Model techniques map
Techniquesoptimizationtraining parallelism

implementation detail · filed under optimization

Window-Bounded KV Exchange under Context Parallelism

An implementation detail for sliding-window layers that exchanges only the key-value data reachable by the queries, bounding traffic by the window size.

source
1
models
2
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.

the sliding-window layers exchange only the KV their queries can reach—at most a window-sized segment—so per-layer traffic is bounded by the window rather than the sequence length.

usedsoftware implementationin MiMo-V2.6Xiaomi

Filed alongside

Other methods under optimization :: training parallelism.