[CI] Various CI fixes (dmlc#11196)

hcho3 · trivialfis · razdoburdin · web-flow · commit fc32798b651c · 2025-02-03T21:14:57.000-08:00
* Fix dmlc#10752 (dmlc#10972) * Fix dmlc#10752 * [CI] Replace Mambaforge -> Miniforge3 * Fix formatting * Increase timeout for Multi-GPU approx tests. (dmlc#11111) * Fix tests with the latest scikit-learn. (dmlc#11086) * Fix tests with the latest scikit-learn. * dask. * Remove scikit-learn pin --------- Co-authored-by: Philip Hyunsu Cho <chohyu01@cs.washington.edu> * Revert base class change * SYCL. Fix CI failures (dmlc#10987) * [CI] Use Python 3.10 + latest sklearn in Windows * Increase timeout in C++ tests from 1 to 5 seconds. (dmlc#10756) To avoid CI failures on FreeBSD. * [CI] Pin sklearn to 1.5.2 * [CI] Apply part of dmlc#10434 --------- Co-authored-by: Jiaming Yuan <jm.yuan@outlook.com> Co-authored-by: Dmitry Razdoburdin <d.razdoburdin@gmail.com>
diff --git a/.github/workflows/jvm_tests.yml b/.github/workflows/jvm_tests.yml
@@ -30,7 +30,7 @@ jobs:
 
     - uses: conda-incubator/setup-miniconda@a4260408e20b96e80095f42ff7f1a15b27dd94ca  # v3.0.4
       with:
-        miniforge-variant: Mambaforge
+        miniforge-variant: Miniforge3
         miniforge-version: latest
         activate-environment: jvm_tests
         environment-file: tests/ci_build/conda_env/jvm_tests.yml
diff --git a/.github/workflows/main.yml b/.github/workflows/main.yml
@@ -81,7 +81,7 @@ jobs:
         submodules: 'true'
     - uses: conda-incubator/setup-miniconda@a4260408e20b96e80095f42ff7f1a15b27dd94ca  # v3.0.4
       with:
-        miniforge-variant: Mambaforge
+        miniforge-variant: Miniforge3
         miniforge-version: latest
         activate-environment: linux_sycl_test
         environment-file: tests/ci_build/conda_env/linux_sycl_test.yml
@@ -123,7 +123,7 @@ jobs:
         submodules: 'true'
     - uses: conda-incubator/setup-miniconda@a4260408e20b96e80095f42ff7f1a15b27dd94ca  # v3.0.4
       with:
-        miniforge-variant: Mambaforge
+        miniforge-variant: Miniforge3
         miniforge-version: latest
         activate-environment: cpp_test
         environment-file: tests/ci_build/conda_env/cpp_test.yml
diff --git a/.github/workflows/python_tests.yml b/.github/workflows/python_tests.yml
@@ -26,7 +26,7 @@ jobs:
         submodules: 'true'
     - uses: conda-incubator/setup-miniconda@a4260408e20b96e80095f42ff7f1a15b27dd94ca  # v3.0.4
       with:
-        miniforge-variant: Mambaforge
+        miniforge-variant: Miniforge3
         miniforge-version: latest
         activate-environment: python_lint
         environment-file: tests/ci_build/conda_env/python_lint.yml
@@ -58,7 +58,7 @@ jobs:
         submodules: 'true'
     - uses: conda-incubator/setup-miniconda@a4260408e20b96e80095f42ff7f1a15b27dd94ca  # v3.0.4
       with:
-        miniforge-variant: Mambaforge
+        miniforge-variant: Miniforge3
         miniforge-version: latest
         activate-environment: sdist_test
         environment-file: tests/ci_build/conda_env/sdist_test.yml
@@ -130,7 +130,7 @@ jobs:
 
