Skip to content

[NPU] Extend functional tests to cover sub-byte element types #32279

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 3 commits into from
Oct 8, 2025
Merged

[NPU] Extend functional tests to cover sub-byte element types #32279

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
ca47ee7
[NPU] Extend ZeroTensorTests
StefaniaHergane Sep 29, 2025
40da6dd
[NPU] Update func tests OVInferRequestIOTensorSetPrecisionTestNPU
StefaniaHergane Oct 6, 2025
4c6572c
[NPU] Update skip config with supported precision tests
StefaniaHergane Oct 6, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
18 changes: 17 additions & 1 deletion src/plugins/intel_npu/tests/functional/internal/backend/zero_tensor_tests.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -12,9 +12,25 @@
using namespace ov::test::behavior;

const std::vector<ov::AnyMap> configsInferRequestRunTests = {{}};
const std::vector<ov::element::Type> defaultTensorDataType = {ov::element::f32};

INSTANTIATE_TEST_SUITE_P(compatibility_smoke_BehaviorTest,
ZeroTensorTests,
::testing::Combine(::testing::Values(ov::test::utils::DEVICE_NPU),
::testing::ValuesIn(configsInferRequestRunTests)),
::testing::ValuesIn(configsInferRequestRunTests),
::testing::ValuesIn(defaultTensorDataType)),
ZeroTensorTests::getTestCaseName);

const std::vector<ov::element::Type> supportedTensorDataTypes = {
ov::element::f32, ov::element::f16, ov::element::bf16, ov::element::f8e4m3, ov::element::f8e5m2,
ov::element::f8e8m0, ov::element::nf4, ov::element::u2, ov::element::u4, ov::element::i4,
ov::element::u8, ov::element::i8, ov::element::u16, ov::element::i16, ov::element::u32,
ov::element::i32, ov::element::u64, ov::element::i64, ov::element::f64, ov::element::boolean,
};

INSTANTIATE_TEST_SUITE_P(compatibility_smoke_BehaviorTest,
ZeroTensorTestsCheckDataType,
::testing::Combine(::testing::Values(ov::test::utils::DEVICE_NPU),
::testing::ValuesIn(configsInferRequestRunTests),
::testing::ValuesIn(supportedTensorDataTypes)),
ZeroTensorTestsCheckDataType::getTestCaseName);
122 changes: 68 additions & 54 deletions src/plugins/intel_npu/tests/functional/internal/backend/zero_tensor_tests.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,9 +33,10 @@
#include "zero_backend.hpp"
#include "zero_tensor.hpp"

using CompilationParams = std::tuple<std::string, // Device name
ov::AnyMap // Config
>;
using CompilationParamsAndTensorDataType = std::tuple<std::string, // Device name
ov::AnyMap, // Config
ov::element::Type // Tensor data type
>;

using ::testing::AllOf;
using ::testing::HasSubstr;
Expand All @@ -44,19 +45,21 @@ namespace ov {
namespace test {
namespace behavior {
class ZeroTensorTests : public ov::test::behavior::OVPluginTestBase,
public testing::WithParamInterface<CompilationParams> {
public testing::WithParamInterface<CompilationParamsAndTensorDataType> {
protected:
std::shared_ptr<ov::Core> core = utils::PluginCache::get().core();
ov::AnyMap configuration;
ov::element::Type element_type;
std::shared_ptr<::intel_npu::ZeroInitStructsHolder> init_struct;
std::shared_ptr<::intel_npu::OptionsDesc> options = std::make_shared<::intel_npu::OptionsDesc>();
::intel_npu::Config npu_config = ::intel_npu::Config(options);

public:
static std::string getTestCaseName(testing::TestParamInfo<CompilationParams> obj) {
static std::string getTestCaseName(const testing::TestParamInfo<CompilationParamsAndTensorDataType>& obj) {
std::string targetDevice;
ov::AnyMap configuration;
std::tie(targetDevice, configuration) = obj.param;
ov::element::Type type;
std::tie(targetDevice, configuration, type) = obj.param;
std::replace(targetDevice.begin(), targetDevice.end(), ':', '_');
targetDevice = ov::test::utils::getTestsPlatformFromEnvironmentOr(ov::test::utils::DEVICE_NPU);

Expand All @@ -69,12 +72,15 @@ class ZeroTensorTests : public ov::test::behavior::OVPluginTestBase,
configItem.second.print(result);
}
}
if (!type.get_type_name().empty()) {
result << "tensorDataType=" << type.get_type_name() << "_";
}

return result.str();
}

void SetUp() override {
std::tie(target_device, configuration) = this->GetParam();
std::tie(target_device, configuration, element_type) = this->GetParam();

