Open Models Almanac
Methodsmodel architecturetoken mixersoftmax attentionsliding window attention

single mention · model architecture

Dynamic Attention Window Size Training

Trains with a dynamic attention-window size to balance inference with real-time prefill caching and offline audio understanding.

The sources on this page only mention it in passing. The line above is the extraction's one-line gloss. See sliding window attention for the methods filed beside it.

Filed under model architecture › token mixer › softmax attention › sliding window attention