jaimetang / book-code Goto Github PK

Training...
Batch 500,Model1 Train Loss:0.0951,Model1 Train ACC:96.0000,Model2 Train Loss:0.5438,Model2 Train ACC:82.0000, Blending_Model ACC:96.0000
Batch 1000,Model1 Train Loss:0.0716,Model1 Train ACC:97.0000,Model2 Train Loss:0.4565,Model2 Train ACC:88.0000, Blending_Model ACC:97.0000
Epoch, Model1 Loss:0.0713, Model1 Acc:97.0000%, Model2 Loss:0.4449, Model2 Acc:88.0000%,Blending_Model ACC:97.0000
Validing...
Epoch, Model1 Loss:0.0491, Model1 Acc:98.0000%, Model2 Loss:0.2798, Model2 Acc:96.0000%,Blending_Model ACC:98.0000
Epoch 1/4
Training...
Batch 500,Model1 Train Loss:35.3260,Model1 Train ACC:96.0000,Model2 Train Loss:0.2745,Model2 Train ACC:95.0000, Blending_Model ACC:96.0000
Batch 1000,Model1 Train Loss:nan,Model1 Train ACC:73.0000,Model2 Train Loss:0.2537,Model2 Train ACC:95.0000, Blending_Model ACC:73.0000
Epoch, Model1 Loss:nan, Model1 Acc:71.0000%, Model2 Loss:0.2503, Model2 Acc:95.0000%,Blending_Model ACC:71.0000
Validing...

tensor.sub_(mean[:, None, None]).div_(std[:, None, None])

RuntimeError: output with shape [1, 28, 28] doesn't match the broadcast shape [3, 28, 28]

以上是报错，似乎是mnist数据集是灰度图？然后代码中处理时将均值设置的是对rgb的图像进行的

迁移VGG16部分
running_loss = 0.0
running_corrects = 0
for batch, data in enumerate(dataloader[phase], 1):

    X, y = data
    if Use_gpu:
        X, y = Variable(X.cuda()), Variable(y.cuda())
    else:
        X, y = Variable(X), Variable(y)
        
    y_pred = model(X)
    _, pred = torch.max(y_pred.data, 1)
    optimizer.zero_grad()
    loss = loss_f(y_pred, y)
    
    if phase == "train":
        loss.backward()
        optimizer.step()
        running_loss += loss.data[0]
        running_corrects += torch.sum(pred == y.data)
        
        if batch%500 == 0 and phase =="train":
            print("Batch {}, Train Loss:{:.4f}, Train ACC:{:.4f}".format(batch, running_loss/batch, 100*running_corrects(16*batch)))

epoch_loss = running_loss*16/len(image_datasets[phase])
epoch_acc = 100*running_corrects/len(image_datasets[phase])
print("{} Loss:{:.4f} Acc:{:.4f}%".format(phase, epoch_loss,epoch_acc))

这些行前面全部少了一个缩进

C:\Users\3057.conda\envs\pytorch11\python.exe C:/Users/lucifer/pytorch11/chp8_aut.py
Epoch:50 Style Loss: 881.223694 Content Loss:0.000000
Epoch:100 Style Loss: 881.223694 Content Loss:0.000000
Epoch:150 Style Loss: 881.223694 Content Loss:0.000000
Epoch:200 Style Loss: 881.223694 Content Loss:0.000000
Epoch:250 Style Loss: 881.223694 Content Loss:0.000000
Epoch:300 Style Loss: 881.223694 Content Loss:0.000000

Process finished with exit code 0

如题，问下训练完成后，从哪里输出迁移后的结果？

可以告知下示例代码吗？

第八章风格迁移运行的时候出现这个错误，请问该怎么解决呢？？
Traceback (most recent call last):
File "C:/Users/lucifer/pytorch11/chapter_8.py", line 142, in
optimizer.step(closure)
File "C:\Users\3057.conda\envs\pytorch11\lib\site-packages\torch\optim\lbfgs.py", line 103, in step
orig_loss = closure()
File "C:/Users/lucifer/pytorch11/chapter_8.py", line 133, in closure
for cl in content_loss:
TypeError: 'Content_loss' object is not iterable

Process finished with exit code 1

It seems many code are deprecated in latest version of pytorch. Such as variables

121页
（1） 使用Numpy中的onse可以创建维度指定且元素全为1的数组

onse应该改成ones

第6章代码
loss.backward()
optimizer.step()
running_loss += loss.data[0]
running_correct += torch.sum(pred == y_train.data)

会报错
改成
loss.backward()
optimizer.step()
running_loss += loss.data
running_correct += torch.sum(pred == y_train.data)