jung-geun/PSO
1
0
Fork 0
mirror of https://github.com/jung-geun/PSO.git synced 2025-12-19 20:44:39 +09:00
Code Issues Packages Projects Releases Wiki Activity

23-07-06

Browse Source 
dev container 실행 코드 추가
This commit is contained in:
jung-geun
2023-07-06 22:04:42 +09:00
parent c7384cdf7b
commit c163de6cb6
11 changed files with 192 additions and 132 deletions
Download Patch File Download Diff File Expand all files Collapse all files

85
pso/optimizer.py
View File

@@ -15,10 +15,13 @@ gpus = tf.config.experimental.list_physical_devices("GPU")
if gpus:
try:
# tf.config.experimental.set_visible_devices(gpus[0], "GPU")
print(tf.config.experimental.get_visible_devices("GPU"))
tf.config.experimental.set_memory_growth(gpus[0], True)
print("set memory growth")
except RuntimeError as e:
print(e)
class Optimizer:
"""
particle swarm optimization
@@ -41,7 +44,7 @@ class Optimizer:
):
"""
particle swarm optimization
Args:
model (keras.models): 모델 구조
loss (str): 손실함수
@@ -55,7 +58,7 @@ class Optimizer:
"""
np.random.seed(np_seed)
tf.random.set_seed(tf_seed)
self.model = model # 모델 구조
self.loss = loss # 손실함수
self.n_particles = n_particles # 파티클 개수
@@ -66,11 +69,12 @@ class Optimizer:
self.w_max = w_max # 최대 관성 수치
self.negative_swarm = negative_swarm # 최적해와 반대로 이동할 파티클 비율 - 0 ~ 1 사이의 값
self.mutation_swarm = mutation_swarm # 관성을 추가로 사용할 파티클 비율 - 0 ~ 1 사이의 값
self.g_best_score = [0 , np.inf] # 최고 점수 - 시작은 0으로 초기화
self.g_best_score = [0, np.inf] # 최고 점수 - 시작은 0으로 초기화
self.g_best = None # 최고 점수를 받은 가중치
self.g_best_ = None # 최고 점수를 받은 가중치 - 값의 분산을 위한 변수
self.avg_score = 0 # 평균 점수
self.save_path = None # 저장 위치
negative_count = 0
for i in tqdm(range(self.n_particles), desc="Initializing Particles"):
@@ -81,16 +85,18 @@ class Optimizer:
m.set_weights(self._decode(w_, sh_, len_))
m.compile(loss=self.loss, optimizer="sgd", metrics=["accuracy"])
self.particles[i] = Particle(
m,
loss,
negative=True if i < negative_swarm * self.n_particles else False,
m,
loss,
negative=True if i < negative_swarm * self.n_particles else False,
mutation=mutation_swarm,
)
)
if i < negative_swarm * self.n_particles:
negative_count += 1
print(f"negative swarm : {negative_count} / {self.n_particles}")
print(f"mutation swarm : {mutation_swarm * self.n_particles} / {self.n_particles}")
print(
f"mutation swarm : {mutation_swarm * self.n_particles} / {self.n_particles}"
)
gc.collect()
@@ -110,11 +116,10 @@ class Optimizer:
del self.avg_score
gc.collect()
def _encode(self, weights):
"""
가중치를 1차원으로 풀어서 반환
Args:
weights (list) : keras model의 가중치
Returns:
@@ -132,15 +137,14 @@ class Optimizer:
w_gpu = np.append(w_gpu, w_)
del weights
return w_gpu, shape, length
return w_gpu, shape, length
def _decode(self, weight, shape, length):
"""
_encode 로 인코딩된 가중치를 원본 shape으로 복원
파라미터는 encode의 리턴값을 그대로 사용을 권장
Args:
weight (numpy array): 가중치 - 1차원으로 풀어서 반환
shape (list): 가중치의 원본 shape
@@ -156,7 +160,7 @@ class Optimizer:
w_ = np.reshape(w_, shape[i])
weights.append(w_)
start = end
del weight
del shape
del length
@@ -183,7 +187,6 @@ class Optimizer:
else:
return 1 + np.abs(score)
def fit(
self,
x,
@@ -204,7 +207,7 @@ class Optimizer:
save : bool - True : save, False : not save
save_path : str ex) "./result",
renewal : str ex) "acc" or "loss",
empirical_balance : bool - True :
empirical_balance : bool - True :
Dispersion : bool - True : g_best 의 값을 분산시켜 전역해를 찾음, False : g_best 의 값만 사용
