Announcing Collimator 2.0

We’re very excited to announce that Collimator version 2.0 has been released! This release marks a major overhaul to our modeling and simulation engine, laying the groundwork for significant performance and functionality improvements. Feel free to jump right in and try it out. If you run into any issues, check out our tips for migrating your existing models, hit us up in the support chat, or send an email to help@collimator.ai.

Here’s an overview of what’s new:

JAX support

We have fully integrated JAX into our backend, much improving simulation run times for large models, parallel simulations, and optimization tasks. JAX also adds auto-differentiation to the Collimator toolkit, allowing us to better integrate machine learning workflows, so keep an eye on this blog for more updates on that front. JAX is amazing, but if you’re not ready to migrate your Python code, you can choose NumPy as your numerical backend on a case-by-case basis—look for the setting in your Python block properties. You can find out more information about JAX here.

‍

PyCollimator update

Not only have we overhauled our modeling and simulation engine in the cloud application, we’ve also massively updated our Python package, PyCollimator. In addition to being able to connect to Collimator remotely to get information and run simulations, you can also run Collimator locally: build models, run simulations, perform optimization tasks, and more, all in your local Python environment. If you have advanced use cases, want to run things offline, or simply prefer code, now you can do it with PyCollimator. Oh, and since it’s Collimator, PyCollimator is also powered by JAX. Find out more here, and explore some sample notebooks.

…and lots more

We’ve also made a number of smaller improvements throughout the app: usability improvements to the Dashboard, better reporting of status for long-running simulations, clarified simulation settings, new blocks, and more.

We’re just getting started with the updates, so keep your eyes on this blog for even more news. Happy modeling!