specific method · filed under inference & serving
Projected-input caching for speculative KDA rollback
Caches projected inputs so accepted-token states can be rebuilt on-chip during speculative KDA rollback.
Also called cache only these projected inputs.
- source
- 1
- model
- 1
- lab adopt it
- 1
- strongest
- core
How sources treat it
One count per evidence span, weakest treatment to strongest.
core 1
Documented in
Evidence
1 span quoted from the sources, strongest treatment first.
We therefore cache only these projected inputs, rebuild the states of accepted tokens on-chip, and write back the states of the verified and bonus tokens
coreinference servingin Kimi K3Moonshot AI
Filed alongside
Other methods under inference & serving :: KV cache management.
Chunked prefillPrefix cachingPrompt cachingSWA Bounded ReplayLanguage-model-only serving modePeriodic cache checkpointingCompressed KV cachingKDA-aware prefix-cache managementRadixCacheZero SWA caching8-bit Mamba cache quantizationAsynchronous cache offloading and restorationAutomatic cacheBlock-based KV cacheCross-layer KV-cache reuseCustomized heterogeneous KV-cache layoutDisable prefix caching for benchmarkingDynamic allocation of fixed-size state-cache poolsEncoder SWA bounded replayExact SWA KV reconstruction via full multi-layer replayFine-grained prefix hashingHierarchical KV cachingInference-side KV-cache reset on weight synchronizationKDA with prefill cache