Use AbstractPPL AD interface

.github/workflows/CI.yml

@@ -80,3 +80,37 @@ jobs:
           files: lcov.info
           token: ${{ secrets.CODECOV_TOKEN }}
           fail_ci_if_error: true
+
+  reversediff:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v6
+
+      - uses: julia-actions/setup-julia@v3
+        with:
+          version: "1"
+
+      - uses: julia-actions/cache@v3
+
+      - name: Run AD with ReverseDiff on demo models
+        working-directory: test/ext/DynamicPPLReverseDiffExt
+        run: |
+          julia --project=. --color=yes -e 'using Pkg; Pkg.instantiate()'
+          julia --project=. --color=yes main.jl
+
+  marginallogdensities:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v6
+
+      - uses: julia-actions/setup-julia@v3
+        with:
+          version: "1"
+
+      - uses: julia-actions/cache@v3
+
+      - name: Run MarginalLogDensities integration tests
+        working-directory: test/ext/DynamicPPLMarginalLogDensitiesExt
+        run: |
+          julia --project=. --color=yes -e 'using Pkg; Pkg.instantiate()'
+          julia --project=. --color=yes main.jl

.github/workflows/Enzyme.yml

@@ -29,7 +29,7 @@ jobs:
       - uses: julia-actions/cache@v3
 
       - name: Run AD with Enzyme on demo models
-        working-directory: test/integration/enzyme
+        working-directory: test/ext/DynamicPPLEnzymeCoreExt
         run: |
           julia --project=. --color=yes -e 'using Pkg; Pkg.instantiate()'
           julia --project=. --color=yes main.jl

AGENTS.md

@@ -14,6 +14,7 @@ DynamicPPL builds on AbstractPPL.jl for shared PPL interfaces such as `VarName`,
 
   - CI also runs Aqua.jl quality checks and doctests.
   - Test files are self-contained: use package imports, not relative imports or `include()`, so they run individually with TestPicker.jl.
+  - Always refresh each environment (`Pkg.update()` / `up`) before tests or doc builds — a stale manifest can cause subtle resolution and loading issues.
   - Formatting is JuliaFormatter v1 (Blue style), enforced by CI:
 
     ```bash

Project.toml

@@ -1,6 +1,6 @@
 name = "DynamicPPL"
 uuid = "366bfd00-2699-11ea-058f-f148b4cae6d8"
-version = "0.41.8"
+version = "0.42"
 
 [deps]
 ADTypes = "47edcb42-4c32-4615-8424-f2b9edc5f35b"
@@ -12,7 +12,6 @@ Bijectors = "76274a88-744f-5084-9051-94815aaf08c4"
 Chairmarks = "0ca39b1e-fe0b-4e98-acfc-b1656634c4de"
 Compat = "34da2185-b29b-5c13-b0c7-acf172513d20"
 ConstructionBase = "187b0558-2788-49d3-abe0-74a17ed4e7c9"
-DifferentiationInterface = "a0c0ee7d-e4b9-4e03-894e-1c5f64a51d63"
 Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
 DocStringExtensions = "ffbed154-4ef7-542d-bbb7-c09d3a79fcae"
 FillArrays = "1a297f60-69ca-5386-bcde-b61e274b549b"
@@ -44,22 +43,21 @@ DynamicPPLComponentArraysExt = ["ComponentArrays"]
 DynamicPPLEnzymeCoreExt = ["EnzymeCore"]
 DynamicPPLForwardDiffExt = ["ForwardDiff"]
 DynamicPPLMCMCChainsExt = ["MCMCChains"]
+DynamicPPLMooncakeExt = ["Mooncake"]
 DynamicPPLMarginalLogDensitiesExt = ["MarginalLogDensities"]
-DynamicPPLMooncakeExt = ["Mooncake", "DifferentiationInterface"]
 DynamicPPLReverseDiffExt = ["ReverseDiff"]
 
