====== Lab 03 ======

===== Factor Graphs =====

During this lab, you should get a bit deeper understanding of the factorgraph concepts presented during the lectures.

{{ :courses:aro:tutorials:labs_03_templates.zip |}} \\ {{ :courses:aro:tutorials:labs_03_program.pdf |}} 

===== Pytorch installation =====

<note important>
You will need Pytorch and matplotlib installed on your PC for this (and only this) lab (either inside or outside the ARO Apptainer).
</note>

If you have any working Pytorch and matplotlib installation on your PC, then you're ready for the lab.

If you don't, follow this tutorial (for Ubuntu Linux):

<note>
Please, make sure to be outside any Conda environments when following this tutorial.
</note>

====== Outside ARO Apptainer ======

<code>
sudo apt update && sudo apt install python3-venv
python3 -m venv --system-site-packages ~/pytorch_venv  # or any other folder in your home
source ~/pytorch_venv/bin/activate
pip install torch matplotlib  # This downloads about 900 MB of data and takes 7 GB on disk
</code>

====== Inside ARO Apptainer ======

<code>
python3 -m virtualenv --system-site-packages ~/pytorch_venv  # or any other folder in your home
source ~/pytorch_venv/bin/activate
pip install torch  # This downloads about 900 MB of data and takes 7 GB on disk
</code>

<note>
If you won't need PyTorch for anything else and you don't need GPU support (this lab does not need it), you can install the CPU-only version of PyTorch. It only takes up 1 GB on disk.

<code>
# pip install torch matplotlib  # instead of this, do the following two commands
pip install matplotlib
pip install torch --index-url https://download.pytorch.org/whl/cpu
</code>
</note>

====== Using the virtualenv ======

When you leave this console or reboot your computer, you get the virtualenv back by calling ''source ~/pytorch_venv/bin/activate''.

<note>
If you're on WSL, the installation of Pytorch took too much space and deleting the virtualenv doesn't free up the space, please follow [[https://superuser.com/a/1612289|this guide]] to reclaim the taken space.
</note>