for resnet12 and resnet18

[DanqingZ]

did your implementations train from scratch or load the pretrained weights like below?

import torchvision.models as models
resnet18 = models.resnet18(pretrained=True)

[fmu2]

This implementation assumes randomly initialized ResNets. The conv layer in the stem has smaller kernel size since the input size is 84 (which is typical for meta-learning/few-shot learning), so pre-trained models would not work here.