Detailed Description

Class to manage multiple tensors with string-based lookup.

This class provides a central repository for caching tensors, allowing easy retrieval by name and automatic memory management. All tensors added to the cache are owned by the cache and will be automatically released when the cache is destroyed.

Definition at line 511 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

Public Member Functions
	TensorCache ()
	Default constructor. More...

	~TensorCache ()
	Destructor that automatically releases all cached tensors. More...

bool	addTensor (Tensor *tensor)
	Add a tensor to the cache. More...

Tensor *	getTensor (const std::string &name) const
	Get a tensor by name. More...

void *	getTensorBuffer (const std::string &name) const
	Get a tensor's buffer by name. More...

size_t	getTensorSize (const std::string &name) const
	Get a tensor's size by name. More...

std::vector< int >	getTensorShape (const std::string &name) const
	Get a tensor's shape by name. More...

int	getTensorGpuId (const std::string &name) const
	Get a tensor's gpu_id by name. More...

int	getTensorFrameNum (const std::string &name) const
	Get a tensor's frame_num by name. More...

torch::Tensor	getTorchTensor (const std::string &name) const
	Get a tensor directly as PyTorch tensor by name. More...

bool	updateTensorBuffer (const std::string &name, void *newBuffer)
	Update only a tensor's buffer pointer, keeping all other properties the same. More...

Tensor *	cloneTensorCPU (const std::string &name)
	Clone a tensor to CPU memory. More...

torch::Tensor	cloneTorchTensorCPU (const std::string &name)
	Clone a tensor to CPU memory and convert to PyTorch tensor. More...

bool	hasTensor (const std::string &name) const
	Check if a tensor with the given name exists. More...

bool	removeTensor (const std::string &name)
	Remove a tensor from the cache and release its resources. More...

void	releaseAll ()
	Release all tensors and clear the cache. More...

size_t	size () const
	Get the number of tensors in the cache. More...

std::vector< std::string >	getTensorNames () const
	Get a list of all tensor names in the cache. More...

const std::unordered_map< std::string, Tensor * > &	getTensors () const
	Get all tensors in the cache. More...

void	display (bool print_data=false) const
	Display information about all cached tensors. More...

void	displayTensor (const std::string &name, bool print_data=false) const
	Display information about given tensor. More...

	TensorCache ()
	Default constructor. More...

	~TensorCache ()
	Destructor that automatically releases all cached tensors. More...

bool	addTensor (Tensor *tensor)
	Add a tensor to the cache. More...

Tensor *	getTensor (const std::string &name) const
	Get a tensor by name. More...

void *	getTensorBuffer (const std::string &name) const
	Get a tensor's buffer by name. More...

size_t	getTensorSize (const std::string &name) const
	Get a tensor's size by name. More...

std::vector< int >	getTensorShape (const std::string &name) const
	Get a tensor's shape by name. More...

int	getTensorGpuId (const std::string &name) const
	Get a tensor's gpu_id by name. More...

int	getTensorFrameNum (const std::string &name) const
	Get a tensor's frame_num by name. More...

torch::Tensor	getTorchTensor (const std::string &name) const
	Get a tensor directly as PyTorch tensor by name. More...

bool	updateTensorBuffer (const std::string &name, void *newBuffer)
	Update only a tensor's buffer pointer, keeping all other properties the same. More...

Tensor *	cloneTensorCPU (const std::string &name)
	Clone a tensor to CPU memory. More...

torch::Tensor	cloneTorchTensorCPU (const std::string &name)
	Clone a tensor to CPU memory and convert to PyTorch tensor. More...

bool	hasTensor (const std::string &name) const
	Check if a tensor with the given name exists. More...

bool	removeTensor (const std::string &name)
	Remove a tensor from the cache and release its resources. More...

void	releaseAll ()
	Release all tensors and clear the cache. More...

size_t	size () const
	Get the number of tensors in the cache. More...

std::vector< std::string >	getTensorNames () const
	Get a list of all tensor names in the cache. More...

const std::unordered_map< std::string, Tensor * > &	getTensors () const
	Get all tensors in the cache. More...

void	display (bool print_data=false) const
	Display information about all cached tensors. More...

void	displayTensor (const std::string &name, bool print_data=false) const
	Display information about given tensor. More...

