AAAMLP starter

Starter kit for projects from Approaching Almost Any Machine Learning problem book by Abhishek Thakur

Note 👷

This starter is very much work in progress. I am updating key modules as I am making my way through the book.

Prereqs

install miniconda
conda create -n NAME_FOR_YOUR_CONDA_ENV python=3.7.6 (if you are on M1/M2, you might need to run this before conda config --env --set subdir osx-64)
conda activate NAME_FOR_YOUR_CONDA_ENV
conda env create -f environment.yml
OR conda env create -f environment_osx.yml on Mac
conda activate ml
download mnist_train.csv from Abhishek's Kaggle, and save it to input directory

Getting started

This template gets you started with a basic skeleton for your ML app.

Creating folds

cd src && python create_folds.py --folds 10 (default is 5 folds)

Adding/editing available models

model_dispatcher.py contains a dictionary with models available in your application

Training your model

To train your model, use

python train.py --fold 0 --model desision_tree_entropy