This repo provides an example of how to incorporate popular machine learning tools such as DVC, MLflow, and Hydra in your machine learning project. I use my project on predicting aggressive tweets as an example.

Find the article on how to use MLflow and Hydra here

Find the article on how to use DVC here

DVC is a data version control tool. To install DVC, run

pip install dvc

With Hydra, you can compose your configuration dynamically. To install Hydra, simply run

pip install hydra-core --upgrade

MLflow is a platform to manage the ML lifecycle, including experimentation, reproducibility, and deployment. Install MLflow with

pip install mlflow

• src: file for source code
• mlruns: file for mlflow runs
• configs: to keep config files
• outputs: results from the runs of Hydra. Each time you run your function nested inside Hydra's decoration, the output will be saved here. If you want to change the directory in mlflow folder, use

import mlflow
import hydra
from hydra import utils

mlflow.set_tracking_uri('file://' + utils.get_original_cwd() + '/mlruns')

• src/preprocessing.py: file for preprocessing
• src/train_pipeline.py: training's pipeline
• src/train.py: file for training and saving model
• src/predict.py: file for prediction and loading model

Pull the data from Google Drive

dvc pull 

To run the configs and see how these experiments are displayed on MLflow's server, clone this repo and run

python src/train.py

Once the run is completed, you can access to MLflow's server with

mlflow ui

Access http://localhost:5000/ from the same directory that you run the file, you should be able to see your experiment like this