File size: 2,716 Bytes
9860743 | 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 | import torch
import numpy as np
from cvxpylayers.torch import CvxpyLayer
from utils_sudoku import setup_cvx_qp_problem, get_sudoku_matrix
from models_sudoku import get_default_sudoku_params
import cvxpy as cp
def debug_infeasible_solver():
nBatch = 1
A = torch.load("sudoku/bad_A.pt")
z0_a = torch.load("sudoku/bad_z0_a.pt")
b = torch.load("sudoku/bad_b.pt")
p = torch.load("sudoku/bad_p.pt")
print("A shape: ", A.shape)
print("z0_a shape: ", z0_a.shape)
print("b shape: ", b.shape)
print("p shape: ", p.shape)
rec_b = torch.matmul(A,z0_a)
diff = torch.norm(b-rec_b).item()
print("diff b: ", diff)
assert(diff==0)
n = 2
y_dim = (n**2)**3
num_ineq = y_dim
param_vals = get_default_sudoku_params(n, Qpenalty=0.1, get_equality=True)
num_ineq = y_dim
num_eq = A.shape[0]
Q = param_vals["Q"]**0.5
G = param_vals["G"]
h = param_vals["h"]
problem, objective, ineq_functions, eq_functions, params, variables = setup_cvx_qp_problem(opt_var_dim=y_dim, num_ineq=num_ineq, num_eq=num_eq)
Q_cp, p_cp, G_cp, h_cp, A_cp, b_cp = params
y_cp = variables[0]
Q_cp.value = Q.double().cpu().numpy()
p_cp.value = p.reshape(-1).double().cpu().numpy()
G_cp.value = G.double().cpu().numpy()
h_cp.value = h.double().cpu().numpy()
A_cp.value = A.double().cpu().numpy()
b_cp.value = b.double().cpu().numpy()
y_cp = problem.solve(solver=cp.GUROBI)
print("status: ", problem.status)
print("optimal value: ", problem.value)
print("optimal var: ", y_cp.value)
print("optimal var shape: ", y_cp.value.shape)
print("check eq constraints (should be close to 0): ", np.linalg.norm(A_cp.value @ y_cp.value - b_cp.value))
print("check ineq constraints (should be <=0): ", G_cp.value @ y_cp.value - h_cp.value)
print("p: ", p_cp.value)
# optlayer = CvxpyLayer(problem, parameters=params, variables=variables)
# Q_batched = Q.unsqueeze(0).expand(nBatch, -1, -1) # (batch, y_dim, y_dim)
# G_batched = G.unsqueeze(0).expand(nBatch, -1, -1) # (batch, num_ineq, y_dim)
# h_batched = h.unsqueeze(0).expand(nBatch, -1) # (batch, num_ineq)
# A_batched = A.unsqueeze(0).expand(nBatch, -1, -1) # (batch, num_eq, y_dim)
# b_batched = b.unsqueeze(0).expand(nBatch, -1) # (batch, num_eq)
# params_batched = [Q_batched, p, G_batched, h_batched, A_batched, b_batched]
# sol, = optlayer(*params_batched)
# rec_z0_a = torch.matmul(torch.linalg.inv(A), b)
# diff = torch.norm(z0_a-rec_z0_a).item()
# assert(diff==0)
if __name__=="__main__":
debug_infeasible_solver() |