diff --git a/docs/OperatorKernels.md b/docs/OperatorKernels.md
index 5a71864c7a..8bb10aa10f 100644
--- a/docs/OperatorKernels.md
+++ b/docs/OperatorKernels.md
@@ -122,7 +122,8 @@ Do not modify directly.*
|Greater|*in* A:**T**
*in* B:**T**
*out* C:**T1**|13+|**T** = tensor(double), tensor(float), tensor(int32), tensor(int64)
**T1** = tensor(bool)|
|||[9, 12]|**T** = tensor(double), tensor(float), tensor(int32), tensor(int64)
**T1** = tensor(bool)|
|||[7, 8]|**T** = tensor(double), tensor(float)
**T1** = tensor(bool)|
-|GreaterOrEqual|*in* A:**T**
*in* B:**T**
*out* C:**T1**|12+|**T** = tensor(double), tensor(float), tensor(int32), tensor(int64)
**T1** = tensor(bool)|
+|GreaterOrEqual|*in* A:**T**
*in* B:**T**
*out* C:**T1**|16+|**T** = tensor(double), tensor(float), tensor(int32), tensor(int64)
**T1** = tensor(bool)|
+|||[12, 15]|**T** = tensor(double), tensor(float), tensor(int32), tensor(int64)
**T1** = tensor(bool)|
|GridSample|*in* X:**T1**
*in* grid:**T1**
*out* Y:**T2**|16+|**T1** = tensor(float)
**T2** = tensor(float)|
|HardSigmoid|*in* X:**T**
*out* Y:**T**|6+|**T** = tensor(float)|
|Hardmax|*in* input:**T**
*out* output:**T**|13+|**T** = tensor(float)|
@@ -146,11 +147,13 @@ Do not modify directly.*
|LSTM|*in* X:**T**
*in* W:**T**
*in* R:**T**
*in* B:**T**
*in* sequence_lens:**T1**
*in* initial_h:**T**
*in* initial_c:**T**
*in* P:**T**
*out* Y:**T**
*out* Y_h:**T**
*out* Y_c:**T**|14+|**T** = tensor(double), tensor(float)
**T1** = tensor(int32)|
|||[7, 13]|**T** = tensor(double), tensor(float)
**T1** = tensor(int32)|
|LayerNormalization|*in* X:**T**
*in* Scale:**T**
*in* B:**T**
*out* Y:**T**
*out* Mean:**U**
*out* InvStdDev:**U**|1+|**T** = tensor(double), tensor(float)|
-|LeakyRelu|*in* X:**T**
*out* Y:**T**|6+|**T** = tensor(float)|
+|LeakyRelu|*in* X:**T**
*out* Y:**T**|16+|**T** = tensor(float)|
+|||[6, 15]|**T** = tensor(float)|
|Less|*in* A:**T**
*in* B:**T**
*out* C:**T1**|13+|**T** = tensor(double), tensor(float), tensor(int32), tensor(int64)
**T1** = tensor(bool)|
|||[9, 12]|**T** = tensor(double), tensor(float), tensor(int32), tensor(int64)
**T1** = tensor(bool)|
|||[7, 8]|**T** = tensor(double), tensor(float)
**T1** = tensor(bool)|
-|LessOrEqual|*in* A:**T**
*in* B:**T**
*out* C:**T1**|12+|**T** = tensor(double), tensor(float), tensor(int32), tensor(int64)
**T1** = tensor(bool)|
+|LessOrEqual|*in* A:**T**
*in* B:**T**
*out* C:**T1**|16+|**T** = tensor(double), tensor(float), tensor(int32), tensor(int64)
**T1** = tensor(bool)|
+|||[12, 15]|**T** = tensor(double), tensor(float), tensor(int32), tensor(int64)
**T1** = tensor(bool)|
|Log|*in* input:**T**
*out* output:**T**|13+|**T** = tensor(double), tensor(float)|
|||[6, 12]|**T** = tensor(double), tensor(float)|
|LogSoftmax|*in* input:**T**
*out* output:**T**|13+|**T** = tensor(double), tensor(float)|
@@ -204,7 +207,8 @@ Do not modify directly.*
|OptionalGetElement|*in* input:**O**
*out* output:**V**|15+|**O** = optional(seq(tensor(bfloat16))), optional(seq(tensor(bool))), optional(seq(tensor(double))), optional(seq(tensor(float))), optional(seq(tensor(float16))), optional(seq(tensor(int16))), optional(seq(tensor(int32))), optional(seq(tensor(int64))), optional(seq(tensor(int8))), optional(seq(tensor(string))), optional(seq(tensor(uint16))), optional(seq(tensor(uint32))), optional(seq(tensor(uint64))), optional(seq(tensor(uint8))), optional(tensor(bfloat16)), optional(tensor(bool)), optional(tensor(double)), optional(tensor(float)), optional(tensor(float16)), optional(tensor(int16)), optional(tensor(int32)), optional(tensor(int64)), optional(tensor(int8)), optional(tensor(string)), optional(tensor(uint16)), optional(tensor(uint32)), optional(tensor(uint64)), optional(tensor(uint8))
**V** = seq(tensor(bfloat16)), seq(tensor(bool)), seq(tensor(double)), seq(tensor(float)), seq(tensor(float16)), seq(tensor(int16)), seq(tensor(int32)), seq(tensor(int64)), seq(tensor(int8)), seq(tensor(string)), seq(tensor(uint16)), seq(tensor(uint32)), seq(tensor(uint64)), seq(tensor(uint8)), tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(string), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
|OptionalHasElement|*in* input:**O**
*out* output:**B**|15+|**B** = tensor(bool)
**O** = optional(seq(tensor(bfloat16))), optional(seq(tensor(bool))), optional(seq(tensor(double))), optional(seq(tensor(float))), optional(seq(tensor(float16))), optional(seq(tensor(int16))), optional(seq(tensor(int32))), optional(seq(tensor(int64))), optional(seq(tensor(int8))), optional(seq(tensor(string))), optional(seq(tensor(uint16))), optional(seq(tensor(uint32))), optional(seq(tensor(uint64))), optional(seq(tensor(uint8))), optional(tensor(bfloat16)), optional(tensor(bool)), optional(tensor(double)), optional(tensor(float)), optional(tensor(float16)), optional(tensor(int16)), optional(tensor(int32)), optional(tensor(int64)), optional(tensor(int8)), optional(tensor(string)), optional(tensor(uint16)), optional(tensor(uint32)), optional(tensor(uint64)), optional(tensor(uint8))|
|Or|*in* A:**T**
*in* B:**T**
*out* C:**T1**|7+|**T** = tensor(bool)
**T1** = tensor(bool)|
-|PRelu|*in* X:**T**
*in* slope:**T**
*out* Y:**T**|9+|**T** = tensor(float)|
+|PRelu|*in* X:**T**
*in* slope:**T**
*out* Y:**T**|16+|**T** = tensor(float)|
