AttributeError: This 'Pipeline' has no attribute 'predict_proba'

I tried to use auto_sklearn2 library for oil fingerprint identificvation from GCXGC image in jpeg. However, I got the following attribution errors "AttributeError: 'LinearSVC' object has no attribute 'predict_proba'" and
"AttributeError: This 'Pipeline' has no attribute 'predict_proba'" when I use the function

` p = self.rf.predict_proba(np.array([[d,nok,n1,n2]]))[0]`

This happened after defining self.rf using AutoSklearnClassifier according to the class defnition:

```
class GCxGC():

self.rf = AutoSklearnClassifier(time_limit=30, random_state=42)
```

I tried to restrict probabilistic classifiers that support predict_proba in the following manner:

```
# Restrict to classifiers that support predict_proba

X = np.array([[d, nok, n1, n2]])
p = safe_predict_proba(self.rf, X)[0]

#########################################################################
def safe_predict_proba(clf, X):
# Always return probability-like predictions from a pipeline.
# Works for Auto-Sklearn pipelines even if classifier is LinearSVC or other non-probabilistic.
# Get the last step of the pipeline


try:
# Try to get probabilities from a model in a robust way.
# Falls back to decision_function or predict if needed.
# If clf is a pipeline, get the final step
if isinstance(clf, Pipeline):
model = clf.steps[-1][1]
else:
model = clf

# Case 1: model has predict_proba
if hasattr(model, "predict_proba"):
return model.predict_proba(X)

# Case 2: model has decision_function
elif hasattr(model, "decision_function"):
scores = model.decision_function(X)
if scores.ndim == 1:
from scipy.special import expit
probs = expit(scores)
return np.vstack([1 - probs, probs]).T
else:
from sklearn.utils.extmath import softmax
return softmax(scores)

# Case 3: fallback → use predict() and convert to one-hot
elif hasattr(model, "predict"):
preds = model.predict(X)
if hasattr(model, "classes_"):
classes = model.classes_
else:
classes = np.unique(preds)

one_hot = np.zeros((len(preds), len(classes)))
for i, c in enumerate(preds):
one_hot[i, np.where(classes == c)[0][0]] = 1.0
return one_hot

else:
raise RuntimeError(f"Model {type(model)} has no method to produce probabilities")

except Exception as e:
raise RuntimeError(f"Failed to compute probabilities: {e}")
#######################################################################
probabilistic_classifiers = [
"random_forest",
"gradient_boosting",
"k_nearest_neighbors",
"logistic_regression",
"multinomial_nb",
"qda",
"sgd", # only if loss is 'log' (logistic regression)
"extra_trees",
"mlp"
]
class GCxGC():

self.rf = AutoSklearnClassifier(time_limit=30, include={"classifier": probabilistic_classifiers},random_state=42) # error

```

However, it is not working since it has given the following error:

> TypeError: AutoSklearnClassifier.__init__() got an unexpected keyword argument 'include'
This is critical since I want to restrict probabilistic_classifiers of AutoSklearnClassifier to this choice
```
# Restrict to classifiers that support predict_proba
probabilistic_classifiers = [
"random_forest",
"gradient_boosting",
"k_nearest_neighbors",
"logistic_regression",
"multinomial_nb",
"qda",
"sgd", # only if loss is 'log' (logistic regression)
"extra_trees",
"mlp"
]
```

Any sugggeston would be appreciated