search

google-research / big_vision

Official codebase used to develop Vision Transformer, SigLIP, MLP-Mixer, LiT and more.
Apache License 2.0
2.2k stars 147 forks source link

augmentation and regularization used in MLP-Mixer #5

Closed Ga-Lee closed 2 years ago

Ga-Lee commented 2 years ago

Hello! Thank you for your work! In the paper of MLP-Mixer, when training mixer-B/16 on imagenet1k from scratch, it is said extra regularization is applied to gain the the accuracy of 76% , I want to know what detailed augmentation and regularization strategy is used for the experiment? Is there any config file can be found? Thank you for your help! : )

akolesnikoff commented 2 years ago

Hi!

I've just submitted the MLP-Mixer code, as well as the configuration that achieves 76% top-1 ImageNet with MLP-Mixer-B/16 model: https://github.com/google-research/big_vision/blob/main/big_vision/configs/mlp_mixer_i1k.py. Let me know if you have any further questions.

Ga-Lee commented 2 years ago

Hi!

I've just submitted the MLP-Mixer code, as well as the configuration that achieves 76% top-1 ImageNet with MLP-Mixer-B/16 model: https://github.com/google-research/big_vision/blob/main/big_vision/configs/mlp_mixer_i1k.py. Let me know if you have any further questions.

Got it ! Thank you for your reply!