python - Individual trees in tf.BoostedTreesClassifer?

Question

I am trying to familiarize myself with tensorflow tf.BoostedTreesClassifer.

Is there a way of viewing the individual trees of the model?

Read the 2017 paper (TF Boosted Trees: A scalable TensorFlow based framework for gradient boosting) and gone through the basic tutorial https://www.tensorflow.org/tutorials/estimator/boosted_trees. I would love to be able to examine the actual trees built by the training algorithm. A bit like get_dump in XGBoost.

Is there such a thing? I have exported the SavedModel, loaded it and tried examining the predict signature on tensorboard (using tf.summary) to generate logdir. However, I am having difficulties in connecting the ops in the graph to the trees with weights.

question from:https://stackoverflow.com/questions/65878532/individual-trees-in-tf-boostedtreesclassifer

Answer 1

Waitting for answers