我有这段代码可以使用 HDF5Sharp 将数组写入 Hdf5。 但问题是,我需要将数据写入 1 x 100 x 500 的块,而不是 100k x 100 x 500,而且我不知道该怎么做。
using HDF5CSharp;
// array of 100k x 100 x 500
float[,,] traindata = new float[100000, 100, 500];
// ...
// fill the array with data
// ...
long fileId = Hdf5.CreateFile("d:\\Temp\\Test.H5");
var groupId = Hdf5.CreateOrOpenGroup(fileId, "data");
Hdf5.WriteDataset(groupId, "data", traindata);
Hdf5.CloseGroup(groupId);
Hdf5.CloseFile(fileId);
不确定如何使用您指定的库完成此操作,但使用 HDFql(一种将您从 HDF5 的低级细节中抽象出来的高级语言),您的用例可以在 C# 中按以下方式解决:
// use HDFql namespace (make sure it can be found by the C# compiler)
using AS.HDFql;
// declare variables
float[,,] traindata = new float[1, 100, 500];
int number;
int i;
// create an HDF5 file named 'Test.h5' and use (i.e. open) it
HDFql.Execute("create and use file Test.h5");
// create a chunked dataset named 'data' (within a group named 'grp') of data type float with three dimensions (the first dimension is extendible)
HDFql.Execute("create chunked(1, 100, 500) dataset grp/data as float(0 to unlimited, 100, 500)");
// register variable 'traindata'
number = HDFql.VariableRegister(traindata);
// loop 100000 times
for(i = 0; i < 100000; i++)
{
// populate variable 'traindata'
// (...)
// insert (i.e. write) data stored in 'traindata' into the last position of 'data' (using a hyperslab selection)
HDFql.Execute("insert into grp/data[-1:::] values from memory " + number);
// alter (i.e. extend) first dimension of 'data' plus one unit
HDFql.Execute("alter dimension grp/data to +1");
}