yosubshin/walton-mm-mathinstruct-mmr1

这是一个结合了 Walton 和 MMR1 的视觉语言数据集，经过 data-preproc 包处理，用于视觉语言模型的训练。数据集包含了输入序列的token化ID、序列的注意力掩码、语言建模的标签、PIL图像对象、原始对话消息和处理元数据等特征。数据集共有10000个样本，处理成功率为100%。

This is a combined vision-language dataset of Walton and MMR1, processed using the data-preproc package for training vision-language models. The dataset includes tokenized input IDs, attention masks for the sequences, labels for language modeling, PIL Image objects, original conversation messages, and processing metadata. There are a total of 10,000 samples in the dataset with a processing success rate of 100%.