Missed TensorFlow World? Check out the recap. Learn more

tf.contrib.estimator.build_raw_supervised_input_receiver_fn

View source on GitHub

Build a supervised_input_receiver_fn for raw features and labels.

Aliases:

tf.contrib.estimator.build_raw_supervised_input_receiver_fn(
    features,
    labels,
    default_batch_size=None
)

This function wraps tensor placeholders in a supervised_receiver_fn with the expectation that the features and labels appear precisely as the model_fn expects them. Features and labels can therefore be dicts of tensors, or raw tensors.

Args:

  • features: a dict of string to Tensor or Tensor.
  • labels: a dict of string to Tensor or Tensor.
  • default_batch_size: the number of query examples expected per batch. Leave unset for variable batch size (recommended).

Returns:

A supervised_input_receiver_fn.

Raises:

  • ValueError: if features and labels have overlapping keys.