max/poet
1
0
Fork 0
mirror of https://git.gfz-potsdam.de/naaice/poet.git synced 2025-12-16 04:48:23 +01:00
Code Issues Packages Projects Releases Wiki Activity

add serializer/deserializer functionality for training and target data

Browse Source
This commit is contained in:
Hannes Signer 2024-12-23 23:28:24 +01:00
parent f76e438c30
commit 05e8f11d82
2 changed files with 83 additions and 76 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

118
src/Chemistry/SurrogateModels/serializer.cpp
View File

@ -3,6 +3,8 @@
#include <Eigen/src/Core/Matrix.h>
#include <cstddef>
#include <cstdio>
#include <cstring>
#include <vector>
using namespace std;
namespace poet{
@ -32,6 +34,12 @@ size_t calculateStructSize(void *struct_pointer, char type){
}
} else if (type == 'T') {
struct_size += sizeof(size_t); // number of vectors
struct_size += sizeof(size_t); // length of vector
for (const std::vector<double> &vector:
*static_cast<std::vector<std::vector<double>>*>(struct_pointer)){
struct_size += vector.size() * sizeof(double);
}
}
return struct_size;
@ -78,60 +86,6 @@ int serializeModelWeights(const EigenModel *model, char *memory){
return 0;
}
// EigenModel deserializeModelWeights(char *memory, size_t buffer_size){
// EigenModel deserializedModel;
// size_t num_matrices;
// size_t size_counter = 0;
// std::memcpy(&num_matrices, memory, sizeof(size_t));
// fprintf(stdout, "number of matrices: %zu\n", num_matrices);
// memory += sizeof(size_t);
// size_counter += sizeof(size_t);
// deserializedModel.weight_matrices.resize(num_matrices);
// for (Eigen::MatrixXd &matrix : deserializedModel.weight_matrices) {
// size_t rows, cols;
// std::memcpy(&rows, memory, sizeof(size_t));
// memory += sizeof(size_t);
// size_counter += sizeof(size_t);
// std::memcpy(&cols, memory, sizeof(size_t));
// fprintf(stdout, "rows: %zu, cols: %zu\n", rows, cols);
// memory += sizeof(size_t);
// size_counter += sizeof(size_t);
// fprintf(stdout, "rows before: %td, cols before: %td\n", matrix.rows(), matrix.cols());
// matrix.resize(rows, cols);
// std::memcpy(matrix.data(), memory, rows * cols * sizeof(double));
// memory += rows * cols * sizeof(double);
// size_counter += rows * cols * sizeof(double);
// }
// fprintf(stdout, "deserialized size of matrices: %zu\n", size_counter);
// size_t num_biases;
// std::memcpy(&num_biases, memory, sizeof(size_t));
// memory += sizeof(size_t);
// size_counter += sizeof(size_t);
// fprintf(stdout, "number of biases: %zu\n", num_biases);
// deserializedModel.biases.resize(num_biases);
// for (Eigen::VectorXd &bias : deserializedModel.biases) {
// size_t size;
// std::memcpy(&size, memory, sizeof(size_t));
// fprintf(stdout, "bias length: %zu\n", size);
// memory += sizeof(size_t);
// size_counter += sizeof(size_t);
// bias.resize(size);
// std::memcpy(bias.data(), memory, size * sizeof(double));
// memory += size * sizeof(double);
// size_counter += size * sizeof(double);
// }
// fprintf(stdout, "deserialized size: %zu\n", size_counter);
// if(size_counter > buffer_size){
// fprintf(stderr, "buffer corrupted!\n");
// }
// return deserializedModel;
// }
EigenModel deserializeModelWeights(char *memory, size_t buffer_size) {
EigenModel deserializedModel;
size_t size_counter = 0;
@ -147,11 +101,11 @@ EigenModel deserializeModelWeights(char *memory, size_t buffer_size) {
size_counter += sizeof(size_t);
deserializedModel.weight_matrices.resize(num_matrices);
// Matrizen deserialisieren
// matrix deserialization
for (Eigen::MatrixXd &matrix : deserializedModel.weight_matrices) {
size_t rows, cols;
// Buffer-Check
// buffer check
if (size_counter + 2 * sizeof(size_t) > buffer_size) {
fprintf(stderr, "Buffer too small for matrix dimensions.\n");
return deserializedModel;
@ -170,15 +124,15 @@ EigenModel deserializeModelWeights(char *memory, size_t buffer_size) {
return deserializedModel;
}
// Kopiere Daten in neue Matrix
// interpret memory as Eigen::MatrixXd (more efficient than memcpy?)
Eigen::MatrixXd temp = Eigen::Map<Eigen::MatrixXd>(
reinterpret_cast<double*>(memory), rows, cols);
matrix = temp; // Kopieren der Daten
matrix = temp; // copy data to matrix
memory += rows * cols * sizeof(double);
size_counter += rows * cols * sizeof(double);
}
// Anzahl Biases
// number of bias vectors
size_t num_biases;
if (size_counter + sizeof(size_t) > buffer_size) {
fprintf(stderr, "Buffer too small for biases.\n");
@ -189,7 +143,7 @@ EigenModel deserializeModelWeights(char *memory, size_t buffer_size) {
size_counter += sizeof(size_t);
deserializedModel.biases.resize(num_biases);
// Biases deserialisieren
// deserialization of bias vectors
for (Eigen::VectorXd &bias : deserializedModel.biases) {
size_t size;
if (size_counter + sizeof(size_t) > buffer_size) {
@ -206,7 +160,8 @@ EigenModel deserializeModelWeights(char *memory, size_t buffer_size) {
return deserializedModel;
}
// Kopiere Daten in neuen Vektor
// same procedure as for the matrices
// TODO: delete temp variable
Eigen::VectorXd temp = Eigen::Map<Eigen::VectorXd>(
reinterpret_cast<double*>(memory), size);
bias = temp; // Kopieren der Daten
@ -218,4 +173,45 @@ EigenModel deserializeModelWeights(char *memory, size_t buffer_size) {
}
int serializeTrainingData(std::vector<std::vector<double>> data, char *memory){
size_t num_vectors = data.size();
std::memcpy(memory, &num_vectors, sizeof(size_t));
memory += sizeof(size_t);
for (const std::vector<double> &vector : data) {
size_t size = vector.size();
std::memcpy(memory, &size, sizeof(size_t));
memory += sizeof(size_t);
std::memcpy(memory, vector.data(), size * sizeof(double));
memory += size * sizeof(double);
}
return 0;
}
std::vector<std::vector<double>> deserializeTrainingData(char* data){
std::vector<std::vector<double>> deserialized_data;
size_t num_vectors;
std::memcpy(&num_vectors, data, sizeof(size_t));
data += sizeof(size_t);
for (size_t i = 0; i < num_vectors; i++) {
size_t size;
std::memcpy(&size, data, sizeof(size_t));
data += sizeof(size_t);
std::vector<double> vector(size);
std::memcpy(vector.data(), data, size * sizeof(double));
data += size * sizeof(double);
deserialized_data.push_back(vector);
}
return deserialized_data;
}
}

