cpp/precommit: fix device to improve reproducibility (#87)

Author: Wovchena

.github/workflows/test_precommit.yml

@@ -138,7 +138,8 @@ jobs:
       # .\w_openvino_toolkit_windows_2023.0.0.10926.b4452d56304_x86_64\setupvars.bat exits with 0 code without moving to a next command. Set PATH manually
       run: |
         set PATH=opencv\opencv\build\x64\vc16\bin;w_openvino_toolkit_windows_2023.0.0.10926.b4452d56304_x86_64\runtime\bin\intel64\Release;w_openvino_toolkit_windows_2023.0.0.10926.b4452d56304_x86_64\runtime\3rdparty\tbb\bin;%PATH%
-        .\build\Release\test_sanity.exe -d data -p tests\cpp\precommit\public_scope.json && .\build\Release\test_model_config -d data
+        .\build\Release\test_sanity.exe -d data -p tests\cpp\precommit\public_scope.json
+        .\build\Release\test_model_config -d data
   serving_api:
     strategy:
       fail-fast: false

tests/cpp/precommit/test_model_config.cpp

@@ -42,7 +42,7 @@ class MockAdapter : public OpenVINOInferenceAdapter {
             : OpenVINOInferenceAdapter() {
             auto core = ov::Core();
             auto model = core.read_model(modelPath);
-            loadModel(model, core, "AUTO");
+            loadModel(model, core, "CPU");
         }
 };
 
@@ -76,7 +76,8 @@ std::string string_format(const std::string &fmt, Args... args) {
 
 TEST_P(ClassificationModelParameterizedTest, TestClassificationDefaultConfig) {
     auto model_path = string_format(MODEL_PATH_TEMPLATE, GetParam().name.c_str(), GetParam().name.c_str());
-    auto model = ClassificationModel::create_model(DATA_DIR + "/" + model_path);
+    bool preload = true;
+    auto model = ClassificationModel::create_model(DATA_DIR + "/" + model_path, {}, preload, "CPU");
 
     auto ov_model = model->getModel();
 
@@ -98,7 +99,8 @@ TEST_P(ClassificationModelParameterizedTest, TestClassificationCustomConfig) {
         {"resize_type", "fit_to_window"},
         {"labels", mock_labels}
     };
-    auto model = ClassificationModel::create_model(DATA_DIR + "/" + model_path, configuration);
+    bool preload = true;
+    auto model = ClassificationModel::create_model(DATA_DIR + "/" + model_path, configuration, preload, "CPU");
 
     auto ov_model = model->getModel();
 
@@ -121,14 +123,15 @@ TEST_P(ClassificationModelParameterizedTestSaveLoad, TestClassificationCorrectne
     }
 
     auto model_path = string_format(MODEL_PATH_TEMPLATE, GetParam().name.c_str(), GetParam().name.c_str());
-    auto model = ClassificationModel::create_model(DATA_DIR + "/" + model_path);
+    bool preload = true;
+    auto model = ClassificationModel::create_model(DATA_DIR + "/" + model_path, {}, preload, "CPU");
 
     auto ov_model = model->getModel();
     ov::serialize(ov_model, TMP_MODEL_FILE);
 
     auto result = model->infer(image)->topLabels;
 
-    auto model_restored = ClassificationModel::create_model(TMP_MODEL_FILE);
+    auto model_restored = ClassificationModel::create_model(TMP_MODEL_FILE, {}, preload, "CPU");
     auto result_data = model_restored->infer(image);
     auto result_restored = result_data->topLabels;
 
@@ -143,7 +146,8 @@ TEST_P(ClassificationModelParameterizedTestSaveLoad, TestClassificationCorrectne
     }
 
     auto model_path = string_format(MODEL_PATH_TEMPLATE, GetParam().name.c_str(), GetParam().name.c_str());
-    auto model = ClassificationModel::create_model(DATA_DIR + "/" + model_path);
+    bool preload = true;
+    auto model = ClassificationModel::create_model(DATA_DIR + "/" + model_path, {}, preload, "CPU");
     auto ov_model = model->getModel();
     ov::serialize(ov_model, TMP_MODEL_FILE);
     auto result = model->infer(image)->topLabels;
@@ -159,7 +163,8 @@ TEST_P(ClassificationModelParameterizedTestSaveLoad, TestClassificationCorrectne
 
 TEST_P(SSDModelParameterizedTest, TestDetectionDefaultConfig) {
     auto model_path = string_format(MODEL_PATH_TEMPLATE, GetParam().name.c_str(), GetParam().name.c_str());
-    auto model = DetectionModel::create_model(DATA_DIR + "/" + model_path);
+    bool preload = true;
+    auto model = DetectionModel::create_model(DATA_DIR + "/" + model_path, {}, "", preload, "CPU");
 
     auto ov_model = model->getModel();
 
