lin = nn.Linear(1, 1)
xx = torch.randn(5, 1)
print(torch.allclose(lin(xx), xx @ lin.weight.T + lin.bias))    # True