deep_residual_network

最新推荐文章于 2023-11-25 21:05:14 发布
最新推荐文章于 2023-11-25 21:05:14 发布
阅读量33 收藏
点赞数
分类专栏： 神经网络与深度学习 Pytorch教程 文章标签： 深度学习 pytorch python ResNet 神经网络
版权声明：本文为博主原创文章，遵循 CC 4.0 BY-SA 版权协议，转载请附上原文出处链接和本声明。
本文链接：https://blog.csdn.net/AIHUBEI/article/details/134605302
版权

Deep_Residual_Network

# Import necessary packages.
import torch
import torch.nn as nn
import torchvision
import torchvision.transforms as transforms

# Device Configuration.
device = torch.device('cuda' if torch.cuda.is_available() else 'cpu')
print(device)

cuda

# Hyper-parameters setting.
num_epochs = 80
batch_size = 100
learning_rate = 0.001

# Image preprocessing modules.
transform = transforms.Compose([
    transforms.Pad(4),
    transforms.RandomHorizontalFlip(),
    transforms.RandomCrop(32),
    transforms.ToTensor()])

# Load The CIFAR-10 dataset.
train_dataset = torchvision.datasets.CIFAR10(root='../../data/',
                                             train=True,
                                             transform=transform,
                                            download=True)

test_dataset = torchvision.datasets.CIFAR10(root='../../data/',
                                            train=False,
                                            transform=transforms.ToTensor(),
                                            download=False)                                             

Files already downloaded and verified

# Define Data Loader.
train_loader = torch.utils.data.DataLoader(dataset=train_dataset,
                                           batch_size=batch_size,
                                           shuffle=True)

test_loader = torch.utils.data.DataLoader(dataset=test_dataset,
                                          batch_size=batch_size,
                                          shuffle=False)

# 3x3 convolution.
def conv3x3(in_channels, out_channels, stride=1):
    return nn.Conv2d(in_channels, out_channels, kernel_size=3,
                     stride=stride, padding=1, bias=False)

# Residual block.
class ResidualBlock(nn.Module):
    def __init__(self, in_channels, out_channels, stride=1, downsample=None):
        super(ResidualBlock, self).__init__()
        self.conv1 = conv3x3(in_channels, out_channels, stride)
        self.bn1 = nn.BatchNorm2d(out_channels)
        self.relu = nn.ReLU(inplace=True)
        self.conv2 = conv3x3(out_channels, out_channels)
        self.bn2 = nn.BatchNorm2d(out_channels)
        self.downsample = downsample
    
    def forward(self, x):
        residual = x
        out = self.conv1(x)
        out = self.bn1(out)
        out = self.relu(out)
        out = self.conv2(out)
        out = self.bn2(out)
        if self.downsample:
            residual = self.downsample(x)
        out += residual
        out = self.relu(out)
        return out

# ResNet.
class ResNet(nn.Module):
    def __init__(self, block, layers, num_classes=10):
        super(ResNet, self).__init__()
        self.in_channels = 16
        self.conv = conv3x3(3, 16)
        self.bn = nn.BatchNorm2d(16)
        self.relu = nn.ReLU(inplace=True)
        self.layer1 = self.make_layer(block, 16, layers[0])
        self.layer2 = self.make_layer(block, 32, layers[1], 2)
        self.layer3 = self.make_layer(block, 64, layers[2], 2)
        self.avg_pool = nn.AvgPool2d(8)
        self.fc = nn.Linear(64, num_classes)

    def make_layer(self, block, out_channels, blocks, stride=1):
        downsample = None
        if (stride != 1) or (self.in_channels != out_channels):
            downsample = nn.Sequential(
                conv3x3(self.in_channels, out_channels, stride=stride),
                nn.BatchNorm2d(out_channels))
        layers = []
        layers.append(block(self.in_channels, out_channels, stride, downsample))
        self.in_channels = out_channels
        for i in range(1, blocks):
            layers.append(block(out_channels, out_channels))
        return nn.Sequential(*layers)
    
    def forward(self, x):
        out = self.conv(x)
        out = self.bn(out)
        out = self.relu(out)
        out = self.layer1(out)
        out = self.layer2(out)
        out = self.layer3(out)
        out = self.avg_pool(out)
        out = out.view((out.size(0)), -1)
        out = self.fc(out)
        return out

model = ResNet(ResidualBlock, [2, 2, 2]).to(device)

# Loss and optimizer.
criterion = nn.CrossEntropyLoss()
optimizer = torch.optim.Adam(model.parameters(), lr=learning_rate)

# For updating learning rate.
def update_lr(optimizer, lr):
    for param_group in optimizer.param_groups:
        param_group['lr'] = lr

# Train the model
total_step = len(train_loader)
curr_lr = learning_rate
for epoch in range(num_epochs):
    for i, (images, labels) in enumerate(train_loader):
        images = images.to(device)
        labels = labels.to(device)

        # Forward pass.
        outputs = model(images)
        loss = criterion(outputs, labels)

        # Backward and optimize.
        optimizer.zero_grad()
        loss.backward()
        optimizer.step()

        # Set an output conunter.
        if (i+1) % 100 == 0:
            print('Epoch [{}/{}], Step [{}/{}], Loss: {:.4f}'
                  .format(epoch+1, num_epochs, i+1, total_step, loss.item()))
        
