test: of loss function expression on workers

src/Configure.jl

@@ -163,9 +163,8 @@ function move_functions_to_workers(
                 continue
             end
             ops = (options.loss_function_expression,)
-            example_inputs = (
-                create_expression(zero(T), options, dataset), dataset, options
-            )
+            ex = create_expression(zero(T), options, dataset)
+            example_inputs = (ex, dataset, options)
         elseif function_set == :complexity_mapping
             if !(options.complexity_mapping isa Function)
                 continue

test/runtests.jl

@@ -146,6 +146,10 @@ include("test_mlj.jl")
     include("test_custom_operators_multiprocessing.jl")
 end
 
+@testitem "Testing whether we can move loss function expression to workers." tags = [:part2] begin
+    include("test_loss_function_expression_multiprocessing.jl")
+end
+
 @testitem "Test whether the precompilation script works." tags = [:part2] begin
     include("test_precompilation.jl")
 end

test/test_loss_function_expression_multiprocessing.jl

@@ -0,0 +1,48 @@
+using SymbolicRegression
+using Test
+
+defs = quote
+    using SymbolicRegression
+
+    early_stop(loss, c) = ((loss <= 1e-10) && (c <= 4))
+    function my_loss_expression(ex::Expression, dataset::Dataset, options::Options)
+        prediction, complete = eval_tree_array(ex, dataset.X, options)
+        if !complete
+            return Inf
+        end
+        return sum((prediction .- dataset.y) .^ 2) / dataset.n
+    end
+end
+
+# This is needed as workers are initialized in `Core.Main`!
+if (@__MODULE__) != Core.Main
+    Core.eval(Core.Main, defs)
+    eval(:(using Main: my_loss_expression))
+else
+    eval(defs)
+end
+
+X = randn(Float32, 5, 100)
+y = @. 2 * cos(X[4, :])
+
+options = SymbolicRegression.Options(;
+    binary_operators=[*, +],
+    unary_operators=[cos],
+    early_stop_condition=early_stop,
+    loss_function_expression=my_loss_expression,
+)
+
+hof = equation_search(
+    X,
+    y;
+    weights=ones(Float32, 100),
+    options=options,
+    niterations=1_000_000_000,
+    numprocs=2,
+    parallelism=:multiprocessing,
+)
+
+@test any(
+    early_stop(member.loss, length(get_tree(member.tree))) for
+    member in hof.members[hof.exists]
+)