Move Enzyme AD tests to their own group and environment (#1261)

Enzyme v0.13.182 fails to differentiate through the BLAS `dot` lowering
introduced in Julia 1.13 ("No augmented forward pass found for
ijl_lazy_load_and_lookup"), breaking the Core CI job on the `pre` channel.
The upstream fix (Enzyme 23ff630) is not yet in a registered release.

Split the Enzyme-dependent AD tests into their own `AD` test group, folder,
and environment:

- test/ADtests.jl and test/AD_performance_regression.jl move to test/AD/.
- The AD group gets its own env (test/AD/Project.toml), so its Enzyme-family
  dependencies (Enzyme, Tracker, Mooncake, BenchmarkTools, Ipopt,
  OptimizationMOI, OptimizationLBFGSB) live only there and are removed from the
  main test target, simplifying the main [extras]/[targets]/[compat].
- test_groups.toml declares the AD group with versions = ["lts", "1"] so CI
  runs it on stable releases but excludes it from the `pre` channel until a
  fixed Enzyme is registered.

Verified locally: `GROUP=AD` resolves/instantiates the new env and passes
(AD Tests 123/123, AD Performance Regression 3/3) on Julia 1.12.4; the main
test env also builds without the moved dependencies.

See #1260.

Co-authored-by: ChrisRackauckas-Claude <accounts@chrisrackauckas.com>

Author: ChrisRackauckas-Claude

Project.toml

@@ -20,12 +20,6 @@ TerminalLoggers = "5d786b92-1e48-4d6f-9151-6b4477ca9bed"
 [sources.OptimizationBase]
 path = "lib/OptimizationBase"
 
-[sources.OptimizationLBFGSB]
-path = "lib/OptimizationLBFGSB"
-
-[sources.OptimizationMOI]
-path = "lib/OptimizationMOI"
-
 [sources.OptimizationOptimJL]
 path = "lib/OptimizationOptimJL"
 
@@ -35,17 +29,14 @@ path = "lib/OptimizationOptimisers"
 [compat]
 ADTypes = "1.18"
 ArrayInterface = "7.10"
-BenchmarkTools = "1"
 Boltz = "1"
 ComponentArrays = ">= 0.13.9"
 ConsoleProgressMonitor = "0.1.1"
 DiffEqFlux = "2, 3, 4"
 DocStringExtensions = "0.9.5"
-Enzyme = "0.13"
 FiniteDiff = "2"
 Flux = "0.13, 0.14, 0.15, 0.16"
 ForwardDiff = "0.10, 1"
-Ipopt = "1"
 IterTools = "1.3"
 LinearAlgebra = "1.10"
 LinearSolve = "3.64.0"
@@ -54,12 +45,9 @@ LoggingExtras = "0.4, 1"
 Lux = "1.12.4"
 MLUtils = "0.4"
 ModelingToolkit = "11"
-Mooncake = "0.4.138, 0.5"
 Optim = ">= 1.4.1"
 Optimisers = ">= 0.2.5"
 OptimizationBase = "5"
-OptimizationLBFGSB = "1.2"
-OptimizationMOI = "1"
 OptimizationOptimJL = "0.4.10"
 OptimizationOptimisers = "0.3.16"
 OrdinaryDiffEqTsit5 = "1, 2"
@@ -77,30 +65,23 @@ SparseArrays = "1.10"
 Symbolics = "7"
 TerminalLoggers = "0.1"
 Test = "1.10"
-Tracker = "0.2"
 Zygote = "0.6, 0.7"
 julia = "1.10"
 
