Graph neural networks with configuration cross-attention for tensor compilers

TL;DR

This paper introduces TGraph, a neural network architecture with configuration cross-attention that efficiently screens for optimal tensor configurations in computational graphs, enhancing inference speed and reducing carbon emissions.

Contribution

The paper presents TGraph, a novel neural graph architecture that acts as an AI tensor compiler, outperforming traditional heuristics-based methods in configuration screening.

Findings

Improved mean Kendall's τ from 29.8% to 67.4%.

Potential to reduce CO₂ emissions by over 50% in AI data centers.

Demonstrated effectiveness on TPU graph configurations.

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 the traditional heuristics-based compilers. The proposed solution improves mean Kendall's $\tau$ across layout collections of TpuGraphs from 29.8% of the reliable baseline to 67.4% of TGraph. We estimate the potential CO$_2$ 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.