DreamTeam/Valis
Jump to navigation
Jump to search
About
Valis is a shared compute resource for training machine learning models. Specifically, Valis excels at tasks that require GPU compute. Valis is intended for research and educational purposes only (not for profit).
Specs
- Intel 4114 10-core CPU
- 4x Nvidia Titan V GPUs
- 196GB ECC RAM
- 2x 2TB SSDs
Access
Valis is accessed through a VPN. Follow the instructions below to receive and utilize a VPN user account.
- Let the community know about your project by sending a message to the ml list or join the Dream Team Neurohackers in person on Wednesday nights at 8pm.
- Make a note on Talk:Valis with your GPG public key fingerprint. You will be emailed an encrypted archive containing your cert, key, OpenVPN configuration file, and OpenVPN username/password for connecting to the VPN plus a Jupyterhub username/password.
- Jupyterhub and OpenVPN currently use two different user access systems.
- Configure your computer to connect to our OpenVPN server
- On Debian this can be configured using network-manager-openvpn.
- On OS X Tunnelblick works well.
- Connect to the VPN and run #verifications.
- If the verifications pass, then continue. If not, see #Debugging/FAQ.
- Change your VPN password here.
- Please note that all TLS certs on the internal network are signed by an internal CA. Your browser will not trust this CA, nor the certs that it has signed by default.
- Login to our Jupyterhub instance.
- Open a terminal session and #change your linux user password.
- Open a Python 3 notebook and begin hacking.
verifications
$ ping 172.16.1.1 # DNS server $ ping 172.16.1.2 # Valis $ ping valis.ml # Also Valis, demonstrates your using the internal DNS
change your linux user password
$ passwd
Scheduling
Notes on how/what to check before running a process
Debugging/FAQ
Services
URL | Comments |
---|---|
https://pfsense.localdomain | Firewall, DNS, OpenVPN server, etc. |
https://jupyterhub.valis.ml | Notebook server |
https://chronograf.valis.ml | Metrics dashboards |