coreml: Add a few missing operators

Author: borg323

onnxruntime/core/providers/coreml/builders/impl/activation_op_builder.cc

@@ -118,6 +118,11 @@ Status ActivationOpBuilder::AddToModelBuilderImpl(ModelBuilder& model_builder,
     } else if (op_type == "PRelu") {
       coreml_op_type = "prelu";
       add_alpha = true;
+    } else if (op_type == "Softplus") {
+      coreml_op_type = "softplus";
+    } else if (op_type == "Elu") {
+      coreml_op_type = "elu";
+      add_alpha = true;
     } else {
       return ORT_MAKE_STATUS(ONNXRUNTIME, INVALID_ARGUMENT,
                              "ActivationOpBuilder::AddToModelBuilderImpl, unknown op: ", op_type);
@@ -141,7 +146,7 @@ Status ActivationOpBuilder::AddToModelBuilderImpl(ModelBuilder& model_builder,
         }
       } else {
         NodeAttrHelper helper(node);
-        const auto alpha = helper.Get("alpha", 0.01f);
+        const auto alpha = helper.Get("alpha", "Elu" == op_type ? 1.0f : 0.01f);
 
         if (input_dtype == ONNX_NAMESPACE::TensorProto_DataType_FLOAT) {
           AddOperationInput(*op, "alpha", model_builder.AddScalarConstant(op->type(), "alpha", alpha));
@@ -259,8 +264,10 @@ bool ActivationOpBuilder::IsOpSupportedImpl(const Node& node, const OpBuilderInp
                                             const logging::Logger& logger) const {
   const auto& op_type = node.OpType();
 
-  if (op_type == "Gelu" && !input_params.create_mlprogram) {
-    return false;
+  if (!input_params.create_mlprogram) {
+    if (op_type == "Gelu" || op_type == "Softplus" || op_type == "Elu") {
+      return false;
+    }
   }
   if (op_type == "PRelu") {
     return IsPReluOpSupported(node, input_params, logger);
@@ -269,8 +276,13 @@ bool ActivationOpBuilder::IsOpSupportedImpl(const Node& node, const OpBuilderInp
   return true;
 }
 
-int ActivationOpBuilder::GetMinSupportedOpSet(const Node& /* node */) const {
-  // All ops opset 5- uses consumed_inputs attribute which is not supported for now
+int ActivationOpBuilder::GetMinSupportedOpSet(const Node& node) const {
+  const auto& op_type(node.OpType());
+  // Softplus was unmodified from opset 1 to 21 (with no attributes).
+  if (op_type == "Softplus") {
+    return 1;
+  }
+  // All other ops opset 5- uses consumed_inputs attribute which is not supported for now.
   return 6;
 }
 
@@ -286,6 +298,8 @@ void CreateActivationOpBuilder(const std::string& op_type, OpBuilderRegistration
           "PRelu",
           "LeakyRelu",
           "Gelu",
+          "Softplus",
+          "Elu",
       };
 
   op_registrations.builders.push_back(std::make_unique<ActivationOpBuilder>());

onnxruntime/core/providers/coreml/builders/impl/base_op_builder.cc

@@ -105,7 +105,7 @@ bool BaseOpBuilder::HasSupportedInputsImpl(const Node& node, const OpBuilderInpu
 bool BaseOpBuilder::HasSupportedOpSet(const Node& node, const logging::Logger& logger) const {
   auto since_version = node.SinceVersion();
   if (since_version < GetMinSupportedOpSet(node) || since_version > GetMaxSupportedOpSet(node)) {
-    LOGS(logger, VERBOSE) << node.OpType() << "is only supported for opset ["
+    LOGS(logger, VERBOSE) << node.OpType() << " is only supported for opset ["
                           << GetMinSupportedOpSet(node) << ", "
                           << GetMaxSupportedOpSet(node) << "]";
     return false;

