vit-small training params for imagenet

[shakedpe]

Hi,
I want to train a baseline using vit-small, what should be the training params? batch_size, learning_rate, epochs, etc etc?

[rwightman]

Something like the b16_rpn or maxvit/convnext nano should work well for vit small on imagenet: https://gist.github.com/rwightman/943c0fe59293b44024bbd2d5d23e6303