CompatHelper: bump compat for AbstractPPL to 0.6 for package test, (k…

…eep existing compat) (#469)

* Fixed a typo in tutorial (#451)

* CompatHelper: bump compat for Turing to 0.24 for package turing, (keep existing compat) (#450)

This pull request changes the compat entry for the `Turing` package from `0.21` to `0.21, 0.24` for package turing.
This keeps the compat entries for earlier versions.



Note: I have not tested your package with this new compat entry.
It is your responsibility to make sure that your package tests pass before you merge this pull request.

Co-authored-by: Hong Ge <[email protected]>

* Some minor utility improvements (#452)

This PR does the following:
- Moves the `varname_leaves` from `TestUtils` to main module.
  - It can be very useful in Turing.jl for constructing `Chains` and the like, so I think it's a good idea to make it part of the main module rather than keeping it "hidden" there.
- Makes the default `varinfo` in the constructor of `LogDensityFunction` be `model.context` rather than a new `DynamicPPL.DefaultContext`.
  - The `context` pass to `evaluate!!` will override the leaf-context in `model.context`, and so the current default constructor always uses `DefaultContext` as the leaf-context, even if the `Model` has been `contextualize`d with some other leaf-context, e.g. `PriorContext`. This PR fixes this issue.

* Always run CI  (#453)

I find the current `bors` workflow a bit tedious. Most of the time, I summon `bors` to see the CI results (see e.g. #438). Given that most `CI` tests are quick (< 10mins), we can always run them by default. 

The most time-consuming `IntegrationTests` is still run by `bors` to avoid excessive CI runs.

* Compat with new Bijectors.jl (#454)

This PR makes DPPL compatible with the changes to come in TuringLang/Bijectors.jl#214.

Tests are passing locally.

Closes #455 Closes #456

* Another Bijectors.jl compat bound bump (#457)

* CompatHelper: bump compat for MCMCChains to 6 for package test, (keep existing compat) (#467)

This pull request changes the compat entry for the `MCMCChains` package from `4.0.4, 5` to `4.0.4, 5, 6` for package test.
This keeps the compat entries for earlier versions.



Note: I have not tested your package with this new compat entry.
It is your responsibility to make sure that your package tests pass before you merge this pull request.

Co-authored-by: Hong Ge <[email protected]>

* CompatHelper: bump compat for AbstractPPL to 0.6 for package test, (keep existing compat)

---------

Co-authored-by: Hong Ge <[email protected]>
Co-authored-by: github-actions[bot] <[email protected]>
Co-authored-by: Tor Erlend Fjelde <[email protected]>

.github/workflows/CI.yml

@@ -9,6 +9,9 @@ on:
       - trying
       # Build the master branch.
       - master
+  pull_request:
+    branches:
+      - master
 
 jobs:
   test:

Project.toml

@@ -1,6 +1,6 @@
 name = "DynamicPPL"
 uuid = "366bfd00-2699-11ea-058f-f148b4cae6d8"
-version = "0.21.5"
+version = "0.22.1"
 
 [deps]
 AbstractMCMC = "80f14c24-f653-4e6a-9b94-39d6b0f70001"
@@ -24,7 +24,7 @@ ZygoteRules = "700de1a5-db45-46bc-99cf-38207098b444"
 AbstractMCMC = "2, 3.0, 4"
 AbstractPPL = "0.5.3, 0.6"
 BangBang = "0.3"
-Bijectors = "0.5.2, 0.6, 0.7, 0.8, 0.9, 0.10"
+Bijectors = "0.11, 0.12"
 ChainRulesCore = "0.9.7, 0.10, 1"
 ConstructionBase = "1"
 Distributions = "0.23.8, 0.24, 0.25"

docs/src/tutorials/prob-interface.md

@@ -20,7 +20,7 @@ end
 nothing # hide
 ```
 
-We generate some data using `μ = 0` and `σ = 1`:
+We generate some data using `μ = 0`:
 
 ```@example probinterface
 Random.seed!(1776)
@@ -35,7 +35,7 @@ Conditioning takes a variable and fixes its value as known.
 We do this by passing a model and a collection of conditioned variables to [`|`](@ref) or its alias [`condition`](@ref):
 
 ```@example probinterface
-model = gdemo(length(dataset)) | (x=dataset, μ=0, σ=1)
+model = gdemo(length(dataset)) | (x=dataset, μ=0)
 nothing # hide
 ```
 

src/abstract_varinfo.jl

@@ -405,7 +405,7 @@ end
 
 # Vector-based ones.
 function link!!(
-    t::StaticTransformation{<:Bijectors.Bijector{1}},
+    t::StaticTransformation{<:Bijectors.Transform},
     vi::AbstractVarInfo,
     spl::AbstractSampler,
     model::Model,
@@ -420,7 +420,7 @@ function link!!(
 end
 
 function invlink!!(
-    t::StaticTransformation{<:Bijectors.Bijector{1}},
+    t::StaticTransformation{<:Bijectors.Transform},
     vi::AbstractVarInfo,
     spl::AbstractSampler,
     model::Model,
@@ -452,9 +452,8 @@ julia> using DynamicPPL, Distributions, Bijectors
 julia> @model demo() = x ~ Normal()
 demo (generic function with 2 methods)
 
