Experiments Tracking & Tuning¶
Research projects often involve many runs with different code versions, datasets, and hyperparameters. These tools help record, compare, and optimize those experiments, from logging metrics to running large hyperparameter searches on the clusters.
-
Track and compare experiments in a web dashboard on Comet.
-
Run hyperparameter searches in parallel across many cluster jobs.
-
Log metrics, system usage, and artifacts from training runs, and compare them in a web dashboard.
-
Monitor compute usage of Slurm jobs, and log these metrics.