fix get_attr and new tensorflow saver

mmdnn/conversion/common/IR/IR_graph.py

@@ -60,8 +60,13 @@ class IRGraphNode(GraphNode):
             attr = self.layer.attr[name]
             field = attr.WhichOneof('value')
             val = getattr(attr, field) if field else default_value
+            if not val:
+                return val
             if isinstance(val, AttrValue.ListValue):
-                return list(val.ListFields()[0][1])
+                if val.ListFields():
+                    return list(val.ListFields()[0][1])
+                else:
+                    return val.ListFields()
             else:
                 return val.decode('utf-8') if isinstance(val, bytes) else val
         else:

mmdnn/conversion/examples/tensorflow/extractor.py

@@ -174,9 +174,6 @@ class tensorflow_extractor(base_extractor):
 
         init = tf.global_variables_initializer()
         with tf.Session() as sess:
-            # tf.train.export_meta_graph("kit.meta", as_text=True)
-            # writer = tf.summary.FileWriter('./graphs', sess.graph)
-            # writer.close()
             sess.run(init)
             saver = tf.train.Saver()
             saver.restore(sess, path + cls.architecture_map[architecture]['filename'])

mmdnn/conversion/examples/tensorflow/imagenet_test.py

@@ -59,10 +59,31 @@ class TestTF(TestKit):
     def dump(self, path = None):
         if path is None: path = self.args.dump
         with tf.Session() as sess:
-            init = tf.global_variables_initializer()
+            sess.run(tf.global_variables_initializer())
-            sess.run(init)
+
-            saver = tf.train.Saver()
+            builder = tf.saved_model.builder.SavedModelBuilder(path)
-            save_path = saver.save(sess, self.args.dump)
+
+            tensor_info_input = tf.saved_model.utils.build_tensor_info(self.input)
+            tensor_info_output = tf.saved_model.utils.build_tensor_info(self.model)
+
+            prediction_signature = (
+                tf.saved_model.signature_def_utils.build_signature_def(
+                    inputs={'input': tensor_info_input},
+                    outputs={'output': tensor_info_output},
+                    method_name=tf.saved_model.signature_constants.PREDICT_METHOD_NAME
+                )
+            )
+
+            builder.add_meta_graph_and_variables(
+                sess,
+                [tf.saved_model.tag_constants.TRAINING],
+                signature_def_map={
+                    tf.saved_model.signature_constants.DEFAULT_SERVING_SIGNATURE_DEF_KEY: prediction_signature
+                }
+            )
+
+            save_path = builder.save()
+
             print ('Tensorflow file is saved as [{}], generated by [{}.py] and [{}].'.format(
                 save_path, self.args.n, self.args.w))
 

mmdnn/conversion/tensorflow/saver.py

@@ -4,9 +4,30 @@ import tensorflow as tf
 def save_model(MainModel, network_filepath, weight_filepath, dump_filepath):
     input, model = MainModel.KitModel(weight_filepath)
     with tf.Session() as sess:
-        init = tf.global_variables_initializer()
+        sess.run(tf.global_variables_initializer())
-        sess.run(init)
+
-        saver = tf.train.Saver()
+            builder = tf.saved_model.builder.SavedModelBuilder(path)
-        save_path = saver.save(sess, dump_filepath)
+
+            tensor_info_input = tf.saved_model.utils.build_tensor_info(input)
+            tensor_info_output = tf.saved_model.utils.build_tensor_info(model)
+
+            prediction_signature = (
+                tf.saved_model.signature_def_utils.build_signature_def(
+                    inputs={'input': tensor_info_input},
+                    outputs={'output': tensor_info_output},
+                    method_name=tf.saved_model.signature_constants.PREDICT_METHOD_NAME
+                )
+            )
+
+            builder.add_meta_graph_and_variables(
+                sess,
+                [tf.saved_model.tag_constants.TRAINING],
+                signature_def_map={
+                    tf.saved_model.signature_constants.DEFAULT_SERVING_SIGNATURE_DEF_KEY: prediction_signature
+                }
+            )
+
+            save_path = builder.save()
+
         print('Tensorflow file is saved as [{}], generated by [{}.py] and [{}].'.format(
             save_path, network_filepath, weight_filepath))

tests/test_conversion_imagenet.py

@@ -689,24 +689,27 @@ class TestModels(CorrectnessTest):
             os.remove(converted_file + '.npy')
 
