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

Thank you very much for your quick and clear explanation.