Hello, I have a relative complex function, which I want use backward()
a = torch.zeros([3,3])
a[0,0] =1.0
a[1,2] =1.0
a[2,1] =1.0
b1 = torch.tensor([[1.0,2.0,3.0],[2.0,3.0,4.0],[3.0, 4.0,5.0]])
b2 = torch.tensor([[3.0,2.0,1.0],[4.0,3.0,2.0],[5.0, 4.0,3.0]])
c1 = torch.tensor([[1.0],[2.0],[3.0]])
c2 = torch.tensor([[4.0],[5.0],[6.0]])
c1 = Variable(c1, requires_grad=True)
c2 = Variable(c2, requires_grad=True)
b = c1.mul(b1) + c2.mul(b2)
b = Variable(b, requires_grad=True)
c = torch.div(a, b) - 1
c = Variable(c, requires_grad=True)
ref = torch.tensor([[1.0,0.0,0.0],[1.0,0.0,0.0],[1.0, 0.0,0.0]])
c.backward(ref, retain_graph=True)
d = b.grad
I want calculate the gradient of b and c1, but I alwazs got d = None. Please help me with this problem
Thank you