jihuny/tldr_1.4b_10k_gopt_policy_norm_penul_mean_1_iter0

该数据集包含了提示文本(prompt)、索引(prompt_idx)、选中的文本(chosen)、被拒绝的文本(rejected)及其索引，还有选中文本和拒绝文本的分数、对数概率、完成原因等信息。此外，还包含了选中和拒绝文本的嵌入向量差异，以及最佳分数和排名。数据集分为训练集，共有10000个示例，大小为180,372,974字节。

The dataset includes prompt text (prompt), index (prompt_idx), chosen text (chosen), rejected text (rejected) and their indices, as well as scores, log probabilities, finish reasons for the chosen and rejected texts. In addition, it contains the embedding vector differences for the chosen and rejected texts, optimal score, and rank. The dataset is split into a training set with a total of 10,000 examples, and it is 180,372,974 bytes in size.