GPU servers (Goggins)

There are currently 3 GPU equipped servers available for the lab members:

Server Address CPU Memory (GB) GPU(s)
Goggins 01 decmtcsaiml01.cs.man.ac.uk Intel Xeon Gold 5217 (16 cores HT) 188 2x Quadro RTX 6000 (24GB VRAM)
Goggins 02 decmtcsaiml02.cs.man.ac.uk | | | | 1x Quadro RTX 6000 (24GB VRAM)
Goggins 04 decmtcsaiml04.cs.man.ac.uk | | | | 1x RTX A6000 (48GB VRAM)

The servers operate on a free time sharing basis, meaning you can use them at any time, unless there is a request for exclusive use, which will be evaluated per case and if approved will be informed on the servers slack channel.

To request access, please contact Danilo Carvalho (Goggins 01/02) or Alex Bogatu (Goggins 04), providing the following information:

  • Full name
  • University username (a string of 8 letters and numbers)

Access is done via SSH.

ssh <username>@decmtcsaiml<num>.cs.man.ac.uk

where <username> is the university username and <num> is the server number.
The password is the same as your university IT systems one.

Storage

The servers have 3 storage spaces available to the users:

  • Home: Very small space available (~5GB). Use for config files and project/library symlinks.
  • Scratch: SSD storage, ~100GB p/ user, accessible by $HOME/scratch/. Use for code or data files requiring speed.
  • Data: HDD storage, ~300GB p/ user, accessible by $HOME/data/. Use for bigger data files or archival.

While storage quotas are not strictly enforced, exceeding them may result in loss of data.

Runtimes & Libraries

All servers have the build-essentials (C/C++), python (>= 3.6) and OpenJRE 11 (java) packages installed.

To set up a python environment for ML with libraries, we suggest a user install of Anaconda with any needed packages in one or more virtual envs.

If you need an application or library that is not available in the server, please contact Danilo Carvalho (Goggins 01/02) or Alex Bogatu (Goggins 04), with relevant information about the software needed.

Idiap Computational Resources

(*Idiap members only)

Idiap provides researchers with access to a centralised computational facility.

You can find a complete description of the available resources along with guidelines to use them on the Idiap Intranet.

A very useful documentation on how to setup your working environment (including Pytorch) is available here.