n.問題

super(Add, self).__init__()
super().__init__()

知るところ

torchvision.transforms.Grayscale(num_output_channels=3)
# grayscale의 1채널 영상을 3채널로 동일한 값으로 확장함

PYTORCH

Transfer learning

model.save

ウェイトのみ保存(Save Weights Only)

toch.save(model.state_dict(), os.path.join(MODEL_PATH, 'model.pt'))

new_model = model() # 모델정의먼저하고
new_model.load_state_dict(os.path.join(MODEL_PATH, 'model.pt') # weight로드

モデル、ウェイトの保存

torch.save(model, os.path.join(MODEL_PATH, 'model.pt'))

new_model = torch.load(os.path.join(MODEL_PATH, 'model.pt'))

checkpoints

# 현재 모델의 epoch, weight, optimizer, loss 저장
torch.save({'epoch': e,
            'model_state_dict': model.state_dict(),
            'optimizer_state_dict': optimizer.state_dict(),
            'loss': epoch_loss,
            },
            f"saved/checkpoint_model_{e}_{epoch_loss/len(dataloader)}_{epoch_acc/len(dataloader)}.pt")


checkpoint = torch.load(PATH) 
model.load_state_dict(checkpoint['model_state_dict'])
optimizer.load_state_dict(checkpoint['optimizer_state_dict'])
epoch = checkpoint['epoch']
loss = checkpoint['loss']

Transter learning

import torch
from torchvision import models

device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
vgg = models.vgg16(pretrained=True).to(device)

for name, layer in vgg.named_modules():
    print(name, layer)

vgg = models.vgg16(pretrained=True).to(device)
class MyNewNet(nn.Module):
    def __init__(self):
        super(MyNewNet, self).__init__()
        self.vgg19 = models.vgg19(pretrained=True)
        self.linear_layers = nn.Linear(1000, 1)

   def forward(self, x):
      x = self.vgg19(x)
     return self.linear_layers(x)

for param in my_model.parameters():
    param.requires_grad = False
    
for param in my_model.linear_layers.parameters():
    param.requires_grad = True

Monitoring tools for PyTorch

Tensorboard

Weight&biases