specific method · filed under inference & serving
EAGLE-3-style draft-model fine-tuning
Converts a pretrained MTP layer into an EAGLE-3-style draft model while freezing the target model.
Also called Draft Model Fine-Tuning.
- 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.
As the draft model of EAGLE-3 comprises a single decoder layer whose structure matches the MTP layer, we fine-tune the pre-trained MTP layer into an EAGLE-3-style draft model, with the target model frozen and only the draft layer and its feature-fusion projection updated.
usedunclearin Kimi K3Moonshot AI
Filed alongside
Other methods under inference & serving :: decoding strategy.
Speculative decodingMulti-Token PredictionDSparkEAGLEDFlashPresence PenaltyBest-of-N scaffoldingNEXTN speculative decodingMulti-layer EAGLESpeculative samplingStandardized sampling configurationMulti-stage candidate filteringRecursive shared MTP-head draftingTask-specific sampling parametersChat Prefix CompletionConcurrency-aware draft-length tuningDistribution-matched draft-model fine-tuningFused recurrent replay kernelGrammar-constrained decodingKV-cache sharing between drafter and targetLongest-trace selectionMTP-1 speculative decodingSame-checkpoint target and draft weightsThroughput-aware dynamic verification-length scheduling