+|||[9, 15]|**T** = tensor(float)|
|||[7, 8]|**T** = tensor(float)|
|Pad|*in* data:**T**
*in* pads:**tensor(int64)**
*in* constant_value:**T**
*out* output:**T**
or
*in* data:**T**
*out* output:**T**|13+|**T** = tensor(bool), tensor(double), tensor(float), tensor(int32), tensor(int64), tensor(int8), tensor(uint32), tensor(uint64), tensor(uint8)|
|||[11, 12]|**T** = tensor(double), tensor(float), tensor(int32), tensor(int64), tensor(int8), tensor(uint32), tensor(uint64), tensor(uint8)|
@@ -275,7 +279,8 @@ Do not modify directly.*
|Round|*in* X:**T**
*out* Y:**T**|11+|**T** = tensor(double), tensor(float), tensor(float16)|
|Scale|*in* input:**T**
*out* output:**T**|1+|**T** = tensor(float)|
|ScaledTanh|*in* input:**T**
*out* output:**T**|1+|**T** = tensor(float)|
-|Scan|*in* initial_state_and_scan_inputs:**V**
*out* final_state_and_scan_outputs:**V**
or
*in* sequence_lens:**I**
*in* initial_state_and_scan_inputs:**V**
*out* final_state_and_scan_outputs:**V**|11+|**V** = tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(string), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
+|Scan|*in* initial_state_and_scan_inputs:**V**
*out* final_state_and_scan_outputs:**V**
or
*in* sequence_lens:**I**
*in* initial_state_and_scan_inputs:**V**
*out* final_state_and_scan_outputs:**V**|16+|**V** = tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(string), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
+|||[11, 15]|**V** = tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(string), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
|||[9, 10]|**V** = tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(string), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
|||8|**I** = tensor(int64)
**V** = tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(string), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
|Scatter|*in* data:**T**
*in* indices:**Tind**
*in* updates:**T**
*out* output:**T**|[9, 10]|**T** = tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(string), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)
**Tind** = tensor(int32), tensor(int64)|
@@ -539,7 +544,8 @@ Do not modify directly.*
|Greater|*in* A:**T**
*in* B:**T**
*out* C:**T1**|13+|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint32), tensor(uint64)
**T1** = tensor(bool)|
|||[9, 12]|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint32), tensor(uint64)|
|||[7, 8]|**T** = tensor(double), tensor(float), tensor(float16)|
-|GreaterOrEqual|*in* A:**T**
*in* B:**T**
*out* C:**T1**|12+|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint32), tensor(uint64)
**T1** = tensor(bool)|
+|GreaterOrEqual|*in* A:**T**
*in* B:**T**
*out* C:**T1**|16+|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint32), tensor(uint64)
**T1** = tensor(bool)|
+|||[12, 15]|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint32), tensor(uint64)
**T1** = tensor(bool)|
|HardSigmoid|*in* X:**T**
*out* Y:**T**|6+|**T** = tensor(double), tensor(float), tensor(float16)|
|Identity|*in* input:**T**
*out* output:**T**
or
*in* input:**V**
*out* output:**V**|14+|**V** = seq(tensor(bfloat16)), seq(tensor(bool)), seq(tensor(double)), seq(tensor(float)), seq(tensor(float16)), 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(int16), tensor(int32), tensor(int64), tensor(int8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
|||13|**T** = tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
@@ -554,11 +560,13 @@ Do not modify directly.*
|LSTM|*in* X:**T**
*in* W:**T**
*in* R:**T**
*in* B:**T**
*in* sequence_lens:**T1**
*in* initial_h:**T**
*in* initial_c:**T**
*in* P:**T**
*out* Y:**T**
*out* Y_h:**T**
*out* Y_c:**T**|14+|**T** = tensor(double), tensor(float), tensor(float16)
**T1** = tensor(int32)|
|||[7, 13]|**T** = tensor(double), tensor(float), tensor(float16)
**T1** = tensor(int32)|
|LayerNormalization|*in* X:**T**
*in* Scale:**T**
*in* B:**T**
*out* Y:**T**
*out* Mean:**U**
*out* InvStdDev:**U**|1+|**T** = tensor(bfloat16), tensor(double), tensor(float), tensor(float16)
**U** = tensor(double), tensor(float)|
-|LeakyRelu|*in* X:**T**
*out* Y:**T**|6+|**T** = tensor(double), tensor(float), tensor(float16)|
+|LeakyRelu|*in* X:**T**
*out* Y:**T**|16+|**T** = tensor(double), tensor(float), tensor(float16)|
+|||[6, 15]|**T** = tensor(double), tensor(float), tensor(float16)|
|Less|*in* A:**T**
*in* B:**T**
*out* C:**T1**|13+|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint32), tensor(uint64)
**T1** = tensor(bool)|
|||[9, 12]|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint32), tensor(uint64)|
|||[7, 8]|**T** = tensor(double), tensor(float), tensor(float16)|
-|LessOrEqual|*in* A:**T**
*in* B:**T**
*out* C:**T1**|12+|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint32), tensor(uint64)
**T1** = tensor(bool)|
+|LessOrEqual|*in* A:**T**
*in* B:**T**
*out* C:**T1**|16+|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint32), tensor(uint64)
**T1** = tensor(bool)|
+|||[12, 15]|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint32), tensor(uint64)
**T1** = tensor(bool)|
|Log|*in* input:**T**
*out* output:**T**|13+|**T** = tensor(double), tensor(float), tensor(float16)|
|||[6, 12]|**T** = tensor(double), tensor(float), tensor(float16)|
