ReactiveAI/TinyStories-Interaction-SFT

这是一个基于roneneldan/TinyStories数据集的监督微调数据集，用于Reactive Transformer架构的第二个训练阶段的验证。数据集包含查询和答案字段，适用于反应模型的监督微调。数据集是简单的、小型的、合成的，仅用于研究。它不包含现实世界的知识，只包含关于合成生成的简单故事的问题和答案。数据集分为训练集和验证集，每行包含两个文本字段：查询和答案，两个文本字段组合的最大长度为256个标记。

This is a supervised fine-tuning dataset based on the roneneldan/TinyStories dataset, made for the validation of the second training stage of the Reactive Transformer architecture. The dataset contains query and answer fields and is suitable for supervised fine-tuning of reactive models. The dataset is simple, small, and synthetic, intended for research use only. It does not include real-world knowledge but only questions and answers about synthetically generated simple stories. The dataset is split into training and validation sets, with each row containing two text fields: query and answer, with a combined maximum of 256 tokens.