[Opset 10] updates for thresholded relu (#308)

onnx · Jun 4, 2019 · 0b883fe · 0b883fe
1 parent 5e70d8e
commit 0b883fe
Show file tree

Hide file tree

Showing 2 changed files with 19 additions and 4 deletions.
diff --git a/onnxmltools/convert/coreml/operator_converters/neural_network/Activation.py b/onnxmltools/convert/coreml/operator_converters/neural_network/Activation.py
@@ -43,11 +43,11 @@ def convert_activation(scope, operator, container):
     elif activation_type =='scaledTanh':
         apply_scaled_tanh(scope, inputs[0], outputs[0], container, operator_name=attrs['name'],
                           alpha=params.scaledTanh.alpha, beta=params.scaledTanh.beta)
+    elif activation_type == 'thresholdedReLU':
+        apply_thresholded_relu(scope, inputs, outputs, container, operator_name=attrs['name'],
+                               alpha=params.thresholdedReLU.alpha)
     else:
-        if activation_type == 'thresholdedReLU':
-            op_type = 'ThresholdedRelu'
-            attrs['alpha'] = params.thresholdedReLU.alpha
-        elif activation_type == 'softsign':
+        if activation_type == 'softsign':
             op_type = 'Softsign'
         elif activation_type == 'softplus':
             op_type = 'Softplus'

diff --git a/onnxutils/onnxconverter_common/onnx_ops.py b/onnxutils/onnxconverter_common/onnx_ops.py
@@ -580,6 +580,21 @@ def apply_sum(scope, input_names, output_name, container, operator_name=None):
 def apply_tanh(scope, input_name, output_name, container, operator_name=None):
     _apply_unary_operation(scope, 'Tanh', input_name, output_name, container, operator_name)
 
+def apply_thresholded_relu(scope, input_name, output_name, container, operator_name=None, alpha=None):
+    if alpha == None:
+        alpha = [1.0]
+
+    name = _create_name_or_use_existing_one(scope, 'ThresholdedRelu', operator_name)
+    attrs = {'name': name, 'alpha': alpha[0]}
+    if container.target_opset < 10:
+    # ThresholdedRelu graduated from an experimental op to a full op in opset 10
+    # onnxruntime maintains support in the ONNX domain for ThresholdedRelu as a contrib op
+        attrs['op_domain'] = "ai.onnx"
+        op_version = 1
+    else:
+        op_version = 10
+    container.add_node('ThresholdedRelu', input_name, output_name, op_version=op_version, **attrs)
+
 def apply_tile(scope, input_name, output_name, container, operator_name=None, repeats=None):
     name = _create_name_or_use_existing_one(scope, 'Tile', operator_name)