Skip to content

Instantly share code, notes, and snippets.

@harsh183
Created November 5, 2018 08:24
Show Gist options
  • Save harsh183/38ea87c2bdeef66360a914030e7d623d to your computer and use it in GitHub Desktop.
Save harsh183/38ea87c2bdeef66360a914030e7d623d to your computer and use it in GitHub Desktop.
A noob's guide to fast.ai on VectorDash

VectorDash guide for Fasth AI (for Noobs, by Noobs, of Noobs)

Last updated: 2018-11-4

Okay I saw this on reddit a while back here

It's like digital ocean but GPU, neat and really simple to setup. My notes on this and some bash scripts I guess.

I'm a bit salty there is no persistent storage, but the net is fast so I think I can work around that.

Setup

SSH key

Warnings like WARNING: UNPROTECTED PRIVATE KEY FILE! etc.

Basically it's with the file permissions of the ssh keys, so just change file permissions so that only the current user can access it while groups and others have no access to it.

The datasets

fast.ai notebooks

Getting Juypter notebook to work - Port forwarding

Basically I couldn't figure out how to get the jupyter notebook thing to work over the net so ssh port forward.

  1. Start your jupyter notebook
ubuntu@remote:~$ jupyter notebook
  1. Port forward on your own machine. (Note that the key file is just hostname and -key after it)
you@localmachine:~$: ssh -i <YOUR-MACHINE-NAME>-key -N -f -L localhost:8888:localhost:8888 ubuntu@167.99.156.225 -p <YOUR PORT>
  1. Open up the link that the vm shows after you run jupyter because it has the access token. Something like this (token differs)
http://localhost:8888/?token=4da5ec85bc5a47bb467f78d6b5a56deedd9c38d41b40c1d2

TODO: A lot more just getting started

They've nicely given a bash script for downloading datasets so we'll use that. Plus there is the fast.ai bash script. I think I'll just try making a combination of both.

Enjoy

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment