def train_population(pop: Population, games: list, unused_cpu: int = 2):
"""Evaluate the given population on a training set."""
multi_env = get_multi_env(pop=pop, game_config=pop.config)
multi_env.set_games(games, noise=False)
pool = mp.Pool(mp.cpu_count() - unused_cpu)
manager = mp.Manager()
return_dict = manager.dict()
pbar = tqdm(total=len(pop.population), desc="Evaluating")
def update(*_):
pbar.update()
for genome_id, genome in pop.population.items():
pool.apply_async(func=multi_env.eval_genome,
args=((genome_id, genome), return_dict),
callback=update)
pool.close() # Close the pool
pool.join() # Postpone continuation until everything is finished
pbar.close()
# Calculate the fitness from the given return_dict
fitness = calc_pop_fitness(
fitness_cfg=pop.config.evaluation,
game_cfg=pop.config.game,
game_obs=return_dict,
gen=pop.generation,
)
for i, genome in pop.population.items():
genome.fitness = fitness[i]
# Save the results
pop.generation += 1
pop.save()
def evaluate_population(pop: Population, cfg: Config, cpu: int, experiment_id: int):
"""Evaluate the given population."""
pop.log(f"{pop.name} - Evaluating the population...")
_, game_ids_eval = get_game_ids(experiment_id=experiment_id)
multi_env = get_multi_env(pop=pop, game_config=cfg)
multi_env.set_games(game_ids_eval, noise=False)
pool = mp.Pool(mp.cpu_count() - cpu)
manager = mp.Manager()
return_dict = manager.dict()
pbar = tqdm(total=len(pop.population), desc="Evaluating")
def update(*_):
pbar.update()
for genome_id, genome in pop.population.items():
pool.apply_async(func=multi_env.eval_genome, args=((genome_id, genome), return_dict), callback=update)
pool.close() # Close the pool
pool.join() # Postpone continuation until everything is finished
pbar.close()
# Calculate the fitness from the given return_dict
pop.log(f"{pop.name} - Calculating fitness scores...")
fitness = calc_pop_fitness(
fitness_cfg=pop.config.evaluation,
game_cfg=cfg.game,
game_obs=return_dict,
gen=pop.generation,
)
for i, genome in pop.population.items():
genome.fitness = fitness[i]
# Get the fittest genome
best = None
for g in pop.population.values():
if best is None or g.fitness > best.fitness: best = g
pop.best_genome = best
# Save the results
pop.save()
# Visualize most fit genome
visualize_genome(
debug=True,
genome=best,
population=pop,
)
# Trace the most fit genome
trace_most_fit(
debug=False,
games=game_ids_eval,
genome=best,
population=pop,
unused_cpu=cpu,
)
def evaluate_same_games_and_evolve(
self,
games: list,
pop: Population,
n: int = 1,
parallel=True,
save_interval: int = 1,
):
"""
Evaluate the population on the same games.
:param games: List of games used for training
:param pop: Population object
:param n: Number of generations
:param parallel: Parallel the code (disable parallelization for debugging purposes)
:param save_interval: Indicates how often a population gets saved
"""
multi_env = get_multi_env(pop=pop, game_config=self.game_config)
msg = f"Repetitive evaluating games: {games}"
pop.log(msg, print_result=False)
multi_env.set_games(games)
# Iterate and evaluate over the games
saved = True
for iteration in range(n):
single_evaluation(
multi_env=multi_env,
parallel=parallel,
pop=pop,
unused_cpu=self.unused_cpu,
)
# Save the population
if (iteration + 1) % save_interval == 0:
pop.save()
saved = True
else:
saved = False
# Make sure that last iterations saves
if not saved: pop.save()
def evaluate_and_evolve(
self,
pop: Population,
n: int = 1,
parallel=True,
save_interval: int = 1,
):
"""
Evaluate the population for a single evaluation-process.