|LogSoftmax|*in* input:**T**
*out* output:**T**|13+|**T** = tensor(double), tensor(float), tensor(float16)|
@@ -594,7 +602,8 @@ Do not modify directly.*
|Not|*in* X:**T**
*out* Y:**T**|1+|**T** = tensor(bool)|
|OneHot|*in* indices:**T1**
*in* depth:**T2**
*in* values:**T3**
*out* output:**T3**|11+|**T1** = tensor(int32), tensor(int64)
**T2** = tensor(int32), tensor(int64)
**T3** = tensor(float), tensor(float16), tensor(int64)|
|Or|*in* A:**T**
*in* B:**T**
*out* C:**T1**|7+|**T** = tensor(bool)
**T1** = tensor(bool)|
-|PRelu|*in* X:**T**
*in* slope:**T**
*out* Y:**T**|9+|**T** = tensor(double), tensor(float), tensor(float16)|
+|PRelu|*in* X:**T**
*in* slope:**T**
*out* Y:**T**|16+|**T** = tensor(double), tensor(float), tensor(float16)|
+|||[9, 15]|**T** = tensor(double), tensor(float), tensor(float16)|
|||[7, 8]|**T** = tensor(double), tensor(float), tensor(float16)|
|Pad|*in* data:**T**
*in* pads:**tensor(int64)**
*in* constant_value:**T**
*out* output:**T**
or
*in* data:**T**
*out* output:**T**|13+|**T** = tensor(bool), tensor(double), tensor(float), tensor(float16)|
|||[11, 12]|**T** = tensor(double), tensor(float), tensor(float16)|
@@ -661,7 +670,8 @@ Do not modify directly.*
|RoiAlign|*in* X:**T1**
*in* rois:**T1**
*in* batch_indices:**T2**
*out* Y:**T1**|10+|**T1** = tensor(double), tensor(float)
**T2** = tensor(int64)|
|Round|*in* X:**T**
*out* Y:**T**|11+|**T** = tensor(double), tensor(float), tensor(float16)|
|ScaledTanh|*in* input:**T**
*out* output:**T**|1+|**T** = tensor(double), tensor(float), tensor(float16)|
-|Scan|*in* initial_state_and_scan_inputs:**V**
*out* final_state_and_scan_outputs:**V**
or
*in* sequence_lens:**I**
*in* initial_state_and_scan_inputs:**V**
*out* final_state_and_scan_outputs:**V**|11+|**V** = tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
+|Scan|*in* initial_state_and_scan_inputs:**V**
*out* final_state_and_scan_outputs:**V**
or
*in* sequence_lens:**I**
*in* initial_state_and_scan_inputs:**V**
*out* final_state_and_scan_outputs:**V**|16+|**V** = tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
+|||[11, 15]|**V** = tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
|||[9, 10]|**V** = tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
|||8|**I** = tensor(int64)
**V** = tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(string), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
|Scatter|*in* data:**T**
*in* indices:**Tind**
*in* updates:**T**
*out* output:**T**|[9, 10]|**T** = tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)
**Tind** = tensor(int32), tensor(int64)|
@@ -728,7 +738,8 @@ Do not modify directly.*
|||[1, 10]|**T** = tensor(bfloat16), tensor(bool), tensor(double), tensor(float), tensor(float16), tensor(int16), tensor(int32), tensor(int64), tensor(int8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(uint8)|
|Upsample|*in* X:**T**
*in* scales:**tensor(float)**
*out* Y:**T**
or
*in* X:**T**
*out* Y:**T**|9|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(uint8)|
|||[7, 8]|**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(uint8)|
-|Where|*in* condition:**B**
*in* X:**T**
*in* Y:**T**
*out* output:**T**|9+|**B** = tensor(bool)
**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint8)|
+|Where|*in* condition:**B**
*in* X:**T**
*in* Y:**T**
*out* output:**T**|16+|**B** = tensor(bool)
**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint8)|
+|||[9, 15]|**B** = tensor(bool)
**T** = tensor(double), tensor(float), tensor(float16), tensor(int32), tensor(int64), tensor(uint8)|
|Xor|*in* A:**T**
*in* B:**T**
*out* C:**T1**|7+|**T** = tensor(bool)
**T1** = tensor(bool)|
| |
| |
diff --git a/onnxruntime/core/providers/cpu/activation/activations.cc b/onnxruntime/core/providers/cpu/activation/activations.cc
index 5409f63e53..e16bec0880 100644
--- a/onnxruntime/core/providers/cpu/activation/activations.cc
+++ b/onnxruntime/core/providers/cpu/activation/activations.cc
@@ -34,7 +34,7 @@ namespace onnxruntime {
REGISTER_UNARY_ELEMENTWISE_KERNEL(Elu, 6);
REGISTER_UNARY_ELEMENTWISE_KERNEL(HardSigmoid, 6);
-REGISTER_UNARY_ELEMENTWISE_KERNEL(LeakyRelu, 6);
+REGISTER_VERSIONED_UNARY_ELEMENTWISE_KERNEL(LeakyRelu, 6, 15);
REGISTER_VERSIONED_UNARY_ELEMENTWISE_TYPED_KERNEL(Relu, 6, 12, float);
REGISTER_VERSIONED_UNARY_ELEMENTWISE_TYPED_KERNEL(Relu, 6, 12, double);
REGISTER_VERSIONED_UNARY_ELEMENTWISE_TYPED_KERNEL(Relu, 13, 13, float);
@@ -57,6 +57,9 @@ REGISTER_UNARY_ELEMENTWISE_TYPED_KERNEL(Tanh, 13, float);
REGISTER_UNARY_ELEMENTWISE_TYPED_KERNEL(Tanh, 13, double);
REGISTER_UNARY_ELEMENTWISE_KERNEL(ThresholdedRelu, 10);
+// Opset-16 adds BFloat16 to allowed types for the LeakyRelu operator
+REGISTER_UNARY_ELEMENTWISE_KERNEL(LeakyRelu, 16);
+
namespace functors {
template
Status ElementWiseRangedTransform::Create(const std::string& type, const NodeAttributes& attributes,
diff --git a/onnxruntime/core/providers/cpu/controlflow/scan_9.cc b/onnxruntime/core/providers/cpu/controlflow/scan_9.cc
index bc7d7009cf..7ff4321058 100644
--- a/onnxruntime/core/providers/cpu/controlflow/scan_9.cc
+++ b/onnxruntime/core/providers/cpu/controlflow/scan_9.cc
@@ -503,8 +503,18 @@ ONNX_CPU_OPERATOR_VERSIONED_KERNEL(Scan,
Scan<9>);
// Opset 11 starts to support Neg Axis.
