Installation environment win10 +tensorflow 1.7.0 +cuda9.0+cuDNN v7.0
When installing tensorflow, it is a library added directly in pycharm
Due to this installation method, tensorboard is not affiliated with the tensorflow installation path, and their paths are in a parallel relationship, which will cause tensorboard to be prompted if
ensorboard –logdir='logs/' is entered in the console when using tensorboard: command not found, in fact, it is not that tensorboard is not installed. At this time, you need to find the path of tensorflow, press Win+R to enter python,
and then enter import tensorflow as tf and tf.__path__ in turn, and then you will get the path of tensorflow.
Open this path and you will find:
Both tensorboard and tensorflow are under sit-packages. Open the tensorboard folder and there is no tensorboard.py file. In fact, the main.py file here is tensorboard,py
After figuring out this problem, you also need to figure out the routine you want to call tensorboard, which generates a log path after running. Take my environment as an example,
For example, in the project where my program is located, after running the program, there will be a logs folder in the project file, which contains the files generated by the program. In addition, there is another place where the same file will be generated, that is, the logs under the installation path of your original python environment will also have the same file.
Finally, you can enter this in the cmd console: python C:\Users\your own path\Python\Python36\site-packages\tensorboard\main.py --logdir=your own project path\logs
Copy the address below to Google Chrome or Firefox and it's ok.
Reference: https://blog.csdn.net/appleml/article/details/76976006
thank you very much!