AML & DAAI 2024/2025 Project - Task Arithmetic Under Bias-Variance Trade-offs

Official repository for the "Task Arithmetic Under Bias-Variance Trade-offs" project - Advanced Machine Learning & Data Analysis and Artificial Intelligence Courses 2024/2025 @ PoliTo

Getting Started

Make sure to have a CUDA capable device, supporting at least CUDA 11.8, installed and correctly configured on your system.

(The base code of this project has been produced using CUDA 11.8 and Python 3.10.9)

Follow https://pytorch.org/get-started/locally/ to setup PyTorch (note that PyTorch comes pre-installed on Google Colab's environments, so you can skip this step)

Once you have properly setup everything, make sure you are in the correct directory and run from the command line:

pip install -r requirements.txt

Dataset

Download to your disk/drive the datasets from the provided drive folder (see project report). Then, unzip them to some location.

At this point you should be able to run and edit the base code provided.

NOTE: please, do not push the datasets into your github repository / your exam submission.

Base Code Structure

The starting code should already provide everything needed to easily extend it. Read carefully the specifications in the project report.

In the following, you can find a brief description of the included files.

File/Folder	Description
args.py	contains the function responsible for parsing each command line argument.
datasets/	contains the files with code to load data, build splits and dataloaders.
utils.py	contains several utilities to correctly setup the pipeline.
task_vectors.py	contains the code for building task vectors and/or load checkpoints.
modeling.py	contains the backbone architectures and modules used in the project.
heads.py	contains the logic to build and store the open-vocabulary classifiers used in the project.

Running The Experiments

To run the experiments you can use, copy and modify the provided launch script launch_scripts/base.sh, which should give you an idea on how to design the implementation of the missing files. As an example, after producing the three missing files, you should be able to launch the experiments as

./launch_scripts/base.sh

NOTE: you should upload with your code also the launch scripts as they are fundamental to reproduce your results.

In the following, you can find a brief description of the relevant command line arguments when launching the experiments. They should suffice. However, if needed, you can add additional arguments by editing the args.py file.

Basic Command Line Arguments

Argument	Description
--model	the name of the architecture used in the experiments (keep it as ViT-B-32)
--batch-size	batch size used in the optimization procedure (default: 32)
--lr	learning rate used in the optimization procedure (default: 1e-4)
--wd	weight decay of the optimizer (default: 0.0)
--data-location	path to the folder containing your unzipped dataset folders.
--save	path to the folder where to save your results.

Bug Reporting

You are encouraged to share in the project's dedicated Slack channel any bugs you discover in the code. It's incredibly valuable for everyone involved to be aware of any issues as soon as they arise, helping to address them promptly. Your contribution is greatly appreciated!