Fix for inverse spec (#2706)

* Fixed spec for input rank

* Full-rank enforce

* Updated spec

* Updated spec for behavior

* Type constrain updated

* Refactor desc

* Shape inference changes

* Fail for invalid shapes

* float type only

* Typo

docs/Changelog.md

@@ -14445,8 +14445,10 @@ This version of the operator has been available since version 12 of the default
 
   Calculates inverse of a square matrix or batches of square matrices.
   Inverse takes one input tensor of shape `[*, M, M]`, where `*` is zero or more batch dimensions,
-  and the inner-most 2 dimensions form square matrices.
-  The output is a tensor of shape `[*, M, M]`, containing the individual inverses of all input submatrices.
+  and the inner-most 2 dimensions form square matrices. These matrices must be invertible (full-rank).
+  The behavior where one of the matrices is not invertible is undefined. The implementation can choose
+  to throw an error or output (garbage) results as is. The output is a tensor of shape `[*, M, M]`,
+  containing the individual inverses of all input submatrices.
 
 #### Version
 
@@ -14456,21 +14458,21 @@ This version of the operator has been available since version 12 of the default
 
 <dl>
 <dt><tt>X</tt> : T</dt>
-<dd>Input tensor</dd>
+<dd>Input tensor. Every matrix in the batch must be invertible.</dd>
 </dl>
 
 #### Outputs
 
 <dl>
 <dt><tt>Y</tt> : T</dt>
-<dd>Output tensor of the same type as input.</dd>
+<dd>Output tensor of the same type and shape as the input tensor.</dd>
 </dl>
 
 #### Type Constraints
 
 <dl>
-<dt><tt>T</tt> : tensor(uint8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(int8), tensor(int16), tensor(int32), tensor(int64), tensor(float16), tensor(float), tensor(double)</dt>
-<dd>Constrain input and output types to all numerical tensor types.</dd>
+<dt><tt>T</tt> : tensor(float16), tensor(float), tensor(double)</dt>
+<dd>Constrain input and output types to float tensors.</dd>
 </dl>
 
 ### <a name="LessOrEqual-12"></a>**LessOrEqual-12**</a>

docs/Operators.md

@@ -7368,8 +7368,10 @@ expect(node, inputs=[x, s, bias], outputs=[y],
 
   Calculates inverse of a square matrix or batches of square matrices.
   Inverse takes one input tensor of shape `[*, M, M]`, where `*` is zero or more batch dimensions,
-  and the inner-most 2 dimensions form square matrices.
-  The output is a tensor of shape `[*, M, M]`, containing the individual inverses of all input submatrices.
+  and the inner-most 2 dimensions form square matrices. These matrices must be invertible (full-rank).
+  The behavior where one of the matrices is not invertible is undefined. The implementation can choose
+  to throw an error or output (garbage) results as is. The output is a tensor of shape `[*, M, M]`,
+  containing the individual inverses of all input submatrices.
 
 #### Version
 
@@ -7379,21 +7381,21 @@ This version of the operator has been available since version 12 of the default
 
 <dl>
 <dt><tt>X</tt> : T</dt>
-<dd>Input tensor</dd>
+<dd>Input tensor. Every matrix in the batch must be invertible.</dd>
 </dl>
 
 #### Outputs
 
 <dl>
 <dt><tt>Y</tt> : T</dt>
-<dd>Output tensor of the same type as input.</dd>
+<dd>Output tensor of the same type and shape as the input tensor.</dd>
 </dl>
 
 #### Type Constraints
 
 <dl>
-<dt><tt>T</tt> : tensor(uint8), tensor(uint16), tensor(uint32), tensor(uint64), tensor(int8), tensor(int16), tensor(int32), tensor(int64), tensor(float16), tensor(float), tensor(double)</dt>
-<dd>Constrain input and output types to all numerical tensor types.</dd>
+<dt><tt>T</tt> : tensor(float16), tensor(float), tensor(double)</dt>
+<dd>Constrain input and output types to float tensors.</dd>
 </dl>
 
 

onnx/defs/math/defs.cc

@@ -2297,8 +2297,10 @@ ONNX_OPERATOR_SET_SCHEMA(
 static const char* Inverse_ver12_doc = R"DOC(
 Calculates inverse of a square matrix or batches of square matrices.
 Inverse takes one input tensor of shape `[*, M, M]`, where `*` is zero or more batch dimensions,
-and the inner-most 2 dimensions form square matrices.
-The output is a tensor of shape `[*, M, M]`, containing the individual inverses of all input submatrices.
+and the inner-most 2 dimensions form square matrices. These matrices must be invertible (full-rank).
+The behavior where one of the matrices is not invertible is undefined. The implementation can choose
+to throw an error or output (garbage) results as is. The output is a tensor of shape `[*, M, M]`,
+containing the individual inverses of all input submatrices.
 )DOC";
 
 ONNX_OPERATOR_SET_SCHEMA(
@@ -2306,12 +2308,14 @@ ONNX_OPERATOR_SET_SCHEMA(
     12,
     OpSchema()
         .SetDoc(Inverse_ver12_doc)
-        .Input(0, "X", "Input tensor", "T")
-        .Output(0, "Y", "Output tensor of the same type as input.", "T")
+        .Input(0, "X", "Input tensor. Every matrix in the batch must be invertible.", "T")
+        .Output(0, "Y", "Output tensor of the same type and shape as the input tensor.", "T")
         .TypeConstraint(
             "T",
-            OpSchema::all_numeric_types(),
-            "Constrain input and output types to all numerical tensor types.")
+            {"tensor(float16)",
+             "tensor(float)",
+             "tensor(double)"},
+            "Constrain input and output types to float tensors.")
         .TypeAndShapeInferenceFunction([](InferenceContext& ctx) {
           // Type inference
           propagateElemTypeFromInputToOutput(ctx, 0, 0);

onnx/test/shape_inference_test.py

@@ -1284,6 +1284,25 @@ class TestShapeInference(unittest.TestCase):
             [])
         self._assert_inferred(graph, [make_tensor_value_info('z', TensorProto.FLOAT, (4, 5, 6))])
 
+    def test_inverse_float(self):  # type: () -> None
+        graph = self._make_graph(
+            [('x', TensorProto.FLOAT16, (3, 4, 4))],
+            [make_node('Inverse', ['x'], 'y')],
+            [])
+        self._assert_inferred(graph, [make_tensor_value_info('y', TensorProto.FLOAT16, (3, 4, 4))])
+
+        graph = self._make_graph(
+            [('x', TensorProto.FLOAT, (2, 5, 5))],
+            [make_node('Inverse', ['x'], 'y')],
+            [])
+        self._assert_inferred(graph, [make_tensor_value_info('y', TensorProto.FLOAT, (2, 5, 5))])
+
+        graph = self._make_graph(
+            [('x', TensorProto.DOUBLE, (5, 5))],
+            [make_node('Inverse', ['x'], 'y')],
+            [])
+        self._assert_inferred(graph, [make_tensor_value_info('y', TensorProto.DOUBLE, (5, 5))])
+
     def test_logsoftmax_2d(self):  # type: () -> None
         graph = self._make_graph(
             [('x', TensorProto.FLOAT, (4, 5))],