nn_adaptive_log_softmax_with_loss function

AdaptiveLogSoftmaxWithLoss module

Efficient softmax approximation as described in Efficient softmax approximation for GPUs by Edouard Grave, Armand Joulin, Moustapha Cissé, David Grangier, and Hervé Jégou

nn_adaptive_log_softmax_with_loss(
  in_features,
  n_classes,
  cutoffs,
  div_value = 4,
  head_bias = FALSE
)

Arguments

• in_features: (int): Number of features in the input tensor
• n_classes: (int): Number of classes in the dataset
• cutoffs: (Sequence): Cutoffs used to assign targets to their buckets
• div_value: (float, optional): value used as an exponent to compute sizes of the clusters. Default: 4.0
• head_bias: (bool, optional): If True, adds a bias term to the 'head' of the adaptive softmax. Default: False

Returns

NamedTuple with output and loss fields:

• output is a Tensor of size N containing computed target log probabilities for each example
• loss is a Scalar representing the computed negative log likelihood loss

Details

Adaptive softmax is an approximate strategy for training models with large output spaces. It is most effective when the label distribution is highly imbalanced, for example in natural language modelling, where the word frequency distribution approximately follows the Zipf's law.

Adaptive softmax partitions the labels into several clusters, according to their frequency. These clusters may contain different number of targets each.

Additionally, clusters containing less frequent labels assign lower dimensional embeddings to those labels, which speeds up the computation. For each minibatch, only clusters for which at least one target is present are evaluated.

The idea is that the clusters which are accessed frequently (like the first one, containing most frequent labels), should also be cheap to compute -- that is, contain a small number of assigned labels. We highly recommend taking a look at the original paper for more details.

• cutoffs should be an ordered Sequence of integers sorted in the increasing order. It controls number of clusters and the partitioning of targets into clusters. For example setting cutoffs = c(10, 100, 1000)

  means that first 10 targets will be assigned to the 'head' of the adaptive softmax, targets 11, 12, ..., 100 will be assigned to the first cluster, and targets 101, 102, ..., 1000 will be assigned to the second cluster, while targets 1001, 1002, ..., n_classes - 1 will be assigned to the last, third cluster.

• div_value is used to compute the size of each additional cluster, which is given as $\left\lfloor\frac{\mbox{in\_features}}{\mbox{div\_value}^{idx}}\right\rfloor$, where $idx$ is the cluster index (with clusters for less frequent words having larger indices, and indices starting from $1$).

• head_bias if set to True, adds a bias term to the 'head' of the adaptive softmax. See paper for details. Set to False in the official implementation.

Note

This module returns a NamedTuple with output

and loss fields. See further documentation for details.

To compute log-probabilities for all classes, the log_prob

method can be used.

Warning

Labels passed as inputs to this module should be sorted according to their frequency. This means that the most frequent label should be represented by the index 0, and the least frequent label should be represented by the index n_classes - 1.

Shape

• input: $(N, \mbox{in\_features})$
• target: $(N)$ where each value satisfies $0 \<= \mbox{target[i]} \<= \mbox{n\_classes}$
• output1: $(N)$
• output2: Scalar