Constructor & Destructor Documentation

◆ TensorCache() [1/2]

TensorCache::TensorCache ( )

inline

Default constructor.

Definition at line 519 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ ~TensorCache() [1/2]

TensorCache::~TensorCache ( )

inline

Destructor that automatically releases all cached tensors.

Ensures that all memory allocated for tensors is properly freed when the cache is destroyed.

Definition at line 527 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

References releaseAll().

◆ TensorCache() [2/2]

TensorCache::TensorCache ( )

inline

Default constructor.

Definition at line 519 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ ~TensorCache() [2/2]

TensorCache::~TensorCache ( )

inline

Destructor that automatically releases all cached tensors.

Ensures that all memory allocated for tensors is properly freed when the cache is destroyed.

Definition at line 527 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

References releaseAll().

Member Function Documentation

◆ addTensor() [1/2]

bool TensorCache::addTensor ( Tensor * tensor )

inline

Add a tensor to the cache.

Parameters

tensor Pointer to tensor object (ownership transferred to cache)

Returns: True if tensor was added, false if tensor is nullptr

Note: The cache takes ownership of the tensor and will delete it when appropriate

Definition at line 538 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

References Tensor::name.

◆ addTensor() [2/2]

bool TensorCache::addTensor ( Tensor * tensor )

inline

Add a tensor to the cache.

Parameters

tensor Pointer to tensor object (ownership transferred to cache)

Returns: True if tensor was added, false if tensor is nullptr

Note: The cache takes ownership of the tensor and will delete it when appropriate

Definition at line 538 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

References Tensor::name.

◆ cloneTensorCPU() [1/2]

Tensor* TensorCache::cloneTensorCPU ( const std::string & name )

inline

Clone a tensor to CPU memory.

Creates a new Tensor object with a CPU buffer and copies data from the existing tensor. Works regardless of whether the source tensor is on CPU or GPU.

Parameters

name	Name of the tensor to clone

Returns: Pointer to new tensor on CPU, or nullptr if source tensor not found Caller is responsible for deleting the returned tensor

Definition at line 675 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

References Tensor::buffer, Tensor::data_type, Tensor::frame_num, getTensor(), Tensor::gpu_id, hasTensor(), Tensor::name, Tensor::shape, and Tensor::size.

Referenced by cloneTorchTensorCPU().

◆ cloneTensorCPU() [2/2]

Tensor* TensorCache::cloneTensorCPU ( const std::string & name )

inline

Clone a tensor to CPU memory.

Creates a new Tensor object with a CPU buffer and copies data from the existing tensor. Works regardless of whether the source tensor is on CPU or GPU.

Parameters

name	Name of the tensor to clone

Returns: Pointer to new tensor on CPU, or nullptr if source tensor not found Caller is responsible for deleting the returned tensor

Definition at line 675 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

References Tensor::buffer, Tensor::data_type, Tensor::frame_num, getTensor(), Tensor::gpu_id, hasTensor(), Tensor::name, Tensor::shape, and Tensor::size.

◆ cloneTorchTensorCPU() [1/2]

torch::Tensor TensorCache::cloneTorchTensorCPU ( const std::string & name )

inline

Clone a tensor to CPU memory and convert to PyTorch tensor.

Creates a CPU copy of the specified tensor and converts it to a PyTorch tensor. Works regardless of whether the source tensor is on CPU or GPU. The function handles memory management internally, freeing the cloned tensor after creating the PyTorch tensor.

Parameters

name	Name of the tensor to clone

Returns: torch::Tensor PyTorch tensor on CPU, or empty tensor if source tensor not found

Definition at line 749 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

References cloneTensorCPU(), and Tensor::to_torch_tensor().

