Aratako/iterative-dpo-data-for-ORPO-iter3
收藏Hugging Face2024-12-17 更新2024-12-21 收录
下载链接：
https://hf-mirror.com/datasets/Aratako/iterative-dpo-data-for-ORPO-iter3
下载链接
链接失效反馈官方服务：
资源简介：
该数据集是基于合成instruction数据创建的日语偏好数据集。使用了开发中的模型生成回答，并通过另一个模型对回答进行评分，最终将评分最高的回答标记为chosen，最低的标记为rejected。数据集包含多个特征，如prompt、chosen、rejected等，并提供了训练集的分割信息。
This dataset is a Japanese Preference dataset generated based on [Aratako/Self-Instruct-Qwen2.5-72B-Instruct-60k](https://huggingface.co/datasets/Aratako/Self-Instruct-Qwen2.5-72B-Instruct-60k). The creation process involves using the [Aratako/Llama-Gemma-2-27b-CPO_SimPO-iter2](https://huggingface.co/Aratako/Llama-Gemma-2-27b-CPO_SimPO-iter2) model to generate responses and the [Qwen/Qwen2.5-72B-Instruct-GPTQ-Int8](https://huggingface.co/Qwen/Qwen2.5-72B-Instruct-GPTQ-Int8) model to score the responses. The dataset structure includes multiple fields such as id, prompt, chosen, rejected, etc.
提供机构：
Aratako
