Repeated args when iterating over the optimizer

[FelipeMLopez]

Hi, I have a problem when I try to iterate over the optimizer. This is a chunk of my code:

      repeat = True
      pre_eval = []
      while repeat:
  
          search = GlobalOptimizer(
              search_func,
              lower_bounds=loaded_lb,
              upper_bounds=loaded_ub,
              evaluations=pre_eval,
              maximize=True,
              epsilon=.01
          )
  
          num_function_calls = 50
          search.run(num_function_calls)
  
          pos_eval = search.evaluations
  
          pos_eval_x = [dicc for dicc,val in pos_eval if val > 1.0]
  
          # If we don't have enough logs -> repeat
          if 20 < len(pos_eval_x):
              repeat = False
          else:
              pre_eval = search.evaluations
              print(f'\n\n\nREPEAT! {n_opt} {len(pos_eval)} {len(pos_eval_x)}\n\n\n')
  
          del(search)
          gc.collect()

Between the first and the second iteration I do not detect any problem. From the second iteration onwards the optimizer always uses the same arguments. E.g.:

Iteration #1
#1.1 args = [ 1, 2, 3 ]
#1.2 args = [ 3, 12, 5 ]
...
#1.50 args = [ 40, 1, 7 ]

Iteration #2
#2.1 args = [ 3, 7, 1 ]
#2.2 args = [ 1, 4, 6 ]
...
#2.50 args = [ 9, 5, 81 ]

Iteration #3
#3.1 args = [ 3, 7, 1 ]
#3.2 args = [ 1, 4, 6 ]
...
#3.50 args = [ 9, 5, 81 ]

Iteration #4
#4.1 args = [ 3, 7, 1 ]
#4.2 args = [ 1, 4, 6 ]
...
#4.50 args = [ 9, 5, 81 ]

I've managed to avoid the problem by indicating a new random_state at each iteration:

      search = GlobalOptimizer(
          search_func,
          lower_bounds=loaded_lb,
          upper_bounds=loaded_ub,
          evaluations=pre_eval,
          maximize=True,
          random_state=np.random.randint(1, 1000),
          epsilon=.01
      )

If I'm updating "evaluations=" by adding the new evaluations performed, why does the optimizer keep repeating the arguments? I tried to look at the code myself but maybe you can give me a quicker answer.

Thank very much for the help!