        # Decay learning rate.
        if (epoch+1) % 20 ==0:
            curr_lr /= 3
            update_lr(optimizer, curr_lr)

Epoch [1/80], Step [100/500], Loss: 1.6813
Epoch [1/80], Step [200/500], Loss: 1.3309
Epoch [1/80], Step [300/500], Loss: 1.3600
Epoch [1/80], Step [400/500], Loss: 1.3821
Epoch [1/80], Step [500/500], Loss: 1.2320
Epoch [2/80], Step [100/500], Loss: 0.9961
Epoch [2/80], Step [200/500], Loss: 1.0894
Epoch [2/80], Step [300/500], Loss: 0.8659
Epoch [2/80], Step [400/500], Loss: 1.0653
Epoch [2/80], Step [500/500], Loss: 0.8834
Epoch [3/80], Step [100/500], Loss: 0.9407
Epoch [3/80], Step [200/500], Loss: 0.9502
Epoch [3/80], Step [300/500], Loss: 0.9662
Epoch [3/80], Step [400/500], Loss: 0.7733
Epoch [3/80], Step [500/500], Loss: 0.6249
Epoch [4/80], Step [100/500], Loss: 0.7448
Epoch [4/80], Step [200/500], Loss: 0.6193
Epoch [4/80], Step [300/500], Loss: 0.6481
Epoch [4/80], Step [400/500], Loss: 0.8609
Epoch [4/80], Step [500/500], Loss: 0.8430
Epoch [5/80], Step [100/500], Loss: 0.8937
Epoch [5/80], Step [200/500], Loss: 0.6558
Epoch [5/80], Step [300/500], Loss: 0.6467
Epoch [5/80], Step [400/500], Loss: 0.8017
Epoch [5/80], Step [500/500], Loss: 0.6142
Epoch [6/80], Step [100/500], Loss: 0.5732
Epoch [6/80], Step [200/500], Loss: 0.6271
Epoch [6/80], Step [300/500], Loss: 0.6341
Epoch [6/80], Step [400/500], Loss: 0.6480
Epoch [6/80], Step [500/500], Loss: 0.6894
Epoch [7/80], Step [100/500], Loss: 0.7035
Epoch [7/80], Step [200/500], Loss: 0.7022
Epoch [7/80], Step [300/500], Loss: 0.4513
Epoch [7/80], Step [400/500], Loss: 0.5571
Epoch [7/80], Step [500/500], Loss: 0.6292
Epoch [8/80], Step [100/500], Loss: 0.4555
Epoch [8/80], Step [200/500], Loss: 0.6679
Epoch [8/80], Step [300/500], Loss: 0.5793
Epoch [8/80], Step [400/500], Loss: 0.5330
Epoch [8/80], Step [500/500], Loss: 0.4378
Epoch [9/80], Step [100/500], Loss: 0.4883
Epoch [9/80], Step [200/500], Loss: 0.5861
Epoch [9/80], Step [300/500], Loss: 0.4735
Epoch [9/80], Step [400/500], Loss: 0.5172
Epoch [9/80], Step [500/500], Loss: 0.4156
Epoch [10/80], Step [100/500], Loss: 0.4031
Epoch [10/80], Step [200/500], Loss: 0.6231
Epoch [10/80], Step [300/500], Loss: 0.4010
Epoch [10/80], Step [400/500], Loss: 0.4843
Epoch [10/80], Step [500/500], Loss: 0.6768
Epoch [11/80], Step [100/500], Loss: 0.4757
Epoch [11/80], Step [200/500], Loss: 0.5781
Epoch [11/80], Step [300/500], Loss: 0.4803
Epoch [11/80], Step [400/500], Loss: 0.4234
Epoch [11/80], Step [500/500], Loss: 0.5142
Epoch [12/80], Step [100/500], Loss: 0.5684
Epoch [12/80], Step [200/500], Loss: 0.3083
Epoch [12/80], Step [300/500], Loss: 0.4764
Epoch [12/80], Step [400/500], Loss: 0.4499
Epoch [12/80], Step [500/500], Loss: 0.5892
Epoch [13/80], Step [100/500], Loss: 0.3176
Epoch [13/80], Step [200/500], Loss: 0.3424
Epoch [13/80], Step [300/500], Loss: 0.5316
Epoch [13/80], Step [400/500], Loss: 0.3902
Epoch [13/80], Step [500/500], Loss: 0.4355
Epoch [14/80], Step [100/500], Loss: 0.3610
Epoch [14/80], Step [200/500], Loss: 0.3949
Epoch [14/80], Step [300/500], Loss: 0.3674
Epoch [14/80], Step [400/500], Loss: 0.5186
Epoch [14/80], Step [500/500], Loss: 0.3936
Epoch [15/80], Step [100/500], Loss: 0.3408
Epoch [15/80], Step [200/500], Loss: 0.4585
Epoch [15/80], Step [300/500], Loss: 0.4356
Epoch [15/80], Step [400/500], Loss: 0.3548
Epoch [15/80], Step [500/500], Loss: 0.4019
Epoch [16/80], Step [100/500], Loss: 0.4717
Epoch [16/80], Step [200/500], Loss: 0.4037
Epoch [16/80], Step [300/500], Loss: 0.4970
