sungyub/kodcode-v1-verl
收藏Hugging Face2025-11-07 更新2025-11-15 收录
下载链接：
https://hf-mirror.com/datasets/sungyub/kodcode-v1-verl
下载链接
链接失效反馈官方服务：
资源简介：
KodCode-V1 VERL Format是基于KodCode/KodCode-V1数据集的预处理版本，适用于VERL（Versatile Reinforcement Learning）训练。数据集包含高质量的Python编码挑战，具有全面的测试用例和质量指标。数据集采用VERL兼容的架构，并保留了原始数据集中的质量指标，如通过率、测试覆盖率等。数据集的大小为434,896个训练示例，语言为Python代码生成。
This dataset is a preprocessed version of KodCode/KodCode-V1 formatted for VERL (Versatile Reinforcement Learning) training. It contains high-quality Python coding challenges with comprehensive test cases and quality metrics. The dataset follows the VERL training schema with a consistent dictionary key ordering and retains original quality metrics such as pass percentage and test coverage. It consists of 434,896 training examples and is designed for Python code generation.
提供机构：
sungyub
