NAACL 2024industry109 citations

SOLAR 10.7B: Scaling Large Language Models with Simple yet Effective Depth Up-Scaling

Sanghoon Kim, Dahyun Kim, Chanjun Park, Wonsung Lee, Wonho Song, Yunsu Kim, Hyeonwoo Kim, Yungi Kim

Abstract

We introduce SOLAR 10.7B, a large language model (LLM) with 10.7 billion parameters, demonstrating superior performance in various natural language processing (NLP) tasks. Inspired by recent efforts to efficiently up-scale LLMs, we present a method for scaling LLMs called depth up-scaling (DUS), which encompasses depthwise scaling and continued pretraining. In contrast to other LLM up-scaling methods that use mixture-of-experts, DUS does not require complex changes to train and inference efficiently. We show experimentally that DUS is simple yet effective in scaling up high-performance LLMs from small ones. Building on the DUS model, we additionally present SOLAR 10.7B-Instruct, a variant fine-tuned for instruction-following capabilities, surpassing Mixtral-8x7B-Instruct. SOLAR 10.7B is publicly available under the Apache 2.0 license, promoting broad access and application in the LLM field.

BibTeX
@inproceedings{kim-etal-2024-solar,
    title = "{SOLAR} 10.7{B}: Scaling Large Language Models with Simple yet Effective Depth Up-Scaling",
    author = "Kim, Sanghoon  and
      Kim, Dahyun  and
      Park, Chanjun  and
      Lee, Wonsung  and
      Song, Wonho  and
      Kim, Yunsu  and
      Kim, Hyeonwoo  and
      Kim, Yungi  and
      Lee, Hyeonju  and
      Kim, Jihoo  and
      Ahn, Changbae  and
      Yang, Seonghoon  and
      Lee, Sukyung  and
      Park, Hyunbyung  and
      Gim, Gyoungjin  and
      Cha, Mikyoung  and
      Lee, Hwalsuk  and
      Kim, Sunghun",
    editor = "Yang, Yi  and
      Davani, Aida  and
      Sil, Avi  and
      Kumar, Anoop",
    booktitle = "Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 6: Industry Track)",
    month = jun,
    year = "2024",
    address = "Mexico City, Mexico",
    publisher = "Association for Computational Linguistics",
    url = "https://aclanthology.org/2024.naacl-industry.3/",
    doi = "10.18653/v1/2024.naacl-industry.3",
    pages = "23--35"
}