【pytorch】冻结、固定部分参数

我这里用resnet18举例，加入这一块代码即可。
如果不知道你的神经网络模型中有哪些层，可以像我一样先print（model_ft）看看。
比如我这里冻结了前1、2、3层，你也可以冻结其他的层。

    print(model_ft)
    frozen_layers = [model_ft.layer1,model_ft.layer2,model_ft.layer3]
    for layer in frozen_layers:
        for name, value in layer.named_parameters():
            value.requires_grad = False
    params = filter(lambda p: p.requires_grad, model_ft.parameters())
    optimizer_ft = torch.optim.SGD(params, lr=0.001)

整体代码：

from __future__ import print_function, division
from torchvision.datasets import ImageFolder
from torch.utils.data import DataLoader
import torch.nn as nn
import torch
import torch.optim as optim
from torch.optim import lr_scheduler
from torch.autograd import Variable
import numpy as np
import torch.nn as nn
import torch.nn.functional as F
import torch.optim as optim
from torchvision import datasets, transforms
import time
import os
import copy
from torchvision import models
import matplotlib.pyplot as plt


data_transforms = {
    'train': transforms.Compose([
        transforms.RandomResizedCrop(224),
        transforms.RandomHorizontalFlip(),
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
    ]),
    'valid': transforms.Compose([
        transforms.Resize(256),
        transforms.CenterCrop(224),
        transforms.ToTensor(),
        transforms.Normalize([0.485, 0.456, 0.406], [0.229, 0.224, 0.225])
    ]),
}

data_dir = 'maize'
image_datasets = {x: datasets.ImageFolder(os.path.join(data_dir, x),
                                          data_transforms[x])
                  for x in ['train', 'valid']}
dataloaders = {x: torch.utils.data.DataLoader(image_datasets[x], batch_size=4,
                                             shuffle=True, num_workers=1)
              for x in ['train', 'valid']}
dataset_sizes = {x: len(image_datasets[x]) for x in ['train', 'valid']}
class_names = image_datasets['train'].classes

device = torch.device("cuda:0" if torch.cuda.is_available() else "cpu")

#更新学习率
def exp_lr_scheduler(optimizer, epoch, init_lr=0.001, lr_decay_epoch=7):
    """Decay learning rate by a factor of 0.1 every lr_decay_epoch epochs."""
    lr = init_lr * (0.1 ** (epoch // lr_decay_epoch))

    if epoch % lr_decay_epoch == 0:
        print('LR is set to {}'.format(lr))

    for param_group in optimizer.param_groups:
        param_group['lr'] = lr

    return optimizer


def train_model(model, criterion, optimizer, scheduler, num_epochs):
    since = time.time()

    best_model_wts = copy.deepcopy(model.state_dict())
    best_acc = 0.0

    for epoch in range(num_epochs):
        print('Epoch {}/{}'.format(epoch+1, num_epochs))
        print('-' * 10)

        # Each epoch has a training and validation phase
        for phase in ['train', 'valid']:
            if phase == 'train':
      #        scheduler.step()
               optimizer=scheduler(optimizer,epoch)
               model.train(True)  # Set model to training mode
            else:
                model.train(False)  # Set model to evaluate mode

            running_loss = 0.0
            running_corrects = 0
            # Iterate over data.
            for data in dataloaders[phase]:
                # get the inputs
                inputs, labels = data

                inputs, labels = Variable(inputs), Variable(labels)

                # zero the parameter gradients
                optimizer.zero_grad()

                # forward
                outputs = model(inputs)
                _, preds = torch.max(outputs.data, 1)
                loss = criterion(outputs, labels)

                # backward + optimize only if in training phase
                if phase == 'train':
                    loss.backward()
                    optimizer.step()
               #     scheduler.step()
                # statistics
                running_loss += loss.item() * inputs.size(0) #loss.data[0]
                running_corrects += torch.sum(preds == labels.data)

            epoch_loss = running_loss / dataset_sizes[phase]
            epoch_acc = (100 * running_corrects )/ dataset_sizes[phase]

            print('{} Loss: {:.4f} Acc: {:.4f}%'.format(
                phase, epoch_loss, epoch_acc))

            # deep copy the model
            if phase == 'valid' and epoch_acc > best_acc:
                best_acc = epoch_acc
                best_model_wts = copy.deepcopy(model.state_dict())
                torch.save(model, 'model(frozen_resnet).pkl')
        print()
    time_elapsed = time.time() - since
    print('Training complete in {:.0f}m {:.0f}s'.format(
        time_elapsed // 60, time_elapsed % 60))
    print('Best val Acc: {:4f}'.format(best_acc))
    # load best model weights
    model.load_state_dict(best_model_wts)
    torch.save(model, 'model(frozen123_RES_best).pkl')
    return model


# Finetuning the convnet
if __name__ ==  '__main__':
    model_ft = models.resnet18(pretrained=False)
    pre = torch.load('resnet18-5c106cde.pth')
    model_ft.load_state_dict(pre)
    
    print(model_ft)
    frozen_layers = [model_ft.layer1,model_ft.layer2,model_ft.layer3]
    for layer in frozen_layers:
        for name, value in layer.named_parameters():
            value.requires_grad = False
    params = filter(lambda p: p.requires_grad, model_ft.parameters())
    optimizer_ft = torch.optim.SGD(params, lr=0.001)
    

    num_ftrs = model_ft.fc.in_features
    model_ft.fc = nn.Linear(num_ftrs, 4)
    criterion = nn.CrossEntropyLoss()

    model_ft = train_model(model_ft, criterion, optimizer_ft, exp_lr_scheduler, num_epochs=10)

查看是否冻结成功可以参考这篇文章
查看神经网络的层是否冻结成功