There is a newer version of the record available.

Published November 25, 2024 | Version v1
Software Open

ATTNChecker: Highly-Optimized Fault Tolerant Attention for Large Language Model Training

Authors/Creators

  • 1. ROR icon University of Alabama at Birmingham
  • 2. ROR icon Pacific Northwest National Laboratory
  • 3. ROR icon William & Mary

Description

Artifacts for the "ATTNChecker: Highly-Optimized Fault Tolerant Attention for Large Language Model Training" accepted at PPoPP 2025.

Files

ATTNChecker.zip

Files (372.2 MB)

Name Size
md5:933b79e8e677de69f57b0f99bbf63d9e
372.2 MB Preview Download