jihuny/tldr_1.4b_10k_gopt_policy_norm_penul_prev_1e-6_iter0
收藏Hugging Face2025-11-03 更新2025-11-15 收录
下载链接：
https://hf-mirror.com/datasets/jihuny/tldr_1.4b_10k_gopt_policy_norm_penul_prev_1e-6_iter0
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了一系列与提示（prompt）相关的信息，包括选择的答案（chosen）和被拒绝的答案（rejected），以及它们各自的索引、得分、对数概率和完成原因等。此外，还包括了嵌入向量的差异。数据集被划分为训练集，大小为182254225字节，共有10000个示例。数据集的目的是用于训练模型以进行文本选择或评估。
The dataset contains a series of information related to prompts, including chosen answers and rejected answers, along with their indices, scores, log probabilities, and finish reasons. It also includes the difference in embedding vectors. The dataset is split into a training set, which is 182254225 bytes in size and contains 10,000 examples. The purpose of the dataset is to train models for text selection or evaluation.
提供机构：
jihuny
