Work

Batch Size and Calling Interface in Gradient Boosted Tree Serving: replication package

Year: 2026

Type: software

Abstract: Replication package for a study measuring per-row inference latency of LightGBM, XGBoost and CatBoost across nine serving interfaces and thirteen batch sizes at three model sizes, on a single thread. The full grid was executed twice on separately allocated cloud instances, giving 702 measurements per run.

Source: Zenodo (CERN European Organization for Nuclear Research)

DOI: 10.5281/zenodo.22373725

Authors: Abu-Zanona, Marwan, Aldairi, Maryam, Rahamatalla, Lamia Hassan, Almousa, Najla Abdulaziz, Abubakr, Hebah Abdullah, Zia, Muhammad Ahmad, Ikram, Atif

Institutions: King Faisal University, University of Lahore, Universiti Malaysia Terengganu

Language: English

Topic: Parallel Computing and Optimization Techniques

Subfield: Hardware and Architecture

Field: Computer Science

Domain: Physical Sciences

Open Access status: green

OpenAlex ID: W7208818511

Cites:

Cited by:

Related to: 10