Epoch [16/80], Step [400/500], Loss: 0.4643
Epoch [16/80], Step [500/500], Loss: 0.4482
Epoch [17/80], Step [100/500], Loss: 0.4007
Epoch [17/80], Step [200/500], Loss: 0.5095
Epoch [17/80], Step [300/500], Loss: 0.2736
Epoch [17/80], Step [400/500], Loss: 0.4718
Epoch [17/80], Step [500/500], Loss: 0.3577
Epoch [18/80], Step [100/500], Loss: 0.4523
Epoch [18/80], Step [200/500], Loss: 0.3902
Epoch [18/80], Step [300/500], Loss: 0.2361
Epoch [18/80], Step [400/500], Loss: 0.2933
Epoch [18/80], Step [500/500], Loss: 0.2130
Epoch [19/80], Step [100/500], Loss: 0.3324
Epoch [19/80], Step [200/500], Loss: 0.3762
Epoch [19/80], Step [300/500], Loss: 0.2952
Epoch [19/80], Step [400/500], Loss: 0.3768
Epoch [19/80], Step [500/500], Loss: 0.2550
Epoch [20/80], Step [100/500], Loss: 0.4499
Epoch [20/80], Step [200/500], Loss: 0.4516
Epoch [20/80], Step [300/500], Loss: 0.3245
Epoch [20/80], Step [400/500], Loss: 0.3915
Epoch [20/80], Step [500/500], Loss: 0.2645
Epoch [21/80], Step [100/500], Loss: 0.2207
Epoch [21/80], Step [200/500], Loss: 0.4940
Epoch [21/80], Step [300/500], Loss: 0.4244
Epoch [21/80], Step [400/500], Loss: 0.4214
Epoch [21/80], Step [500/500], Loss: 0.2739
Epoch [22/80], Step [100/500], Loss: 0.2975
Epoch [22/80], Step [200/500], Loss: 0.2993
Epoch [22/80], Step [300/500], Loss: 0.3349
Epoch [22/80], Step [400/500], Loss: 0.4027
Epoch [22/80], Step [500/500], Loss: 0.4266
Epoch [23/80], Step [100/500], Loss: 0.3436
Epoch [23/80], Step [200/500], Loss: 0.3375
Epoch [23/80], Step [300/500], Loss: 0.3030
Epoch [23/80], Step [400/500], Loss: 0.3841
Epoch [23/80], Step [500/500], Loss: 0.4496
Epoch [24/80], Step [100/500], Loss: 0.5373
Epoch [24/80], Step [200/500], Loss: 0.2922
Epoch [24/80], Step [300/500], Loss: 0.4524
Epoch [24/80], Step [400/500], Loss: 0.2701
Epoch [24/80], Step [500/500], Loss: 0.3375
Epoch [25/80], Step [100/500], Loss: 0.4769
Epoch [25/80], Step [200/500], Loss: 0.3759
Epoch [25/80], Step [300/500], Loss: 0.4362
Epoch [25/80], Step [400/500], Loss: 0.3946
Epoch [25/80], Step [500/500], Loss: 0.4517
Epoch [26/80], Step [100/500], Loss: 0.3163
Epoch [26/80], Step [200/500], Loss: 0.2814
Epoch [26/80], Step [300/500], Loss: 0.2830
Epoch [26/80], Step [400/500], Loss: 0.3459
Epoch [26/80], Step [500/500], Loss: 0.3376
Epoch [27/80], Step [100/500], Loss: 0.5180
Epoch [27/80], Step [200/500], Loss: 0.3052
Epoch [27/80], Step [300/500], Loss: 0.2578
Epoch [27/80], Step [400/500], Loss: 0.4357
Epoch [27/80], Step [500/500], Loss: 0.3561
Epoch [28/80], Step [100/500], Loss: 0.2215
Epoch [28/80], Step [200/500], Loss: 0.4192
Epoch [28/80], Step [300/500], Loss: 0.3859
Epoch [28/80], Step [400/500], Loss: 0.2494
Epoch [28/80], Step [500/500], Loss: 0.3063
Epoch [29/80], Step [100/500], Loss: 0.2807
Epoch [29/80], Step [200/500], Loss: 0.3785
Epoch [29/80], Step [300/500], Loss: 0.2530
Epoch [29/80], Step [400/500], Loss: 0.2523
Epoch [29/80], Step [500/500], Loss: 0.2863
Epoch [30/80], Step [100/500], Loss: 0.3900
Epoch [30/80], Step [200/500], Loss: 0.4359
Epoch [30/80], Step [300/500], Loss: 0.4164
Epoch [30/80], Step [400/500], Loss: 0.3966
Epoch [30/80], Step [500/500], Loss: 0.4372
Epoch [31/80], Step [100/500], Loss: 0.3702
Epoch [31/80], Step [200/500], Loss: 0.4615
Epoch [31/80], Step [300/500], Loss: 0.3059
Epoch [31/80], Step [400/500], Loss: 0.3323
Epoch [31/80], Step [500/500], Loss: 0.2572
Epoch [32/80], Step [100/500], Loss: 0.2867
Epoch [32/80], Step [200/500], Loss: 0.6127
Epoch [32/80], Step [300/500], Loss: 0.2086
Epoch [32/80], Step [400/500], Loss: 0.2994
