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

Apply "lintrunner -a" to WindowsAI #18982

Merged
merged 1 commit into from
Jan 3, 2024
Merged

Apply "lintrunner -a" to WindowsAI #18982

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
59 changes: 29 additions & 30 deletions onnxruntime/test/contrib_ops/matmul_integer_to_float_test.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -103,7 +103,6 @@ void TestMatMulIntegerToFloat(const std::vector<int64_t>& A_dims,
} else {
test.Run(OpTester::ExpectResult::kExpectSuccess, "", {kCpuExecutionProvider});
}

}

template <typename IType, typename WType, typename OType, bool HasZeroPoint, bool HasBias>
Expand All @@ -113,43 +112,43 @@ void RunMatMulIntegerToFloatTest(const string& model_path) {
std::vector<int64_t> Y_dims{4, 128};

TestMatMulIntegerToFloat<IType, WType, OType>(
A_dims,
B_dims,
model_path,
false, /*is_matrix_b_constant*/
false, /*per_column*/
HasZeroPoint, /*has_zp*/
HasBias /*has_bias*/
A_dims,
B_dims,
model_path,
false, /*is_matrix_b_constant*/
false, /*per_column*/
HasZeroPoint, /*has_zp*/
HasBias /*has_bias*/
);

TestMatMulIntegerToFloat<IType, WType, OType>(
A_dims,
B_dims,
model_path,
true, /*is_matrix_b_constant*/
false, /*per_column*/
HasZeroPoint, /*has_zp*/
HasBias /*has_bias*/
A_dims,
B_dims,
model_path,
true, /*is_matrix_b_constant*/
false, /*per_column*/
HasZeroPoint, /*has_zp*/
HasBias /*has_bias*/
);

TestMatMulIntegerToFloat<IType, WType, OType>(
A_dims,
B_dims,
model_path,
false, /*is_matrix_b_constant*/
true, /*per_column*/
HasZeroPoint, /*has_zp*/
HasBias /*has_bias*/
A_dims,
B_dims,
model_path,
false, /*is_matrix_b_constant*/
true, /*per_column*/
HasZeroPoint, /*has_zp*/
HasBias /*has_bias*/
);

TestMatMulIntegerToFloat<IType, WType, OType>(
A_dims,
B_dims,
model_path,
true, /*is_matrix_b_constant*/
true, /*per_column*/
HasZeroPoint, /*has_zp*/
HasBias /*has_bias*/
A_dims,
B_dims,
model_path,
true, /*is_matrix_b_constant*/
true, /*per_column*/
HasZeroPoint, /*has_zp*/
HasBias /*has_bias*/
);
}

Expand All @@ -171,7 +170,7 @@ TEST(MatMulIntegerToFloat, HasZeroPoint_NoBias_test_S8S8_FP16) {
TEST(MatMulIntegerToFloat, NoZeroPoint_HasBias_test_S8S8_FP16) {
RunMatMulIntegerToFloatTest<int8_t, int8_t, MLFloat16, false, true>("testdata/matmul_integer_to_float16_int8_int8_bias.onnx");
}
#endif // USE_DML
#endif // USE_DML

TEST(MatMulIntegerToFloat, HasZeroPoint_NoBias_test_U8X8) {
RunMatMulIntegerToFloatTest<uint8_t, int8_t, float, true, false>("testdata/matmul_integer_to_float_int8.onnx");
Expand Down
6 changes: 3 additions & 3 deletions onnxruntime/test/contrib_ops/quantize_attention_op_test.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -114,7 +114,7 @@ void RunQAttention(const std::vector<float>& input_data,
execution_providers.push_back(DefaultCpuExecutionProvider());
} else if constexpr (ep == EP::DML) {
execution_providers.push_back(DefaultDmlExecutionProvider());
} else{ // onednn ep
} else { // onednn ep
execution_providers.push_back(DefaultDnnlExecutionProvider());
}

Expand Down Expand Up @@ -322,8 +322,8 @@ static void RunQAttentionAll(
batch_size, sequence_length, hidden_size, number_of_heads,
use_special_quantize_parameter, is_unidirectional, input_hidden_size);
RunQAttentionDML(input_data, weight_data, bias_data, mask_index_data, output_data,
batch_size, sequence_length, hidden_size, number_of_heads,
use_special_quantize_parameter, is_unidirectional, input_hidden_size);
batch_size, sequence_length, hidden_size, number_of_heads,
use_special_quantize_parameter, is_unidirectional, input_hidden_size);
}

// ONEDNN EP only supports 2D raw mask
Expand Down
4 changes: 2 additions & 2 deletions onnxruntime/test/optimizer/graph_transform_test.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5661,15 +5661,15 @@ TEST_F(GraphTransformationTests, MatMulIntegerToFloatTest) {
}

