mlfoundations-dev/seed_math_nvidia_math_exploit

该数据集包含四个字段：指令种子(instruction_seed)、生成解决方案(generated_solution)、响应种子(response_seed)和问题来源(problem_source)，均为字符串类型。数据集有一个训练集部分，包含100万个示例，总大小为约1.1GB。提供了一种默认配置，指定了训练数据的文件路径。

The dataset includes four fields: instruction_seed, generated_solution, response_seed, and problem_source, all of which are string types. The dataset has a training set part with 1,000,000 examples, totaling approximately 1.1GB in size. A default configuration is provided, specifying the file path for the training data.