SKIP_IF_CURRENT_TEST_IS_DISABLED()
OVPluginTestBase::SetUp();
Expand All @@ -95,12 +101,12 @@ TEST_P(ZeroTensorTests, AllocateDeleteAllocateZeroTensor) {
SKIP_IF_CURRENT_TEST_IS_DISABLED()

auto shape = Shape{1, 2, 2, 2};
auto zero_tensor = std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, element::f32, shape, true);
auto zero_tensor = std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, element_type, shape, true);
ASSERT_TRUE(::intel_npu::zeroUtils::memory_was_allocated_in_the_same_l0_context(init_struct->getContext(),
zero_tensor->data()));

zero_tensor = {};
zero_tensor = std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, element::f32, shape, false);
zero_tensor = std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, element_type, shape, false);
ASSERT_TRUE(::intel_npu::zeroUtils::memory_was_allocated_in_the_same_l0_context(init_struct->getContext(),
zero_tensor->data()));

Expand All @@ -115,10 +121,10 @@ TEST_P(ZeroTensorTests, CheckSetSmallerShape) {

auto shape = Shape{1, 20, 20, 20};
auto shape_size = ov::shape_size(shape);
auto zero_tensor = std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, element::f32, shape, true);
auto zero_tensor = std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, element_type, shape, true);
EXPECT_EQ(shape, zero_tensor->get_shape());
EXPECT_EQ(shape_size, zero_tensor->get_size());
EXPECT_EQ(shape_size * sizeof(ov::element::f32), zero_tensor->get_byte_size());
EXPECT_EQ(shape_size * element_type.size(), zero_tensor->get_byte_size());

auto data = zero_tensor->data();

Expand All @@ -128,7 +134,7 @@ TEST_P(ZeroTensorTests, CheckSetSmallerShape) {
zero_tensor->set_shape(new_shape);
EXPECT_EQ(new_shape, zero_tensor->get_shape());
EXPECT_EQ(new_shape_size, zero_tensor->get_size());
EXPECT_EQ(new_shape_size * sizeof(ov::element::f32), zero_tensor->get_byte_size());
EXPECT_EQ(new_shape_size * element_type.size(), zero_tensor->get_byte_size());
EXPECT_EQ(data, zero_tensor->data());
ASSERT_TRUE(::intel_npu::zeroUtils::memory_was_allocated_in_the_same_l0_context(init_struct->getContext(),
zero_tensor->data()));
Expand All @@ -139,10 +145,10 @@ TEST_P(ZeroTensorTests, CheckSetBiggerShape) {

auto shape = Shape{1, 20, 20, 20};
auto shape_size = ov::shape_size(shape);
auto zero_tensor = std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, element::f32, shape, false);
auto zero_tensor = std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, element_type, shape, false);
EXPECT_EQ(shape, zero_tensor->get_shape());
EXPECT_EQ(shape_size, zero_tensor->get_size());
EXPECT_EQ(shape_size * sizeof(ov::element::f32), zero_tensor->get_byte_size());
EXPECT_EQ(shape_size * element_type.size(), zero_tensor->get_byte_size());

auto new_shape = Shape{1, 50, 50, 50};
auto new_shape_size = ov::shape_size(new_shape);
Expand All @@ -152,7 +158,7 @@ TEST_P(ZeroTensorTests, CheckSetBiggerShape) {
zero_tensor->set_shape(new_shape);
EXPECT_EQ(new_shape, zero_tensor->get_shape());
EXPECT_EQ(new_shape_size, zero_tensor->get_size());
EXPECT_EQ(new_shape_size * sizeof(ov::element::f32), zero_tensor->get_byte_size());
EXPECT_EQ(new_shape_size * element_type.size(), zero_tensor->get_byte_size());
ASSERT_TRUE(zero_tensor->memory_address_changed());
ASSERT_TRUE(::intel_npu::zeroUtils::memory_was_allocated_in_the_same_l0_context(init_struct->getContext(),
zero_tensor->data()));
Expand All @@ -164,122 +170,118 @@ TEST_P(ZeroTensorTests, CheckSetBiggerShape) {
TEST_P(ZeroTensorTests, CheckIsContinuousZeroTensorScalar) {
SKIP_IF_CURRENT_TEST_IS_DISABLED()

auto zero_tensor =
std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, ov::element::f32, Shape{}, true);
auto zero_tensor = std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, element_type, Shape{}, true);
auto data = zero_tensor->data();
auto strides = zero_tensor->get_strides();

ov::Tensor view_tensor;

view_tensor = ov::Tensor(ov::element::f32, ov::Shape{}, data, strides);
view_tensor = ov::Tensor(element_type, ov::Shape{}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), true);
}