#ifdef USE_DML
TEST_F(GraphTransformationTests, MatMulIntegerToFloat16Test) {
TEST_F(GraphTransformationTests, MatMulIntegerToFloat16Test) {
constexpr const ORTCHAR_T* model_uri = MODEL_FOLDER "fusion/matmul_integer_to_float16_int8.onnx";
std::shared_ptr<Model> p_model;
ASSERT_STATUS_OK(Model::Load(model_uri, p_model, nullptr, *logger_));
Graph& graph = p_model->MainGraph();

for (auto& node : graph.Nodes()) {
node.SetExecutionProviderType(kDmlExecutionProvider);
}
}
onnxruntime::GraphTransformerManager graph_transformation_mgr{5};
ASSERT_STATUS_OK(graph_transformation_mgr.Register(std::make_unique<MatMulIntegerToFloatFusion>(), TransformerLevel::Level2));
ASSERT_STATUS_OK(graph_transformation_mgr.ApplyTransformers(graph, TransformerLevel::Level2, *logger_));
Expand Down
74 changes: 65 additions & 9 deletions onnxruntime/test/testdata/matmul_integer_to_float.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -13,7 +13,13 @@ def GenerateModel(model_name, sign_i, sign_w, output_type_fp16, has_zp=True, bia
"MatMulInteger",
),
helper.make_node("Mul", ["a_scale", "b_scale"], ["multiplier"], "mul_right"),
helper.make_node("Cast", ["matmul_output_int32"], ["matmul_output_float"], "cast", to=TensorProto.FLOAT16 if output_type_fp16 else TensorProto.FLOAT),
helper.make_node(
"Cast",
["matmul_output_int32"],
["matmul_output_float"],
"cast",
to=TensorProto.FLOAT16 if output_type_fp16 else TensorProto.FLOAT,
),
helper.make_node(
"Mul",
["matmul_output_float", "multiplier"],
Expand Down Expand Up @@ -48,14 +54,22 @@ def GenerateModel(model_name, sign_i, sign_w, output_type_fp16, has_zp=True, bia
if bias:
nodes.extend([helper.make_node("Add", ["mul_bottom_output", "bias"], ["Y"], "add")])

inputs.extend([helper.make_tensor_value_info("bias", TensorProto.FLOAT16 if output_type_fp16 else TensorProto.FLOAT, ["N"])])
inputs.extend(
[
helper.make_tensor_value_info(
"bias", TensorProto.FLOAT16 if output_type_fp16 else TensorProto.FLOAT, ["N"]
)
]
)

graph = helper.make_graph(
nodes,
"DynamicQuantizeMatMul_fusion", # name
inputs,
[ # outputs
helper.make_tensor_value_info("Y", TensorProto.FLOAT16 if output_type_fp16 else TensorProto.FLOAT, ["M", "N"]),
helper.make_tensor_value_info(
"Y", TensorProto.FLOAT16 if output_type_fp16 else TensorProto.FLOAT, ["M", "N"]
),
],
)

Expand All @@ -66,16 +80,58 @@ def GenerateModel(model_name, sign_i, sign_w, output_type_fp16, has_zp=True, bia
if __name__ == "__main__":
GenerateModel("matmul_integer_to_float16_int8.onnx", sign_i=False, sign_w=True, output_type_fp16=True)
GenerateModel("matmul_integer_to_float16_uint8.onnx", sign_i=False, sign_w=False, output_type_fp16=True)
GenerateModel("matmul_integer_to_float16_int8_bias.onnx", sign_i=False, sign_w=True, output_type_fp16=True, has_zp=False, bias=True)
GenerateModel("matmul_integer_to_float16_uint8_bias.onnx", sign_i=False, sign_w=False, output_type_fp16=True, has_zp=False, bias=True)
GenerateModel(
"matmul_integer_to_float16_int8_bias.onnx",
sign_i=False,
sign_w=True,
output_type_fp16=True,
has_zp=False,
bias=True,
)
GenerateModel(
"matmul_integer_to_float16_uint8_bias.onnx",
sign_i=False,
sign_w=False,
output_type_fp16=True,
has_zp=False,
bias=True,
)

GenerateModel("matmul_integer_to_float16_int8_int8.onnx", sign_i=True, sign_w=True, output_type_fp16=True)
GenerateModel("matmul_integer_to_float16_int8_int8_bias.onnx", sign_i=True, sign_w=True, output_type_fp16=True, has_zp=False, bias=True)
GenerateModel(
"matmul_integer_to_float16_int8_int8_bias.onnx",
sign_i=True,
sign_w=True,
output_type_fp16=True,
has_zp=False,
bias=True,
)

GenerateModel("matmul_integer_to_float_int8.onnx", sign_i=False, sign_w=True, output_type_fp16=False)
GenerateModel("matmul_integer_to_float_uint8.onnx", sign_i=False, sign_w=False, output_type_fp16=False)
GenerateModel("matmul_integer_to_float_int8_bias.onnx", sign_i=False, sign_w=True, output_type_fp16=False, has_zp=False, bias=True)
GenerateModel("matmul_integer_to_float_uint8_bias.onnx", sign_i=False, sign_w=False, output_type_fp16=False, has_zp=False, bias=True)
GenerateModel(
"matmul_integer_to_float_int8_bias.onnx",
sign_i=False,
sign_w=True,
output_type_fp16=False,
has_zp=False,
bias=True,
)
GenerateModel(
"matmul_integer_to_float_uint8_bias.onnx",
sign_i=False,
sign_w=False,
output_type_fp16=False,
has_zp=False,
bias=True,
)

GenerateModel("matmul_integer_to_float_int8_int8.onnx", sign_i=True, sign_w=True, output_type_fp16=False)
GenerateModel("matmul_integer_to_float_int8_int8_bias.onnx", sign_i=True, sign_w=True, output_type_fp16=False, has_zp=False, bias=True)
GenerateModel(
"matmul_integer_to_float_int8_int8_bias.onnx",
sign_i=True,
sign_w=True,
output_type_fp16=False,
has_zp=False,
bias=True,
)
2 changes: 1 addition & 1 deletion onnxruntime/test/testdata/transform/fusion/matmul_integer_to_float.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -104,4 +104,4 @@ def GenerateModel(model_name): # noqa: N802


if __name__ == "__main__":
GenerateModel("matmul_integer_to_float.onnx")
GenerateModel("matmul_integer_to_float.onnx")
Loading