also set alpha in golden master

tests/glm/test_glm.py

@@ -430,7 +430,10 @@ def test_glm_fit_intercept_argument(estimator, fit_intercept):
 )
 def test_glm_solver_argument(estimator, solver, l1_ratio, y, X):
     """Test GLM for invalid solver argument."""
-    glm = estimator(solver=solver, l1_ratio=l1_ratio, alpha=1.0)
+    kwargs = {"solver": solver, "l1_ratio": l1_ratio}
+    if estimator == GeneralizedLinearRegressor:
+        kwargs["alpha"] = 1.0
+    glm = estimator(**kwargs)
     with pytest.raises(ValueError):
         glm.fit(X, y)
 

tests/glm/test_golden_master.py

@@ -100,25 +100,32 @@ def expected_all():
 
 
 gm_model_parameters = {
-    "default": {},  # default params
-    "half-regularization": {"alpha": 0.5},  # regularization (other than alpha = 1)
-    "elastic-net": {"l1_ratio": 0.5},  # elastic-net
-    "lasso": {"l1_ratio": 1},  # lasso
+    "regularization": {"alpha": 1.0},  # regularization with alpha = 1
+    "half-regularization": {"alpha": 0.5},  # regularization with alpha = 0
+    "elastic-net": {"l1_ratio": 0.5, "alpha": 1.0},  # elastic-net
+    "lasso": {"l1_ratio": 1, "alpha": 1.0},  # lasso
     "variable_p1": {
         "l1_ratio": 1,
         "P1": np.arange(30) / 10,
+        "alpha": 1.0,
     },  # lasso with variable penalty
     "variable_p2": {
         "l1_ratio": 0,
         "P2": _make_P2(),
+        "alpha": 1.0,
     },  # ridge with Tikhonov regularization
     "variable_p1_p2": {
         "l1_ratio": 0.5,
         "P1": np.arange(30) / 10,
         "P2": _make_P2(),
+        "alpha": 1.0,
     },  # elastic net with P1 and P2 variable penalty
-    "fit_intercept": {"fit_intercept": False},  # do not fit the intercept
-    "bounds": {"lower_bounds": np.full(30, 0), "upper_bounds": np.full(30, 0.4)},
+    "fit_intercept": {"fit_intercept": False, "alpha": 1.0},  # do not fit the intercept
+    "bounds": {
+        "lower_bounds": np.full(30, 0),
+        "upper_bounds": np.full(30, 0.4),
+        "alpha": 1.0,
+    },
 }