java.lang.IllegalArgumentException: Attribute 'xgboost.sklearn.XGBClassifier._le' has an unsupported value (Python class sklearn.preprocessing._label.LabelEncoder)

[itzikjan]

Hi,

We are using this package for a long time at production with python 2.7 with the following code:

params2 = {'n_estimators': 100,
'learning_rate': 0.5,
'seed': 0,
'subsample': 0.8,
'n_jobs': 50,
'colsample_bytree': 0.8,
'objective': 'binary:logistic',
'max_depth': 10,
'min_child_weight': 300,
'gamma': 2,
'max_delta_step': 6
}

estimator = xgb.XGBClassifier(**params2)
mapper = DataFrameMapper([(i, None) if j != 'object' and j != 'bool' else (i,
[CategoricalDomain(
missing_value_treatment="as_value",
invalid_value_treatment="as_missing",
missing_value_replacement=train_x[
i].value_counts().idxmax(),
invalid_value_replacement=train_x[
i].value_counts().idxmax()),
LabelEncoder()])
for i, j in
zip(train_x.columns.values, train_x.dtypes.values)]
, input_df=True, df_out=True)

rf_pipeline = PMMLPipeline([("mapper", mapper), ("classifier", estimator)])
rf_pipeline.fit(train_x, train_y)
sklearn2pmml(rf_pipeline, pmml_model_name, with_repr=True)

pip3 freeze

ai-model-infra==0.1
awscli==1.16.300
beautifulsoup4==4.7.1
boto==2.49.0
boto3==1.10.36
botocore==1.13.36
certifi==2019.11.28
chardet==3.0.4
colorama==0.4.1
Cython==0.29.14
datadog==0.32.0
decorator==4.4.1
docutils==0.15.2
fsspec==0.6.1
idna==2.8
jmespath==0.9.3
joblib==0.14.1
lxml==4.3.0
mysqlclient==1.3.14
nltk==3.4
nose==1.3.4
numpy==1.17.4
ortools==7.4.7247
pandas==0.25.3
pandasql==0.7.3
protobuf==3.11.1
py-dateutil==2.2
pyarrow==0.13.0
pyasn1==0.4.8
python-dateutil==2.8.0
python36-sagemaker-pyspark==1.2.1
pytz==2018.9
PyYAML==3.11
requests==2.22.0
rsa==3.4.2
s3fs==0.4.0
s3transfer==0.2.1
scikit-learn==0.22
scipy==1.3.3
singledispatch==3.4.0.3
six==1.12.0
sklearn==0.0
sklearn-pandas==1.8.0
sklearn2pmml==0.51.0
soupsieve==1.6.2
SQLAlchemy==1.3.11
urllib3==1.25.7
windmill==1.6
xgboost==0.90

We are moving to python 3.6. and we are getting the following error (versions: 0.47.1 and 0.51.0)

1. Standard output is empty
   Standard error:
   Dec 11, 2019 1:17:18 PM org.jpmml.sklearn.Main run
   INFO: Parsing PKL..
   Dec 11, 2019 1:17:18 PM org.jpmml.sklearn.Main run
   INFO: Parsed PKL in 132 ms.
   Dec 11, 2019 1:17:18 PM org.jpmml.sklearn.Main run
   INFO: Converting..
   Dec 11, 2019 1:17:18 PM org.jpmml.sklearn.Main run
   SEVERE: Failed to convert
   java.lang.IllegalArgumentException: Attribute 'xgboost.sklearn.XGBClassifier._le' has an unsupported value (Python class sklearn.preprocessing._label.LabelEncoder)
   at org.jpmml.sklearn.CastFunction.apply(CastFunction.java:43)
   at org.jpmml.sklearn.PyClassDict.get(PyClassDict.java:57)
   at sklearn.LabelEncoderClassifier.getLabelEncoder(LabelEncoderClassifier.java:40)
   at sklearn.LabelEncoderClassifier.getClasses(LabelEncoderClassifier.java:34)
   at sklearn.ClassifierUtil.getClasses(ClassifierUtil.java:32)
   at sklearn2pmml.pipeline.PMMLPipeline.encodePMML(PMMLPipeline.java:128)
   at org.jpmml.sklearn.Main.run(Main.java:145)
   at org.jpmml.sklearn.Main.main(Main.java:94)
   Caused by: java.lang.ClassCastException: Cannot cast net.razorvine.pickle.objects.ClassDict to sklearn.preprocessing.LabelEncoder
   at java.lang.Class.cast(Class.java:3369)
   at org.jpmml.sklearn.CastFunction.apply(CastFunction.java:41)
   ... 7 more

Exception in thread "main" java.lang.IllegalArgumentException: Attribute 'xgboost.sklearn.XGBClassifier._le' has an unsupported value (Python class sklearn.preprocessing._label.LabelEncoder)
at org.jpmml.sklearn.CastFunction.apply(CastFunction.java:43)
at org.jpmml.sklearn.PyClassDict.get(PyClassDict.java:57)
at sklearn.LabelEncoderClassifier.getLabelEncoder(LabelEncoderClassifier.java:40)
at sklearn.LabelEncoderClassifier.getClasses(LabelEncoderClassifier.java:34)
at sklearn.ClassifierUtil.getClasses(ClassifierUtil.java:32)
at sklearn2pmml.pipeline.PMMLPipeline.encodePMML(PMMLPipeline.java:128)
at org.jpmml.sklearn.Main.run(Main.java:145)
at org.jpmml.sklearn.Main.main(Main.java:94)
Caused by: java.lang.ClassCastException: Cannot cast net.razorvine.pickle.objects.ClassDict to sklearn.preprocessing.LabelEncoder
at java.lang.Class.cast(Class.java:3369)
at org.jpmml.sklearn.CastFunction.apply(CastFunction.java:41)
... 7 more

2. In other versions we also got the following error:
   ('Invalid value treatment {0} does not support invalid_value_replacement attribute', 'as_missing')

[vruusmann]

TLDR: During Scikit-Learn version upgrade from 0.21.X to 0.22.X many modules were renamed (typically, by prepending an underscore character to the module name). For example, sklearn.preprocessing.label.LabelEncoder became sklearn.preprocessing._label.LabelEncoder.

If you're using Scikit-Learn 0.22.X (or newer), then you need to upgrade to SkLearn2PMML version 0.51.X (or newer). For example, SkLearn2PMML version 0.51.0, which is based on JPMML-SkLearn version 1.5.25 knows both label and _label modules:
https://github.com/jpmml/jpmml-sklearn/blob/1.5.25/src/main/resources/META-INF/sklearn2pmml.properties#L121-L122

Exception in thread "main" java.lang.IllegalArgumentException: Attribute 'xgboost.sklearn.XGBClassifier._le' has an unsupported value (Python class sklearn.preprocessing._label.LabelEncoder)

Please upgrade to SkLearn2PMML version 0.51.0 (or newer).

'Invalid value treatment {0} does not support invalid_value_replacement attribute', 'as_missing'

This is a legitimate complaint. Older SkLearn2PMML versions did not check for conflicting domain attribute values, whereas newer ones do.

Please update your Python source code. Specifically, remove any domain attribute values that you're not 100% sure about.