Leveraging Pre-trained Checkpoints for Sequence Generation Tasks
This work addresses the gap in using pre-trained models for generation tasks, benefiting NLP practitioners by improving performance and saving compute time.
The paper tackled the problem of applying pre-trained checkpoints to sequence generation tasks, demonstrating that initializing Transformer-based models with BERT, GPT-2, and RoBERTa checkpoints achieves new state-of-the-art results on tasks like Machine Translation and Text Summarization.
Unsupervised pre-training of large neural models has recently revolutionized Natural Language Processing. By warm-starting from the publicly released checkpoints, NLP practitioners have pushed the state-of-the-art on multiple benchmarks while saving significant amounts of compute time. So far the focus has been mainly on the Natural Language Understanding tasks. In this paper, we demonstrate the efficacy of pre-trained checkpoints for Sequence Generation. We developed a Transformer-based sequence-to-sequence model that is compatible with publicly available pre-trained BERT, GPT-2 and RoBERTa checkpoints and conducted an extensive empirical study on the utility of initializing our model, both encoder and decoder, with these checkpoints. Our models result in new state-of-the-art results on Machine Translation, Text Summarization, Sentence Splitting, and Sentence Fusion.