Factory that materializes a Sampler from GenerationParams.
virtual std::unique_ptr< Sampler > create(const GenerationParams ¶ms)=0
Build a configured Sampler for one generation.
Pure-virtual per-generation sampler used by the decode loop.
virtual int32_t sample()=0
Sample one token from the current decode position.
virtual void reset()
Reset internal sampler state.
Configuration structs with defaults.
Activate model on GPU (WARM → ACTIVE).
Generation parameters for a single inference call.