From 19a5cf77014ff5a729f49dc80611598852dd0396 Mon Sep 17 00:00:00 2001
From: Ravin Kohli <kohliravin7@gmail.com>
Date: Mon, 8 Aug 2022 18:16:48 +0200
Subject: [PATCH 1/3] fix update of pipeline config options in fit pipeline

---
 autoPyTorch/api/base_task.py |  2 +-
 test/test_api/test_api.py    | 56 ++++++++++++++++++++++++++++++++++++
 2 files changed, 57 insertions(+), 1 deletion(-)

diff --git a/autoPyTorch/api/base_task.py b/autoPyTorch/api/base_task.py
index c5468eae7..12d451ca0 100644
--- a/autoPyTorch/api/base_task.py
+++ b/autoPyTorch/api/base_task.py
@@ -1632,7 +1632,7 @@ def fit_pipeline(
                              names=[eval_metric] if eval_metric is not None else None,
                              all_supported_metrics=False).pop()
 
-        pipeline_options = self.pipeline_options.copy().update(pipeline_options) if pipeline_options is not None \
+        pipeline_options = {**self.pipeline_options, **pipeline_options} if pipeline_options is not None \
             else self.pipeline_options.copy()
 
         assert pipeline_options is not None
diff --git a/test/test_api/test_api.py b/test/test_api/test_api.py
index 465d74c6b..6a9d9d6ba 100644
--- a/test/test_api/test_api.py
+++ b/test/test_api/test_api.py
@@ -932,6 +932,62 @@ def test_pipeline_fit(openml_id,
             assert not os.path.exists(cv_model_path)
 
 
+@pytest.mark.parametrize('openml_id', (40984,))
+@pytest.mark.parametrize("budget", [1])
+def test_pipeline_fit_pass_pipeline_options(
+    openml_id,
+    backend,
+    budget,
+    n_samples):
+    # Get the data and check that contents of data-manager make sense
+    X, y = sklearn.datasets.fetch_openml(
+        data_id=int(openml_id),
+        return_X_y=True, as_frame=True
+    )
+    X_train, X_test, y_train, y_test = sklearn.model_selection.train_test_split(
+        X[:n_samples], y[:n_samples], random_state=1)
+
+    # Search for a good configuration
+    estimator = TabularClassificationTask(
+        backend=backend,
+        ensemble_size=0
+    )
+
+    dataset = estimator.get_dataset(X_train=X_train,
+                                    y_train=y_train,
+                                    X_test=X_test,
+                                    y_test=y_test)
+
+    configuration = estimator.get_search_space(dataset).get_default_configuration()
+    pipeline, run_info, run_value, dataset = estimator.fit_pipeline(dataset=dataset,
+                                                                    configuration=configuration,
+                                                                    run_time_limit_secs=50,
+                                                                    budget_type='epochs',
+                                                                    budget=budget,
+                                                                    pipeline_options={'early_stopping': 100}
+                                                                    )
+    assert isinstance(dataset, BaseDataset)
+    assert isinstance(run_info, RunInfo)
+    assert isinstance(run_info.config, Configuration)
+
+    assert isinstance(run_value, RunValue)
+    assert 'SUCCESS' in str(run_value.status)
+
+    # Make sure that the pipeline can be pickled
+    dump_file = os.path.join(tempfile.gettempdir(), 'automl.dump.pkl')
+    with open(dump_file, 'wb') as f:
+        pickle.dump(pipeline, f)
+
+    num_run_dir = estimator._backend.get_numrun_directory(
+        run_info.seed, run_value.additional_info['num_run'], budget=float(budget))
+    model_path = os.path.join(num_run_dir, estimator._backend.get_model_filename(
+        run_info.seed, run_value.additional_info['num_run'], budget=float(budget)))
+
+    # We expect the model path always
+    # And the cv model only on 'cv'
+    assert os.path.exists(model_path)
+
+
 @pytest.mark.parametrize('openml_id', (40984,))
 @pytest.mark.parametrize('resampling_strategy,resampling_strategy_args',
                          ((HoldoutValTypes.holdout_validation, {'val_share': 0.8}),

From 09bf3c69c2fb81bd4f3ed2d0bf11ce8a85e87e3b Mon Sep 17 00:00:00 2001
From: Ravin Kohli <kohliravin7@gmail.com>
Date: Mon, 8 Aug 2022 18:18:04 +0200
Subject: [PATCH 2/3] fix flake and test

---
 test/test_api/test_api.py | 3 ++-
 1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/test/test_api/test_api.py b/test/test_api/test_api.py
index 6a9d9d6ba..118bda718 100644
--- a/test/test_api/test_api.py
+++ b/test/test_api/test_api.py
@@ -938,7 +938,8 @@ def test_pipeline_fit_pass_pipeline_options(
     openml_id,
     backend,
     budget,
-    n_samples):
+    n_samples
+):
     # Get the data and check that contents of data-manager make sense
     X, y = sklearn.datasets.fetch_openml(
         data_id=int(openml_id),

From 08bddb68f56859fab842b7aeb5c5e12269729510 Mon Sep 17 00:00:00 2001
From: Ravin Kohli <kohliravin7@gmail.com>
Date: Fri, 12 Aug 2022 13:27:12 +0200
Subject: [PATCH 3/3] suggestions from review

---
 test/test_api/test_api.py | 4 +---
 1 file changed, 1 insertion(+), 3 deletions(-)

diff --git a/test/test_api/test_api.py b/test/test_api/test_api.py
index 118bda718..d95f3943f 100644
--- a/test/test_api/test_api.py
+++ b/test/test_api/test_api.py
@@ -932,8 +932,7 @@ def test_pipeline_fit(openml_id,
             assert not os.path.exists(cv_model_path)
 
 
-@pytest.mark.parametrize('openml_id', (40984,))
-@pytest.mark.parametrize("budget", [1])
+@pytest.mark.parametrize('openml_id,budget', [(40984, 1)])
 def test_pipeline_fit_pass_pipeline_options(
     openml_id,
     backend,
@@ -985,7 +984,6 @@ def test_pipeline_fit_pass_pipeline_options(
         run_info.seed, run_value.additional_info['num_run'], budget=float(budget)))
 
     # We expect the model path always
-    # And the cv model only on 'cv'
     assert os.path.exists(model_path)