mtb_ml_model_t Struct Reference

#include <mtb_ml_model.h>

Data Fields
Model runtime object common fields
char	name [MTB_ML_MODEL_NAME_LEN]
int	model_size
int	buffer_size
int	input_size
int	output_size
int	lib_error
MTB_ML_DATA_T *	output
MTB_ML_DATA_T *	input
void *	tflm_obj
int	model_time_steps
int	recurrent_ts_size
int	input_zero_point
float	input_scale
int	output_zero_point
float	output_scale
mtb_ml_profile_config_t	profiling
uint64_t	m_cpu_cycles
uint32_t	m_sum_frames
uint64_t	m_cpu_sum_cycles
uint32_t	m_cpu_peak_frame
uint64_t	m_cpu_peak_cycles
bool	is_rnn_streaming
COMPONENT_ML_TFLM
Model runtime object fields for TFLM with interpreter
uint8_t *	arena_buffer
COMPONENT_ML_TFLM_LESS
Model runtime object fields for TFLM without interpreter
tflm_rmf_apis_t	rmf_apis

Detailed Description

ML model runtime object structure

Field Documentation

arena_buffer

uint8_t* arena_buffer

pointer of allocated tensor arena buffer

buffer_size

int buffer_size

the size of ML model working buffer

input

MTB_ML_DATA_T* input

pointer of ML inference input buffer

input_scale

float input_scale

scale of input data

input_size

int input_size

array size of input data

input_zero_point

int input_zero_point

zero point of input data

is_rnn_streaming

bool is_rnn_streaming

Is the model an RNN streaming model

lib_error

int lib_error

error code from ML inference library

m_cpu_cycles

uint64_t m_cpu_cycles

CPU profiling cycles

m_cpu_peak_cycles

uint64_t m_cpu_peak_cycles

CPU profiling peak cycles

m_cpu_peak_frame

uint32_t m_cpu_peak_frame

CPU profiling peak frame

m_cpu_sum_cycles

uint64_t m_cpu_sum_cycles

CPU profiling total cycles

m_sum_frames

uint32_t m_sum_frames

profiling frames

model_size

int model_size

the size of ML model

name

char name[MTB_ML_MODEL_NAME_LEN]

the name of ML model

output

MTB_ML_DATA_T* output

pointer of ML inference output buffer

output_scale

float output_scale

scale of output data

output_size

int output_size

array size of output data

output_zero_point

int output_zero_point

zero point of output data

profiling

mtb_ml_profile_config_t profiling

flags of profiling

recurrent_ts_size

int recurrent_ts_size

number of data time steps in NN. 0 if non streaming RNN

rmf_apis

tflm_rmf_apis_t rmf_apis

data structure of Tflite-micro APIs

tflm_obj

void* tflm_obj

pointer of Tflite-micro runtime object

---

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

• mtb_ml_model.h