Ahmadzei's picture
added 3 more tables for large emb model
5fa1a76
It set new performance records on 4 representative vision tasks, including ImageNet-V2 image classification, COCO object detection, ADE20K semantic segmentation, and Kinetics-400 video action classification.