Skip to main content

Le Board, a simple leaderboard generator for ranking your algorithms experiments

Project description

Le Board (beta)

Le Board, a simple leaderboard generator for ranking your algorithms experiments.

Install

Install with pip and configure with init, linking a Google account is the recommended choice.

$ pip install leboard
$ leboard init

sketch of leboard init

Choose an option for Data Handler (1):

  1. Google Firebase (recommended)

  2. Local file system (no shared options)

if option 1)

if option 2)

  • Choose a path to store your boards (~/.leboard):

Usage

import leboard

letask = leboard.TaskBoard("MNIST")

# other members will have access to the MNIST task too
# (restrict a members definition only when a task is set)
letask.members(["rob@gmail.com", "jane@mail.com"])

# do your stuff
# ...

experiment = letask.Entry()

experiment.set("accuracy", accuracy)
experiment.set("loss", loss)
experiment.set("parameters", params, hide=True)
experiment.set("confusion_image", confusion_image)

# commit will save to DataHandler available/configured
# (proposed is local simple file storage, and shared google firebase)
experiment.commit()

Leaderboard

Show all experiments with links to detailed resources, this can open a page hosted in firebase.

  • keep experiment data in order of creation

letask.leaderboard("MNIST", "accuracy")

Advanced?

Resource Typing?

Entry can have different types, this can be usefull for handling resources like image, or HTML pages, to show and interact with them.

from leboard import types as lbt

experiment.set("accuracy", accuracy)
experiment.set("loss", loss)
experiment.set("confusion_image", lbt.Image(confusion_image))
experiment.set("notebook_html", lbt.HTML(notebook_html))

Committing an Entry could automatically save a copy of the current notebook state to the DataHandler.

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

leboard-0.1.1.tar.gz (5.9 kB view hashes)

Uploaded Source

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page