Efficient Training of Language Models using Few-Shot Learning

Sashank J. Reddi, Sobhan Miryoosefi, Stefani Karp, Shankar Krishnan, Satyen Kale, Seungyeon Kim, Sanjiv Kumar
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:14553-14568, 2023.

Abstract

Large deep learning models have achieved state-of-the-art performance across various natural language processing (NLP) tasks and demonstrated remarkable few-shot learning performance. However, training them is often challenging and resource-intensive. In this paper, we study an efficient approach to train language models using few-shot learners. We show that, by leveraging the fast learning nature of few-shot learners, one can train language models efficiently in a stagewise manner. Our main insight is that stacking a good few-shot learner on a good small language model provides a good initializer for a larger language model. Using this insight and building upon progressive stacking approaches, we develop novel approaches for training such networks in a stagewise manner. Furthermore, we also provide a theoretical framework and accompanying empirical studies to support our insights, thereby creating a theoretical foundation for progressive stacking. Finally, we provide empirical results to demonstrate the effectiveness of our approach in reducing the training time of few-shot learners.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-j-reddi23a, title = {Efficient Training of Language Models using Few-Shot Learning}, author = {J. Reddi, Sashank and Miryoosefi, Sobhan and Karp, Stefani and Krishnan, Shankar and Kale, Satyen and Kim, Seungyeon and Kumar, Sanjiv}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {14553--14568}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/j-reddi23a/j-reddi23a.pdf}, url = {https://proceedings.mlr.press/v202/j-reddi23a.html}, abstract = {Large deep learning models have achieved state-of-the-art performance across various natural language processing (NLP) tasks and demonstrated remarkable few-shot learning performance. However, training them is often challenging and resource-intensive. In this paper, we study an efficient approach to train language models using few-shot learners. We show that, by leveraging the fast learning nature of few-shot learners, one can train language models efficiently in a stagewise manner. Our main insight is that stacking a good few-shot learner on a good small language model provides a good initializer for a larger language model. Using this insight and building upon progressive stacking approaches, we develop novel approaches for training such networks in a stagewise manner. Furthermore, we also provide a theoretical framework and accompanying empirical studies to support our insights, thereby creating a theoretical foundation for progressive stacking. Finally, we provide empirical results to demonstrate the effectiveness of our approach in reducing the training time of few-shot learners.} }
Endnote
%0 Conference Paper %T Efficient Training of Language Models using Few-Shot Learning %A Sashank J. Reddi %A Sobhan Miryoosefi %A Stefani Karp %A Shankar Krishnan %A Satyen Kale %A Seungyeon Kim %A Sanjiv Kumar %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-j-reddi23a %I PMLR %P 14553--14568 %U https://proceedings.mlr.press/v202/j-reddi23a.html %V 202 %X Large deep learning models have achieved state-of-the-art performance across various natural language processing (NLP) tasks and demonstrated remarkable few-shot learning performance. However, training them is often challenging and resource-intensive. In this paper, we study an efficient approach to train language models using few-shot learners. We show that, by leveraging the fast learning nature of few-shot learners, one can train language models efficiently in a stagewise manner. Our main insight is that stacking a good few-shot learner on a good small language model provides a good initializer for a larger language model. Using this insight and building upon progressive stacking approaches, we develop novel approaches for training such networks in a stagewise manner. Furthermore, we also provide a theoretical framework and accompanying empirical studies to support our insights, thereby creating a theoretical foundation for progressive stacking. Finally, we provide empirical results to demonstrate the effectiveness of our approach in reducing the training time of few-shot learners.
APA
J. Reddi, S., Miryoosefi, S., Karp, S., Krishnan, S., Kale, S., Kim, S. & Kumar, S.. (2023). Efficient Training of Language Models using Few-Shot Learning. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:14553-14568 Available from https://proceedings.mlr.press/v202/j-reddi23a.html.

Related Material