Struct anira::ModelData

struct ModelData

Container for neural network model data and metadata.

The ModelData struct encapsulates all information necessary to load and identify a neural network model for inference. It supports both binary model data (loaded from files) and string-based model paths, along with backend-specific metadata.

See also

InferenceBackend, InferenceConfig

Public Functions

inline ModelData(void *data, size_t size, InferenceBackend backend, const std::string &model_function = "", bool is_binary = true)

Constructs ModelData with binary data or model path.

Parameters:

  • data – Pointer to model data (binary data) or model path string

  • size – Size of the data in bytes (for binary) or string length (for paths)

  • backend – The inference backend that will use this model

  • model_function – Optional function name within the model (LibTorch only)

  • is_binary – Whether the data represents binary model data (true) or a file path (false)

inline ModelData(const std::string &model_path, InferenceBackend backend, const std::string &model_function = "", bool is_binary = false)

Constructs ModelData from a file path string.

Convenience constructor for creating ModelData from a model file path. The path string is copied internally and managed by the ModelData instance.

Parameters:

  • model_path – Path to the model file on disk

  • backend – The inference backend that will use this model

  • model_function – Optional function name within the model (LibTorch only)

  • is_binary – Whether to treat the path as binary data (typically false for file paths)

inline ModelData(const ModelData &other)

Copy constructor with proper memory management.

Creates a deep copy of the ModelData, ensuring independent memory management for non-binary data while sharing binary data references safely.

Parameters:

other – The ModelData instance to copy from

inline ModelData &operator=(const ModelData &other)

Assignment operator with proper memory management.

Safely assigns one ModelData to another, handling memory deallocation and reallocation as needed for non-binary data.

Parameters:

other – The ModelData instance to assign from

Returns:

Reference to this ModelData instance

inline ~ModelData()

Destructor with automatic memory cleanup.

Automatically frees allocated memory for non-binary data to prevent memory leaks. Binary data is assumed to be managed externally and is not freed.

inline bool operator==(const ModelData &other) const

Equality comparison operator.

Parameters:

other – The ModelData instance to compare with

Returns:

true if all members are equal, false otherwise

inline bool operator!=(const ModelData &other) const

Inequality comparison operator.

Parameters:

other – The ModelData instance to compare with

Returns:

true if any members are not equal, false otherwise

Public Members

void *m_data

Pointer to model data (binary data or string data)

size_t m_size

Size of the model data in bytes.

InferenceBackend m_backend

Target inference backend for this model.

std::string m_model_function

Function name within the model (LibTorch specific)

bool m_is_binary

Whether the data represents binary model data.