[BUG] Return a boolean fitted state from MultiNormalizer

pytorch_forecasting/data/encoders.py

@@ -1568,7 +1568,9 @@ def func(*args, **kwargs):
                                 for key, val in kwargs.items()
                             }
                             results.append(getattr(norm, name)(*new_args, **new_kwargs))
-                        return results
+                        return (
+                            all(results) if name == "__sklearn_is_fitted__" else results
+                        )
 
                     return func
                 else:

tests/test_data/test_encoders.py

@@ -4,6 +4,8 @@
 import numpy as np
 import pandas as pd
 import pytest
+from sklearn.pipeline import Pipeline
+from sklearn.preprocessing import StandardScaler
 from sklearn.utils.validation import NotFittedError, check_is_fitted
 import torch
 
@@ -182,6 +184,25 @@ def test_MultiNormalizer_fitted():
         pytest.fail(f"{NotFittedError}")
 
 
+def test_MultiNormalizer_requires_all_pipeline_normalizers_fitted():
+    data = np.array([[1.0], [2.0]])
+    normalizers = [
+        Pipeline([("scaler", StandardScaler())]),
+        Pipeline([("scaler", StandardScaler())]),
+    ]
+    normalizer = MultiNormalizer(normalizers)
+
+    with pytest.raises(NotFittedError):
+        check_is_fitted(normalizer)
+
+    normalizers[0].fit(data)
+    with pytest.raises(NotFittedError):
+        check_is_fitted(normalizer)
+
+    normalizers[1].fit(data)
+    check_is_fitted(normalizer)
+
+
 def test_TorchNormalizer_dtype_consistency():
     """
     - Ensures that even for float64 `target_scale`, the transformation will not change the prediction dtype.