arXivDaily arXiv每日学术速递 周一至周五更新
​通知:联邦节假日调休:哥伦布日 / 原住民日(Columbus Day / Indigenous Peoples' Day),今日(13日)无更新。
arXiv 2306.11644cs.CLcs.AIcs.LG

教科书就是你所需的全部

Textbooks Are All You Need

Microsoft Research(微软研究院)
Suriya Gunasekar, Yi Zhang, Jyoti Aneja, Caio César Teodoro Mendes, Allie Del Giorno, Sivakanth Gopi, Mojan Javaheripi, Piero Kauffmann, Gustavo de Rosa, Olli S… 展开作者

Suriya Gunasekar, Yi Zhang, Jyoti Aneja, Caio César Teodoro Mendes, Allie Del Giorno, Sivakanth Gopi, Mojan Javaheripi, Piero Kauffmann, Gustavo de Rosa, Olli Saarikivi, Adil Salim, Shital Shah, Harkirat Singh Behl, Xin Wang, Sébastien Bubeck, Ronen Eldan, Adam Tauman Kalai, Yin Tat Lee, Yuanzhi Li

AI总结:

本研究推出仅13亿参数的代码大模型phi-1,通过高质量网页数据与GPT-3.5生成的合成教科书/练习题训练,在小体量下实现了HumanEval 50.6%、MBPP 55.5%的pass@1准确率,还展现出显著涌现特性。

摘要:

我们推出phi-1,一款面向代码领域的新型大语言模型,其规模远小于同类竞品模型:phi-1是基于Transformer架构的模型,参数量为13亿,在8张A100显卡上训练4天,训练数据包含从网页筛选的60亿token“教科书级质量”数据,以及用GPT-3.5生成的10亿token合成教科书与练习题。尽管规模很小,phi-1在HumanEval基准上的pass@1准确率达50.6%,在MBPP基准上达55.5%。与经过代码练习题数据集微调前的基础模型phi-1-base,以及采用相同训练流程、参数量3.5亿、在HumanEval上仍能达到45%准确率的更小模型phi-1-small相比,phi-1还展现出令人惊讶的涌现特性。

英文摘要:

We introduce phi-1, a new large language model for code, with significantly smaller size than competing models: phi-1 is a Transformer-based model with 1.3B parameters, trained for 4 days on 8 A100s, using a selection of ``textbook quality" data from the web (6B tokens) and synthetically generated textbooks and exercises with GPT-3.5 (1B tokens). Despite this small scale, phi-1 attains pass@1 accuracy 50.6% on HumanEval and 55.5% on MBPP. It also displays surprising emergent properties compared to phi-1-base, our model before our finetuning stage on a dataset of coding exercises, and phi-1-small, a smaller model with 350M parameters trained with the same pipeline as phi-1 that still achieves 45% on HumanEval.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