Merge pull request #2 from PaddlePaddle/develop

add

.travis.yml

+language: python
+
+python:
+  - '2.7'
+  - '3.5'
+  - '3.6'
+
+script:
+  - if [[ $TRAVIS_PYTHON_VERSION != 2.7 ]]; then /bin/bash ./tools/check_code_style.sh; fi
+
+notifications:
+  email:
+    on_success: change
+    on_failure: always

FAQ.md

 ## 常见问题
 
 **Q1. TensorFlow模型转换过程中，提示『Unknown shape for input tensor[tensor name: "input"]， Please define shape of input here』？**  
-A：该提示信息表示无法从TensorFlow的pb模型中获取到输入tensor(tensor名为"input:)的shape信息，所以需要用户手动在提示后输入详细的shape信息，如None,224,224,3 其中None表示Batch 
+A：该提示信息表示无法从TensorFlow的pb模型中获取到输入tensor(tensor名为"input:)的shape信息，所以需要用户手动在提示后输入详细的shape信息，如None,224,224,3 其中None表示Batch
 
 
 **Q2. TensorFlow模型转换失败怎么解决?**  

README.md

@@ -14,7 +14,7 @@ paddlepaddle >= 1.5.0
 
 **按需安装以下依赖**  
 tensorflow ： tensorflow == 1.14.0  
-caffe ： caffe == 1.0.0  
+caffe ： 无  
 onnx ： onnx == 1.5.0  pytorch == 1.1.0
 
 ## 安装

add_caffe_custom_layer.md

 ## 如何转换Caffe自定义Layer
 
-本文档介绍如何将Caffe自定义Layer转换为PaddlePaddle模型中的对应实现, 用户可根据自己需要，添加代码实现自定义层，从而支持模型的完整转换。            
-***步骤一 下载代码***        
+本文档介绍如何将Caffe自定义Layer转换为PaddlePaddle模型中的对应实现, 用户可根据自己需要，添加代码实现自定义层，从而支持模型的完整转换。  
+***步骤一 下载代码***  
 此处涉及修改源码，应先卸载x2paddle，并且下载源码，主要有以下两步完成：
 ```
 pip uninstall x2paddle
 pip install git+https://github.com/PaddlePaddle/X2Paddle.git@develop
 ```
 
-***步骤二 编译caffe.proto***   
+***步骤二 编译caffe.proto***  
 该步骤依赖protobuf编译器，其安装过程有以下两种方式：
-> 选择一：pip install protobuf             
+> 选择一：pip install protobuf  
 > 选择二：使用[官方源码](https://github.com/protocolbuffers/protobuf)进行编译
 
