Two-hot/SAM-SGT
收藏官方服务：
资源简介：
SAM-SGT数据集是用于语义生成调优（SGT）的分割训练数据集，存储为5GB的tar分片文件以适应HuggingFace平台限制。该数据集作为SGT方法中的生成代理，支持统一多模态模型的视觉理解和生成能力协同优化。数据具体布局和提取说明可在相关页面查看。
The SAM-SGT dataset is a segmentation training dataset used for Semantic Generative Tuning (SGT), stored as 5GB tar shards to fit HuggingFace limits. It serves as a generative proxy in the SGT method to synergize visual understanding and generation capabilities in unified multimodal models. For data layout and extraction instructions, refer to the relevant page.
提供机构：
Two-hot
