PowerInfer/PowerMath
收藏Hugging Face2025-07-27 更新2025-08-09 收录
下载链接:
https://hf-mirror.com/datasets/PowerInfer/PowerMath
下载链接
链接失效反馈官方服务:
资源简介:
该数据集包含了大约100亿个使用Qwen2.5-14B-Instruct模型生成的预训练数据标记。通过MGA风格的方法论,从MegaMath数据集中创建了多样化的综合训练数据。数据集主要是英文的,并且可能在泛化到现实世界情况时存在限制。
This dataset contains approximately 10 billion pretraining data tokens generated using Qwen2.5-14B-Instruct. The dataset creates diverse and comprehensive training data from the MegaMath dataset using an MGA-style methodology. The dataset is primarily in English and may have limitations in generalizing to real-world scenarios.
提供机构:
PowerInfer



