python segmentation fault

Ongoing

Shlomo Hoch

• August 23, 2020 06:18

Hi,

I'm running on macbook pro 13" 2020, python 3.7.8, cvxpy 1.1.4, gurobi 9.0.3 the following problem:

N, aas = O.shape
P = len(P_sizes)
S = cp.Variable(N, boolean=True)
# A = cp.Variable(aas, boolean=True)
chi = cp.Variable(N, boolean=True)
# Op = cp.Parameter(O.shape, boolean=True)
# Op.value = O

constrains = []

A = S @ O

# for i in range(N):
#     exp = A @ Op[i] - P + (1+P)*(1-chi[i])
#     constrains.append(exp >= 0)
#     constrains.append(exp <= P)
exp = O @ A.T - P + (1+P)*(1-chi)
constrains.append(exp >= 0)
constrains.append(exp <= P)

exp = 2 * S - chi
constrains.append(exp >= 0)
constrains.append(exp <= 1)

lib_sizes = []
ind_lib_sizes = {}
ind_constrains = []
log_ind_constrains = []
pointer = 0
for i in range(P):
    s = cp.sum(A[pointer:pointer + P_sizes[i]])
    lib_sizes.append(s)
    ind_lib_sizes[i] = [cp.Variable(boolean=True) for j in range(P_sizes[i])]
    exp = cp.sum([ind_lib_sizes[i][j] * (j+1) for j in range(P_sizes[i])]) == s
    ind_constrains.append(exp)
    log_ind_constrains.extend([ind_lib_sizes[i][j] * np.log(j) for j in range(1, P_sizes[i])])
    pointer += P_sizes[i]

constrains.extend(ind_constrains)

constrains.append(cp.sum(log_ind_constrains) <= cp.log(limit))

prob = cp.Problem(cp.Maximize(c.T @ S), constrains)

Where O is a numpy boolean matrix.

I'm not sure why trying to run

prob.solve(solver=cp.GUROBI, verbose=True)

causes segmentation fault.

Thank you

0

• Official comment

  

  Matthias Miltenberger

  Gurobi Staff

  • August 24, 2020 07:32

  Hi Shlomo,

  Could you please also share the error message about the segmentation fault? Do you have a (minimal) reproducible example code that you can share so we can inspect this error?

  Thanks for reporting,
  Matthias

Please sign in to leave a comment.