mytestdpo/llama3_gsm8k1_first_wrong_processed_old_format
收藏Hugging Face2024-12-29 更新2025-04-26 收录
下载链接：
https://hf-mirror.com/datasets/mytestdpo/llama3_gsm8k1_first_wrong_processed_old_format
下载链接
链接失效反馈官方服务：
资源简介：
该数据集包含了多个字段，如索引(idx)，提示(prompt)，是否为第一轮(first_round)，真实结果(gt)，奖励(rewards)，个人解决方案(my_solu)，标志(flag)，轮数(turn)，以及对话(conversations)。对话字段中又包含了内容(content)和角色(role)。数据集分为训练集(train)，共有74539个示例，大小为约435MB。数据集的配置信息中，提供了默认配置，以及训练集的数据文件路径。
The dataset includes fields such as index (idx), prompt, first round (first_round), ground truth (gt), rewards, personal solution (my_solu), flag, turn, and conversation (conversations). The conversation field contains content and role. The dataset is split into a training set (train) with a total of 74539 examples and is approximately 435MB in size. The configuration information of the dataset provides a default configuration and the data file path for the training set.
提供机构：
mytestdpo
