Open ben100118 opened 6 months ago
Furthermore,what's the value of mixup?
0.1 must be some bug.
Please check the reference of https://arxiv.org/abs/2102.01243 this paper (i.e., search mixup and find the related paper). This is a data augmentation method.
Hello Yuan,I'm delighted to read your paper and reproduce your work.And I encounter some problems. When empolying the audioset_pretrain,why does the stride as same as the patch_size(overlap == 0). In addition, when I train ast on speechCommands,the acc is lower than your result(value 0.1) Thank you for taking the time to read my question; I look forward to your response!