When I try to do deep learning in a local environment without GPU, it is slow and stress Mach. On the other hand, if you try to do it on AWS where GPU is easy to use, you have to write code without IDE (vim or emacs). So, let's try using jupyter on AWS to make it a little easier.
Can be complemented with tab. that's all! Well, there are things like seaborn that looks graphical, and pandas that makes the table look nice. Easy when talking to others. And
#jupyter installation
pip install jupyter
#Generation of jupyter configuration file etc.
jupyter notebook --generate-config
#If you do it as the root user, you should stop doing it. If you do something else--allow-Add root additionally.
#Local connection settings
vi ~/.jupyter/jupyter_notebook_config.py
#Add the following
c.NotebookApp.ip ='*' #IP control of connection source
c.NotebookApp.port = 8888 #Connection source port control
jupyter notebook --allow-root
After that, if you allow the global IP and port: 8888 of the connection source in the AWS security group, it's OK.
Now you can use caffe, keras, etc. on jypyter as you like and code with GPU.
Recommended Posts