A collection of state-of-the-art algorithms for the training, serving and interpretation of Decision Forest models in Keras.
APACHE-2.0 License
Bot releases are visible (Hide)
Published by achoum over 2 years ago
contrib
module for contributed, non-core functionality.contrib.scikit_learn_model_converter
, which facilitates convertinghessian_split_score_subtract_parent
to subtract thetree.pretty()
or str(tree)
.Published by achoum over 2 years ago
pd_dataframe_to_tf_dataset
on Pandas dataframe not containing arrays.Published by achoum almost 3 years ago
validation_interval_in_trees
,keep_non_leaf_label_distribution
and 'random_seed' hyper-parameters.batch_size
argument in the pd_dataframe_to_tf_dataset
utility.num_threads=None
.try_resume_training
to facilitate resumingcheck_dataset=False
.fit
method now passes the validation_data
to the Yggdrasilverbose
constructor argument andfit
parameter of the model.num_candidate_attributes
is not ignored anymore whennum_candidate_attributes_ratio=-1
.max_num_scanned_rows_to_accumulate_statistics
parameter tofit_on_dataset_path
.Published by achoum almost 3 years ago
Published by achoum almost 3 years ago
predict_single_probability_for_binary_classification
compute_permutation_variable_importance
parameter.max_vocab_count
argument to the model constructor. The existingmax_vocab_count
argument in FeatureUsage
objects take precedence.max_vocab_count
in a FeatureUsage
with typeCATEGORICAL_SET
.Published by achoum about 3 years ago
validation_ratio=0
).undefined symbol
error if you install this version with a TensorFlow version different than 2.6. Previous versions were compiled for TF 2.5.Published by arvnds about 3 years ago
Major change : Tensorflow 2.6 compatibility
This release is currently being tested and will be updated to be the latest version in PyPI soon, in the meantime users who need the fixes below can install this version directly from the wheels below, i.e. pip install tensorflow_decision_forests-0.1.9-cp39-cp39-manylinux_2_12_x86_64.manylinux2010_x86_64.whl
for python 3.9.
Fix failure from Github Issue #45 where the wrong field was accessed for leaf node distributions.
Fix incorrect handling of CART pruning when validation set is empty. Previously, the whole tree would be erroneously pruned. Now, pruning is disabled if the validation set is not specified.
Fix saving of categorical features specification in the Builder.
Migration to Tensorflow 2.6. You will see an undefined symbol error if you install this version with a TensorFlow version different than 2.6. Previous versions were compiled for TF 2.5.
Published by achoum about 3 years ago
name
argument in the model classes constructors.postprocessing
model constructor argument to easy applyextract_all_trees
method in the model inspector to efficiently exactnum_threads
constructor argument to control the number of trainingimport_dataspec
constructor argument to the model builder toassets
directory is entirely removed before the export (instead of thePublished by achoum over 3 years ago
pd_dataframe_to_tf_dataset
can automatically rename feature to make themmodel.save(...)
can override an existing model.Published by achoum over 3 years ago
sorting_strategy
to disable the computation of thePublished by achoum over 3 years ago
pd_dataframe_to_tf_dataset
's task.Published by achoum over 3 years ago
model.fit
support training callbacks and a validation dataset.Published by achoum over 3 years ago
Published by achoum over 3 years ago
Initial Release of TensorFlow Decision Forests.