A Guide for Creating a Server for Deep Learning: Assembly, Server & Library Setup

Just finished a write up on how to create your own Server for training models which can be found here:

The part spec’ing, component build, server configuration, and DL library/environment setup is thoroughly documented. Anyone can build their own server, and I wanted to share my progress in making my own.


Hey thanks for this! I have a question… did you consider ssh’ing through a router VPN? Do you think port forwarding is safe? I am thinking of buying a router and installing a VPN with dd-wrt

I didn’t consider using a VPN as SSH imo is secure enough. From doing some quick googling, it seems like the consensus is that remote connecting into a server with either SSH or VPN is pretty much the same strength security wise. According to this guy, VPN is a bit safer because it “works at the transport layer, supports both UDP and TCP and allows secure access to multiple resources.”

Personally, I’m comfortable with using SSH, but if you’re willing to go the extra step and use VPN then it only helps!