Skip to content
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

[GPU] Fix lack of support for 3D (and smaller) shapes in TransposeFusion pass #26440

Merged
merged 1 commit into from
Sep 5, 2024
Merged

[GPU] Fix lack of support for 3D (and smaller) shapes in TransposeFusion pass #26440

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
12 changes: 6 additions & 6 deletions src/plugins/intel_gpu/src/plugin/transformations/increase_position_ids_precision.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -29,8 +29,8 @@ IncreasePositionIdsPrecision::IncreasePositionIdsPrecision() {
using namespace ov::pass::pattern;
using ov::pass::pattern::op::Or;

auto gemm = wrap_type<ov::intel_gpu::op::Gemm>();
auto concat = wrap_type<ov::op::v0::Concat>({gemm, gemm});
auto gemm_or_matmul = wrap_type<ov::intel_gpu::op::Gemm, ov::op::v0::MatMul>();
auto concat = wrap_type<ov::op::v0::Concat>({gemm_or_matmul, gemm_or_matmul});
auto sin = wrap_type<ov::op::v0::Sin>({concat});
auto cos = wrap_type<ov::op::v0::Cos>({concat});

Expand All @@ -50,15 +50,15 @@ IncreasePositionIdsPrecision::IncreasePositionIdsPrecision() {
ov::matcher_pass_callback callback = [OV_CAPTURE_CPY_AND_THIS](ov::pass::pattern::Matcher& m) {
const auto& pattern_map = m.get_pattern_value_map();

auto gemm_node = std::dynamic_pointer_cast<ov::intel_gpu::op::Gemm>(pattern_map.at(gemm).get_node_shared_ptr());
auto matmul_node = std::dynamic_pointer_cast<ov::op::v0::MatMul>(pattern_map.at(gemm_or_matmul).get_node_shared_ptr());
auto cos_node = std::dynamic_pointer_cast<ov::op::v0::Cos>(pattern_map.at(cos).get_node_shared_ptr());
auto sin_node = std::dynamic_pointer_cast<ov::op::v0::Sin>(pattern_map.at(sin).get_node_shared_ptr());

if (!gemm_node || transformation_callback(gemm_node))
if (!matmul_node || transformation_callback(matmul_node))
return false;

const auto desired_et = ov::element::f32;
const auto original_et = gemm_node->get_output_element_type(0);
const auto original_et = matmul_node->get_output_element_type(0);
if (original_et == desired_et)
return false;

Expand Down Expand Up @@ -112,7 +112,7 @@ IncreasePositionIdsPrecision::IncreasePositionIdsPrecision() {
}
};

bool is_changed = insert_converts_before_if_needed(gemm_node);
bool is_changed = insert_converts_before_if_needed(matmul_node);

if (is_changed) {
insert_converts_after_if_needed(cos_node);
Expand Down
7 changes: 7 additions & 0 deletions src/plugins/intel_gpu/src/plugin/transformations/transpose_fusion.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -52,6 +52,13 @@ bool has_optimized_version(const ov::Output<ov::Node>& output, bool supports_imm
{3, 0, 1, 2},
};

const auto expected_dims_num = 4;
const auto original_dims_num = transpose_order.size();
if (original_dims_num < expected_dims_num) {
transpose_order.resize(expected_dims_num);
std::iota(transpose_order.begin() + original_dims_num, transpose_order.end(), original_dims_num);
}

if (!cldnn::one_of(transpose_order, allowed_orders))
return false;

Expand Down
46 changes: 46 additions & 0 deletions src/plugins/intel_gpu/tests/unit/transformations/increase_position_ids_precision_test.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -130,3 +130,49 @@ TEST_F(TransformationTestsF, IncreasePositionIdsPrecisionWithUnsqueeze) {
}
comparator.enable(FunctionsComparator::CmpValues::ATTRIBUTES);
}

