nayeeee/loss_per_token_wikitext
收藏Hugging Face2025-03-03 更新2025-04-12 收录
下载链接：
https://hf-mirror.com/datasets/nayeeee/loss_per_token_wikitext
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含四个字段：id_sequence（整型序列ID），id_token（整型序列令牌），loss_standard（标准损失，浮点型序列）和loss_strassen（斯特拉斯滕损失，浮点型序列）。数据集分为训练集、验证集和测试集，分别包含1,043,868个、2,205个和2,538个示例。数据集的总下载大小为1,848,690,137字节，解压后大小为2,070,773,468字节。
The dataset includes four fields: id_sequence (integer sequence ID), id_token (integer sequence token), loss_standard (standard loss, floating-point sequence), and loss_strassen (Strassen loss, floating-point sequence). The dataset is split into training, validation, and test sets, containing 1,043,868, 2,205, and 2,538 examples respectively. The total download size of the dataset is 1,848,690,137 bytes, and the size after decompression is 2,070,773,468 bytes.
提供机构：
nayeeee