39
src/Chemistry/SurrogateModels/serializer.hpp
View File

@ -6,6 +6,17 @@
namespace poet{
/**
* @brief Calculates the size of stored elements in the EigenModel and TrainData
* structs
*
* @param struct_pointer: pointer to the struct
* @param type: determines the struct type given: E for EigenModel and T
* training data vector structures
* @return size_t: size of stored elements
*/
size_t calculateStructSize(void* struct_pointer, char type);
/**
* @brief Serialize the weights and biases of the model into a memory location
* to send them via RDMA
@ -14,6 +25,9 @@ namespace poet{
* model
* @param memory: Pointer to the memory location where the serialized data will
* be stored
* The serialized data looks like this:
* |# matrices|# cols of matrix 1|# rows of matrix 1|matrix 1 data|# cols of matrix 2|...
* |# bias vectors|length of bias vector 1|bias vector 1 data|length of bias vector 2|...
* @return int: 0 if the serialization was successful, -1 otherwise
*/
int serializeModelWeights(const EigenModel *model, char *memory);
@ -31,26 +45,23 @@ EigenModel deserializeModelWeights(char* memory, size_t buffer_size);
*
* @param data Struct of TrainingData containing the training data
* @param memory
* The serialized data looks like this:
* |# of vectors|length of vector 1.1|vector 1.1 data|length of vector 1.2|...
* |# of vectors|length of vector 2.1|vector 2.1 data|length of vector 2.2|...
* |length of vector |vector 3 data|
* n_training_runs
* @return std::vector<char>
*/
int serializeTrainingData(const TrainingData& data, void *memory);
int serializeTrainingData(std::vector<std::vector<double>> *data, char *memory);
/**
* @brief Deserialize the training data from a memory location
*
* @param data Pointer to the memory location where the serialized data is stored
* @return TrainingData struct containing the training data
* @param data Pointer to the memory location where the serialized data is
* stored
* @return std::vector<std::vector<double>> containing n vectors for each
* species with m training elements
*/
TrainingData deserializeTrainingData(void* data);
/**
* @brief Calculates the size of stored elements in the EigenModel and TrainData
* structs
*
* @param struct_pointer: pointer to the struct
* @param type: determines the struct type given: E for EigenModel and T TrainData
* @return size_t: size of stored elements
*/
size_t calculateStructSize(void* struct_pointer, char type);
std::vector<std::vector<double>> deserializeTrainingData(char* data);
}
#endif