bert2BERT: Towards Reusable Pretrained Language Models

By Cheng Chen, Yichun Yin, Lifeng Shang, Xin Jiang, Yujia Qin, Fengyu Wang, Zhi Wang, Xiao Chen, Zhiyuan Liu, Qun Liu. 2021

87 citations recorded by the source corpus.

In recent years, researchers tend to pre-train ever-larger language models to explore the upper limit of deep models. However, large language model pre-training costs intensive computational resources, and most of the models are trained from scratch without reusing the existing pre-trained models, which is wasteful. In this paper, we propose bert2BERT, which can effectively transfer the knowledge of an existing smaller pre-trained model to a large model through parameter initialization and significantly improve the pre-training efficiency of the large model. Specifically, we extend the previous function-preserving method proposed in computer vision on the Transformer-based language model, and further improve it by proposing a novel method, advanced knowledge for large model’s initialization. In addition, a two-stage learning method is proposed to further accelerate the pre-training. We conduct extensive experiments on representative PLMs (e.g., BERT and GPT) and demonstrate that (1) our method can save a significant amount of training cost compared with baselines including learning from scratch, StackBERT and MSLT; (2) our method is generic and applicable to different types of pre-trained models. In particular, bert2BERT saves about 45% and 47% computational cost of pre-training BERT_{\rm BASE} and GPT_{\rm BASE} by reusing the models of almost their half sizes.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Zhiyuan Liu
2. Maosong Sun
3. Yankai Lin
4. Peng Li
5. Jie Zhou
6. Yujia Qin
7. Jiajie Zhang
8. Minlie Huang

In Computer Vision and Pattern Recognition

1. Zhiyuan Liu
2. Maosong Sun
3. Peng Li
4. Yankai Lin
5. Jie Zhou
6. Jiajie Zhang
7. Yujia Qin
8. Hannaneh Hajishirzi

Citations in the corpus

• ELLE: Efficient Lifelong Pre-training for Emerging Data (2022, methodology): “Different from Chen et al. (2021), we additionally introduce random noises δ i into the newly copied parameters of W (cid:48) during initialization.”
• Knowledge Inheritance for Pre-trained Language Models (2021, unknown): “Another line of methods (Gong et al., 2019; Gu et al., 2021; Chen et al., 2022; Qin et al., 2022) proposes to pre-train larger PLMs progressively.”