Use Random.default_rng() instead of Random.GLOBAL_RNG

Author: devmotion

README.md

@@ -289,7 +289,7 @@ function sample(
 Draw `n_samples` samples using the kernel `κ` under the Hamiltonian system `h`
 
 - The randomness is controlled by `rng`.
-  - If `rng` is not provided, `GLOBAL_RNG` will be used.
+  - If `rng` is not provided, the default random number generator (`Random.default_rng()`) will be used.
 - The initial point is given by `θ`.
 - The adaptor is set by `adaptor`, for which the default is no adaptation.
   - It will perform `n_adapts` steps of adaptation, for which the default is `1_000` or 10% of `n_samples`, whichever is lower. 

research/src/riemannian_hmc.jl

@@ -145,7 +145,7 @@ Base.rand(rng::AbstractRNG, metric::AbstractMetric, kinetic, θ) =
     _rand(rng, metric, kinetic)    # this disambiguity is required by Random.rand
 Base.rand(rng::AbstractVector{<:AbstractRNG}, metric::AbstractMetric, kinetic, θ) =
     _rand(rng, metric, kinetic)
-Base.rand(metric::AbstractMetric, kinetic, θ) = rand(GLOBAL_RNG, metric, kinetic)
+Base.rand(metric::AbstractMetric, kinetic, θ) = rand(Random.default_rng(), metric, kinetic)
 
 ### metric.jl
 

src/AdvancedHMC.jl

@@ -7,7 +7,7 @@ using LinearAlgebra:
     Symmetric, UpperTriangular, mul!, ldiv!, dot, I, diag, cholesky, UniformScaling
 using StatsFuns: logaddexp, logsumexp
 import Random
-using Random: GLOBAL_RNG, AbstractRNG
+using Random: AbstractRNG
 using ProgressMeter: ProgressMeter
 using SimpleUnPack: @unpack
 

src/metric.jl

@@ -132,7 +132,7 @@ Base.rand(
     kinetic::AbstractKinetic,
 ) = _rand(rng, metric, kinetic)
 Base.rand(metric::AbstractMetric, kinetic::AbstractKinetic) =
-    rand(GLOBAL_RNG, metric, kinetic)
+    rand(Random.default_rng(), metric, kinetic)
 
 # ignore θ by default unless defined by the specific kinetic (i.e. not position-dependent)
 Base.rand(

src/sampler.jl

@@ -118,7 +118,7 @@ sample(
     progress::Bool = false,
     (pm_next!)::Function = pm_next!,
 ) = sample(
-    GLOBAL_RNG,
+    Random.default_rng(),
     h,
     κ,
     θ,
@@ -146,7 +146,7 @@ sample(
     )
 Sample `n_samples` samples using the proposal `κ` under Hamiltonian `h`.
 - The randomness is controlled by `rng`. 
-    - If `rng` is not provided, `GLOBAL_RNG` will be used.
+    - If `rng` is not provided, the default random number generator (`Random.default_rng()`) will be used.
 - The initial point is given by `θ`.
 - The adaptor is set by `adaptor`, for which the default is no adaptation.
     - It will perform `n_adapts` steps of adaptation, for which the default is the minimum of `1_000` and 10% of `n_samples`

src/trajectory.jl

@@ -3,7 +3,7 @@
 ####
 #### Developers' Notes
 ####
-#### Not all functions that use `rng` require a fallback function with `GLOBAL_RNG`
+#### Not all functions that use `rng` require a fallback function with `Random.default_rng()`
 #### as default. In short, only those exported to other libries need such a fallback
 #### function. Internal uses shall always use the explict `rng` version. (Kai Xu 6/Jul/19)
 
@@ -241,10 +241,10 @@ $(SIGNATURES)
 
 Make a MCMC transition from phase point `z` using the trajectory `τ` under Hamiltonian `h`.
 
-NOTE: This is a RNG-implicit fallback function for `transition(GLOBAL_RNG, τ, h, z)`
+NOTE: This is a RNG-implicit fallback function for `transition(Random.default_rng(), τ, h, z)`
 """
 function transition(τ::Trajectory, h::Hamiltonian, z::PhasePoint)
-    return transition(GLOBAL_RNG, τ, h, z)
+    return transition(Random.default_rng(), τ, h, z)
 end
 
 ###
@@ -834,7 +834,7 @@ function find_good_stepsize(
     θ::AbstractVector{<:AbstractFloat};
     max_n_iters::Int = 100,
 )
-    return find_good_stepsize(GLOBAL_RNG, h, θ; max_n_iters = max_n_iters)
+    return find_good_stepsize(Random.default_rng(), h, θ; max_n_iters = max_n_iters)
 end
 
 "Perform MH acceptance based on energy, i.e. negative log probability."

test/integrator.jl

@@ -39,7 +39,7 @@ using Statistics: mean
             @test AdvancedHMC.nom_step_size(lf) == ϵ0
             @test AdvancedHMC.step_size(lf) == ϵ0
 
-            lf2 = AdvancedHMC.jitter(Random.GLOBAL_RNG, lf)
+            lf2 = AdvancedHMC.jitter(Random.default_rng(), lf)
             @test lf2 === lf
             @test AdvancedHMC.nom_step_size(lf2) == ϵ0
             @test AdvancedHMC.step_size(lf2) == ϵ0
@@ -53,7 +53,7 @@ using Statistics: mean
             @test lf.ϵ == ϵ0
             @test AdvancedHMC.step_size(lf) == lf.ϵ
 
-            lf2 = AdvancedHMC.jitter(Random.GLOBAL_RNG, lf)
+            lf2 = AdvancedHMC.jitter(Random.default_rng(), lf)
             @test lf2.ϵ0 == ϵ0
             @test AdvancedHMC.nom_step_size(lf2) == ϵ0
             @test lf2.ϵ != ϵ0