Ludwig

A declarative deep learning framework

Ludwig is a deep learning framework that allows training models and use them for prediction by writing simple declarative configurations. It is built on top of PyTorch and it uses a data-type based abstraction that allows for really large number of applications to be developed (from NLP to computer vision to time series forecasting to regression and categorization, question answering, dialogue systems and so on). Thanks to the declarative nature of its configuration files, Ludwig allows for extremely fast prototyping and iteration on models. It is usable both by novices to train deep learning models without knowing all the details of PyTorch and of deep learning in general, but also allows experienced users to be much more productive, reducing tasks that would require months to minutes. I developed it over the course of two years (2017 and 2018) at Uber AI, with the help of Yaroslav Dudin (who contributed code reorganizations and image feature mostly) and Sai Sumanth Miryala (who developed the test suite). On the 11th of February 2019 it was released as open source.
The code was released on GitHub together with the accompanying documentation and a blogpost that introduces it in more detail.
An additional blogpost was released on 24th July 2019 about the release of v0.2. Since then several improved versions have been release and announced on the (link https://predibase.com/blog text: Predibase blog).

It has been used in more than ten projects inside Uber, consisting of text classification, entity recognition, image classification, information extraction, dialogue systems, language generation, timeseries forecasting, and many more. Ludwig was also highlighted as one of the main contributions of Uber AI in its 2019 recap blogpost.

Here you can read the paper describing it and the slides of the presentation.
Download Paper Download Slides

An introductory video was also shot and announced on the UberEng blog.

A more in depth presentation, delivered at the Uber Open Source Summit Sofia 2019 is also available.

The presentation I gave when proposing the project for incubation at the Linux Foundation AI is available here (the Ludwig presentation starts at 6:20).

News

FIrst page on HackerNews
VentureBeat
VentureBeat on Ludwig v0.2
Computer Business Review
No 1 position on toutiao (chinese news app) as it was recommended by jiqizhixin (most famous chinese AI platform)
PureAi
InfoQ
Neurohive
Weixin
Sohu
Mc.ai
Devcalss
Synced
CTOVision
AITopics
Jaxenter
Analytics India Magazine
Hackaday
Qiita post
https://www.marktechpost.com/2020/10/16/uber-open-sources-ludwig-v0-3-the-third-update-to-its-code-free-deep-learning-toolbox-built-on-top-of-tensorflow/ text Mrktechpost
KDNuggets
MCAI
Science Wiki
https://www.machine-learning.news/example/21867140 text ML-News

Full articles

Science article on no-code machine learning featuring Ludwig
Forbes
Ludwig wins InfoWorld's BOSSIE award 2019 (screenshot for the non-subscribers)
Computer Business Review
Towards Data Science Introductory article
Another Towards Data Science Introduction article
Towards Data Science Ludwig applications article
Towards Data Science on the impressive uber deep learning stack
Towards Data Science Codeless Deep Learning Pipelines with Ludwig and Comet.ml
Comet.ml partners with Uber on Ludwig
Domino Data Lab Ludwig practitioner guide
Odbms.org: Interview to Piero and section about Ludwig
puntoinformatico.it: Interview to Piero and section about Ludwig
In depth article on dev.to
Hackernoon's 10 Must-Try Open Source Tools for Machine Learning
Algoritmia's Six open-source machine learning tools you should know
Coveo's writeup on how they used Ludwig for query suggestions
Automated Intent Classification Using Ludwig - Part 1
Automated Intent Classification Using Ludwig - Part 2
How to evaluate content quality with Ludwig
Identifying Pneumonia in chest X-rays usingLudwig
Codemotion Magazine Interview
zdnet article about Chris Ré's presentation on Software 2.0 which includes Ludwig

Podcast interviews

InfoQ Podcast with Wesley Reisz

Videos

Siraj Raval’s Deep Learning with No Code

Siraj Raval’s AutoML

Gilbert Tanner’s Uber Ludwig Tutorial #1 - What is Ludwig and how does it work

Suneel Marthi's Beaming Deep Learning with Ludwig

InfoQ.ai recording

Recording of 1h presentation at QCon.ai 2019 in San Francisco

Invited Presentations


Collaborators

Yarosalv Dudin

Sai Sumanth Miryala