I am using MixUp via callback like so
learn.fit(1, cbs=[MixUp(0.2)])
Will learn.tta()
be set up to apply mixup by default ? If not, how do I ensure mixup is applied in TTA ?
I am using MixUp via callback like so
learn.fit(1, cbs=[MixUp(0.2)])
Will learn.tta()
be set up to apply mixup by default ? If not, how do I ensure mixup is applied in TTA ?
MixUp is a augmentation / regularization technique. that means by using MixUp during training, you can avoid overfitting, train your model longer and thus increase the accuracy. This is done by combining two imputs / outputs during training. see https://docs.fast.ai/callback.mixup.html#MixUp
MixUp can not be applied during inference / prediction. Hope that helps.
Florian