◆ cloneTorchTensorCPU() [2/2]

torch::Tensor TensorCache::cloneTorchTensorCPU ( const std::string & name )

inline

Clone a tensor to CPU memory and convert to PyTorch tensor.

Creates a CPU copy of the specified tensor and converts it to a PyTorch tensor. Works regardless of whether the source tensor is on CPU or GPU. The function handles memory management internally, freeing the cloned tensor after creating the PyTorch tensor.

Parameters

name	Name of the tensor to clone

Returns: torch::Tensor PyTorch tensor on CPU, or empty tensor if source tensor not found

Definition at line 749 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

References cloneTensorCPU(), and Tensor::to_torch_tensor().

◆ display() [1/2]

void TensorCache::display ( bool print_data = false ) const

inline

Display information about all cached tensors.

Parameters

print_data Whether to print the tensor data values

Definition at line 846 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ display() [2/2]

void TensorCache::display ( bool print_data = false ) const

inline

Display information about all cached tensors.

Parameters

print_data Whether to print the tensor data values

Definition at line 846 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ displayTensor() [1/2]

void TensorCache::displayTensor	(	const std::string &	name,
		bool	print_data = `false`
	)		const

inline

Display information about given tensor.

Parameters

name	Name of the tensor to display
print_data	Whether to print the tensor data values

Definition at line 862 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ displayTensor() [2/2]

void TensorCache::displayTensor	(	const std::string &	name,
		bool	print_data = `false`
	)		const

inline

Display information about given tensor.

Parameters

name	Name of the tensor to display
print_data	Whether to print the tensor data values

Definition at line 862 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ getTensor() [1/2]

Tensor* TensorCache::getTensor ( const std::string & name ) const

inline

Get a tensor by name.

Parameters

name	Name of the tensor to retrieve

Returns: Pointer to the tensor, or nullptr if not found

Note: The returned tensor is still owned by the cache

Definition at line 553 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ getTensor() [2/2]

Tensor* TensorCache::getTensor ( const std::string & name ) const

inline

Get a tensor by name.

Parameters

name	Name of the tensor to retrieve

Returns: Pointer to the tensor, or nullptr if not found

Note: The returned tensor is still owned by the cache

Definition at line 553 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

Referenced by cloneTensorCPU(), getTorchTensor(), and updateTensorBuffer().

◆ getTensorBuffer() [1/2]

void* TensorCache::getTensorBuffer ( const std::string & name ) const

inline

Get a tensor's buffer by name.

Parameters

name	Name of the tensor whose buffer to retrieve

Returns: void* Pointer to the tensor buffer, or nullptr if tensor not found or has no buffer

Definition at line 567 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ getTensorBuffer() [2/2]

void* TensorCache::getTensorBuffer ( const std::string & name ) const

inline

Get a tensor's buffer by name.

Parameters

name	Name of the tensor whose buffer to retrieve

Returns: void* Pointer to the tensor buffer, or nullptr if tensor not found or has no buffer

Definition at line 567 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ getTensorFrameNum() [1/2]

int TensorCache::getTensorFrameNum ( const std::string & name ) const

inline

Get a tensor's frame_num by name.

Parameters

name	Name of the tensor whose frame_num to retrieve

Returns: int Frame number of the tensor, or -1 if tensor not found

Definition at line 623 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ getTensorFrameNum() [2/2]

int TensorCache::getTensorFrameNum ( const std::string & name ) const

inline

Get a tensor's frame_num by name.

Parameters

name	Name of the tensor whose frame_num to retrieve

Returns: int Frame number of the tensor, or -1 if tensor not found

Definition at line 623 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ getTensorGpuId() [1/2]

int TensorCache::getTensorGpuId ( const std::string & name ) const

inline

Get a tensor's gpu_id by name.

Parameters

name	Name of the tensor whose gpu_id to retrieve

Returns: int GPU ID of the tensor, or -1 if tensor not found or is on CPU

Definition at line 609 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ getTensorGpuId() [2/2]

int TensorCache::getTensorGpuId ( const std::string & name ) const

inline

Get a tensor's gpu_id by name.

