sgns::sgprocessing::MNN_Llm¶

#include <processing_processor_mnn_llm.hpp>

Inherits from sgns::sgprocessing::ProcessingProcessor

Public Functions¶

	Name
	MNN_Llm()
	~MNN_Llm() override
virtual ProcessingResult	StartProcessing(std::vector< std::vector< uint8_t > > & chunkhashes, const sgns::IoDeclaration & proc, std::vector< char > & promptData, std::vector< char > & modelFile, const std::vector< sgns::Parameter > * parameters, const ExecutionContext & execCtx) override

Additional inherited members¶

Public Functions inherited from sgns::sgprocessing::ProcessingProcessor

	Name
virtual	~ProcessingProcessor() =default
virtual float	GetProgress() const
void	RunTeardown()

Protected Functions inherited from sgns::sgprocessing::ProcessingProcessor

	Name
void	PushTeardown(std::function< void()> fn)

Protected Attributes inherited from sgns::sgprocessing::ProcessingProcessor

	Name
std::atomic< float >	m_progress
sgns::sgprocmanager::Logger	m_logger

Public Functions Documentation¶

function MNN_Llm¶

inline MNN_Llm()

Create an LLM processor

function ~MNN_Llm¶

inline ~MNN_Llm() override

function StartProcessing¶

virtual ProcessingResult StartProcessing(
    std::vector< std::vector< uint8_t > > & chunkhashes,
    const sgns::IoDeclaration & proc,
    std::vector< char > & promptData,
    std::vector< char > & modelFile,
    const std::vector< sgns::Parameter > * parameters,
    const ExecutionContext & execCtx
) override

Parameters:

chunkhashes - Reference to vector to store chunk hashes
proc - Input/output declaration with processing parameters
promptData - Input prompt text as character vector
modelFile - MNN LLM model file data (materialized to a temp directory before MNN::Transformer::Llm::createLLM(), which – unlike MNN::Interpreter::createFromBuffer – requires a directory path, not an in-memory buffer)

Reimplements: sgns::sgprocessing::ProcessingProcessor::StartProcessing

Start processing data – autoregressive LLM text generation.

Updated on 2026-09-17 at 06:29:14 +0000