when using the pre training selfie2anime model ，torch.Size is not matched

[Xccdd]

I encountered the following problems when using the selfie2anime pre training model:

RuntimeError: Error(s) in loading state_dict for ResnetGenerator_our:
	size mismatch for deconv3_content.weight: copying a param with shape torch.Size([30, 64, 7, 7]) from checkpoint, the shape in current model is torch.Size([27, 64, 7, 7]).
	size mismatch for deconv3_content.bias: copying a param with shape torch.Size([30]) from checkpoint, the shape in current model is torch.Size([27]).

Has anyone ever had this problem? Thank you！！