TEST_P(ZeroTensorTests, CheckIsContinuousHostTensor1Dimension) {
SKIP_IF_CURRENT_TEST_IS_DISABLED()

auto zero_tensor =
std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, ov::element::f32, Shape{128}, true);
std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, element_type, Shape{128}, true);

auto data = zero_tensor->data();
auto strides = zero_tensor->get_strides();

ov::Tensor view_tensor;

view_tensor = ov::Tensor(ov::element::f32, ov::Shape{128}, data, strides);
view_tensor = ov::Tensor(element_type, ov::Shape{128}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), true);

view_tensor = ov::Tensor(ov::element::f32, ov::Shape{16}, data, strides);
view_tensor = ov::Tensor(element_type, ov::Shape{16}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), true);
}

TEST_P(ZeroTensorTests, CheckIsContinuousZeroTensor2Dimensions) {
SKIP_IF_CURRENT_TEST_IS_DISABLED()

auto zero_tensor =
std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, ov::element::f32, Shape{32, 128}, true);
std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, element_type, Shape{32, 128}, true);
auto data = zero_tensor->data();
auto strides = zero_tensor->get_strides();

ov::Tensor view_tensor;

view_tensor = ov::Tensor(ov::element::f32, Shape{16, 128}, data, strides);
view_tensor = ov::Tensor(element_type, Shape{16, 128}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), true);

view_tensor = ov::Tensor(ov::element::f32, Shape{1, 128}, data, strides);
view_tensor = ov::Tensor(element_type, Shape{1, 128}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), true);

view_tensor = ov::Tensor(ov::element::f32, Shape{1, 16}, data, strides);
view_tensor = ov::Tensor(element_type, Shape{1, 16}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), true);

view_tensor = ov::Tensor(ov::element::f32, Shape{2, 16}, data, strides);
view_tensor = ov::Tensor(element_type, Shape{2, 16}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), false);
}

TEST_P(ZeroTensorTests, CheckIsContinuousZeroTensor3Dimensions) {
SKIP_IF_CURRENT_TEST_IS_DISABLED()

auto zero_tensor =
std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, ov::element::f32, Shape{5, 32, 128}, true);
std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, element_type, Shape{5, 32, 128}, true);
auto data = zero_tensor->data();
auto strides = zero_tensor->get_strides();

ov::Tensor view_tensor;

view_tensor = ov::Tensor(ov::element::f32, Shape{2, 32, 128}, data, strides);
view_tensor = ov::Tensor(element_type, Shape{2, 32, 128}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), true);

view_tensor = ov::Tensor(ov::element::f32, Shape{2, 16, 128}, data, strides);
view_tensor = ov::Tensor(element_type, Shape{2, 16, 128}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), false);

view_tensor = ov::Tensor(ov::element::f32, Shape{1, 1, 128}, data, strides);
view_tensor = ov::Tensor(element_type, Shape{1, 1, 128}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), true);

view_tensor = ov::Tensor(ov::element::f32, Shape{1, 1, 64}, data, strides);
view_tensor = ov::Tensor(element_type, Shape{1, 1, 64}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), true);

view_tensor = ov::Tensor(ov::element::f32, Shape{1, 16, 128}, data, strides);
view_tensor = ov::Tensor(element_type, Shape{1, 16, 128}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), true);
}

TEST_P(ZeroTensorTests, CheckIsContinuousZeroTensor4Dimensions) {
SKIP_IF_CURRENT_TEST_IS_DISABLED()

auto zero_tensor = std::make_shared<::intel_npu::ZeroTensor>(init_struct,
npu_config,
ov::element::f32,
Shape{3, 5, 32, 128},
true);
auto zero_tensor =
std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, element_type, Shape{3, 5, 32, 128}, true);
auto data = zero_tensor->data();
auto strides = zero_tensor->get_strides();

ov::Tensor view_tensor;

view_tensor = ov::Tensor(ov::element::f32, Shape{1, 2, 32, 128}, data, strides);
view_tensor = ov::Tensor(element_type, Shape{1, 2, 32, 128}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), true);

view_tensor = ov::Tensor(ov::element::f32, Shape{2, 5, 32, 128}, data, strides);
view_tensor = ov::Tensor(element_type, Shape{2, 5, 32, 128}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), true);

view_tensor = ov::Tensor(ov::element::f32, Shape{2, 2, 32, 128}, data, strides);
view_tensor = ov::Tensor(element_type, Shape{2, 2, 32, 128}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), false);

view_tensor = ov::Tensor(ov::element::f32, Shape{1, 2, 5, 128}, data, strides);
view_tensor = ov::Tensor(element_type, Shape{1, 2, 5, 128}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), false);

view_tensor = ov::Tensor(ov::element::f32, Shape{3, 5, 32, 64}, data, strides);
view_tensor = ov::Tensor(element_type, Shape{3, 5, 32, 64}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), false);

view_tensor = ov::Tensor(ov::element::f32, Shape{1, 1, 16, 128}, data, strides);
view_tensor = ov::Tensor(element_type, Shape{1, 1, 16, 128}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), true);

