Add tests with multi-threading/processing

Author: lhnguyen-vn

test/Project.toml

@@ -1,4 +1,5 @@
 [deps]
+ComputationalResources = "ed09eef8-17a6-5b46-8889-db040fac31e3"
 Distributions = "31c24e10-a181-5473-b8eb-7969acd0382f"
 EvoTrees = "f6006082-12f8-11e9-0c9c-0d5d367ab1e5"
 MLJBase = "a7f614a8-145f-11e9-1d2a-a57a1082229d"

test/runtests.jl

@@ -1,6 +1,7 @@
 using ParticleSwarmOptimization
 using Random
 using Test
+using ComputationalResources
 using Distributions
 using EvoTrees
 using MLJBase

test/strategies/basic.jl

@@ -14,46 +14,50 @@
     @test ps.prob_shift == 0.25
 end
 
-@testset "EvoTree Tuning" begin
-    rng = StableRNG(123)
-    features = rand(rng, 10_000) .* 5 .- 2
-    X = MLJBase.table(reshape(features, (size(features)[1], 1)))
-    y = sin.(features) .* 0.5 .+ 0.5
-    y = EvoTrees.logit(y) + randn(rng, size(y))
-    y = EvoTrees.sigmoid(y)
+for acceleration in (CPU1(), CPUProcesses(), CPUThreads())
+    @testset "EvoTree Tuning with $(typeof(acceleration))" begin
+        rng = StableRNG(123)
+        features = rand(rng, 10_000) .* 5 .- 2
+        X = MLJBase.table(reshape(features, (size(features)[1], 1)))
+        y = sin.(features) .* 0.5 .+ 0.5
+        y = EvoTrees.logit(y) + randn(rng, size(y))
+        y = EvoTrees.sigmoid(y)
 
-    tree = VERSION ≥ v"1.4" ? EvoTreeRegressor(rng=rng) : EvoTreeRegressor(seed=123)
-    r1 = range(tree, :max_depth; values=[3:7;])
-    r2 = range(tree, :η; lower=-2, upper=0, scale=exp10)
+        tree = EvoTreeRegressor(rng=rng)
+        r1 = range(tree, :max_depth; values=[3:7;])
+        r2 = range(tree, :η; lower=-2, upper=0, scale=exp10)
 
-    baseline_self_tuning_tree = TunedModel(
-        model=tree,
-        tuning=RandomSearch(rng=StableRNG(1234)),
-        # tuning=ParticleSwarm(n_particles=3, rng=rng),
-        resampling=CV(nfolds=5, rng=StableRNG(8888)),
-        range=[r1, r2],
-        measure=(ŷ, y) -> mean(abs.(ŷ .- y)),
-        n=15
-    )
-    baseline_mach = machine(baseline_self_tuning_tree, X, y)
-    fit!(baseline_mach, verbosity=0)
-    baseline_rep = report(baseline_mach)
-    baseline_best_loss = baseline_rep.best_history_entry.measurement[1]
+        baseline_self_tuning_tree = TunedModel(
+            model=tree,
+            tuning=RandomSearch(rng=StableRNG(1234)),
+            # tuning=ParticleSwarm(n_particles=3, rng=rng),
+            resampling=CV(nfolds=5, rng=StableRNG(8888)),
+            range=[r1, r2],
+            measure=(ŷ, y) -> mean(abs.(ŷ .- y)),
+            n=15,
+            acceleration=acceleration
+        )
+        baseline_mach = machine(baseline_self_tuning_tree, X, y)
+        fit!(baseline_mach, verbosity=2)
+        baseline_rep = report(baseline_mach)
+        baseline_best_loss = baseline_rep.best_history_entry.measurement[1]
 
-    self_tuning_tree = TunedModel(
-        model=tree,
-        tuning=ParticleSwarm(rng=StableRNG(1234)),
-        # tuning=ParticleSwarm(n_particles=3, rng=rng),
-        resampling=CV(nfolds=5, rng=StableRNG(8888)),
-        range=[r1, r2],
-        measure=(ŷ, y) -> mean(abs.(ŷ .- y)),
-        n=15
-    )
-    mach = machine(self_tuning_tree, X, y)
-    fit!(mach, verbosity=0)
-    rep = report(mach)
-    best_loss = rep.best_history_entry.measurement[1]
+        self_tuning_tree = TunedModel(
+            model=tree,
+            tuning=ParticleSwarm(rng=StableRNG(1234)),
+            resampling=CV(nfolds=5, rng=StableRNG(8888)),
+            range=[r1, r2],
+            measure=(ŷ, y) -> mean(abs.(ŷ .- y)),
+            n=15,
+            acceleration=acceleration
+        )
+        mach = machine(self_tuning_tree, X, y)
+        fit!(mach, verbosity=2)
+        rep = report(mach)
+        best_loss = rep.best_history_entry.measurement[1]
 
-    # Compare with random search result with the same settings
-    @test best_loss < baseline_best_loss || isapprox(best_loss, baseline_best_loss, 1e-3)
+        # Compare with random search result with the same settings
+        @test best_loss < baseline_best_loss ||
+              isapprox(best_loss, baseline_best_loss, 1e-3)
+    end
 end