Unify transition also in external samplers (#2030)

* Transition

* Revert "Transition"

This reverts commit 71c8097.

* bug

* repeated functions

* move Transition to inference

* default get_stat

* bug

* Update src/inference/Inference.jl

Co-authored-by: David Widmann <[email protected]>

* Update src/inference/Inference.jl

Co-authored-by: David Widmann <[email protected]>

* Update src/inference/Inference.jl

Co-authored-by: David Widmann <[email protected]>

* rest of david changes

* bring back Transition(a,b)

---------

Co-authored-by: David Widmann <[email protected]>

Author: JaimeRZP

Project.toml

@@ -39,7 +39,6 @@ Tracker = "9f7883ad-71c0-57eb-9f7f-b5c9e6d3789c"
 [compat]
 AbstractMCMC = "4"
 AdvancedHMC = "0.3.0, 0.4"
-AdvancedMH = "0.6.8, 0.7"
 AdvancedPS = "0.4"
 AdvancedVI = "0.2"
 BangBang = "0.3"

src/contrib/inference/abstractmcmc.jl

@@ -3,28 +3,13 @@ struct TuringState{S,F}
     logdensity::F
 end
 
-struct TuringTransition{T,NT<:NamedTuple,F<:AbstractFloat}
-    θ::T
-    lp::F
-    stat::NT
-end
-
-function TuringTransition(vi::AbstractVarInfo, t)
-    theta = tonamedtuple(vi)
-    lp = getlogp(vi)
-    return TuringTransition(theta, lp, getstats(t))
-end
-
-metadata(t::TuringTransition) = merge((lp = t.lp,), t.stat)
-DynamicPPL.getlogp(t::TuringTransition) = t.lp
-
 state_to_turing(f::DynamicPPL.LogDensityFunction, state) = TuringState(state, f)
 function transition_to_turing(f::DynamicPPL.LogDensityFunction, transition)
     θ = getparams(transition)
     varinfo = DynamicPPL.unflatten(f.varinfo, θ)
     # TODO: `deepcopy` is overkill; make more efficient.
     varinfo = DynamicPPL.invlink!!(deepcopy(varinfo), f.model)
-    return TuringTransition(varinfo, transition)
+    return Transition(varinfo, transition)
 end
 
 # NOTE: Only thing that depends on the underlying sampler.

src/inference/Inference.jl

@@ -123,6 +123,9 @@ end
 ######################
 # Default Transition #
 ######################
+# Default
+# Extended in contrib/inference/abstractmcmc.jl
+getstats(t) = nothing
 
 struct Transition{T, F<:AbstractFloat, S<:Union{NamedTuple, Nothing}}
     θ     :: T
@@ -132,10 +135,10 @@ end
 
 Transition(θ, lp) = Transition(θ, lp, nothing)
 
-function Transition(vi::AbstractVarInfo; nt::NamedTuple=NamedTuple())
+function Transition(vi::AbstractVarInfo, t=nothing; nt::NamedTuple=NamedTuple())
     θ = merge(tonamedtuple(vi), nt)
     lp = getlogp(vi)
-    return Transition(θ, lp, nothing)
+    return Transition(θ, lp, getstats(t))
 end
 
 function metadata(t::Transition)
@@ -664,9 +667,7 @@ function transitions_from_chain(
         model(rng, vi, sampler)
 
         # Convert `VarInfo` into `NamedTuple` and save.
-        theta = DynamicPPL.tonamedtuple(vi)
-        lp = Turing.getlogp(vi)
-        Transition(theta, lp)
+        Transition(vi)
     end
 
     return transitions