Data-Driven Offline Optimization For Architecting Hardware Accelerators

Abstract

Industry has gradually moved towards application-specific hardware accelerators in order to attain higher efficiency. While such a paradigm shift is already starting to show promising results, designers need to spend considerable manual effort and perform a large number of time-consuming simulations to find accelerators that can accelerate multiple target applications while obeying design constraints. Moreover, such a "simulation-driven" approach must be re-run from scratch every time the set of target applications or design constraints change. An alternative paradigm is to use a "data-driven", offline approach that utilizes logged simulation data, to architect hardware accelerators, without needing any form of simulations. Such an approach not only alleviates the need to run time-consuming simulation, but also enables data reuse and applies even when set of target applications changes. In this paper, we develop such a data-driven offline optimization method for designing hardware accelerators, dubbed PRIME, that enjoys all of these properties. Our approach learns a conservative, robust estimate of the desired cost function, utilizes infeasible points, and optimizes the design against this estimate without any additional simulator queries during optimization. PRIME architects accelerators -- tailored towards both single and multiple applications -- improving performance upon state-of-the-art simulation-driven methods by about 1.54x and 1.20x, while considerably reducing the required total simulation time by 93% and 99%, respectively. In addition, PRIME also architects effective accelerators for unseen applications in a zero-shot setting, outperforming simulation-based methods by 1.26x.

First two authors contributed equally; published at ICLR 2022

References in corpus (17)

• Practical Bayesian Optimization of Machine Learning Algorithms
• Language Models are Few-Shot Learners
• Sequence Transduction with Recurrent Neural Networks
• Conservative Q-Learning for Offline Reinforcement Learning
• Scalable Bayesian Optimization Using Deep Neural Networks
• Bayesian Optimization with Unknown Constraints
• Chip Placement with Deep Reinforcement Learning
• Mind Mappings: Enabling Efficient Algorithm-Accelerator Mapping Space Search
• Accelerator-aware Neural Network Design using AutoML
• Marvel: A Data-centric Compiler for DNN Operators on Spatial Accelerators
• Model Inversion Networks for Model-Based Optimization
• Autofocused oracles for model-based design
• Apollo: Transferable Architecture Exploration
• Preventing Posterior Collapse with delta-VAEs
• Conservative Objective Models for Effective Offline Model-Based Optimization
• Learned Hardware/Software Co-Design of Neural Accelerators
• Offline Model-Based Optimization via Normalized Maximum Likelihood Estimation