jitubutwal144 / mlflow-project-demo

Uses mlflow to track experiments for each ML lifecycle steps

How to run on localhost?

• On MAC, CMD + P
• Rebuild Container
• Create a folder with kaggle credentials .kaggle/kaggle.json, which is required to get dataset from kaggle
• Terminal >> New Terminal
• mlflow ui
• In other terminal run experiments using command: mlflow run . --env-manager=local --experiment-name fraud -P n_trials=1

Things to check after a successfull run

• After running command from terminal (mlflow ui), it should open a user interface under http://127.0.0.1:5000/#/experiments
• There should be data folder created within this project and following subfolders: raw, processed
• There should be a folder called mlruns within this project
• mlflow UI should contain experiments runs with predicted metrices generated using catboosting alogorithm
• There should be catboost_info folder generated within root director