mlfoundations-dev/instruction_filtering_fasttext_per_domain_seed_data_code

该数据集包含了对话指令种子（instruction_seed）、来源（source）、70B模型蒸馏后的响应（r1_distill_70b_response）等字段。数据集被划分为训练集，其中包含2500个样本，大小为约102.56MB。此外，数据集还提供了原始行索引（__original_row_idx）和对话信息（conversations），对话信息中包含了对话的发送者和内容。

The dataset includes fields such as instruction seed (instruction_seed), source (source), and response after distillation of the 70B model (r1_distill_70b_response). The dataset is split into a training set, which contains 2500 samples and is about 102.56MB in size. In addition, the dataset provides the original row index (__original_row_idx) and conversation information (conversations), which includes the sender and content of the conversation.