:param pop: Population object
:param n: Number of generations
:param parallel: Parallel the code (disable parallelization for debugging purposes)
:param save_interval: Indicates how often a population gets saved
"""
multi_env = get_multi_env(pop=pop, game_config=self.game_config)
saved = True
for iteration in range(n):
# Set random set of games
self.sample_games(multi_env, pop.log)
# Evaluate the population on the newly sampled games
single_evaluation(
multi_env=multi_env,
parallel=parallel,
pop=pop,
unused_cpu=self.unused_cpu,
)
# Save the population
if (iteration + 1) % save_interval == 0:
pop.save()
saved = True
else:
saved = False
# Make sure that last iterations saves
if not saved: pop.save()
def train(
population: Population,
game_config: Config,
games: list,
iterations: int,
unused_cpu: int = 0,
save_interval: int = 10,
):
"""Train the population on the requested number of iterations. Manual adaptation of main's train()."""
population.log("\n===> TRAINING <===\n")
multi_env = get_multi_env(pop=population, game_config=game_config)
msg = f"Repetitive evaluating on games: {games} for {iterations} iterations"
population.log(msg, print_result=False)
# Iterate and evaluate over the games
saved = True
for iteration in range(iterations):
# Set and randomize the games
multi_env.set_games(games, noise=True)
# Prepare the generation's reporters for the generation
population.reporters.start_generation(gen=population.generation,
logger=population.log)
# Fetch the dictionary of genomes
genomes = list(iteritems(population.population))
# Initialize the evaluation-pool
pool = mp.Pool(mp.cpu_count() - unused_cpu)
manager = mp.Manager()
return_dict = manager.dict()
for genome in genomes:
pool.apply_async(func=multi_env.eval_genome,
args=(genome, return_dict))
pool.close() # Close the pool
pool.join() # Postpone continuation until everything is finished
# Calculate the fitness from the given return_dict
fitness = calc_pop_fitness(
fitness_cfg=population.config.evaluation,
game_cfg=game_config.game,
game_obs=return_dict,
gen=population.generation,
)
for i, genome in genomes:
genome.fitness = fitness[i]
# Gather and report statistics
best = None
for g in itervalues(population.population):
if best is None or g.fitness > best.fitness: best = g
population.reporters.post_evaluate(population=population.population,
species=population.species,
best_genome=best,
logger=population.log)
# Update the population's best_genome
genomes = sorted(population.population.items(),
key=lambda x: x[1].fitness,
reverse=True)
population.best_fitness[population.generation] = genomes[0][1].fitness
population.best_genome_hist[population.generation] = genomes[0]
population.best_genome = best
# Let population evolve
population.evolve()
# Update the genomes such all have one hidden node
for g in population.population.values():
n_hidden, _ = g.size()
while n_hidden < 1:
g.mutate_add_connection(population.config.genome)
n_hidden, _ = g.size()
# End generation
population.reporters.end_generation(population=population.population,
name=str(population),
species_set=population.species,
logger=population.log)
# Save the population
if (iteration + 1) % save_interval == 0:
population.save()
saved = True
else:
saved = False
# Make sure that last iterations saves
if not saved: population.save()
def main(
fitness,
prob_gru: float,
prob_sru: float,
prob_lstm: float,
version=0,
unused_cpu=1,
):
"""
Run a population's configuration.
:param fitness: Fitness function used to evaluate the population
:param prob_gru: Probability of mutating towards a GRU-node
:param prob_sru: Probability of mutating towards a SRU-node
:param prob_lstm: Probability of mutating towards a LSTM-node
:param version: Version of the model
:param unused_cpu: Number of CPUs not used during training
"""
# Re-configure the config-file
cfg = Config()
cfg.bot.angular_dir = []
cfg.bot.delta_dist_enabled = False
cfg.bot.dist_enabled = True
cfg.game.duration = 60 # 60 seconds should be enough to reach the target from each starting orientation
cfg.population.pop_size = 512
# Let inputs apply to configuration
cfg.genome.rnn_prob_gru = prob_gru
cfg.genome.rnn_prob_simple_rnn = prob_sru
cfg.genome.rnn_prob_lstm = prob_lstm
cfg.evaluation.fitness = fitness
cfg.update()
# Copy population over from experiment1
name = get_name(cfg=cfg, version=version)
path_exp1 = f'population/storage/experiment1/{name}/'
if not os.path.exists(path_exp1):
raise Exception(
f"Experiment 1 must be executed first for population {name}, terminating experiment 2..."
)
# Population exists in experiment1, copy over to experiment2 (change experiment1 population's folder and save)
pop = Population(name=name,
config=cfg,
folder_name=get_folder(experiment_id=1),
use_backup=False)
assert pop.generation > 0 # Population is not new (redundant check)
folder = get_folder(experiment_id=2)
pop.folder_name = folder
pop.save() # Overrides pre-existing populations!
