How to keep track of your model experiments?
Experiment Management - Infrastructure and Tooling
- Getting your models to perform well is a very iterative process. If you don’t have a system for managing your experiments, it quickly gets out of control.
- TensorBoard is a TensorFlow extension that allows you to easily monitor your model in a browser.
- Losswise provides ML practitioners with a Python API and accompanying dashboard to visualize progress within and across training sessions.
- Comet.ml is another platform that enables engineers and data scientists to efficiently maintain their preferred workflow and tools, track previous work, and collaborate throughout the iterative process.
- Weights & Biases is an experiment tracking tool for deep learning that allows you to (1) store all the hyper-parameters and output metrics in one place; (2) explore and compare every experiment with training/inference visualizations; and (3) create beautiful reports that showcase your work.
- MLflow is an open-source platform for the entire machine learning lifecycle started by Databricks. Its MLflow Tracking component is an API and UI for logging parameters, code versions, metrics, and output files from your model training process.