When I reproduced the relevant code of the pggan paper, I encountered a problem. There is a data visualization thing in it, that is, when the loss function is displayed with the training picture, that is, tensorboard is different from the previous way of saving pictures and loss function images. So this article talks about how to use tensorboard.
I don’t want to explain too much about the use of tensorboard. I will write about the specific use later. This article mainly talks about how to display tensorboard.
First find similar files as shown below:
Then enter cmd on the command line
That is, use the mouse to click on the line in the screenshot above
Enter tensorboard --logdir=E:\desktop\paper\PGGAN\ProGAN\logs\gan1
Please note here that please change the address of this folder, because everyone's folder location is different, you only need to copy the above folder address again. If the following error is displayed
Please don't worry, I have encountered this kind of problem before, this is because I use anaconda, just enter the above command in the scripts folder in the virtual environment you built, and then a URL will appear, copy the URL Use to call out tensorboard. If you don't know how to use anaconda or the like, I may write about the construction method of the virtual environment and the correspondence between torch and python versions in future articles, because when I first started learning, this stuck for a long time. After that, the knowledge and code related to cyclegan and pggan may be updated. Hope to bring you help. I also hope that you can leave me a message if there is something you don’t know. If I see it, I can answer it for everyone.