# Copy over all generations as well, since these are used during population evaluation
path = f"population{'_backup' if pop.use_backup else ''}/storage/{pop.folder_name}/{pop}/"
copy_tree(f"{path_exp1}generations", f"{path}generations")
# Give overview of population
gru = cfg.genome.rnn_prob_gru
sru = cfg.genome.rnn_prob_simple_rnn
lstm = cfg.genome.rnn_prob_lstm
msg = f"\n\n\n\n\n===> RUNNING EXPERIMENT 2 FOR THE FOLLOWING CONFIGURATION: <===" \
f"\n\t> fitness: {cfg.evaluation.fitness}" \
f"\n\t> GRU enabled: {gru > 0} (probability={round(gru, 2)})" \
f"\n\t> SRU enabled: {sru > 0} (probability={round(sru, 2)})" \
f"\n\t> LSTM enabled: {lstm > 0} (probability={round(lstm, 2)})" \
f"\n\t> Saving under folder: {folder}\n"
pop.log(msg)
# Set games used for evaluation
_, games_eval = get_game_ids(experiment_id=2)
# Execute the requested segments
try:
# Evaluate the trained population
evaluate(
games=games_eval,
population=pop,
unused_cpu=unused_cpu,
)
except Exception as e:
pop.log(traceback.format_exc(), print_result=False)
raise e
finally:
process_killer('run_population.py') # Close all the terminated files
def main(pop_name: str,
version: int,
unused_cpu: int = 2,
use_backup: bool = False):
# Check if valid population name
if pop_name not in SUPPORTED:
raise Exception(f"Population '{pop_name}' not supported!")
# Create the population
cfg = get_config()
cfg.population.specie_elitism = 1
folder = get_folder(experiment_id=7)
pop = Population(
name=f'{pop_name}/v{version}',
config=cfg,
folder_name=folder,
use_backup=use_backup,
)
# Replace the population's initial population with the requested topologies genomes
if pop.generation == 0:
for g_id in pop.population.keys():
pop.population[g_id] = get_topology(pop_name, gid=g_id, cfg=cfg)
pop.species.speciate(config=pop.config,
population=pop.population,
generation=pop.generation,
logger=pop.log)
pop.log(f"\n\n\n===> RUNNING EXPERIMENT 7 <===\n")
# Set games and environment used for training and evaluation
games_train, games_eval = get_game_ids(experiment_id=7)
train_env = get_multi_env(config=cfg)
eval_env = get_multi_env(config=cfg)
eval_env.set_games(games_eval, noise=False)
solution_found = False
while not solution_found:
# Train the population for a single iteration
pop.log("\n===> TRAINING <===")
train_env.set_games(games_train, noise=True)
# Prepare the generation's reporters for the generation
pop.reporters.start_generation(gen=pop.generation, logger=pop.log)
# Fetch the dictionary of genomes
genomes = list(iteritems(pop.population))
# Initialize the evaluation-pool
pool = mp.Pool(mp.cpu_count() - unused_cpu)
manager = mp.Manager()
return_dict = manager.dict()
for genome in genomes:
pool.apply_async(func=train_env.eval_genome,
args=(genome, return_dict))
pool.close() # Close the pool
pool.join() # Postpone continuation until everything is finished
# Calculate the fitness from the given return_dict
fitness = calc_pop_fitness(
fitness_cfg=pop.config.evaluation,
game_cfg=cfg.game,
game_obs=return_dict,
gen=pop.generation,
)
for i, genome in genomes:
genome.fitness = fitness[i]
# Update the population's best_genome
best = None
for g in itervalues(pop.population):
if best is None or g.fitness > best.fitness: best = g
pop.reporters.post_evaluate(population=pop.population,
species=pop.species,
best_genome=best,
logger=pop.log)
# Update the population's best_genome
genomes = sorted(pop.population.items(),
key=lambda x: x[1].fitness,
reverse=True)
pop.best_fitness[pop.generation] = genomes[0][1].fitness
pop.best_genome_hist[pop.generation] = genomes[0]
pop.best_genome = best
# Let population evolve
evolve(pop, pop_name)
# End generation
pop.reporters.end_generation(population=pop.population,
name=str(pop),
species_set=pop.species,
logger=pop.log)
# Test if evaluation finds a solution for the new generation, impossible if fitness < 0.7
if pop.best_genome.fitness > 0.7 or pop.generation % 10 == 0:
pop.log("\n===> EVALUATING <===")
genomes = list(iteritems(pop.population))
pool = mp.Pool(mp.cpu_count() - unused_cpu)
manager = mp.Manager()
return_dict = manager.dict()
for genome in genomes:
pool.apply_async(func=eval_env.eval_genome,
args=(genome, return_dict))
pool.close() # Close the pool
pool.join() # Postpone continuation until everything is finished
# Calculate the fitness from the given return_dict
finished = calc_finished_ratio(
fitness_cfg=cfg.evaluation,
game_obs=return_dict,
)
best = None
for i, genome in genomes:
genome.fitness = finished[i]
if best is None or finished[i] > best.fitness: best = genome
pop.log(f"Best genome:\n{best}\n{best.nodes[2]}")
# Solution is found
if best.fitness == 1:
pop.best_genome = best
pop.log(f"Solution found!")
