Wanda++: Pruning Large Language Models via Regional Gradients
This work addresses the challenge of efficient model compression for LLM deployment, representing an incremental improvement over existing pruning methods.
The paper tackles the problem of pruning large language models for inference speedup while minimizing accuracy degradation by introducing Wanda++, a pruning framework that uses decoder-block-level regional gradients. The result is a 32% perplexity improvement over the previous state-of-the-art method Wanda in language modeling tasks, with pruning completed in under 10 minutes on a single GPU.
Large Language Models (LLMs) pruning seeks to remove unimportant weights for inference speedup with minimal accuracy impact. However, existing methods often suffer from accuracy degradation without full-model sparsity-aware fine-tuning. This paper presents Wanda++, a novel pruning framework that outperforms the state-of-the-art methods by utilizing decoder-block-level \textbf{regional} gradients. Specifically, Wanda++ improves the pruning score with regional gradients for the first time and proposes an efficient regional optimization method to minimize pruning-induced output discrepancies between the dense and sparse decoder output. Notably, Wanda++ improves perplexity by up to 32\% over Wanda in the language modeling task and generalizes effectively to downstream tasks. Moreover, despite updating weights with regional optimization, Wanda++ remains orthogonal to sparsity-aware fine-tuning, further reducing perplexity with LoRA in great extend. Our approach is lightweight, pruning a 7B LLaMA model in under 10 minutes on a single H100 GPU.