Hula0401/cad-sft

CAD-SFT数据集是一个用于CAD代码生成模型监督微调的数据集，包含两个子集：cad-recode-20k和text2cad/。cad-recode-20k包含20,000个样本，每个样本包括CadQuery Python源代码和渲染的4视图PNG图像。text2cad/是重新格式化的Text2CAD语料库，包含171,177个CadQuery .py文件，这些文件从单行紧凑格式重新布局为多行链式格式，以便于更友好的语言模型训练。数据集的上游来源包括CAD-Recode v1.5和Text2CAD语料库，许可证为cc-by-nc-4.0。

The CAD-SFT dataset is designed for supervised fine-tuning of CAD code-generation models, consisting of two subsets: cad-recode-20k and text2cad/. cad-recode-20k contains 20,000 samples, each including CadQuery Python source code and a rendered 4-view PNG image. text2cad/ is a reformatted Text2CAD corpus comprising 171,177 CadQuery .py files, which have been relaid from single-line compact form into multi-line chained form for more friendly LM training. The datasets upstream sources include CAD-Recode v1.5 and the Text2CAD corpus, and it is licensed under cc-by-nc-4.0.