Transformer-based NLP models are trained using hundreds of millions or even billions of parameters, limiting their applicability in computationally constrained environments. While the number of parameters generally correlates with performance, it is not clear whether the entire network is required for a downstream task. Motivated by the recent work on pruning and distilling pre-trained models, we explore strategies to drop layers in pre-trained models, and observe the effect of pruning on downstream GLUE tasks. We were able to prune BERT, RoBERTa and XLNet models up to 40%, while maintaining up to 98% of their original performance. Additionally we show that our pruned models are on par with those built using knowledge distillation, both in terms of size and performance. Our experiments yield interesting observations such as: (i) the lower layers are most critical to maintain downstream task performance, (ii) some tasks such as paraphrase detection and sentence similarity are more robust to the dropping of layers, and (iii) models trained using different objective function exhibit different learning patterns and w.r.t the layer dropping. © 2022 Elsevier Ltd
- Efficient transfer learning,
- Interpretation and analysis,
- Pre-trained transformer models,
- Down-stream,
- Efficient transfer learning,
- Interpretation and analyse,
- Objective functions,
- Performance,
- Pre-trained transformer model,
- Sentence similarity,
- Task performance,
- Transfer learning,
- Transformer modeling
https://doi.org/10.1016/j.csl.2022.101429
IR Deposit conditions:
OA version (pathway b) Accepted version
24 months embargo
License: CC BY NC-ND
Must link to publisher version with DOI