GRAM: Fast Fine-tuning of Pre-trained Language Models for Content-based Collaborative Filtering

TL;DR

GRAM introduces a gradient accumulation technique that accelerates training of pre-trained language model-based content-based collaborative filtering, reducing resource consumption while maintaining performance.

Contribution

It proposes Single-step and Multi-step GRAM methods that significantly speed up training with less memory, enabling efficient fine-tuning of PLM-based CCF models.

Findings

Up to 146x training speedup on multiple datasets.

Maintains theoretical equivalence to end-to-end training.

Reduces GPU memory usage substantially.

Abstract

Content-based collaborative filtering (CCF) predicts user-item interactions based on both users' interaction history and items' content information. Recently, pre-trained language models (PLM) have been used to extract high-quality item encodings for CCF. However, it is resource-intensive to train a PLM-based CCF model in an end-to-end (E2E) manner, since optimization involves back-propagating through every content encoding within a given user interaction sequence. To tackle this issue, we propose GRAM (GRadient Accumulation for Multi-modality in CCF), which exploits the fact that a given item often appears multiple times within a batch of interaction histories. Specifically, Single-step GRAM aggregates each item encoding's gradients for back-propagation, with theoretic equivalence to the standard E2E training. As an extension of Single-step GRAM, we propose Multi-step GRAM, which increases the gradient update latency, achieving a further speedup with drastically less GPU memory. GRAM significantly improves training efficiency (up to 146x) on five datasets from two task domains of Knowledge Tracing and News Recommendation. Our code is available at https://github.com/yoonseok312/GRAM.