Weighted Grouped Query Attention in Transformers
This work addresses the problem of reducing inference time for large language models, which is crucial for deployment under hardware constraints, but it is incremental as it builds directly on existing GQA methods.
The paper tackles the high inference costs in transformer language models by proposing Weighted Grouped-Query Attention (WGQA), a variation of Grouped-Query Attention that introduces learnable parameters for weighted averaging during finetuning, achieving an average 0.53% improvement over GQA with no additional inference overhead.
The attention mechanism forms the foundational blocks for transformer language models. Recent approaches show that scaling the model achieves human-level performance. However, with increasing demands for scaling and constraints on hardware memory, the inference costs of these models remain high. To reduce the inference time, Multi-Query Attention (MQA) and Grouped-Query Attention (GQA) were proposed in (Shazeer, 2019) and (Ainslieet al., 2023) respectively. In this paper, we propose a variation of Grouped-Query Attention, termed Weighted Grouped-Query Attention (WGQA). We introduced new learnable parameters for each key and value head in the T5 decoder attention blocks, enabling the model to take a weighted average during finetuning. Our model achieves an average of 0.53% improvement over GQA, and the performance converges to traditional Multi-head attention (MHA) with no additional overhead during inference. We evaluated the introduction of these parameters and subsequent finetuning informs the model about the grouping mechanism during training, thereby enhancing performance. Additionally, we demonstrate the scaling laws in our analysis by comparing the results between T5-small and T5-base architecture.