+ONNX_CPU_OPERATOR_VERSIONED_KERNEL(Scan,
+ 11,
+ 15,
+ KernelDefBuilder()
+ // 'I' is in the ONNX spec but is not actually used for any inputs or outputs
+ //.TypeConstraint("I", DataTypeImpl::GetTensorType())
+ .TypeConstraint("V", DataTypeImpl::AllTensorTypes()),
+ Scan<9>);
+
+// Opset 16 starts to support BFloat16 type for the type constraint "V"
ONNX_CPU_OPERATOR_KERNEL(Scan,
- 11,
+ 16,
KernelDefBuilder()
// 'I' is in the ONNX spec but is not actually used for any inputs or outputs
//.TypeConstraint("I", DataTypeImpl::GetTensorType())
diff --git a/onnxruntime/core/providers/cpu/cpu_execution_provider.cc b/onnxruntime/core/providers/cpu/cpu_execution_provider.cc
index 7cb43c01c7..029b481533 100644
--- a/onnxruntime/core/providers/cpu/cpu_execution_provider.cc
+++ b/onnxruntime/core/providers/cpu/cpu_execution_provider.cc
@@ -29,7 +29,7 @@ namespace onnxruntime {
class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 10, Clip);
class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, Elu);
class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, HardSigmoid);
-class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, LeakyRelu);
+class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 15, LeakyRelu);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12, float, Relu);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, 12, double, Relu);
class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 6, Selu);
@@ -288,7 +288,7 @@ class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOn
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 12, int64_t, MatMul);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 13, float, BatchNormalization);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 13, double, BatchNormalization);
-class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, PRelu);
+class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 15, PRelu);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 9, float, Upsample);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 9, int32_t, Upsample);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 9, 9, int8_t, Upsample);
@@ -385,7 +385,7 @@ class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOn
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12, double, Softmax);
class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12, Loop);
class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12, DepthToSpace);
-class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, Scan);
+class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 15, Scan);
class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12, Flatten);
class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, Compress);
class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 11, 12, Concat);
@@ -476,14 +476,14 @@ class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOn
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 12, double_double, Dropout);
class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, Celu);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, float, GreaterOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, double, GreaterOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, int32_t, GreaterOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, int64_t, GreaterOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, float, LessOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, double, LessOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, int32_t, LessOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, int64_t, LessOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15, float, GreaterOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15, double, GreaterOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15, int32_t, GreaterOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15, int64_t, GreaterOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15, float, LessOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15, double, LessOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15, int32_t, LessOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 12, 15, int64_t, LessOrEqual);
// opset 13
class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 13, float, Erf);
@@ -730,7 +730,17 @@ class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain,
class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, int32_t, Where);
class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, int64_t, Where);
class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, uint8_t, Where);
-
+class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, LeakyRelu);
+class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, PRelu);
+class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, Scan);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, float, GreaterOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, double, GreaterOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, int32_t, GreaterOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, int64_t, GreaterOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, float, LessOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, double, LessOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, int32_t, LessOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCpuExecutionProvider, kOnnxDomain, 16, int64_t, LessOrEqual);
// !!PLEASE READ BELOW!! Following that, add new entries above this comment
@@ -780,7 +790,7 @@ Status RegisterOnnxOperatorKernels(KernelRegistry& kernel_registry) {
Clip)>,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
- BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
@@ -1170,17 +1180,17 @@ Status RegisterOnnxOperatorKernels(KernelRegistry& kernel_registry) {
