FelixFester/OpenHermes-2.5

OpenHermes 2.5数据集是Open Hermes 1数据集的延续，规模更大、多样性更丰富、质量更高，包含100万条主要通过合成生成的指令和聊天样本。该数据集支撑了Open Hermes 2/2.5和Nous Hermes 2模型的显著进步。数据集遵循sharegpt结构，每个条目包含一个对话列表，每个对话回合有“from”字段表示角色和“value”字段包含实际文本。数据集来源包括Airoboros 2.2、CamelAI领域专家数据集、ChatBot Arena（仅GPT-4）、Collective Cognition等多个开源数据集和自定义合成数据集。

The OpenHermes 2.5 dataset is a continuation of the Open Hermes 1 dataset, at a much larger scale, much more diverse, and much higher quality compilation, reaching 1M, primarily synthetically generated instruction and chat samples. The Open Hermes 2/2.5 and Nous Hermes 2 models have made significant advancements of SOTA LLMs over recent months, and are underpinned by this exact compilation and curation of many open source datasets and custom created synthetic datasets. The dataset follows a sharegpt structure, with each entry containing a list of conversations, each turn having a from field denoting the role and a value field containing the actual text. The dataset sources include Airoboros 2.2, CamelAI Domain Expert Datasets, ChatBot Arena (GPT-4 Only), Collective Cognition, and many other open-source datasets and custom synthetic datasets.