Merge branch 'master' into sync-condition-2023May15

2023-05-23 01:07:50 +02:00 · 2023-05-23 01:07:50 +02:00 · b994a38b28
commit b994a38b28
parent 3a834c4992 3913ff470f
8 changed files with 66 additions and 51 deletions
--- a/python-package/xgboost/testing/init.py
+++ b/python-package/xgboost/testing/init.py
@ -603,26 +603,6 @@ sparse_datasets_strategy = strategies.sampled_from(
    ]
 )

-_unweighted_datasets_strategy = strategies.sampled_from(
-    [
-        TestDataset(
-            "calif_housing", get_california_housing, "reg:squarederror", "rmse"
-        ),
-        TestDataset(
-            "calif_housing-l1", get_california_housing, "reg:absoluteerror", "mae"
-        ),
-        TestDataset("cancer", get_cancer, "binary:logistic", "logloss"),
-        TestDataset("sparse", get_sparse, "reg:squarederror", "rmse"),
-        TestDataset("sparse-l1", get_sparse, "reg:absoluteerror", "mae"),
-        TestDataset(
-            "empty",
-            lambda: (np.empty((0, 100)), np.empty(0)),
-            "reg:squarederror",
-            "rmse",
-        ),
-    ]
-)
-

 def make_datasets_with_margin(
    unweighted_strategy: strategies.SearchStrategy,
@ -664,7 +644,28 @@ def make_datasets_with_margin(

 # A strategy for drawing from a set of example datasets. May add random weights to the
 # dataset
-dataset_strategy = make_datasets_with_margin(_unweighted_datasets_strategy)()
+@memory.cache
+def make_dataset_strategy() -> Callable:
+    _unweighted_datasets_strategy = strategies.sampled_from(
+        [
+            TestDataset(
+                "calif_housing", get_california_housing, "reg:squarederror", "rmse"
+            ),
+            TestDataset(
+                "calif_housing-l1", get_california_housing, "reg:absoluteerror", "mae"
+            ),
+            TestDataset("cancer", get_cancer, "binary:logistic", "logloss"),
+            TestDataset("sparse", get_sparse, "reg:squarederror", "rmse"),
+            TestDataset("sparse-l1", get_sparse, "reg:absoluteerror", "mae"),
+            TestDataset(
+                "empty",
+                lambda: (np.empty((0, 100)), np.empty(0)),
+                "reg:squarederror",
+                "rmse",
+            ),
+        ]
+    )
+    return make_datasets_with_margin(_unweighted_datasets_strategy)()


 _unweighted_multi_datasets_strategy = strategies.sampled_from(
--- a/tests/python-gpu/test_gpu_linear.py
+++ b/tests/python-gpu/test_gpu_linear.py
@ -28,8 +28,7 @@ def train_result(param, dmat, num_rounds):


 class TestGPULinear:
-    @given(parameter_strategy, strategies.integers(10, 50),
-           tm.dataset_strategy)
+    @given(parameter_strategy, strategies.integers(10, 50), tm.make_dataset_strategy())
    @settings(deadline=None, max_examples=20, print_blob=True)
    def test_gpu_coordinate(self, param, num_rounds, dataset):
        assume(len(dataset.y) > 0)
@ -45,7 +44,7 @@ class TestGPULinear:
    @given(
        parameter_strategy,
        strategies.integers(10, 50),
-        tm.dataset_strategy,
+        tm.make_dataset_strategy(),
        strategies.floats(1e-5, 0.8),
        strategies.floats(1e-5, 0.8)
    )
--- a/tests/python-gpu/test_gpu_prediction.py
+++ b/tests/python-gpu/test_gpu_prediction.py
@ -265,8 +265,9 @@ class TestGPUPredict:
        base_margin = cudf.Series(rng.randn(rows))
        self.run_inplace_base_margin(booster, dtrain, X, base_margin)

-    @given(strategies.integers(1, 10),
-           tm.dataset_strategy, shap_parameter_strategy)
+    @given(
+        strategies.integers(1, 10), tm.make_dataset_strategy(), shap_parameter_strategy
+    )
    @settings(deadline=None, max_examples=20, print_blob=True)
    def test_shap(self, num_rounds, dataset, param):
        if dataset.name.endswith("-l1"):  # not supported by the exact tree method
@ -281,8 +282,9 @@ class TestGPUPredict:
        assume(len(dataset.y) > 0)
        assert np.allclose(np.sum(shap, axis=len(shap.shape) - 1), margin, 1e-3, 1e-3)

