Additional cleanup

Project.toml

@@ -1,7 +1,7 @@
 name = "ThermodynamicIntegration"
 uuid = "1022446e-a4a4-4a46-8bce-0ffd39f68cd3"
 authors = ["Theo Galy-Fajou <theo.galyfajou@gmail.com> and contributors"]
-version = "0.2.5"
+version = "0.2.6"
 
 [deps]
 AdvancedHMC = "0bf59076-c3b1-5ca4-86bd-e02cd72cde3d"

README.md

@@ -20,6 +20,7 @@ For a different way of computing the evidence integral see also my [BayesianQuad
 
 A simple package to compute Thermodynamic Integration for computing the evidence in a Bayesian setting.
 You need to provide the `logprior` and the `loglikelihood` as well as an initial sample:
+
 ```julia
     using Distributions, ThermodynamicIntegration
     D = 5
@@ -36,7 +37,8 @@ You need to provide the `logprior` and the `loglikelihood` as well as an initial
     # -8.211990123364176
 ```
 
-You can also simply pass a Turing model :
+You can also simply pass a Turing model:
+
 ```julia
     using Turing
     @model function gauss(y)
@@ -53,11 +55,21 @@ You can also simply pass a Turing model :
 ## Parallel sampling
 
 The algorithm also works on multiple threads by calling :
+
 ```julia
     alg = ThermInt(n_samples=5000) 
-    logZ = alg(logprior, loglikelihood, rand(prior), TIParallelThreads())
+    logZ = alg(logprior, loglikelihood, rand(prior), TIThreads())
 ```
 
+or on multiple processes:
+
+```julia
+    alg = ThermInt(n_samples=5000) 
+    logZ = alg(logprior, loglikelihood, rand(prior), TIDistributed())
+```
+
+Note that you need to load `ThermodynamicIntegration` and other necessary external packages on your additional processes via `@everywhere`.
+
 ## Sampling methods
 
 Right now sampling is based on [`AdvancedHMC.jl`](https://github.com/TuringLang/AdvancedHMC.jl), with the `ForwardDiff` AD backend.

src/ThermodynamicIntegration.jl

@@ -4,16 +4,14 @@ using AdvancedHMC
 using Distributed
 using ForwardDiff
 using ProgressMeter
-using Random
+using Random: Random, AbstractRNG, default_rng
 using Requires
 using Statistics
 using Trapz
 
 export ThermInt
 export TISerial, TIThreads, TIDistributed
 
-const GLOBAL_RNG = Random.MersenneTwister(42)
-
 const ADBACKEND = Ref(:ForwardDiff)
 
 set_adbackend(ad::String) = set_adbackend(Symbol(ad))

src/thermint.jl

@@ -7,12 +7,13 @@
 `(1:n_steps) ./ n_steps).^5`
 
 A `ThermInt` object can then be used as a function:
+
 ```julia
-alg = ThermInt(30)
-alg(loglikelihood, logprior, x_init)
+    alg = ThermInt(30)
+    alg(loglikelihood, logprior, x_init)
 ```
 """
-struct ThermInt{AD,TRNG,V}
+struct ThermInt{AD,TRNG<:AbstractRNG,V}
     schedule::V
     n_samples::Int
     n_warmup::Int
@@ -26,7 +27,7 @@ function ThermInt(rng::AbstractRNG, schedule; n_samples::Int=2000, n_warmup::Int
 end
 
 function ThermInt(schedule; n_samples::Int=2000, n_warmup::Int=500)
-    return ThermInt(GLOBAL_RNG, schedule; n_samples=n_samples, n_warmup=n_warmup)
+    return ThermInt(default_rng(), schedule; n_samples=n_samples, n_warmup=n_warmup)
 end
 
 function ThermInt(rng::AbstractRNG; n_steps::Int, n_samples::Int=2000, n_warmup::Int=500)
@@ -37,7 +38,7 @@ end
 
 function ThermInt(; n_steps::Int=30, n_samples::Int=2000, n_warmup::Int=500)
     return ThermInt(
-        GLOBAL_RNG, range(0, 1; length=n_steps) .^ 5; n_samples=n_samples, n_warmup=n_warmup
+        default_rng(), range(0, 1; length=n_steps) .^ 5; n_samples=n_samples, n_warmup=n_warmup
     )
 end