Comet.ml brings sharing ML models into the space age
Stop sharing machine learning models like it’s the stone age. Comet.ml intends to become the GitHub of ML, offering data scientists a simple, easy to use tool to share, compare, and optimize their machine learning models.
Machine learning has a lot going on for it, but sharing ML results is like living in the stone age. There’s only two good options: handing over a physical storage item such as a thumb drive, or emailing them like it’s 1995! Thanks to Comet.ml, a new tool that aims to be the GitHub of ML, all of this is in the past.
Comet.ml is a tool that allows data science teams and individuals to automatically track datasets, code changes, experimentation history, and production models for machine learning models. Thanks to its modern dashboard, users are able to efficiently train optimized models and reproduce efforts again and again.
This free tool is simple to use – data scientists can get started in seconds with tracking their training code. All a developer needs to do it to add the Comet.ml tracking code to their machine learning app of choice and run their experiments as normal. That’s it.
Data scientists can keep their preferred routine and stick with their existing workflows and development tools, since Comet.ml can be integrated with popular machine learning frameworks, like Keras API, TensorFlow, Scikit Learn, Pytorch, and more. Since Comet.ml is vendor neutral, developers can share access to other team members or publicly without any problems. (No more awkward dropbox shares!)
Comet.ml has a number of comparison tools, making it easier for developers and data scientists alike to choose the most efficient and optimized machine learning model. Devs can compare different experiments and see the variances in code, hyper-params, and other data points.
Helpfully, they have excellent Git integration. When everything is good and ready, developers can export finished projects over to GitHub and other git providers. Developers can create a pull request with the best machine learning model straight to GitHub. (However, Comet.ml is still working on an export function, which should arrive shortly.)
How to get Comet.ml
Interested? Comet.ml is now open for all developers looking to give it a whirl. Right now, there is a free tier that gives users one private project and unlimited public projects. Academics are eligible to access to the paid tiers. And, just like GitHub, there are a number of paid tiers for teams that want to keep their data private, each with a free 30 day trial.
For more information, head on over to Comet.ml and see if this tool is right for you!