61ccde700f
* Refactored VPU tests not to use old interfaces * Added except of exceptions * Commented failing part of HDDL tests
260 lines
9.4 KiB
C++
260 lines
9.4 KiB
C++
// Copyright (C) 2018-2020 Intel Corporation
|
|
// SPDX-License-Identifier: Apache-2.0
|
|
//
|
|
|
|
#include "myriad_layers_tests.hpp"
|
|
|
|
using namespace InferenceEngine;
|
|
|
|
using myriadInferTests_nightly = myriadLayersTests_nightly;
|
|
|
|
TEST_F(myriadInferTests_nightly, NCHW_Input) {
|
|
std::string model = R"V0G0N(
|
|
<net name="Power" version="2" batch="1">
|
|
<layers>
|
|
<layer name="data" type="Input" precision="FP16" id="1">
|
|
<output>
|
|
<port id="1">
|
|
<dim>1</dim>
|
|
<dim>3</dim>
|
|
<dim>600</dim>
|
|
<dim>800</dim>
|
|
</port>
|
|
</output>
|
|
</layer>
|
|
<layer name="power" type="Power" precision="FP16" id="2">
|
|
<power_data power="1" scale="1" shift="0"/>
|
|
<input>
|
|
<port id="2">
|
|
<dim>1</dim>
|
|
<dim>3</dim>
|
|
<dim>600</dim>
|
|
<dim>800</dim>
|
|
</port>
|
|
</input>
|
|
<output>
|
|
<port id="3">
|
|
<dim>1</dim>
|
|
<dim>3</dim>
|
|
<dim>600</dim>
|
|
<dim>800</dim>
|
|
</port>
|
|
</output>
|
|
</layer>
|
|
</layers>
|
|
<edges>
|
|
<edge from-layer="1" from-port="1" to-layer="2" to-port="2"/>
|
|
</edges>
|
|
</net>
|
|
)V0G0N";
|
|
|
|
ASSERT_NO_THROW(readNetwork(model));
|
|
|
|
const auto& network = _cnnNetwork;
|
|
|
|
_inputsInfo = network.getInputsInfo();
|
|
_inputsInfo["data"]->setPrecision(Precision::FP16);
|
|
|
|
_outputsInfo = network.getOutputsInfo();
|
|
_outputsInfo["power"]->setPrecision(Precision::FP16);
|
|
|
|
ASSERT_NO_THROW(_exeNetwork = _vpuPluginPtr->LoadNetwork(network));
|
|
ASSERT_NO_THROW(_inferRequest = _exeNetwork.CreateInferRequest());
|
|
|
|
auto dims = _inputsInfo["data"]->getTensorDesc().getDims();
|
|
|
|
auto tensorDescNHWC = TensorDesc(Precision::FP16, dims, Layout::NHWC);
|
|
auto inputNHWC = make_shared_blob<ie_fp16>(tensorDescNHWC);
|
|
ASSERT_NO_THROW(inputNHWC->allocate());
|
|
|
|
auto outputNHWC = make_shared_blob<ie_fp16>(tensorDescNHWC);
|
|
ASSERT_NO_THROW(outputNHWC->allocate());
|
|
|
|
auto tensorDescNCHW = TensorDesc(Precision::FP16, dims, Layout::NCHW);
|
|
auto inputNCHW = make_shared_blob<ie_fp16>(tensorDescNCHW);
|
|
ASSERT_NO_THROW(inputNCHW->allocate());
|
|
|
|
auto outputNCHW = make_shared_blob<ie_fp16>(tensorDescNCHW);
|
|
ASSERT_NO_THROW(outputNCHW->allocate());
|
|
|
|
ASSERT_NO_THROW(GenRandomData(inputNHWC));
|
|
|
|
for (size_t i = 0; i < inputNHWC->size(); i++) {
|
|
inputNCHW->buffer().as<ie_fp16*>()[tensorDescNCHW.offset(i)] = inputNHWC->cbuffer().as<const ie_fp16*>()[tensorDescNHWC.offset(i)];
|
|
}
|
|
|
|
ASSERT_NO_THROW(_inferRequest.SetBlob("data", inputNHWC));
|
|
ASSERT_NO_THROW(_inferRequest.SetBlob("power", outputNHWC));
|
|
ASSERT_NO_THROW(_inferRequest.Infer());
|
|
ASSERT_NO_THROW(_inferRequest.SetBlob("data", inputNCHW));
|
|
ASSERT_NO_THROW(_inferRequest.SetBlob("power", outputNCHW));
|
|
ASSERT_NO_THROW(_inferRequest.Infer());
|
|
|
|
CompareCommonAbsolute(outputNHWC, outputNCHW, 0.0);
|
|
}
|
|
|
|
TEST_F(myriadInferTests_nightly, AddOutputToConvWithReLU) {
|
|
const std::string conv_model = R"V0G0N(
|
|
<Net name="conv_model" version="2" batch="1">
|
|
<layers>
|
|
<layer name="input" type="Input" precision="FP16" id="1">
|
|
<output>
|
|
<port id="1">
|
|
<dim>1</dim>
|
|
<dim>64</dim>
|
|
<dim>56</dim>
|
|
<dim>56</dim>
|
|
</port>
|
|
</output>
|
|
</layer>
|
|
<layer name="conv" type="Convolution" precision="FP16" id="2">
|
|
<convolution_data
|
|
stride-x="1" stride-y="1"
|
|
pad-x="0" pad-y="0"
|
|
kernel-x="1" kernel-y="1"
|
|
output="64"
|
|
group="1"/>
|
|
<input>
|
|
<port id="2">
|
|
<dim>1</dim>
|
|
