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 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.
@article{arxiv.2405.16623,
title = {Graph neural networks with configuration cross-attention for tensor compilers},
author = {Dmitrii Khizbullin and Eduardo Rocha de Andrade and Thanh Hau Nguyen and Matheus Pedroza Ferreira and David R. Pugh},
journal= {arXiv preprint arXiv:2405.16623},
year = {2024}
}