Hey all, Caliban author here. Caliban is a tool that helps ML researchers launch and track their numerical experiments in an isolated, reproducible computing environment. We built this at Google to try and remove the heavy friction that always accompanies a move from interactive development -> local model training -> Cloud.<p>Caliban will dynamically build Docker containers for your code, without you having to know anything about Docker, and then execute that container for you in a variety of different modes.<p>Full docs live here: <a href="https://caliban.readthedocs.io/en/latest/" rel="nofollow">https://caliban.readthedocs.io/en/latest/</a>
We wrote a small tutorial here: <a href="https://github.com/google/caliban#getting-started-with-caliban" rel="nofollow">https://github.com/google/caliban#getting-started-with-calib...</a><p>Hope this helps folks out there with their ML research! Happy to answer any questions that come up.