very hard to replicate your notebooks.
Are there any recent updates to this question/response? I’m very comfortable with Swift and totally uncomfortable with Python… sadly… but I would love to be able to develop in Swift and just convert models to CoreML. Is that already a reasonable possibility?
@sgugger
How to specify padding shape (x,y) in conv2D of swift4TF?
Pytorch and tensorflow conv2d support this feature.
Second question is how to find shape of output after each layer of model in s4tf?
Hi Hiromi
Slightly off topic but if you are using Colab you need to copy the Package.swift and main.swift on the Colab server. Please see my blog GCE on conwyn.github.io
Regards Conwyn