solution_found = True # End the outer while-loop
# Save the population with their evaluation results
pop.save()
def set_population(
pop: Population,
hops: float,
weight_range: float,
mutate_bias: bool = False,
mutate_reset: bool = False,
mutate_update: bool = False,
mutate_candidate: bool = False,
):
"""Set the given population as mutations of the given genome."""
assert weight_range > hops
r = int(weight_range / hops)
# Re-initialize the weight around the provided genome
pop.population = dict()
genome_key = 0
# Create genome-mutations
if mutate_bias:
for i in range(3):
for a in range(-r, r + 1):
new_genome = copy.deepcopy(pop.best_genome)
new_genome.nodes[2].bias_h[
i] = new_genome.nodes[2].bias_h[i] + a * hops
new_genome.key = genome_key
pop.population[genome_key] = new_genome
genome_key += 1
if mutate_reset:
for a in range(-r, r + 1):
for b in range(-r, r + 1):
new_genome = copy.deepcopy(pop.best_genome)
new_genome.nodes[2].weight_xh_full[
0, 0] = new_genome.nodes[2].weight_xh_full[0, 0] + a * hops
new_genome.nodes[2].weight_hh[
0, 0] = new_genome.nodes[2].weight_hh[0, 0] + b * hops
new_genome.key = genome_key
pop.population[genome_key] = new_genome
genome_key += 1
if mutate_update:
for a in range(-r, r + 1):
for b in range(-r, r + 1):
new_genome = copy.deepcopy(pop.best_genome)
new_genome.nodes[2].weight_xh_full[
1, 0] = new_genome.nodes[2].weight_xh_full[1, 0] + a * hops
new_genome.nodes[2].weight_hh[
1, 0] = new_genome.nodes[2].weight_hh[1, 0] + b * hops
new_genome.key = genome_key
pop.population[genome_key] = new_genome
genome_key += 1
if mutate_candidate:
for a in range(-r, r + 1):
for b in range(-r, r + 1):
new_genome = copy.deepcopy(pop.best_genome)
new_genome.nodes[2].weight_xh_full[
2, 0] = new_genome.nodes[2].weight_xh_full[2, 0] + a * hops
new_genome.nodes[2].weight_hh[
2, 0] = new_genome.nodes[2].weight_hh[2, 0] + b * hops
new_genome.key = genome_key
pop.population[genome_key] = new_genome
genome_key += 1
# Save the updated population
pop.save()
def evaluate_and_evolve(
self,
pop: Population,
n: int = 1,
parallel=True,
save_interval: int = 1,
):
"""
Evaluate the population on the same set of games.
:param pop: Population object
:param n: Number of generations
:param parallel: Parallel the code (disable parallelization for debugging purposes)
:param save_interval: Indicates how often a population gets saved
"""
multi_env = get_multi_env(pop=pop, game_config=self.game_config)
msg = f"Repetitive evaluating on games: {self.games} for {n} iterations"
pop.log(msg, print_result=False)
# Iterate and evaluate over the games
saved = True
for iteration in range(n):
# Set and randomize the games
multi_env.set_games(self.games, noise=True)
# Prepare the generation's reporters for the generation
pop.reporters.start_generation(gen=pop.generation, logger=pop.log)
# Fetch the dictionary of genomes
genomes = list(iteritems(pop.population))
if parallel:
# Initialize the evaluation-pool
pool = mp.Pool(mp.cpu_count() - self.unused_cpu)
manager = mp.Manager()
return_dict = manager.dict()
for genome in genomes:
pool.apply_async(func=multi_env.eval_genome,
args=(genome, return_dict))
pool.close() # Close the pool
pool.join(
) # Postpone continuation until everything is finished
else:
return_dict = dict()
for genome in tqdm(genomes, desc="sequential training"):
multi_env.eval_genome(genome, return_dict)
# Calculate the fitness from the given return_dict
fitness = calc_pop_fitness(
fitness_cfg=pop.config.evaluation,
game_cfg=self.game_config.game,
game_obs=return_dict,
gen=pop.generation,
)
for i, genome in genomes:
genome.fitness = fitness[i]
# Gather and report statistics
best = None
for g in itervalues(pop.population):
if best is None or g.fitness > best.fitness: best = g
pop.reporters.post_evaluate(population=pop.population,
species=pop.species,
best_genome=best,
logger=pop.log)
# Update the population's best_genome
genomes = sorted(pop.population.items(),
key=lambda x: x[1].fitness,
reverse=True)
pop.best_fitness[pop.generation] = genomes[0][1].fitness
pop.best_genome_hist[pop.generation] = genomes[0]
pop.best_genome = best
# Let population evolve
pop.evolve()
# End generation
pop.reporters.end_generation(population=pop.population,
name=str(pop),
species_set=pop.species,
logger=pop.log)
# Save the population
if (iteration + 1) % save_interval == 0:
pop.save()
saved = True
else:
saved = False
# Make sure that last iterations saves
if not saved: pop.save()
def create_genomes(genome: Genome, pop: Population, d: int, range_width: int):
"""Create mutations of the provided genome and inject these in the given population.."""
