Support alpha=0 in Ridge

When `alpha = 0`, `Ridge` is equivalent to a `LinearRegression`.
Previously we checked if alpha was positive, scikit-learn instead
requires that alpha is non-negative. This updates the check and adds a
test.

python/cuml/cuml/linear_model/ridge.pyx

@@ -1,5 +1,5 @@
 #
-# Copyright (c) 2019-2024, NVIDIA CORPORATION.
+# Copyright (c) 2019-2025, NVIDIA CORPORATION.
 #
 # Licensed under the Apache License, Version 2.0 (the "License");
 # you may not use this file except in compliance with the License.
@@ -241,9 +241,8 @@ class Ridge(UniversalBase,
         self.intercept_value = 0.0
 
     def _check_alpha(self, alpha):
-        if alpha <= 0.0:
-            msg = "alpha value has to be positive"
-            raise TypeError(msg.format(alpha))
+        if alpha < 0.0:
+            raise ValueError(f"alpha must be non-negative, got {alpha}")
 
     def _get_algorithm_int(self, algorithm):
         if self.solver not in ['svd', 'eig', 'cd']:

python/cuml/cuml/tests/test_linear_model.py

@@ -417,6 +417,17 @@ def test_ridge_regression_model(datatype, algorithm, nrows, column_info):
         )
 
 
+def test_ridge_and_least_squares_equal_when_alpha_is_0():
+    X, y = make_regression(n_samples=5, n_features=4, random_state=0)
+
+    ridge = cuRidge(alpha=0.0, fit_intercept=False)
+    ols = cuLinearRegression(fit_intercept=False)
+
+    ridge.fit(X, y)
+    ols.fit(X, y)
+    assert array_equal(ridge.coef_, ols.coef_)
+
+
 @pytest.mark.parametrize("datatype", [np.float32, np.float64])
 @pytest.mark.parametrize("algorithm", ["eig", "svd"])
 @pytest.mark.parametrize(