Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

MiniCPM: Unveiling the Potential of Small Language Models with Scalable Training Strategies

About

The burgeoning interest in developing Large Language Models (LLMs) with up to trillion parameters has been met with concerns regarding resource efficiency and practical expense, particularly given the immense cost of experimentation. This scenario underscores the importance of exploring the potential of Small Language Models (SLMs) as a resource-efficient alternative. In this context, we introduce MiniCPM, specifically the 1.2B and 2.4B non-embedding parameter variants, not only excel in their respective categories but also demonstrate capabilities on par with 7B-13B LLMs. While focusing on SLMs, our approach exhibits scalability in both model and data dimensions for future LLM research. Regarding model scaling, we employ extensive model wind tunnel experiments for stable and optimal scaling. For data scaling, we introduce a Warmup-Stable-Decay (WSD) learning rate scheduler (LRS), conducive to continuous training and domain adaptation. We present an in-depth analysis of the intriguing training dynamics that occurred in the WSD LRS. With WSD LRS, we are now able to efficiently study data-model scaling law without extensive retraining experiments on both axes of model and data, from which we derive the much higher compute optimal data-model ratio than Chinchilla Optimal. Additionally, we introduce MiniCPM family, including MiniCPM-DPO, MiniCPM-MoE and MiniCPM-128K, whose excellent performance further cementing MiniCPM's foundation in diverse SLM applications. MiniCPM models are available publicly at https://github.com/OpenBMB/MiniCPM .

Shengding Hu, Yuge Tu, Xu Han, Chaoqun He, Ganqu Cui, Xiang Long, Zhi Zheng, Yewei Fang, Yuxiang Huang, Weilin Zhao, Xinrong Zhang, Zheng Leng Thai, Kaihuo Zhang, Chongyi Wang, Yuan Yao, Chenyang Zhao, Jie Zhou, Jie Cai, Zhongwu Zhai, Ning Ding, Chao Jia, Guoyang Zeng, Dahai Li, Zhiyuan Liu, Maosong Sun• 2024

Related benchmarks

TaskDatasetResultRank
Multimodal EvaluationMME
Score1.81e+3
658
Multimodal UnderstandingMMBench--
637
Multimodal UnderstandingMM-Vet
MM-Vet Score41
531
Science Question AnsweringScienceQA--
502
Multimodal ReasoningMM-Vet
MM-Vet Score41
431
Video UnderstandingMVBench
Accuracy67.44
425
Multitask Language UnderstandingMMLU
Accuracy49.6
413
Mathematical ReasoningMathVista
Score38.7
385
Diagram UnderstandingAI2D
Accuracy62.9
247
Diagram Question AnsweringAI2D
AI2D Accuracy63.8
232
Showing 10 of 63 rows

Other info

Follow for update