CodeNet
收藏资源简介:
CodeNet是由IBM研究院创建的一个大规模数据集,包含超过1400万个代码样本和约5亿行代码,涵盖55种不同的编程语言。数据集主要从AIZU和AtCoder两个在线评判系统中收集,用于支持AI在编程领域的学习和研究。CodeNet不仅规模庞大,还提供了丰富的标注数据,用于代码相似性、分类、跨语言代码翻译以及代码性能优化等任务的研究。此外,数据集还提供了样本输入和输出测试集,用于确定代码的正确性,并可能指导代码质量改进的强化学习。CodeNet的应用领域广泛,旨在解决提高软件开发效率和现代化遗留应用程序的关键问题。
CodeNet is a large-scale dataset created by IBM Research, containing over 14 million code samples and approximately 500 million lines of code across 55 distinct programming languages. The dataset is primarily collected from two online judging systems, AIZU and AtCoder, to support AI-related learning and research in the field of programming. Not only does CodeNet boast a massive scale, but it also provides richly annotated data for research on tasks such as code similarity detection, code classification, cross-language code translation, and code performance optimization. Additionally, the dataset offers paired input and output test sets to verify the correctness of code, and may guide reinforcement learning for code quality improvement. CodeNet has a wide range of application scenarios, aiming to address critical issues including improving software development efficiency and modernizing legacy applications.




