About loss visualization

[ChenSiwiI]

I would like to ask if there is a txt file to save the loss during training and testing, and can it be visualized

[cuteboyqq]

l update the train.py code 2023-04-22 12:50 updated :

1. add log.txt file
   • include nc, model information
   • include Dloss and Gloss and auc and lr at each epoch

[ChenSiwiI]

l update the train.py code 2023-04-22 12:50 updated :

1. add log.txt file

   • include nc, model information
   • include Dloss and Gloss and auc and lr at each epoch

thanks！

[cuteboyqq]

In def train_epochs(model,train_loader,test_loader,args): at train.py , try to implement draw plot of auc and loss at each epochs, but failed...

#============================
# Initialize list for draw plots
#============================
GlossTr_list=[] #New add code
auc_val_list=[] #New add code
epoches = [] #New add code
#==================================
for epoch in range(1, args.epoch+1):
    train_loss = 0.0
    G_loss = 0
    D_loss = 0
    ....
    GlossTr_list.append(avg_g_loss)   #New add code
    epoches.append(epoch) #New add code
    ...

   #calculate auc
    auc = model.test(test_loader)
    auc_val_list.append(auc)  #New add code
    print('auc = {:.6f}'.format(auc))
    #==========================================================
    # Below code implement : Plot the figures at each epochs (New add code)
    #=============================================================
    fig1, ax1 = plt.subplots(figsize=(11, 8))

    ax1.plot(epoches, GlossTr_list)
    ax1.set_title("Average training Gloss vs epochs")
    ax1.set_xlabel("Epochs")
    ax1.set_ylabel("Current Gloss")
    print(args.save_dir)
    plt.savefig(args.save_dir,"loss_vs_epochs.png")

    plt.clf()

    fig2, ax2 = plt.subplots(figsize=(11, 8))

    ax2.plot(epoches, auc_val_list, label="Val auc")
    ax2.set_title("Average AUC vs epochs")
    ax2.set_xlabel("Epochs")
    ax2.set_ylabel("Current AUC")
    plt.legend(loc='lower right')

    plt.savefig(args.save_dir + "auc_vs_epochs.png")

    plt.close('all')

up code try to draw plot of loss-epoch and auc-epoch, but have errors, l also update this code, but noted (do not use now) sorry

[phonver]

l update the train.py code 2023-04-22 12:50 updated :l更新train.py代码2023-04-22 12：50更新：

1. add log.txt file 添加log.txt文件

   • include nc, model information 包括NC、模型信息
   • include Dloss and Gloss and auc and lr at each epoch包括在每个时期Dloss和Gloss以及auc和lr

thanks！ 谢谢！

In def train_epochs(model,train_loader,test_loader,args): at train.py , try to implement draw plot of auc and loss at each epochs, but failed...在def train_epochs（model，train_loader，test_loader，args）中：在train.py，尝试实现绘制每个epoch的auc和loss的图，但失败了...

#============================
# Initialize list for draw plots
#============================
GlossTr_list=[] #New add code
auc_val_list=[] #New add code
epoches = [] #New add code
#==================================
for epoch in range(1, args.epoch+1):
    train_loss = 0.0
    G_loss = 0
    D_loss = 0
    ....
    GlossTr_list.append(avg_g_loss)   #New add code
    epoches.append(epoch) #New add code
    ...

   #calculate auc
    auc = model.test(test_loader)
    auc_val_list.append(auc)  #New add code
    print('auc = {:.6f}'.format(auc))
    #==========================================================
    # Below code implement : Plot the figures at each epochs (New add code)
    #=============================================================
    fig1, ax1 = plt.subplots(figsize=(11, 8))

    ax1.plot(epoches, GlossTr_list)
    ax1.set_title("Average training Gloss vs epochs")
    ax1.set_xlabel("Epochs")
    ax1.set_ylabel("Current Gloss")
    print(args.save_dir)
    plt.savefig(args.save_dir,"loss_vs_epochs.png")

    plt.clf()

    fig2, ax2 = plt.subplots(figsize=(11, 8))

    ax2.plot(epoches, auc_val_list, label="Val auc")
    ax2.set_title("Average AUC vs epochs")
    ax2.set_xlabel("Epochs")
    ax2.set_ylabel("Current AUC")
    plt.legend(loc='lower right')

    plt.savefig(args.save_dir + "auc_vs_epochs.png")

    plt.close('all')

up code try to draw plot of loss-epoch and auc-epoch, but have errors, l also update this code, but noted (do not use now) sorry上代码尝试绘制损失时期和auc-epoch图，但有错误，我也更新了此代码，但注意到（现在不使用）抱歉

Thank you. But each test will encounter the following situations: 3 Start normal AE: loss : 2.254371166229248 24 Start normal AE: loss : 5.585793495178223 25 Start normal AE: loss : 4.096798419952393 26 Start normal AE: loss : 88.05529022216797 Traceback (most recent call last): File "F:/Skip-CBAM-SelfAttention-GANomaly-Pytorch-main/test.py", line 212, in main() File "F:/Skip-CBAM-SelfAttention-GANomaly-Pytorch-main/test.py", line 47, in main test(args) File "F:/Skip-CBAM-SelfAttention-GANomaly-Pytorch-main/test.py", line 84, in test 'normal',device,args) File "F:/Skip-CBAM-SelfAttention-GANomaly-Pytorch-main/test.py", line 146, in infer images, labels = next(dataiter) File "G:\Users\Administrator\anaconda3\envs\ganomaly\lib\site-packages\torch\utils\data\dataloader.py", line 628, in next data = self._next_data() File "G:\Users\Administrator\anaconda3\envs\ganomaly\lib\site-packages\torch\utils\data\dataloader.py", line 670, in _next_data index = self._next_index() # may raise StopIteration File "G:\Users\Administrator\anaconda3\envs\ganomaly\lib\site-packages\torch\utils\data\dataloader.py", line 618, in _next_index return next(self._sampler_iter) # may raise StopIteration StopIteration

I want to know if the program was forced to interrupt at a certain step? And there is no record of loss in the test~

[cuteboyqq]

did you set the batch_size 1 ? For example command : python test.py --batch-size 1 --normal xxx --abnormal xxx ...