weight

Definition

A learned numerical parameter — collectively, the weights are the model. "Open-weight" means the trained parameters are downloadable and runnable locally, a meaningfully different thing from open source, since training data and code are usually not included. The word has a second, unrelated sense: the coefficient placed on a term in a loss function or on a sample in a dataset, as in loss weighting.

Explainers using this term

• Paper Walkthrough: WHALE — Training the Weights and the Harness in TurnsWHALE: A Simple Recipe for Joint Harness-Weight Optimization