specific method · filed under model architecture
Pure Sliding Window Attention
Sliding-window attention used without another attention mechanism in the specified layers.
- source
- 1
- model
- 1
- lab adopt it
- 1
- strongest
- used
How sources treat it
One count per evidence span, weakest treatment to strongest.
used 1
Documented in
Evidence
1 span quoted from the sources, strongest treatment first.
For the first two layers, we use pure sliding window attention.
usedunclearin DeepSeek-V4.1-FlashDeepSeek
Filed alongside
Other methods under model architecture :: token mixer :: softmax attention :: sliding window attention.
Sliding Window AttentionHybrid Sliding Window Attention512-token Sliding Window AttentionDecoder SWA Bounded ReplayAlternating Row-Major and Column-Major Token SerializationDynamic Attention Window Size TrainingFixed Sliding Window AttentionPer-Head GatingSink-Augmented Sliding Window AttentionSWA-1024SWA-128