TEST_F(TransformationTestsF, IncreasePositionIdsMatmulWithoutUnsqueeze) {
{
auto input = std::make_shared<ov::op::v0::Parameter>(ov::element::i64, ov::PartialShape{ -1, -1 });
auto input_convert = std::make_shared<ov::op::v0::Convert>(input, ov::element::i32);
auto input_unsqueeze = std::make_shared<ov::op::v0::Unsqueeze>(input_convert, std::make_shared<ov::op::v0::Constant>(ov::element::i64, ov::Shape{1}, std::vector<int64_t>{1}));
auto input_convert_fp = std::make_shared<ov::op::v0::Convert>(input_unsqueeze, ov::element::f16);
auto rotary_embd_const = std::make_shared<ov::op::v0::Constant>(ov::element::f16, ov::Shape{1, 64, 1});

auto matmul = std::make_shared<ov::op::v0::MatMul>(input_convert_fp, rotary_embd_const);
auto concat = std::make_shared<ov::op::v0::Concat>(ov::OutputVector{matmul, matmul}, 2);

auto cos = std::make_shared<ov::op::v0::Cos>(concat);
auto sin = std::make_shared<ov::op::v0::Sin>(concat);

auto rope_input = std::make_shared<ov::op::v0::Parameter>(ov::element::f16, ov::PartialShape::dynamic(4));
auto rope = std::make_shared<ov::op::internal::RoPE>(ov::OutputVector{rope_input, cos, sin}, ov::op::internal::RoPE::Config());

model = std::make_shared<ov::Model>(ov::NodeVector{ rope }, ov::ParameterVector{ input, rope_input });
manager.register_pass<IncreasePositionIdsPrecision>();
}
{
auto input = std::make_shared<ov::op::v0::Parameter>(ov::element::i64, ov::PartialShape{ -1, -1 });
auto input_convert = std::make_shared<ov::op::v0::Convert>(input, ov::element::i32);
auto input_unsqueeze = std::make_shared<ov::op::v0::Unsqueeze>(input_convert, std::make_shared<ov::op::v0::Constant>(ov::element::i64, ov::Shape{1}, std::vector<int64_t>{1}));
auto rotary_embd_const = std::make_shared<ov::op::v0::Constant>(ov::element::f16, ov::Shape{1, 64, 1});

auto input_convert_f32 = std::make_shared<ov::op::v0::Convert>(input_unsqueeze, ov::element::f32);
auto rotary_embd_const_convert_f32 = std::make_shared<ov::op::v0::Convert>(rotary_embd_const, ov::element::f32);

auto matmul = std::make_shared<ov::op::v0::MatMul>(input_convert_f32, rotary_embd_const_convert_f32);
auto concat = std::make_shared<ov::op::v0::Concat>(ov::OutputVector{matmul, matmul}, 2);

auto cos = std::make_shared<ov::op::v0::Cos>(concat);
auto sin = std::make_shared<ov::op::v0::Sin>(concat);

auto cos_convert = std::make_shared<ov::op::v0::Convert>(cos, ov::element::f16);
auto sin_convert = std::make_shared<ov::op::v0::Convert>(sin, ov::element::f16);

auto rope_input = std::make_shared<ov::op::v0::Parameter>(ov::element::f16, ov::PartialShape::dynamic(4));
auto rope = std::make_shared<ov::op::internal::RoPE>(ov::OutputVector{rope_input, cos_convert, sin_convert}, ov::op::internal::RoPE::Config());

model_ref = std::make_shared<ov::Model>(ov::NodeVector{ rope }, ov::ParameterVector{ input, rope_input });
}
comparator.enable(FunctionsComparator::CmpValues::ATTRIBUTES);
}
58 changes: 58 additions & 0 deletions src/plugins/intel_gpu/tests/unit/transformations/transpose_matmul_fusion_test.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -124,6 +124,64 @@ TEST_F(TransformationTestsF, TranposeMatmulFusion4) {
}
}