Epoch [32/80], Step [500/500], Loss: 0.3686
Epoch [33/80], Step [100/500], Loss: 0.5084
Epoch [33/80], Step [200/500], Loss: 0.2910
Epoch [33/80], Step [300/500], Loss: 0.5251
Epoch [33/80], Step [400/500], Loss: 0.3583
Epoch [33/80], Step [500/500], Loss: 0.2018
Epoch [34/80], Step [100/500], Loss: 0.3070
Epoch [34/80], Step [200/500], Loss: 0.4379
Epoch [34/80], Step [300/500], Loss: 0.3893
Epoch [34/80], Step [400/500], Loss: 0.3778
Epoch [34/80], Step [500/500], Loss: 0.4583
Epoch [35/80], Step [100/500], Loss: 0.3852
Epoch [35/80], Step [200/500], Loss: 0.2713
Epoch [35/80], Step [300/500], Loss: 0.5001
Epoch [35/80], Step [400/500], Loss: 0.2437
Epoch [35/80], Step [500/500], Loss: 0.3665
Epoch [36/80], Step [100/500], Loss: 0.4438
Epoch [36/80], Step [200/500], Loss: 0.2112
Epoch [36/80], Step [300/500], Loss: 0.3104
Epoch [36/80], Step [400/500], Loss: 0.3022
Epoch [36/80], Step [500/500], Loss: 0.2445
Epoch [37/80], Step [100/500], Loss: 0.3935
Epoch [37/80], Step [200/500], Loss: 0.3391
Epoch [37/80], Step [300/500], Loss: 0.5324
Epoch [37/80], Step [400/500], Loss: 0.4255
Epoch [37/80], Step [500/500], Loss: 0.3942
Epoch [38/80], Step [100/500], Loss: 0.2929
Epoch [38/80], Step [200/500], Loss: 0.3554
Epoch [38/80], Step [300/500], Loss: 0.2800
Epoch [38/80], Step [400/500], Loss: 0.3467
Epoch [38/80], Step [500/500], Loss: 0.3648
Epoch [39/80], Step [100/500], Loss: 0.2709
Epoch [39/80], Step [200/500], Loss: 0.2959
Epoch [39/80], Step [300/500], Loss: 0.2654
Epoch [39/80], Step [400/500], Loss: 0.3282
Epoch [39/80], Step [500/500], Loss: 0.3584
Epoch [40/80], Step [100/500], Loss: 0.3200
Epoch [40/80], Step [200/500], Loss: 0.4697
Epoch [40/80], Step [300/500], Loss: 0.5168
Epoch [40/80], Step [400/500], Loss: 0.3246
Epoch [40/80], Step [500/500], Loss: 0.4181
Epoch [41/80], Step [100/500], Loss: 0.5672
Epoch [41/80], Step [200/500], Loss: 0.3359
Epoch [41/80], Step [300/500], Loss: 0.4046
Epoch [41/80], Step [400/500], Loss: 0.2741
Epoch [41/80], Step [500/500], Loss: 0.3885
Epoch [42/80], Step [100/500], Loss: 0.3990
Epoch [42/80], Step [200/500], Loss: 0.3618
Epoch [42/80], Step [300/500], Loss: 0.2712
Epoch [42/80], Step [400/500], Loss: 0.3771
Epoch [42/80], Step [500/500], Loss: 0.3654
Epoch [43/80], Step [100/500], Loss: 0.4265
Epoch [43/80], Step [200/500], Loss: 0.3484
Epoch [43/80], Step [300/500], Loss: 0.2833
Epoch [43/80], Step [400/500], Loss: 0.3779
Epoch [43/80], Step [500/500], Loss: 0.2916
Epoch [44/80], Step [100/500], Loss: 0.4146
Epoch [44/80], Step [200/500], Loss: 0.3840
Epoch [44/80], Step [300/500], Loss: 0.3005
Epoch [44/80], Step [400/500], Loss: 0.3404
Epoch [44/80], Step [500/500], Loss: 0.4090
Epoch [45/80], Step [100/500], Loss: 0.4705
Epoch [45/80], Step [200/500], Loss: 0.4325
Epoch [45/80], Step [300/500], Loss: 0.4477
Epoch [45/80], Step [400/500], Loss: 0.3278
Epoch [45/80], Step [500/500], Loss: 0.5257
Epoch [46/80], Step [100/500], Loss: 0.2231
Epoch [46/80], Step [200/500], Loss: 0.3204
Epoch [46/80], Step [300/500], Loss: 0.4188
Epoch [46/80], Step [400/500], Loss: 0.3421
Epoch [46/80], Step [500/500], Loss: 0.2937
Epoch [47/80], Step [100/500], Loss: 0.4077
Epoch [47/80], Step [200/500], Loss: 0.4088
Epoch [47/80], Step [300/500], Loss: 0.4501
Epoch [47/80], Step [400/500], Loss: 0.3707
Epoch [47/80], Step [500/500], Loss: 0.2656
Epoch [48/80], Step [100/500], Loss: 0.2942
Epoch [48/80], Step [200/500], Loss: 0.3595
Epoch [48/80], Step [300/500], Loss: 0.3710
Epoch [48/80], Step [400/500], Loss: 0.3372
