Description
In this talk we’ll present a toolchain for conducting and organizing computational experiments consisting of Sacred -- the core framework -- and two supporting tools: Labwatch and Sacredboard. These tools are agnostic of the methods and libraries used, and instead focus on solving the universal everyday problems of running computational experiments like reproducing results, bookkeeping, tuning hyperparameters, and organizing and analyising the runs and results. Attendees will be introduced to the core features of these libraries, and learn how they can form the basis for an effective and efficient workflow for machine learning research.