Awesome

👋 Welcome, Owl is a large language model designed for AIOps field to handle IT operations related tasks (fault diagnosis, log analysis, etc). 😊 Owl is an open source project and we welcome everyone to participate and contribute 🌟.

Abstract

With the rapid advancement of IT operations, managing and analyzing large data volumes efficiently for practical applications has become increasingly critical. Natural Language Processing (NLP) techniques have demonstrated remarkable capabilities in various tasks, including named entity recognition, machine translation, and dialogue systems. Recently, Large Language Models (LLMs) have achieved significant improvements across various domain-specific areas. However, there is a noticeable gap in the development of specialized Large Language Models (LLMs) tailored for IT operations. In this paper, we introduce the OWL, a large language model trained on our constructed OWL-Instruct with a wide range of IT-related information. Specifically, limited by the maximum input length, we propose the Homogeneous Markov Context Extension method (HMCE). The mixture-of-adapter strategy is leveraged to improve the parameter-efficient tuning across different domains or tasks. Further, we evaluate the performance of OWL on the OWL-Bench established by us and open IT-related benchmarks. OWL demonstrates superior performance results on IT tasks, which outperforms existing models by significant margins. Moreover, we hope that the findings of our work will provide more insights to revolutionize the techniques of IT operations with specialized LLMs.

Updated News

Our paper has been accepted by ICLR 2024!👏👏

Owl-Instruction

Four phases of constructing OWL-Instruct and how we train our OWL: (1) Data Generation; (2) GPT4 Screening; (3) Mannual Verification; and (4) Supervised Fine-tuning. We provide the bilingual instruction data (ops001).

Owl-Bench

Owl-Bench consist of two parts: Multiple_Choice && Question_Answer.

Usage

For multiple choice test, please refer to the MC_readme. To use question answer test, please refer to the QA_readme. Here the example validation data is provided, we will updata it soon.

Citation

Feel free to cite us if you like Owl.

@inproceedings{
guo2024owl,
title={{OWL}: A Large Language Model for {IT} Operations},
author={Hongcheng Guo and Jian Yang and Jiaheng Liu and Liqun Yang and Linzheng Chai and Jiaqi Bai and Junran Peng and Xiaorong Hu and Chao Chen and Dongfeng Zhang and xu Shi and Tieqiao Zheng and liangfan zheng and Bo Zhang and Ke Xu and Zhoujun Li},
booktitle={The Twelfth International Conference on Learning Representations},
year={2024},
url={https://openreview.net/forum?id=SZOQ9RKYJu}
}

Contact

We are happy to see 'Owl' being applied in the real world and willing to contribute to the community. Feel free to contact us if you have any question or cooperation! Email: hongchengguo@buaa.edu.cn