[onert/odc] Auto-compilation. Tests (#14435)

This PR introduces tests for `odc:auto-compilation`. For [Issue]( #13288). From [Draft](#13530). ONE-DCO-1.0-Signed-off-by: Evgenii Maltsev [email protected]
Samsung · Dec 11, 2024 · c73dd4c · c73dd4c
1 parent a850110
commit c73dd4c
Show file tree

Hide file tree

Showing 3 changed files with 213 additions and 0 deletions.
diff --git a/tests/nnfw_api/CMakeLists.txt b/tests/nnfw_api/CMakeLists.txt
@@ -65,3 +65,7 @@ install(DIRECTORY ${NNPACKAGE_MODEL_DIR} DESTINATION ${NNPACKAGE_INSTALL_TARGET}
 # Install nnpackage test model (mobilenet)
 set(NNPACKAGE_MODEL_DIR ${NNAS_PROJECT_SOURCE_DIR}/runtime/contrib/TFLiteSharp/TFLiteTestApp/res/)
 install(DIRECTORY ${NNPACKAGE_MODEL_DIR} DESTINATION ${NNPACKAGE_INSTALL_TARGET}/mobilenet_v1_1.0_224)
+
+# Install nnpackage test model (conv2d)
+set(NNPACKAGE_MODEL_DIR ${NNAS_PROJECT_SOURCE_DIR}/tests/nnfw_api/res/autocompilation/)
+install(DIRECTORY ${NNPACKAGE_MODEL_DIR} DESTINATION ${NNPACKAGE_INSTALL_TARGET}/conv2d)
diff --git a/tests/nnfw_api/res/autocompilation/conv2d.circle b/tests/nnfw_api/res/autocompilation/conv2d.circle
diff --git a/tests/nnfw_api/src/OdcAutoCompilation.test.cc b/tests/nnfw_api/src/OdcAutoCompilation.test.cc
@@ -0,0 +1,209 @@
+/*
+ * Copyright (c) 2024 Samsung Electronics Co., Ltd. All Rights Reserved
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *    http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <gtest/gtest.h>
+#include <gmock/gmock.h>
+
+#include <nnfw_internal.h>
+
+#include "common.h"
+#include "fixtures.h"
+#include "CircleGen.h"
+#include "GenModelTest.h"
+#include "NNPackages.h"
+
+#include <chrono>
+using ::testing::FloatNear;
+using ::testing::Matcher;
+
+Matcher<std::vector<float>> FloatArrayNear(const std::vector<float> &values, float max_abs_error)
+{
+  std::vector<Matcher<float>> matchers;
+  matchers.reserve(values.size());
+  for (const float v : values)
+  {
+    matchers.emplace_back(FloatNear(v, max_abs_error));
+  }
+  return ElementsAreArray(matchers);
+}
+
+const std::string model_name = "conv2d";
+
+std::vector<std::vector<float>> input_tensors = {
+  {1, 2, 3, 4, 5.5, 6.7, 1, 2, 3, 4, 5.5, 6.7, 1, 2, 3, 4, 5.5, 6.7},
+  {7.1, 8, 9, 9.4, 1.2, 12.7, 1, 2, 3, 4, 5.5, 6.7, 7.1, 8, 9, 5.2, 3.2, 2.7},
+  {6.3, 2, 3.2, 4.7, 5.5, 3.7, 1, 2, 3.5, 4.2, 5.5, 6.7, 1.7, 2.2, 3.5, 4.5, 5.5, 6.7},
+  {8.4, 2.2, 3.7, 4.4, 5.5, 6.7, 1.2, 2.6, 3.3, 4.7, 5.5, 6.7, 1.3, 2, 3.2, 4.4, 5.5, 6.7},
+  {1.2, 2.6, 3.3, 4, 5.5, 6.7, 1, 2, 3.4, 4, 5.5, 6.7, 1, 2.6, 3, 4, 5.5, 6.7}};
+
+// Test for running a model with auto compilation
+TEST(TestOdcAutoCompilation, AutoCompilation_test)
+{
+  EXPECT_TRUE(input_tensors.size());
+
+  auto model_path = NNPackages::get().getModelAbsolutePath(model_name.c_str());
+
+  // setup session and load model
+  nnfw_session *session = nullptr;
+  nnfw_create_session(&session);
+
+  nnfw_load_model_from_modelfile(session, (model_path + ".circle").c_str());
+  nnfw_set_available_backends(session, "cpu");
+  nnfw_prepare(session);
+
+  // Delete minmax file
+  nnfw_odc_delete_minmax_file(session);
+
+  std::string compile_model_extension = "tvn";
+
+  // Delete previuos quantized and compiled model
+  std::string quantized_model_name = model_path + std::string(".q.circle");
+  std::string compiled_model_name = model_path + std::string(".") + compile_model_extension;
+  std::remove(quantized_model_name.c_str());
+  std::remove(compiled_model_name.c_str());
+
+  // setup ODC parameters
+  nnfw_set_quantized_model_path(session, quantized_model_name.c_str());
+  nnfw_set_quantization_type(session, NNFW_QUANTIZE_TYPE::NNFW_QUANTIZE_TYPE_U8_ASYM);
+
+  nnfw_set_codegen_model_path(session, compiled_model_name.c_str());
+
+  const int RUNS_COUNT_FOR_QUANTIZATION = input_tensors.size();
+  nnfw_set_odc_param_minmax_records_count(session, RUNS_COUNT_FOR_QUANTIZATION);
+
+  std::vector<std::vector<float>> float_model_output_tensors;
