TheRock/rocm-libraries/dnn-providers/miopen-provider/tests/engines/plans/TestMiopenBatchnormFwdTrainingPlan.cpp

// Copyright © Advanced Micro Devices, Inc., or its affiliates.
// SPDX-License-Identifier:  MIT

#include "engines/plans/MiopenBatchnormFwdTrainingPlan.hpp"
#include <gtest/gtest.h>
#include <hipdnn_flatbuffers_sdk/flatbuffer_utilities/GraphWrapper.hpp>
#include <hipdnn_test_sdk/utilities/FlatbufferGraphTestUtils.hpp>

using namespace miopen_plugin;

TEST(TestBatchnormFwdTrainingParams, InitializesRequiredTensorsFromValidGraph)
{
    auto builder = hipdnn_test_sdk::utilities::createValidBatchnormFwdTrainingGraph();
    const hipdnn_flatbuffers_sdk::flatbuffer_utilities::GraphWrapper graph(
        builder.GetBufferPointer(), builder.GetSize());

    const auto& node = graph.getNode(0);
    auto* attrs = node.attributes_as_BatchnormAttributes();
    ASSERT_NE(attrs, nullptr);

    const BatchnormFwdTrainingParams params(*attrs, graph.getTensorMap());

    // All required tensors should be initialized
    EXPECT_NO_THROW(params.x());
    EXPECT_NO_THROW(params.y());
    EXPECT_NO_THROW(params.scale());
    EXPECT_NO_THROW(params.bias());
}

TEST(TestBatchnormFwdTrainingParams, ExtractsEpsilonValueCorrectly)
{
    auto builder = hipdnn_test_sdk::utilities::createValidBatchnormFwdTrainingGraph();
    const hipdnn_flatbuffers_sdk::flatbuffer_utilities::GraphWrapper graph(
        builder.GetBufferPointer(), builder.GetSize());

    const auto& node = graph.getNode(0);
    auto* attrs = node.attributes_as_BatchnormAttributes();
    ASSERT_NE(attrs, nullptr);

    const BatchnormFwdTrainingParams params(*attrs, graph.getTensorMap());

    // Epsilon should be extracted as double
    EXPECT_NEAR(params.epsilonValue(nullptr, 0), 1e-5, 1e-10);
}

TEST(TestBatchnormFwdTrainingParams, HandlesMeanVariancePresent)
{
    auto builder = hipdnn_test_sdk::utilities::createValidBatchnormFwdTrainingGraph(
        {1, 3, 14, 14}, {1, 3, 14, 14}, true);
    const hipdnn_flatbuffers_sdk::flatbuffer_utilities::GraphWrapper graph(
        builder.GetBufferPointer(), builder.GetSize());

    const auto& node = graph.getNode(0);
    auto* attrs = node.attributes_as_BatchnormAttributes();
    ASSERT_NE(attrs, nullptr);

    const BatchnormFwdTrainingParams params(*attrs, graph.getTensorMap());

    EXPECT_TRUE(params.hasSaveMeanVariance());
    EXPECT_NO_THROW(params.mean());
    EXPECT_NO_THROW(params.invVariance());
}

TEST(TestBatchnormFwdTrainingParams, HandlesMeanVarianceMissing)
{
    auto builder = hipdnn_test_sdk::utilities::createValidBatchnormFwdTrainingGraph(
        {1, 3, 14, 14}, {1, 3, 14, 14}, false);
    const hipdnn_flatbuffers_sdk::flatbuffer_utilities::GraphWrapper graph(
        builder.GetBufferPointer(), builder.GetSize());

    const auto& node = graph.getNode(0);
    auto* attrs = node.attributes_as_BatchnormAttributes();
    ASSERT_NE(attrs, nullptr);

    const BatchnormFwdTrainingParams params(*attrs, graph.getTensorMap());

    EXPECT_FALSE(params.hasSaveMeanVariance());
}

TEST(TestBatchnormFwdTrainingParams, HasRunningStatsReturnsFalseWhenNotProvided)
{
    auto builder = hipdnn_test_sdk::utilities::createValidBatchnormFwdTrainingGraph();
    const hipdnn_flatbuffers_sdk::flatbuffer_utilities::GraphWrapper graph(
        builder.GetBufferPointer(), builder.GetSize());

    const auto& node = graph.getNode(0);
    auto* attrs = node.attributes_as_BatchnormAttributes();
    ASSERT_NE(attrs, nullptr);

    const BatchnormFwdTrainingParams params(*attrs, graph.getTensorMap());

    EXPECT_FALSE(params.hasRunningStats());
}