新闻
Scaling Laws for Mixture Pretraining Under Data Constraints
Apple Machine Learning Research · 发布于 · 阅读约3分钟
30秒读懂
- 发生了什么
- Apple researchers studied how to mix scarce target data with abundant generic data during language model pretraining, finding optimal repetition rates.
- 为何重要
- Engineers training models on low-resource languages or specialized domains with limited data need guidance on mixture ratios and repetition.
- 注意
- Results span 2000 runs but optimal repetition rates vary by target data size, compute budget, and model scale, requiring case-by-case analysis.
收听本摘要
- language model
这条新闻背后的模式
每个模式都讲清楚技术如何运作、何时值得投入，以及在哪里会失效。
The Agent Architect
每周一个模式、一个权衡、一个生产事故案例。为构建智能体系统的人准备的每周简报。
每周一封邮件,一键退订。您的地址仅用于发送简报。