BuildKernelCreateInfo,
BuildKernelCreateInfo,
+ Where)>,
BuildKernelCreateInfo,
+ Where)>,
BuildKernelCreateInfo,
+ Where)>,
BuildKernelCreateInfo,
+ Where)>,
BuildKernelCreateInfo,
+ Where)>,
BuildKernelCreateInfo,
+ Where)>,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
- BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
- BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
@@ -1535,22 +1545,22 @@ Status RegisterOnnxOperatorKernels(KernelRegistry& kernel_registry) {
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
// opset 13
BuildKernelCreateInfo,
@@ -1915,6 +1925,25 @@ Status RegisterOnnxOperatorKernels(KernelRegistry& kernel_registry) {
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
};
for (auto& function_table_entry : function_table) {
diff --git a/onnxruntime/core/providers/cpu/math/element_wise_ops.cc b/onnxruntime/core/providers/cpu/math/element_wise_ops.cc
index de9876f9cf..612c66a06d 100644
--- a/onnxruntime/core/providers/cpu/math/element_wise_ops.cc
+++ b/onnxruntime/core/providers/cpu/math/element_wise_ops.cc
@@ -352,15 +352,27 @@ REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(Equal, 13, int64_t, Equal);
REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(Equal, 13, float, Equal);
REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(Equal, 13, double, Equal);
-REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(LessOrEqual, 12, float, LessOrEqual);
-REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(LessOrEqual, 12, double, LessOrEqual);
-REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(LessOrEqual, 12, int32_t, LessOrEqual);
-REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(LessOrEqual, 12, int64_t, LessOrEqual);
+REG_ELEMENTWISE_LOGICALOP_VERSIONED_TYPED_KERNEL(LessOrEqual, 12, 15, float, LessOrEqual);
+REG_ELEMENTWISE_LOGICALOP_VERSIONED_TYPED_KERNEL(LessOrEqual, 12, 15, double, LessOrEqual);
+REG_ELEMENTWISE_LOGICALOP_VERSIONED_TYPED_KERNEL(LessOrEqual, 12, 15, int32_t, LessOrEqual);
+REG_ELEMENTWISE_LOGICALOP_VERSIONED_TYPED_KERNEL(LessOrEqual, 12, 15, int64_t, LessOrEqual);
-REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(GreaterOrEqual, 12, float, GreaterOrEqual);
-REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(GreaterOrEqual, 12, double, GreaterOrEqual);
-REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(GreaterOrEqual, 12, int32_t, GreaterOrEqual);
-REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(GreaterOrEqual, 12, int64_t, GreaterOrEqual);
+// Opset-16 adds BFloat16 to allowed types for the LessOrEqual operator
+REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(LessOrEqual, 16, float, LessOrEqual);
+REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(LessOrEqual, 16, double, LessOrEqual);
+REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(LessOrEqual, 16, int32_t, LessOrEqual);
+REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(LessOrEqual, 16, int64_t, LessOrEqual);
+
+REG_ELEMENTWISE_LOGICALOP_VERSIONED_TYPED_KERNEL(GreaterOrEqual, 12, 15, float, GreaterOrEqual);
+REG_ELEMENTWISE_LOGICALOP_VERSIONED_TYPED_KERNEL(GreaterOrEqual, 12, 15, double, GreaterOrEqual);
+REG_ELEMENTWISE_LOGICALOP_VERSIONED_TYPED_KERNEL(GreaterOrEqual, 12, 15, int32_t, GreaterOrEqual);
+REG_ELEMENTWISE_LOGICALOP_VERSIONED_TYPED_KERNEL(GreaterOrEqual, 12, 15, int64_t, GreaterOrEqual);
+
+// Opset-16 adds BFloat16 to allowed types for the GreaterOrEqual operator
+REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(GreaterOrEqual, 16, float, GreaterOrEqual);
+REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(GreaterOrEqual, 16, double, GreaterOrEqual);
+REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(GreaterOrEqual, 16, int32_t, GreaterOrEqual);
+REG_ELEMENTWISE_LOGICALOP_TYPED_KERNEL(GreaterOrEqual, 16, int64_t, GreaterOrEqual);
REG_ELEMENTWISE_VERSIONED_TYPED_KERNEL(Mean, 6, 7, float, Mean_6);
REG_ELEMENTWISE_VERSIONED_TYPED_KERNEL(Mean, 8, 12, float, Mean_8);
@@ -1448,9 +1460,17 @@ ONNX_CPU_OPERATOR_VERSIONED_KERNEL(
KernelDefBuilder().TypeConstraint("T", DataTypeImpl::GetTensorType()),
PRelu);
-ONNX_CPU_OPERATOR_KERNEL(
+ONNX_CPU_OPERATOR_VERSIONED_KERNEL(
PRelu,
9,
+ 15,
+ KernelDefBuilder().TypeConstraint("T", DataTypeImpl::GetTensorType()),
+ PRelu);
+
+// Opset-16 adds BFloat16 to allowed types for the PRelu operator
+ONNX_CPU_OPERATOR_KERNEL(
+ PRelu,
+ 16,
KernelDefBuilder().TypeConstraint("T", DataTypeImpl::GetTensorType()),
PRelu);
diff --git a/onnxruntime/core/providers/cuda/activation/activations.cc b/onnxruntime/core/providers/cuda/activation/activations.cc
index df02e333b3..8ddce55eae 100644
--- a/onnxruntime/core/providers/cuda/activation/activations.cc
+++ b/onnxruntime/core/providers/cuda/activation/activations.cc
@@ -72,7 +72,7 @@ namespace cuda {
UNARY_ACTIVATION_OP_HFD(Elu, 6);
UNARY_ACTIVATION_OP_HFD(HardSigmoid, 6);
-UNARY_ACTIVATION_OP_HFD(LeakyRelu, 6);
+UNARY_ACTIVATION_OP_VERSIONED_HFD(LeakyRelu, 6, 15);
UNARY_ACTIVATION_OP_HFD(Relu, 14);
UNARY_ACTIVATION_OP_VERSIONED_HFD_WITH_BF16(Relu, 13, 13);
UNARY_ACTIVATION_OP_VERSIONED_HFD(Relu, 6, 12);
@@ -85,5 +85,8 @@ UNARY_ACTIVATION_OP_HFD(Tanh, 13);
UNARY_ACTIVATION_OP_VERSIONED_HFD(Tanh, 6, 12);
UNARY_ACTIVATION_OP_HFD(ThresholdedRelu, 10);
+// Opset-16 adds BFloat16 to allowed types for the LeakyRelu operator
+UNARY_ACTIVATION_OP_HFD(LeakyRelu, 16);
+
} // namespace cuda
} // namespace onnxruntime
diff --git a/onnxruntime/core/providers/cuda/controlflow/scan.cc b/onnxruntime/core/providers/cuda/controlflow/scan.cc
index 37a88e9765..33300d0fbd 100644
--- a/onnxruntime/core/providers/cuda/controlflow/scan.cc
+++ b/onnxruntime/core/providers/cuda/controlflow/scan.cc
@@ -88,9 +88,21 @@ ONNX_OPERATOR_VERSIONED_KERNEL_EX(Scan,
Scan<9>);
// Opset 11 starts to support Neg Axis.