-    @given(strategies.integers(1, 10),
-           tm.dataset_strategy, shap_parameter_strategy)
+    @given(
+        strategies.integers(1, 10), tm.make_dataset_strategy(), shap_parameter_strategy
+    )
    @settings(deadline=None, max_examples=10, print_blob=True)
    def test_shap_interactions(self, num_rounds, dataset, param):
        if dataset.name.endswith("-l1"):  # not supported by the exact tree method
@ -335,7 +337,7 @@ class TestGPUPredict:

        np.testing.assert_equal(cpu_leaf, gpu_leaf)

-    @given(predict_parameter_strategy, tm.dataset_strategy)
+    @given(predict_parameter_strategy, tm.make_dataset_strategy())
    @settings(deadline=None, max_examples=20, print_blob=True)
    def test_predict_leaf_gbtree(self, param, dataset):
        # Unsupported for random forest
@ -346,7 +348,7 @@ class TestGPUPredict:
        param['tree_method'] = 'gpu_hist'
        self.run_predict_leaf_booster(param, 10, dataset)

-    @given(predict_parameter_strategy, tm.dataset_strategy)
+    @given(predict_parameter_strategy, tm.make_dataset_strategy())
    @settings(deadline=None, max_examples=20, print_blob=True)
    def test_predict_leaf_dart(self, param: dict, dataset: tm.TestDataset) -> None:
        # Unsupported for random forest
--- a/tests/python-gpu/test_gpu_updaters.py
+++ b/tests/python-gpu/test_gpu_updaters.py
@ -48,7 +48,9 @@ class TestGPUUpdatersMulti:
 class TestGPUUpdaters:
    cputest = test_up.TestTreeMethod()

-    @given(hist_parameter_strategy, strategies.integers(1, 20), tm.dataset_strategy)
+    @given(
+        hist_parameter_strategy, strategies.integers(1, 20), tm.make_dataset_strategy()
+    )
    @settings(deadline=None, max_examples=50, print_blob=True)
    def test_gpu_hist(self, param, num_rounds, dataset):
        param["tree_method"] = "gpu_hist"
@ -150,7 +152,7 @@ class TestGPUUpdaters:
    @given(
        hist_parameter_strategy,
        strategies.integers(1, 20),
-        tm.dataset_strategy
+        tm.make_dataset_strategy(),
    )
    @settings(deadline=None, max_examples=20, print_blob=True)
    def test_gpu_hist_device_dmatrix(
@ -171,7 +173,7 @@ class TestGPUUpdaters:
    @given(
        hist_parameter_strategy,
        strategies.integers(1, 3),
-        tm.dataset_strategy
+        tm.make_dataset_strategy(),
    )
    @settings(deadline=None, max_examples=10, print_blob=True)
    def test_external_memory(self, param, num_rounds, dataset):
@ -213,7 +215,7 @@ class TestGPUUpdaters:
        np.testing.assert_allclose(predictions, 0.0, 1e-6)

    @pytest.mark.mgpu
-    @given(tm.dataset_strategy, strategies.integers(0, 10))
+    @given(tm.make_dataset_strategy(), strategies.integers(0, 10))
    @settings(deadline=None, max_examples=10, print_blob=True)
    def test_specified_gpu_id_gpu_update(self, dataset, gpu_id):
        param = {'tree_method': 'gpu_hist', 'gpu_id': gpu_id}
--- a/tests/python/test_linear.py
+++ b/tests/python/test_linear.py
@ -28,8 +28,12 @@ def train_result(param, dmat, num_rounds):


 class TestLinear:
-    @given(parameter_strategy, strategies.integers(10, 50),
-           tm.dataset_strategy, coord_strategy)
+    @given(
+        parameter_strategy,
+        strategies.integers(10, 50),
+        tm.make_dataset_strategy(),
+        coord_strategy
+    )
    @settings(deadline=None, max_examples=20, print_blob=True)
    def test_coordinate(self, param, num_rounds, dataset, coord_param):
        param['updater'] = 'coord_descent'
@ -45,7 +49,7 @@ class TestLinear:
    @given(
        parameter_strategy,
        strategies.integers(10, 50),
-        tm.dataset_strategy,
+        tm.make_dataset_strategy(),
        coord_strategy,
        strategies.floats(1e-5, 0.8),
        strategies.floats(1e-5, 0.8)
@ -61,8 +65,9 @@ class TestLinear:
        note(result)
        assert tm.non_increasing([result[0], result[-1]])

