learning_ai

A curated repository of AI learning notes, hands-on notebooks, and small experiments.

What this repo contains

This repository currently focuses on a first course sequence about language modeling fundamentals. The notebooks move from simple probability distributions to n-gram models, dataset preparation, and training a small language model.

Repository structure

• course_1/: a notebook-based learning path covering probability distributions, n-gram models, dataset preparation, and small language model training.
• requirements.txt: common local dependencies for the current notebook set.
• .gitignore: keeps notebook checkpoints and local environment files out of version control.

Current learning path

Course 1: Language modeling foundations

Start with the notebooks in order:

1. gdm_lab_1_1_create_your_own_probability_distribution.ipynb
2. gdm_lab_1_2_experiment_with_n_gram_models.ipynb
3. gdm_lab_1_3_compare_n_gram_models_and_transformer_language_models.ipynb
4. gdm_lab_1_4_prepare_the_dataset_for_training_a_slm.ipynb
5. gdm_lab_1_5_train_your_own_small_language_model.ipynb

See course_1/README.md for notebook-by-notebook details.

Setup

Option 1: Google Colab

This is the easiest path for the current notebooks, especially because some notebooks install extra dependencies directly inside the notebook.

Option 2: Local Jupyter environment

1. Create and activate a virtual environment.
2. Install the shared dependencies:

pip install -r requirements.txt

3. Open the notebooks in Jupyter.
4. Leave the notebook install cells enabled, because they install the ai_foundations package and a few notebook-specific extras.

Environment notes

The notebooks use Python 3 and rely on packages such as:

• pandas
• ipython
• tensorflow
• keras
• ai_foundations

Notebook 1_3 also installs extra JAX and checkpointing dependencies inside the notebook itself. Notebook 1_5 uses Keras with a JAX backend, so Colab is likely to be the smoothest way to run the full sequence.

Recommended workflow

• Read each notebook in sequence.
• Run cells and keep your own notes on observations, errors, and experiments.
• Use this repo to track what you learned, not just what you copied.
• Add new courses as separate folders with their own README files for consistency.

Improvement ideas

• Add short reflection notes after each notebook.
• Add screenshots or output samples once you start customizing the labs.
• Split future courses into separate folders like course_2/, vision/, or agents/.
• Add a progress tracker if you want this repo to become your public learning log.