Can we use pretrained models with iresnet backbone?

[ZubairKhan001]

Seems like pertained ms1mv3 arcface models cant be pretrained with a smaller custom dataset. Is that correct.

def _iresnet(arch, block, layers, pretrained, progress, **kwargs):
model = IResNet(block, layers, **kwargs)
if pretrained:
raise ValueError()
return model

[nttstar]

Yes, why not?

[ZubairKhan001]

please, how can we do this in train_v2 file in arcface_torch as it only has --resume configuration option.