pop.log(f"{pop.name} - Setting up genomes...")
pbar = tqdm(range(((range_width + range_width + 1) ** 2) * 3), desc="Generating genomes")
genome_key = 0
gru_node_id = None
for node_id, node in genome.nodes.items():
if type(node) == GruNodeGene:
gru_node_id = node_id
break
# Create the reset-mutated genomes
for a in range(-range_width, range_width + 1):
w_xh = np.asarray([[a / d], [0], [0]])
for b in range(-range_width, range_width + 1):
w_hh = np.asarray([[b / d], [0], [0]])
# Add the specified genome to the population
new_genome = copy.deepcopy(genome)
new_genome.nodes[gru_node_id].weight_xh_full += w_xh
new_genome.nodes[gru_node_id].weight_hh += w_hh
new_genome.fitness = None
new_genome.key = genome_key
pop.population[genome_key] = new_genome
genome_key += 1
pbar.update()
# Create the update-mutated genomes
for a in range(-range_width, range_width + 1):
w_xh = np.asarray([[0], [a / d], [0]])
for b in range(-range_width, range_width + 1):
w_hh = np.asarray([[0], [b / d], [0]])
# Add the specified genome to the population
new_genome = copy.deepcopy(genome)
new_genome.nodes[gru_node_id].weight_xh_full += w_xh
new_genome.nodes[gru_node_id].weight_hh += w_hh
new_genome.fitness = None
new_genome.key = genome_key
pop.population[genome_key] = new_genome
genome_key += 1
pbar.update()
# Create the candidate-mutated genomes
for a in range(-range_width, range_width + 1):
w_xh = np.asarray([[0], [0], [a / d]])
for b in range(-range_width, range_width + 1):
w_hh = np.asarray([[0], [0], [b / d]])
# Add the specified genome to the population
new_genome = copy.deepcopy(genome)
new_genome.nodes[gru_node_id].weight_xh_full += w_xh
new_genome.nodes[gru_node_id].weight_hh += w_hh
new_genome.fitness = None
new_genome.key = genome_key
pop.population[genome_key] = new_genome
genome_key += 1
pbar.update()
pbar.close()
assert len(pop.population) == (((range_width - -range_width + 1) ** 2) * 3)
pop.save()
def main(topology_id: int,
batch_size: int = 1000,
train_batch: int = 3,
min_finished: float = MIN_FINISHED,
unused_cpu: int = 2,
save_pop: bool = False,
use_backup: bool = False):
"""Run a population infinitely long and store all its good genomes."""
