Doxygen Book
Public Attributes | List of all members
_GstTensorFilterProperties Struct Reference

GstTensorFilter's properties for NN framework (internal data structure) More...

#include <nnstreamer_plugin_api_filter.h>

Collaboration diagram for _GstTensorFilterProperties:
Collaboration graph
[legend]

Public Attributes

const char * fwname
 
int fw_opened
 
const char ** model_files
 
int num_models
 
int input_configured
 
GstTensorsInfo input_meta
 
tensors_layout input_layout
 
unsigned int input_ranks [NNS_TENSOR_SIZE_LIMIT]
 
int output_configured
 
GstTensorsInfo output_meta
 
tensors_layout output_layout
 
unsigned int output_ranks [NNS_TENSOR_SIZE_LIMIT]
 
const char * custom_properties
 
accl_hw * hw_list
 
int num_hw
 
const char * accl_str
 
char * shared_tensor_filter_key
 
int latency
 
int throughput
 
int invoke_dynamic
 
uint32_t suspend
 

Detailed Description

GstTensorFilter's properties for NN framework (internal data structure)

Because custom filters of tensor_filter may need to access internal data of GstTensorFilter, we define this data structure here.

Definition at line 112 of file nnstreamer_plugin_api_filter.h.

Member Data Documentation

◆ accl_str

const char* _GstTensorFilterProperties::accl_str

accelerator configuration passed in as parameter, use in GstTensorFilterFramework V0 only

Definition at line 132 of file nnstreamer_plugin_api_filter.h.

◆ custom_properties

const char* _GstTensorFilterProperties::custom_properties

sub-plugin specific custom property values in string

Definition at line 129 of file nnstreamer_plugin_api_filter.h.

◆ fw_opened

int _GstTensorFilterProperties::fw_opened

TRUE IF open() is called or tried. Use int instead of gboolean because this is referred by custom plugins.

Definition at line 115 of file nnstreamer_plugin_api_filter.h.

◆ fwname

const char* _GstTensorFilterProperties::fwname

The name of NN Framework

Definition at line 114 of file nnstreamer_plugin_api_filter.h.

◆ hw_list

accl_hw* _GstTensorFilterProperties::hw_list

accelerators supported by framework intersected with user provided accelerator preference, use in GstTensorFilterFramework V1 only

Definition at line 130 of file nnstreamer_plugin_api_filter.h.

◆ input_configured

int _GstTensorFilterProperties::input_configured

TRUE if input tensor is configured. Use int instead of gboolean because this is referred by custom plugins.

Definition at line 119 of file nnstreamer_plugin_api_filter.h.

◆ input_layout

tensors_layout _GstTensorFilterProperties::input_layout

data layout info provided as a property to tensor_filter for the input, defaults to _NNS_LAYOUT_ANY for all the tensors

Definition at line 121 of file nnstreamer_plugin_api_filter.h.

◆ input_meta

GstTensorsInfo _GstTensorFilterProperties::input_meta

configured input tensor info

Definition at line 120 of file nnstreamer_plugin_api_filter.h.

◆ input_ranks

unsigned int _GstTensorFilterProperties::input_ranks[NNS_TENSOR_SIZE_LIMIT]

the rank list of input tensors, it is calculated based on the dimension string.

Definition at line 122 of file nnstreamer_plugin_api_filter.h.

◆ invoke_dynamic

int _GstTensorFilterProperties::invoke_dynamic

True for supporting invoke with flexible output.

Definition at line 137 of file nnstreamer_plugin_api_filter.h.

◆ latency

int _GstTensorFilterProperties::latency

The average latency over the recent 10 inferences in microseconds

Definition at line 135 of file nnstreamer_plugin_api_filter.h.

◆ model_files

const char** _GstTensorFilterProperties::model_files

File path to the model file (as an argument for NNFW). char instead of gchar for non-glib custom plugins

Definition at line 116 of file nnstreamer_plugin_api_filter.h.

◆ num_hw

int _GstTensorFilterProperties::num_hw

number of hardware accelerators in the hw_list supported by the framework

Definition at line 131 of file nnstreamer_plugin_api_filter.h.

◆ num_models

int _GstTensorFilterProperties::num_models

number of model files. Some frameworks need multiple model files to initialize the graph (caffe, caffe2)

Definition at line 117 of file nnstreamer_plugin_api_filter.h.

◆ output_configured

int _GstTensorFilterProperties::output_configured

TRUE if output tensor is configured. Use int instead of gboolean because this is referred by custom plugins.

Definition at line 124 of file nnstreamer_plugin_api_filter.h.

◆ output_layout

tensors_layout _GstTensorFilterProperties::output_layout

data layout info provided as a property to tensor_filter for the output, defaults to _NNS_LAYOUT_ANY for all the tensors

Definition at line 126 of file nnstreamer_plugin_api_filter.h.

◆ output_meta

GstTensorsInfo _GstTensorFilterProperties::output_meta

configured output tensor info

Definition at line 125 of file nnstreamer_plugin_api_filter.h.

◆ output_ranks

unsigned int _GstTensorFilterProperties::output_ranks[NNS_TENSOR_SIZE_LIMIT]

the rank list of output tensors, it is calculated based on the dimension string.

Definition at line 127 of file nnstreamer_plugin_api_filter.h.

◆ shared_tensor_filter_key

char* _GstTensorFilterProperties::shared_tensor_filter_key

the shared instance key to use same model representation

Definition at line 133 of file nnstreamer_plugin_api_filter.h.

◆ suspend

uint32_t _GstTensorFilterProperties::suspend

Timeout (ms) for suspend. (Unload the framework)

Definition at line 139 of file nnstreamer_plugin_api_filter.h.

◆ throughput

int _GstTensorFilterProperties::throughput

The average throughput in the number of outputs per second

Definition at line 136 of file nnstreamer_plugin_api_filter.h.

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