+ONNX_OPERATOR_VERSIONED_KERNEL_EX(Scan,
+ kOnnxDomain,
+ 11,
+ 15,
+ kCudaExecutionProvider,
+ (*KernelDefBuilder::Create())
+ // 'I' is in the ONNX spec but is not used for any inputs or outputs
+ // .TypeConstraint("I", DataTypeImpl::GetTensorType())
+ .TypeConstraint("V", DataTypeImpl::AllFixedSizeTensorTypes()),
+ Scan<9>);
+
+// Opset 16 starts to support BFloat16 type for the type constraint "V"
ONNX_OPERATOR_KERNEL_EX(Scan,
kOnnxDomain,
- 11,
+ 16,
kCudaExecutionProvider,
(*KernelDefBuilder::Create())
// 'I' is in the ONNX spec but is not used for any inputs or outputs
diff --git a/onnxruntime/core/providers/cuda/cuda_execution_provider.cc b/onnxruntime/core/providers/cuda/cuda_execution_provider.cc
index 2f244efddf..099e6c633b 100755
--- a/onnxruntime/core/providers/cuda/cuda_execution_provider.cc
+++ b/onnxruntime/core/providers/cuda/cuda_execution_provider.cc
@@ -464,9 +464,9 @@ class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain,
class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, float, HardSigmoid);
class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, double, HardSigmoid);
class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, MLFloat16, HardSigmoid);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, float, LeakyRelu);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, double, LeakyRelu);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, MLFloat16, LeakyRelu);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 15, float, LeakyRelu);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 15, double, LeakyRelu);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 15, MLFloat16, LeakyRelu);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, float, Relu);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, double, Relu);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Relu);
@@ -497,9 +497,9 @@ class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kO
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, float, PRelu);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, double, PRelu);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 8, MLFloat16, PRelu);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, float, PRelu);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, double, PRelu);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, MLFloat16, PRelu);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, float, PRelu);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, double, PRelu);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, MLFloat16, PRelu);
class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, bool, And);
class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, bool, Or);
class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, bool, Xor);
@@ -523,20 +523,20 @@ class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kO
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, float, Greater);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, double, Greater);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, MLFloat16, Greater);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, int32_t, GreaterOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, int64_t, GreaterOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, uint32_t, GreaterOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, uint64_t, GreaterOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, float, GreaterOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, double, GreaterOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, MLFloat16, GreaterOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, int32_t, LessOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, int64_t, LessOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, uint32_t, LessOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, uint64_t, LessOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, float, LessOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, double, LessOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, MLFloat16, LessOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, int32_t, GreaterOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, int64_t, GreaterOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, uint32_t, GreaterOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, uint64_t, GreaterOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, float, GreaterOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, double, GreaterOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, MLFloat16, GreaterOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, int32_t, LessOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, int64_t, LessOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, uint32_t, LessOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, uint64_t, LessOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, float, LessOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, double, LessOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 12, 15, MLFloat16, LessOrEqual);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, int32_t, Add);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, int64_t, Add);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 7, 12, uint32_t, Add);
@@ -765,12 +765,12 @@ class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kO
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, MLFloat16, Less);
class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, EyeLike);
class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 10, Scatter);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, MLFloat16, Where);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, float, Where);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, double_t, Where);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, int32_t, Where);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, int64_t, Where);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, uint8_t, Where);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, MLFloat16, Where);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, float, Where);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, double_t, Where);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, int32_t, Where);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, int64_t, Where);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 15, uint8_t, Where);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, bool, NonZero);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, uint8_t, NonZero);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 9, 12, int32_t, NonZero);
@@ -871,7 +871,7 @@ class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kO
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, float, ReduceSumSquare);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, double, ReduceSumSquare);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, MLFloat16, ReduceSumSquare);
-class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, Scan);
+class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 15, Scan);
class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, ScatterElements);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, int32_t, Slice);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 11, 12, int64_t, Slice);
@@ -1263,6 +1263,35 @@ class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain,
class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 15, MLFloat16, BatchNormalization);
class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 15, Shape);
+// Opset 16
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, float, LeakyRelu);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, double, LeakyRelu);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, MLFloat16, LeakyRelu);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, float, PRelu);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, double, PRelu);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, MLFloat16, PRelu);