Parameters

name	Name of the tensor whose gpu_id to retrieve

Returns: int GPU ID of the tensor, or -1 if tensor not found or is on CPU

Definition at line 609 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ getTensorNames() [1/2]

std::vector<std::string> TensorCache::getTensorNames ( ) const

inline

Get a list of all tensor names in the cache.

Returns: Vector of tensor names

Definition at line 823 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ getTensorNames() [2/2]

std::vector<std::string> TensorCache::getTensorNames ( ) const

inline

Get a list of all tensor names in the cache.

Returns: Vector of tensor names

Definition at line 823 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ getTensors() [1/2]

const std::unordered_map<std::string, Tensor*>& TensorCache::getTensors ( ) const

inline

Get all tensors in the cache.

Returns: Reference to the map containing all tensor name-pointer pairs

Definition at line 837 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ getTensors() [2/2]

const std::unordered_map<std::string, Tensor*>& TensorCache::getTensors ( ) const

inline

Get all tensors in the cache.

Returns: Reference to the map containing all tensor name-pointer pairs

Definition at line 837 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ getTensorShape() [1/2]

std::vector<int> TensorCache::getTensorShape ( const std::string & name ) const

inline

Get a tensor's shape by name.

Parameters

name	Name of the tensor whose shape to retrieve

Returns: std::vector<int> Shape of the tensor (dimensions), or empty vector if tensor not found

Definition at line 595 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ getTensorShape() [2/2]

std::vector<int> TensorCache::getTensorShape ( const std::string & name ) const

inline

Get a tensor's shape by name.

Parameters

name	Name of the tensor whose shape to retrieve

Returns: std::vector<int> Shape of the tensor (dimensions), or empty vector if tensor not found

Definition at line 595 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ getTensorSize() [1/2]

size_t TensorCache::getTensorSize ( const std::string & name ) const

inline

Get a tensor's size by name.

Parameters

name	Name of the tensor whose size to retrieve

Returns: size_t Size of the tensor buffer in bytes, or 0 if tensor not found

Definition at line 581 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ getTensorSize() [2/2]

size_t TensorCache::getTensorSize ( const std::string & name ) const

inline

Get a tensor's size by name.

Parameters

name	Name of the tensor whose size to retrieve

Returns: size_t Size of the tensor buffer in bytes, or 0 if tensor not found

Definition at line 581 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ getTorchTensor() [1/2]

torch::Tensor TensorCache::getTorchTensor ( const std::string & name ) const

inline

Get a tensor directly as PyTorch tensor by name.

Parameters

name	Name of the tensor to retrieve

Returns: torch::Tensor PyTorch tensor created from the cached tensor, or empty tensor if not found

Note: The returned PyTorch tensor shares memory with the original tensor buffer

Definition at line 638 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

References Tensor::buffer, getTensor(), and Tensor::to_torch_tensor().

◆ getTorchTensor() [2/2]

torch::Tensor TensorCache::getTorchTensor ( const std::string & name ) const

inline

Get a tensor directly as PyTorch tensor by name.

Parameters

name	Name of the tensor to retrieve

Returns: torch::Tensor PyTorch tensor created from the cached tensor, or empty tensor if not found

Note: The returned PyTorch tensor shares memory with the original tensor buffer

Definition at line 638 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

References Tensor::buffer, getTensor(), and Tensor::to_torch_tensor().

◆ hasTensor() [1/2]

bool TensorCache::hasTensor ( const std::string & name ) const

inline

Check if a tensor with the given name exists.

Parameters

name	Name of the tensor to check

Returns: True if tensor exists, false otherwise

Definition at line 766 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

Referenced by cloneTensorCPU().

◆ hasTensor() [2/2]

bool TensorCache::hasTensor ( const std::string & name ) const

inline

Check if a tensor with the given name exists.

Parameters

name	Name of the tensor to check

Returns: True if tensor exists, false otherwise

Definition at line 766 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ releaseAll() [1/2]

void TensorCache::releaseAll ( )

inline

Release all tensors and clear the cache.