     - uses: conda-incubator/setup-miniconda@a4260408e20b96e80095f42ff7f1a15b27dd94ca  # v3.0.4
       with:
-        miniforge-variant: Mambaforge
+        miniforge-variant: Miniforge3
         miniforge-version: latest
         activate-environment: macos_cpu_test
         environment-file: tests/ci_build/conda_env/macos_cpu_test.yml
@@ -227,7 +227,7 @@ jobs:
 
     - uses: conda-incubator/setup-miniconda@a4260408e20b96e80095f42ff7f1a15b27dd94ca  # v3.0.4
       with:
-        miniforge-variant: Mambaforge
+        miniforge-variant: Miniforge3
         miniforge-version: latest
         activate-environment: linux_cpu_test
         environment-file: tests/ci_build/conda_env/linux_cpu_test.yml
@@ -280,7 +280,7 @@ jobs:
 
     - uses: conda-incubator/setup-miniconda@a4260408e20b96e80095f42ff7f1a15b27dd94ca  # v3.0.4
       with:
-        miniforge-variant: Mambaforge
+        miniforge-variant: Miniforge3
         miniforge-version: latest
         activate-environment: linux_sycl_test
         environment-file: tests/ci_build/conda_env/linux_sycl_test.yml
diff --git a/.github/workflows/python_wheels.yml b/.github/workflows/python_wheels.yml
@@ -34,7 +34,7 @@ jobs:
       run: brew install libomp
     - uses: conda-incubator/setup-miniconda@a4260408e20b96e80095f42ff7f1a15b27dd94ca  # v3.0.4
       with:
-        miniforge-variant: Mambaforge
+        miniforge-variant: Miniforge3
         miniforge-version: latest
         python-version: 3.9
         use-mamba: true
diff --git a/plugin/sycl/common/hist_util.cc b/plugin/sycl/common/hist_util.cc
@@ -9,7 +9,7 @@
 #include "../data/gradient_index.h"
 #include "hist_util.h"
 
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 
 namespace xgboost {
 namespace sycl {
diff --git a/plugin/sycl/common/hist_util.h b/plugin/sycl/common/hist_util.h
@@ -15,7 +15,7 @@
 #include "../../src/common/hist_util.h"
 #include "../data/gradient_index.h"
 
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 
 namespace xgboost {
 namespace sycl {
diff --git a/plugin/sycl/common/partition_builder.h b/plugin/sycl/common/partition_builder.h
@@ -25,7 +25,7 @@
 #include "../data/gradient_index.h"
 #include "../tree/expand_entry.h"
 
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 
 namespace xgboost {
 namespace sycl {
diff --git a/plugin/sycl/common/row_set.h b/plugin/sycl/common/row_set.h
@@ -15,7 +15,7 @@
 
 #include "../data.h"
 
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 
 namespace xgboost {
 namespace sycl {
diff --git a/plugin/sycl/data.h b/plugin/sycl/data.h
@@ -22,7 +22,7 @@
 
 #include "../../src/common/threading_utils.h"
 
-#include "CL/sycl.hpp"
+#include <sycl/sycl.hpp>
 
 namespace xgboost {
 namespace sycl {
diff --git a/plugin/sycl/data/gradient_index.cc b/plugin/sycl/data/gradient_index.cc
@@ -8,7 +8,7 @@
 
 #include "gradient_index.h"
 
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 
 namespace xgboost {
 namespace sycl {
diff --git a/plugin/sycl/data/gradient_index.h b/plugin/sycl/data/gradient_index.h
@@ -10,7 +10,7 @@
 #include "../data.h"
 #include "../../src/common/hist_util.h"
 
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 
 namespace xgboost {
 namespace sycl {
diff --git a/plugin/sycl/device_manager.h b/plugin/sycl/device_manager.h
@@ -10,7 +10,7 @@
 #include <string>
 #include <unordered_map>
 
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 
 #pragma GCC diagnostic push
 #pragma GCC diagnostic ignored "-Wtautological-constant-compare"
diff --git a/plugin/sycl/objective/multiclass_obj.cc b/plugin/sycl/objective/multiclass_obj.cc
@@ -23,7 +23,7 @@
 #include "../../../src/objective/multiclass_param.h"
 
