Skip to content

[ONNX][Attention] Enable conversion of ONNX Attention by function #33316

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
mlukasze merged 6 commits into from
Jan 29, 2026
Merged

[ONNX][Attention] Enable conversion of ONNX Attention by function #33316

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
6 commits
Select commit Hold shift + click to select a range
56afe0d
Enable ONNX Attention by function
mitruska Dec 19, 2025
2f5bb71
Port ONNX version 1.18.0 upgrade
mitruska Dec 19, 2025
34e3d17
Merge branch 'master' into mitruska/attention_onnx_func
mvafin Jan 8, 2026
426f4db
Merge branch 'master' into mitruska/attention_onnx_func
mvafin Jan 9, 2026
6daab4a
Merge branch 'master' into mitruska/attention_onnx_func
mvafin Jan 16, 2026
0c6715e
Merge branch 'master' into mitruska/attention_onnx_func
mvafin Jan 28, 2026
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
2 changes: 1 addition & 1 deletion src/frontends/onnx/frontend/CMakeLists.txt
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -77,7 +77,7 @@ ov_add_frontend(NAME onnx
FILEDESCRIPTION "FrontEnd to load and convert ONNX file format"
LINK_LIBRARIES openvino_onnx_common openvino::core::dev)

set(ONNX_OPSET_VERSION 21 CACHE INTERNAL "Supported version of ONNX operator set")
set(ONNX_OPSET_VERSION 23 CACHE INTERNAL "Supported version of ONNX operator set")
target_compile_definitions(${TARGET_NAME} PRIVATE ONNX_OPSET_VERSION=${ONNX_OPSET_VERSION})

if(BUILD_SHARED_LIBS)
Expand Down
3 changes: 1 addition & 2 deletions src/frontends/onnx/frontend/src/core/transform.cpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -94,8 +94,7 @@ void ov::frontend::onnx::transform::expand_onnx_functions(ModelProto& model_prot

// Retrieve the operation schema from ONNX library
int opset_version = static_cast<int>(get_opset_version(model_proto, node.domain()));
const auto* schema_registry = OpSchemaRegistry::Instance();
const auto node_op_schema = schema_registry->GetSchema(node.op_type(), opset_version, node.domain());
const auto node_op_schema = OpSchemaRegistry::Schema(node.op_type(), opset_version, node.domain());

// Check if operation schema found
if (!node_op_schema) {
Expand Down
2 changes: 1 addition & 1 deletion src/frontends/onnx/frontend/src/core/transform.hpp
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,7 +13,7 @@ namespace transform {

using ::ONNX_NAMESPACE::ModelProto;

static const std::vector<std::string> onnx_functions_to_expand = {"Bernoulli"};
static const std::vector<std::string> onnx_functions_to_expand = {"Bernoulli", "Attention"};

/// \brief Replace nodes with expanded body of ONNX functions
///
Expand Down
1 change: 0 additions & 1 deletion src/frontends/onnx/tests/__init__.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -177,7 +177,6 @@ def xfail_test(reason="Mark the test as expected to fail", strict=True):
skip_issue_119896 = pytest.mark.skip(reason="Unsupported element type: FLOAT8")

# ONNX 1.18
xfail_issue_171766 = pytest.mark.skip(reason="Unsupported feature: Attention")
xfail_issue_171767 = pytest.mark.skip(reason="Unsupported element type: FLOAT4E2M1")
xfail_issue_171768 = pytest.mark.skip(reason="Unsupported feature: RMSNormalization")
xfail_issue_171770 = pytest.mark.skip(reason="Unsupported feature: RotaryEmbedding")
Expand Down
51 changes: 0 additions & 51 deletions src/frontends/onnx/tests/tests_python/test_backend.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -71,7 +71,6 @@
xfail_issue_139936,
xfail_issue_139937,
xfail_issue_139938,
xfail_issue_171766,
xfail_issue_171767,
xfail_issue_171768,
xfail_issue_171770,
Expand Down Expand Up @@ -629,56 +628,6 @@ def expect_fail(test_case_path, xfail): # type: (str) -> None
"OnnxBackendNodeModelTest.test_qlinearmatmul_2D_int8_float32_cpu",
"OnnxBackendNodeModelTest.test_qlinearmatmul_2D_uint8_float16_cpu",
),
(
xfail_issue_171766,
"OnnxBackendNodeModelTest.test_attention_3d_attn_mask_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_causal_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_diff_heads_sizes_attn_mask_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_diff_heads_sizes_causal_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_diff_heads_sizes_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_diff_heads_sizes_scaled_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_diff_heads_sizes_softcap_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_diff_heads_with_past_and_present_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_gqa_attn_mask_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_gqa_causal_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_gqa_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_gqa_scaled_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_gqa_softcap_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_gqa_with_past_and_present_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_scaled_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_softcap_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_with_past_and_present_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_with_past_and_present_qk_matmul_bias_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_with_past_and_present_qk_matmul_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_with_past_and_present_qk_matmul_softcap_cpu",
"OnnxBackendNodeModelTest.test_attention_3d_with_past_and_present_qk_matmul_softmax_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_attn_mask_bool_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_attn_mask_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_causal_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_diff_heads_sizes_attn_mask_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_diff_heads_sizes_causal_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_diff_heads_sizes_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_diff_heads_sizes_scaled_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_diff_heads_sizes_softcap_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_diff_heads_with_past_and_present_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_gqa_attn_mask_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_gqa_causal_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_gqa_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_gqa_scaled_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_gqa_softcap_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_gqa_with_past_and_present_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_scaled_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_softcap_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_with_past_and_present_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_with_past_and_present_qk_matmul_bias_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_with_past_and_present_qk_matmul_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_with_qk_matmul_bias_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_with_qk_matmul_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_with_qk_matmul_softcap_cpu",
"OnnxBackendNodeModelTest.test_attention_4d_with_qk_matmul_softmax_cpu",
),
(
xfail_issue_171767,
"OnnxBackendNodeModelTest.test_cast_FLOAT16_to_FLOAT4E2M1_cpu",
Expand Down
Loading