Epoch [48/80], Step [500/500], Loss: 0.5289
Epoch [49/80], Step [100/500], Loss: 0.3020
Epoch [49/80], Step [200/500], Loss: 0.3487
Epoch [49/80], Step [300/500], Loss: 0.4004
Epoch [49/80], Step [400/500], Loss: 0.4134
Epoch [49/80], Step [500/500], Loss: 0.4896
Epoch [50/80], Step [100/500], Loss: 0.4403
Epoch [50/80], Step [200/500], Loss: 0.4179
Epoch [50/80], Step [300/500], Loss: 0.2329
Epoch [50/80], Step [400/500], Loss: 0.3027
Epoch [50/80], Step [500/500], Loss: 0.4491
Epoch [51/80], Step [100/500], Loss: 0.4268
Epoch [51/80], Step [200/500], Loss: 0.3180
Epoch [51/80], Step [300/500], Loss: 0.2959
Epoch [51/80], Step [400/500], Loss: 0.2462
Epoch [51/80], Step [500/500], Loss: 0.3458
Epoch [52/80], Step [100/500], Loss: 0.3443
Epoch [52/80], Step [200/500], Loss: 0.2158
Epoch [52/80], Step [300/500], Loss: 0.2663
Epoch [52/80], Step [400/500], Loss: 0.3248
Epoch [52/80], Step [500/500], Loss: 0.3340
Epoch [53/80], Step [100/500], Loss: 0.4008
Epoch [53/80], Step [200/500], Loss: 0.4801
Epoch [53/80], Step [300/500], Loss: 0.4733
Epoch [53/80], Step [400/500], Loss: 0.4176
Epoch [53/80], Step [500/500], Loss: 0.3105
Epoch [54/80], Step [100/500], Loss: 0.3254
Epoch [54/80], Step [200/500], Loss: 0.3082
Epoch [54/80], Step [300/500], Loss: 0.2672
Epoch [54/80], Step [400/500], Loss: 0.3272
Epoch [54/80], Step [500/500], Loss: 0.3505
Epoch [55/80], Step [100/500], Loss: 0.3498
Epoch [55/80], Step [200/500], Loss: 0.4057
Epoch [55/80], Step [300/500], Loss: 0.3522
Epoch [55/80], Step [400/500], Loss: 0.4080
Epoch [55/80], Step [500/500], Loss: 0.5481
Epoch [56/80], Step [100/500], Loss: 0.3302
Epoch [56/80], Step [200/500], Loss: 0.4734
Epoch [56/80], Step [300/500], Loss: 0.3435
Epoch [56/80], Step [400/500], Loss: 0.3337
Epoch [56/80], Step [500/500], Loss: 0.4469
Epoch [57/80], Step [100/500], Loss: 0.3973
Epoch [57/80], Step [200/500], Loss: 0.2889
Epoch [57/80], Step [300/500], Loss: 0.2850
Epoch [57/80], Step [400/500], Loss: 0.3346
Epoch [57/80], Step [500/500], Loss: 0.2428
Epoch [58/80], Step [100/500], Loss: 0.3499
Epoch [58/80], Step [200/500], Loss: 0.3708
Epoch [58/80], Step [300/500], Loss: 0.3457
Epoch [58/80], Step [400/500], Loss: 0.3043
Epoch [58/80], Step [500/500], Loss: 0.2127
Epoch [59/80], Step [100/500], Loss: 0.4164
Epoch [59/80], Step [200/500], Loss: 0.5447
Epoch [59/80], Step [300/500], Loss: 0.3270
Epoch [59/80], Step [400/500], Loss: 0.2835
Epoch [59/80], Step [500/500], Loss: 0.3678
Epoch [60/80], Step [100/500], Loss: 0.5183
Epoch [60/80], Step [200/500], Loss: 0.3077
Epoch [60/80], Step [300/500], Loss: 0.3788
Epoch [60/80], Step [400/500], Loss: 0.5352
Epoch [60/80], Step [500/500], Loss: 0.2007
Epoch [61/80], Step [100/500], Loss: 0.3592
Epoch [61/80], Step [200/500], Loss: 0.3335
Epoch [61/80], Step [300/500], Loss: 0.3052
Epoch [61/80], Step [400/500], Loss: 0.3914
Epoch [61/80], Step [500/500], Loss: 0.3509
Epoch [62/80], Step [100/500], Loss: 0.5136
Epoch [62/80], Step [200/500], Loss: 0.3592
Epoch [62/80], Step [300/500], Loss: 0.2378
Epoch [62/80], Step [400/500], Loss: 0.2732
Epoch [62/80], Step [500/500], Loss: 0.3470
Epoch [63/80], Step [100/500], Loss: 0.3089
Epoch [63/80], Step [200/500], Loss: 0.3117
Epoch [63/80], Step [300/500], Loss: 0.4159
Epoch [63/80], Step [400/500], Loss: 0.4553
Epoch [63/80], Step [500/500], Loss: 0.3710
Epoch [64/80], Step [100/500], Loss: 0.4569
Epoch [64/80], Step [200/500], Loss: 0.4581
Epoch [64/80], Step [300/500], Loss: 0.3520
Epoch [64/80], Step [400/500], Loss: 0.3495
Epoch [64/80], Step [500/500], Loss: 0.3255
