E-LANG: Energy-Based Joint Inferencing of Super and Swift Language Models
This addresses the problem of computational inefficiency in deploying large language models for NLP practitioners, offering an architecture-agnostic solution that is more flexible than existing methods.
The paper tackles the high computational cost of large language models by proposing E-LANG, a dynamic inference approach that routes inputs between large Super-models and light-weight Swift models based on energy characteristics, achieving speed-ups of up to 3.3x on GLUE and 3.2x less computations with BERT-based SOTA.
Building huge and highly capable language models has been a trend in the past years. Despite their great performance, they incur high computational cost. A common solution is to apply model compression or choose light-weight architectures, which often need a separate fixed-size model for each desirable computational budget, and may lose performance in case of heavy compression. This paper proposes an effective dynamic inference approach, called E-LANG, which distributes the inference between large accurate Super-models and light-weight Swift models. To this end, a decision making module routes the inputs to Super or Swift models based on the energy characteristics of the representations in the latent space. This method is easily adoptable and architecture agnostic. As such, it can be applied to black-box pre-trained models without a need for architectural manipulations, reassembling of modules, or re-training. Unlike existing methods that are only applicable to encoder-only backbones and classification tasks, our method also works for encoder-decoder structures and sequence-to-sequence tasks such as translation. The E-LANG performance is verified through a set of experiments with T5 and BERT backbones on GLUE, SuperGLUE, and WMT. In particular, we outperform T5-11B with an average computations speed-up of 3.3$\times$ on GLUE and 2.9$\times$ on SuperGLUE. We also achieve BERT-based SOTA on GLUE with 3.2$\times$ less computations. Code and demo are available in the supplementary materials.