aeolian83/HuggingFaceH4_ultrafeedback_binarized_filtered_10k_sampled
收藏Hugging Face2025-04-09 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/aeolian83/HuggingFaceH4_ultrafeedback_binarized_filtered_10k_sampled
下载链接
链接失效反馈官方服务：
资源简介：
这是一个用于Merge-Up SLM训练的数据集，包含prompt、prompt_id、选中的(chosen)和被拒绝的(rejected)内容及其角色，以及对话消息(messages)。数据集中的每个样本都有评分和token数量，且数据来源于特定来源。数据集在采样前有7799个样本，经过英语过滤和按token长度比例采样后，用于训练。
This dataset is for Merge-Up SLM training, containing prompt, prompt_id, chosen and rejected content with their roles, and conversation messages. Each sample in the dataset has a score and token count, and the data source is from a specific origin. The dataset originally had 7799 samples, which were filtered for English language and sampled proportionally by token length for training.
提供机构：
aeolian83