Epoch [65/80], Step [100/500], Loss: 0.2694
Epoch [65/80], Step [200/500], Loss: 0.2800
Epoch [65/80], Step [300/500], Loss: 0.2462
Epoch [65/80], Step [400/500], Loss: 0.2442
Epoch [65/80], Step [500/500], Loss: 0.3629
Epoch [66/80], Step [100/500], Loss: 0.4210
Epoch [66/80], Step [200/500], Loss: 0.3035
Epoch [66/80], Step [300/500], Loss: 0.5971
Epoch [66/80], Step [400/500], Loss: 0.3247
Epoch [66/80], Step [500/500], Loss: 0.4722
Epoch [67/80], Step [100/500], Loss: 0.3194
Epoch [67/80], Step [200/500], Loss: 0.3479
Epoch [67/80], Step [300/500], Loss: 0.4012
Epoch [67/80], Step [400/500], Loss: 0.3244
Epoch [67/80], Step [500/500], Loss: 0.3968
Epoch [68/80], Step [100/500], Loss: 0.4607
Epoch [68/80], Step [200/500], Loss: 0.2946
Epoch [68/80], Step [300/500], Loss: 0.5424
Epoch [68/80], Step [400/500], Loss: 0.3642
Epoch [68/80], Step [500/500], Loss: 0.3861
Epoch [69/80], Step [100/500], Loss: 0.4021
Epoch [69/80], Step [200/500], Loss: 0.3440
Epoch [69/80], Step [300/500], Loss: 0.2665
Epoch [69/80], Step [400/500], Loss: 0.3696
Epoch [69/80], Step [500/500], Loss: 0.4682
Epoch [70/80], Step [100/500], Loss: 0.3457
Epoch [70/80], Step [200/500], Loss: 0.3573
Epoch [70/80], Step [300/500], Loss: 0.2991
Epoch [70/80], Step [400/500], Loss: 0.4386
Epoch [70/80], Step [500/500], Loss: 0.3190
Epoch [71/80], Step [100/500], Loss: 0.4371
Epoch [71/80], Step [200/500], Loss: 0.4293
Epoch [71/80], Step [300/500], Loss: 0.4279
Epoch [71/80], Step [400/500], Loss: 0.4478
Epoch [71/80], Step [500/500], Loss: 0.2983
Epoch [72/80], Step [100/500], Loss: 0.3917
Epoch [72/80], Step [200/500], Loss: 0.3143
Epoch [72/80], Step [300/500], Loss: 0.2839
Epoch [72/80], Step [400/500], Loss: 0.4507
Epoch [72/80], Step [500/500], Loss: 0.3409
Epoch [73/80], Step [100/500], Loss: 0.2883
Epoch [73/80], Step [200/500], Loss: 0.2914
Epoch [73/80], Step [300/500], Loss: 0.3873
Epoch [73/80], Step [400/500], Loss: 0.3090
Epoch [73/80], Step [500/500], Loss: 0.2304
Epoch [74/80], Step [100/500], Loss: 0.3744
Epoch [74/80], Step [200/500], Loss: 0.3707
Epoch [74/80], Step [300/500], Loss: 0.3504
Epoch [74/80], Step [400/500], Loss: 0.3842
Epoch [74/80], Step [500/500], Loss: 0.2762
Epoch [75/80], Step [100/500], Loss: 0.3845
Epoch [75/80], Step [200/500], Loss: 0.2869
Epoch [75/80], Step [300/500], Loss: 0.4739
Epoch [75/80], Step [400/500], Loss: 0.2564
Epoch [75/80], Step [500/500], Loss: 0.4151
Epoch [76/80], Step [100/500], Loss: 0.2882
Epoch [76/80], Step [200/500], Loss: 0.2850
Epoch [76/80], Step [300/500], Loss: 0.4033
Epoch [76/80], Step [400/500], Loss: 0.3802
Epoch [76/80], Step [500/500], Loss: 0.2550
Epoch [77/80], Step [100/500], Loss: 0.3178
Epoch [77/80], Step [200/500], Loss: 0.3310
Epoch [77/80], Step [300/500], Loss: 0.3325
Epoch [77/80], Step [400/500], Loss: 0.3344
Epoch [77/80], Step [500/500], Loss: 0.2777
Epoch [78/80], Step [100/500], Loss: 0.3008
Epoch [78/80], Step [200/500], Loss: 0.5229
Epoch [78/80], Step [300/500], Loss: 0.2749
Epoch [78/80], Step [400/500], Loss: 0.3293
Epoch [78/80], Step [500/500], Loss: 0.3426
Epoch [79/80], Step [100/500], Loss: 0.2453
Epoch [79/80], Step [200/500], Loss: 0.3240
Epoch [79/80], Step [300/500], Loss: 0.3560
Epoch [79/80], Step [400/500], Loss: 0.2577
Epoch [79/80], Step [500/500], Loss: 0.3159
Epoch [80/80], Step [100/500], Loss: 0.2661
Epoch [80/80], Step [200/500], Loss: 0.3118
Epoch [80/80], Step [300/500], Loss: 0.2826
Epoch [80/80], Step [400/500], Loss: 0.4284
Epoch [80/80], Step [500/500], Loss: 0.3187

