Public Member Functions
	__init__ (self, filepath)

	load (self)

	feed_random_inputs (self)

	run (self)

	get_outputs (self)

Data Fields
	filepath

	session

	inputs

	inputs_size

	inputs_data

	outputs

	outputs_size

	outs

Detailed Description

Definition at line 54 of file validate_onnx2circle.py.

Constructor & Destructor Documentation

◆ init()

validate_onnx2circle.OnnxRunner.__init__	(	self,
		filepath
	)

Definition at line 55 of file validate_onnx2circle.py.

    def __init__(self, filepath):
        self.filepath = filepath
        self.session = None
        self.inputs = None
        self.inputs_size = None
        self.inputs_data = None
        self.outputs = None
        self.outputs_size = None
 

Member Function Documentation

◆ feed_random_inputs()

validate_onnx2circle.OnnxRunner.feed_random_inputs ( self )

Definition at line 77 of file validate_onnx2circle.py.

    def feed_random_inputs(self):
        self.inputs = self.session.get_inputs()
        self.inputs_size = len(self.inputs)
        # reset input dictionary
        self.inputs_data = {}
        for in_idx in range(self.inputs_size):
            input_shape = self.inputs[in_idx].shape
            input_type = self.inputs[in_idx].type
            if input_type == 'tensor(float)':
                torch_type = torch.float32
            else:
                # TODO support other dtype
                raise SystemExit("Unsupported input dtype")
 
            x = torch.randn(input_shape, dtype=torch_type)
            input_npa = to_numpy(x)
            self.inputs_data.update({self.inputs[in_idx].name: input_npa})
 
            # save NHWC form of input for luci-interpreter
            input_npa_nhwc = to_nhwc(input_npa)
            input_npa_nhwc.tofile(circle_filepath + ".input" + str(in_idx))
 

References Operation.inputs, Request.inputs, circlechef::CircleImport.inputs(), crew::Part.inputs, luci::CircleReader.inputs(), luci::PGroup.inputs, luci::pass::Expression.inputs, mio::circle::Reader.inputs(), moco::ModelSignature.inputs(), tflinspect::Reader.inputs(), tflchef::TFliteImport.inputs(), tflread::Reader.inputs(), luci_interpreter::CircleReader.inputs(), onert_micro::core::reader::OMCircleReader.inputs(), moco::tf::test::TFNodeBuildTester.inputs(), moco::test::TFNodeBuildTester.inputs(), moco::test::TFNodeBuildTester.inputs(), luci::CircleFakeQuant.inputs(), moco::TFFakeQuantWithMinMaxVars.inputs(), luci::CircleAddN.inputs(), luci::CircleCustom.inputs(), luci::CircleAddN.inputs(), luci::CircleCustom.inputs(), loco::Graph.inputs(), nnkit::support::onnx::Runner.inputs(), ann::Operation.inputs(), loco::Graph.inputs(), luci::CircleFakeQuant.inputs(), moco::TFFakeQuantWithMinMaxVars.inputs(), nnc::sir::CallFunction.inputs, validate_onnx2circle.OnnxRunner.inputs, onert_micro::execute::OMRuntimeKernel.inputs, nnfw_custom_kernel_params.inputs, package.infer.session.inputs, onert::exec::IODescription.inputs, validate_onnx2circle.OnnxRunner.inputs_data, onert_micro::execute::OMRuntimeKernel.inputs_data, validate_onnx2circle.OnnxRunner.inputs_size, validate_onnx2circle.OnnxRunner.session, NNFW_SESSION.session, validate_onnx2circle.str, validate_onnx2circle.to_nhwc(), and validate_onnx2circle.to_numpy().

