UniAttn: Reducing Inference Costs via Softmax Unification for Post-Training LLMs

TL;DR

UniAttn introduces a novel post-training method that unifies Softmax activations across transformer blocks, significantly reducing inference costs while maintaining model performance.

Contribution

The paper proposes Softmax UniAttn, a new approach that unifies Softmax activations to lower inference costs in post-trained LLMs, outperforming existing methods.

Findings

Significant reduction in inference latency with UniAttn.

Maintains comparable performance to standard post-training.

Outperforms existing efficient architectures during post-training.

Abstract

Post-training is essential for adapting Large Language Models (LLMs) to real-world applications. Deploying post-trained models faces significant challenges due to substantial memory overhead and noticeable inference latency. Existing work has identified significant redundancies in LLMs and proposed efficient architectures, namely intra-layer KV sharing and cross-layer KV sharing. However, these methods still result in high inference time overhead, remaining suboptimal for post-training pre-trained LLMs. In this paper, we identify that the \texttt{Softmax} operation is a primary bottleneck for LLM inference and discover that it is actually highly redundant during post-training. We propose Softmax \textbf{Uni}fication in \textbf{Att}e\textbf{n}tion (\textbf{UniAttn}), a novel post-training method that unifies Softmax activations across transformer blocks to reduce LLM inference costs. Additionally, UniAttn adopts a linear projection to compensate for the errors induced by Softmax unification. Experiments show that UniAttn matches the performance of standard post-training while significantly reducing inference costs, outperforming existing efficient architectures during post-training.