TEST_F(TransformationTestsF, TranposeMatmulFusion5) {
{
auto input_a = std::make_shared<ov::op::v0::Parameter>(ov::element::f32, ov::PartialShape::dynamic(3));
auto input_b = std::make_shared<ov::op::v0::Parameter>(ov::element::f32, ov::PartialShape::dynamic(3));
auto matmul = std::make_shared<ov::op::v0::MatMul>(input_a, input_b);
auto tranpose_c_const = ov::op::v0::Constant::create(ov::element::i64, ov::Shape{3}, {0, 2, 1});
auto tranpose_c = std::make_shared<ov::op::v1::Transpose>(matmul, tranpose_c_const);

model = std::make_shared<ov::Model>(ov::NodeVector{ tranpose_c }, ov::ParameterVector{ input_a, input_b });

const auto supports_immad = false;
manager.register_pass<TransposeFusion>(supports_immad);
}
{
std::vector<int64_t> order_a = {0, 1, 2};
std::vector<int64_t> order_b = {0, 1, 2};
std::vector<int64_t> order_c = {0, 2, 1};
auto input_a = std::make_shared<ov::op::v0::Parameter>(ov::element::f32, ov::PartialShape::dynamic(3));
auto input_b = std::make_shared<ov::op::v0::Parameter>(ov::element::f32, ov::PartialShape::dynamic(3));
auto gemm = std::make_shared<ov::intel_gpu::op::Gemm>(input_a, input_b, order_a, order_b, order_c, ov::element::undefined);

model_ref = std::make_shared<ov::Model>(ov::NodeVector{ gemm }, ov::ParameterVector{ input_a, input_b });
comparator.enable(FunctionsComparator::ATTRIBUTES);
}
}

TEST_F(TransformationTestsF, TranposeMatmulFusion6) {
auto input_a = std::make_shared<ov::op::v0::Parameter>(ov::element::f32, ov::PartialShape::dynamic(2));
auto input_b = std::make_shared<ov::op::v0::Parameter>(ov::element::f32, ov::PartialShape::dynamic(2));
auto matmul = std::make_shared<ov::op::v0::MatMul>(input_a, input_b);
auto tranpose_c_const = ov::op::v0::Constant::create(ov::element::i64, ov::Shape{2}, {1, 0});
auto tranpose_c = std::make_shared<ov::op::v1::Transpose>(matmul, tranpose_c_const);

model = std::make_shared<ov::Model>(ov::NodeVector{ tranpose_c }, ov::ParameterVector{ input_a, input_b });

const auto supports_immad = false;
manager.register_pass<TransposeFusion>(supports_immad);

model_ref = model->clone();
comparator.enable(FunctionsComparator::ATTRIBUTES);
}

TEST_F(TransformationTestsF, TranposeMatmulFusion7) {
auto input_a = std::make_shared<ov::op::v0::Parameter>(ov::element::f32, ov::PartialShape{2, 4});
auto input_b = std::make_shared<ov::op::v0::Parameter>(ov::element::f32, ov::PartialShape{4, 2});
auto matmul = std::make_shared<ov::op::v0::MatMul>(input_a, input_b);
auto tranpose_c_const = ov::op::v0::Constant::create(ov::element::i64, ov::Shape{2}, {1, 0});
auto tranpose_c = std::make_shared<ov::op::v1::Transpose>(matmul, tranpose_c_const);

model = std::make_shared<ov::Model>(ov::NodeVector{ tranpose_c }, ov::ParameterVector{ input_a, input_b });

const auto supports_immad = false;
manager.register_pass<TransposeFusion>(supports_immad);

model_ref = model->clone();
comparator.enable(FunctionsComparator::ATTRIBUTES);
}

TEST_F(TransformationTestsF, TranposeMatmulFusion_Illegal_1) {
{
auto input_a = std::make_shared<ov::op::v0::Parameter>(ov::element::f32, ov::PartialShape{10, 20});
Expand Down
Loading