Skip to content

Add int8/uint8 support for Max and Min operators #26386

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 4 commits into from
Oct 29, 2025
Merged

Add int8/uint8 support for Max and Min operators #26386

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
16fec6a
Add int8/uint8/int16/uint16 support for Max and Min operators
invalid-email-address Oct 22, 2025
e657cb1
Update doc
snnn Oct 22, 2025
188a0a4
Merge remote-tracking branch 'origin/main' into users/snnn/int8
Oct 27, 2025
74ec33b
Remove int16/uint16 support from Max/Min operators
Oct 27, 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
8 changes: 4 additions & 4 deletions docs/OperatorKernels.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -242,8 +242,8 @@ Do not modify directly.*
|||[9, 12]|**T** = tensor(double), tensor(float), tensor(int32), tensor(int64), tensor(uint32), tensor(uint64)|
|||[1, 8]|**T** = tensor(double), tensor(float)|
|MatMulInteger|*in* A:**T1**<br> *in* B:**T2**<br> *in* a_zero_point:**T1**<br> *in* b_zero_point:**T2**<br> *out* Y:**T3**|10+|**T1** = tensor(int8), tensor(uint8)<br/> **T2** = tensor(int8), tensor(uint8)<br/> **T3** = tensor(int32)|
|Max|*in* data_0:**T**<br> *out* max:**T**|13+|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint32), tensor(uint64)|
|||12|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint32), tensor(uint64)|
|Max|*in* data_0:**T**<br> *out* max:**T**|13+|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(int8), tensor(uint32), tensor(uint64), tensor(uint8)|
|||12|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(int8), tensor(uint32), tensor(uint64), tensor(uint8)|
|||[8, 11]|**T** = tensor(double), tensor(float)|
|||[6, 7]|**T** = tensor(float)|
|MaxPool|*in* X:**T**<br> *out* Y:**T**<br><br>or<br><br>*in* X:**T**<br> *out* Y:**T**<br> *out* Indices:**I**|22+|**I** = tensor(int64)<br/> **T** = tensor(double), tensor(float), tensor(int8), tensor(uint8)|
Expand All @@ -263,8 +263,8 @@ Do not modify directly.*
|MelWeightMatrix|*in* num_mel_bins:**T1**<br> *in* dft_length:**T1**<br> *in* sample_rate:**T1**<br> *in* lower_edge_hertz:**T2**<br> *in* upper_edge_hertz:**T2**<br> *out* output:**T3**|17+|**T1** = tensor(int32), tensor(int64)<br/> **T2** = tensor(float)<br/> **T3** = tensor(double), tensor(float), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
|MemcpyFromHost|*in* X:**T**<br> *out* Y:**T**|1+|**T** = seq(tensor(bfloat16)), seq(tensor(bool)), seq(tensor(double)), seq(tensor(float)), seq(tensor(float16)), seq(tensor(float8e4m3fn)), seq(tensor(float8e4m3fnuz)), seq(tensor(float8e5m2)), seq(tensor(float8e5m2fnuz)), seq(tensor(int16)), seq(tensor(int32)), seq(tensor(int64)), seq(tensor(int8)), seq(tensor(uint16)), seq(tensor(uint32)), seq(tensor(uint64)), seq(tensor(uint8)), tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(float8e4m3fn), tensor(float8e4m3fnuz), tensor(float8e5m2), tensor(float8e5m2fnuz), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
|MemcpyToHost|*in* X:**T**<br> *out* Y:**T**|1+|**T** = seq(tensor(bfloat16)), seq(tensor(bool)), seq(tensor(double)), seq(tensor(float)), seq(tensor(float16)), seq(tensor(float8e4m3fn)), seq(tensor(float8e4m3fnuz)), seq(tensor(float8e5m2)), seq(tensor(float8e5m2fnuz)), seq(tensor(int16)), seq(tensor(int32)), seq(tensor(int64)), seq(tensor(int8)), seq(tensor(uint16)), seq(tensor(uint32)), seq(tensor(uint64)), seq(tensor(uint8)), tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(float8e4m3fn), tensor(float8e4m3fnuz), tensor(float8e5m2), tensor(float8e5m2fnuz), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
|Min|*in* data_0:**T**<br> *out* min:**T**|13+|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint32), tensor(uint64)|
|||12|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint32), tensor(uint64)|
|Min|*in* data_0:**T**<br> *out* min:**T**|13+|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(int8), tensor(uint32), tensor(uint64), tensor(uint8)|
|||12|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(int8), tensor(uint32), tensor(uint64), tensor(uint8)|
|||[8, 11]|**T** = tensor(double), tensor(float)|
|||[6, 7]|**T** = tensor(float)|
|Mod|*in* A:**T**<br> *in* B:**T**<br> *out* C:**T**|13+|**T** = tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
Expand Down
12 changes: 8 additions & 4 deletions onnxruntime/core/providers/cpu/math/element_wise_ops.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -20,14 +20,16 @@ namespace op_kernel_type_control {
ORT_SPECIFY_OP_KERNEL_ARG_DEFAULT_TYPES(kCpuExecutionProvider, kOnnxDomain, Max, 8, Input, 0, float, double);

