|View source on GitHub|
Parses the input features using the given feature columns.
tfl.estimators.transform_features( features, feature_columns=None )
Used in the notebooks
|Used in the tutorials|
This function can be used to parse input features when constructing a custom estimator. When using this function, you will not need to wrap categorical features with dense feature embeddings, and the resulting tensors will not be concatenated, making it easier to use the features in the calibration layers.
features: A dict from feature names to tensors.
feature_columns: A list of FeatureColumn objects to be used for parsing. If not provided, the input features are assumed to be already parsed.
collections.OrderedDict mapping feature names to parsed tensors.