-使用脚本./tools/compile.sh将caffe.proto（包含所需的自定义Layer信息）编译成我们所需的目标语言（Python）         
+使用脚本./tools/compile.sh将caffe.proto（包含所需的自定义Layer信息）编译成我们所需的目标语言（Python）  
 使用方式：
 ```
 bash ./toos/compile.sh /home/root/caffe/src/caffe/proto
@@ -25,14 +25,14 @@ bash ./toos/compile.sh /home/root/caffe/src/caffe/proto
 - 仿照./x2paddle/op_mapper/caffe_custom_layer中的其他文件，在mylayer.py中主要需要实现3个函数，下面以roipooling.py为例分析代码：
   1. `def roipooling_shape(input_shape, pooled_w=None, pooled_h=None)`  
      参数：
-     1. input_shape（list）：其中每个元素代表该层每个输入数据的shape，为必须传入的参数   
+     1. input_shape（list）：其中每个元素代表该层每个输入数据的shape，为必须传入的参数  
      2. pooled_w（int）：代表ROI Pooling的kernel的宽，其命名与.prototxt中roi_pooling_param中的key一致
-     3. pooled_h（int）：代表ROI Pooling的kernel的高，其命名与.prototxt中roi_pooling_param中的key一致          
+     3. pooled_h（int）：代表ROI Pooling的kernel的高，其命名与.prototxt中roi_pooling_param中的key一致  
 
-     功能：计算出进行ROI Pooling后的shape         
-     返回：一个list，其中每个元素代表每个输出数据的shape，由于ROI Pooling的输出数据只有一个，所以其list长度为1     
+     功能：计算出进行ROI Pooling后的shape  
+     返回：一个list，其中每个元素代表每个输出数据的shape，由于ROI Pooling的输出数据只有一个，所以其list长度为1  
 
-  2. `def roipooling_layer(inputs, input_shape=None, name=None, pooled_w=None, pooled_h=None, spatial_scale=None)` 
+  2. `def roipooling_layer(inputs, input_shape=None, name=None, pooled_w=None, pooled_h=None, spatial_scale=None)`
 
      参数：
      1. inputs（list）：其中每个元素代表该层每个输入数据，为必须传入的参数
@@ -40,9 +40,9 @@ bash ./toos/compile.sh /home/root/caffe/src/caffe/proto
      3. name（str）：ROI Pooling层的名字，为必须传入的参数  
      4. pooled_w（int）：代表ROI Pooling的kernel的宽，其命名与.prototxt中roi_pooling_param中的key一致
      5. pooled_h（int）：代表ROI Pooling的kernel的高，其命名与.prototxt中roi_pooling_param中的key一致
-     6. spatial_scale（float）：用于将ROI坐标从输入比例转换为池化时使用的比例，其命名与.prototxt中roi_pooling_param中的key一致            
+     6. spatial_scale（float）：用于将ROI坐标从输入比例转换为池化时使用的比例，其命名与.prototxt中roi_pooling_param中的key一致  
 
-     功能：运用PaddlePaddle完成组网来实现`roipooling_layer`的功能         
+     功能：运用PaddlePaddle完成组网来实现`roipooling_layer`的功能  
      返回：一个Variable，为组网后的结果
 
   3. `def roipooling_weights(name, data=None)`  
@@ -51,7 +51,7 @@ bash ./toos/compile.sh /home/root/caffe/src/caffe/proto
      1. name（str）：ROI Pooling层的名字，为必须传入的参数
      2. data（list）：由Caffe模型.caffemodel获得的关于roipooling的参数，roipooling的参数为None
 
-     功能：为每个参数（例如kernel、bias等）命名；同时，若Caffe中该层参数与PaddlePaddle中参数的格式不一致，则变换操作也在该函数中实现。     
+     功能：为每个参数（例如kernel、bias等）命名；同时，若Caffe中该层参数与PaddlePaddle中参数的格式不一致，则变换操作也在该函数中实现。  
      返回：一个list，包含每个参数的名字。
 
 - 在roipooling.py中注册`roipooling`，主要运用下述代码实现：
@@ -70,9 +70,9 @@ bash ./toos/compile.sh /home/root/caffe/src/caffe/proto
 # 在X2Paddle目录下安装x2paddle
 python setup.py install
 # 运行转换代码
-x2paddle --framework=caffe 
-         --prototxt=deploy.proto 
-         --weight=deploy.caffemodel 
-         --save_dir=pd_model 
+x2paddle --framework=caffe
+         --prototxt=deploy.proto
+         --weight=deploy.caffemodel
+         --save_dir=pd_model
          --caffe_proto=/home/root/caffe/src/caffe/proto/caffe_pb2.py
 ```

caffe_custom_layer.md

@@ -10,5 +10,3 @@
 | Normalize | [code](https://github.com/weiliu89/caffe/blob/ssd/src/caffe/layers/normalize_layer.cpp) |
 | ROIPooling | [code](https://github.com/rbgirshick/caffe-fast-rcnn/blob/0dcd397b29507b8314e252e850518c5695efbb83/src/caffe/layers/roi_pooling_layer.cpp) |
 | Axpy | [code](https://github.com/hujie-frank/SENet/blob/master/src/caffe/layers/axpy_layer.cpp) |
-
-

export_tf_model.md

@@ -28,7 +28,7 @@ def freeze_model(sess, output_tensor_names, freeze_model_path):
         f.write(out_graph.SerializeToString())
 
     print("freeze model saved in {}".format(freeze_model_path))
-    
+
 # 加载模型参数
 sess = tf.Session()
 inputs = tf.placeholder(dtype=tf.float32,

requirements.txt

+pre-commit
+yapf == 0.28.0

tools/check_code_style.sh

+#!/bin/bash
+function abort(){
+    echo "Your change doesn't follow X2Paddle's code style." 1>&2
+    echo "Please use pre-commit to check what is wrong." 1>&2
+    exit 1
+}
+
+trap 'abort' 0
+set -e
+cd $TRAVIS_BUILD_DIR
+export PATH=/usr/bin:$PATH
+pre-commit install
+
+if ! pre-commit run -a ; then
+    git diff
+    exit 1
+fi
+
+trap : 0

tools/check_for_lite.py

-import urllib
+from six.moves import urllib
 import sys
 from paddle.fluid.framework import Program
 

x2paddle/__init__.py

-__version__ = "0.4.5"
+__version__ = "0.5.0"

x2paddle/convert.py

@@ -106,6 +106,8 @@ def tf2paddle(model_path,
         # optimizer below is experimental
         optimizer.merge_activation()
         optimizer.merge_bias()
+        optimizer.merge_batch_norm()
+        optimizer.merge_prelu()
     else:
         mapper = TFOpMapperNHWC(model)
         optimizer = TFOptimizer(mapper)
@@ -177,6 +179,9 @@ def main():
             x2paddle.__version__))
         return
 