@@ -181,7 +186,8 @@ TEST_P(SSDModelParameterizedTest, TestDetectionCustomConfig) {
         {"resize_type", "fit_to_window"},
         {"labels", mock_labels}
     };
-    auto model = DetectionModel::create_model(DATA_DIR + "/" + model_path, configuration);
+    bool preload = true;
+    auto model = DetectionModel::create_model(DATA_DIR + "/" + model_path, configuration, "", preload, "CPU");
 
     auto ov_model = model->getModel();
 
@@ -204,7 +210,8 @@ TEST_P(DetectionModelParameterizedTestSaveLoad, TestDetctionCorrectnessAfterSave
     }
 
     auto model_path = string_format(MODEL_PATH_TEMPLATE, GetParam().name.c_str(), GetParam().name.c_str());
-    auto model = DetectionModel::create_model(DATA_DIR + "/" + model_path);
+    bool preload = true;
+    auto model = DetectionModel::create_model(DATA_DIR + "/" + model_path, {}, "", preload, "CPU");
 
     auto ov_model = model->getModel();
     ov::serialize(ov_model, TMP_MODEL_FILE);
@@ -215,7 +222,7 @@ TEST_P(DetectionModelParameterizedTestSaveLoad, TestDetctionCorrectnessAfterSave
     if (!image.data) {
         throw std::runtime_error{"Failed to read the image"};
     }
-    auto model_restored = DetectionModel::create_model(TMP_MODEL_FILE);
+    auto model_restored = DetectionModel::create_model(TMP_MODEL_FILE, {}, "", preload, "CPU");
     auto result_data = model_restored->infer(image);
     auto result_restored = result_data->objects;
 
@@ -227,8 +234,6 @@ TEST_P(DetectionModelParameterizedTestSaveLoad, TestDetctionCorrectnessAfterSave
         ASSERT_EQ(result[i].width, result_restored[i].width);
         ASSERT_EQ(result[i].height, result_restored[i].height);
     }
-
-    SUCCEED();
 }
 
 TEST_P(DetectionModelParameterizedTestSaveLoad, TestDetctionCorrectnessAfterSaveLoadWithAdapter) {
@@ -238,7 +243,8 @@ TEST_P(DetectionModelParameterizedTestSaveLoad, TestDetctionCorrectnessAfterSave
     }
 
     auto model_path = string_format(MODEL_PATH_TEMPLATE, GetParam().name.c_str(), GetParam().name.c_str());
-    auto model = DetectionModel::create_model(DATA_DIR + "/" + model_path);
+    bool preload = true;
+    auto model = DetectionModel::create_model(DATA_DIR + "/" + model_path, {}, "", preload, "CPU");
     auto ov_model = model->getModel();
     ov::serialize(ov_model, TMP_MODEL_FILE);
     auto result = model->infer(image)->objects;
@@ -261,8 +267,6 @@ TEST_P(DetectionModelParameterizedTestSaveLoad, TestDetctionCorrectnessAfterSave
         ASSERT_EQ(result[i].width, result_restored[i].width);
         ASSERT_EQ(result[i].height, result_restored[i].height);
     }
-
-    SUCCEED();
 }
 
 

tests/cpp/precommit/test_sanity.cpp

@@ -75,16 +75,19 @@ TEST_P(ModelParameterizedTest, SynchronousInference)
     }
 
     if ("DetectionModel" == GetParam().type) {
-        auto model = DetectionModel::create_model(DATA_DIR + "/" + model_path);
+        bool preload = true;
+        auto model = DetectionModel::create_model(DATA_DIR + "/" + model_path, {}, "", preload, "CPU");
         auto result = model->infer(image);
         EXPECT_GT(result->objects.size(), 0);
     } else if ("ClassificationModel" == GetParam().type) {
-        auto model = ClassificationModel::create_model(DATA_DIR + "/" + model_path);
+        bool preload = true;
+        auto model = ClassificationModel::create_model(DATA_DIR + "/" + model_path, {}, preload, "CPU");
         std::unique_ptr<ClassificationResult> result = model->infer(image);
         ASSERT_GT(result->topLabels.size(), 0);
         EXPECT_GT(result->topLabels.front().score, 0.0f);
     } else if ("SegmentationModel" == GetParam().type) {
-        auto model = SegmentationModel::create_model(DATA_DIR + "/" + model_path);
+        bool preload = true;
+        auto model = SegmentationModel::create_model(DATA_DIR + "/" + model_path, {}, preload, "CPU");
         auto result = model->infer(image)->asRef<ImageResultWithSoftPrediction>();
         ASSERT_GT(model->getContours(result).size(), 0);
     }