D-Robotics Official Website
D-Robotics Developer Community
D-Robotics Official Website
D-Robotics Developer Community
English

hmct.api.ORTExecutor

Interface Description

The model inference function is provided by HMCT to support inference on intermediate models generated during the model conversion process.

Interface Form

class ORTExecutor(ORTExecutorBase): def __init__(self, model:ModelProto): def create_session(self): -> InferenceSession def get_support_devices(cls) -> list[str] def to(self, device:Union[str, list[str]]) -> None def get_inputs(self) -> list[str] def get_outputs(self) -> list[str] def inference(self, inputs：Dict[str, np.ndarray])->Dict[str, np.ndarray]

Member Function

Member FunctionDescriptionReturn ValueRange of Legal Parameter Values
ORTExecutor.init(self, model: ModelProto)Class initialization function, passing in an onnx ModelProto object.None.An onnx ModelProto object.
def create_session(self): -> InferenceSessionCreate a session for inference.InferenceSession for inference.Parameterless.
def get_support_devices(cls) -> list[str]Get all supported devices for the current ORTExecutor.String list of currently supported devices.Parameterless.
def to(self, device:Union[str, list[str]]) -> NoneModify model inference is to run DEVICE.None.'cuda', 'cpu' or a list of both.
def get_inputs(self) -> list[str]

Get a list composed of NodeArg class of model inputs, where NodeArg class has three member variables: name represents the input name, type represents the input datatype, and shape represents the input shape.

class NodeArg: def __init__(self, name: str, type: int, shape: Sequence[Union[int, str]]): self.name = name self.type = type self.shape = shape
List composed of NodeArg class of model inputs.Parameterless.
def get_outputs(self) -> list[str]Get a list composed of NodeArg class of model outputs, where NodeArg class has three member variables: name represents the output name, type represents the output datatype, and shape represents the output shape, NodeArg definition is the same as the description in get_inputs.List composed of NodeArg class of model outputs.Parameterless.
def inference(self, inputs：Dict[str, np.ndarray])->Dict[str, np.ndarray]Perform one forward inference using the input data, get the inference result and return it.A Dict, key is the model output name str, value is the output result np.ndarray

A dict with key as the string of the input name and value as the input to this inference

{ 'input_name0': np.ndarray, 'input_name1': np.ndarray, ... }