Investigating the Characteristics of a Transformer in a Few-Shot Setup: Does Freezing Layers in RoBERTa Help?

Digvijay Ingle, Rishabh Tripathi, Ayush Kumar, Kevin Patel, Jithendra Vepa


Abstract
Transformer based language models have been widely adopted by industrial and research organisations in developing machine learning applications in the presence of limited annotated data. While these models show remarkable results, their functioning in few-shot settings is still poorly understood. Hence, we perform an investigative study to understand the characteristics of such models fine-tuned in few-shot setups. Specifically, we compare the intermediate layer representations obtained from a few-shot model and a pre-trained language model. We observe that pre-trained and few-shot models show similar representations over initial layers, whereas the later layers show a stark deviation. Based on these observations, we propose to freeze the initial Transformer layers to fine-tune the model in a constrained text classification setup with K annotated data points per class, where K ranges from 8 to 64. In our experiments across six benchmark sentence classification tasks, we discover that freezing initial 50% Transformer layers not only reduces training time but also surprisingly improves Macro F1 (upto 8%) when compared to fully trainable layers in few-shot setup. We also observe that this idea of layer freezing can very well be generalized to state-of-the-art few-shot text classification techniques, like DNNC and LM-BFF, leading to significant reduction in training time while maintaining comparable performance.
Anthology ID:
2022.blackboxnlp-1.19
Volume:
Proceedings of the Fifth BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP
Month:
December
Year:
2022
Address:
Abu Dhabi, United Arab Emirates (Hybrid)
Editors:
Jasmijn Bastings, Yonatan Belinkov, Yanai Elazar, Dieuwke Hupkes, Naomi Saphra, Sarah Wiegreffe
Venue:
BlackboxNLP
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
238–248
Language:
URL:
https://aclanthology.org/2022.blackboxnlp-1.19
DOI:
10.18653/v1/2022.blackboxnlp-1.19
Bibkey:
Cite (ACL):
Digvijay Ingle, Rishabh Tripathi, Ayush Kumar, Kevin Patel, and Jithendra Vepa. 2022. Investigating the Characteristics of a Transformer in a Few-Shot Setup: Does Freezing Layers in RoBERTa Help?. In Proceedings of the Fifth BlackboxNLP Workshop on Analyzing and Interpreting Neural Networks for NLP, pages 238–248, Abu Dhabi, United Arab Emirates (Hybrid). Association for Computational Linguistics.
Cite (Informal):
Investigating the Characteristics of a Transformer in a Few-Shot Setup: Does Freezing Layers in RoBERTa Help? (Ingle et al., BlackboxNLP 2022)
Copy Citation:
PDF:
https://aclanthology.org/2022.blackboxnlp-1.19.pdf
Video:
 https://aclanthology.org/2022.blackboxnlp-1.19.mp4