 #include "../device_manager.h"
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 
 namespace xgboost {
 namespace sycl {
diff --git a/plugin/sycl/objective/regression_obj.cc b/plugin/sycl/objective/regression_obj.cc
@@ -24,12 +24,13 @@
 #pragma GCC diagnostic push
 #pragma GCC diagnostic ignored "-Wtautological-constant-compare"
 #include "../../../src/objective/regression_loss.h"
+#include "../../../src/objective/adaptive.h"
 #pragma GCC diagnostic pop
 #include "../../../src/objective/regression_param.h"
 
 #include "../device_manager.h"
 
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 
 namespace xgboost {
 namespace sycl {
diff --git a/plugin/sycl/predictor/predictor.cc b/plugin/sycl/predictor/predictor.cc
@@ -10,7 +10,7 @@
 #include <limits>
 #include <mutex>
 
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 
 #include "../data.h"
 
diff --git a/plugin/sycl/tree/param.h b/plugin/sycl/tree/param.h
@@ -19,7 +19,7 @@
 #include "../src/tree/param.h"
 #pragma GCC diagnostic pop
 
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 
 namespace xgboost {
 namespace sycl {
diff --git a/plugin/sycl/tree/split_evaluator.h b/plugin/sycl/tree/split_evaluator.h
@@ -21,7 +21,7 @@
 #include "../../src/common/math.h"
 #include "../../src/tree/param.h"
 
-#include <CL/sycl.hpp>
+#include <sycl/sycl.hpp>
 
 namespace xgboost {
 namespace sycl {
diff --git a/tests/ci_build/Dockerfile.gpu b/tests/ci_build/Dockerfile.gpu
@@ -6,7 +6,7 @@ ARG NCCL_VERSION_ARG
 
 # Environment
 ENV DEBIAN_FRONTEND noninteractive
-SHELL ["/bin/bash", "-c"]   # Use Bash as shell
+SHELL ["/bin/bash", "-c"]
 
 # Install all basic requirements
 RUN \
diff --git a/tests/ci_build/conda_env/linux_sycl_test.yml b/tests/ci_build/conda_env/linux_sycl_test.yml
@@ -19,3 +19,4 @@ dependencies:
 - pytest-cov
 - dpcpp_linux-64
 - onedpl-devel
+- intel-openmp
diff --git a/tests/ci_build/conda_env/win64_test.yml b/tests/ci_build/conda_env/win64_test.yml
@@ -2,11 +2,11 @@ name: win64_env
 channels:
 - conda-forge
 dependencies:
-- python=3.8
+- python=3.10
 - numpy
 - scipy
 - matplotlib
-- scikit-learn
+- scikit-learn<=1.5.2
 - pandas
 - pytest
 - boto3
diff --git a/tests/cpp/collective/test_worker.h b/tests/cpp/collective/test_worker.h
@@ -150,13 +150,13 @@ template <typename WorkerFn>
 void TestDistributedGlobal(std::int32_t n_workers, WorkerFn worker_fn, bool need_finalize = true,
                            std::chrono::seconds test_timeout = std::chrono::seconds{30}) {
   system::SocketStartup();
-  std::chrono::seconds timeout{1};
+  std::chrono::seconds poll_timeout{5};
 
   std::string host;
   auto rc = GetHostAddress(&host);
   SafeColl(rc);
 
-  RabitTracker tracker{MakeTrackerConfig(host, n_workers, timeout)};
+  RabitTracker tracker{MakeTrackerConfig(host, n_workers, poll_timeout)};
   auto fut = tracker.Run();
 
   std::vector<std::thread> workers;
@@ -165,7 +165,7 @@ void TestDistributedGlobal(std::int32_t n_workers, WorkerFn worker_fn, bool need
   for (std::int32_t i = 0; i < n_workers; ++i) {
     workers.emplace_back([=] {
       auto fut = std::async(std::launch::async, [=] {
-        auto config = MakeDistributedTestConfig(host, port, timeout, i);
+        auto config = MakeDistributedTestConfig(host, port, poll_timeout, i);
         Init(config);
         worker_fn();
         if (need_finalize) {
diff --git a/tests/cpp/common/test_quantile.cc b/tests/cpp/common/test_quantile.cc
@@ -283,22 +283,22 @@ void TestColSplitQuantile(size_t rows, size_t cols) {
 }
 }  // anonymous namespace
 