-julia> # By subtyping `Bijector{1}`, we inherit the `(inv)link!!` defined for
-       # bijectors which acts on 1-dimensional arrays, i.e. vectors.
-       struct MyBijector <: Bijectors.Bijector{1} end
+julia> # By subtyping `Transform`, we inherit the `(inv)link!!`.
+       struct MyBijector <: Bijectors.Transform end
 
 julia> # Define some dummy `inverse` which will be used in the `link!!` call.
        Bijectors.inverse(f::MyBijector) = identity

src/logdensityfunction.jl

@@ -10,7 +10,7 @@ $(FIELDS)
 ```jldoctest
 julia> using Distributions
 
-julia> using DynamicPPL: LogDensityFunction
+julia> using DynamicPPL: LogDensityFunction, contextualize
 
 julia> @model function demo(x)
            m ~ Normal()
@@ -36,6 +36,12 @@ julia> # By default it uses `VarInfo` under the hood, but this is not necessary.
 
 julia> LogDensityProblems.logdensity(f, [0.0])
 -2.3378770664093453
+
+julia> # This also respects the context in `model`.
+       f_prior = LogDensityFunction(contextualize(model, DynamicPPL.PriorContext()), VarInfo(model));
+
+julia> LogDensityProblems.logdensity(f_prior, [0.0]) == logpdf(Normal(), 0.0)
+true
 ```
 """
 struct LogDensityFunction{V,M,C}
@@ -60,7 +66,7 @@ end
 function LogDensityFunction(
     model::Model,
     varinfo::AbstractVarInfo=VarInfo(model),
-    context::AbstractContext=DefaultContext(),
+    context::AbstractContext=model.context,
 )
     return LogDensityFunction(varinfo, model, context)
 end

src/simple_varinfo.jl

@@ -648,7 +648,7 @@ Distributions.loglikelihood(model::Model, θ) = loglikelihood(model, SimpleVarIn
 
 # Allow usage of `NamedBijector` too.
 function link!!(
-    t::StaticTransformation{<:Bijectors.NamedBijector},
+    t::StaticTransformation{<:Bijectors.NamedTransform},
     vi::SimpleVarInfo{<:NamedTuple},
     spl::AbstractSampler,
     model::Model,
@@ -663,7 +663,7 @@ function link!!(
 end
 
 function invlink!!(
-    t::StaticTransformation{<:Bijectors.NamedBijector},
+    t::StaticTransformation{<:Bijectors.NamedTransform},
     vi::SimpleVarInfo{<:NamedTuple},
     spl::AbstractSampler,
     model::Model,

src/test_utils.jl

@@ -10,26 +10,8 @@ using Random: Random
 using Bijectors: Bijectors
 using Setfield: Setfield
 
-"""
-    varname_leaves(vn::VarName, val)
-
-Return iterator over all varnames that are represented by `vn` on `val`,
-e.g. `varname_leaves(@varname(x), rand(2))` results in an iterator over `[@varname(x[1]), @varname(x[2])]`.
-"""
-varname_leaves(vn::VarName, val::Real) = [vn]
-function varname_leaves(vn::VarName, val::AbstractArray{<:Union{Real,Missing}})
-    return (
-        VarName(vn, DynamicPPL.getlens(vn) ∘ Setfield.IndexLens(Tuple(I))) for
-        I in CartesianIndices(val)
-    )
-end
-function varname_leaves(vn::VarName, val::AbstractArray)
-    return Iterators.flatten(
-        varname_leaves(
-            VarName(vn, DynamicPPL.getlens(vn) ∘ Setfield.IndexLens(Tuple(I))), val[I]
-        ) for I in CartesianIndices(val)
-    )
-end
+# For backwards compat.
+using DynamicPPL: varname_leaves
 
 """
     update_values!!(vi::AbstractVarInfo, vals::NamedTuple, vns)
@@ -704,7 +686,7 @@ Simple model for which [`default_transformation`](@ref) returns a [`StaticTransf
 end
 
 function DynamicPPL.default_transformation(::Model{typeof(demo_static_transformation)})
-    b = Bijectors.stack(Bijectors.Exp{0}(), Bijectors.Identity{0}())
+    b = Bijectors.stack(Bijectors.elementwise(exp), identity)
     return DynamicPPL.StaticTransformation(b)
 end
 

src/utils.jl

@@ -740,3 +740,49 @@ infer_nested_eltype(::Type{<:AbstractDict{<:Any,ET}}) where {ET} = infer_nested_
 
