specific method · filed under model architecture
512-token Sliding Window Attention
Sliding Window Attention using a 512-token window.
Also called + SWA-512.
- sources
- 2
- models
- 2
- lab adopt it
- 1
- strongest
- used
How sources treat it
One count per evidence span, weakest treatment to strongest.
evaluated 1used 1
Documented in
Evidence
2 spans quoted from the sources, strongest treatment first.
Of the 66 decoder layers, 55 use local attention with a small 512-token window.
usedmodel architecturein InklingThinking Machines Lab
Filed alongside
Other methods under model architecture :: token mixer :: softmax attention :: sliding window attention.
Sliding Window AttentionHybrid Sliding Window AttentionDecoder SWA Bounded ReplayAlternating Row-Major and Column-Major Token SerializationDynamic Attention Window Size TrainingFixed Sliding Window AttentionPer-Head GatingPure Sliding Window AttentionSink-Augmented Sliding Window AttentionSWA-1024SWA-128