jihuny/tldr_1.4b_10k_gopt_policy_unnorm_penul_mean_1e-3_iter0

该数据集包含文本提示及其相关索引，选择的文本和被拒绝的文本及其相关索引，分数、对数概率、完成原因、嵌入向量差异、最优分数和排名等信息。数据集分为训练集，大小为180301441字节，包含10000个样本。

The dataset includes text prompts and their related indices, chosen and rejected texts with their related indices, scores, log probabilities, finish reasons, embedding vector differences, optimal score, and rank. The dataset is split into a training set, which is 180301441 bytes in size and contains 10,000 samples.