Ryan edits 1

.gitignore

@@ -10,3 +10,4 @@ coverage
 docs/build/
 env
 node_modules
+LocalPreferences.toml

Project.toml

@@ -4,22 +4,31 @@ version = "0.1.0"
 authors = ["Ryan Senne <rsenne@bu.edu>"]
 
 [deps]
+ADTypes = "47edcb42-4c32-4615-8424-f2b9edc5f35b"
 AbstractMCMC = "80f14c24-f653-4e6a-9b94-39d6b0f70001"
+CUDA = "052768ef-5323-5732-b1bb-66c8b64840ba"
 DifferentiationInterface = "a0c0ee7d-e4b9-4e03-894e-1c5f64a51d63"
-Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
+Enzyme = "7da242da-08ed-463a-9acd-ee780be4f1d9"
 LinearAlgebra = "37e2e46d-f89d-539d-b4ee-838fcccc9c8e"
-LogExpFunctions = "2ab3a3ac-af41-5b50-aa03-7779005ae688"
 MCMCChains = "c7f686f2-ff18-58e9-bc7b-31028e88f75d"
 Mooncake = "da2b9cff-9c12-43a0-ae48-6db2b0edb7d6"
 Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
 Statistics = "10745b16-79ce-11e8-11f9-7d13ad32a3b2"
 
+[weakdeps]
+LogDensityProblems = "6fdf6af0-433a-55f7-b3ed-c6c6e0b8df7c"
+
+[extensions]
+LogDensityProblemsExt = "LogDensityProblems"
+
 [compat]
+ADTypes = "1.21.0"
 AbstractMCMC = "5.10.0"
+CUDA = "5.11.0"
 DifferentiationInterface = "0.7.13"
-Distributions = "0.25.122"
+Enzyme = "0.13.131"
 LinearAlgebra = "1.12.0"
-LogExpFunctions = "0.3.29"
+LogDensityProblems = "2"
 MCMCChains = "7.7.0"
 Mooncake = "0.4.192"
 Random = "1.11.0"

ext/LogDensityProblemsExt.jl

@@ -0,0 +1,54 @@
+module LogDensityProblemsExt
+
+using ParallelMCMC
+import LogDensityProblems
+
+"""
+    DensityModel(ld)
+
+Construct a `DensityModel` from any object implementing the
+[LogDensityProblems](https://github.com/tpapp/LogDensityProblems.jl) interface.
+
+`ld` must support:
+- `LogDensityProblems.capabilities(ld)` returning at least
+  `LogDensityProblems.LogDensityOrder{1}` (i.e. gradient available).
+- `LogDensityProblems.dimension(ld)` → `Int`
+- `LogDensityProblems.logdensity_and_gradient(ld, x)` → `(logp, grad)`
+
+# Turing.jl / DynamicPPL example
+```julia
+using Turing, LogDensityProblems, LogDensityProblemsAD, Mooncake, ParallelMCMC, MCMCChains
+
+@model function mymodel(y)
+    μ ~ Normal(0, 1)
+    y ~ Normal(μ, 0.5)
+end
+
+obs = 1.5
+ld  = DynamicPPL.LogDensityFunction(mymodel(obs))
+ldg = LogDensityProblemsAD.ADgradient(Mooncake.Extras.MooncakeAD(), ld)
+
+model = DensityModel(ldg)
+chain = sample(model, AdaptiveMALASampler(0.3; n_warmup=500), 2_000;
+               chain_type=MCMCChains.Chains, progress=true)
+```
+"""
+function ParallelMCMC.DensityModel(ld)
+    caps = LogDensityProblems.capabilities(ld)
+    caps isa LogDensityProblems.LogDensityOrder{0} &&
+        error("LogDensityProblems model must support gradients (LogDensityOrder{1} or higher). " *
+              "Wrap it with LogDensityProblemsAD.ADgradient first.")
+
+    dim = LogDensityProblems.dimension(ld)
+
+    logp(x) = LogDensityProblems.logdensity(ld, x)
+
+    function gradlogp(x)
+        _, g = LogDensityProblems.logdensity_and_gradient(ld, x)
+        return g
+    end
+
+    return ParallelMCMC.DensityModel(logp, gradlogp, dim)
+end
+
+end # module