Error nicely in KMeans if n_samples < n_clusters

python/cuml/cuml/cluster/kmeans.pyx

@@ -399,6 +399,11 @@ class KMeans(Base,
                     "while a minimum of 1 is required by KMeans."
                 )
 
+        if n_rows < self.n_clusters:
+            raise ValueError(
+                f"n_samples={n_rows} should be >= n_clusters={self.n_clusters}."
+            )
+
         self._seed = check_random_seed(self.random_state)
         self.n_features_in_ = n_cols
         self.feature_names_in_ = X_m.index

python/cuml/cuml_accel_tests/upstream/scikit-learn/xfail-list.yaml

@@ -1147,7 +1147,6 @@
   tests:
   - "sklearn.cluster.tests.test_k_means::test_relocating_with_duplicates[elkan-dense]"
   - "sklearn.cluster.tests.test_k_means::test_relocating_with_duplicates[lloyd-dense]"
-  - "sklearn.cluster.tests.test_k_means::test_wrong_params[param0-n_samples.* should be >= n_clusters-KMeans]"
   - "sklearn.cluster.tests.test_k_means::test_wrong_params[param1-The shape of the initial centers .* does not match the number of clusters-KMeans]"
   - "sklearn.cluster.tests.test_k_means::test_wrong_params[param3-The shape of the initial centers .* does not match the number of features of the data-KMeans]"
 - reason: cuml.accel deviates in predict output for KMeans

python/cuml/tests/test_kmeans.py

@@ -449,3 +449,13 @@ def test_kmeans_empty_x():
     y = np.ones(shape=10)
     with pytest.raises(ValueError, match=r"Found array with 0 feature\(s\)"):
         model.fit(X, y)
+
+
+def test_kmeans_n_samples_less_than_n_clusters():
+    model = cuml.KMeans(n_clusters=8)
+
+    X = np.array([[1, 2, 3], [4, 5, 6]], dtype=np.float32)
+    with pytest.raises(
+        ValueError, match="n_samples=2 should be >= n_clusters=8"
+    ):
+        model.fit(X)