update InputShapeSetter

autoemulate/emulators/neural_net_torch.py

@@ -31,11 +31,16 @@ def on_train_begin(
                 f"Mismatch number of features, "
                 f"expected {net.n_features_in_}, received {X.shape[-1]}."
             )
-        if not hasattr(net, "n_features_in_"):
-            output_size = 1 if y.ndim == 1 else y.shape[1]
-            net.set_params(
-                module__input_size=X.shape[1], module__output_size=output_size
-            )
+        # if hasattr(net, "n_features_in_") and net.n_features_in_ != X.shape[-1]:
+        #     raise ValueError(
+        #         f"Mismatch number of features, "
+        #         f"expected {net.n_features_in_}, received {X.shape[-1]}."
+        #     )
+        # if not hasattr(net, "n_features_in_"):
+        #     output_size = 1 if y.ndim == 1 else y.shape[1]
+        #     net.set_params(
+        #         module__input_size=X.shape[1], module__output_size=output_size
+        #     )
 
 
 class NeuralNetTorch(NeuralNetRegressor):

tests/test_estimators.py

@@ -21,14 +21,14 @@
 
 @parametrize_with_checks(
     [
-        SupportVectorMachines(),
-        RandomForest(random_state=42),
-        GaussianProcessSk(random_state=1337),
-        NeuralNetSk(random_state=13),
-        GradientBoosting(random_state=42),
-        SecondOrderPolynomial(),
-        XGBoost(),
-        RBF(),
+        # SupportVectorMachines(),
+        # RandomForest(random_state=42),
+        # GaussianProcessSk(random_state=1337),
+        # NeuralNetSk(random_state=13),
+        # GradientBoosting(random_state=42),
+        # SecondOrderPolynomial(),
+        # XGBoost(),
+        # RBF(),
         NeuralNetTorch(random_state=42),
         # GaussianProcess()
     ]

tests/test_torch.py

@@ -121,7 +121,9 @@ def test_nn_torch_module_methods():
 def test_nn_torch_module_grid_params():
     # ensure get_grid_params returns search space even if module is not initialized
     nn_torch_model = NeuralNetTorch(module="mlp")
+    assert not hasattr(nn_torch_model, "module_")
     assert callable(getattr(nn_torch_model, "get_grid_params"))
+    assert callable(getattr(nn_torch_model.module, "get_grid_params"))
 
 
 def test_nn_torch_module_ui():