search

microsoft / unilm

Large-scale Self-supervised Pre-training Across Tasks, Languages, and Modalities
https://aka.ms/GeneralAI
MIT License
20.2k stars 2.55k forks source link

question ablout table 6 of the paper #856

Closed cobraheleah closed 2 years ago

cobraheleah commented 2 years ago

Hey, Thank you publishing the paper for BEiT2. I wonder know, in table 6 of the paper, what is Performance of teacher models means ?

pengzhiliang commented 2 years ago

Yes, we finetune the teacher with the BEiT fine-tuning recipe. Of course, we sweep enough large scope of hyper-parameters, to get a high performance. Because the original papers (CLIP/DINO) do not report the fine-tuning results.

Thanks, Zhiliang.