YuLan-Mini: An Open Data-efficient Language Model

TL;DR

YuLan-Mini is a 2.42B parameter language model trained on 1.08 trillion tokens, achieving top-tier performance with efficient data pipeline, robust optimization, and targeted data selection techniques.

Contribution

The paper introduces novel training strategies and data handling methods that enable high performance of a smaller language model with less data and resources.

Findings

YuLan-Mini achieves performance comparable to larger models.

Efficient training is possible with optimized data pipeline and techniques.

Open release of data composition details for reproducibility.

Abstract

Effective pre-training of large language models (LLMs) has been challenging due to the immense resource demands and the complexity of the technical processes involved. This paper presents a detailed technical report on YuLan-Mini, a highly capable base model with 2.42B parameters that achieves top-tier performance among models of similar parameter scale. Our pre-training approach focuses on enhancing training efficacy through three key technical contributions: an elaborate data pipeline combines data cleaning with data schedule strategies, a robust optimization method to mitigate training instability, and an effective annealing approach that incorporates targeted data selection and long context training. Remarkably, YuLan-Mini, trained on 1.08T tokens, achieves performance comparable to industry-leading models that require significantly more data. To facilitate reproduction, we release the full details of the data composition for each training phase. Project details can be accessed at the following link: https://github.com/RUC-GSAI/YuLan-Mini.