Unknown

Dataset Information

0

Graph neural networks with configuration cross-attention for tensor compilers.

ABSTRACT: With the recent popularity of neural networks comes the need for efficient serving of inference workloads. A neural network inference workload can be represented as a computational graph with nodes as operators transforming multidimensional tensors. The tensors can be transposed and/or tiled in a combinatorially large number of ways, some configurations leading to accelerated inference. We propose TGraph, a neural graph architecture that allows screening for fast configurations of the target computational graph, thus representing an artificial intelligence (AI) tensor compiler in contrast to traditional heuristic-based compilers. The proposed solution improves mean Kendall's τ across layout collections of TpuGraphs from 29.8% of the reliable baseline to 67.4% of TGraph. We estimate the potential CO2 emission reduction associated with our work to be equivalent to over 50% of the total household emissions in the areas hosting AI-oriented data centers.

SUBMITTER: Khizbullin D 

PROVIDER: S-EPMC12406497 | biostudies-literature | 2025

REPOSITORIES: biostudies-literature

Json Xml
altmetric image

Publications

Graph neural networks with configuration cross-attention for tensor compilers.

Khizbullin Dmitrii D   de Andrade Eduardo Rocha ER   Nguyen Thanh Hau TH   Ferreira Matheus Pedroza MP   Pugh David R DR  

Frontiers in artificial intelligence 20250820

With the recent popularity of neural networks comes the need for efficient serving of inference workloads. A neural network inference workload can be represented as a computational graph with nodes as operators transforming multidimensional tensors. The tensors can be transposed and/or tiled in a combinatorially large number of ways, some configurations leading to accelerated inference. We propose TGraph, a neural graph architecture that allows screening for fast configurations of the target com  ...[more]

Similar Datasets

Unknown Graph Attention Site Prediction (GrASP): Identifying Druggable Binding Sites Using Graph Neural Networks with Attention.
| S-EPMC10402091 | biostudies-literature
Unknown Graph Attention Site Prediction (GrASP): Identifying Druggable Binding Sites Using Graph Neural Networks with Attention.
| S-EPMC11182664 | biostudies-literature
Unknown Predicting functional effect of missense variants using graph attention neural networks.
| S-EPMC10361701 | biostudies-literature
Unknown Integrating scRNA-seq and scATAC-seq with inter-type attention heterogeneous graph neural networks.
| S-EPMC11725394 | biostudies-literature
Unknown Cross-attention PHV: Prediction of human and virus protein-protein interactions using cross-attention-based neural networks.
| S-EPMC9546503 | biostudies-literature
Unknown Evaluating graph neural networks under graph sampling scenarios.
| S-EPMC9044246 | biostudies-literature
Unknown Evaluating the generalizability of graph neural networks for predicting collision cross section.
| S-EPMC11363525 | biostudies-literature
Unknown Predicting ADMET Properties from Molecule SMILE: A Bottom-Up Approach Using Attention-Based Graph Neural Networks
| S-EPMC11207804 | biostudies-literature
Unknown Predicting Toxicity toward Nitrifiers by Attention-Enhanced Graph Neural Networks and Transfer Learning from Baseline Toxicity.
| S-EPMC11912336 | biostudies-literature
Unknown Highly accurate and large-scale collision cross sections prediction with graph neural networks.
| S-EPMC10319785 | biostudies-literature