+    assert args.framework is not None, "--framework is not defined(support tensorflow/caffe/onnx)"
+    assert args.save_dir is not None, "--save_dir is not defined"
+
     try:
         import paddle
         v0, v1, v2 = paddle.__version__.split('.')
@@ -185,8 +190,6 @@ def main():
             return
     except:
         print("paddlepaddle not installed, use \"pip install paddlepaddle\"")
-    assert args.framework is not None, "--framework is not defined(support tensorflow/caffe/onnx)"
-    assert args.save_dir is not None, "--save_dir is not defined"
 
     if args.framework == "tensorflow":
         assert args.model is not None, "--model should be defined while translating tensorflow model"

x2paddle/core/fluid_code.py

@@ -13,8 +13,9 @@
 # limitations under the License.
 
 from x2paddle.core.graph import GraphNode
-import collections
 from x2paddle.core.util import *
+import collections
+import six
 
 
 class Layer(object):
@@ -28,7 +29,7 @@ class Layer(object):
     def get_code(self):
         layer_code = ""
         if self.output is not None:
-            if isinstance(self.output, str):
+            if isinstance(self.output, six.string_types):
                 layer_code = self.output + " = "
             else:
                 layer_code = self.output.layer_name + " = "
@@ -47,7 +48,7 @@ class Layer(object):
                                     "[{}]".format(input.index) + ", ")
                     else:
                         in_list += (input.layer_name + ", ")
-                elif isinstance(input, str):
+                elif isinstance(input, six.string_types):
                     in_list += (input + ", ")
                 else:
                     raise Exception(
@@ -72,7 +73,7 @@ class Layer(object):
                                "[{}]".format(self.inputs.index) + ", ")
             else:
                 layer_code += (self.inputs.layer_name + ", ")
-        elif isinstance(self.inputs, str):
+        elif isinstance(self.inputs, six.string_types):
             layer_code += (self.inputs + ", ")
         else:
             raise Exception("Unknown type of inputs.")
@@ -119,6 +120,6 @@ class FluidCode(object):
         for layer in self.layers:
             if isinstance(layer, Layer):
                 codes.append(layer.get_code())
-            elif isinstance(layer, str):
+            elif isinstance(layer, six.string_types):
                 codes.append(layer)
         return codes

x2paddle/core/graph.py

@@ -12,6 +12,8 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
+from __future__ import print_function
+from __future__ import division
 import collections
 import copy as cp
 
@@ -98,8 +100,3 @@ class Graph(object):
             raise Exception("node[{}] not in graph".format(dst))
         self.node_map[dst].inputs.append(src)
         self.node_map[src].outputs.append(dst)
-
-    def print(self):
-        for i, tmp in enumerate(self.topo_sort):
-            print(tmp, self.node_map[tmp].layer_type, self.node_map[tmp].inputs,
-                  self.node_map[tmp].outputs)

x2paddle/core/op_mapper.py

@@ -11,6 +11,7 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
+import paddle.fluid as fluid
 from paddle.fluid.proto import framework_pb2
 from x2paddle.core.util import *
 import inspect
@@ -46,6 +47,28 @@ def export_paddle_param(param, param_name, dir):
     fp.close()
 
 
+# This func will copy to generate code file
+def run_net(param_dir="./"):
+    import os
+    inputs, outputs = x2paddle_net()
+    for i, out in enumerate(outputs):
+        if isinstance(out, list):
+            for out_part in out:
+                outputs.append(out_part)
+            del outputs[i]
+    exe = fluid.Executor(fluid.CPUPlace())
+    exe.run(fluid.default_startup_program())
+
+    def if_exist(var):
+        b = os.path.exists(os.path.join(param_dir, var.name))
+        return b
+
+    fluid.io.load_vars(exe,
+                       param_dir,
+                       fluid.default_main_program(),
+                       predicate=if_exist)
+
+
 class OpMapper(object):
     def __init__(self):
         self.paddle_codes = ""

x2paddle/core/util.py

@@ -11,8 +11,6 @@
 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 # See the License for the specific language governing permissions and
 # limitations under the License.
-
-import paddle.fluid as fluid
 import numpy
 import math
 import os
@@ -20,25 +18,3 @@ import os
 
