Depth-Adaptive Transformer
This addresses the problem of computational waste in large-scale sequence tasks for AI practitioners, offering a more efficient model design.
The paper tackles the inefficiency of fixed-computation sequence-to-sequence models by introducing a depth-adaptive Transformer that adjusts computation and model capacity per sequence, achieving comparable accuracy to a baseline Transformer on IWSLT German-English translation while using less than a quarter of the decoder layers.
State of the art sequence-to-sequence models for large scale tasks perform a fixed number of computations for each input sequence regardless of whether it is easy or hard to process. In this paper, we train Transformer models which can make output predictions at different stages of the network and we investigate different ways to predict how much computation is required for a particular sequence. Unlike dynamic computation in Universal Transformers, which applies the same set of layers iteratively, we apply different layers at every step to adjust both the amount of computation as well as the model capacity. On IWSLT German-English translation our approach matches the accuracy of a well tuned baseline Transformer while using less than a quarter of the decoder layers.