about episodic training and resnet

[HowCuteIsBee2002018]

Problem

1. in episodic training notebook, the model initialization state it uses resnet12, but the comment is written stated using resnet18 from pytorch?

2. also can i ask if the resnet predesigned modules given in easyfsl, what are the differences with them and the pytorch models of resnet? and why the resnet12 parameters has planes of [64, 160, 320, 640] but not other resnets?

3. will i get the similar results if different resnets being used in Prototypical Network in the original paper, because i am not sure which can be similar to compare to the model of the original paper used, can u provide some advices on this?

[ebennequin]

The comment about ResNet18 is a typo because the notebook previously used ResNet18 but was later changed toi better fit the common practices in the community. The easyfsl.modules.resnet12 follows the implementation from here, so by using it you should be able to reproduce the results for Protoypical Networks from the associated paper.

Note that there is no built-in ResNet12 in PyTorch, as it is rarely used outside of the Few-Shot Learning community.