fused re-parameterisation, inverted bottleneck and partial spatial convolutions
please see details in the corresponding MIDL submission
trainer python files are to be copied into lib/python3.9/site-packages/nnunet/training/network_training/nnUNet_variants/architectural_variants
please note that we restricted the #epochs to 150
watch short video: https://youtu.be/EgJZW0jRdOk