Orに対応

python/src/nnabla/utils/converter/onnx/ONNXOpCoverage.md

@@ -61,7 +61,7 @@ that indicates if each operator can be converted to NNP.
 |Mul|:yellow_heart:|broadcast will be converted to a BroadcastTo|
 |Neg|:black_heart:||
 |Not|:green_heart:||
-|Or|:black_heart:||
+|Or|:yellow_heart:|broadcast will be converted to a BroadcastTo|
 |PRelu|:black_heart:||
 |Pad|:black_heart:||
 |Pow|:black_heart:||

python/src/nnabla/utils/converter/onnx/exporter.py

@@ -49,6 +49,7 @@
     "Mean": "ReduceMean",
     "Mul2": "Mul",
     "LogicalAnd": "And",
+    "LogicalOr": "Or",
     # optype that gets converted
     "Identity": "Dropout",
     "Affine": "Gemm",
@@ -270,19 +271,15 @@ def convert_to_nodes(func, variables, input_types, output_types, broadcast_targe
         bp = func.broadcast_to_param
         broadcast_target[func.output[0]] = (func.input[1], bp.axis)
         # we do not append node here because BroadcastTo should disappear
-    elif func.type == "Add2":
+    elif (func.type == "Add2" or
+          func.type == "Mul2"):
         # Check if the second input is a brodcast target.
         bt = func.input[1]
         if bt in broadcast_target:
             merge_broadcast(n, func, bt, broadcast_target)
         nl.append(n)
-    elif func.type == "Mul2":
-        # Check if the second input is a brodcast target.
-        bt = func.input[1]
-        if bt in broadcast_target:
-            merge_broadcast(n, func, bt, broadcast_target)
-        nl.append(n)
-    elif func.type == "LogicalAnd":
+    elif (func.type == "LogicalAnd" or
+          func.type == "LogicalOr"):
         # Store the input/output tensor's name and convert it to boolean
         input_types[n.input[0]] = TensorProto.BOOL
         output_types[n.output[0]] = TensorProto.BOOL

python/src/nnabla/utils/converter/onnx/reader.py

@@ -66,6 +66,7 @@
     "ReduceMax": "Max",
     "ReduceProd": "Prod",
     "And": "LogicalAnd",
+    "Or": "LogicalOr",
     # Constant does not get converted to a function
     # but we list it here so we can accept it
     "Constant": ""
@@ -519,13 +520,10 @@ def convert_to_functions(pb, network, node, base_name, initializers,
         if len(func.input) > 2:
             raise ValueError("Sum operations with more than two input is currently not supported")
         func_list.append(func)
-    elif node.op_type == "Add":
-        convert_broadcasting_operator(func_list, node, func, base_name, func_counter)
-        func_list.append(func)
-    elif node.op_type == "Mul":
-        convert_broadcasting_operator(func_list, node, func, base_name, func_counter)
-        func_list.append(func)
-    elif node.op_type == "And":
+    elif (node.op_type == "Add" or
+          node.op_type == "Mul" or
+          node.op_type == "And" or
+          node.op_type == "Or"):
         convert_broadcasting_operator(func_list, node, func, base_name, func_counter)
         func_list.append(func)
     elif node.op_type == "Constant":

python/test/utils/conversion/test_conversion.py

@@ -515,6 +515,33 @@ def test_nnp_onnx_conversion_and_broadcast_axis1(tmpdir, nnp_fixture):
                                     "and_broadcast_axis1.onnx",
                                     "out_data_1", "exec_0")
 
+
+def test_onnx_nnp_conversion_or_no_broadcast(tmpdir, nnp_fixture):
+    convert_onnx_to_nnp_and_compare(tmpdir, TEST_DATA_DIR,
+                                    "or_no_broadcast.onnx",
+                                    "or_no_broadcast.nnp",
+                                    "out_data_1", "exec_0")
+
+
+def test_nnp_onnx_conversion_or_no_broadcast(tmpdir, nnp_fixture):
+    convert_nnp_to_onnx_and_compare(tmpdir, TEST_DATA_DIR,
+                                    "or_no_broadcast.nnp",
+                                    "or_no_broadcast.onnx",
+                                    "out_data_1", "exec_0")
+
+
+def test_onnx_nnp_conversion_or_broadcast_axis1(tmpdir, nnp_fixture):
+    convert_onnx_to_nnp_and_compare(tmpdir, TEST_DATA_DIR,
+                                    "or_broadcast_axis1.onnx",
+                                    "or_broadcast_axis1.nnp",
+                                    "out_data_1", "exec_0")
+
+
+def test_nnp_onnx_conversion_or_broadcast_axis1(tmpdir, nnp_fixture):
+    convert_nnp_to_onnx_and_compare(tmpdir, TEST_DATA_DIR,
+                                    "or_broadcast_axis1.nnp",
+                                    "or_broadcast_axis1.onnx",
+                                    "out_data_1", "exec_0")
 # These following tests are invalidated due to a
 # backend bug? decribed in the following issue:
 # https://github.com/Microsoft/CNTK/issues/3127