Hello,
I have a custom Encoder-Decoder network. The Encoder I am using is a Resnet18. However, I can only use batch_size=1
because of the memory constraints.
Therefore, I replaced the BatchNorm layers with GroupNorm+Standardized_Weights as mentioned in the GroupNorm paper.
However, my question is whether I can use the pre-trained weights (pre-trained model) provided by PyTorch? If not, Can you please let me know whether we have a pre-trained ResNet model with GroupNorm + Standardized_Weights layers?
Thank You.