This repository contains source code and material produced for the exam of the course Deep Learning, attended in the second semester of the academic year 2022-2023 of master's degree in Physics of Complex Systems at the University of Turin. Content of the repository at the time of the exam is pointed by tag exam.
The course is teached by prof. Matteo Osella.
As part of the exam I discuss the Transformer model, implementing the code from paper Attention Is All You Need with the help of the related YouTube video by Umar Jamil and other sources.
Python 3.10 is used for the code base.
Source code is stored and described in file project.ipynb. An effort to follow PEP 8 is done when writing code in notebook cells, but no automatic helper tool is used.
Figures used in the notebook are stored in folder figures.
In this folder, files matching the pattern ModalNet-*.png are copies of the original figures used in the paper, hence they are creation of the authors of the article. Other files are my own creation.
A Python module can be extracted from notebook project.ipynb by running the following command in the same directory:
python -m convert
Cells identified with tag convert-module are extracted to the output file.
Note that the converted code is likely not PEP 8 compliant. Additional post-processing could be required, as running autopep8 or black.