MLflow Extend

Extend MLflow's functionality.

Installation

From PyPI

pip install mlflow-extend

From GitHub (development version)

pip install git+https://github.com/harupy/mlflow-extend.git

Example

import numpy as np
import pandas as pd
import matplotlib.pyplot as plt
from plotly import graph_objects as go

from mlflow_extend import mlflow

with mlflow.start_run():
    # mlflow native APIs
    mlflow.log_param('param', 0)
    mlflow.log_metric('metric', 1.0)

    ##### new APIs mlflow_extend provides #####

    # flatten dict
    mlflow.log_params_flatten({"a": {"b": 0}})
    mlflow.log_metrics_flatten({"a": {"b": 0.0}})

    # dict
    mlflow.log_dict({'a': 0}, 'dict.json')

    # numpy array
    mlflow.log_numpy(np.array([0]), 'array.npy')

    # pandas dataframe
    mlflow.log_df(pd.DataFrame({'a': [0]}), 'df.csv')

    # matplotlib figure
    fig, ax = plt.subplots()
    ax.plot([0, 1], [0, 1])
    mlflow.log_figure(fig, 'figure.png')

    # plotly figure
    fig = go.Figure(data=[go.Bar(x=[1, 2, 3], y=[1, 3, 2])])
    mlflow.log_figure(fig, 'figure.html')

    # confusion matrix
    mlflow.log_confusion_matrix([[1, 2], [3, 4]])

    # run "mlflow ui" and see the result.

Creating a Development Environment

conda create -n mlflow-extend python=3.6
conda activate mlflow-extend
pip install -r requirements.txt -r requirements-dev.txt

Formatting Code

After formatting, verify all the checks pass by running:

./dev/lint.sh

Running Type Check

mypy .

Running Unit Tests

# Run all the unit tests.
./dev/test.sh

# Save figures generated during the unit tests in '.pytest_basetemp'.
./dev/test.sh --savefig

Building Documentation

cd docs
make html

# Remove everything under 'docs/build' and run 'make html'.
make clean html

The generated files will be stored in docs/build/html. Open docs/build/html/index.html on the browser to check if the documentation is built properly.

Releasing New Version

1. Make a pull request to update __version__ in mlflow-extend/version.py to the next version.

- __version__ = "1.2.2"  # current version
+ __version__ = "1.2.3"  # next version

2. After the pull request is merged, create a new tag and push it to the remote.

git tag v1.2.3
git push origin v1.2.3

3. Open the release page and publish a new release.

4. Upload distribution archives to PyPI using twine.

# Remove old distribution archives.
rm -r dist/*

# Generate new distribution archives.
python setup.py sdist bdist_wheel

# Upload to Test PyPI and verify everything looks right.
twine upload --repository-url https://test.pypi.org/legacy/ dist/*

# Upload to PyPI (THIS CAN NOT BE UNDONE).
twine upload dist/*