4 ms·
Yup, there's plenty of ML architectures that try to save on parameters size, achieving better generalization (less overfitting) at the expense of slightly costl
by zozbot234 5y ago
Yup, there's plenty of ML architectures that try to save on parameters size, achieving better generalization (less overfitting) at the expense of slightly costlier training and inference. The memory constraints on Tesla Dojo might not be a big deal after all.