VARGPT-family/VARGPT_datasets
收藏Hugging Face2025-04-09 更新2025-02-15 收录
下载链接：
https://hf-mirror.com/datasets/VARGPT-family/VARGPT_datasets
下载链接
链接失效反馈官方服务：
资源简介：
VARGPT_datasets是用于训练VARGPT模型的多个数据集的集合。它包括三个阶段的指令微调数据集：阶段一包含1.28M预训练指令微调数据集；阶段二包括来自LLaVA-1.5和LLaVA-OneVision数据集的样本，以及从ImageNet-Instruct-130k数据集采样的数据；阶段三包含ImageNet-Instruct-130k和ImageNet-Instruct-1270k数据集。此外，还有关于图像数据下载、处理以及如何将指令微调数据路径与图像路径匹配的说明。
VARGPT_datasets is a collection of datasets for training the VARGPT model. It includes instruction fine-tuning datasets for three stages: stage one contains 1.28M pre-training instruction fine-tuning datasets; stage two includes samples from LLaVA-1.5 and LLaVA-OneVision datasets, as well as data sampled from the ImageNet-Instruct-130k dataset; stage three contains the ImageNet-Instruct-130k and ImageNet-Instruct-1270k datasets. Moreover, there are instructions on image data download, processing, and how to match instruction fine-tuning data paths with image paths.
提供机构：
VARGPT-family
