GraSS: Scalable Data Attribution with Gradient Sparsification and Sparse Projection
This work addresses the computational bottleneck for researchers and practitioners using data attribution in large-scale machine learning models, representing an incremental improvement over existing methods.
The paper tackles the scalability problem of gradient-based data attribution methods by proposing GraSS, a gradient compression algorithm that leverages sparsity to reduce computational and memory costs, achieving up to 165% faster throughput on billion-scale models.
Gradient-based data attribution methods, such as influence functions, are critical for understanding the impact of individual training samples without requiring repeated model retraining. However, their scalability is often limited by the high computational and memory costs associated with per-sample gradient computation. In this work, we propose GraSS, a novel gradient compression algorithm and its variants FactGraSS for linear layers specifically, that explicitly leverage the inherent sparsity of per-sample gradients to achieve sub-linear space and time complexity. Extensive experiments demonstrate the effectiveness of our approach, achieving substantial speedups while preserving data influence fidelity. In particular, FactGraSS achieves up to 165% faster throughput on billion-scale models compared to the previous state-of-the-art baselines. Our code is publicly available at https://github.com/TRAIS-Lab/GraSS.