FelixFester/CodeFeedback-Filtered-Instruction

CodeFeedback-Filtered-Instruction是一个精选的代码指令查询集合，提取自四个著名的开源代码指令调优数据集：Magicoder-OSS-Instruct、Python code subset of ShareGPT、Magicoder-Evol-Instruct和Evol-Instruct-Code。最初从这些数据集中聚合了287k查询，通过使用Qwen-72B-Chat开源聊天模型进行选择性过滤，仅保留了复杂度评分为4或5的查询，最终形成了156k高质量单轮代码指令的集合。在后续处理步骤中，除了单轮打包外，仅使用了查询而不考虑响应，但在此数据集中保留了所有响应以便用户更方便地使用。

CodeFeedback-Filtered-Instruction is a curated collection of code instruction queries extracted from four prominent open-source code instruction tuning datasets: Magicoder-OSS-Instruct, Python code subset of ShareGPT, Magicoder-Evol-Instruct, and Evol-Instruct-Code. Initially, 287k queries were aggregated from these datasets. To isolate the most intricate and informative instructions, a rigorous filtering process was employed using the Qwen-72B-Chat open-source chat model, retaining only those rated 4 or 5 in complexity, resulting in a final collection of 156k high-quality single-turn code instructions. In subsequent processing steps, besides Single-turn Packing, only queries were used without considering responses, but here all responses were retained to provide users with more convenient usage options.