[computerex]

In short, no. Not for practically large models used in common tasks like image classification or speech to text.