tf.contrib.learn.multi_head

tf.contrib.learn.multi_head(
    heads,
    loss_weights=None
)

Defined in tensorflow/contrib/learn/python/learn/estimators/head.py.

Creates a MultiHead stemming from same logits/hidden layer. (deprecated)

THIS FUNCTION IS DEPRECATED. It will be removed in a future version. Instructions for updating: Please switch to tf.contrib.estimator.*_head.

Args:

  • heads: list of Head objects.
  • loss_weights: optional list of weights to be used to merge losses from each head. All losses are weighted equally if not provided.

Returns:

A instance of Head that merges multiple heads.

Raises:

  • ValueError: if heads and loss_weights have different size.