pulsatrix
Loading...
Searching...
No Matches
pulsatrix::Dataset Class Referenceabstract

Random-access dataset abstraction – pulsatrix's analogue of PyTorch's torch.utils.data.Dataset / LibTorch's torch::data::Dataset (len/__getitem__). Every concrete modality (CsvDataset, MnistDatasetAdapter, future image/text/audio readers) subclasses this; DataLoader depends only on this interface. More...

#include <dataset.hpp>

Inheritance diagram for pulsatrix::Dataset:

Public Member Functions

virtual ~Dataset ()=default
 
virtual int64_t size () const =0
 Number of samples in this dataset.
 
virtual Sample get (int64_t index) const =0
 Loads one sample by index.
 

Detailed Description

Random-access dataset abstraction – pulsatrix's analogue of PyTorch's torch.utils.data.Dataset / LibTorch's torch::data::Dataset (len/__getitem__). Every concrete modality (CsvDataset, MnistDatasetAdapter, future image/text/audio readers) subclasses this; DataLoader depends only on this interface.

Note
NOT a Module – a Dataset produces Tensors, it does not compute over them, so it has no forward()/backward()/propagate_relevance() (charter non-negotiable #5's LRP requirement applies to Module subclasses only, per campaign_exai_dl_library_data_pipeline's explicit scope note).

Constructor & Destructor Documentation

◆ ~Dataset()

virtual pulsatrix::Dataset::~Dataset ( )
virtualdefault

Member Function Documentation

◆ get()

virtual Sample pulsatrix::Dataset::get ( int64_t  index) const
pure virtual

Loads one sample by index.

Parameters
indexSample index, must be in [0, size()).
Returns
The sample at index.
Exceptions
std::out_of_rangeif index is out of bounds – external boundary: the index is chosen by a Sampler/DataLoader from possibly-external configuration (shuffle seed, resumed epoch offset), not a compile-time-known loop bound.

Implemented in pulsatrix::AudioFolderDataset, pulsatrix::CsvDataset, pulsatrix::ImageFolderDataset, pulsatrix::MnistDatasetAdapter, pulsatrix::TextDataset, pulsatrix::TransformDataset, and pulsatrix::VideoFrameDirectoryDataset.

◆ size()

virtual int64_t pulsatrix::Dataset::size ( ) const
pure virtual

The documentation for this class was generated from the following file: