problem with sckit-learn GradientBoostingRegressor

[doolingdavid]

When trying to write out a pipeline with a GradientBoostingRegressor, I get an error. For example, using the example code from the previous now closed issue:

import numpy as np
from sklearn.ensemble import GradientBoostingRegressor
from sklearn.pipeline import Pipeline
from nyoka import skl_to_pmml
X = np.array([572,565,647,514,571,447,529,586,763,572]).reshape(-1, 1)
y = np.array([513, 593, 565, 571, 517, 586, 517, 460, 521, 620])

est = GradientBoostingRegressor(n_estimators=1, max_depth=1, random_state=1)
pipe = Pipeline([('est', est)])
pipe.fit(X, y)

skl_to_pmml(pipe, ['feature1'], 'est', "..\\nyoka_out.pmml")

yields this error:

---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)
<ipython-input-210-9b2c81fdcff7> in <module>
----> 1 skl_to_pmml(pipe, ['feature1'], 'est', "..\\nyoka_out.pmml")

/opt/anaconda3/lib/python3.7/site-packages/nyoka/skl/skl_to_pmml.py in skl_to_pmml(pipeline, col_names, target_name, pmml_f_name, model_name, description)
     64                                       mining_imp_val,
     65                                       categoric_values,
---> 66                                       model_name)
     67 
     68         pmml = pml.PMML(

/opt/anaconda3/lib/python3.7/site-packages/nyoka/skl/skl_to_pmml.py in get_PMML_kwargs(model, derived_col_names, col_names, target_name, mining_imp_val, categoric_values, model_name)
    185                                                           mining_imp_val,
    186                                                           categoric_values,
--> 187                                                           model_name)}
    188     elif any_in(neurl_netwk_model_names, skl_mdl_super_cls_names):
    189         algo_kwargs = {'NeuralNetwork': get_neural_models(model,

/opt/anaconda3/lib/python3.7/site-packages/nyoka/skl/skl_to_pmml.py in get_ensemble_models(model, derived_col_names, col_names, target_name, mining_imp_val, categoric_values, model_name)
   1033     model_kwargs = get_model_kwargs(model, col_names, target_name, mining_imp_val,categoric_values)
   1034     if model.__class__.__name__ == 'GradientBoostingRegressor':
-> 1035         model_kwargs['Targets'] = get_targets(model, target_name)
   1036 
   1037     mining_models = list()

/opt/anaconda3/lib/python3.7/site-packages/nyoka/skl/skl_to_pmml.py in get_targets(model, target_name)
   1067                 pml.Target(
   1068                     field=target_name,
-> 1069                     rescaleConstant="{:.16f}".format(model.init_.mean),
   1070                     rescaleFactor="{:.16f}".format(model.learning_rate)
   1071                 )

AttributeError: 'DummyRegressor' object has no attribute 'mean'

I get the same error when trying to do the same with a different GRB pipeline

[Nirmal-Neel]

Hi @doolingdavid, currently Nyoka has support for scikit-learn with version <= 0.20.4. We are planning to release a new version of Nyoka which will support latest version of scikit-learn.