harithoppil/ml-swe-prompts

ML SWE Prompts是一个用于OPD蒸馏训练的机器学习/训练相关软件工程提示的统一集合。所有提示都是英文的。数据集过滤了一些核心的ML仓库，如huggingface、numpy、Lightning-AI和ray-project，排除了pandas-dev、qiskit等其他仓库。数据集包含多个配置，如all、swe_bench_ml、swe_dev_sft、swe_dev_rft、terminal_bench_verified和terminal_bench_trajectories，每个配置有不同的来源、行数和描述。数据集的模式（schema）包括source和prompt等字段，不同来源的字段有所差异。ML SWE Prompts is a unified collection of ML/training-related software engineering prompts for OPD distillation training. All prompts are in English. Filtered to core ML repos: huggingface, numpy, Lightning-AI, ray-project. Excludes pandas-dev, qiskit, open-mmlab, scipy, tensorflow, spaCy. The dataset includes multiple configs such as all, swe_bench_ml, swe_dev_sft, swe_dev_rft, terminal_bench_verified, and terminal_bench_trajectories, each with different sources, rows, and descriptions. The schema includes fields like source and prompt, with additional fields varying by source.