tf.contrib.nn.sampled_sparse_softmax_loss

Computes and returns the sampled sparse softmax training loss.

tf.contrib.nn.sampled_sparse_softmax_loss(
    weights, biases, labels, inputs, num_sampled, num_classes, sampled_values=None,
    remove_accidental_hits=True, partition_strategy='mod',
    name='sampled_sparse_softmax_loss'
)

This is a faster way to train a softmax classifier over a huge number of classes.

This operation is for training only. It is generally an underestimate of the full softmax loss.

A common use case is to use this method for training, and calculate the full softmax loss for evaluation or inference. In this case, you must set partition_strategy="div" for the two losses to be consistent, as in the following example:

if mode == "train":
  loss = tf.nn.sampled_sparse_softmax_loss(
      weights=weights,
      biases=biases,
      labels=labels,
      inputs=inputs,
      ...,
      partition_strategy="div")
elif mode == "eval":
  logits = tf.matmul(inputs, tf.transpose(weights))
  logits = tf.nn.bias_add(logits, biases)
  loss = tf.nn.sparse_softmax_cross_entropy_with_logits(
      labels=tf.squeeze(labels),
      logits=logits)

See our Candidate Sampling Algorithms Reference

Also see Section 3 of Jean et al., 2014 (pdf) for the math.