Glossary term
Glossary term
Training and Fine-Tuning
A Python-first configuration library that sets the values of functions and classes without invasive code or infrastructure. In the case of Pax—and other ML codebases—these functions and classes represent models and training hyperparameters.
Fiddle assumes that machine learning codebases are typically divided into:
Library code, which defines the layers and optimizers.
Dataset "glue" code, which calls the libraries and wires everything together.
Fiddle captures the call structure of the glue code in an unevaluated and mutable form.
Created for this library
A research team uses Fiddle to express experiment configurations in Python so configuration overrides remain typed and discoverable.
An ML platform team standardizes on Fiddle so experiment configurations stay refactor-safe across the codebase.
A research lab adopts Fiddle to express hierarchical configurations cleanly across many experiments without manual YAML merges.
Definition source: Google for Developers Machine Learning Glossary | Creative Commons Attribution 4.0 License