MoE-I$^2$: Compressing Mixture of Experts Models through Inter-Expert Pruning and Intra-Expert Low-Rank Decomposition
This work addresses the problem of efficient deployment for MoE LLMs, offering a domain-specific solution that is incremental in nature.
The paper tackles the high deployment costs of Mixture of Experts (MoE) LLMs by introducing a two-stage compression method that reduces model size and computational cost while maintaining performance in zero-shot tasks, as demonstrated on models like Qwen1.5-MoE-A2.7B and Mixtral-8x7B.
The emergence of Mixture of Experts (MoE) LLMs has significantly advanced the development of language models. Compared to traditional LLMs, MoE LLMs outperform traditional LLMs by achieving higher performance with considerably fewer activated parameters. Despite this efficiency, their enormous parameter size still leads to high deployment costs. In this paper, we introduce a two-stage compression method tailored for MoE to reduce the model size and decrease the computational cost. First, in the inter-expert pruning stage, we analyze the importance of each layer and propose the Layer-wise Genetic Search and Block-wise KT-Reception Field with the non-uniform pruning ratio to prune the individual expert. Second, in the intra-expert decomposition stage, we apply the low-rank decomposition to further compress the parameters within the remaining experts. Extensive experiments on Qwen1.5-MoE-A2.7B, DeepSeek-V2-Lite, and Mixtral-8$\times$7B demonstrate that our proposed methods can both reduce the model size and enhance inference efficiency while maintaining performance in various zero-shot tasks. The code will be available at \url{https://github.com/xiaochengsky/MoEI-2.git}