FineVision HuggingFaceM4
Fine Vision FineVision is a massive collection of datasets with 17.3M images, 24.3M samples, 88.9M turns, and 9.5B answer tokens, designed for training state-of-the-art open Vision-Language-Models. More detail can be found in the blog post: https://huggingface.co/spaces/HuggingFaceM4/FineVision Load the data from datasets import load_dataset, get_dataset_config_names # Get all subset names and load the first one available_subsets =… See the full description on the dataset page: https://huggingface.co/datasets/HuggingFaceM4/FineVision.
- 类型
- dataset
- 下载量
- 112,161
- 点赞
- 512
- 访问
- public
- 文件
- 0
标签
- 多模态数据集
- 视觉语言模型
- 指令微调
- 对话数据
- 图像数据集
- 多模态
- 文本生成
- 视觉问答
摘要
FineVision 是 Hugging Face 团队打造的超大规模视觉语言模型（VLM）训练数据集，包含 1730 万张图像、2430 万样本、8890 万轮对话和 95 亿答案 token，覆盖图表、化学、电路、图表、文档、数学、音乐、营养、表格等多样化的视觉领域。该数据集带有人工标注的相关性、视觉依赖度、图像对应度、格式质量等多维度评分，适合用于微调和训练开源视觉语言模型，解决多模态理解与视觉问答任务。适用于科研与工业界训练高性能 VLM，尤其适合需要细粒度视觉-文本对齐的场景。
README
--- dataset_info: - config_name: CoSyn_400k_chart features: - name: images list: image - name: texts list: - name: user dtype: string - name: assistant dtype: string - name: source dtype: string - name: relevance_ratings list: int64 - name: relevance_min dtype: int64 - name: visual_dependency_ratin…