 # No need + causes issues for some AD backends, e.g. Zygote.
 ChainRulesCore.@non_differentiable infer_nested_eltype(x)
+
+"""
+    varname_leaves(vn::VarName, val)
+
+Return an iterator over all varnames that are represented by `vn` on `val`.
+
+# Examples
+```jldoctest
+julia> using DynamicPPL: varname_leaves
+
+julia> foreach(println, varname_leaves(@varname(x), rand(2)))
+x[1]
+x[2]
+
+julia> foreach(println, varname_leaves(@varname(x[1:2]), rand(2)))
+x[1:2][1]
+x[1:2][2]
+
+julia> x = (y = 1, z = [[2.0], [3.0]]);
+
+julia> foreach(println, varname_leaves(@varname(x), x))
+x.y
+x.z[1][1]
+x.z[2][1]
+```
+"""
+varname_leaves(vn::VarName, ::Real) = [vn]
+function varname_leaves(vn::VarName, val::AbstractArray{<:Union{Real,Missing}})
+    return (
+        VarName(vn, getlens(vn) ∘ Setfield.IndexLens(Tuple(I))) for
+        I in CartesianIndices(val)
+    )
+end
+function varname_leaves(vn::VarName, val::AbstractArray)
+    return Iterators.flatten(
+        varname_leaves(VarName(vn, getlens(vn) ∘ Setfield.IndexLens(Tuple(I))), val[I]) for
+        I in CartesianIndices(val)
+    )
+end
+function varname_leaves(vn::DynamicPPL.VarName, val::NamedTuple)
+    iter = Iterators.map(keys(val)) do sym
+        lens = Setfield.PropertyLens{sym}()
+        varname_leaves(vn ∘ lens, get(val, lens))
+    end
+    return Iterators.flatten(iter)
+end

test/Project.toml

@@ -22,17 +22,17 @@ Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
 
 [compat]
 AbstractMCMC = "2.1, 3.0, 4"
-AbstractPPL = "0.5.1, 0.6"
-Bijectors = "0.9.5, 0.10"
+AbstractPPL = "0.5, 0.6"
+Bijectors = "0.11, 0.12"
 Distributions = "0.25"
 DistributionsAD = "0.6.3"
 Documenter = "0.26.1, 0.27"
 ForwardDiff = "0.10.12"
 LogDensityProblems = "2"
-MCMCChains = "4.0.4, 5"
+MCMCChains = "4.0.4, 5, 6"
 MacroTools = "0.5.5"
 Setfield = "0.7.1, 0.8, 1"
 StableRNGs = "1"
-Tracker = "0.2.11"
+Tracker = "0.2.23"
 Zygote = "0.5.4, 0.6"
 julia = "1.6"

test/simple_varinfo.jl

@@ -64,6 +64,7 @@
         @testset "$(typeof(vi))" for vi in (
             SimpleVarInfo(Dict()), SimpleVarInfo(values_constrained), VarInfo(model)
         )
+            vi = SimpleVarInfo(values_constrained)
             for vn in DynamicPPL.TestUtils.varnames(model)
                 vi = DynamicPPL.setindex!!(vi, get(values_constrained, vn), vn)
             end
@@ -108,6 +109,8 @@
 
     @testset "SimpleVarInfo on $(nameof(model))" for model in
                                                      DynamicPPL.TestUtils.DEMO_MODELS
+        model = DynamicPPL.TestUtils.demo_dot_assume_matrix_dot_observe_matrix()
+
         # We might need to pre-allocate for the variable `m`, so we need
         # to see whether this is the case.
         svi_nt = SimpleVarInfo(rand(NamedTuple, model))

test/test_util.jl

@@ -13,11 +13,8 @@ function test_model_ad(model, logp_manual)
     x = DynamicPPL.getall(vi)
 
     # Log probabilities using the model.
-    function logp_model(x)
-        new_vi = VarInfo(vi, SampleFromPrior(), x)
-        model(new_vi)
-        return getlogp(new_vi)
-    end
+    ℓ = DynamicPPL.LogDensityFunction(model, vi)
+    logp_model = Base.Fix1(LogDensityProblems.logdensity, ℓ)
 
     # Check that both functions return the same values.
     lp = logp_manual(x)

test/turing/Project.toml

@@ -6,5 +6,5 @@ Turing = "fce5fe82-541a-59a6-adf8-730c64b5f9a0"
 
 [compat]
 DynamicPPL = "0.20, 0.21"
-Turing = "0.21"
+Turing = "0.21, 0.22, 0.23, 0.24"
 julia = "1.6"

test/turing/compiler.jl

@@ -70,7 +70,7 @@
         x = Float64[1 2]
 
         @model function gauss(x)
-            priors = TArray{Float64}(2)
+            priors = Array{Float64}(undef, 2)
             priors[1] ~ InverseGamma(2, 3)         # s
             priors[2] ~ Normal(0, sqrt(priors[1])) # m
             for i in 1:length(x)