◆ get_outputs()

validate_onnx2circle.OnnxRunner.get_outputs ( self )

Definition at line 102 of file validate_onnx2circle.py.

    def get_outputs(self):
        self.outputs = self.session.get_outputs()
        self.outputs_size = len(self.outputs)
 
 
# Run ONNX model

References validate_onnx2circle.OnnxRunner.get_outputs(), Operation.outputs, Request.outputs, circlechef::CircleImport.outputs(), crew::Part.outputs, luci::CircleReader.outputs(), mio::circle::Reader.outputs(), moco::ModelSignature.outputs(), tflinspect::Reader.outputs(), tflchef::TFliteImport.outputs(), tflread::Reader.outputs(), luci_interpreter::CircleReader.outputs(), onert_micro::core::reader::OMCircleReader.outputs(), mio::circle::Reader.outputs(), loco::Graph.outputs(), nnkit::support::onnx::Runner.outputs(), ann::Operation.outputs(), loco::Graph.outputs(), luci::PGroup.outputs, nnc::sir::CallFunction.outputs, validate_onnx2circle.OnnxRunner.outputs, onert_micro::execute::OMRuntimeKernel.outputs, nnfw_custom_kernel_params.outputs, package.infer.session.outputs, onert::exec::IODescription.outputs, validate_onnx2circle.OnnxRunner.outputs_size, validate_onnx2circle.OnnxRunner.session, and NNFW_SESSION.session.

Referenced by validate_onnx2circle.OnnxRunner.get_outputs().

◆ load()

validate_onnx2circle.OnnxRunner.load ( self )

Definition at line 64 of file validate_onnx2circle.py.

    def load(self):
        model = onnx.load(self.filepath)
        onnx.checker.check_model(model)
        options = ort.SessionOptions()
        # NOTE this is needed for U18.04
        # referenced: https://github.com/microsoft/onnxruntime/issues/10113
        options.intra_op_num_threads = 4
        # NOTE set `providers` for https://github.com/microsoft/onnxruntime/issues/17631
        providers = ort.get_available_providers()
        self.session = ort.InferenceSession(self.filepath,
                                            sess_options=options,
                                            providers=providers)
 

References validate_onnx2circle.OnnxRunner.filepath, validate_onnx2circle.OnnxRunner.session, and NNFW_SESSION.session.

◆ run()

validate_onnx2circle.OnnxRunner.run ( self )

Definition at line 99 of file validate_onnx2circle.py.

    def run(self):
        self.outs = self.session.run(None, self.inputs_data)
 

Referenced by package.infer.session.inference().

Field Documentation

◆ filepath

validate_onnx2circle.OnnxRunner.filepath

Definition at line 56 of file validate_onnx2circle.py.

Referenced by validate_onnx2circle.OnnxRunner.load().

◆ inputs

validate_onnx2circle.OnnxRunner.inputs

Definition at line 58 of file validate_onnx2circle.py.

Referenced by validate_onnx2circle.OnnxRunner.feed_random_inputs(), and package.infer.session.set_inputs().

◆ inputs_data

validate_onnx2circle.OnnxRunner.inputs_data

Definition at line 60 of file validate_onnx2circle.py.

Referenced by validate_onnx2circle.OnnxRunner.feed_random_inputs().

◆ inputs_size

validate_onnx2circle.OnnxRunner.inputs_size

Definition at line 59 of file validate_onnx2circle.py.

Referenced by validate_onnx2circle.OnnxRunner.feed_random_inputs().

◆ outputs

validate_onnx2circle.OnnxRunner.outputs

Definition at line 61 of file validate_onnx2circle.py.

Referenced by validate_onnx2circle.OnnxRunner.get_outputs(), package.infer.session.inference(), and package.infer.session.set_outputs().

◆ outputs_size

validate_onnx2circle.OnnxRunner.outputs_size

Definition at line 62 of file validate_onnx2circle.py.

Referenced by validate_onnx2circle.OnnxRunner.get_outputs().

◆ outs

validate_onnx2circle.OnnxRunner.outs

Definition at line 100 of file validate_onnx2circle.py.

◆ session

validate_onnx2circle.OnnxRunner.session

Definition at line 57 of file validate_onnx2circle.py.

Referenced by validate_onnx2circle.OnnxRunner.feed_random_inputs(), validate_onnx2circle.OnnxRunner.get_outputs(), and validate_onnx2circle.OnnxRunner.load().

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

compiler/one-cmds/validate-onnx2circle/validate_onnx2circle.py

Public Member Functions

Data Fields