A logical, reasonably standardized, but flexible project structure for doing and sharing machine learning work.
Note: template tested on Python 3.8
Run:
$ pip install requirements.txtRun:
cookiecutter https://github.com/mihail911/e2eml-cookiecutter
The directory structure of your new project looks like this:
├── LICENSE
├── README.md <- The top-level README for this project
│
├── requirements.txt <- The requirements file for reproducing the analysis environment, e.g.
│ generated with `pip freeze > requirements.txt`
│
├── assets <- Stores any images, PDFs, or other static assets
│
├── data
│ ├── processed <- The final, processed data sets for modeling
│ └── raw <- The original, immutable data dump
│
├── deploy <- Stores any files, configurations related to deploying your model (Dockerfile, etc.)
│
├── model_checkpoints <- Trained and serialized models, model predictions, model summaries, config files
│
├── notebooks <- Jupyter notebooks for doing exploratory data analysis, analyzing model outputs, etc.
│
├── scripts <- Single-purpose scripts for functionality such as processing and cleaning data
│
├── tests <- Tests for the various aspects of the project (data cleanliness, data processing, model training code, etc.)
│
├── {{cookiecutter.repo_name}} <- Source code for use in this project
│ ├── __init__.py <- Makes {{cookiecutter.repo_name}} a Python module
│ │
│ ├── model <- Stores any relevant modeling code, interfaces, and definitions
│ │ └── __init__.py
│ │
│ ├── server <- Stores deployment and inference server code
│ │ ├── __init__.py
│ │ └── main.py <- Main module for running server
│ │
│ ├── utils <- Stores various utilities used in project
│ │ ├── __init__.py
│ │
│ ├── train.py <- Script to run model training
│ └── eval.py <- Script to run trained model evaluation