Skip to content
ar-toPublic

About

repo for ML related projects and tests

Resources

Stars

0 stars

Watchers

1 watching

Forks

Repository files navigation

devML

repo for ML related projects and tests

Getting Started with a Python Model

note some steps may not yet exist or be implemented (e.g. check results directory) but left here for reference as this repo is improved.

To get started with a Python machine learning model, follow these steps:

Prerequisites

Ensure you have the following installed:

  • Python 3.8 or higher
  • pip (Python package manager)

Installation

  1. Clone the repository:
    git clone https://github.com/yourusername/devML.git
    cd devML
  2. Create a virtual environment:
    python -m venv venv
    source venv/bin/activate  # On Windows, use `venv\Scripts\activate`
  3. Install dependencies:
    pip install -r requirements.txt
  4. Upgrade Packages
    pip install pip-review
    pip-review # similar to `pip list --outdated`
    pip-review --auto #upgrade all because pip is not easy for this
    pip freeze > requirements.txt # update
  5. Install internal modules Modules like utils are self contained modules that live within this project alone to help with helper methods like monitoring. This project follows a flat structure so it remains flexible to organizing. But custom modules like utils requires treating this project like a package and install it on the python interpreter local to this project via venv above. Simply run:
    pip install -e .

Restart the notebook kernel. Then open a file using it such as explorations/utils_checks.ipynb and run the cell that imports it to test.

Running a Model

  1. Prepare your dataset and place it in the data/ directory.
  2. Run the model script:
    python src/train_model.py
  3. Check the output in the results/ directory.

Notes

  • Modify config.yaml to adjust model parameters.
  • Refer to the docs/ folder for detailed explanations of each module.

Knowledge Graph (graphify)

This repo has a graphify knowledge graph of its code, notebooks and docs. Claude Code and its subagents can query it with /graphify query "<question>" instead of re-reading the whole repo. Rebuild it with /graphify ., or run /graphify . --update to re-extract only changed files.

  • Outputs are in graphify-out/. A log of each build is in docs/graphify_changelog.md.
  • graphify can't read .ipynb files, so notebooks are converted to .py files in graphify-out/notebooks/ before each build.
  • Git: .gitignore excludes everything in graphify-out/ except graph.json and GRAPH_REPORT.md. Those two files are committed so a fresh clone, or a subagent, can query the graph without rebuilding it. The cache, manifest, cost log, notebook conversions and graph.html stay local.

About

repo for ML related projects and tests

Resources

Stars

0 stars

Watchers

1 watching

Forks

Releases

Packages

Contributors

Languages