chai-ppo-rm-implicit
收藏资源简介:
该数据集用于训练RP奖励模型,包含三个特征:chosen(字符串类型)、rejected(字符串类型)和margin(浮点数类型)。数据集分为训练集和测试集,训练集包含162819个样本,测试集包含1024个样本。数据集的总下载大小为314960195字节,总数据集大小为566584477.0字节。数据集的配置名为'default',数据文件路径分别为'data/train-*'和'data/test-*'。更好的格式化有助于训练更快收敛并达到更高的最终准确度。
This dataset is designed for training an RP reward model, and includes three features: chosen (string type), rejected (string type), and margin (float type). The dataset is split into training and test sets, with the training set containing 162,819 samples and the test set containing 1,024 samples. The total download size of the dataset is 314,960,195 bytes, while the total dataset size is 566,584,477.0 bytes. The configuration name of the dataset is "default", and the data file paths are "data/train-*" and "data/test-*" respectively. Proper formatting helps accelerate training convergence and achieve higher final accuracy.
数据集概述
数据集信息
-
特征:
- chosen: 类型为字符串
- rejected: 类型为字符串
- margin: 类型为浮点数
-
分割:
- train:
- 字节数: 563043388.8580104
- 样本数: 162819
- test:
- 字节数: 3541088.141989588
- 样本数: 1024
- train:
-
下载大小: 314960195
-
数据集大小: 566584477.0
配置
- 配置名称: default
- 数据文件:
- train: data/train-*
- test: data/test-*
- 数据文件:
用途
- 用于训练RP奖励模型。




