RSVP for your your local TensorFlow Everywhere event today!


Uses a Module to construct a dense representation from a text feature.

Used in the notebooks

Used in the tutorials

This feature column can be used on an input feature whose values are strings of arbitrary size.

The result of this feature column is the result of passing its input through the module m instantiated from module_spec, as per result = m(input). The result must have dtype float32 and shape [batch_size, num_features] with a known value of num_features.


  comment = hub.text_embedding_column("comment", "/tmp/text-module")
  feature_columns = [comment, ...]
  features = {
    "comment": np.array(["wow, much amazing", "so easy", ...]),
  labels = np.array([[1], [0], ...])
  # If running TF 2.x, use `tf.compat.v1.estimator.inputs.numpy_input_fn`
  input_fn = tf.estimator.inputs.numpy_input_fn(features, labels,
  estimator = tf.estimator.DNNClassifier(hidden_units, feature_columns)
  estimator.train(input_fn, max_steps=100)

key A string or _FeatureColumn identifying the text feature.
module_spec A ModuleSpec defining the Module to instantiate or a path where to load a ModuleSpec via load_module_spec
trainable Whether or not the Module is trainable. False by default, meaning the pre-trained weights are frozen. This is different from the ordinary tf.feature_column.embedding_column(), but that one is intended for training from scratch.

_DenseColumn that converts from text input.

ValueError if module_spec is not suitable for use in this feature column.