# Ollama
# First launch of Ollama
Go to https://qcg.pcss.plcloud.pl (opens new window)
# Submit Task
Click "Submit Task"

Find the Ollama template

Submit using the template

Set the appropriate General options (you can leave them by default)

Set the resource options as needed

When all options are set, click "Submit"


When the environment is ready, go to the Overview tab and click VNC

You will be redirected to the OpenWebui interface, where you need to create an account

Use it
# Deleting a task

For the appropriate task, click Cancel Task



Delete it


# Using an Environment from a Pre-Existing Job
Click "Submit Task"

Find the Ollama template

Submit using the template

Set the appropriate General options

Change the Model Environment to Use Existing

Go to home@altair

Find a similar path to the model environment and confirm



In the Webui environment similarly select use existing





Set the resource options as needed

When all options are set, click "Submit"


When the environment is ready, go to the Overview tab and click VNC

You will be redirected to the OpenWebui interface, where you need to create an account

Use it
# Searching the web with OpenWebui
Go to admin panel

Go to settings
Go to web search

Make sure web search is allowed
Expand the list of search engines
And choose the right one for you

Configure based on the appropriate tutorial
https://docs.openwebui.com/category/-web-search/ (opens new window)
Configure the rest of the parameters (you can leave it default)

Go to chat
Indicate that you want to search the internet

Type a query.
# Ollama command line
Symbolic linking of 2 directories: ~/.singularity and ~/.ollama
ln -s <grant_id>/project_data/ollama/ ~/.ollama
ln -s <grant_id>/project_data/singularity/ ~/.singularity
Log in to a GPU node and select the number of available GPUs
srun --mem 100G --gpus-per-node 4 -p proxima --pty bash
Create a definition file ollama.def
Bootstrap: docker
From: ollama/ollama
%startscript
ollama serve
Build the image
sudo singularity build ollama.sif ollama.def
Run the container
# ollama runs on port 11434
singularity instance start --nv --bind /mnt/:/mnt/ ollama.sif ollama
Enter the container
singularity shell instance://ollama
Test the selected model
ollama run deepseek-r1:7b