NyanDoggo/spider-cot-reasoning

Spider-Reasoning数据集是基于Spider Text-to-SQL数据集的一个扩展版本，通过OpenAI的gpt-4o模型生成了推理步骤。该数据集以alpaca风格的指令格式进行格式化，适用于微调目的。具体来说，instruction包含问题和数据库的模式（包括表、列和样本数据），而输出包含由gpt-4o生成的推理步骤以及用<final_answer>标签标记的黄金SQL。

The Spider-Reasoning dataset is an augmented version of the Spider Text-to-SQL dataset, with reasoning steps generated by OpenAIs gpt-4o model. This dataset is formatted in alpaca-style instructions, ready for finetuning purposes. Specifically, the instruction contains the question and the schema of the database (including tables, columns, and sample data), and the output contains the reasoning steps generated by gpt-4o along with the gold SQL enclosed in <final_answer> tags.