single mention · model architecture
Gated attention with reduced KV heads
An attention configuration specifying 32 query heads and 2 key-value heads, without evidence establishing a more specific method name.
The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See gated delta network for the methods filed beside it.
Filed under model architecture › token mixer › linear attention & state space › gated delta network