diff --git a/x2paddle/decoder/onnx_decoder.py b/x2paddle/decoder/onnx_decoder.py
index 5ecd3c6e7787df71e2fd557423dcc10d5558ed18..ad66a1430b189aaa1ba4b1e59dbfd52bbd11893a 100644
--- a/x2paddle/decoder/onnx_decoder.py
+++ b/x2paddle/decoder/onnx_decoder.py
@@ -181,6 +181,7 @@ class ONNXGraph(Graph):
                     layer,
                     layer_name=layer.name,
                     is_global_input=is_place_holder)
+
         #set data node's weight
         for name, weight in self.graph_weights(self.model):
             if name in self.node_map:
@@ -405,8 +406,7 @@ class ONNXDecoder(object):
         for item in graph.value_info:
             item.name = self.make_variable_name(item.name)
         for node in graph.node:
-            if node.name == '':
-                node.name = node.output[0]
+            node.name = node.output[0]
             node.name = self.make_variable_name(node.name)
             for i in range(len(node.input)):
                 node.input[i] = self.make_variable_name(node.input[i])
diff --git a/x2paddle/op_mapper/onnx_directly_map.py b/x2paddle/op_mapper/onnx_directly_map.py
index 75c2a842f8074761b6a1a4cbacd3248fb8254526..c6edbfba4c124e49d7ff30ed9e9528d1c45639fc 100644
--- a/x2paddle/op_mapper/onnx_directly_map.py
+++ b/x2paddle/op_mapper/onnx_directly_map.py
@@ -45,7 +45,9 @@ default_op_mapping = {
         'reduce_mean', ['X'], ['Out'],
         dict(axes='dim', keepdims='keep_dim'),
         dict(keep_dim=1)
-    ]
+    ],
+    'LeakyRelu': ['leaky_relu', ['X'], ['Out'],
+                  dict(), dict(alpha=.01)]
 }
 
 default_ioa_constraint = {
diff --git a/x2paddle/op_mapper/onnx_op_mapper.py b/x2paddle/op_mapper/onnx_op_mapper.py
index e15ec2191b28d9fce7b3e65965ae16a3561de73d..c1e213258ee3436c26f5d87e6f24fc9b91c92d7a 100644
--- a/x2paddle/op_mapper/onnx_op_mapper.py
+++ b/x2paddle/op_mapper/onnx_op_mapper.py
@@ -36,6 +36,14 @@ def _const_weight_or_none(node):
     return None
 
 
+def get_same_padding(in_size, kernel_size, stride):
+    new_size = int(math.ceil(in_size * 1.0 / stride))
+    pad_size = (new_size - 1) * stride + kernel_size - in_size
+    pad0 = int(pad_size / 2)
+    pad1 = pad_size - pad0
+    return [pad0, pad1]
+
+
 class ONNXOpMapper(OpMapper):
     def __init__(self, decoder):
         super(ONNXOpMapper, self).__init__()
@@ -397,9 +405,8 @@ class ONNXOpMapper(OpMapper):
 
     def AveragePool(self, node):
         val_x = self.graph.get_node(node.layer.input[0], copy=True)
-        assert node.get_attr(
-            'auto_pad',
-            'NOTSET') == 'NOTSET', 'only auto_pad = NOTSET is supported'
+
+        auto_pad = node.get_attr('auto_pad', 'NOTSET')
         kernel_shape = node.get_attr("kernel_shape")
         poolnd = len(kernel_shape)
         strides = node.get_attr("strides")
@@ -408,8 +415,16 @@ class ONNXOpMapper(OpMapper):
         pads = node.get_attr('pads', [0] * (poolnd * 2))
         fluid_op = 'pool{}d'.format(poolnd)
         assert 2 <= poolnd <= 3, 'only pool2d and pool3d is supported'
-
         paddings, val_x = self._pad_if_asymmetric(node, pads, val_x)
+
+        input_shape = val_x.out_shapes
+        if auto_pad == "SAME_UPPER" or auto_pad == "SAME_UPPER":
+            pad_h = get_same_padding(input_shape[2], kernel_shape[0],
+                                     strides[0])
+            pad_w = get_same_padding(input_shape[3], kernel_shape[1],
+                                     strides[1])
+            attr = {"paddings": pad_h + pad_w, "pad_value": 0.0}
+
         attr = {
             "pool_size": kernel_shape,
             "pool_type": string('avg'),
@@ -621,10 +636,8 @@ class ONNXOpMapper(OpMapper):
 
     def MaxPool(self, node):
         val_x = self.graph.get_node(node.layer.input[0], copy=True)
-        assert node.get_attr(
-            'auto_pad', 'NOTSET'
-        ) == 'NOTSET', 'only auto_pad = NOTSET is supported'  # optional
 
+        auto_pad = node.get_attr('auto_pad', 'NOTSET')
         assert node.get_attr(
             "dilations") is None, 'only dilations = 0 is supported'  # optional
 
@@ -637,6 +650,15 @@ class ONNXOpMapper(OpMapper):
         fluid_op = 'pool{}d'.format(poolnd)
         assert 2 <= poolnd <= 3, 'only pool2d and pool3d is supported'
         paddings, val_x = self._pad_if_asymmetric(node, pads, val_x)
+
+        input_shape = val_x.out_shapes
+        if auto_pad == "SAME_UPPER" or auto_pad == "SAME_UPPER":
+            pad_h = get_same_padding(input_shape[2], kernel_shape[0],
+                                     strides[0])
+            pad_w = get_same_padding(input_shape[3], kernel_shape[1],
+                                     strides[1])
+            attr = {"paddings": pad_h + pad_w, "pad_value": 0.0}
+
         attr = {
             "pool_size": kernel_shape,
             "pool_type": string("max"),
diff --git a/x2paddle_model_zoo.md b/x2paddle_model_zoo.md
index 6aaf7499762dca1d44f1050b0b07767de7ca85d9..c0511118b8a7c3fcdfb6f8754b7518d6417cc08d 100644
--- a/x2paddle_model_zoo.md
+++ b/x2paddle_model_zoo.md
@@ -42,17 +42,21 @@
 | Alexnet | [torchvison.model.alexnet](https://github.com/pytorch/vision/blob/master/torchvision/models/alexnet.py) |9|
 | Shufflenet | [onnx official](https://github.com/onnx/models/tree/master/vision/classification/shufflenet) |9|
 | Inception_v2 | [onnx official](https://github.com/onnx/models/tree/master/vision/classification/inception_and_googlenet/inception_v2) |9|
+| Mobilenet_v2 | [pytorch(personal practice)](https://github.com/tonylins/pytorch-mobilenet-v2) |9|
 
-目前onnx2paddle主要支持onnx operator version 9，关于如何使用torchvison的model:
+目前onnx2paddle主要支持onnx operator version 9；
+如何将torchvison或者个人开发者写的pytroch model转换成onnx model:
 ```
 import torch
 import torchvision
 
 #根据不同模型调整输入的shape
 dummy_input = torch.randn(1, 3, 224, 224)
+
+#预训练后的pytorch model
 resnet18 = torchvision.models.resnet18(pretrained=True)
 
-#"resnet18.onnx"为onnx model的存储路径
+#"resnet18.onnx"为onnx model的存储路径，1.1
 torch.onnx.export(resnet18, dummy_input, "resnet18.onnx",verbose=True)
 
 ```