41 ms·
Optimization problems in PyTorch are always parameterized with a smooth manifold structure so they're always differentiable, otherwise gradient descent wouldn't
by haltist 3y ago
Optimization problems in PyTorch are always parameterized with a smooth manifold structure so they're always differentiable, otherwise gradient descent wouldn't work.