Data

Dataset definitions and loading are handled by datamaestro-IR. XPMIR builds on top of those types with adapters for subsampling, folding, and transforming datasets to fit training and evaluation pipelines.