 [compat]
 ADTypes = "1"
 AbstractMCMC = "5.14"
-AbstractPPL = "0.14.1"
+AbstractPPL = "0.15"
 Accessors = "0.1"
 BangBang = "0.4.1"
-Bijectors = "0.15.17"
+Bijectors = "0.16"
 Chairmarks = "1.3.1"
 Compat = "4"
 ComponentArrays = "0.15"
 ConstructionBase = "1.5.4"
-DifferentiationInterface = "0.6.41, 0.7"
 Distributions = "0.25"
 DocStringExtensions = "0.9"
 EnzymeCore = "0.6 - 0.8"

benchmarks/Project.toml

@@ -21,12 +21,12 @@ DynamicPPL = {path = ".."}
 
 [compat]
 ADTypes = "1.14.0"
-AbstractPPL = "0.14"
-Bijectors = "0.15.17"
+AbstractPPL = "0.15"
+Bijectors = "0.16"
 Chairmarks = "1.3.1"
 DifferentiationInterface = "0.7"
 Distributions = "0.25.117"
-DynamicPPL = "0.41"
+DynamicPPL = "0.42"
 Enzyme = "0.13"
 ForwardDiff = "1"
 LogDensityProblems = "2.1.2"

docs/Project.toml

@@ -25,18 +25,18 @@ StatsFuns = "4c63d2b9-4356-54db-8cca-17b64c39e42c"
 [compat]
 ADTypes = "1"
 AbstractMCMC = "5"
-AbstractPPL = "0.14"
+AbstractPPL = "0.15"
 Accessors = "0.1"
 BangBang = "0.4"
-Bijectors = "0.15.17"
+Bijectors = "0.16"
 Chairmarks = "1"
 ChangesOfVariables = "0.1"
 DimensionalData = "0.30"
 Distributions = "0.25"
 Documenter = "1"
 DocumenterInterLinks = "1"
 DocumenterMermaid = "0.1, 0.2"
-DynamicPPL = "0.41"
+DynamicPPL = "0.42"
 FillArrays = "0.13, 1"
 ForwardDiff = "0.10, 1"
 LogDensityProblems = "2"

ext/DynamicPPLMarginalLogDensitiesExt.jl

@@ -158,7 +158,7 @@ accs = DynamicPPL.OnlyAccsVarInfo((
     DynamicPPL.RawValueAccumulator(false),
     # ... whatever else you need
 ))
-_, accs = DynamicPPL.init!!(rng, model, oavi, init_strategy, DynamicPPL.UnlinkAll())
+_, accs = DynamicPPL.init!!(rng, model, accs, init_strategy, DynamicPPL.UnlinkAll())
 ```
 
 You can then extract all the updated data from `accs` using DynamicPPL's existing API (see
@@ -192,7 +192,7 @@ retcode: Success
 u: 1-element Vector{Float64}:
  4.88281250001733e-5
 
-julia> # Get the an initialisation strategy representing the mode of `y`.
+julia> # Get an initialisation strategy representing the mode of `y`.
        init_strategy = InitFromVector(mld, opt_solution.u);
 
 julia> # Evaluate the model with this initialisation strategy.

ext/DynamicPPLMooncakeExt.jl

@@ -1,6 +1,7 @@
 module DynamicPPLMooncakeExt
 
 using DynamicPPL: DynamicPPL, is_transformed
+using AbstractPPL: AbstractPPL
 using Mooncake: Mooncake
 
 # These are purely optimisations (although quite significant ones sometimes, especially for
@@ -15,17 +16,21 @@ Mooncake.@zero_derivative Mooncake.DefaultCtx Tuple{
 
 using DynamicPPL: @model, LinkAll, getlogjoint_internal, LogDensityFunction
 using ADTypes: AutoMooncake
-import DifferentiationInterface
 using Distributions: Normal, InverseGamma, Beta
 using PrecompileTools: @setup_workload, @compile_workload
 @setup_workload begin
     @compile_workload begin
-        for dist in (Normal(), InverseGamma(2, 3), Beta(2, 2))
-            @model f() = x ~ dist
-            ldf = LogDensityFunction(
-                f(), getlogjoint_internal, LinkAll(); adtype=AutoMooncake()
-            )
-            DynamicPPL.LogDensityProblems.logdensity_and_gradient(ldf, [0.5])
+        # Julia does not guarantee transitive extensions are loaded while this
+        # extension precompiles, so skip the workload unless Mooncake's
+        # AbstractPPL methods are already available.
+        if !isnothing(Base.get_extension(AbstractPPL, :AbstractPPLMooncakeExt))
+            for dist in (Normal(), InverseGamma(2, 3), Beta(2, 2))
+                @model f() = x ~ dist
+                ldf = LogDensityFunction(
+                    f(), getlogjoint_internal, LinkAll(); adtype=AutoMooncake()
+                )
+                DynamicPPL.LogDensityProblems.logdensity_and_gradient(ldf, [0.5])
+            end
         end
     end
 end