`model.eval()` vs `torch.no_grad()`

• model.eval(): set the model to eval mode so that layers such as batch norm or drop out work in eval mode.
• torch.no_grad(): disable the autograd engine so that gradients are not computed and stored.