# Get the CSV used to store the results in
csv_path, csv_name, added = get_csv_path(topology_id, use_backup=use_backup, batch_size=batch_size)
# Create the population
name = csv_name if save_pop else 'dummy'
cfg = get_config()
folder = get_folder(experiment_id=6)
pop = Population(
name=name,
config=cfg,
folder_name=folder,
use_backup=use_backup,
overwrite=True, # Every iteration, create a new population from scratch
)
# Replace the population's initial population with the requested topologies genomes
for g_id in pop.population.keys(): pop.population[g_id] = get_genome(topology_id, g_id=g_id, cfg=cfg)
pop.species.speciate(config=pop.config,
population=pop.population,
generation=pop.generation,
logger=pop.log)
# Set games and environment used for training and evaluation
pop.log(f"\n\n\n===> RUNNING EXPERIMENT 6 <===\n")
games_train, games_eval = get_game_ids(experiment_id=6)
train_env = get_multi_env(config=cfg)
eval_env = get_multi_env(config=cfg)
eval_env.set_games(games_eval, noise=False)
# Keep training and evolving the network until the complete CSV is filled
last_saved = pop.generation
try:
while added < batch_size:
t = time.localtime()
pop.log(f"\n\n===> Selective genome creation at {added / batch_size * 100}%, "
f"storing in csv '{csv_path.split('/')[-1]}' "
f"({t.tm_hour:02d}h-{t.tm_min:02d}m-{t.tm_sec:02d}s) <===")
# Train the population
pop.log("\n===> Training <===")
for _ in tqdm(range(train_batch), desc="Training"):
train_env.set_games(games_train, noise=True)
genomes = list(iteritems(pop.population))
# Initialize the evaluation-pool
pool = mp.Pool(mp.cpu_count() - unused_cpu)
manager = mp.Manager()
return_dict = manager.dict()
for genome in genomes:
pool.apply_async(func=train_env.eval_genome, args=(genome, return_dict))
pool.close() # Close the pool
pool.join() # Postpone continuation until everything is finished
# Calculate the fitness from the given return_dict
fitness = calc_pop_fitness(
fitness_cfg=pop.config.evaluation,
game_cfg=cfg.game,
game_obs=return_dict,
gen=pop.generation,
)
for i, genome in genomes:
genome.fitness = fitness[i]
# Update the population's best_genome
best = None
for g in itervalues(pop.population):
if best is None or g.fitness > best.fitness: best = g
genomes = sorted(pop.population.items(), key=lambda x: x[1].fitness, reverse=True)
pop.best_fitness[pop.generation] = genomes[0][1].fitness
pop.best_genome_hist[pop.generation] = genomes[0]
pop.best_genome = best
pop.log(f"Best training fitness: {best.fitness}")
# Let population evolve
pop.evolve()
# Constraint each of the population's new genomes to the given topology
for g in pop.population.values():
enforce_topology(g, topology_id=topology_id)
# Save the population after training
if pop.generation - last_saved >= 100:
pop.save()
last_saved = pop.generation
# Evaluate the current population as was done in experiment6
pop.log("\n===> EVALUATING <===")
genomes = list(iteritems(pop.population))
pool = mp.Pool(mp.cpu_count() - unused_cpu)
manager = mp.Manager()
return_dict = manager.dict()
for genome in genomes:
pool.apply_async(func=eval_env.eval_genome, args=(genome, return_dict))
pool.close() # Close the pool
pool.join() # Postpone continuation until everything is finished
# Calculate the fitness from the given return_dict
finished = calc_finished_ratio(
fitness_cfg=cfg.evaluation,
game_obs=return_dict,
)
best = None
for i, genome in genomes:
genome.fitness = finished[i]
if best is None or finished[i] > best.fitness: best = genome
# Give evaluation overview of population
pop.log(f"Best evaluation finish ratio: {round(best.fitness, 2)}")
best_str = str(best).replace("\n", "\n\t")
best_str += "\n\t" + str(best.nodes[2]).replace("\n", "\n\t")
pop.log(f"Best genome: \n\t{best_str}")
sids = list(iterkeys(pop.species.species))
sids.sort()
msg = f"\nPopulation '{name}' has {len(pop.species.species):d} species:" \
f"\n\t specie age size finished stag " \
f"\n\t======== ===== ====== ========== ======"
pop.log(msg) if pop.log else print(msg)
for sid in sids:
s = pop.species.species[sid]
a = pop.generation - s.created
n = len(s.members)
sf = [g.fitness for g in s.members.values() if g.fitness]
f = "--" if len(sf) == 0 else f"{max(sf):.2f}"
st = pop.generation - s.last_improved
msg = f"\t{sid:^8} {a:^5} {n:^6} {f:^10} {st:^6}"
pop.log(msg) if pop.log else print(msg)
# Write the result to CSV
with open(csv_path, 'a', newline='') as f:
writer = csv.writer(f)
for _, g in genomes:
# Only write the genomes that exceed the minimum 'finished ratio' threshold!
if g.fitness >= min_finished:
writer.writerow(get_genome_parameters(g, topology_id=topology_id))
added += 1
finally:
# Remove the dummy population if it exists
pop.save()
path = f"population{'_backup' if use_backup else ''}/storage/{pop.folder_name}/dummy/"
if os.path.exists(path):
shutil.rmtree(path)