Definition at line 793 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

Referenced by ~TensorCache().

◆ releaseAll() [2/2]

void TensorCache::releaseAll ( )

inline

Release all tensors and clear the cache.

Definition at line 793 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ removeTensor() [1/2]

bool TensorCache::removeTensor ( const std::string & name )

inline

Remove a tensor from the cache and release its resources.

Parameters

name	Name of the tensor to remove

Returns: True if tensor was found and removed, false otherwise

Definition at line 776 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ removeTensor() [2/2]

bool TensorCache::removeTensor ( const std::string & name )

inline

Remove a tensor from the cache and release its resources.

Parameters

name	Name of the tensor to remove

Returns: True if tensor was found and removed, false otherwise

Definition at line 776 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ size() [1/2]

size_t TensorCache::size ( ) const

inline

Get the number of tensors in the cache.

Returns: Number of tensors

Definition at line 814 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ size() [2/2]

size_t TensorCache::size ( ) const

inline

Get the number of tensors in the cache.

Returns: Number of tensors

Definition at line 814 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

◆ updateTensorBuffer() [1/2]

bool TensorCache::updateTensorBuffer	(	const std::string &	name,
		void *	newBuffer
	)

inline

Update only a tensor's buffer pointer, keeping all other properties the same.

Parameters

name	Name of the tensor to update
newBuffer	Pointer to the new buffer (must be GPU memory with same size/layout)

Returns: True if tensor was found and updated, false otherwise

Definition at line 653 of file 9.2/sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

References Tensor::buffer, and getTensor().

◆ updateTensorBuffer() [2/2]

bool TensorCache::updateTensorBuffer	(	const std::string &	name,
		void *	newBuffer
	)

inline

Update only a tensor's buffer pointer, keeping all other properties the same.

Parameters

name	Name of the tensor to update
newBuffer	Pointer to the new buffer (must be GPU memory with same size/layout)

Returns: True if tensor was found and updated, false otherwise

Definition at line 653 of file sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h.

References Tensor::buffer, and getTensor().

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

sources/gst-plugins/gst-nvdsvideotemplate/ds-sparse4d/utils/tensor.h


NVIDIA DeepStream SDK API Reference	9.2 Release

NVIDIA DeepStream SDK API Reference

9.2 Release

Detailed Description

Public Member Functions

Constructor & Destructor Documentation

◆ TensorCache() [1/2]

◆ ~TensorCache() [1/2]

◆ TensorCache() [2/2]

◆ ~TensorCache() [2/2]

Member Function Documentation

◆ addTensor() [1/2]

◆ addTensor() [2/2]

◆ cloneTensorCPU() [1/2]

◆ cloneTensorCPU() [2/2]

◆ cloneTorchTensorCPU() [1/2]

◆ cloneTorchTensorCPU() [2/2]

◆ display() [1/2]

◆ display() [2/2]

◆ displayTensor() [1/2]

◆ displayTensor() [2/2]

◆ getTensor() [1/2]

◆ getTensor() [2/2]

◆ getTensorBuffer() [1/2]

◆ getTensorBuffer() [2/2]

◆ getTensorFrameNum() [1/2]

◆ getTensorFrameNum() [2/2]

◆ getTensorGpuId() [1/2]

◆ getTensorGpuId() [2/2]

◆ getTensorNames() [1/2]

◆ getTensorNames() [2/2]

◆ getTensors() [1/2]

◆ getTensors() [2/2]

◆ getTensorShape() [1/2]

◆ getTensorShape() [2/2]

◆ getTensorSize() [1/2]

◆ getTensorSize() [2/2]

◆ getTorchTensor() [1/2]

◆ getTorchTensor() [2/2]

◆ hasTensor() [1/2]

◆ hasTensor() [2/2]

◆ releaseAll() [1/2]

◆ releaseAll() [2/2]

◆ removeTensor() [1/2]

◆ removeTensor() [2/2]

◆ size() [1/2]

◆ size() [2/2]

◆ updateTensorBuffer() [1/2]

◆ updateTensorBuffer() [2/2]