keep_elitism not holding elites between instances of ga_instance.run() #344

Activity

1. ahmedfgad

   added

   help wantedExtra attention is neededExtra attention is needed

   on Mar 28, 2026
2. ahmedfgad commented on Mar 28, 2026

   

   ahmedfgad

   on Mar 28, 2026

   Owner

   More actions

   The saved pickled files are not loaded properly.

   Assuming there are pickled files already saved, the code still builds a new GA instance and run it from scratch.

   Each time you build the GeneticAlgorithmInstance object using the pygad.GA class, generations_completed is set to 0.

   The following if statement is never executed then because generations_completed = 0.

   if (GeneticAlgorithmInstance.generations_completed > num_generations-1):

   Therefore, no GA is loaded.

   You proceed normally to call the run() method which uses the randomly initialized population. In other words, the pre-computed elites in the previous run are not used.

   GeneticAlgorithmInstance.run()

   To make things simple, I created 2 scripts. This is the first script that creates the GA and run it.

   import pygad
   import numpy as np
   
   def fitnessfunction(ga_instance, solution, solutionID):
       fitness = np.sum(solution)
       return fitness
   
   def on_generation(ga_instance):
       print (ga_instance.generations_completed, max(ga_instance.cal_pop_fitness()))
   
   num_generations = 100
   GeneticAlgorithmInstance = pygad.GA(random_seed=6, num_genes=100, num_generations=num_generations,
                                       sol_per_pop=100, num_parents_mating=5, parent_selection_type="sus", on_generation=on_generation,
                                       gene_type=float, fitness_func=fitnessfunction, keep_elitism=5, mutation_probability=0.2)
   
   GeneticAlgorithmInstance.run()
   GeneticAlgorithmInstance.save("E://phonysaves//"+str(GeneticAlgorithmInstance.generations_completed))
   highest = max(GeneticAlgorithmInstance.cal_pop_fitness())
   
   print(highest)

   The second script just loads the pickled file and it prints the highest fitness achieved previously.

   import pygad
   import numpy as np
   
   def fitnessfunction(ga_instance, solution, solutionID):
       fitness = np.sum(solution)
       return fitness
   
   def on_generation(ga_instance):
       print (ga_instance.generations_completed, max(ga_instance.cal_pop_fitness()))
   
   GeneticAlgorithmInstance = pygad.load("E://phonysaves//"+str(100))
   highest = max(GeneticAlgorithmInstance.cal_pop_fitness())
   print(highest)

   Hope this clarifies your doubts.