I could only read the abstract. This is obviously cool in terms of having a dedicated high-speed inference unit. What I'm curious about is how well this extend

I could only read the abstract. This is obviously cool in terms of having a dedicated high-speed inference unit. What I'm curious about is how well this extends to actual modern NN architecture. They talk about image classification, but is it just with a MLP, or is there a pathway to implementing convolution layers or things like batch-norm. There is that "MLP mixer" network, so maybe it would be viable for many tasks even if an MLP is all they have...