I have a neural network that applies torchvision.transforms.GaussianBlur at the end of the forward method.
If I try to convert the network to TensorRT engine with Pytorch/TensorRT it does not work because it is not compatible with torchvision transformations.
I saw a interesting approach in the GeneralizedRCNNTransformation module, which seems to be compatible with Pytorch/TensorRT as is torch based and I would like to do something similar with the GaussianBlur.
Could you suggest me the way to do? Should I re-code the source code of GaussianBlur? Could be Kornia compatible?