'model.eval()' vs 'with torch.no_grad()'

Thank you for the explanation!