NTK-approximating MLP Fusion for Efficient Language Model Fine-tuning

Tianxin Wei, Zeming Guo, Yifan Chen, Jingrui He
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:36821-36838, 2023.

Abstract

Fine-tuning a pre-trained language model (PLM) emerges as the predominant strategy in many natural language processing applications. However, even fine-tuning the PLMs and doing inference are expensive, especially on edge devices with low computing power. Some general approaches (e.g. quantization and distillation) have been widely studied to reduce the compute/memory of PLM fine-tuning, while very few one-shot compression techniques are explored. In this paper, we investigate the neural tangent kernel (NTK)–which reveals the gradient descent dynamics of neural networks–of the multilayer perceptrons (MLP) modules in a PLM and propose to coin a lightweight PLM through NTK-approximating MLP fusion. To achieve this, we reconsider the MLP as a bundle of sub-MLPs, and cluster them into a given number of centroids, which can then be restored as a compressed MLP and surprisingly shown to well approximate the NTK of the original PLM. Extensive experiments of PLM fine-tuning on both natural language understanding (NLU) and generation (NLG) tasks are provided to verify the effectiveness of the proposed method MLP fusion. Our code is available at https://github.com/weitianxin/MLP_Fusion.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-wei23b, title = {{NTK}-approximating {MLP} Fusion for Efficient Language Model Fine-tuning}, author = {Wei, Tianxin and Guo, Zeming and Chen, Yifan and He, Jingrui}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {36821--36838}, 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/wei23b/wei23b.pdf}, url = {https://proceedings.mlr.press/v202/wei23b.html}, abstract = {Fine-tuning a pre-trained language model (PLM) emerges as the predominant strategy in many natural language processing applications. However, even fine-tuning the PLMs and doing inference are expensive, especially on edge devices with low computing power. Some general approaches (e.g. quantization and distillation) have been widely studied to reduce the compute/memory of PLM fine-tuning, while very few one-shot compression techniques are explored. In this paper, we investigate the neural tangent kernel (NTK)–which reveals the gradient descent dynamics of neural networks–of the multilayer perceptrons (MLP) modules in a PLM and propose to coin a lightweight PLM through NTK-approximating MLP fusion. To achieve this, we reconsider the MLP as a bundle of sub-MLPs, and cluster them into a given number of centroids, which can then be restored as a compressed MLP and surprisingly shown to well approximate the NTK of the original PLM. Extensive experiments of PLM fine-tuning on both natural language understanding (NLU) and generation (NLG) tasks are provided to verify the effectiveness of the proposed method MLP fusion. Our code is available at https://github.com/weitianxin/MLP_Fusion.} }
Endnote
%0 Conference Paper %T NTK-approximating MLP Fusion for Efficient Language Model Fine-tuning %A Tianxin Wei %A Zeming Guo %A Yifan Chen %A Jingrui He %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-wei23b %I PMLR %P 36821--36838 %U https://proceedings.mlr.press/v202/wei23b.html %V 202 %X Fine-tuning a pre-trained language model (PLM) emerges as the predominant strategy in many natural language processing applications. However, even fine-tuning the PLMs and doing inference are expensive, especially on edge devices with low computing power. Some general approaches (e.g. quantization and distillation) have been widely studied to reduce the compute/memory of PLM fine-tuning, while very few one-shot compression techniques are explored. In this paper, we investigate the neural tangent kernel (NTK)–which reveals the gradient descent dynamics of neural networks–of the multilayer perceptrons (MLP) modules in a PLM and propose to coin a lightweight PLM through NTK-approximating MLP fusion. To achieve this, we reconsider the MLP as a bundle of sub-MLPs, and cluster them into a given number of centroids, which can then be restored as a compressed MLP and surprisingly shown to well approximate the NTK of the original PLM. Extensive experiments of PLM fine-tuning on both natural language understanding (NLU) and generation (NLG) tasks are provided to verify the effectiveness of the proposed method MLP fusion. Our code is available at https://github.com/weitianxin/MLP_Fusion.
APA
Wei, T., Guo, Z., Chen, Y. & He, J.. (2023). NTK-approximating MLP Fusion for Efficient Language Model Fine-tuning. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:36821-36838 Available from https://proceedings.mlr.press/v202/wei23b.html.

Related Material