Add Greater and others for v9 (#270)

Add v9 support for Greater, Constant, Flatten, Gemm,
MatMul, and PRelu. No logic changes seem needed, just to
declare support of v9.

Author: chinhuang007

doc/support_status.md

@@ -22,7 +22,7 @@ ______
 |Ceil|1, 6|
 |Clip|1, 6|
 |Concat|1, 4|
-|Constant|1|
+|Constant|1, 9|
 |ConstantFill|1|
 |ConstantLike|9|
 |Conv|1|
@@ -38,17 +38,17 @@ ______
 |Exp|1, 6|
 |Expand|8|
 |EyeLike|N/A|
-|Flatten|1|
+|Flatten|1, 9|
 |Floor|1, 6|
 |GRU|1, 3, 7|
 |GRUUnit|N/A|
 |Gather|1|
-|Gemm|1, 6, 7|
+|Gemm|1, 6, 7, 9|
 |GivenTensorFill|N/A|
 |GlobalAveragePool|1|
 |GlobalLpPool|1, 2|
 |GlobalMaxPool|1|
-|Greater|1, 7|
+|Greater|1, 7, 9|
 |HardSigmoid|1, 6|
 |Hardmax|1|
 |Identity|1|
@@ -64,7 +64,7 @@ ______
 |Loop|N/A|
 |LpNormalization|1|
 |LpPool|N/A|
-|MatMul|1|
+|MatMul|1, 9|
 |Max|1, 6, 8|
 |MaxPool|1, 8|
 |MaxRoiPool|N/A|
@@ -76,7 +76,7 @@ ______
 |Neg|1, 6|
 |Not|1|
 |Or|1, 7|
-|PRelu|1, 6, 7|
+|PRelu|1, 6, 7, 9|
 |Pad|1, 2|
 |ParametricSoftplus|N/A|
 |Pow|1, 7|

onnx_tf/handlers/backend/constant.py

@@ -12,11 +12,19 @@
 class Constant(BackendHandler):
 
   @classmethod
-  def version_1(cls, node, **kwargs):
+  def _common(cls, node, **kwargs):
     attr_value = node.attrs["value"]
     dtype = data_type.onnx2tf(attr_value.data_type)
     value = numpy_helper.to_array(attr_value)
     return [
         cls.make_tensor_from_onnx_node(
             node, inputs=[value], attrs={"dtype": dtype})
     ]
+
+  @classmethod
+  def version_1(cls, node, **kwargs):
+    return cls._common(node, **kwargs)
+
+  @classmethod
+  def version_9(cls, node, **kwargs):
+    return cls._common(node, **kwargs)

onnx_tf/handlers/backend/flatten.py

@@ -10,7 +10,7 @@
 class Flatten(BackendHandler):
 
   @classmethod
-  def version_1(cls, node, **kwargs):
+  def _common(cls, node, **kwargs):
     x = kwargs["tensor_dict"][node.inputs[0]]
     shape = tf.shape(x)
     x_rank = len(x.shape)
@@ -25,3 +25,12 @@ def version_1(cls, node, **kwargs):
       cal_shape = (tf.reduce_prod(shape[0:axis]),
                    tf.reduce_prod(shape[axis:tf.size(shape)]))
     return [tf.reshape(x, cal_shape)]
+
+  @classmethod
+  def version_1(cls, node, **kwargs):
+    return cls._common(node, **kwargs)
+
+  @classmethod
+  def version_9(cls, node, **kwargs):
+    return cls._common(node, **kwargs)
+

onnx_tf/handlers/backend/gemm.py

@@ -33,3 +33,7 @@ def version_6(cls, node, **kwargs):
   @classmethod
   def version_7(cls, node, **kwargs):
     return cls._common(node, **kwargs)
+
+  @classmethod
+  def version_9(cls, node, **kwargs):
+    return cls._common(node, **kwargs)

