gai-lab-ese

Generative-AI lab experiments (GAI Lab ESE) as a pip package: the exact code of every notebook as a .txt/.py file, a full theory + code walkthrough + viva questions for each experiment, the original notebooks, and music.wav.

Install

pip install --upgrade gai-lab-ese              # code + theory
pip install --upgrade "gai-lab-ese[all]"       # + everything needed to run all experiments
pip install --upgrade "gai-lab-ese[tf]"        # Exp 1, 2, 5, 9
pip install --upgrade "gai-lab-ese[hf]"        # Exp 3, Prac
pip install --upgrade "gai-lab-ese[audio]"     # Exp 6
pip install --upgrade "gai-lab-ese[cyclegan]"  # Exp 7

Colab / Jupyter: put ! in front, e.g. !pip install --upgrade gai-lab-ese and !gai-lab-ese --list.

Command line

gai-lab-ese --list                         # list experiments
gai-lab-ese --show vae                    # title + theory + full source
gai-lab-ese --theory vae                  # only theory / viva explanation
gai-lab-ese --code vae                    # only code
gai-lab-ese --save vae --out vae.py
gai-lab-ese --save-all --out gai_code       # all: .py, .txt, theory and original notebooks
gai-lab-ese --data --out gai_data           # datasets
gai-lab-ese --all --out gai_all             # code + theory + datasets in one folder
gai-lab-ese --run vae                    # run an experiment

Python

from gai_lab_ese import list_experiments, show_experiment, get_source, get_theory, save_experiment, save_all, run_experiment, download_all

list_experiments()
show_experiment("vae")              # title, theory, then the full Python source
print(get_theory("vae"))            # viva explanation only
code = get_source("vae")
save_experiment("vae", "vae.py")
download_all("gai_all")

Experiments

Experiments E1-E3 (Markov chain text, genetic-algorithm image, evolutionary image) are 100% offline and need only random / numpy / matplotlib. Original notebooks: gai-lab-ese --notebook vae (saves Exp1.ipynb). In Colab, Colab already has TensorFlow, PyTorch and transformers, so pip install gai-lab-ese is enough.

See PUBLISHING.md for how to upload this package to PyPI.

License

MIT