Virtualization Technology News and Information
Article
RSS
Comet ML Debuts Collaborative Workspaces for Data Science and MLOps Teams

Comet ML announced updates to Comet Workspaces, including the introduction of Interactive Reports, ML Templates and the industry's first workflow for proactively considering carbon emissions as part of the machine learning process. Today's updates further empower data scientists and teams to build better models faster, while ensuring that organizations can continue to operate in an environmentally responsible manner.

One of the most pressing challenges facing machine learning and artificial intelligence teams today is the difficulty of delivering quality, trained models from experiment to production. Recent studies have shown that as many as 55 percent of companies never take their models to production, and nearly 87 percent of machine learning projects fail. Despite having cutting-edge technologies to build machine learning models, tools that enable enterprise machine learning teams to implement a consistent MLOps process, workflows and reporting have lagged behind.

"While much has been said about the potential of AI and machine learning for business, a majority of that innovation hasn't translated into value yet," said Gideon Mendels, Co-founder and CEO, Comet. "The challenges range from lack of defined workflow and processes to inability to collaborate and share insights across teams. That's why MLOps has arisen as a key concept-defining the people, processes and technologies that will drive wide-spread success with machine learning and AI at scale. But this must also be done responsibly, in a way that considers and addresses significant computing requirements and emissions."

Comet Enterprise automates experiment and model management, automatically tracking data sets, code changes, experimentation history, and models all at scale. One key component is Comet Workspaces. Since its inception, Comet Workspaces has provided a one-stop-shop for data science and machine learning teams to consolidate, control, and collaborate on machine learning projects and experiments.

With additions of Interactive Reports, ML Templates, and the CodeCarbon Panel, Comet Workspaces deliver an integrated approach to managing ML teams and development-from planning to delivering to reporting the status and results of machine learning projects-all within the context of environmental impact.

"CodeCarbon is an open source tool that estimates the amount of carbon dioxide (CO2) produced by computing resources both locally and on the cloud," said Sasha Luccioni, postdoctoral researcher at Mila. "Comet ML has been a great partner as we've worked together to help researchers and developers understand and reduce emissions. With Comet ML's new CodeCarbon Panel and workflow, developers will be able to incorporate those decisions directly into the experiment and model training process."

Today's updates include:

  • Interactive Reports -- share and report the results of your experiments internally or externally via an intuitive and fully interactive user interface (UI) which supports fully customizable code panels. Visit the Comet Report Library.
  • ML Templates -- use pre-built interactive templates that accelerate planning and reporting for the most machine learning common needs - such as project initiation and business stakeholder reports.
  • CodeCarbon Panel & Emissions Tracking Template -- interactive panel to proactively incorporate and consider the carbon emissions of your projects, ensuring that models can be optimized while being environmentally responsible.

"Our goal at Comet is to solve the challenges that organizations face when getting from experimentation to production," continued Mendels. "The biggest challenge AI teams have today isn't DevOps or infrastructure but actually building models that meet the business KPI. That's why our focus is on making experimentation and model management streamlined and predictable as you go through the research process. The more automated and intuitive you can make the process by delivering thoughtful and pre-built tools for data scientists, the more likely it becomes that organizations can drive real value with machine learning and AI."

The CodeCarbon Panel and new Emissions Tracking Template were made possible by CodeCarbon, a joint initiative between Mila, BCG GAMMA, Haverford College, and Comet ML.

Published Monday, December 07, 2020 10:43 AM by David Marshall
Filed under:
Comments
There are no comments for this post.
To post a comment, you must be a registered user. Registration is free and easy! Sign up now!
Calendar
<December 2020>
SuMoTuWeThFrSa
293012345
6789101112
13141516171819
20212223242526
272829303112
3456789