Legateboost: A Simple Gradient Boosting Library for Supercomputers
Monday, May 13, 2024 3:00 PM to Wednesday, May 15, 2024 4:00 PM · 2 days 1 hr. (Europe/Berlin)
Foyer D-G - 2nd floor
Research Poster
ML Systems and ToolsScalable Application Frameworks
Information
Poster is on display and will be presented at the poster pitch session.
We present the machine learning library Legateboost, an implementation of gradient boosting based on the legate/legion/realm parallel programming framework. Legateboost is a highly extensible boosting framework, implementing decision tree models, kernel ridge regression, linear models and neural networks that can be run on GPU or CPU clusters. Existing software packages such as XGBoost or LightGBM require tens of thousands of lines of carefully tuned C++ code to achieve high levels of parallel performance and implement state of the art features. Legateboost's implementation within the legate parallel programming framework is dramatically simpler, more extensible and more maintainable, yet providing comparable performance compared to existing libraries.
We present the machine learning library Legateboost, an implementation of gradient boosting based on the legate/legion/realm parallel programming framework. Legateboost is a highly extensible boosting framework, implementing decision tree models, kernel ridge regression, linear models and neural networks that can be run on GPU or CPU clusters. Existing software packages such as XGBoost or LightGBM require tens of thousands of lines of carefully tuned C++ code to achieve high levels of parallel performance and implement state of the art features. Legateboost's implementation within the legate parallel programming framework is dramatically simpler, more extensible and more maintainable, yet providing comparable performance compared to existing libraries.
Format
On-site