 
+
+    # In case of odd number add the extra padding at the end for SAME_UPPER(eg. pads:[0, 2, 2, 0, 0, 3, 3, 0]) and at the beginning for SAME_LOWER(eg. pads:[0, 3, 3, 0, 0, 2, 2, 0])
+
     exception_tabel = {
-        'cntk_Keras_resnet18',                      # Cntk Padding is SAME_UPPER, but Keras Padding is SAME_LOWER, in first convolution layer.
+        'cntk_Keras_resnet18',                      # Cntk Padding is SAME_LOWER, but Keras Padding is SAME_UPPER, in first convolution layer.
-        'cntk_Keras_resnet152',                     # Cntk Padding is SAME_UPPER, but Keras Padding is SAME_LOWER, in first convolution layer.
+        'cntk_Keras_resnet152',                     # Cntk Padding is SAME_LOWER, but Keras Padding is SAME_UPPER, in first convolution layer.
-        'cntk_Tensorflow_resnet18',                 # Cntk Padding is SAME_UPPER, but Keras Padding is SAME_LOWER, in first convolution layer.
+        'cntk_Tensorflow_resnet18',                 # Cntk Padding is SAME_LOWER, but Keras Padding is SAME_UPPER, in first convolution layer.
-        'cntk_Tensorflow_resnet152',                # Cntk Padding is SAME_UPPER, but Keras Padding is SAME_LOWER, in first convolution layer.
+        'cntk_Tensorflow_resnet152',                # Cntk Padding is SAME_LOWER, but Keras Padding is SAME_UPPER, in first convolution layer.
-        'tensorflow_Cntk_inception_v1',             # TODO
+        'tensorflow_Cntk_inception_v1',             # Cntk Padding is SAME_LOWER, but Tensorflow Padding is SAME_UPPER, in first convolution layer.
-        'tensorflow_Cntk_resnet_v1_50',             # TODO
+        'tensorflow_Cntk_resnet_v1_50',             # Cntk Padding is SAME_LOWER, but Tensorflow Padding is SAME_UPPER, in first convolution layer.
-        'tensorflow_Cntk_resnet_v2_50',             # TODO
+        'tensorflow_Cntk_resnet_v2_50',             # Cntk Padding is SAME_LOWER, but Tensorflow Padding is SAME_UPPER, in first convolution layer.
-        'tensorflow_Cntk_resnet_v1_152',            # TODO
+        'tensorflow_Cntk_resnet_v1_152',            # Cntk Padding is SAME_LOWER, but Tensorflow Padding is SAME_UPPER, in first convolution layer.
-        'tensorflow_Cntk_resnet_v2_152',            # TODO
+        'tensorflow_Cntk_resnet_v2_152',            # Cntk Padding is SAME_LOWER, but Tensorflow Padding is SAME_UPPER, in first convolution layer.
-        'tensorflow_Cntk_mobilenet_v1_1.0',         # TODO
+        'tensorflow_Cntk_mobilenet_v1_1.0',         # Cntk Padding is SAME_LOWER, but Tensorflow Padding is SAME_UPPER, in first convolution layer.
-        'tensorflow_frozen_MXNet_inception_v1',     # TODO
+        'tensorflow_frozen_MXNet_inception_v1',     # different after AvgPool. AVG POOL padding difference between these two framework. MXNet AVGPooling Padding is SAME_LOWER, Tensorflow AVGPooling Padding is SAME_UPPER
         'tensorflow_MXNet_inception_v3',            # different after "InceptionV3/InceptionV3/Mixed_5b/Branch_3/AvgPool_0a_3x3/AvgPool". AVG POOL padding difference between these two framework.
         'caffe_Pytorch_inception_v1',               # TODO
         'caffe_Pytorch_alexnet',                    # TODO
         'caffe_Pytorch_inception_v4',               # TODO, same with caffe_Cntk_inception_v4
-        'darknet_Keras_yolov2',                     # TODO,
+        'darknet_Keras_yolov2',                     # accumulation of small difference
-        'darknet_Keras_yolov3',                     # TODO,
+        'darknet_Keras_yolov3',                     # accumulation of small difference
     }