Internvl3 Vanilla Inference

class InternVL3VanillaInference : public trt_edgellm::safety::ModelInference

Request orchestration using Runner-selected contexts and its synchronous preparation callback. SafetyRuntime checks its fatal admission gate before entry and reports any returned failure.

Public Functions

inline InternVL3VanillaInference(

ModelInferenceDefinition const &definition,

llm::RuntimeData const &llmData,

vlm::RequestOperations &visualOperations,

LMStageContext const &visualContext

) noexcept

virtual SafetyStatus run(

ModelInferenceRunner &runner,

ProcessingRequest const &request,

TokenIdBuffer output,

ModelInferenceResult &result

) const noexcept override

Public Static Functions

static SafetyStatus create(

ModelInferenceDefinition const &definition,

llm::ModelLocalData const &data,

std::vector<LMStageContext> const &contexts,

llm::RequestOperations &llmOperations,

vlm::RequestOperations &visualOperations,

std::unique_ptr<ModelInference> &output

) noexcept

Initialization-only binding to validated Runner contexts and model-local storage. Contexts and operations outlive this object; processing setup and warmup precede request admission.