CompatHelper: bump compat for AdvancedHMC to 0.7, (keep existing compat) (#2503)

* CompatHelper: bump compat for AdvancedHMC to 0.7, (keep existing compat)

* Relax two test tolerances a bit

* Adjust some tolerances and sample counts in tests

* Adjust more test sample counts

---------

Co-authored-by: CompatHelper Julia <[email protected]>
Co-authored-by: Markus Hauru <[email protected]>

Author: 3 people

Project.toml

@@ -51,7 +51,7 @@ TuringOptimExt = "Optim"
 ADTypes = "1.9"
 AbstractMCMC = "5.5"
 Accessors = "0.1"
-AdvancedHMC = "0.3.0, 0.4.0, 0.5.2, 0.6"
+AdvancedHMC = "0.3.0, 0.4.0, 0.5.2, 0.6, 0.7"
 AdvancedMH = "0.8"
 AdvancedPS = "0.6.0"
 AdvancedVI = "0.2"

test/mcmc/Inference.jl

@@ -115,7 +115,7 @@ using Turing
         )
         check_gdemo(chn3)
 
-        chn3_contd = sample(StableRNG(seed), gdemo_default, alg3, 2_000; resume_from=chn3)
+        chn3_contd = sample(StableRNG(seed), gdemo_default, alg3, 5_000; resume_from=chn3)
         check_gdemo(chn3_contd)
     end
 

test/mcmc/gibbs.jl

@@ -573,7 +573,7 @@ end
         # is not ideal
         # Issue ref: https://github.com/TuringLang/Turing.jl/issues/2402
         @test isapprox(mean(num_ms), 8.6087; atol=0.8)
-        @test isapprox(std(num_ms), 1.8865; atol=0.02)
+        @test isapprox(std(num_ms), 1.8865; atol=0.03)
     end
 
     # The below test used to sample incorrectly before
@@ -655,7 +655,7 @@ end
             # Run the Gibbs sampler and NUTS on the same model, compare statistics of the
             # chains.
             @testset "comparison with 'gold-standard' samples" begin
-                num_iterations = 1_000
+                num_iterations = 2_000
                 thinning = 10
                 num_chains = 4
 

test/mcmc/hmc.jl

@@ -172,7 +172,7 @@ using Turing
 
     @testset "nuts inference" begin
         alg = NUTS(1000, 0.8; adtype=adbackend)
-        res = sample(StableRNG(seed), gdemo_default, alg, 500)
+        res = sample(StableRNG(seed), gdemo_default, alg, 5_000)
         check_gdemo(res)
     end
 
@@ -251,7 +251,7 @@ using Turing
         gdemo_default_prior = DynamicPPL.contextualize(
             demo_hmc_prior(), DynamicPPL.PriorContext()
         )
-        chain = sample(gdemo_default_prior, alg, 500; initial_params=[3.0, 0.0])
+        chain = sample(gdemo_default_prior, alg, 5_000; initial_params=[3.0, 0.0])
         check_numerical(
             chain, [:s, :m], [mean(truncated(Normal(3, 1); lower=0)), 0]; atol=0.2
         )