mlflow-tools

Some tools for MLflow.

Overview

• Dislay tools - show details and list MLflow objects
  • Superceded by https://github.com/amesar/mlflow-reports. Sample JSON output
    • Superceded by this.
• Manipulate object tools - manipulate MLflow objects - delete, rename, etc.
• Other tools
  • Failed runs - Save run details for MLflow rate limited exceptions and replay later.
  • Model version validation tools
• Databricks notebooks
• Tests

MLflow Tools

Command line scripts

Display (list and dump) MLflow objects

• README
• List: experiments, registered models and model versions
• Dump: run, experiment and registered model
• JSON samples of MLflow object display dumps

Model Version Validation

• See README_check_version.
• Tools to:
  • Validate a version's MLflow model.
  • Compare two versions' MLflow models.

Helper Tools

• README
• Find best run of an experiment.
• Find model artifact paths of a run
• Find matching artifacts
• Download model artifacts.
• Call MLflow model server.
• Registered model tools
  • Register a run's model as a registered model.
  • Delete registered model.
  • Delete model stages.
• Call http_client - either MLflow API or Databricks API.

Databricks notebooks

• Notebook (README) versions of command line scripts
• Sample notebook screenshots: list registered models , dump model and list model versions

Other tools

README:

• MLflow Spark UDF Workaound
• Failed Run Replayer
• Seldon MLflow MLServer

Setup

Step 1. Create a virtual environment.

python -m venv mlflow-tools
source mlflow-tools/bin/activate

Step 2. pip install

pip install from github

pip install git+https:///github.com/amesar/mlflow-tools/#egg=mlflow-tools

or pip install in editable mode

git clone https://github.com/amesar/mlflow-tools
cd mlflow-tools
pip install -e .