Skip to content

Latest commit

 

History

History
40 lines (28 loc) · 1.24 KB

README.md

File metadata and controls

40 lines (28 loc) · 1.24 KB
Raw

MLflow

In this repo only two components of MLflow are used:

  • tracking - for tracking parameters, metrics and artefacts
  • projects - for separating environments for executing steps in pipeline

Setup

Local tracking server

MLflow can be used locally without sharing experiments with the team.
To run locally:

  1. install necessary packages pip install -r requirements.txt
  2. in one terminal run mlflow server --default-artifact-root <PATH_WHERE_TO_STORE_ARTIFACTS>

Remote tracking server

To set up simple remote tracking server on EC2 follow instructions in tracking_server directory
To use such server:

  1. (install and) configure aws cli
  2. change tracking_uri in config.yaml to point to tracking server

How to use this repo

After preparing tracking server:

  • execute to run full pipeline:
cd <REPO_ROOT>
mlflow run .
  • open tracking_uri in browser to see MLflow UI

All necessary pipeline configuration options can be found in config.yaml.
Options specific to model are stored in src/train_model/model_config.json


Helper script

download_artifact.py is a helper script showing how artifacts can be downloaded