+class ONNX_OPERATOR_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, Scan);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, MLFloat16, Where);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, float, Where);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, double_t, Where);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, int32_t, Where);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, int64_t, Where);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, uint8_t, Where);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, int32_t, GreaterOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, int64_t, GreaterOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, uint32_t, GreaterOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, uint64_t, GreaterOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, float, GreaterOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, double, GreaterOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, MLFloat16, GreaterOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, int32_t, LessOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, int64_t, LessOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, uint32_t, LessOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, uint64_t, LessOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, float, LessOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, double, LessOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kCudaExecutionProvider, kOnnxDomain, 16, MLFloat16, LessOrEqual);
+
template <>
KernelCreateInfo BuildKernelCreateInfo() {
return {};
@@ -1306,9 +1335,9 @@ static Status RegisterCudaKernels(KernelRegistry& kernel_registry) {
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
@@ -1339,9 +1368,9 @@ static Status RegisterCudaKernels(KernelRegistry& kernel_registry) {
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
@@ -1365,20 +1394,20 @@ static Status RegisterCudaKernels(KernelRegistry& kernel_registry) {
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
@@ -1606,12 +1635,12 @@ static Status RegisterCudaKernels(KernelRegistry& kernel_registry) {
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
@@ -1716,7 +1745,7 @@ static Status RegisterCudaKernels(KernelRegistry& kernel_registry) {
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
- BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
@@ -2022,7 +2051,7 @@ static Status RegisterCudaKernels(KernelRegistry& kernel_registry) {
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
-
+
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
@@ -2104,6 +2133,35 @@ static Status RegisterCudaKernels(KernelRegistry& kernel_registry) {
BuildKernelCreateInfo,
BuildKernelCreateInfo,
+ // Opset 16
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+
};
for (auto& function_table_entry : function_table) {
diff --git a/onnxruntime/core/providers/cuda/math/binary_elementwise_ops.cc b/onnxruntime/core/providers/cuda/math/binary_elementwise_ops.cc
index e98e639d67..888f540c88 100644
--- a/onnxruntime/core/providers/cuda/math/binary_elementwise_ops.cc
+++ b/onnxruntime/core/providers/cuda/math/binary_elementwise_ops.cc
@@ -117,6 +117,17 @@ Status BinaryElementwise::Prepare(OpKernelContext* context, Bin
(*KernelDefBuilder::Create()).TypeConstraint("T", DataTypeImpl::GetTensorType()).TypeConstraint("T1", DataTypeImpl::GetTensorType()), \
x);
+#define BINARY_ELEMENTWISE_LOGICALOP_REGISTER_KERNEL_VERSIONED_TYPED(x, startver, endver, T) \
+ ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_EX( \
+ x, \
+ kOnnxDomain, \
+ startver, \
+ endver, \
+ T, \
+ kCudaExecutionProvider, \
+ (*KernelDefBuilder::Create()).TypeConstraint("T", DataTypeImpl::GetTensorType()).TypeConstraint("T1", DataTypeImpl::GetTensorType()), \
+ x);
+
#define BINARY_ELEMENTWISE_REGISTER_KERNEL_VERSIONED_TYPED(x, startver, endver, T) \
ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_EX( \
x, \
@@ -255,6 +266,16 @@ Status BinaryElementwise::Prepare(OpKernelContext* context, Bin
BINARY_ELEMENTWISE_LOGICALOP_REGISTER_KERNEL_TYPED(name, ver, double) \
BINARY_ELEMENTWISE_LOGICALOP_REGISTER_KERNEL_TYPED(name, ver, BFloat16)
+#define BINARY_LOGICALOP_REGISTER_VERSIONED_UZILHFD(name, startver, endver) \
+ BINARY_ELEMENTWISE_LOGICALOP_REGISTER_KERNEL_VERSIONED_TYPED(name, startver, endver, uint32_t) \
+ BINARY_ELEMENTWISE_LOGICALOP_REGISTER_KERNEL_VERSIONED_TYPED(name, startver, endver, uint64_t) \
+ BINARY_ELEMENTWISE_LOGICALOP_REGISTER_KERNEL_VERSIONED_TYPED(name, startver, endver, int32_t) \
+ BINARY_ELEMENTWISE_LOGICALOP_REGISTER_KERNEL_VERSIONED_TYPED(name, startver, endver, int64_t) \
+ BINARY_ELEMENTWISE_LOGICALOP_REGISTER_KERNEL_VERSIONED_TYPED(name, startver, endver, MLFloat16) \
+ BINARY_ELEMENTWISE_LOGICALOP_REGISTER_KERNEL_VERSIONED_TYPED(name, startver, endver, float) \
+ BINARY_ELEMENTWISE_LOGICALOP_REGISTER_KERNEL_VERSIONED_TYPED(name, startver, endver, double) \
+ BINARY_ELEMENTWISE_LOGICALOP_REGISTER_KERNEL_VERSIONED_TYPED(name, startver, endver, BFloat16)
+
#define BINARY_OP_REGISTER_VERSIONED_HFD(name, startver, endver) \
BINARY_ELEMENTWISE_REGISTER_KERNEL_VERSIONED_TYPED(name, startver, endver, MLFloat16) \
BINARY_ELEMENTWISE_REGISTER_KERNEL_VERSIONED_TYPED(name, startver, endver, float) \
@@ -294,7 +315,9 @@ BINARY_LOGICALOP_TYPED(And, 7, bool)
BINARY_LOGICALOP_TYPED(Or, 7, bool)
BINARY_LOGICALOP_TYPED(Xor, 7, bool)
BINARY_OP_VERSIONED_HFD(PRelu, 7, 8)
-BINARY_OP_HFD(PRelu, 9)
+BINARY_OP_VERSIONED_HFD(PRelu, 9, 15)
+// Opset-16 adds BFloat16 to allowed types for the PRelu operator
+BINARY_OP_HFD(PRelu, 16)
// Pow since version 12
ONNX_OPERATOR_VERSIONED_KERNEL_EX(
@@ -509,8 +532,14 @@ BINARY_OP_REGISTER_VERSIONED_HFD(Greater, 7, 8)
BINARY_LOGICALOP_REGISTER_UZILHFD(Less, 13)
BINARY_OP_REGISTER_VERSIONED_UZILHFD(Less, 9, 12)
BINARY_OP_REGISTER_VERSIONED_HFD(Less, 7, 8)
-BINARY_LOGICALOP_REGISTER_UZILHFD(GreaterOrEqual, 12)
-BINARY_LOGICALOP_REGISTER_UZILHFD(LessOrEqual, 12)
+BINARY_LOGICALOP_REGISTER_VERSIONED_UZILHFD(GreaterOrEqual, 12, 15)
+BINARY_LOGICALOP_REGISTER_VERSIONED_UZILHFD(LessOrEqual, 12, 15)
+
+// Opset-16 adds BFloat16 to allowed types for the GreaterOrEqual operator
+BINARY_LOGICALOP_REGISTER_UZILHFD(GreaterOrEqual, 16)
+
+// Opset-16 adds BFloat16 to allowed types for the LessOrEqual operator
+BINARY_LOGICALOP_REGISTER_UZILHFD(LessOrEqual, 16)
} // namespace cuda
} // namespace onnxruntime
diff --git a/onnxruntime/core/providers/cuda/tensor/where.cc b/onnxruntime/core/providers/cuda/tensor/where.cc
index e718ba0876..432f436aac 100644
--- a/onnxruntime/core/providers/cuda/tensor/where.cc
+++ b/onnxruntime/core/providers/cuda/tensor/where.cc
@@ -10,10 +10,21 @@ namespace cuda {
// kernel builder functions
#define WHERE_TYPED_KERNEL_WITH_TYPE_NAME(T, TName) \
- ONNX_OPERATOR_TYPED_KERNEL_EX( \
+ ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_EX( \
Where, \
kOnnxDomain, \
9, \
+ 15, \
+ TName, \
+ kCudaExecutionProvider, \
+ (*KernelDefBuilder::Create()) \
+ .TypeConstraint("B", DataTypeImpl::GetTensorType()) \
+ .TypeConstraint("T", DataTypeImpl::GetTensorType()), \
+ Where); \
+ ONNX_OPERATOR_TYPED_KERNEL_EX( \
+ Where, \
+ kOnnxDomain, \
+ 16, \
TName, \
kCudaExecutionProvider, \
(*KernelDefBuilder::Create()) \
diff --git a/onnxruntime/core/providers/rocm/rocm_execution_provider.cc b/onnxruntime/core/providers/rocm/rocm_execution_provider.cc
index 9198351bda..5121c1e626 100644
--- a/onnxruntime/core/providers/rocm/rocm_execution_provider.cc
+++ b/onnxruntime/core/providers/rocm/rocm_execution_provider.cc
@@ -398,9 +398,9 @@ class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain,
class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 6, float, HardSigmoid);
class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 6, double, HardSigmoid);
class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 6, MLFloat16, HardSigmoid);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 6, float, LeakyRelu);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 6, double, LeakyRelu);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 6, MLFloat16, LeakyRelu);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 6, 15, float, LeakyRelu);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 6, 15, double, LeakyRelu);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 6, 15, MLFloat16, LeakyRelu);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 6, 12, float, Relu);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 6, 12, double, Relu);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 6, 12, MLFloat16, Relu);
