ApproxABFT: Approximate Algorithm-Based Fault Tolerance for Vision Transformers. Vision Transformers (ViTs) with outstanding performance becomes a popular backbone of deep learning models for the main-stream vision tasks including classification, object detection, and segmentation.
Feb 21, 2023
This work uses the inherent fault-tolerance of ViTs and proposes an approximate ABFT, namely ApproxABFT, to invoke the error recovery procedure only when ...
ApproxABFT: Approximate Algorithm-Based Fault. Tolerance for Vision ... algorithm-based fault tolerance (ABFT) [14] strategy based on checksum with much ...
According to our experiments, ApproxABFT reduces the computing overhead by 25.92% to 81.62% and improves the model accuracy by 2.63% to 72.56% compared to the ...
People also ask
What is algorithm-based fault tolerance?
What is a vision transformer for image prediction?
ApproxABFT: Approximate algorithm-based fault tolerance for vision transformers. ... ApproxABFT: Approximate algorithm-based fault tolerance for vision ...
Though tremendous progress has been made, existing ABFT approaches for one-sided matrix decomposition on heterogeneous systems with GPUs have following ...
This article proposes several systematic ABFT schemes based on checksum techniques and analyzes their fault protection ability and runtime thoroughly, ...
Feb 23, 2023 · According to our experiments, ApproxABFT reduces the computing overhead by 25.92% to 81.62% and improves the model accuracy by 2.63% to 72.56% ...
Soft Error Reliability Analysis of Vision Transformers. IEEE Trans ... ApproxABFT: Approximate Algorithm-Based Fault Tolerance for Vision Transformers.
Soft Error Reliability Analysis of Vision Transformers. IEEE Trans ... ApproxABFT: Approximate Algorithm-Based Fault Tolerance for Vision Transformers.