This is base project to track, reproduce and versioning experiments. This implements a gridsearch on hiper parameters of a collection of tree based models from scikit-learn + xgboost, train and test and log the results with MLflow.
- Docker installed.
run:
docker build -t mlops:latest .
Edit the configuration file in data/hiper_parameters/parameters.json to use different parameters/models.
Edit the configuration file in data/configurations/configurations.json with the experiment name you are about to run.
run bash run_container.sh
visualize results in results/[experiment_name].csv or run sudo docker run -it --privileged -v $(pwd):/home --network host mlops:latest mlflow ui and acess
Create a tag for the experiment e.g:
git add results/* data/*
git commit -m "results from base parameters"
git tag -a v0.0.1 -m "results from base parameters"
git push origin v0.0.1