How can I insert an Input Convex Nerual Network as an objective function into the cvxpylayer?

[lk1983823]

The cvxpylayer uses the Cvxpy form of objective function as its objective in the problem, such as:

objective = cp.Minimize(0.5 * cp.pnorm(A @ x - b, p=1))
problem = cp.Problem(objective, constraints)
assert problem.is_dpp()
cvxpylayer = CvxpyLayer(problem, parameters=[A, b], variables=[x])
A_tch = torch.randn(m, n, requires_grad=True)
b_tch = torch.randn(m, requires_grad=True)

However, I want to use a more complicated convex objective to do the optimization. For example, a trained input convex neural network with pytorch code may be a suitable objective function. So how can I do this?
Thanks for the help.