ORT_SPECIFY_OP_KERNEL_ARG_DEFAULT_TYPES(kCpuExecutionProvider, kOnnxDomain, Max, 12, Input, 0,
float, double, MLFloat16, int32_t, uint32_t, int64_t, uint64_t);
float, double, MLFloat16, int8_t, int32_t, uint32_t,
int64_t, uint8_t, uint64_t);
ORT_SPECIFY_OP_KERNEL_ARG_REQUIRED_TYPES(kCpuExecutionProvider, kOnnxDomain, Max, 12, Input, 0,
int32_t, int64_t);

// Min
ORT_SPECIFY_OP_KERNEL_ARG_DEFAULT_TYPES(kCpuExecutionProvider, kOnnxDomain, Min, 8, Input, 0, float, double);
ORT_SPECIFY_OP_KERNEL_ARG_DEFAULT_TYPES(kCpuExecutionProvider, kOnnxDomain, Min, 12, Input, 0,
float, double, MLFloat16, int32_t, uint32_t, int64_t, uint64_t);
float, double, MLFloat16, int8_t, int32_t, uint32_t,
int64_t, uint8_t, uint64_t);
ORT_SPECIFY_OP_KERNEL_ARG_REQUIRED_TYPES(kCpuExecutionProvider, kOnnxDomain, Min, 12, Input, 0,
int32_t, int64_t);

Expand Down Expand Up @@ -989,7 +991,8 @@ Status Min_8::Compute(OpKernelContext* context) const {
return MinMaxMLFloat16<true>(*this, context);
break;
default:
utils::MLTypeCallDispatcher<float, double, int32_t, uint32_t, int64_t, uint64_t>
utils::MLTypeCallDispatcher<float, double, int8_t, int32_t, uint32_t,
int64_t, uint8_t, uint64_t>
t_disp(dt_type);
return t_disp.InvokeRet<Status, ComputeImpl>(*this, context);
}
Expand Down Expand Up @@ -1055,7 +1058,8 @@ Status Max_8::Compute(OpKernelContext* context) const {
return MinMaxMLFloat16<false>(*this, context);
break;
default:
utils::MLTypeCallDispatcher<float, double, int32_t, uint32_t, int64_t, uint64_t>
utils::MLTypeCallDispatcher<float, double, int8_t, int32_t, uint32_t,
int64_t, uint8_t, uint64_t>
t_disp(dt_type);
return t_disp.InvokeRet<Status, ComputeImpl>(*this, context);
}
Expand Down
68 changes: 68 additions & 0 deletions onnxruntime/test/providers/cpu/math/element_wise_ops_test.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -3700,6 +3700,74 @@ TEST(MathOpTest, Equal_multidirectional_broadcastAB_bool) {
test.Run();
}

TEST(MathOpTest, Max_12_Int8) {
OpTester test("Max", 12);
test.AddInput<int8_t>("data_0", {1, 3},
{1, 2, 3});
test.AddInput<int8_t>("data_2", {3, 3},
{10, 20, 30,
40, 50, 60,
70, 80, 90});
test.AddInput<int8_t>("data_1", {3, 1},
{-1, -2, 127});
test.AddOutput<int8_t>("max", {3, 3},
{10, 20, 30,
40, 50, 60,
127, 127, 127});
test.Run(OpTester::ExpectResult::kExpectSuccess, "", {kTensorrtExecutionProvider, kOpenVINOExecutionProvider});
}

TEST(MathOpTest, Max_12_UInt8) {
OpTester test("Max", 12);
test.AddInput<uint8_t>("data_0", {1, 3},
{1, 20, 30});
test.AddInput<uint8_t>("data_2", {3, 3},
{10, 20, 30,
40, 50, 60,
70, 80, 90});
test.AddInput<uint8_t>("data_1", {3, 1},
{100, 20, 30});
test.AddOutput<uint8_t>("max", {3, 3},
{100, 100, 100,
40, 50, 60,
70, 80, 90});
test.Run(OpTester::ExpectResult::kExpectSuccess, "", {kTensorrtExecutionProvider});
}

TEST(MathOpTest, Min_12_Int8) {
OpTester test("Min", 12);
test.AddInput<int8_t>("data_0", {1, 3},
{1, 2, 3});
test.AddInput<int8_t>("data_2", {3, 3},
{10, 20, 30,
40, 50, 60,
-70, -80, -90});
test.AddInput<int8_t>("data_1", {3, 1},
{-1, 20, 127});
test.AddOutput<int8_t>("min", {3, 3},
{-1, -1, -1,
1, 2, 3,
-70, -80, -90});
test.Run(OpTester::ExpectResult::kExpectSuccess, "", {kTensorrtExecutionProvider, kOpenVINOExecutionProvider});
}

TEST(MathOpTest, Min_12_UInt8) {
OpTester test("Min", 12);
test.AddInput<uint8_t>("data_0", {1, 3},
{1, 20, 30});
test.AddInput<uint8_t>("data_2", {3, 3},
{10, 20, 30,
40, 50, 60,
70, 80, 90});
test.AddInput<uint8_t>("data_1", {3, 1},
{1, 20, 30});
test.AddOutput<uint8_t>("min", {3, 3},
{1, 1, 1,
1, 20, 20,
1, 20, 30});
test.Run(OpTester::ExpectResult::kExpectSuccess, "", {kTensorrtExecutionProvider});
}

TEST(MathOpTest, Mean_6) {
OpTester test("Mean", 6);
std::vector<int64_t> dims{3, 3};
Expand Down
Loading