8f43927575
examples/ is based on KasperskyLab/knp at 55470ed6 with local changes to the mnist-learn altai/blifat models, cuda-backend CMake and model_loader. Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
119 lines
4.6 KiB
C++
119 lines
4.6 KiB
C++
/**
|
|
* @file network_functions.h
|
|
* @brief Network functions for specific model types.
|
|
* @kaspersky_support D. Postnikov
|
|
* @date 03.02.2026
|
|
* @license Apache 2.0
|
|
* @copyright © 2026 AO Kaspersky Lab
|
|
*
|
|
* Licensed under the Apache License, Version 2.0 (the "License");
|
|
* you may not use this file except in compliance with the License.
|
|
* You may obtain a copy of the License at
|
|
*
|
|
* http://www.apache.org/licenses/LICENSE-2.0
|
|
*
|
|
* Unless required by applicable law or agreed to in writing, software
|
|
* distributed under the License is distributed on an "AS IS" BASIS,
|
|
* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
* See the License for the specific language governing permissions and
|
|
* limitations under the License.
|
|
*/
|
|
|
|
#pragma once
|
|
|
|
#include <knp/neuron-traits/all_traits.h>
|
|
|
|
#include <memory>
|
|
|
|
// cppcheck-suppress missingInclude
|
|
#include "annotated_network.h"
|
|
// cppcheck-suppress missingInclude
|
|
#include "dataset.h"
|
|
// cppcheck-suppress missingInclude
|
|
#include "model_desc.h"
|
|
|
|
|
|
/**
|
|
* @brief Generic network construction function with fallback error handling.
|
|
*
|
|
* @details This template provides the generic interface for network construction across all supported neuron
|
|
* models. When a specific neuron model specialization is not implemented, this generic version throws a runtime
|
|
* error indicating the neuron type is not supported. Each supported neuron model should implement a template
|
|
* specialization for proper network construction.
|
|
*
|
|
* @tparam Neuron neuron type.
|
|
*
|
|
* @param model_desc model description containing configuration parameters.
|
|
*
|
|
* @return constructed `AnnotatedNetwork` for the specified neuron type.
|
|
*
|
|
* @throws `std::runtime_error` if the neuron type is not supported.
|
|
*/
|
|
template <typename Neuron>
|
|
AnnotatedNetwork construct_network(const ModelDescription& model_desc)
|
|
{
|
|
throw std::runtime_error("Not supported neuron type.");
|
|
}
|
|
|
|
|
|
/**
|
|
* @brief Generic network preparation function for inference with fallback error handling.
|
|
*
|
|
* @details This template provides the generic interface for preparing networks for inference operations across all
|
|
* supported neuron models. When a specific neuron model specialization is not implemented, this generic version
|
|
* throws a runtime error indicating the neuron type is not supported. Each supported neuron model should implement
|
|
* a template specialization for proper inference preparation.
|
|
*
|
|
* @tparam Neuron neuron type.
|
|
*
|
|
* @param backend shared pointer to the computational backend for inference execution.
|
|
* @param model_desc model description containing configuration parameters and paths.
|
|
* @param network annotated network structure to prepare for inference.
|
|
*
|
|
* @throws `std::runtime_error` if the neuron type is not supported.
|
|
*/
|
|
template <typename Neuron>
|
|
void prepare_network_for_inference(
|
|
const std::shared_ptr<knp::core::Backend>& backend, const ModelDescription& model_desc, AnnotatedNetwork& network)
|
|
{
|
|
throw std::runtime_error("Not supported neuron type.");
|
|
}
|
|
|
|
|
|
/**
|
|
* @brief Generic training labels spike generator function with fallback error handling.
|
|
*
|
|
* @details This template provides the generic interface for creating training label spike generators across all supported
|
|
* neuron models. When a specific neuron model specialization is not implemented, this generic version throws a runtime error
|
|
* indicating the neuron type is not supported. Each supported neuron model should implement a template specialization for
|
|
* proper label spike generation.
|
|
*
|
|
* @tparam Neuron neuron type.
|
|
*
|
|
* @param dataset dataset containing training labels and configuration.
|
|
*
|
|
* @return callable function that generates spike data for training labels.
|
|
*
|
|
* @throws `std::runtime_error` if the neuron type is not supported.
|
|
*/
|
|
template <typename Neuron>
|
|
std::function<knp::core::messaging::SpikeData(knp::core::Step)> make_training_labels_spikes_generator(
|
|
const Dataset& dataset)
|
|
{
|
|
throw std::runtime_error("Not supported neuron type.");
|
|
}
|
|
|
|
|
|
/**
|
|
* @brief Replace WTA with direct projection connections.
|
|
*
|
|
* @details The AltAI neuron model does not natively support WTA operations. To make a network built with WTA compatible
|
|
* with AltAI during inference, this function replaces each WTA sender/receiver pair with an equivalent direct projection.
|
|
* The original WTA data structures are cleared after the conversion.
|
|
*
|
|
* @note This function is used by both AltAI and BLIFAT inference preparation pipelines (BLIFAT calls it for completeness,
|
|
* even though its networks typically do not contain WTA structures).
|
|
*
|
|
* @param network annotated network.
|
|
*/
|
|
void replace_wta_with_projections(AnnotatedNetwork& network);
|