실시간 그래프 코드

[신민서]

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81

import matplotlib.pyplot as plt   class LivePlot:     def __init__(self):         plt.ion()  # 인터랙티브 모드 활성화         self.fig = plt.figure(figsize=(12, 5))         self.ax1 = self.fig.add_subplot(1, 2, 1)         self.ax2 = self.fig.add_subplot(1, 2, 2)           self.x_arr = []         self.loss_hist_train = []         self.loss_hist_valid = []         self.accuracy_hist_train = []         self.accuracy_hist_valid = []       def update_loss_train(self, epoch, loss_train): # 훈련 손실 그래프 업데이트         if len(self.loss_hist_train) == 0 or self.x_arr[-1] != epoch: # 에포크 추가 조건문             self.x_arr.append(epoch)  # 에포크 추가             # 위 조건문을 넣지 않으면 x, y의 오류가 생김 -> x값인 에포크만 다른 함수들이 계속 추가하면 너무 커짐짐           self.loss_hist_train.append(loss_train)         self.plot_graphs()       def update_loss_valid(self, epoch, loss_valid): # 검증 손실 그래프 업데이트         if len(self.loss_hist_train) == 0 or self.x_arr[-1] != epoch:             self.x_arr.append(epoch)  # 에포크 추가           self.loss_hist_valid.append(loss_valid)         self.plot_graphs()       def update_accuracy_train(self, epoch, acc_train): # 훈련 정확도 그래프 업데이트         if len(self.loss_hist_train) == 0 or self.x_arr[-1] != epoch:             self.x_arr.append(epoch)  # 에포크 추가                      self.accuracy_hist_train.append(acc_train)         self.plot_graphs()       def update_accuracy_valid(self, epoch, acc_valid): # 검증 정확도 그래프 업데이트         if len(self.loss_hist_train) == 0 or self.x_arr[-1] != epoch:             self.x_arr.append(epoch)  # 에포크 추가                  self.accuracy_hist_valid.append(acc_valid)         self.plot_graphs()       def update_all(self, epoch, loss_train, loss_valid, acc_train, acc_valid): # 모든 그래프 업데이트         if len(self.loss_hist_train) == 0 or self.x_arr[-1] != epoch:             self.x_arr.append(epoch)  # 에포크 추가           self.loss_hist_train.append(loss_train)         self.loss_hist_valid.append(loss_valid)         self.accuracy_hist_train.append(acc_train)         self.accuracy_hist_valid.append(acc_valid)         self.plot_graphs()       def plot_graphs(self): # 그래프 그리는 함수         self.ax1.clear()         self.ax1.plot(self.x_arr[:len(self.loss_hist_train)], self.loss_hist_train, '-', label='Train loss')         self.ax1.plot(self.x_arr[:len(self.loss_hist_valid)], self.loss_hist_valid, '--', label='Validation loss')         self.ax1.set_xlabel('Epoch', size=15)         self.ax1.set_ylabel('Loss', size=15)         self.ax1.legend(fontsize=15)           self.ax2.clear()         self.ax2.plot(self.x_arr[:len(self.accuracy_hist_train)], self.accuracy_hist_train, '-', label='Train acc.')         self.ax2.plot(self.x_arr[:len(self.accuracy_hist_valid)], self.accuracy_hist_valid, '--', label='Validation acc.')         self.ax2.set_xlabel('Epoch', size=15)         self.ax2.set_ylabel('Accuracy', size=15)         self.ax2.legend(fontsize=15)           plt.draw()         plt.pause(0.1)                  print("train : ", len(self.loss_hist_train))         print("val : ", len(self.loss_hist_valid))         print("에폭 리스트 : ", self.x_arr[:len(self.loss_hist_valid)],"손실값" ,self.loss_hist_valid)         print("에폭 리스트 : ", self.x_arr[:len(self.accuracy_hist_train)], "정확도", self.accuracy_hist_train)                def show(self): # 출력         plt.savefig('image.png') # 사진 저장 savefig()         plt.show() Colored by Color Scripter

