SOMIL366/4D4T

这是一个名为4D4T的数据集，是一个经过整理的约60GB语料库，分为四个平衡的领域，专门用于训练小型语言模型。四个领域包括：数学（来源于openbmb/UltraData-Math）、历史新闻（来源于allenai/c4的科学部分）、科学（来源于sentence-transformers/s2orc）和通用领域（来源于HuggingFaceFW/fineweb-edu）。每个领域的数据以JSONL.gz格式存储，便于加载和使用。A curated ~60GB corpus split into four balanced domains for training small language models. The domains include: Math (from openbmb/UltraData-Math), History News (from allenai/c4 realnewslike), Science (from sentence-transformers/s2orc), and General (from HuggingFaceFW/fineweb-edu). Data is stored in JSONL.gz format for easy loading and usage.