XGB predict_proba and predict method not defined self.X

[MatufA]

After training new XGBClassifier I tried to load and predict using predict_proba and predict methods and got the error:

AttributeError: 'XGBClassifier' object has no attribute 'X'

After debugging the source code I noticed they not initialized correctly the self.X attribute.
for example in predict_proba implementation:

VerticaPy/verticapy/machine_learning/vertica/base.py

Lines 4595 to 4596 in 4d84228

	# Inititalization
	X = format_type(X, dtype=list, na_out=self.X)

self.X never define before (same for predict).

the only time the methods work fine for me was after fit, due to initialization in fit I suspect.

VerticaPy/verticapy/machine_learning/vertica/base.py

Lines 2313 to 2319 in 4d84228

	# Initialization
	if self.overwrite_model:
	self.drop()
	else:
	self._is_already_stored(raise_error=True)
	X = format_type(X, dtype=list)
	self.X = quote_ident(X)

[oualib]

Thanks for reporting this issue, I'll do some tests and come back to you.

[oualib]

You can not predict if you did not train a model.

I just trained and load a XGBClassifier and I got no issue.

from verticapy.machine_learning.vertica import XGBClassifier

table = "titanic"
X = ["fare", "age"]
y = "survived"
model = XGBClassifier("model_test_260924")
model.drop()
model.fit(table, X, y)
model.X

Result:

['"fare"', '"age"']

Loading the model and checking attribute X:

from verticapy.machine_learning.vertica import load_model

model2 = load_model("model_test_260924")
model2.X

Result:

['fare', 'age']

Besides, it is obvious that attributes X and y are created when fitting the model. It is always advise to write entirely the table name, X and y whenever making a prediction AS the variables name might not be the same in the new table.