Beckschen / TransMix

[CVPR 2022] This repository includes the official project for the paper: TransMix: Attend to Mix for Vision Transformers.
Apache License 2.0
155 stars 12 forks source link

baseline results #12

Open mingkai-zheng opened 2 years ago

mingkai-zheng commented 2 years ago

I just checked the configuration file, it seems that some of the training strategies are quite different from the original DeiT training recipe, (e.g. Batch size, learning rate scheduler, model ema ...) So I'm wondering what would be the baseline result for this configuration?