AMLB-specific test case added

fedot/api/main.py

@@ -442,6 +442,12 @@ def get_metrics(self,
                                            data_type=self.train_data.data_type)
             else:
                 self.test_data.target = target[:len(self.prediction.predict)]
+        elif not len(self.test_data.target):
+            self.test_data = self.data_processor.define_data(
+                target=self.train_data.target[:len(self.test_data.features)],
+                features=self.test_data.features,
+                is_predict=True,
+            )
 
         metrics = ensure_wrapped_in_sequence(metric_names) if metric_names else self.metrics
         metric_names = [str(metric) for metric in metrics]

fedot/preprocessing/preprocessing.py

@@ -445,7 +445,8 @@ def apply_inverse_target_encoding(self, column_to_transform: np.ndarray) -> np.n
                 # There is no need to perform converting (it was performed already)
                 return column_to_transform
             # It is needed to apply fitted encoder to apply inverse transformation
-            transformed = self.target_encoders[main_target_source_name].inverse_transform(column_to_transform)
+            transformed = self.target_encoders[main_target_source_name].inverse_transform(
+                column_to_transform.astype(int))
 
             # Convert one-dimensional array into column
             if len(transformed.shape) == 1:

test/integration/api/test_main_api.py

@@ -1,6 +1,7 @@
 import os
 import shutil
 from copy import deepcopy
+from pathlib import Path
 
 import numpy as np
 import pandas as pd
@@ -14,6 +15,7 @@
 from fedot.core.pipelines.node import PipelineNode
 from fedot.core.pipelines.pipeline import Pipeline
 from fedot.core.repository.tasks import TsForecastingParams
+from fedot.core.utils import fedot_project_root
 from test.data.datasets import get_dataset, get_multimodal_ts_data, load_categorical_unimodal, \
     load_categorical_multidata
 from test.unit.common_tests import is_predict_ignores_target
@@ -324,3 +326,30 @@ def test_forecast_with_not_ts_problem():
     model.fit(train_data, predefined_model='auto')
     with pytest.raises(ValueError):
         model.forecast(pre_history=test_data)
+
+
+@pytest.mark.parametrize(
+    'initial_assumption, timeout', [
+        (Pipeline(
+            PipelineNode(
+                'catboost', nodes_from=[
+                    PipelineNode(
+                        'resample', nodes_from=[
+                            PipelineNode('scaling')])])), 0.001), (None, 5.0)])
+def test_api_for_amlb(initial_assumption, timeout):
+    amlb_data = Path(fedot_project_root(), 'test', 'data', 'amlb')
+    x_train = np.load(str(Path(amlb_data, 'train_australian_fold7.npy')))
+    y_train = np.load(str(Path(amlb_data, 'target_y.npy')), allow_pickle=True)  # real target from AMLB
+    x_test = np.load(str(Path(amlb_data, 'test_australian_fold7.npy')))
+    # TODO resample add
+    training_params = {"preset": "best_quality", "n_jobs": -1}
+
+    fedot = Fedot(problem='classification', timeout=timeout, metric='roc_auc', seed=0,
+                  max_pipeline_fit_time=1, **training_params, initial_assumption=initial_assumption)
+
+    fedot.fit(features=x_train, target=y_train)
+
+    predictions = fedot.predict(features=x_test)
+    probabilities = fedot.predict_proba(features=x_test, probs_for_all_classes=True)
+    assert predictions is not None
+    assert probabilities is not None

test/unit/api/test_main_api.py

@@ -122,6 +122,25 @@ def test_baseline_with_api():
     assert baseline_metrics['f1'] > 0
 
 
+def test_specific_baseline_with_api():
+    """
+    Check that ``get_metrics`` works correctly if ``target`` is defined as a string and ``test_data`` doesn't contain it
+    """
+    df = pd.read_csv(fedot_project_root().joinpath("test/data/simple_classification.csv"))
+    train_data, test_data = np.array_split(df, 2)
+    test_data.drop("Y", axis=1, inplace=True)
+
+    baseline_model = Fedot(problem="classification", metric=["f1"])
+
+    baseline_model.fit(features=train_data, target="Y", predefined_model="auto")
+
+    prediction = baseline_model.predict(features=test_data)
+    assert len(prediction) == len(test_data)
+
+    baseline_metrics = baseline_model.get_metrics()
+    assert baseline_metrics["f1"]
+
+
 def test_forecast_with_multivariate_ts():
     forecast_length = 2