 [extras]
-BenchmarkTools = "6e4b80f9-dd63-53aa-95a3-0cdb28fa8baf"
 Boltz = "4544d5e4-abc5-4dea-817f-29e4c205d9c8"
 ComponentArrays = "b0b7db55-cfe3-40fc-9ded-d10e2dbeff66"
 DiffEqFlux = "aae7a2af-3d4f-5e19-a356-7da93b79d9d0"
-Enzyme = "7da242da-08ed-463a-9acd-ee780be4f1d9"
 FiniteDiff = "6a86dc24-6348-571c-b903-95158fe2bd41"
 Flux = "587475ba-b771-5e3f-ad9e-33799f191a9c"
 ForwardDiff = "f6369f11-7733-5829-9624-2563aa707210"
-Ipopt = "b6b21f68-93f8-5de0-b562-5493be1d77c9"
 IterTools = "c8e1da08-722c-5040-9ed9-7db0dc04731e"
 LinearSolve = "7ed4a6bd-45f5-4d41-b270-4a48e9bafcae"
 Lux = "b2108857-7c20-44ae-9111-449ecde12c47"
 MLUtils = "f1d291b0-491e-4a28-83b9-f70985020b54"
 ModelingToolkit = "961ee093-0014-501f-94e3-6117800e7a78"
-Mooncake = "da2b9cff-9c12-43a0-ae48-6db2b0edb7d6"
 Optim = "429524aa-4258-5aef-a3af-852621145aeb"
 Optimisers = "3bd65402-5787-11e9-1adc-39752487f4e2"
-OptimizationLBFGSB = "22f7324a-a79d-40f2-bebe-3af60c77bd15"
-OptimizationMOI = "fd9f6733-72f4-499f-8506-86b2bdd0dea1"
 OptimizationOptimJL = "36348300-93cb-4f02-beb5-3c3902f8871e"
 OptimizationOptimisers = "42dfb2eb-d2b4-4451-abcd-913932933ac1"
 OrdinaryDiffEqTsit5 = "b1df2697-797e-41e3-8120-5422d3b24e4a"
@@ -114,8 +95,7 @@ SciMLTesting = "09d9d899-5365-40a9-917a-5f67fddea283"
 SparseArrays = "2f01184e-e22b-5df5-ae63-d93ebab69eaf"
 Symbolics = "0c5d862f-8b57-4792-8d23-62f2024744c7"
 Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
-Tracker = "9f7883ad-71c0-57eb-9f7f-b5c9e6d3789c"
 Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
 
 [targets]
-test = ["BenchmarkTools", "Boltz", "ComponentArrays", "DiffEqFlux", "Enzyme", "FiniteDiff", "Flux", "ForwardDiff", "Ipopt", "IterTools", "LinearSolve", "Lux", "MLUtils", "ModelingToolkit", "Optim", "OptimizationLBFGSB", "OptimizationMOI", "OptimizationOptimJL", "OptimizationOptimisers", "OrdinaryDiffEqTsit5", "Pkg", "Random", "ReverseDiff", "SafeTestsets", "SciMLSensitivity", "SciMLTesting", "SparseArrays", "Symbolics", "Test", "Tracker", "Zygote", "Mooncake", "SciMLLogging"]
+test = ["Boltz", "ComponentArrays", "DiffEqFlux", "FiniteDiff", "Flux", "ForwardDiff", "IterTools", "LinearSolve", "Lux", "MLUtils", "ModelingToolkit", "Optim", "OptimizationOptimJL", "OptimizationOptimisers", "OrdinaryDiffEqTsit5", "Pkg", "Random", "ReverseDiff", "SafeTestsets", "SciMLSensitivity", "SciMLTesting", "SparseArrays", "Symbolics", "Test", "Zygote", "SciMLLogging"]

test/AD_performance_regression.jl test/AD/AD_performance_regression.jltest/AD_performance_regression.jl renamed to test/AD/AD_performance_regression.jl