onnxruntime/core/providers/coreml/builders/impl/unary_op_builder.cc

@@ -37,6 +37,10 @@ Status UnaryOpBuilder::AddToModelBuilderImpl(ModelBuilder& model_builder, const
       coreml_op_type = "erf";
     } else if (op_type == "Round") {
       coreml_op_type = "round";
+    } else if (op_type == "Exp") {
+      coreml_op_type = "exp";
+    } else if (op_type == "Log") {
+      coreml_op_type = "log";
     } else {
       return ORT_MAKE_STATUS(ONNXRUNTIME, INVALID_ARGUMENT,
                              "UnaryOpBuilder::AddToModelBuilderImpl, unexpected op: ", op_type);
@@ -79,8 +83,10 @@ Status UnaryOpBuilder::AddToModelBuilderImpl(ModelBuilder& model_builder, const
 
 bool UnaryOpBuilder::IsOpSupportedImpl(const Node& node, const OpBuilderInputParams& input_params,
                                        const logging::Logger& /*logger*/) const {
-  if (!input_params.create_mlprogram && (node.OpType() == "Erf" || node.OpType() == "Round")) {
-    return false;
+  if (!input_params.create_mlprogram) {
+    if (node.OpType() == "Erf" || node.OpType() == "Round" || node.OpType() == "Exp" || node.OpType() == "Log") {
+      return false;
+    }
   }
   return true;
 }

onnxruntime/core/providers/coreml/builders/op_builder_factory.cc

@@ -22,12 +22,16 @@ static OpBuilderRegistrations CreateOpBuilderRegistrations() {
   CreateActivationOpBuilder("PRelu", op_registrations);
   CreateActivationOpBuilder("LeakyRelu", op_registrations);
   CreateActivationOpBuilder("Gelu", op_registrations);
+  CreateActivationOpBuilder("Softplus", op_registrations);
+  CreateActivationOpBuilder("Elu", op_registrations);
 
   // Unary ops
   CreateUnaryOpBuilder("Erf", op_registrations);
   CreateUnaryOpBuilder("Reciprocal", op_registrations);
   CreateUnaryOpBuilder("Round", op_registrations);
   CreateUnaryOpBuilder("Sqrt", op_registrations);
+  CreateUnaryOpBuilder("Exp", op_registrations);
+  CreateUnaryOpBuilder("Log", op_registrations);
 
   // Binary elementwise ops
   CreateBinaryOpBuilder("Add", op_registrations);

tools/ci_build/github/apple/coreml_supported_mlprogram_ops.md

@@ -13,7 +13,9 @@ Keep in sync with doco generated from /docs/execution-providers/CoreML-Execution
 |ai.onnx:ConvTranspose|Weight and bias must be constant.<br/>padding_type of SAME_UPPER/SAME_LOWER is not supported.<br/>kernel_shape must have default values.<br/>output_shape is not supported.<br/>output_padding must have default values.|
 |ai.onnx:DepthToSpace|If 'mode' is 'CRD' the input must have a fixed shape.|
 |ai.onnx:Div||
+|ai.onnx:Elu||
 |ai.onnx:Erf||
+|ai.onnx:Exp||
 |ai.onnx:Gemm|Input B must be constant.|
 |ai.onnx:Gelu||
 |ai.onnx:GlobalAveragePool|Only 2D Pool is supported currently. 3D and 5D support can be added if needed.|
@@ -23,6 +25,7 @@ Keep in sync with doco generated from /docs/execution-providers/CoreML-Execution
 |ai.onnx:InstanceNormalization||
 |ai.onnx:LayerNormalization||
 |ai.onnx:LeakyRelu||
+|ai.onnx:Log||
 |ai.onnx:MatMul|Only support for transA == 0, alpha == 1.0 and beta == 1.0 is currently implemented.|
 |ai.onnx:MaxPool|Only 2D Pool is supported currently. 3D and 5D support can be added if needed.|
 |ai.onnx:Max||
@@ -39,6 +42,7 @@ Keep in sync with doco generated from /docs/execution-providers/CoreML-Execution
 |ai.onnx:Round||
 |ai.onnx:Shape||
 |ai.onnx:Slice|starts/ends/axes/steps must be constant initializers.|
+|ai.onnx:Softplus||
 |ai.onnx:Split|If provided, `splits` must be constant.|
 |ai.onnx:Sub||
 |ai.onnx:Sigmoid||