mlfoundations-dev/oh_v1.2_sin_unnatural_instructions_diversity
收藏Hugging Face2024-12-01 更新2024-12-14 收录
下载链接：
https://hf-mirror.com/datasets/mlfoundations-dev/oh_v1.2_sin_unnatural_instructions_diversity
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含多个特征字段，如对话内容、分片ID、输出、3-gram唯一性、熵、基尼指数、自BLEU分数、嵌入向量、K均值惯性、投影梯度等。数据集包含一个训练集分割，共有821,594个样本，总大小为12,819,596,192字节，下载大小为8,031,876,758字节。
The dataset contains multiple feature fields such as conversations, shard ID, output, 3-gram uniqueness, entropy, Gini index, self-BLEU score, embeddings, K-means inertia, projected gradients, etc. The dataset includes a training split with 821,594 samples, a total size of 12,819,596,192 bytes, and a download size of 8,031,876,758 bytes.
提供机构：
mlfoundations-dev
