GRASPrune: Global Gating for Budgeted Structured Pruning of Large Language Models

TL;DR

GRASPrune is a structured pruning method for large language models that reduces parameters and maintains performance by applying global gating during post-training, without additional fine-tuning.

Contribution

It introduces a global gating framework that enforces a strict budget during pruning, using lightweight gate scores and calibration to produce efficient, high-performing smaller models.

Findings

Removes 50% of parameters on LLaMA-2-7B with minimal perplexity increase.

Maintains competitive zero-shot accuracy across five benchmarks.

Operates efficiently on a single GPU without full model fine-tuning.

Abstract

Large language models (LLMs) are expensive to serve because model parameters, attention computation, and KV caches impose substantial memory and latency costs. We present GRASPrune, a structured pruning framework applied after pretraining that jointly prunes FFN channels and KV head groups under a single global budget. Instead of learning importance scores without constraints and applying the budget only after training, GRASPrune learns lightweight gate scores with a projected straight-through estimator that enforces a hard mask satisfying the budget at every step while keeping the backbone weights frozen. After the mask is fixed, we calibrate scaling factors on the retained units to mitigate scale mismatch caused by pruning, and fold these factors into the pruned weights to obtain a smaller dense checkpoint with no extra parameters at inference. On LLaMA-2-7B, GRASPrune removes 50% of parameters and achieves 12.18 perplexity on WikiText-2 while maintaining competitive average zero-shot accuracy on five benchmarks, using four epochs on 512 unlabeled calibration sequences on a single NVIDIA A100 80GB GPU without any full model fine-tuning.