oumi-ai/limo-vis-mid-resize
收藏Hugging Face2025-07-09 更新2025-08-09 收录
下载链接：
https://hf-mirror.com/datasets/oumi-ai/limo-vis-mid-resize
下载链接
链接失效反馈官方服务：
资源简介：
limo-vis-mid-resize数据集保留了原始数据集的结构，并通过令牌长度和图像质量进行了过滤。该数据集使用data-preproc包进行处理，用于视觉语言模型的训练。数据集特征包括令牌化的输入序列、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。数据集共有598个样本，处理成功率为100%。
The limo-vis-mid-resize dataset preserves the original structure of the dataset and is filtered by token length and image quality. This dataset was processed using the data-preproc package for vision-language model training. Dataset features include tokenized input sequences, attention masks for sequences, labels for language modeling, PIL Image objects, original conversation messages, and processing metadata. The dataset contains a total of 598 samples with a success rate of 100%.
提供机构：
oumi-ai
