Invalid progress was set by the TerminationCriterion.

[ffent]

Hi,

I' currently using the NSGA2 algorithm to solve an ElementwiseProblem, but encounter an AssertionError after the 9th generation.

Error

Exception has occurred: AssertionError
Invalid progress was set by the TerminationCriterion.
  File "/home/mn-man.biz/k6803/Projects/man/Perception/man-devkit/scripts/split_scenes.py", line 439, in optimize_splits
    res = minimize(
  File "/home/mn-man.biz/k6803/Projects/man/Perception/man-devkit/scripts/split_scenes.py", line 612, in main
    indices = optimize_splits(list(elements.values()), list(splits.values()), objectives, seed=42)
  File "/home/mn-man.biz/k6803/Projects/man/Perception/man-devkit/scripts/split_scenes.py", line 628, in <module>
    main(src=args.src, dst=args.dst)
AssertionError: Invalid progress was set by the TerminationCriterion.

Settings

algorithm = NSGA2(
        pop_size=100,
        eliminate_duplicates=True,
        sampling=IntegerRandomSampling(),
        mutation=PM(prob=1.0, eta=3.0, vtype=float, repair=RoundingRepair()),
        crossover=SBX(prob=1.0, eta=3.0, vtype=float, repair=RoundingRepair())
    )

termination = MaximumGenerationTermination(n_max_gen=10)

res = minimize(
        problem,
        algorithm,
        termination,
        verbose=True,
        seed=seed
    )

However, I'm a little bit confused about the fact that a simple MaximumGenerationTermination can encounter an error like "Invalid progress was set by the TerminationCriterion". Can anyone give me a hint?

Stack trace

Click to expand

==========================================================================================
n_gen  |  n_eval  | n_nds  |     cv_min    |     cv_avg    |      eps      |   indicator  
==========================================================================================
     1 |      100 |      1 |  0.3749000000 |  0.5946500000 |             - |             -
     2 |      200 |      1 |  0.3499000000 |  0.4839000000 |             - |             -
     3 |      300 |      1 |  0.2749000000 |  0.4164000000 |             - |             -
     4 |      400 |      1 |  0.1999000000 |  0.3529000000 |             - |             -
     5 |      500 |      1 |  0.1749000000 |  0.2949000000 |             - |             -
     6 |      600 |      1 |  0.0749000000 |  0.2306500000 |             - |             -
     7 |      700 |      1 |  0.0249000000 |  0.1736500000 |             - |             -
     8 |      800 |      1 |  0.0249000000 |  0.1176500000 |             - |             -
/venv/lib/python3.8/site-packages/pymoo/operators/survival/rank_and_crowding/metrics.py:97: RuntimeWarning: invalid value encountered in subtract
  dist = np.row_stack([F, np.full(n_obj, np.inf)]) - np.row_stack([np.full(n_obj, -np.inf), F])
/venv/lib/python3.8/site-packages/pymoo/operators/survival/rank_and_crowding/metrics.py:105: RuntimeWarning: invalid value encountered in divide
  dist_to_last, dist_to_next = dist_to_last[:-1] / norm, dist_to_next[1:] / norm
     9 |      900 |      5 |  0.000000E+00 |  0.0679050000 |             - |             -
/venv/lib/python3.8/site-packages/pymoo/termination/ftol.py:14: RuntimeWarning: invalid value encountered in subtract
  return np.max(np.abs((a - b) / norm))
venv/lib/python3.8/site-packages/pymoo/util/normalization.py:86: RuntimeWarning: invalid value encountered in subtract
  N[..., neither_nan] = (X[..., neither_nan] - xl[neither_nan]) / (xu[neither_nan] - xl[neither_nan])
/venv/lib/python3.8/site-packages/pymoo/util/normalization.py:86: RuntimeWarning: invalid value encountered in divide
  N[..., neither_nan] = (X[..., neither_nan] - xl[neither_nan]) / (xu[neither_nan] - xl[neither_nan])

[blankjul]

Can you quickly check if either the objective values or the constraint values are NaN?
This is what I assume is going on looking at the error message.