# Test the model.
model.eval()
with torch.no_grad():
    correct = 0
    total = 0
    for images, labels in test_loader:
        images = images.to(device)
        labels = labels.to(device)
        outputs = model(images)
        _, predicted = torch.max(outputs.data, 1)
        total += labels.size(0)
        correct += (predicted == labels).sum().item()

    print("Accuracy of the model no the test images: {} %".format(100 * correct / total))

Accuracy of the model no the test images: 84.61 %

# Save the model checkpoint.
torch.save(model.state_dict(), 'model_param.ckpt')

# Load the model checkpoint.
model.load_state_dict(torch.load('model_param.ckpt'))

<All keys matched successfully>
AIHUBEI
关注
  • 0
    点赞
  • 踩
  • 0
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
专栏目录
DRPNN-Deep-Residual-Pan-sharpening-Neural-Network
05-18
DRPNN-深残留潘锐化神经网络这是期刊文章中提出的DRPNN的源代码： 魏Y，袁Q，沉H和张L，“通过学习深度残差网络提高多光谱图像泛锐化的准确性”，在IEEE地质科学和遥感快报中，第1卷。 14号2017年10月，第10卷，第...
A Deep Network with Composite Residual Structure for Handwritten Character Recognition
02-07
A Deep Network with Composite Residual Structure for Handwritten Character Recognition
The deep residual shrinkage network is a variant of deep r.zip
最新发布
04-25
这里为你收集整理了关于AI，机器学习，深度学习相关的资料一份，质量非常高，如果你投入时间去研究几天相信肯定对你有很大的帮助。到时候你会回来感谢我的。 本资源是经过本地编译测试、可打开、可运行的文件或源码...
Wide-Residual-Networks:凯拉斯的广泛残留网络
05-10
可以通过导入wide_residial_network脚本并使用create_wide_residual_network（）方法来使用它。 可以更改几个参数以增加网络的深度或宽度。 请注意，可以通过以下公式计算层数： nb_layers = 4 + 6 * N 因此，N可以...
Residual dense network for intensity-guided depth map enhancement
02-08
Recently, by introducing the guidance from the color image, deep convolutional neural network (DCNN) shows significant improvements for depth map enhancement. However, most DCNN-based methods do not ...
2020年Tensorflow2-GPU 完整安装过程总结
热门推荐
AIHUBEI的博客
03-23 1万+
2020年Tensorflow2-GPU 完整安装过程总结 Tensorflow2.1将是最后一个支持python2的版本。tensorflow2.1增加了对TPU的支持，同时tf.keras和tf.data这两个常用的API也得到了很多新的更新。Tensorflow2.1的cuda版本为10.1，cuDNN版本为7.6. Windows 和 Linux 系统的 TensorFlow pip 版本...
nltk.download(‘stopwords‘)报错，解决方案
AIHUBEI的博客
08-12 9153
nltk.download(‘stopwords’)报错，解决方案如下 ＃报错信息如下 [nltk_data] Error loading stopwords: <urlopen error [Errno 11004] [nltk_data] getaddrinfo failed> False 1.解决方案 手动下载数据，并放在指定位置 下载地址：https://files-cdn.cnblogs.com/files/douzujun/stopwords.zip 下载解压之后放在哪里
神经网络&波士顿房价预测
AIHUBEI的博客
02-25 8935
使用Numpy构建神经网络 百度免费课程学习笔记 本节将使用Python语言和Numpy库来构建神经网络模型，向读者展示神经网络的基本概念和工作过程。 构建神经网络/深度学习模型的基本步骤 如之前的介绍，应用于不同场景的深度学习模型具备一定的通用性，均可以从下述五个步骤来完成模型的构建和训练。 数据处理：从本地文件或网络地址读取数据，并做预处理操作，如校验数据的正确性等。 模型设计：完成网络结构...
多层感知机、常见激活函数
AIHUBEI的博客
06-17 3641
3.8 多层感知机 xiaoyao 动手学深度学习 tensorflow2 前面学习了线性回归和softmax回归在内的单层神经网络。然而深度学习主要关注多层模型。在本节中，将学习多层感知机（multilayer perceptron，MLP）。 3.8.1 隐藏层 多层感知机在单层神经网络的基础上引入了一个或多个隐藏层（hidden layer).隐藏层位于输入层和输出层之间。 如下图所示： 上图所示的多层感知机中，输入和输出个数分别为4和3，中间的隐藏层中包含了5个隐藏单元（hidden unit）
利用Tensorflow2构建CNN并用于图像分类
AIHUBEI的博客
