Struct emulator::inference::InferenceConfig

ClassList > emulator > inference > InferenceConfig

Configuration options for inference backends. More...

• #include <inference_backend.hpp>

Public Attributes

Type	Name
BackendType	backend = BackendType::STUB Backend implementation to use.
int	device_id = -1 GPU device ID (-1 for CPU)
bool	dry_run = false If true, validate config and exit without running.
std::vector< std::string >	expected_input_vars Expected input variable names.
std::vector< std::string >	expected_output_vars Expected output variable names.
int	grid_height = 0 Height dimension (for spatial_mode)
int	grid_width = 0 Width dimension (for spatial_mode)
int	input_channels = 44 Number of input features per sample.
std::string	model_path Path to model file (TorchScript .pt for LibTorch)
int	output_channels = 50 Number of output features per sample.
bool	spatial_mode = /* multi line expression */ If true, reshape to [N, C, H, W] for CNN models.
bool	use_fp16 = false Use half precision (requires CUDA)
bool	verbose = false Enable verbose output (for debugging)

Detailed Description

Contains all parameters needed to initialize an inference backend, including model path, device selection, and tensor dimensions.

Public Attributes Documentation

variable backend

Backend implementation to use.

BackendType emulator::inference::InferenceConfig::backend;

---

variable device_id

GPU device ID (-1 for CPU)

int emulator::inference::InferenceConfig::device_id;

---

variable dry_run

If true, validate config and exit without running.

bool emulator::inference::InferenceConfig::dry_run;

---

variable expected_input_vars

Expected input variable names.

std::vector<std::string> emulator::inference::InferenceConfig::expected_input_vars;

---

variable expected_output_vars

Expected output variable names.

std::vector<std::string> emulator::inference::InferenceConfig::expected_output_vars;

---

variable grid_height

Height dimension (for spatial_mode)

int emulator::inference::InferenceConfig::grid_height;

---

variable grid_width

Width dimension (for spatial_mode)

int emulator::inference::InferenceConfig::grid_width;

---

variable input_channels

Number of input features per sample.

int emulator::inference::InferenceConfig::input_channels;

---

variable model_path

Path to model file (TorchScript .pt for LibTorch)

std::string emulator::inference::InferenceConfig::model_path;

---

variable output_channels

Number of output features per sample.

int emulator::inference::InferenceConfig::output_channels;

---

variable spatial_mode

If true, reshape to [N, C, H, W] for CNN models.

bool emulator::inference::InferenceConfig::spatial_mode;

---

variable use_fp16

Use half precision (requires CUDA)

bool emulator::inference::InferenceConfig::use_fp16;

---

variable verbose

Enable verbose output (for debugging)

bool emulator::inference::InferenceConfig::verbose;

---

---

The documentation for this class was generated from the following file components/emulator_comps/common/src/inference/inference_backend.hpp