DistilBERT, a distilled version of BERT: smaller, faster, cheaper and lighter

TL;DR

DistilBERT is a smaller, faster, and cheaper version of BERT created through knowledge distillation during pre-training, achieving near-original performance with reduced size and improved efficiency for NLP tasks.

Contribution

The paper introduces a novel pre-training distillation method to produce a compact BERT model, DistilBERT, that retains most of BERT's capabilities while being significantly more efficient.

Findings

DistilBERT reduces model size by 40%.

It retains 97% of BERT's language understanding.

It is 60% faster in inference.

Abstract

As Transfer Learning from large-scale pre-trained models becomes more prevalent in Natural Language Processing (NLP), operating these large models in on-the-edge and/or under constrained computational training or inference budgets remains challenging. In this work, we propose a method to pre-train a smaller general-purpose language representation model, called DistilBERT, which can then be fine-tuned with good performances on a wide range of tasks like its larger counterparts. While most prior work investigated the use of distillation for building task-specific models, we leverage knowledge distillation during the pre-training phase and show that it is possible to reduce the size of a BERT model by 40%, while retaining 97% of its language understanding capabilities and being 60% faster. To leverage the inductive biases learned by larger models during pre-training, we introduce a triple loss combining language modeling, distillation and cosine-distance losses. Our smaller, faster and lighter model is cheaper to pre-train and we demonstrate its capabilities for on-device computations in a proof-of-concept experiment and a comparative on-device study.