Grouped-query attention (GQA)
A form of multi-head attention in which the heads are split into groups, and each group shares one set of keys and values. It is nearly as fast as multi-query attention while keeping nearly all the quality of standard attention.
Introduced by
Grouped-query attention, 2023
Used by
- Llama 2, 2023
- Mistral 7B, 2023
- Llama 3, 2024
- Qwen2, 2024
- Llama 3.1 405B, 2024