https://github.com/facultyai/hiplot-mlflow

HiPlot fetcher for experiments logged with MLflow
https://github.com/facultyai/hiplot-mlflow

hiplot machine-learning mlflow plotting visualisation

Last synced: about 1 month ago
JSON representation

HiPlot fetcher for experiments logged with MLflow

• Host: GitHub
• URL: https://github.com/facultyai/hiplot-mlflow
• Owner: facultyai
• Created: 2020-05-01T14:50:54.000Z (over 4 years ago)
• Default Branch: master
• Last Pushed: 2022-05-11T08:57:42.000Z (over 2 years ago)
• Last Synced: 2024-04-26T12:45:45.852Z (8 months ago)
• Topics: hiplot, machine-learning, mlflow, plotting, visualisation
• Language: Python
• Homepage: https://pypi.org/project/hiplot-mlflow/
• Size: 1.72 MB
• Stars: 12
• Watchers: 12
• Forks: 2
• Open Issues: 1
• Metadata Files:
  • Readme: README.rst
  • License: LICENSE-2.0.txt

Awesome Lists containing this project

README

        
hiplot-mlflow

=============

A  `HiPlot `_ experiment

fetcher plugin for `MLflow `_, to help visualise your

tracked experiments.

Installation

------------

Install this library with ``pip`` as:

.. code-block:: sh

   pip install hiplot_mlflow

Usage

-----

You can visualise experiments either in a Jupyter notebook or using HiPlot's

built in server.

Notebook

~~~~~~~~

In a Jupyter notebook, use ``hiplot_mlflow.fetch`` to retrieve an MLflow

experiment by name, and display it with HiPlot:

.. code-block:: python

   import hiplot_mlflow

   experiments = hiplot_mlflow.fetch("my-lovely-experiment")

   experiments.display(force_full_width=True)

You can also retrieve experiments by their MLflow experiment ID:

.. code-block:: python

   experiment = hiplot_mlflow.fetch_by_id(0)

By default, MLflow tags are not shown (only MLflow metrics and parameters are

shown). To display them, pass ``include_tag=True`` to either of the fetch

functions, for example:

.. code-block:: python

   experiment = hiplot_mlflow.fetch("my-lovely-experiment", include_tags=True)

.. image:: images/notebook_name.png

   :alt: Loading HiPlot in a notebook

See more about what you can do with the returned ``hiplot.Experiment`` values

in the `HiPlot documentation

`_.

HiPlot Server

~~~~~~~~~~~~~

To use `HiPlot's built in webserver

`_ with

``hiplot-mlflow``, you can start it up with the custom `experiment fetcher

`_

implemented by this package:

.. code-block:: sh

   hiplot hiplot_mlflow.fetch_by_uri

You can then use the ``mlflow://`` schema to access MLflow experiments in

HiPlot by either experiment or name, for example::

   mlflow://name/experiment-name

   mlflow://id/0

.. image:: images/server_name.png

   :alt: Loading HiPlot server with experiment name

You can also add ``tags=yes`` as a query string parameter to include tags in

the output, for example::

   mlflow://name/experiment-name?tags=yes

You can also use the `multiple experiments

`_

loading syntax. Either the dictionary format (to define your own labels)::

   multi://{

       "first-experiment": "mlflow://id/1",

       "another-experiment": "mlflow://name/another-experiment?tags=yes"

   }

or list format::

   multi://[

       "mlflow://id/1",

       "mlflow://name/another-experiment?tags=yes"

   ]

.. image:: images/server_multi.png

   :alt: Multiple experiments in HiPlot server