wa-hls4ml: A Benchmark and Surrogate Models for hls4ml Resource and Latency Estimation

TL;DR

This paper introduces wa-hls4ml, a comprehensive benchmark and dataset for evaluating ML accelerator resource and latency estimation models, along with novel GNN and transformer-based surrogate models that accurately predict hardware metrics.

Contribution

It provides the first large-scale benchmark dataset and evaluates new GNN and transformer models for resource and latency prediction in FPGA-based ML accelerators.

Findings

Models predict latency and resources within several percent of actual synthesized values.

The dataset includes over 680,000 neural network configurations targeting Xilinx FPGAs.

Surrogate models perform well across diverse ML architectures from scientific domains.

Abstract

As machine learning (ML) is increasingly implemented in hardware to address real-time challenges in scientific applications, the development of advanced toolchains has significantly reduced the time required to iterate on various designs. These advancements have solved major obstacles, but also exposed new challenges. For example, processes that were not previously considered bottlenecks, such as hardware synthesis, are becoming limiting factors in the rapid iteration of designs. To mitigate these emerging constraints, multiple efforts have been undertaken to develop an ML-based surrogate model that estimates resource usage of ML accelerator architectures. We introduce wa-hls4ml, a benchmark for ML accelerator resource and latency estimation, and its corresponding initial dataset of over 680,000 fully connected and convolutional neural networks, all synthesized using hls4ml and targeting Xilinx FPGAs. The benchmark evaluates the performance of resource and latency predictors against several common ML model architectures, primarily originating from scientific domains, as exemplar models, and the average performance across a subset of the dataset. Additionally, we introduce GNN- and transformer-based surrogate models that predict latency and resources for ML accelerators. We present the architecture and performance of the models and find that the models generally predict latency and resources for the 75% percentile within several percent of the synthesized resources on the synthetic test dataset.