I've been using Julia's Flux, it's great for when you have some arbitrary model you want to run gradient descent on that isn't just a bunch of matrix ops, as th

I've been using Julia's Flux, it's great for when you have some arbitrary model you want to run gradient descent on that isn't just a bunch of matrix ops, as the framework overhead is way less than TF or PyTorch due to Julia being 100x faster than pure Python.