06-17 2247
利用Tensorflow2构建CNN并用于图像分类 本文将使用自下而上的方法，从讨论CNN的基本板块构建开始。然后将深入研究CNN的体系结构，并探索如何在TensorFlow中实现CNN。本文包含以下内容： 一维和二维的卷积运算 卷积神经网络结构的构建 使用Tensorflow构建深度卷积神经网络 利用数据扩充技术提高模型泛化性能 实现一个基于人脸图像的CNN分类器来预测个人性别 from IPython.display import Image %matplotlib inline
利用Tensorflow构建生成对抗网络GAN以生成数据
AIHUBEI的博客
07-26 2135
使用生成对抗网络(GAN)生成数据 本文主要内容 介绍了自动编码器的基本原理 比较了生成模型与自动编码器的区别 描述了GAN模型的网络结构 分析了GAN模型的目标核函数以及训练过程 介绍了利用Google Colab进行模型训练的基本步骤 设计并实现了简单的GAN网络，进行了网络训练以及模型评估 from IPython.display import Image %matplotlib inline 生成对抗网络的简单介绍 首先从自动编码器开始–autoencoders Ima
tensorflow数据操作
AIHUBEI的博客
03-23 1644
tensorflow数据操作 xiaoyao 动手学深度学习，tensorflow代码 import tensorflow as tf print(tf.__version__) 2.1.0 在深度学习中，我们通常会频繁地对数据进行操作。作为动手学深度学习的基础，本节将介绍如何对内存中的数据进行操作。 在tensorflow中，tensor是一个类，也是存储和变换数据的主要工具。如果你之前用过...
基于WSL2+Docker+VScode搭建机器学习(深度学习)开发环境
AIHUBEI的博客
11-25 1523
基于WSL2+Docker+VScode搭建机器学习(深度学习)开发环境。
神经网络常用激活函数
AIHUBEI的博客
02-22 1488
神经网络常用激活函数 激活函数：激活函数在神经元中很重要，通常为了增强网络的表示能力和学习能力，激活函数需要具备的性质如下: \qquad（1）连续可导（但也允许少数点上不可导）的非线性函数，其中，可导的激活函数可以直接利用数值优化的方法来学习网络参数。 \qquad（2）激活函数及其导函数要尽可能的简单，这样有利于提高网络的计算效率。 \qquad（3）激活函数的导函数值域要在一个合适的区间内，...
神经网络漫谈(一)：科普篇
AIHUBEI的博客
01-06 1219
本文简要介绍了神经网络训练学习的基本过程，并以图像分类为例，介绍了模型权重更新，模型推理的基本流程。作为神经网络漫谈系列的第一章。
从头开始实现多层人工神经网络
AIHUBEI的博客
05-23 1172
从头开始实现多层人工神经网络 from IPython.display import Image %matplotlib inline 1.使用人工神经网络建模复杂函数 神经网络中的Epoch、Iteration、Batchsize 神经网络中epoch与iteration是不相等的 batchsize：中文翻译为批大小（批尺寸）。在深度学习中，一般采用SGD训练，即每次训练在训练集中取batchsize个样本训练； iteration：中文翻译为迭代，1个iteration等于使用batchsi
利用Tensorflow构建RNN并对序列数据进行建模
AIHUBEI的博客
07-10 973
利用Tensorflow构建RNN并对序列数据进行建模 对文本处理处理任务的方法中，一般将TF-IDF向量作为特征输入。显然的缺陷是：这种方法丢失了输入的文本序列中每个单词的顺序。 对一般的前馈神经网络，比如CNN，通常接受一个定长的向量作为输入。 CNN对文本数据建模的时候，输入变长的字符串或者单词串，然后通过滑动窗口加上池化的方式将原先的输入转换成为一个固定长度的向量表示， 这样做可以捕捉到原始文本中的一些局部特征，但是对于两个单词之间的依赖关系则很难被捕获学习到。 循环神经网络RNN能很好地处理文本数
深度卷积神经网络（AlexNet）--Tensorflow实现
AIHUBEI的博客
09-27 947
(9月27号(组内)–d2l)深度卷积神经网络（AlexNet） ????sec_alexnet 在LeNet提出后，卷积神经网络在计算机视觉和机器学习领域中很有名气。但卷积神经网络并没有主导这些领域。这是因为虽然 LeNet 在小数据集上取得了很好的效果，但是在更大、更真实的数据集上训练卷积神经网络的性能和可行性还有待研究。事实上，在上世纪90年代初到2012年之间的大部分时间里，神经网络往往被其他机器学习方法超越，如支持向量机（support vector machines）。 在计算机视觉中，直接将神
利用转置卷积与WGAN提升图像数据生成的质量
AIHUBEI的博客
07-27 875
通过使用卷积与Wasserstein GAN提升生成图像的质量 本文将实现deep convolutional Gan（DCGAN），同时也将实现Wasserstein GAN（WGAN）。 本文使用到的一些技术： 转置卷积Transposed convolution 批量归一化Batch Normalization WGAN Gradient penalty from IPython.display import Image %matplotlib inline 转置卷积 虽然卷积运
deep residual network
04-11
b'deep residual network'是一种深度神经网络结构，它采用了残差学习的思想，通过引入跨层的连接和残差块（residual block）来减少梯度消失和梯度爆炸问题，从而能够训练深度网络。该结构在图像识别、目标检测、语音...

“相关推荐”对你有帮助么？

  • 非常没帮助
  • 没帮助
  • 一般
  • 有帮助
  • 非常有帮助
提交
评论
添加红包

请填写红包祝福语或标题

个

红包个数最小为10个

元

红包金额最低5元

当前余额3.43元 前往充值 >
需支付：10.00元
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付元
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明：

1.余额是钱包充值的虚拟货币，按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载，可以购买VIP、付费专栏及课程。

余额充值