-    @given(parameter_strategy, strategies.integers(10, 50),
-           tm.dataset_strategy)
+    @given(
+        parameter_strategy, strategies.integers(10, 50), tm.make_dataset_strategy()
+    )
    @settings(deadline=None, max_examples=20, print_blob=True)
    def test_shotgun(self, param, num_rounds, dataset):
        param['updater'] = 'shotgun'
@ -77,9 +82,13 @@ class TestLinear:
            sampled_result = result
        assert tm.non_increasing(sampled_result)

-    @given(parameter_strategy, strategies.integers(10, 50),
-           tm.dataset_strategy, strategies.floats(1e-5, 1.0),
-           strategies.floats(1e-5, 1.0))
+    @given(
+        parameter_strategy,
+        strategies.integers(10, 50),
+        tm.make_dataset_strategy(),
+        strategies.floats(1e-5, 1.0),
+        strategies.floats(1e-5, 1.0)
+    )
    @settings(deadline=None, max_examples=20, print_blob=True)
    def test_shotgun_regularised(self, param, num_rounds, dataset, alpha, lambd):
        param['updater'] = 'shotgun'
--- a/tests/python/test_updaters.py
+++ b/tests/python/test_updaters.py
@ -87,8 +87,9 @@ class TestTreeMethod:
    USE_ONEHOT = np.iinfo(np.int32).max
    USE_PART = 1

-    @given(exact_parameter_strategy, strategies.integers(1, 20),
-           tm.dataset_strategy)
+    @given(
+        exact_parameter_strategy, strategies.integers(1, 20), tm.make_dataset_strategy()
+    )
    @settings(deadline=None, print_blob=True)
    def test_exact(self, param, num_rounds, dataset):
        if dataset.name.endswith("-l1"):
@ -102,7 +103,7 @@ class TestTreeMethod:
        exact_parameter_strategy,
        hist_parameter_strategy,
        strategies.integers(1, 20),
-        tm.dataset_strategy,
+        tm.make_dataset_strategy(),
    )
    @settings(deadline=None, print_blob=True)
    def test_approx(self, param, hist_param, num_rounds, dataset):
@ -141,7 +142,7 @@ class TestTreeMethod:
        exact_parameter_strategy,
        hist_parameter_strategy,
        strategies.integers(1, 20),
-        tm.dataset_strategy
+        tm.make_dataset_strategy()
    )
    @settings(deadline=None, print_blob=True)
    def test_hist(self, param: dict, hist_param: dict, num_rounds: int, dataset: tm.TestDataset) -> None:
--- 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
@ -243,7 +243,7 @@ class TestDistributedGPU:
    @given(
        params=hist_parameter_strategy,
        num_rounds=strategies.integers(1, 20),
-        dataset=tm.dataset_strategy,
+        dataset=tm.make_dataset_strategy(),
        dmatrix_type=strategies.sampled_from(
            [dxgb.DaskDMatrix, dxgb.DaskQuantileDMatrix]
        ),
--- a/tests/test_distributed/test_with_dask/test_with_dask.py
+++ b/tests/test_distributed/test_with_dask/test_with_dask.py
@ -1458,9 +1458,10 @@ class TestWithDask:
        else:
            assert history[-1] < history[0]

-    @given(params=hist_parameter_strategy,
-           dataset=tm.dataset_strategy)
-    @settings(deadline=None, max_examples=10, suppress_health_check=suppress, print_blob=True)
+    @given(params=hist_parameter_strategy, dataset=tm.make_dataset_strategy())
+    @settings(
+        deadline=None, max_examples=10, suppress_health_check=suppress, print_blob=True
+    )
    def test_hist(
            self, params: Dict, dataset: tm.TestDataset, client: "Client"
    ) -> None:
@ -1524,7 +1525,7 @@ class TestWithDask:
        rmse = result["history"]["Valid"]["rmse"][-1]
        assert rmse < 32.0

-    @given(params=hist_parameter_strategy, dataset=tm.dataset_strategy)
+    @given(params=hist_parameter_strategy, dataset=tm.make_dataset_strategy())
    @settings(
        deadline=None, max_examples=10, suppress_health_check=suppress, print_blob=True
    )