leonli66/tokenized_llama_nemotron_math

数据集包含两种类型的配置：思考型和非思考型。每种配置都有对应的训练数据集，数据以.parquet格式存储。每个数据示例由聊天格式的prompt字段和目标target字段组成。The dataset includes two types of configurations: thinking and non-thinking. Each configuration has its corresponding training dataset, stored in the .parquet format. Each data example consists of a chat-formatted prompt field and a target field.