@@ -0,0 +1,49 @@
+[deps]
+ADTypes = "47edcb42-4c32-4615-8424-f2b9edc5f35b"
+BenchmarkTools = "6e4b80f9-dd63-53aa-95a3-0cdb28fa8baf"
+Enzyme = "7da242da-08ed-463a-9acd-ee780be4f1d9"
+FiniteDiff = "6a86dc24-6348-571c-b903-95158fe2bd41"
+ForwardDiff = "f6369f11-7733-5829-9624-2563aa707210"
+Ipopt = "b6b21f68-93f8-5de0-b562-5493be1d77c9"
+Mooncake = "da2b9cff-9c12-43a0-ae48-6db2b0edb7d6"
+Optimization = "7f7a1694-90dd-40f0-9382-eb1efda571ba"
+OptimizationBase = "bca83a33-5cc9-4baa-983d-23429ab6bcbb"
+OptimizationLBFGSB = "22f7324a-a79d-40f2-bebe-3af60c77bd15"
+OptimizationMOI = "fd9f6733-72f4-499f-8506-86b2bdd0dea1"
+OptimizationOptimJL = "36348300-93cb-4f02-beb5-3c3902f8871e"
+Random = "9a3f8284-a2c9-5f02-9a11-845980a1fd5c"
+ReverseDiff = "37e2e3b7-166d-5795-8a7a-e32c996b4267"
+SafeTestsets = "1bc83da4-3b8d-516f-aca4-4fe02f6d838f"
+SciMLTesting = "09d9d899-5365-40a9-917a-5f67fddea283"
+Test = "8dfed614-e22c-5e08-85e1-65c5234f0b40"
+Tracker = "9f7883ad-71c0-57eb-9f7f-b5c9e6d3789c"
+Zygote = "e88e6eb3-aa80-5325-afca-941959d7151f"
+
+[sources]
+Optimization = {path = "../.."}
+OptimizationBase = {path = "../../lib/OptimizationBase"}
+OptimizationLBFGSB = {path = "../../lib/OptimizationLBFGSB"}
+OptimizationMOI = {path = "../../lib/OptimizationMOI"}
+OptimizationOptimJL = {path = "../../lib/OptimizationOptimJL"}
+
+[compat]
+ADTypes = "1.18"
+BenchmarkTools = "1"
+Enzyme = "0.13"
+FiniteDiff = "2"
+ForwardDiff = "0.10, 1"
+Ipopt = "1"
+Mooncake = "0.4.138, 0.5"
+Optimization = "5"
+OptimizationBase = "5"
+OptimizationLBFGSB = "1.2"
+OptimizationMOI = "1"
+OptimizationOptimJL = "0.4.10"
+Random = "1.10"
+ReverseDiff = "1"
+SafeTestsets = "0.1"
+SciMLTesting = "2.1"
+Test = "1.10"
+Tracker = "0.2"
+Zygote = "0.6, 0.7"
+julia = "1.10"

test/ADtests.jl test/AD/ADtests.jltest/ADtests.jl renamed to test/AD/ADtests.jl

@@ -59,8 +59,6 @@ const LIB_DIR = joinpath(dirname(@__DIR__), "lib")
             core = function ()
                 @safetestset "Utils Tests" include("utils.jl")
                 @safetestset "Verbosity Tests" include("verbosity.jl")
-                @safetestset "AD Tests" include("ADtests.jl")
-                @safetestset "AD Performance Regression Tests" include("AD_performance_regression.jl")
                 @safetestset "Optimization" include("native.jl")
                 @safetestset "Mini batching" include("minibatch.jl")
                 # DiffEqFlux test temporarily skipped due to ForwardDiff gradient dispatch
@@ -70,6 +68,19 @@ const LIB_DIR = joinpath(dirname(@__DIR__), "lib")
                 @safetestset "Sense Handling" include("sense_tests.jl")
             end,
             groups = Dict(
+                # The AD tests are their own group with their own environment
+                # (test/AD/Project.toml) so their Enzyme-family dependencies stay out
+                # of the main test env, and so the group can be excluded from the Julia
+                # `pre` channel in test_groups.toml: Enzyme is expected to fail on
+                # prereleases until a fixed Enzyme release lands.
+                # See https://github.com/SciML/Optimization.jl/issues/1260.
+                "AD" => (;
+                    env = joinpath(@__DIR__, "AD"),
+                    body = function ()
+                        @safetestset "AD Tests" include("AD/ADtests.jl")
+                        @safetestset "AD Performance Regression Tests" include("AD/AD_performance_regression.jl")
+                    end,
+                ),
                 "GPU" => (;
                     env = joinpath(@__DIR__, "downstream"),
                     body = function ()

test/AD/Project.toml

@@ -1,5 +1,11 @@
 [Core]
 versions = ["lts", "1", "pre"]
 
+# Enzyme is expected to fail on Julia prereleases (see
+# https://github.com/SciML/Optimization.jl/issues/1260), so the Enzyme-dependent
+# AD tests run their own group and are excluded from the "pre" channel.
+[AD]
+versions = ["lts", "1"]
+
 [QA]
 versions = ["lts", "1"]