fix the problems introduced in previous PRs in operator registration (onnx#1878)

* fix the problems introduced in previous PRs in operator registration

* update the doc

* fix the doc

Author: houseroad

docs/Changelog.md

@@ -9486,6 +9486,51 @@ This version of the operator has been available since version 10 of the default
 <dd>Constrain input and output types to float tensors.</dd>
 </dl>
 
+### <a name="Dropout-10"></a>**Dropout-10**</a>
+
+  Dropout takes one input floating tensor and produces two tensor outputs,
+  output (floating tensor) and mask (`Tensor<bool>`). Depending on whether it is
+  in test mode or not, the output Y will either be a random dropout, or a simple
+  copy of the input. Note that our implementation of Dropout does scaling in
+  the training phase, so during testing nothing needs to be done.
+  This operator has **optional** inputs/outputs. See [the doc](IR.md) for more details about the representation of optional arguments. An empty string may be used in the place of an actual argument's name to indicate a missing argument. Trailing optional arguments (those not followed by an argument that is present) may also be simply omitted.
+
+#### Version
+
+This version of the operator has been available since version 10 of the default ONNX operator set.
+
+#### Attributes
+
+<dl>
+<dt><tt>ratio</tt> : float (default is 0.5)</dt>
+<dd>The ratio of random dropout</dd>
+</dl>
+
+#### Inputs
+
+<dl>
+<dt><tt>data</tt> : T</dt>
+<dd>The input data as Tensor.</dd>
+</dl>
+
+#### Outputs (1 - 2)
+
+<dl>
+<dt><tt>output</tt> : T</dt>
+<dd>The output.</dd>
+<dt><tt>mask</tt> (optional) : T1</dt>
+<dd>The output mask.</dd>
+</dl>
+
+#### Type Constraints
+
+<dl>
+<dt><tt>T</tt> : tensor(float16), tensor(float), tensor(double)</dt>
+<dd>Constrain input and output types to float tensors.</dd>
+<dt><tt>T1</tt> : tensor(bool)</dt>
+<dd>Constrain output mask types to boolean tensors.</dd>
+</dl>
+
 ### <a name="MaxPool-10"></a>**MaxPool-10**</a>
 
   MaxPool consumes an input tensor X and applies max pooling across

docs/Operators.md

@@ -3093,18 +3093,18 @@ expect(node, inputs=[x, y], outputs=[z],
 
 ### <a name="Dropout"></a><a name="dropout">**Dropout**</a>
 
-  Dropout takes one input data (Tensor<float>) and produces two Tensor outputs,
-  output (Tensor<float>) and mask (Tensor<bool>). Depending on whether it is in
-  test mode or not, the output Y will either be a random dropout, or a simple
+  Dropout takes one input floating tensor and produces two tensor outputs,
+  output (floating tensor) and mask (`Tensor<bool>`). Depending on whether it is
+  in test mode or not, the output Y will either be a random dropout, or a simple
   copy of the input. Note that our implementation of Dropout does scaling in
   the training phase, so during testing nothing needs to be done.
   This operator has **optional** inputs/outputs. See [the doc](IR.md) for more details about the representation of optional arguments. An empty string may be used in the place of an actual argument's name to indicate a missing argument. Trailing optional arguments (those not followed by an argument that is present) may also be simply omitted.
 
 #### Version
 
-This version of the operator has been available since version 7 of the default ONNX operator set.
+This version of the operator has been available since version 10 of the default ONNX operator set.
 
-Other versions of this operator: <a href="Changelog.md#Dropout-1">Dropout-1</a>, <a href="Changelog.md#Dropout-6">Dropout-6</a>
+Other versions of this operator: <a href="Changelog.md#Dropout-1">Dropout-1</a>, <a href="Changelog.md#Dropout-6">Dropout-6</a>, <a href="Changelog.md#Dropout-7">Dropout-7</a>
 
 #### Attributes
 
@@ -3125,7 +3125,7 @@ Other versions of this operator: <a href="Changelog.md#Dropout-1">Dropout-1</a>,
 <dl>
 <dt><tt>output</tt> : T</dt>
 <dd>The output.</dd>
-<dt><tt>mask</tt> (optional) : T</dt>
+<dt><tt>mask</tt> (optional) : T1</dt>
 <dd>The output mask.</dd>
 </dl>
 
@@ -3134,6 +3134,8 @@ Other versions of this operator: <a href="Changelog.md#Dropout-1">Dropout-1</a>,
 <dl>
 <dt><tt>T</tt> : tensor(float16), tensor(float), tensor(double)</dt>
 <dd>Constrain input and output types to float tensors.</dd>
+<dt><tt>T1</tt> : tensor(bool)</dt>
+<dd>Constrain output mask types to boolean tensors.</dd>
 </dl>
 