onnx_tf/handlers/backend/greater.py

@@ -17,3 +17,7 @@ def version_1(cls, node, **kwargs):
   @classmethod
   def version_7(cls, node, **kwargs):
     return [cls.make_tensor_from_onnx_node(node, **kwargs)]
+
+  @classmethod
+  def version_9(cls, node, **kwargs):
+    return [cls.make_tensor_from_onnx_node(node, **kwargs)]

onnx_tf/handlers/backend/mat_mul.py

@@ -12,3 +12,7 @@ class MatMul(BackendHandler):
   @classmethod
   def version_1(cls, node, **kwargs):
     return [cls.make_tensor_from_onnx_node(node, **kwargs)]
+
+  @classmethod
+  def version_9(cls, node, **kwargs):
+    return [cls.make_tensor_from_onnx_node(node, **kwargs)]

onnx_tf/handlers/backend/p_relu.py

@@ -32,3 +32,7 @@ def version_6(cls, node, **kwargs):
   @classmethod
   def version_7(cls, node, **kwargs):
     return cls._common(node, **kwargs)
+
+  @classmethod
+  def version_9(cls, node, **kwargs):
+    return cls._common(node, **kwargs)

onnx_tf/handlers/frontend/greater.py

@@ -15,3 +15,7 @@ def version_1(cls, node, **kwargs):
   @classmethod
   def version_7(cls, node, **kwargs):
     return cls.comparison_op(node, **kwargs)
+
+  @classmethod
+  def version_9(cls, node, **kwargs):
+    return cls.comparison_op(node, **kwargs)

onnx_tf/handlers/frontend/matmul.py

@@ -9,7 +9,7 @@
 class Matmul(FrontendHandler):
 
   @classmethod
-  def version_1(cls, node, **kwargs):
+  def _common(cls, node, **kwargs):
     transpose_a = node.attr.get("transpose_a", False)
     transpose_b = node.attr.get("transpose_b", False)
     input_a = node.inputs[0]
@@ -33,3 +33,11 @@ def version_1(cls, node, **kwargs):
       nodes.append(transposed_b)
     nodes.append(cls.make_node_from_tf_node(node, [input_a, input_b]))
     return nodes
+
+  @classmethod
+  def version_1(cls, node, **kwargs):
+    return cls._common(node, **kwargs)
+
+  @classmethod
+  def version_9(cls, node, **kwargs):
+    return cls._common(node, **kwargs)

onnx_tf/opset_version.py

@@ -15,7 +15,7 @@
     'Ceil': [1, 6],
     'Clip': [1, 6],
     'Concat': [1, 4],
-    'Constant': [1],
+    'Constant': [1, 9],
     'ConstantFill': [1],
     'ConstantLike': [9],
     'Conv': [1],
@@ -31,17 +31,17 @@
     'Exp': [1, 6],
     'Expand': [8],
     'EyeLike': [],
-    'Flatten': [1],
+    'Flatten': [1, 9],
     'Floor': [1, 6],
     'GRU': [1, 3, 7],
     'GRUUnit': [],
     'Gather': [1],
-    'Gemm': [1, 6, 7],
+    'Gemm': [1, 6, 7, 9],
     'GivenTensorFill': [],
     'GlobalAveragePool': [1],
     'GlobalLpPool': [1, 2],
     'GlobalMaxPool': [1],
-    'Greater': [1, 7],
+    'Greater': [1, 7, 9],
     'HardSigmoid': [1, 6],
     'Hardmax': [1],
     'Identity': [1],
@@ -57,7 +57,7 @@
     'Loop': [],
     'LpNormalization': [1],
     'LpPool': [],
-    'MatMul': [1],
+    'MatMul': [1, 9],
     'Max': [1, 6, 8],
     'MaxPool': [1, 8],
     'MaxRoiPool': [],
@@ -69,7 +69,7 @@
     'Neg': [1, 6],
     'Not': [1],
     'Or': [1, 7],
-    'PRelu': [1, 6, 7],
+    'PRelu': [1, 6, 7, 9],
     'Pad': [1, 2],
     'ParametricSoftplus': [],
     'Pow': [1, 7],