Pytorch 加载部分预训练模型并冻结某些层

目录

1  pytorch的版本：

2  数据下载地址：

3  原始版本代码下载：

4  直接上代码：

---

 

1  pytorch的版本：

2  数据下载地址：

<https://download.pytorch.org/tutorial/hymenoptera_data.zip>

3  原始版本代码下载：

https://pytorch.org/tutorials/_downloads/transfer_learning_tutorial.py

 

4  直接上代码：

# -*- coding: utf-8 -*-
# @File    : test4.py
# @Blog    : https://blog.csdn.net/caomin1hao

from __future__ import print_function, division

import torch
import torch.nn as nn
import torch.optim as optim
from torch.optim import lr_scheduler
import numpy as np
import torchvision
from torchvision import datasets, models, transforms
import matplotlib.pyplot as plt
import time
import os
import copy


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

plt.ion()   # interactive mode




######################################################################
# 1.定义模型，  2.加载部分预训练数据，  3.冻结部分层
######################################
#1.定义模型
model_conv = models.resnet18()
num_ftrs = model_conv.fc.in_features
model_conv.fc = nn.Linear(num_ftrs, 2)

'''
#打印模型的结构
print('###打印模型model_conv的结构####')
print(model_conv)
print('\n')

print('###打印模型model_conv加载参数前的初始值####')
print(list(model_conv.parameters()))
print('\n')
'''

#############################################
#2.加载部分预训练数据
pretrained_dict = torch.load('./08 transfer_learning/resnet18-5c106cde.pth')
'''
for k,v in pretrained_dict.items():
    print(k)
'''
#删除预训练模型跟当前模型层名称相同，层结构却不同的元素;这里有两个'fc.weight'、'fc.bias'
pretrained_dict.pop('fc.weight')
pretrained_dict.pop('fc.bias')

#自己的模型参数变量
model_dict = model_conv.state_dict()
#去除一些不需要的参数
pretrained_dict = {k: v for k, v in pretrained_dict.items() if k in model_dict}

#参数更新
model_dict.update(pretrained_dict)

# 加载我们真正需要的state_dict
model_conv.load_state_dict(model_dict)

'''
print('###打印模型model_conv加载参数后的参数值####')
print(list(model_conv.parameters()))
print('\n')
'''
#############################################
#3.冻结部分层
#将满足条件的参数的 requires_grad 属性设置为False
for name, value in model_conv.named_parameters():
    if (name != 'fc.weight') and (name != 'fc.bias'):
        value.requires_grad = False
'''
#打印各层的requires_grad属性
print('###打印模型model_conv参数的requires_grad属性####')
for name, param in model_conv.named_parameters():
    print(name,param.requires_grad)
'''


# filter 函数将模型中属性 requires_grad = True 的参数选出来
params_conv = filter(lambda p: p.requires_grad, model_conv.parameters())
model_conv = model_conv.to(device)

criterion = nn.CrossEntropyLoss()

# Observe that only parameters of final layer are being optimized as
# opoosed to before.
optimizer_conv = optim.SGD(params_conv, lr=0.001, momentum=0.9)

# Decay LR by a factor of 0.1 every 7 epochs
exp_lr_scheduler = lr_scheduler.StepLR(optimizer_conv, step_size=7, gamma=0.1)


######################################################################
# Training the model
#编写一个通用函数来训练模型。
# 下面将说明： * 调整学习速率 * 保存最好的模型
#下面的参数scheduler是一个来自 torch.optim.lr_scheduler 的学习速率调整类的对象(LR scheduler object)。


def train_model(model, criterion, optimizer, scheduler, num_epochs=25):
    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, num_epochs - 1))
        print('-' * 10)

        # 每个epoch都有一个训练和验证阶段
        for phase in ['train', 'val']:
            if phase == 'train':
                scheduler.step()
                model.train()  # Set model to training mode
            else:
                model.eval()   # Set model to evaluate mode

            running_loss = 0.0
            running_corrects = 0

            #  迭代数据.
            for inputs, labels in dataloaders[phase]:
                inputs = inputs.to(device)
                labels = labels.to(device)

                # zero the parameter gradients
                optimizer.zero_grad()

                # forward
                # track history if only in train
                with torch.set_grad_enabled(phase == 'train'):
                    outputs = model(inputs)
                    _, preds = torch.max(outputs, 1)
                    loss = criterion(outputs, labels)

                    # 后向+仅在训练阶段进行优化
                    if phase == 'train':
                        loss.backward()
                        optimizer.step()

                # statistics
                running_loss += loss.item() * inputs.size(0)
                running_corrects += torch.sum(preds == labels.data)

            epoch_loss = running_loss / dataset_sizes[phase]
            epoch_acc = running_corrects.double() / dataset_sizes[phase]

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

            # 深度复制mo
            if phase == 'val' and epoch_acc > best_acc:
                best_acc = epoch_acc
                best_model_wts = copy.deepcopy(model.state_dict())

        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))

    # 加载最佳模型权重
    model.load_state_dict(best_model_wts)
    return model

######################################################################
# 可视化部分训练图像，以便了解数据扩充。

def imshow(inp, title=None):
    """Imshow for Tensor."""
    inp = inp.numpy().transpose((1, 2, 0))
    mean = np.array([0.485, 0.456, 0.406])
    std = np.array([0.229, 0.224, 0.225])
    inp = std * inp + mean
    inp = np.clip(inp, 0, 1)
    plt.imshow(inp)
    if title is not None:
        plt.title(title)
    plt.pause(0.001)  # pause a bit so that plots are updated



######################################################################
# Visualizing the model predictions
# 一个通用的展示少量预测图片的函数


def visualize_model(model, num_images=6):
    was_training = model.training
    model.eval()
    images_so_far = 0
    fig = plt.figure()

    with torch.no_grad():
        for i, (inputs, labels) in enumerate(dataloaders['val']):
            inputs = inputs.to(device)
            labels = labels.to(device)

            outputs = model(inputs)
            _, preds = torch.max(outputs, 1)

            for j in range(inputs.size()[0]):
                images_so_far += 1
                ax = plt.subplot(num_images//2, 2, images_so_far)
                ax.axis('off')
                ax.set_title('predicted: {}'.format(class_names[preds[j]]))
                imshow(inputs.cpu().data[j])

                if images_so_far == num_images:
                    model.train(mode=was_training)
                    return
        model.train(mode=was_training)

######################################################################
#训练集数据扩充和归一化
#在验证集上仅需要归一化
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])
    ]),
    'val': 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 = './08 transfer_learning/hymenoptera_data'
image_datasets = {x: datasets.ImageFolder(os.path.join(data_dir, x),
                                          data_transforms[x])
                  for x in ['train', 'val']}
dataloaders = {x: torch.utils.data.DataLoader(image_datasets[x], batch_size=4,
                                              shuffle=True, num_workers=4)
               for x in ['train', 'val']}
dataset_sizes = {x: len(image_datasets[x]) for x in ['train', 'val']}
class_names = image_datasets['train'].classes

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

if __name__ == '__main__':


    # Train and evaluate 2
    # 训练模型 在CPU上，与前一个场景相比，这将花费大约一半的时间，因为不需要为大多数网络计算梯度。但需要计算转发。
    model_conv = train_model(model_conv, criterion, optimizer_conv,
                                 exp_lr_scheduler, num_epochs=11)
    visualize_model(model_conv)


    plt.ioff()
    plt.show()

 

部分运行结果：