<dim>64</dim>
|
|
<dim>56</dim>
|
|
<dim>56</dim>
|
|
</port>
|
|
</input>
|
|
<output>
|
|
<port id="3">
|
|
<dim>1</dim>
|
|
<dim>64</dim>
|
|
<dim>56</dim>
|
|
<dim>56</dim>
|
|
</port>
|
|
</output>
|
|
<weights offset="0" size="8192"/>
|
|
<biases offset="8192" size="128"/>
|
|
</layer>
|
|
</layers>
|
|
<edges>
|
|
<edge from-layer="1" from-port="1" to-layer="2" to-port="2"/>
|
|
</edges>
|
|
</Net>
|
|
)V0G0N";
|
|
|
|
const std::string full_model = R"V0G0N(
|
|
<Net name="full_model" version="2" batch="1">
|
|
<layers>
|
|
<layer name="input" type="Input" precision="FP16" id="1">
|
|
<output>
|
|
<port id="1">
|
|
<dim>1</dim>
|
|
<dim>64</dim>
|
|
<dim>56</dim>
|
|
<dim>56</dim>
|
|
</port>
|
|
</output>
|
|
</layer>
|
|
<layer name="conv" type="Convolution" precision="FP16" id="2">
|
|
<convolution_data
|
|
stride-x="1" stride-y="1"
|
|
pad-x="0" pad-y="0"
|
|
kernel-x="1" kernel-y="1"
|
|
output="64"
|
|
group="1"/>
|
|
<input>
|
|
<port id="2">
|
|
<dim>1</dim>
|
|
<dim>64</dim>
|
|
<dim>56</dim>
|
|
<dim>56</dim>
|
|
</port>
|
|
</input>
|
|
<output>
|
|
<port id="3">
|
|
<dim>1</dim>
|
|
<dim>64</dim>
|
|
<dim>56</dim>
|
|
<dim>56</dim>
|
|
</port>
|
|
</output>
|
|
<weights offset="0" size="8192"/>
|
|
<biases offset="8192" size="128"/>
|
|
</layer>
|
|
<layer name="relu" type="ReLU" precision="FP16" id="3">
|
|
<input>
|
|
<port id="4">
|
|
<dim>1</dim>
|
|
<dim>64</dim>
|
|
<dim>56</dim>
|
|
<dim>56</dim>
|
|
</port>
|
|
</input>
|
|
<output>
|
|
<port id="5">
|
|
<dim>1</dim>
|
|
<dim>64</dim>
|
|
<dim>56</dim>
|
|
<dim>56</dim>
|
|
</port>
|
|
</output>
|
|
</layer>
|
|
</layers>
|
|
<edges>
|
|
<edge from-layer="1" from-port="1" to-layer="2" to-port="2"/>
|
|
<edge from-layer="2" from-port="3" to-layer="3" to-port="4"/>
|
|
</edges>
|
|
</Net>
|
|
)V0G0N";
|
|
|
|
TBlob<uint8_t>::Ptr weights(GenWeights(8320 / sizeof(ie_fp16)));
|
|
|
|
InferenceEngine::Core ie;
|
|
auto conv_network = ie.ReadNetwork(conv_model, weights);
|
|
|
|
auto conv_inputs_info = conv_network.getInputsInfo();
|
|
conv_inputs_info["input"]->setPrecision(Precision::FP16);
|
|
|
|
auto conv_outputs_info = conv_network.getOutputsInfo();
|
|
conv_outputs_info["conv"]->setPrecision(Precision::FP16);
|
|
|
|
Blob::Ptr input = make_shared_blob<ie_fp16>({Precision::FP16, conv_inputs_info["input"]->getTensorDesc().getDims(), Layout::NCHW});
|
|
input->allocate();
|
|
GenRandomData(input);
|
|
|
|
Blob::Ptr conv_output;
|
|
{
|
|
ExecutableNetwork conv_exe;
|
|
ASSERT_NO_THROW(conv_exe = _vpuPluginPtr->LoadNetwork(conv_network));
|
|
|
|
InferRequest conv_req;
|
|
ASSERT_NO_THROW(conv_req = conv_exe.CreateInferRequest());
|
|
ASSERT_NO_THROW(conv_req.SetBlob("input", input));
|
|
ASSERT_NO_THROW(conv_output = conv_req.GetBlob("conv"));
|
|
ASSERT_NO_THROW(conv_req.Infer());
|
|
}
|
|
|
|
auto full_network = ie.ReadNetwork(full_model, weights);
|
|
|
|
full_network.addOutput("conv", 0);
|
|
|
|
auto full_inputs_info = full_network.getInputsInfo();
|
|
full_inputs_info["input"]->setPrecision(Precision::FP16);
|
|
|
|
auto full_outputs_info = full_network.getOutputsInfo();
|
|
full_outputs_info["conv"]->setPrecision(Precision::FP16);
|
|
full_outputs_info["relu"]->setPrecision(Precision::FP16);
|
|
|
|
Blob::Ptr full_output;
|
|
{
|
|
ExecutableNetwork full_exe;
|
|
ASSERT_NO_THROW(full_exe = _vpuPluginPtr->LoadNetwork(full_network));
|
|
|
|
InferRequest full_req;
|
|
ASSERT_NO_THROW(full_req = full_exe.CreateInferRequest());
|
|
ASSERT_NO_THROW(full_req.SetBlob("input", input));
|
|
ASSERT_NO_THROW(full_output = full_req.GetBlob("conv"));
|
|
ASSERT_NO_THROW(full_req.Infer());
|
|
}
|
|
|
|
CompareCommonAbsolute(full_output, conv_output, 0.0f);
|
|
}
|