def __test__():
# global Testobj
sys.setrecursionlimit(30)
file = 'pooltest5.pkl'
filecls = 'pooltest_cls.pkl'
cls = Testobj0
print(cls.param)
if os.path.isfile(file):
objs = ut.load(file)
cls = ut.load(filecls)
print(cls.param)
else:
pool = []
for i in range(1000):
obj = cls(pool)
objs = [x for x in pool]
ut.save(file, objs)
ut.save(filecls, cls)
print(len(objs))
obj = objs[500]
# print(pool[500], len(pool))
print(obj.id, obj.data)
obj()
print(cls.param)
def plot_loss_ex(trainer):
fig, ax = plt.subplots()
for d in ['top', 'right']:
ax.spines[d].set_visible(False)
ylim_low = float('inf')
ylim_upp = 0
try:
with ut.chdir(trainer.out):
if not os.path.isfile('log.json'):
return
log = ut.load('log.json', from_json=True)
for key in ('main/loss', 'val/main/loss'):
a = np.array([l[key] for l in log])
a = np.clip(a, 0, 1e6)
ax.plot(a, label=key)
ylim_upp = max(np.ceil(np.max(a[min(len(a)-1,3):50]))/1000*1000,
ylim_upp)
ylim_low = min(np.min(a)//1000*1000, ylim_low)
ax.set_ylim((ylim_low, ylim_upp))
ax.set_xlabel('epoch')
ax.grid(True)
fig.legend()
fig.savefig('loss1.png')
finally:
plt.close(fig)
sleep(10)
def get_result_file(root='.'):
''' 計算済み最適化ファイルを読み出す
'''
with ut.chdir(root):
files = glob.glob('optimize_*.pkl')
for i, file in enumerate(files):
print(f'[{i}] {file}')
print('select file')
n = int(input())
if n < 0:
return
file = files[n]
print('file:', file)
env, optimizer, history = ut.load(file)
return env, optimizer, history
def get_model(out):
# モデル読み込み
# model_cls = {'nsga2':NSGA2, 'moead':MOEAD}[model]
files = ut.fsort(glob.glob(os.path.join(out, f'*epoch*.pkl')))
for i, file in enumerate(files):
print(f'[{i}]', file)
print('select file')
n = int(input())
if n == -1:
pass
elif n < 0:
return
file = files[n]
print('file:', file)
env, optimizer, history = ut.load(file)
return env, optimizer, history
def main1(out='result', force=False, clear_directory=False):
''' GAテスト & プロット
'''
###
def ln(ind, origin):
if not origin:
return []
return list(zip(origin[0].value, ind.value, origin[1].value))
def dist(ind, origin):
if not origin:
return 0
values = np.array([[ind.value, par.value] for par in origin])
diff = values[:, :, 0] - values[:, :, 1]
dists = np.sqrt(np.sum(diff**2, axis=1))
return np.sum(dists)
fig, ax = plt.subplots()
def plot(pop):
pairs = [(fit.data, fit.data.origin.origin or ()) for fit in pop]
parents = list(chain(*(x[1] for x in pairs)))
lines = [ln(ind, origin) for ind, origin in pairs]
# print(lines)
# exit()
if lines and [x for x in lines if x]:
dists = [dist(ind, origin) for ind, origin in pairs]
print(sum(dists))
ax.cla()
cm = plt.get_cmap('jet')
# print(cm(10))
# exit()
if parents:
x_p, y_p = np.array([ind.value for ind in parents]).T
ax.scatter(x_p, y_p, c='r')
for i, l in enumerate(lines):
if l:
plt.plot(*l, c=cm(i / (len(lines) + 1)), linewidth=0.5)
x, y = np.array([fit.data.value for fit in pop]).T
ax.scatter(x, y, c='b')
plt.pause(1e-10)
###
def local_main():
n_dim = 30
popsize = 100
problem = zdt1
with Environment() as env:
# 個体クラス
indiv_type = Individual
# 初期個体生成クラス
indiv_pool = env.register(indiv_type)
# 遺伝子生成クラス
initializer = UniformInitializer(n_dim)
creator = Creator(initializer, indiv_pool)
# # 適応度クラス
# fit_type = Fitness
# # 初期個体生成クラス
# evaluator = env.register(fit_type)
###
# optimizer = NSGA2(problem=problem, pool=indiv_pool)
optimizer = MOEAD(problem=problem, pool=indiv_pool, ksize=5)
# optimizer.set_initializer(Initializer(3))
# optimizer.setup(problem)
### Additional setting ###
# optimizer.initializer = initializer
# optimizer.n_cycle = None
# optimizer.alternation = 'replace'
##########################
# indivs = [creator() for _ in range(popsize)]
# population = Population([ind.evaluate(problem) for ind in indivs])
# optimizer.calc_fitness(population)
population = optimizer.init_population(creator, popsize=popsize)
history = [population]
# print(population[0].data.origin)
# return
for i in range(100):
population = optimizer(population)
plot(population)
origin = population[0].data.origin.origin or []
# print([x.id for x in origin], '->', population[0].data.id)
history.append(population)
if i % 50 == 50 - 1:
ut.save(f'result/temp{i}.pkl', history)
plt.show()
return env, optimizer, history
###
def resume_main(env, optimizer, history):
print('resume_main')
for i, population in enumerate(history):
plot(population)
origin = population[0].data.origin.origin or []
# print([x.id for x in origin], '->', population[0].data.id)
plt.show()
file = 'test_moead.pkl'
if os.path.exists(file) and not force:
env, optimizer, history = ut.load(file)
resume_main(env, optimizer, history)
else:
env, optimizer, history = local_main()
ut.save(file, (env, optimizer, history))
def __test__():
obj = TestClass('abc')
ut.save('test.pkl', obj)
obj = ut.load('test.pkl')