cs

from (위 코드를 저장한 파일명) import LivePlot 

예시 코드

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137

import torch import torch.nn as nn import torchvision from torchvision import transforms from torch.utils.data import DataLoader, Subset from liveplot import LivePlot   # Device 설정 device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')   # 데이터 전처리 def get_smile(attr):     return attr[31]  # 웃음 레이블 추출   transform_train = transforms.Compose([     transforms.RandomCrop([178, 178]),     transforms.RandomHorizontalFlip(),     transforms.Resize([64, 64]),     transforms.ToTensor(), ])   transform = transforms.Compose([     transforms.CenterCrop([178, 178]),     transforms.Resize([64, 64]),     transforms.ToTensor(), ])   # 데이터셋 로드 image_path = './'   celeba_train_dataset = torchvision.datasets.CelebA(     image_path, split='train',     target_type='attr', download=False,     transform=transform_train, target_transform=get_smile )   celeba_valid_dataset = torchvision.datasets.CelebA(     image_path, split='valid',     target_type='attr', download=False,     transform=transform, target_transform=get_smile )   celeba_test_dataset = torchvision.datasets.CelebA(     image_path, split='test',     target_type='attr', download=False,     transform=transform, target_transform=get_smile )   # 데이터셋 서브셋 가져오기 celeba_train_dataset = Subset(celeba_train_dataset, torch.arange(16000)) celeba_valid_dataset = Subset(celeba_valid_dataset, torch.arange(1000))   # 데이터로더 작성 batch_size = 32 train_dl = DataLoader(celeba_train_dataset, batch_size, shuffle=True) valid_dl = DataLoader(celeba_valid_dataset, batch_size, shuffle=False) test_dl = DataLoader(celeba_test_dataset, batch_size, shuffle=False)     # CNN 모델 정의 model = nn.Sequential(     nn.Conv2d(3, 32, kernel_size=3, padding=1),     nn.ReLU(),     nn.MaxPool2d(kernel_size=2),     nn.Dropout(p=0.5),     nn.Conv2d(32, 64, kernel_size=3, padding=1),     nn.ReLU(),     nn.MaxPool2d(kernel_size=2),     nn.Dropout(p=0.5),     nn.Conv2d(64, 128, kernel_size=3, padding=1),     nn.ReLU(),     nn.MaxPool2d(kernel_size=2),     nn.Conv2d(128, 256, kernel_size=3, padding=1),     nn.ReLU(),     nn.AvgPool2d(kernel_size=8),     nn.Flatten(),     nn.Linear(256, 1),     nn.Sigmoid() ) model.to(device)  # 모델을 GPU로 이동     def train(model, num_epochs, train_dl, valid_dl):     loss_fn = nn.BCELoss()     optimizer = torch.optim.Adam(model.parameters(), lr=0.001)          plotter = LivePlot()  # LivePlot 객체 생성          for epoch in range(num_epochs):         model.train()         loss_tmp = 0         accuracy_tmp = 0                  for x_batch, y_batch in train_dl:             x_batch, y_batch = x_batch.to(device), y_batch.to(device)             optimizer.zero_grad()             pred = model(x_batch)[:, 0]             loss = loss_fn(pred, y_batch.float())             loss.backward()             optimizer.step()                          loss_tmp += loss.item() * y_batch.size(0)             is_correct = ((pred >= 0.5).float() == y_batch).float().cpu()             accuracy_tmp += is_correct.sum()                  loss_train = loss_tmp / len(train_dl.dataset)         acc_train = accuracy_tmp / len(train_dl.dataset)                  # 검증 루프         model.eval‎()         with torch.no_grad():             loss_tmp = 0             accuracy_tmp = 0                          for x_batch, y_batch in valid_dl:                 x_batch, y_batch = x_batch.to(device), y_batch.to(device)                 pred = model(x_batch)[:, 0]                 loss = loss_fn(pred, y_batch.float())                 loss_tmp += loss.item() * y_batch.size(0)                 is_correct = ((pred >= 0.5).float() == y_batch).float().cpu()                 accuracy_tmp += is_correct.sum()                  loss_valid = loss_tmp / len(valid_dl.dataset)         acc_valid = accuracy_tmp / len(valid_dl.dataset)                  plotter.update_all(epoch, loss_train, loss_valid, acc_train, acc_valid)  # 그래프 업데이트         print(f'에포크 {epoch+1} 정확도: {acc_train:.4f} 검증 정확도: {acc_valid:.4f}')          plotter.show()  # 최종 그래프 출력          return loss_train, loss_valid, acc_train, acc_valid       torch.manual_seed(1) num_epochs = 100 hist = train(model, num_epochs, train_dl, valid_dl) Colored by Color Scripter

cs

https://github.com/mxnseo/LivePlottor.git