Crayon: Customized On-Device LLM via Instant Adapter Blending and Edge-Server Hybrid Inference

Jihwan Bang, Juntae Lee, Kyuhong Shim, Seunghan Yang, Simyung Chang


Abstract
The customization of large language models (LLMs) for user-specified tasks gets important. However, maintaining all the customized LLMs on cloud servers incurs substantial memory and computational overheads, and uploading user data can also lead to privacy concerns. On-device LLMs can offer a promising solution by mitigating these issues. Yet, the performance of on-device LLMs is inherently constrained by the limitations of small-scaled models. To overcome these restrictions, we first propose Crayon, a novel approach for on-device LLM customization. Crayon begins by constructing a pool of diverse base adapters, and then we instantly blend them into a customized adapter without extra training. In addition, we develop a device-server hybrid inference strategy, which deftly allocates more demanding queries or non-customized tasks to a larger, more capable LLM on a server. This ensures optimal performance without sacrificing the benefits of on-device customization. We carefully craft a novel benchmark from multiple question-answer datasets, and show the efficacy of our method in the LLM customization.
Anthology ID:
2024.acl-long.204
Volume:
Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers)
Month:
August
Year:
2024
Address:
Bangkok, Thailand
Editors:
Lun-Wei Ku, Andre Martins, Vivek Srikumar
Venue:
ACL
SIG:
Publisher:
Association for Computational Linguistics
Note:
Pages:
3720–3731
Language:
URL:
https://aclanthology.org/2024.acl-long.204
DOI:
10.18653/v1/2024.acl-long.204
Bibkey:
Cite (ACL):
Jihwan Bang, Juntae Lee, Kyuhong Shim, Seunghan Yang, and Simyung Chang. 2024. Crayon: Customized On-Device LLM via Instant Adapter Blending and Edge-Server Hybrid Inference. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers), pages 3720–3731, Bangkok, Thailand. Association for Computational Linguistics.
Cite (Informal):
Crayon: Customized On-Device LLM via Instant Adapter Blending and Edge-Server Hybrid Inference (Bang et al., ACL 2024)
Copy Citation:
PDF:
https://preview.aclanthology.org/nschneid-patch-5/2024.acl-long.204.pdf