@@ -431,9 +431,9 @@ class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kO
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 7, 8, float, PRelu);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 7, 8, double, PRelu);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 7, 8, MLFloat16, PRelu);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, float, PRelu);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, double, PRelu);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, MLFloat16, PRelu);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 15, float, PRelu);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 15, double, PRelu);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 15, MLFloat16, PRelu);
class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 7, bool, And);
class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 7, bool, Or);
class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 7, bool, Xor);
@@ -457,20 +457,20 @@ class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kO
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 12, float, Greater);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 12, double, Greater);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 12, MLFloat16, Greater);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, int32_t, GreaterOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, int64_t, GreaterOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, uint32_t, GreaterOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, uint64_t, GreaterOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, float, GreaterOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, double, GreaterOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, MLFloat16, GreaterOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, int32_t, LessOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, int64_t, LessOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, uint32_t, LessOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, uint64_t, LessOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, float, LessOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, double, LessOrEqual);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, MLFloat16, LessOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, 15, int32_t, GreaterOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, 15, int64_t, GreaterOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, 15, uint32_t, GreaterOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, 15, uint64_t, GreaterOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, 15, float, GreaterOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, 15, double, GreaterOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, 15, MLFloat16, GreaterOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, 15, int32_t, LessOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, 15, int64_t, LessOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, 15, uint32_t, LessOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, 15, uint64_t, LessOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, 15, float, LessOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, 15, double, LessOrEqual);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 12, 15, MLFloat16, LessOrEqual);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 7, 12, int32_t, Add);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 7, 12, int64_t, Add);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 7, 12, uint32_t, Add);
@@ -699,12 +699,12 @@ class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kO
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 12, MLFloat16, Less);
class ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, EyeLike);
class ONNX_OPERATOR_VERSIONED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 10, Scatter);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, MLFloat16, Where);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, float, Where);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, double_t, Where);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, int32_t, Where);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, int64_t, Where);
-class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, uint8_t, Where);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 15, MLFloat16, Where);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 15, float, Where);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 15, double_t, Where);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 15, int32_t, Where);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 15, int64_t, Where);
+class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 15, uint8_t, Where);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 12, bool, NonZero);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 12, uint8_t, NonZero);
class ONNX_OPERATOR_VERSIONED_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 9, 12, int32_t, NonZero);
@@ -1197,6 +1197,34 @@ class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain,
class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 15, MLFloat16, BatchNormalization);
class ONNX_OPERATOR_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 15, Shape);
+// Opset 16
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, float, PRelu);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, double, PRelu);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, MLFloat16, PRelu);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, float, LeakyRelu);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, double, LeakyRelu);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, MLFloat16, LeakyRelu);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, MLFloat16, Where);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, float, Where);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, double_t, Where);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, int32_t, Where);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, int64_t, Where);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, uint8_t, Where);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, int32_t, GreaterOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, int64_t, GreaterOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, uint32_t, GreaterOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, uint64_t, GreaterOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, float, GreaterOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, double, GreaterOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, MLFloat16, GreaterOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, int32_t, LessOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, int64_t, LessOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, uint32_t, LessOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, uint64_t, LessOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, float, LessOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, double, LessOrEqual);
+class ONNX_OPERATOR_TYPED_KERNEL_CLASS_NAME(kRocmExecutionProvider, kOnnxDomain, 16, MLFloat16, LessOrEqual);
+
template <>
KernelCreateInfo BuildKernelCreateInfo() {
return {};
@@ -1204,7 +1232,7 @@ KernelCreateInfo BuildKernelCreateInfo() {
static Status RegisterRocmKernels(KernelRegistry& kernel_registry) {
static const BuildKernelCreateInfoFn function_table[] = {
- BuildKernelCreateInfo, // default entry to avoid the list become empty after ops-reducing
+ BuildKernelCreateInfo, //default entry to avoid the list become empty after ops-reducing
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
@@ -1240,9 +1268,9 @@ static Status RegisterRocmKernels(KernelRegistry& kernel_registry) {
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
@@ -1273,9 +1301,9 @@ static Status RegisterRocmKernels(KernelRegistry& kernel_registry) {
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
@@ -1299,20 +1327,20 @@ static Status RegisterRocmKernels(KernelRegistry& kernel_registry) {
BuildKernelCreateInfo,
BuildKernelCreateInfo,
BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
- BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo,
+ BuildKernelCreateInfo