tf.estimator.experimental.build_raw_supervised_input_receiver_fn

View source on GitHub

Build a supervised_input_receiver_fn for raw features and labels.

tf.estimator.experimental.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.