Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Liu, Andrew, Elaldi, Axel, Franklin, Nicholas T, Russell, Nathan, Atwal, Gurinder S, Ban, Yih-En A, Viessmann, Olivia
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2505.11580
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866912380804399104
author Liu, Andrew
Elaldi, Axel
Franklin, Nicholas T
Russell, Nathan
Atwal, Gurinder S
Ban, Yih-En A
Viessmann, Olivia
author_facet Liu, Andrew
Elaldi, Axel
Franklin, Nicholas T
Russell, Nathan
Atwal, Gurinder S
Ban, Yih-En A
Viessmann, Olivia
contents Invariant Point Attention (IPA) is a key algorithm for geometry-aware modeling in structural biology, central to many protein and RNA models. However, its quadratic complexity limits the input sequence length. We introduce FlashIPA, a factorized reformulation of IPA that leverages hardware-efficient FlashAttention to achieve linear scaling in GPU memory and wall-clock time with sequence length. FlashIPA matches or exceeds standard IPA performance while substantially reducing computational costs. FlashIPA extends training to previously unattainable lengths, and we demonstrate this by re-training generative models without length restrictions and generating structures of thousands of residues. FlashIPA is available at https://github.com/flagshippioneering/flash_ipa.
format Preprint
id arxiv_https___arxiv_org_abs_2505_11580
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Flash Invariant Point Attention
Liu, Andrew
Elaldi, Axel
Franklin, Nicholas T
Russell, Nathan
Atwal, Gurinder S
Ban, Yih-En A
Viessmann, Olivia
Machine Learning
Artificial Intelligence
Biomolecules
Invariant Point Attention (IPA) is a key algorithm for geometry-aware modeling in structural biology, central to many protein and RNA models. However, its quadratic complexity limits the input sequence length. We introduce FlashIPA, a factorized reformulation of IPA that leverages hardware-efficient FlashAttention to achieve linear scaling in GPU memory and wall-clock time with sequence length. FlashIPA matches or exceeds standard IPA performance while substantially reducing computational costs. FlashIPA extends training to previously unattainable lengths, and we demonstrate this by re-training generative models without length restrictions and generating structures of thousands of residues. FlashIPA is available at https://github.com/flagshippioneering/flash_ipa.
title Flash Invariant Point Attention
topic Machine Learning
Artificial Intelligence
Biomolecules
url https://arxiv.org/abs/2505.11580