Getting an Nvidia GPU server running on Ubuntu is made very simple by using a script we've created for you. Grab our GPU server setup script from this link. Run the script by typing:
When it's done, cd to the nbs directory that this script creates, and try out the jupyter notebook using the instructions we've provided. Note that the password we've set up for the notebook is: dl_course
You should only use this script on a computer that hasn't already had its GPU drivers set up. If you've already set up some things on your computer, we'd suggested editing the script to remove any sections that you've already handled.
The script downloads and installs:
- Nvidia's drivers
- CUDA (Nvidia's GPU programming system)
- The anaconda python distribution
- Keras and Theano (deep learning libraries for python)
- Jupyter notebook
Here are a few useful guide that explains how to set up/access a secure linux server running Jupyter notebook in a home environment:
- How to forward ports on your router
- Make your ssh server "secure"
- Use ssh tunneling to visit your Jupyter notebook remotely
- Wake-on-lan setup:
1. Download and install the Depicus wol command line tool 2. Enable wol in Ubuntu 3. Remotely start your ubuntu server whenever you need using "wolcmd [mac address] [ipaddress] [subnet mask] [port number]"