Create a new GPU instance SWITCH
General dashboard
https://cloud.switch.ch/organisations/ec7ac849-8c59-4dcd-9f50-6d697288a589/projects
then, after successful login, click on TI - MLMP as Services -> Dashboard -> again Dashboard in the Compute tile
(choose current auth. session) and finally, you are there 🤓
MLMP specific dashboard
Instructions
First create a floating ip so you can access with ssh:
Project > Network > Floating IPs
Also create an ssh key pair: Project > Compute > Key Pairs
Then create key pair, give it a name and select the ssh type then if im not mistaken a .pem key will be downloaded this is used along the floating ip to access the server (more on that later)
Allocate IP to project > pool: public and then description MLMP_Pub_IP
To actually create the instance go to Project > Compute > Instances
then there 'Launch instance'
Details - Instance Name : whatever was accorded
Source: Boot source image and select ubuntu 24.04, volume size 100-200 GB (or whatever is needed [costs])
Flavor: go to the last one in the available list and look for the GPU that you want
Security Groups: default
Key Pair, the one you created before
then click on launch instance,
After that you will see your instance is being created and to the right you will see an option to attach a floating IP, click it an assign the Floating IP you created at first.
Then after the instance is created you can go to your temrinal and ssh into the instance by:
ssh ubuntu@floating-ip -i path/to/.pem
'ubuntu' is the user created by default, and .pem is the file that was downloaded as you created the key pair.
Then in order to be able to use the GPU you need to install the respective NVIDIA drivers:
sudo apt update
sudo apt install -y linux-headers-$(uname -r) build-essential
sudo ubuntu-drivers autoinstall
sudo apt install -y nvidia-driver-550-open
sudo reboot
nvidia-smi