I am the lead developer of Oríon, an open-source framework developed at Mila for distributed black-box optimization. Its purpose is to serve as a meta-optimizer for machine learning models and training, as well as a flexible experimentation platform for large-scale asynchronous optimization procedures.
For my research projects I developed a framework to automatically set up all available clusters and deploy my experiments. With Mahler, I can wrap the super computer schedulers to gain more control over my workflow, better resiliency, and better automation.
Track is a generic logger backend developed at Mila that can serve as a bridge between different logging libraries for machine learning. The goal is to provide a backend for Oríon so that users can use their own logger without any modification and simply configure Oríon to use the proper bridge with Track.
Kleiṓ is an experiment manager, a logging journal of all data describing your experiments. Its purpose is to provide an automatic tool to log extensive environment information, including a script’s code version, system specification, and script configuration.