Plots not showing sparse split atrribute

[marquisthunder]

Dear ydf developers, Hope this get you well. I have a question about the plot_otree method not showing split condition when using oblique tree model (SPARSE_OBLIQUE)

learner = ydf.GradientBoostedTreesLearner(
   ...:   label='label',
   ...:   split_axis="SPARSE_OBLIQUE",
   ...:   sparse_oblique_normalization="MIN_MAX",
   ...:   sparse_oblique_num_projections_exponent=1.0,
   ...:   ).train(pd.concat([x_train, y_train], axis=1))
 model.get_tree(0)

Out[4]: Tree(root=NonLeaf(value=None, condition=NumericalHigherThanCondition(missing=False, score=0.0, attribute=3, threshold=95.69999694824219), pos_child=Leaf(value=RegressionValue(num_examples=0.0, value=-0.4674617052078247, standard_deviation=None)), neg_child=Leaf(value=RegressionValue(num_examples=0.0, value=0.3032335340976715, standard_deviation=None))))

The split condition is always sparse split instead of real attribute name.

[rstz]

Hi, thank you for reporting this. The name of the attributes and their weights are shown as tooltips if you hover with the mouse over the three dots. Since there are (often) multiple attributes involved, the condition can be quite large and we did not want to clutter the plot too much. Does this work for you?

[marquisthunder]

thanks, it works well.