This blog post was originally published at ClearML’s website. It is reprinted here with the permission of ClearML.
About ClearML, Allegro Trains, and what’s ahead for the MLOps space in 2021
We have three big announcements to our community today, and I wanted to talk to you about them: One, Allegro Trains is changing its name, two, we’re adding a completely new way to use Trains, and three, we’re announcing a bunch of features that make Trains an even better product for you!
The new name for Allegro Trains is ClearML. If you were paying attention last month, we asked the community at large for their opinion on a number of different possible names for Allegro Trains, and we were happy to see that you liked ClearML best, because we did, too.
A name change by itself isn’t much to get excited about. We are changing our name in conjunction, with adding a completely new way to use our ML-Ops toolchain. We are launching a managed service offering for the ClearML open source community. From today, users who want to use ClearML (the artist formerly known as Trains) do not have to self host the ClearML server, but can opt to use our ClearML managed services instead.
Why are we changing the name?
Over the past year, ClearML’s community has grown in leaps and bounds. Today over one thousand organizations globally are using ClearML on a regular basis to manage their AI workflows. We are also seeing more and more of our growing community getting more involved and contributing code and bug fixes. We have a highly involved core community centered around our Slack channel, and we are the leading solution for experiment management & ML Ops orchestration & scheduling.
Of course, we, and our community, loved our former name, Trains. Our community has enjoyed playing word games with “trains” as in training ML / DL models, “trains” as in the transportation vehicle. We also saw people write trAIns to highlight the “AI” part.
However, many people have said it was hard to find us. Search engines assume “trains” is the verb related to training and not the product.
And so, with great sadness we decided we need to part with our former name, with ClearML becoming the clear winner by our community (many of us at Allegro AI are dads, so you’ll have to forgive the Dad Joke Energy here. :D).
Why are we launching a managed service for ClearML?
Organizations from some 50 countries globally have set up and manage ClearML servers. For data science teams with eng/IT support, it’s an easy 20min process. However, oftentimes setting up servers in organizations – even if it’s open source code – requires jumping through IT hoops. And individual practitioners don’t always have the setup for a machine to run a server. We want ClearML to be not just the best toolchain for ML-Ops data science professionals, we also want it to also be the easiest to get started with.
The ClearML community service provides the same functionality as the open source server one can install and self manage, only managed by us as a multi-tenant PaaS / SaaS offering.
Obviously, unlike a self managed server, we have had to place some quotas around the free service tier, since, at the end of the day, it costs us real money to keep this service up and running daily.
But, your data is your data. Your experiments are yours. You can always take your data and models and move them to your own self managed ClearML server. And since it is open source and backed by a flourishing community, you can trust it’ll always be there for you, and always be supported by the larger community.
ClearML Data Hub and More
The last thing we’re announcing are some very powerful features we’ve added recently.First, we’ve launched a whole set of functionality to enable data management, versioning, data lineage and tracking. Our users can now get experiment management and tracking, data versioning and orchestration plus automation, all in a single highly-integrated suite of tools. Or, they can use some of the functionality and easily integrate through our powerful RestAPI to their tool of choice.
Second, we’ve launched remote Jupyter notebooks and dev environments. Users of ClearML can now set up their dev environments on remote machines. Yes, you can debug your work all on a remote machine!
What’s next for ClearML?
We’ve gotten where we are today thanks to you. We’ve added dozens of features, thousands of new users, and many releases, and along the way we have discovered we’re the preferred ML Ops tool for thousands of organizations around the world. We’ve added a slew of new features today and remain at the forefront of the ML Ops field.
Moving forward we have two missions we want to stick to with ClearML:
- Be the leading open source ML Ops tool for the most demanding professionals
- Be the easiest to try, onboard, and scale up for data science teams of all sizes
As you watch what we do in 2021, you’ll hopefully notice we’re focusing on that.
As always, we are driven by our community. If you like what we’re doing, don’t like what we’re doing, or if you’d like to contribute to our code and make ClearML a better project for everyone, come join us on our slack channel and be part of the community.
2020 has been a year with a lot of changes and a lot of challenges for everyone around the world. We’ve been fortunate that we’ve been able to grow and thrive in a challenging year, and I hope we can take the velocity we’ve achieved so far and make tremendous strides in 2021.
Trains has left the station, but ClearML is just getting started.
We hope you’ll come along with us.
Nir Bar Lev
CEO, Allegro AI
