epoch and batchsize for miniImagenet and tieredImagenet

[ghlili]

Hello, sir , I want to know how many epochs and batchsize are used by miniImagenet and tieredImagenet in rotation and S2M2_R, respectively? thanks!!!!

[nupurkmr9]

Hi, for rotation supervision batch size is 64 otherwise 256. MiniImageNet is trained for 400 epochs and tieredImagNet is trained for 100 epochs with rotation self-supervision. Fine-tuning with manifold mixup is done for 100 and 50 more epochs respectively. Data augmentation is always used during backbone training.