Model techniques map
Techniquesoptimizationtraining runtime

implementation detail · filed under optimization

Double-buffered chunked streaming of reference-model weights

Streaming reference weights in chunks between two slots, with one used for computation while the other prefetches.

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.

We stream reference weights into these slots chunk by chunk: one slot is used for the current forward computation while the other prefetches the next chunk

usedpost trainingin Kimi K3Moonshot AI

Filed alongside

Other methods under optimization :: training runtime.