+  std::vector<std::vector<float>> quantized_model_output_tensors;
+
+  // Run FLOAT MODEL
+  // prepare input and output data and run model
+  for (size_t idx = 0; idx < RUNS_COUNT_FOR_QUANTIZATION; idx++)
+  {
+
+    // prepare input
+    nnfw_tensorinfo ti;
+
+    nnfw_input_tensorinfo(session, 0, &ti);
+    nnfw_set_input(session, 0, ti.dtype, input_tensors[idx].data(),
+                   sizeof(float) * input_tensors[idx].size());
+
+    // prepare output
+    nnfw_output_tensorinfo(session, 0, &ti);
+    uint32_t output_elements = 1;
+    for (int32_t i = 0; i < ti.rank; ++i)
+      output_elements *= ti.dims[i];
+
+    std::vector<float> output;
+    output.resize(output_elements);
+    nnfw_set_output(session, 0, ti.dtype, output.data(), sizeof(float) * output_elements);
+
+    // run model
+    NNFW_STATUS status =
+      nnfw_run_with_auto_compilation(session, (compile_model_extension + "-gen").c_str(),
+                                     NNFW_CODEGEN_PREF::NNFW_CODEGEN_PREF_DEFAULT);
+    EXPECT_TRUE(status == NNFW_STATUS_NO_ERROR);
+
+    float_model_output_tensors.push_back(output);
+  }
+
+  // Run COMPILED or QUANTIZED MODEL
+  for (size_t idx = 0; idx < RUNS_COUNT_FOR_QUANTIZATION; idx++)
+  {
+
+    // prepare input
+    nnfw_tensorinfo ti;
+
+    nnfw_input_tensorinfo(session, 0, &ti);
+    nnfw_set_input(session, 0, ti.dtype, input_tensors[idx].data(),
+                   sizeof(float) * input_tensors[idx].size());
+
+    // prepare output
+    nnfw_output_tensorinfo(session, 0, &ti);
+    uint32_t output_elements = 1;
+    for (int32_t i = 0; i < ti.rank; ++i)
+      output_elements *= ti.dims[i];
+
+    std::vector<float> output;
+    output.resize(output_elements);
+    nnfw_set_output(session, 0, ti.dtype, output.data(), sizeof(float) * output_elements);
+
+    // run quantized model
+    NNFW_STATUS status =
+      nnfw_run_with_auto_compilation(session, (compile_model_extension + "-gen").c_str(),
+                                     NNFW_CODEGEN_PREF::NNFW_CODEGEN_PREF_DEFAULT);
+    EXPECT_TRUE(status == NNFW_STATUS_NO_ERROR);
+
+    quantized_model_output_tensors.push_back(output);
+  }
+
+  // results comparison
+  for (size_t idx = 0; idx < quantized_model_output_tensors.size(); idx++)
+  {
+    EXPECT_THAT(float_model_output_tensors[idx],
+                FloatArrayNear(quantized_model_output_tensors[idx], 0.1f));
+  }
+
+  nnfw_close_session(session);
+  SUCCEED();
+}
+
+// Neg test for auto compilation
+TEST(TestOdcAutoCompilation, neg_autoCompilation_no_export_path)
+{
+
+  EXPECT_TRUE(input_tensors.size());
+
+  auto model_path = NNPackages::get().getModelAbsolutePath(model_name.c_str());
+
+  // setup session and load model
+  nnfw_session *session = nullptr;
+  nnfw_create_session(&session);
+
+  nnfw_load_model_from_modelfile(session, (model_path + ".circle").c_str());
+  nnfw_set_available_backends(session, "cpu");
+  nnfw_prepare(session);
+
+  // Delete minmax file
+  nnfw_odc_delete_minmax_file(session);
+
+  const int RUNS_COUNT_FOR_QUANTIZATION = 1;
+  nnfw_set_odc_param_minmax_records_count(session, RUNS_COUNT_FOR_QUANTIZATION);
+
+  // Run FLOAT MODEL
+  // prepare input
+  nnfw_tensorinfo ti;
+
+  nnfw_input_tensorinfo(session, 0, &ti);
+  nnfw_set_input(session, 0, ti.dtype, input_tensors[0].data(),
+                 sizeof(float) * input_tensors[0].size());
+
+  // prepare output
+  nnfw_output_tensorinfo(session, 0, &ti);
+  uint32_t output_elements = 1;
+  for (int32_t i = 0; i < ti.rank; ++i)
+    output_elements *= ti.dims[i];
+
+  std::vector<float> output;
+  output.resize(output_elements);
+  nnfw_set_output(session, 0, ti.dtype, output.data(), sizeof(float) * output_elements);
+
+  // run model
+  NNFW_STATUS status =
+    nnfw_run_with_auto_compilation(session, "", NNFW_CODEGEN_PREF::NNFW_CODEGEN_PREF_DEFAULT);
+  ASSERT_EQ(status, NNFW_STATUS_INVALID_STATE);
+
+  nnfw_close_session(session);
+  SUCCEED();
+}