Extending STAC-ML with Gradient Boosted Tree Models on an ARM-Architecture Cloud Instance

Type: STAC Report

Specs: STAC-ML*

Stack under test:

  • STAC-ML Markets (Inference) Reference Implementation for GBT Models
  • Python 3.12.3; ONNX runtime 1.21.0; NumPy 2.2.3
  • Ubuntu Linux 24.04.1 LTS
    • Numerous realtime tuning options and configurations
  • AWS EC2 c8g.metal-24xl instance:
    • 1 x AWS Graviton4 (ARM Neoverse-V2) Processor
    • 192 GiB memory: 12 x 16 GiB DDR5 DIMM @ 5600 MT/s
    • 600 GiB AWS EBS volume

This STAC Report presents the findings of a Proof of Concept benchmark focused on Gradient-Boosted Tree (GBT) inference for real-time market data analysis. This study evaluates latency performance across three GBT models with varying complexities, comparing X86 and ARM architectures on AWS bare-metal instances using the ONNX runtime.

Please log in to see file attachments. If you are not registered, you may register for no charge.

The STAC-ML Working Group develops benchmark standards for key machine learning (ML) workloads in finance. These benchmarks enable customers, vendors, and STAC to make apples-to-apples comparisons of techniques and technologies.