operato
     | `AutoFiniteDifferences`    | 🔀    | ❌    | ❌     | ✅      | ✅     | ❌      | ❌     | ❌      |
     | `AutoForwardDiff`          | ✅    | ❌    | ✅     | ✅      | ✅     | ✅      | ✅     | ✅      |
     | `AutoGTPSA`                | ✅    | ❌    | ❌     | ✅      | ✅     | ✅      | ✅     | ✅      |
+    | `AutoHyperHessians`        | ❌    | ❌    | ❌     | ❌      | ❌     | ✅      | ✅     | ✅      |
     | `AutoMooncake`             | ❌    | ✅    | ❌     | ❌      | ❌     | ❌      | ❌     | ❌      |
     | `AutoMooncakeForward`      | ✅    | ❌    | ❌     | ❌      | ❌     | ❌      | ❌     | ❌      |
     | `AutoPolyesterForwardDiff` | 🔀    | ❌    | 🔀     | ✅      | ✅     | 🔀      | 🔀     | 🔀      |
@@ -69,6 +71,7 @@ Moreover, each context type is supported by a specific subset of backends:
 | `AutoFiniteDifferences`    | ✅                  | ✅               |
 | `AutoForwardDiff`          | ✅                  | ✅               |
 | `AutoGTPSA`                | ✅                  | ❌               |
+| `AutoHyperHessians`        | ✅                  | ✅               |
 | `AutoMooncake`             | ✅                  | ✅               |
 | `AutoMooncakeForward`      | ✅                  | ✅               |
 | `AutoPolyesterForwardDiff` | ✅                  | ✅               |