HuggingFaceTB/OpenHermes-2.5-H4

该数据集被格式化为与alignment-handbook兼容的格式，用于监督式微调（SFT）。数据集包含多个特征字段，如id、language、conversations、topic、title、system_prompt、category、model_name、model、avatarUrl、hash、source、custom_instruction、skip_prompt_formatting、idx、views和messages。数据集分为train_sft和test_sft两个部分，分别包含951473和50078个示例。

This dataset is formatted to be compatible with the alignment-handbook for supervised fine-tuning (SFT). It includes multiple feature fields such as id, language, conversations, topic, title, system_prompt, category, model_name, model, avatarUrl, hash, source, custom_instruction, skip_prompt_formatting, idx, views, and messages. The dataset is divided into train_sft and test_sft splits, containing 951473 and 50078 examples respectively.