An Automatic and Efficient BERT Pruning for Edge AI Systems

TL;DR

AE-BERT introduces an automatic, efficient pruning framework for BERT models that enhances NLP task performance on resource-limited devices without requiring expert tuning.

Contribution

It presents a novel automatic pruning method that outperforms hand-crafted approaches, enabling better accuracy and higher pruning ratios on NLP benchmarks.

Findings

Outperforms state-of-the-art pruning methods on GLUE tasks.

Achieves higher pruning ratios with minimal accuracy loss.

Provides significant inference speedup on FPGA hardware.

Abstract

With the yearning for deep learning democratization, there are increasing demands to implement Transformer-based natural language processing (NLP) models on resource-constrained devices for low-latency and high accuracy. Existing BERT pruning methods require domain experts to heuristically handcraft hyperparameters to strike a balance among model size, latency, and accuracy. In this work, we propose AE-BERT, an automatic and efficient BERT pruning framework with efficient evaluation to select a "good" sub-network candidate (with high accuracy) given the overall pruning ratio constraints. Our proposed method requires no human experts experience and achieves a better accuracy performance on many NLP tasks. Our experimental results on General Language Understanding Evaluation (GLUE) benchmark show that AE-BERT outperforms the state-of-the-art (SOTA) hand-crafted pruning methods on BERT$_{\mathrm{BASE}}$. On QNLI and RTE, we obtain 75\% and 42.8\% more overall pruning ratio while achieving higher accuracy. On MRPC, we obtain a 4.6 higher score than the SOTA at the same overall pruning ratio of 0.5. On STS-B, we can achieve a 40\% higher pruning ratio with a very small loss in Spearman correlation compared to SOTA hand-crafted pruning methods. Experimental results also show that after model compression, the inference time of a single BERT$_{\mathrm{BASE}}$ encoder on Xilinx Alveo U200 FPGA board has a 1.83$\times$ speedup compared to Intel(R) Xeon(R) Gold 5218 (2.30GHz) CPU, which shows the reasonableness of deploying the proposed method generated subnets of BERT$_{\mathrm{BASE}}$ model on computation restricted devices.