New project: Paniikki GPU machine remote utilisation for AI

TL;DR

Aalto CS-IT is initiating a new project: Paniikki GPU machine remote utilisation for AI. You will be able to use Paniikki machines for your Kaggle projects or Stanford Convolutional Neural Network course from anywhere. If you find this interesting please give us a feedback :)

Background

AI is the new sexy

Alphago is the new GO champion. Everyone is hiring data scientists. Aalto machine learning course is exploding with 600 students and everyone is teaching machine learning and deep learning. Aalto machine learning master’s program, MACADAMIA, is packed with international talents.

While the topic itself is very interesting and there are many resources for studying it, there are some chanllenges that people face: GPU and dependency installations. You could always use a powerful GPU workstation in Paniikki but you may be busy to go there or you may not be familiar with remote SSH.

Can Aalto help you?

YES!!! We are always glad to help you and we want you to get advantage of the luxurious workstations in Paniikki(The NVIDIA Quadro P5000 GPUs in Paniikki costed us €2K/module). We will build a desktop app that allows you to use complete scientific containers on Paniikki machines, remotely without any use of Terminal. We will also prepare containers for famous ML courses so you don’t have to struggle setting the infrastructures.

How are we going to deliver this?

We will distribute a cross-platform app. You install it on your Linux/Windows/Mac. When you start the app, click a container of your choice e.g. Tensorflow_py35. A jupyter notebook server will start on a most idle machine in Paniikki and you will be given an URL & token for the notebook. We are considering of using ElectronJS for the front-end and Singularity for the containers. We use containers because you can just take them if you want to use it somewhere else.

Who could use it?

All Aalto students and staffs.

When will it be available?

2018 spring. Hopefully April.

Last but not least: let’s make Aalto great together!

Our mission is to serve you the best to enhance the science. We work hard but we want to make sure that we are doing it right. So your feedbacks are tremendously invaluable to us. Tell us how you like our ideas and what you want and need, but please be nice ;) We are humans and we have feelings like you. If you would like to collaborate with us you are more than welcome! Please pay us a visit at A243 at CS building or send us an e-mail @ guru@cs.hut.fi.


Appendix

In Paniikki(panic in Finnish), a computer room in the CS building, there are 31 monster machines waiting to rock ‘n roll with you. Here is the spec:

CPU properties Spec
Model Intel(R) Xeon(R) CPU E5-1650 v4 @ 3.60GHz
Architecture x86_64
Byte Order Little Endian
CPU(s) 12
Thread(s) per core 2
MHz 1200.796
max MHz 4000.0000
min MHz 1200.0000
Virtualization VT-x
L1d cache 32K
L1i cache 32K
L2 cache 256K
L3 cache 15360K
Model NVIDIA Quadro P5000
GPU properties Spec
Core GP104GL (Pascal-based)
Core clock 1607 MHz
Memory clock 1251 MHz
Memory size 16384 MiB
Memory type 256-bit GDDR5X
Memory bandwidth 320
CUDA cores 2560
CUDA compute capability 6.1
OpenGL 4.5
OpenCL 1.2
Near GeForce Model GeForce GTX 1080
Memory properties Spec
RAM 32GiB