annotated_deep_learning_paper_implementations
annotated_deep_learning_paper_implementations copied to clipboard
Stride = 2 only for the first block of the entire ResNet
According to this row, ResNet is created only with one convolution with stride = 2:
https://github.com/labmlai/annotated_deep_learning_paper_implementations/blame/ec3c7366b7c7a763b0c06a6f703ded6bc2267638/labml_nn/resnet/init.py#L288
Is it ok?