check_point : int - 저장할 위치 - None : 저장 안함
"""
@@ -226,7 +229,7 @@ class Optimizer:
except ValueError as e:
print(e)
sys.exit(1)
for i in tqdm(range(self.n_particles), desc="Initializing velocity"):
p = self.particles[i]
local_score = p.get_score(x, y, renewal=renewal)
@@ -247,7 +250,7 @@ class Optimizer:
if local_score[1] == None:
local_score[1] = 0
if save:
with open(
f"./{save_path}/{self.day}_{self.n_particles}_{epochs}_{self.c0}_{self.c1}_{self.w_min}_{renewal}.csv",
@@ -258,15 +261,22 @@ class Optimizer:
f.write(", ")
else:
f.write("\n")
f.close()
del local_score
gc.collect()
print(f"initial g_best_score : {self.g_best_score[0] if self.renewal == 'acc' else self.g_best_score[1]}")
print(
f"initial g_best_score : {self.g_best_score[0] if self.renewal == 'acc' else self.g_best_score[1]}"
)
try:
epochs_pbar = tqdm(range(epochs), desc=f"best {self.g_best_score[0]:.4f}|{self.g_best_score[1]:.4f}", ascii=True, leave=True)
epochs_pbar = tqdm(
range(epochs),
desc=f"best {self.g_best_score[0]:.4f}|{self.g_best_score[1]:.4f}",
ascii=True,
leave=True,
)
for epoch in epochs_pbar:
acc = 0
loss = 0
@@ -277,9 +287,16 @@ class Optimizer:
ts = self.c0 + np.random.rand() * (self.c1 - self.c0)
part_pbar = tqdm(range(len(self.particles)), desc=f"acc : {max_score:.4f} loss : {min_loss:.4f}", ascii=True, leave=False)
part_pbar = tqdm(
range(len(self.particles)),
desc=f"acc : {max_score:.4f} loss : {min_loss:.4f}",
ascii=True,
leave=False,
)
for i in part_pbar:
part_pbar.set_description(f"acc : {max_score:.4f} loss : {min_loss:.4f}")
part_pbar.set_description(
f"acc : {max_score:.4f} loss : {min_loss:.4f}"
)
w = self.w_max - (self.w_max - self.w_min) * epoch / epochs
g_, g_sh, g_len = self._encode(self.g_best)
@@ -298,7 +315,7 @@ class Optimizer:
w_g_ = self.f(x, y, self.g_best)
w_p = w_p_ / (w_p_ + w_g_)
w_g = w_p_ / (w_p_ + w_g_)
del w_p_
del w_g_
@@ -315,7 +332,7 @@ class Optimizer:
p_ = np.exp(-1 * p_)
w_p = p_
w_g = 1 - p_
del p_b
del g_a
del l_b
@@ -339,7 +356,9 @@ class Optimizer:
if score[0] < self.g_best_score[1]:
self.g_best_score[1] = score[0]
self.g_best = self.particles[i].get_best_weights()
epochs_pbar.set_description(f"best {self.g_best_score[0]:.4f} | {self.g_best_score[1]:.4f}")
epochs_pbar.set_description(
f"best {self.g_best_score[0]:.4f} | {self.g_best_score[1]:.4f}"
)
elif renewal == "loss":
if score[0] <= self.g_best_score[1]:
if score[0] < self.g_best_score[1]:
@@ -349,7 +368,9 @@ class Optimizer:
if score[1] > self.g_best_score[0]:
self.g_best_score[0] = score[1]
self.g_best = self.particles[i].get_best_weights()
epochs_pbar.set_description(f"best {self.g_best_score[0]:.4f} | {self.g_best_score[1]:.4f}")
epochs_pbar.set_description(
f"best {self.g_best_score[0]:.4f} | {self.g_best_score[1]:.4f}"
)
if score[0] == None:
score[0] = np.inf
@@ -386,7 +407,7 @@ class Optimizer:
self.avg_score = acc / self.n_particles
gc.collect()
except KeyboardInterrupt:
print("Ctrl + C : Stop Training")
except MemoryError:
@@ -407,7 +428,7 @@ class Optimizer:
Returns:
(keras.models): 모델
"""
"""
model = keras.models.model_from_json(self.model.to_json())
model.set_weights(self.g_best)
model.compile(loss=self.loss, optimizer="sgd", metrics=["accuracy"])
@@ -459,7 +480,7 @@ class Optimizer:
"a",
) as f:
json.dump(json_save, f, indent=4)
f.close()
def _check_point_save(self, save_path: str = f"./result/check_point"):