How to solve multiple knapsacks problem with genetic algorithm using PyGAD?

Question

I'm trying to solve the multiple knapsacks problem (MKP), in which I want to fit n items into m containers (knapsacks). Items have their weights and knapsacks capacity that they can hold.

I decided to solve it with genetic algorithm in Python using PyGAD. For that I expressed the problem with binary representation, but despite initial_population being mxn, the solution it's looking for has only one knapsack (1xn), so I get an error.

How to configure PyGAD to look for a m-dimensional solution?

This is my code right now:

import numpy as np
import pygad
import random

knapsacks = np.array([7, 8])
items = np.array([3, 4, 5])

initial_population = np.array([
    [1, 0, 0],  # first knapsack
    [0, 1, 0]   # second knapsack
])

def fitness_func(solution, solution_idx):
    # if sum of items in any knapsack exceeds capacity -> fail
    if np.sum(np.sum(solution*items, axis=1) > knapsacks) != 0:
        fitness = 10e4  # penalty
    # if two items in one column -> fail (an item can't be in two knapsacks)
    elif np.sum(np.argwhere(np.sum(initial_population, axis=0) > 1)):
        fitness = 10e4  # penalty
    else:
        fitness = np.sum(solution*items)
    # the higher fitness (sum of items) the better
    return fitness

ga_instance = pygad.GA(
    num_generations=30,
    num_parents_mating=2,
    fitness_func=fitness_func,
    sol_per_pop=10,
    num_genes=initial_population.size,
    gene_space=[0, 1], # binary representation
    mutation_by_replacement=True,
    gene_type=int,
    parent_selection_type="sss",    # steady_state_selection()
    keep_parents=1,
    crossover_type="single_point",
    mutation_type="random",
    mutation_num_genes=1,  # mutation_percent_genes=10,
    initial_population=initial_population
    )

ga_instance.run()
solution, solution_fitness, solution_idx = ga_instance.best_solution()