llm-jp/llava-instruct-v1_5-en-subset-358k
收藏Hugging Face2024-11-19 更新2024-12-14 收录
下载链接:
https://hf-mirror.com/datasets/llm-jp/llava-instruct-v1_5-en-subset-358k
下载链接
链接失效反馈官方服务:
资源简介:
该数据集是LLaVA-1.5指令数据的一个子集,用于训练llm-jp-3-vila-14b模型。它包含了多个子数据集,包括LLaVA(158K图像)、VQAv2(53K图像)、GQA(46K图像)、OCRVQA(80K图像)和TextVQA(22K图像)。数据集的语言为英语,任务类别为视觉问答,规模在10万到100万之间。数据集的许可证包括Creative Commons Attribution 4.0 License和OpenAI的使用条款。
This dataset is a subset of the LLaVA-1.5 Instruction Data, primarily used for visual question answering tasks. It includes images from multiple datasets such as LLaVA, VQAv2, GQA, OCRVQA, and TextVQA, totaling approximately 358K images. The purpose of this dataset is to train the llm-jp-3-vila-14b model. The dataset is licensed under Creative Commons Attribution 4.0 License and must adhere to OpenAIs terms of use.
提供机构:
llm-jp