-TEST(Quantile, ColSplitBasic) {
+TEST(Quantile, ColumnSplitBasic) {
   constexpr size_t kRows = 10, kCols = 10;
   TestColSplitQuantile<false>(kRows, kCols);
 }
 
-TEST(Quantile, ColSplit) {
+TEST(Quantile, ColumnSplit) {
   constexpr size_t kRows = 4000, kCols = 200;
   TestColSplitQuantile<false>(kRows, kCols);
 }
 
-TEST(Quantile, ColSplitSortedBasic) {
+TEST(Quantile, ColumnSplitSortedBasic) {
   constexpr size_t kRows = 10, kCols = 10;
   TestColSplitQuantile<true>(kRows, kCols);
 }
 
-TEST(Quantile, ColSplitSorted) {
+TEST(Quantile, ColumnSplitSorted) {
   constexpr size_t kRows = 4000, kCols = 200;
   TestColSplitQuantile<true>(kRows, kCols);
 }
diff --git a/tests/cpp/tree/test_approx.cc b/tests/cpp/tree/test_approx.cc
@@ -123,7 +123,7 @@ void TestColumnSplitPartitioner(size_t n_samples, size_t base_rowid, std::shared
 }
 }  // anonymous namespace
 
-TEST(Approx, PartitionerColSplit) {
+TEST(Approx, PartitionerColumnSplit) {
   size_t n_samples = 1024, n_features = 16, base_rowid = 0;
   auto const Xy = RandomDataGenerator{n_samples, n_features, 0}.GenerateDMatrix(true);
   auto hess = GenerateHess(n_samples);
@@ -153,7 +153,7 @@ TEST(Approx, PartitionerColSplit) {
 }
 
 namespace {
-class TestApproxColSplit : public ::testing::TestWithParam<std::tuple<bool, float>> {
+class TestApproxColumnSplit : public ::testing::TestWithParam<std::tuple<bool, float>> {
  public:
   void Run() {
     auto [categorical, sparsity] = GetParam();
@@ -162,9 +162,9 @@ class TestApproxColSplit : public ::testing::TestWithParam<std::tuple<bool, floa
 };
 }  // namespace
 
-TEST_P(TestApproxColSplit, Basic) { this->Run(); }
+TEST_P(TestApproxColumnSplit, Basic) { this->Run(); }
 
-INSTANTIATE_TEST_SUITE_P(ColumnSplit, TestApproxColSplit, ::testing::ValuesIn([]() {
+INSTANTIATE_TEST_SUITE_P(ColumnSplit, TestApproxColumnSplit, ::testing::ValuesIn([]() {
                            std::vector<std::tuple<bool, float>> params;
                            for (auto categorical : {true, false}) {
                              for (auto sparsity : {0.0f, 0.6f}) {
diff --git a/tests/cpp/tree/test_quantile_hist.cc b/tests/cpp/tree/test_quantile_hist.cc
@@ -198,12 +198,12 @@ void TestColumnSplitPartitioner(bst_target_t n_targets) {
 }
 }  // anonymous namespace
 
-TEST(QuantileHist, PartitionerColSplit) { TestColumnSplitPartitioner<CPUExpandEntry>(1); }
+TEST(QuantileHist, PartitionerColumnSplit) { TestColumnSplitPartitioner<CPUExpandEntry>(1); }
 
