HuggingFaceM4/FineVision
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.
mlforge datasets pull HuggingFaceM4/FineVision
Dataset details
About HuggingFaceM4/FineVision
--- datasetinfo: - configname: CoSyn400kchart features: - name: images list: image - name: texts list: - name: user dtype: string - name: assistant dtype: string - name: source dtype: string - name: relevanceratings list: int64 - name: relevancemin dtype: int64 - name: visualdependencyratings list: int64 - name: visualdependencymin dtype: int64 - name: imagecorrespondenceratings list: int64 - name: imagecorrespondencemin dtype: int64 - name: formattingratings list: int64 - name: formattingmin dtype: int64 splits: - name: train numbytes: 25619852113.664 numexamples: 116814 downloadsize: 25239736178 datasetsize: 25619852113.664 - configname: CoSyn400kchemical features: - name: images list: image - name: texts list: - name: user dtype: string - name: assistant dtype: string - name: source dtype: string - name: imagecorrespondenceratings list: int64 - name: imagecorrespondencemin dtype: int64 - name: formattingratings list: int64 - name: formattingmin dtype: int64 - name: relevanceratings list: int64 - name: relevancemin dtype: int64 - name: visualdependencyratings list: int64 - name: visualdependencymin dtype: int64 splits: - name: train numbytes: 284197936.992 numexamples: 8942 downloadsize: 273097193 datasetsize: 284197936.992 - configname: CoSyn400kcircuit features: - name: images list: image - name: texts list: - name: user dtype: string - n