AdaptiveLogSoftmaxWithLoss¶

class
torch.nn.
AdaptiveLogSoftmaxWithLoss
(in_features, n_classes, cutoffs, div_value=4.0, head_bias=False, device=None, dtype=None)[source]¶ Efficient softmax approximation as described in Efficient softmax approximation for GPUs by Edouard Grave, Armand Joulin, Moustapha Cissé, David Grangier, and Hervé Jégou.
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 settingcutoffs = [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 $\left\lfloor\frac{\texttt{in\_features}}{\texttt{div\_value}^{idx}}\right\rfloor$, 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.
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.
Note
This module returns a
NamedTuple
withoutput
andloss
fields. See further documentation for details.Note
To compute logprobabilities for all classes, the
log_prob
method can be used. Parameters
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
output is a Tensor of size
N
containing computed target log probabilities for each exampleloss is a Scalar representing the computed negative log likelihood loss
 Return type
NamedTuple
withoutput
andloss
fields
 Shape:
input: $(N, \texttt{in\_features})$ or $(\texttt{in\_features})$
target: $(N)$ or $()$ where each value satisfies $0 <= \texttt{target[i]} <= \texttt{n\_classes}$
output1: $(N)$ or $()$
output2:
Scalar

log_prob
(input)[source]¶ Computes log probabilities for all $\texttt{n\_classes}$
 Parameters
input (Tensor) – a minibatch of examples
 Returns
logprobabilities of for each class $c$ in range $0 <= c <= \texttt{n\_classes}$, where $\texttt{n\_classes}$ is a parameter passed to
AdaptiveLogSoftmaxWithLoss
constructor.
 Shape:
Input: $(N, \texttt{in\_features})$
Output: $(N, \texttt{n\_classes})$

predict
(input)[source]¶ This is equivalent to self.log_prob(input).argmax(dim=1), but is more efficient in some cases.
 Parameters
input (Tensor) – a minibatch of examples
 Returns
a class with the highest probability for each example
 Return type
output (Tensor)
 Shape:
Input: $(N, \texttt{in\_features})$
Output: $(N)$