Flexattention for efficient high-resolution vision-language modelsPublished in ECCV, 2024Share on Twitter Facebook Google+ LinkedIn Previous Next