view_tensor = ov::Tensor(ov::element::f32, Shape{2, 1, 16, 128}, data, strides);
view_tensor = ov::Tensor(element_type, Shape{2, 1, 16, 128}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), false);

view_tensor = ov::Tensor(ov::element::f32, Shape{1, 1, 1, 128}, data, strides);
view_tensor = ov::Tensor(element_type, Shape{1, 1, 1, 128}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), true);

view_tensor = ov::Tensor(ov::element::f32, Shape{1, 1, 1, 32}, data, strides);
view_tensor = ov::Tensor(element_type, Shape{1, 1, 1, 32}, data, strides);
EXPECT_EQ(view_tensor.is_continuous(), true);
}

Expand All @@ -289,8 +291,8 @@ TEST_P(ZeroTensorTests, CopyDefaultTensorExpectedThrow) {
auto shape = Shape{1, 2, 2, 2};

// shape size is unaligned to standard page size, expect to fail
auto data = static_cast<float*>(::operator new(ov::shape_size(shape) * sizeof(ov::element::f32)));
auto default_tensor = make_tensor(ov::element::f32, shape, data);
auto data = static_cast<float*>(::operator new(ov::shape_size(shape) * element_type.size()));
auto default_tensor = make_tensor(element_type, shape, data);
ASSERT_THROW(auto zero_tensor = std::make_shared<::intel_npu::ZeroTensor>(init_struct, default_tensor, npu_config),
::intel_npu::ZeroTensorException);

Expand All @@ -302,8 +304,7 @@ TEST_P(ZeroTensorTests, CopyZeroTensorAndKeepAlive) {

auto shape = Shape{1, 2, 2, 2};

auto zero_tensor =
std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, ov::element::f32, shape, true);
auto zero_tensor = std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, element_type, shape, true);

auto copy_zero_tensor = std::make_shared<::intel_npu::ZeroTensor>(init_struct, zero_tensor, npu_config);

Expand All @@ -325,8 +326,7 @@ TEST_P(ZeroTensorTests, CopyHostTensorAndKeepAlive) {
auto zero_context = std::make_shared<::intel_npu::RemoteContextImpl>(engine_backend);
auto shape = Shape{1, 2, 2, 2};

auto host_tensor =
std::make_shared<::intel_npu::ZeroHostTensor>(zero_context, init_struct, ov::element::f32, shape);
auto host_tensor = std::make_shared<::intel_npu::ZeroHostTensor>(zero_context, init_struct, element_type, shape);

auto copy_zero_tensor = std::make_shared<::intel_npu::ZeroTensor>(init_struct, host_tensor, npu_config);

Expand All @@ -349,7 +349,7 @@ TEST_P(ZeroTensorTests, CopyRemoteTensorAndKeepAlive) {
auto shape = Shape{1, 2, 2, 2};

auto remote_tensor =
std::make_shared<::intel_npu::ZeroRemoteTensor>(zero_context, init_struct, ov::element::f32, shape);
std::make_shared<::intel_npu::ZeroRemoteTensor>(zero_context, init_struct, element_type, shape);

auto copy_zero_tensor = std::make_shared<::intel_npu::ZeroTensor>(init_struct, remote_tensor, npu_config);

Expand All @@ -375,14 +375,28 @@ TEST_P(ZeroTensorTests, CopyRemoteTensorFromAnotherContextThrow) {
{ov::intel_npu::tensor_type.name(), {ov::intel_npu::TensorType::INPUT}}};

auto context = core->create_context(target_device, params);
auto remote_tensor = context.create_tensor(ov::element::f32, shape);
auto remote_tensor = context.create_tensor(element_type, shape);
auto remote_tensor_impl = get_tensor_impl(remote_tensor);

ASSERT_THROW(
auto zero_tensor = std::make_shared<::intel_npu::ZeroTensor>(init_struct, remote_tensor_impl, npu_config),
::intel_npu::ZeroTensorException);
}

using ZeroTensorTestsCheckDataType = ZeroTensorTests;

TEST_P(ZeroTensorTestsCheckDataType, CopyZeroTensorAndCheckTensorDataType) {
SKIP_IF_CURRENT_TEST_IS_DISABLED()

auto shape = Shape{1, 2, 2, 2};

auto zero_tensor = std::make_shared<::intel_npu::ZeroTensor>(init_struct, npu_config, element_type, shape, true);
EXPECT_EQ(element_type, zero_tensor->get_element_type());

auto copy_zero_tensor = std::make_shared<::intel_npu::ZeroTensor>(init_struct, zero_tensor, npu_config);
EXPECT_EQ(zero_tensor->get_element_type(), copy_zero_tensor->get_element_type());
}

} // namespace behavior
} // namespace test
} // namespace ov
Loading
Loading