nraptisss/telecom-intent-config-sft-10k

这是一个专门用于训练大型语言模型（LLMs）将自然语言网络意图转换为结构化5G/6G配置的数据集。数据集包含10,000个样本，每个样本包含一个自然语言网络意图和一个符合特定模式的JSON配置。数据集覆盖了电信堆栈的四个目标层：CAMARA NetworkSliceBooking、O-RAN A1 Policy、3GPP Intent和O1 NRM Config。数据集格式为ChatML/messages格式，直接兼容于TRL `SFTTrainer`和OpenAI微调。此外，数据集还包含了详细的元数据列，如目标层、切片类型、使用案例、行业部门、地理区域等。数据集是合成的，基于模板生成，且所有样本均为英文。

This is a dataset specifically designed for training large language models (LLMs) to translate natural language network intents into structured 5G/6G configurations. The dataset contains 10,000 samples, each consisting of a natural language network intent paired with a schema-valid JSON configuration. It covers four target layers of the telecom stack: CAMARA NetworkSliceBooking, O-RAN A1 Policy, 3GPP Intent, and O1 NRM Config. The dataset is in ChatML/messages format, directly compatible with TRL `SFTTrainer` and OpenAI fine-tuning. Additionally, it includes detailed metadata columns such as target layer, slice type, use case, sector, region, etc. The dataset is synthetic, generated from templates, and all samples are in English.