This post is co-authored by Angel Diaz, Vice President, Developer Technology and Advocacy; Ruchir Puri, Chief Architect, IBM Watson; and William Lobig, Vice President, IBM Watson Data Platform Development

IBM has been exploring artificial intelligence and machine learning technologies and techniques for decades. From IBM`s victory over Garry Kasparov in chess in 1997 to Watson winning a $1 million prize against Jeopardy champions in 2011, the hunt has always been on for new challenges. And now, as AI technologies and deep learning explodes, data scientists and AI developers are facing a whole new set of challenges.

While the onset of machine learning engines like TensorFlow, PyTorch, and Caffe2 has enabled AI developers to do scalable machine learning using computational data flow graphs, it’s also given rise to multiple machine learning model zoos and formats, each with its own set of syntax and semantics and spread across multiple websites.

We believe machine learning models are at the heart of this AI-fueled renaissance. AI developers need a credible source for these models that not only gives them a one-stop marketplace for multiple free and community-sourced frameworks, but also the metadata and provenance needed to ensure that the origin and quality of these community models is understood. This includes information on the data sets the models were trained on, their accuracy rate, parameters, and so on. In addition, AI developers want to try these models quickly for inferencing and prediction with their own input data. Provenance is especially important to enterprise developers to address a range of concerns, from safety to bias.

Announcing Model Asset eXchange

We are excited to announce one of the first initiatives in this space — an open source enterprise Model Asset eXchange, or “MAX”. MAX is a models app store that aims to ignite a community of data scientists and AI developers, enabling them to easily discover, rate, and deploy machine learning models. MAX is a one-stop exchange for data scientists and AI developers to consume models created using their favorite machine learning engines, like TensorFlow, PyTorch, and Caffe2, and provides a standardized approach to classify, annotate, and deploy these models for prediction and inferencing, including an increasing number of models that can be deployed and customized in IBM’s recently announce AI application development platform, Watson Studio.

This message is very much echoed by Jim Zemlin, Executive Director of The Linux Foundation. “IBM’s release of Model Asset eXchange gives developers a new source for deep learning models. We look forward to seeing the innovation unlocked from the MAX community through collaboration with our forthcoming Acumos AI community.”

Model Asset Exchange

Experiment with MAX models on your system

For trying locally on your machine, MAX offers Docker-based deployment and wraps models with a REST API, which can be used for model inferencing and consumption. The Model API server also provides an interactive Swagger documentation page. When users launch the build, model weights and graphs are downloaded from an object storage into the user’s Docker environment. When the API server is running, you can explore the API and also create test requests.

Train and deploy MAX Models for production workloads using Watson Studio

If you find a model you want to leverage in our recently announced Watson Studio platform, we are packaging a subset of those models into the format specified by newly launched Watson Studio deep learning capabilities. Going forward, an increasing number of models will be made available for use in Watson Studio. Deep Learning as a Service within Watson Studio embraces a wide array of popular open source frameworks like TensorFlow, Caffe, and PyTorch, and offers them truly as a cloud native service on IBM Cloud, with tools to build and manage models throughout their lifecycle. We currently have tutorials for customizing and running MAX community models in Watson Studio. We continue to refine and automate the MAX automation with Studio to provide you the most seamless experience possible. Stay tuned!

Watson Studio

Join MAX and help ignite the community!

We’re just getting started! We’re at the cusp of an AI renaissance, and we believe MAX will help you create and curate best-of-breed models and ignite a community of data scientists and AI developers. Join us and help fuel this renaissance!

Related Links

1 comment on"Igniting a community around deep learning models with Model Asset eXchange (MAX)"

  1. Dr. Thorsten Gressling July 08, 2018

    Really great. How about evolving this with an economy approach to trade this models, maybe introducing billing into a deployed model (via IBM API connect). Another idea: Also adding annotators from IBM text analytics catalog ( as this page still adresses WEX world and contains relevant assets.

Join The Discussion

Your email address will not be published. Required fields are marked *