 

onnx/defs/experiments/defs.cc

@@ -149,63 +149,4 @@ ONNX_OPERATOR_SET_SCHEMA(
              "tensor(double)"},
             "Constrain output types to bool, int32, int64, float16, float, double tensors."));
 
-static const char* ImageScaler_ver1_doc =
-    R"DOC(Scale and bias the input image. Bias values are stored in
-the same ordering as the image pixel format.)DOC";
-
-ONNX_OPERATOR_SET_SCHEMA(
-    ImageScaler,
-    1,
-    OpSchema()
-        .SetSupportLevel(SupportType::EXPERIMENTAL)
-        .SetDoc(ImageScaler_ver1_doc)
-        .Attr(
-            "bias",
-            "Bias applied to each channel, same size as C.",
-            AttributeProto::FLOATS,
-            OPTIONAL)
-        .Attr(
-            "scale",
-            "The scale to apply.",
-            AttributeProto::FLOAT,
-            1.0f)
-        .Input(0, "input", "Input tensor of shape [N,C,H,W]", "T")
-        .Output(0, "output", "Result, has same shape and type as input", "T")
-        .TypeConstraint(
-            "T",
-            {"tensor(float16)", "tensor(float)", "tensor(double)"},
-            "Constrain input and output types to float tensors.")
-        .TypeAndShapeInferenceFunction(propagateShapeAndTypeFromFirstInput));
-
-static const char* Crop_ver1_doc =
-    R"DOC(Crop and image to the specified spatial dimensions. If scale is given,
-then optionally start the crop offset by the left/top border amounts.
-If scale is not provided, crop the borders as provided.)DOC";
-
-ONNX_OPERATOR_SET_SCHEMA(
-    Crop,
-    1,
-    OpSchema()
-        .SetSupportLevel(SupportType::EXPERIMENTAL)
-        .SetDoc(Crop_ver1_doc)
-        .Attr(
-            "border",
-            "A 1-D values of (leftBorder, topBorder, rightBorder, bottomBorder).",
-            AttributeProto::INTS,
-            OPTIONAL)
-        .Attr(
-            "scale",
-            "A 1-D values of (height, width).",
-            AttributeProto::INTS,
-            OPTIONAL)
-        .Input(0, "input", "Input tensor of shape [N,C,H,W]", "T")
-        .Output(
-            0,
-            "output",
-            "Result, has same type as input, with H and W dimensions reduced.",
-            "T")
-        .TypeConstraint(
-            "T",
-            {"tensor(float16)", "tensor(float)", "tensor(double)"},
-            "Constrain input and output types to float tensors."));
 } // namespace ONNX_NAMESPACE

onnx/defs/operator_sets.h

@@ -530,6 +530,7 @@ class ONNX_OPERATOR_SET_SCHEMA_CLASS_NAME(Onnx, 10, MaxPool);
 class ONNX_OPERATOR_SET_SCHEMA_CLASS_NAME(Onnx, 10, AveragePool);
 class ONNX_OPERATOR_SET_SCHEMA_CLASS_NAME(Onnx, 10, Slice);
 class ONNX_OPERATOR_SET_SCHEMA_CLASS_NAME(Onnx, 10, ThresholdedRelu);
+class ONNX_OPERATOR_SET_SCHEMA_CLASS_NAME(Onnx, 10, Dropout);
 
 // Iterate over schema from ai.onnx version 10
 class OpSet_Onnx_ver10 {
@@ -544,11 +545,13 @@ class OpSet_Onnx_ver10 {
     fn(GetOpSchema<ONNX_OPERATOR_SET_SCHEMA_CLASS_NAME(
            Onnx, 10, MaxPool)>());
     fn(GetOpSchema<ONNX_OPERATOR_SET_SCHEMA_CLASS_NAME(
-           Onnx, 10, AveragePool)>());	   
+           Onnx, 10, AveragePool)>());
     fn(GetOpSchema<ONNX_OPERATOR_SET_SCHEMA_CLASS_NAME(
            Onnx, 10, Slice)>());
     fn(GetOpSchema<ONNX_OPERATOR_SET_SCHEMA_CLASS_NAME(
            Onnx, 10, ThresholdedRelu)>());
+    fn(GetOpSchema<ONNX_OPERATOR_SET_SCHEMA_CLASS_NAME(
+           Onnx, 10, Dropout)>());
   }
 };
 

onnx/defs/schema.h

@@ -715,7 +715,6 @@ class OpSchemaRegistry final : public ISchemaRegistry {
           fail_schema(err.str());
         }
 
-        
         m[op_name][op_domain].insert(std::pair<int, OpSchema&&>(ver, std::move(op_schema)));
 
       } catch (const std::exception& e) {