ASDL: A Unified Interface for Gradient Preconditioning in PyTorch
Kazuki Osawa, Satoki Ishikawa, Rio Yokota, Shigang Li, Torsten Hoefler
Code Available — Be the first to reproduce this paper.
ReproduceCode
- github.com/kazukiosawa/asdlOfficialIn paperpytorch★ 192
- github.com/kazukiosawa/asdfghjklpytorch★ 192
Abstract
Gradient preconditioning is a key technique to integrate the second-order information into gradients for improving and extending gradient-based learning algorithms. In deep learning, stochasticity, nonconvexity, and high dimensionality lead to a wide variety of gradient preconditioning methods, with implementation complexity and inconsistent performance and feasibility. We propose the Automatic Second-order Differentiation Library (ASDL), an extension library for PyTorch, which offers various implementations and a plug-and-play unified interface for gradient preconditioning. ASDL enables the study and structured comparison of a range of gradient preconditioning methods.