One-Shot Sensitivity-Aware Mixed Sparsity Pruning for Large Language Models
This work addresses the practical deployment challenge of large language models for real-world applications by improving efficiency through pruning, though it is incremental as it builds on existing pruning techniques.
The authors tackled the high inference latency of large language models by proposing a Hessian sensitivity-aware mixed sparsity pruning method that achieves at least 50% sparsity without retraining, reducing pruning-induced error and enabling compatibility with quantization for further compression.
Various Large Language Models~(LLMs) from the Generative Pretrained Transformer(GPT) family have achieved outstanding performances in a wide range of text generation tasks. However, the enormous model sizes have hindered their practical use in real-world applications due to high inference latency. Therefore, improving the efficiencies of LLMs through quantization, pruning, and other means has been a key issue in LLM studies. In this work, we propose a method based on Hessian sensitivity-aware mixed sparsity pruning to prune LLMs to at least 50% sparsity without the need of any retraining. It allocates sparsity adaptively based on sensitivity, allowing us to reduce pruning-induced error while maintaining the overall sparsity level. The advantages of the proposed method exhibit even more when the sparsity is extremely high. Furthermore, our method is compatible with quantization, enabling further compression of LLMs. We have released the available code.