FeatureBox: Feature Engineering on GPUs for Massive-Scale Ads Systems

TL;DR

FeatureBox is an end-to-end GPU-based framework that accelerates feature extraction and training for large-scale online ad CTR models, significantly reducing training time by optimizing computation and memory management.

Contribution

It introduces a novel GPU-accelerated feature extraction pipeline with a layer-wise scheduling algorithm and dynamic memory management for industrial-scale CTR model training.

Findings

Achieves faster training times compared to traditional CPU-based methods.

Effectively reduces intermediate I/O and memory overhead.

Demonstrates significant performance improvements on real-world ad applications.

Abstract

Deep learning has been widely deployed for online ads systems to predict Click-Through Rate (CTR). Machine learning researchers and practitioners frequently retrain CTR models to test their new extracted features. However, the CTR model training often relies on a large number of raw input data logs. Hence, the feature extraction can take a significant proportion of the training time for an industrial-level CTR model. In this paper, we propose FeatureBox, a novel end-to-end training framework that pipelines the feature extraction and the training on GPU servers to save the intermediate I/O of the feature extraction. We rewrite computation-intensive feature extraction operators as GPU operators and leave the memory-intensive operator on CPUs. We introduce a layer-wise operator scheduling algorithm to schedule these heterogeneous operators. We present a light-weight GPU memory management algorithm that supports dynamic GPU memory allocation with minimal overhead. We experimentally evaluate FeatureBox and compare it with the previous in-production feature extraction framework on two real-world ads applications. The results confirm the effectiveness of our proposed method.