-TEST(QuantileHist, MultiPartitionerColSplit) { TestColumnSplitPartitioner<MultiExpandEntry>(3); }
+TEST(QuantileHist, MultiPartitionerColumnSplit) { TestColumnSplitPartitioner<MultiExpandEntry>(3); }
 
 namespace {
-class TestHistColSplit : public ::testing::TestWithParam<std::tuple<bst_target_t, bool, float>> {
+class TestHistColumnSplit : public ::testing::TestWithParam<std::tuple<bst_target_t, bool, float>> {
  public:
   void Run() {
     auto [n_targets, categorical, sparsity] = GetParam();
@@ -212,9 +212,9 @@ class TestHistColSplit : public ::testing::TestWithParam<std::tuple<bst_target_t
 };
 }  // anonymous namespace
 
-TEST_P(TestHistColSplit, Basic) { this->Run(); }
+TEST_P(TestHistColumnSplit, Basic) { this->Run(); }
 
-INSTANTIATE_TEST_SUITE_P(ColumnSplit, TestHistColSplit, ::testing::ValuesIn([]() {
+INSTANTIATE_TEST_SUITE_P(ColumnSplit, TestHistColumnSplit, ::testing::ValuesIn([]() {
                            std::vector<std::tuple<bst_target_t, bool, float>> params;
                            for (auto categorical : {true, false}) {
                              for (auto sparsity : {0.0f, 0.6f}) {
diff --git a/tests/python-gpu/test_gpu_prediction.py b/tests/python-gpu/test_gpu_prediction.py
@@ -453,7 +453,7 @@ def test_predict_leaf_dart(self, param: dict, dataset: tm.TestDataset) -> None:
     )
     @settings(deadline=None, max_examples=20, print_blob=True)
     def test_predict_categorical_split(self, df):
-        from sklearn.metrics import mean_squared_error
+        from sklearn.metrics import root_mean_squared_error
 
         df = df.astype("category")
         x0, x1 = df["x0"].to_numpy(), df["x1"].to_numpy()
@@ -480,7 +480,7 @@ def test_predict_categorical_split(self, df):
         )
         bst.set_param({"device": "cuda:0"})
         pred = bst.predict(dtrain)
-        rmse = mean_squared_error(y_true=y, y_pred=pred, squared=False)
+        rmse = root_mean_squared_error(y_true=y, y_pred=pred)
         np.testing.assert_almost_equal(
             rmse, eval_history["train"]["rmse"][-1], decimal=5
         )
@@ -576,10 +576,9 @@ def test_dtypes(self):
         # boolean
         orig = cp.random.binomial(1, 0.5, size=rows * cols).reshape(rows, cols)
         predt_orig = booster.inplace_predict(orig)
-        for dtype in [cp.bool8, cp.bool_]:
-            X = cp.array(orig, dtype=dtype)
-            predt = booster.inplace_predict(X)
-            cp.testing.assert_allclose(predt, predt_orig)
+        X = cp.array(orig, dtype=cp.bool_)
+        predt = booster.inplace_predict(X)
+        cp.testing.assert_allclose(predt, predt_orig)
 
         # unsupported types
         for dtype in [
diff --git a/tests/test_distributed/test_gpu_with_dask/test_gpu_with_dask.py b/tests/test_distributed/test_gpu_with_dask/test_gpu_with_dask.py
@@ -19,7 +19,7 @@
 pytestmark = [
     pytest.mark.skipif(**tm.no_dask()),
     pytest.mark.skipif(**tm.no_dask_cuda()),
-    tm.timeout(60),
+    tm.timeout(120),
 ]
 
 from ..test_with_dask.test_with_dask import generate_array
@@ -99,6 +99,8 @@ def is_df(part: T) -> T:
 
     cp.testing.assert_allclose(predt.values.compute(), single_node)
 
+    # Work around https://github.com/dmlc/xgboost/issues/10752
+    X.columns = X.columns.astype("object")
     # Make sure the output can be integrated back to original dataframe
     X["predict"] = predictions
     X["inplace_predict"] = series_predictions