 def string(param):
     return "\'{}\'".format(param)
-
-
-# This func will copy to generate code file
-def run_net(param_dir="./"):
-    import os
-    inputs, outputs = x2paddle_net()
-    for i, out in enumerate(outputs):
-        if isinstance(out, list):
-            for out_part in out:
-                outputs.append(out_part)
-            del outputs[i]
-    exe = fluid.Executor(fluid.CPUPlace())
-    exe.run(fluid.default_startup_program())
-
-    def if_exist(var):
-        b = os.path.exists(os.path.join(param_dir, var.name))
-        return b
-
-    fluid.io.load_vars(exe,
-                       param_dir,
-                       fluid.default_main_program(),
-                       predicate=if_exist)

x2paddle/decoder/caffe_decoder.py

@@ -236,11 +236,7 @@ class CaffeDecoder(object):
         data.MergeFromString(open(self.model_path, 'rb').read())
         pair = lambda layer: (layer.name, self.normalize_pb_data(layer))
         layers = data.layers or data.layer
-        import time
-        start = time.time()
         self.params = [pair(layer) for layer in layers if layer.blobs]
-        end = time.time()
-        print('cost:', str(end - start))
 
     def normalize_pb_data(self, layer):
         transformed = []

x2paddle/op_mapper/caffe_op_mapper.py

@@ -135,7 +135,8 @@ class CaffeOpMapper(OpMapper):
         if isinstance(params.kernel_size, numbers.Number):
             [k_h, k_w] = [params.kernel_size] * 2
         elif len(params.kernel_size) > 0:
-            k_h = params.kernel_h if params.kernel_h > 0 else params.kernel_size[0]
+            k_h = params.kernel_h if params.kernel_h > 0 else params.kernel_size[
+                0]
             k_w = params.kernel_w if params.kernel_w > 0 else params.kernel_size[
                 len(params.kernel_size) - 1]
         elif params.kernel_h > 0 or params.kernel_w > 0:
@@ -156,8 +157,8 @@ class CaffeOpMapper(OpMapper):
             [p_h, p_w] = [params.pad] * 2
         elif len(params.pad) > 0:
             p_h = params.pad_h if params.pad_h > 0 else params.pad[0]
-            p_w = params.pad_w if params.pad_w > 0 else params.pad[len(params.pad) -
-                                                                  1]
+            p_w = params.pad_w if params.pad_w > 0 else params.pad[
+                len(params.pad) - 1]
         elif params.pad_h > 0 or params.pad_w > 0:
             p_h = params.pad_h
             p_w = params.pad_w
@@ -225,12 +226,17 @@ class CaffeOpMapper(OpMapper):
             node.layer_type, params)
         if data is None:
             data = []
-            print('The parameter of {} (type is {}) is not set. So we set the parameters as 0'.format(
-            node.layer_name, node.layer_type))
+            print(
+                'The parameter of {} (type is {}) is not set. So we set the parameters as 0'
+                .format(node.layer_name, node.layer_type))
             input_c = node.input_shape[0][1]
             output_c = channel
-            data.append(np.zeros([output_c, input_c, kernel[0], kernel[1]]).astype('float32'))
-            data.append(np.zeros([output_c,])).astype('float32')
+            data.append(
+                np.zeros([output_c, input_c, kernel[0],
+                          kernel[1]]).astype('float32'))
+            data.append(np.zeros([
+                output_c,
+            ])).astype('float32')
         else:
             data = self.adjust_parameters(node)
         self.weights[node.layer_name + '_weights'] = data[0]
@@ -272,12 +278,17 @@ class CaffeOpMapper(OpMapper):
             node.layer_type, params)
         if data is None:
             data = []
-            print('The parameter of {} (type is {}) is not set. So we set the parameters as 0'.format(
-            node.layer_name, node.layer_type))
+            print(
+                'The parameter of {} (type is {}) is not set. So we set the parameters as 0'
+                .format(node.layer_name, node.layer_type))
             input_c = node.input_shape[0][1]
             output_c = channel
-            data.append(np.zeros([output_c, input_c, kernel[0], kernel[1]]).astype('float32'))
-            data.append(np.zeros([output_c,]).astype('float32'))
+            data.append(
+                np.zeros([output_c, input_c, kernel[0],
+                          kernel[1]]).astype('float32'))
+            data.append(np.zeros([
+                output_c,
+            ]).astype('float32'))
         else:
             data = self.adjust_parameters(node)
         self.weights[node.layer_name + '_weights'] = data[0]
@@ -369,13 +380,17 @@ class CaffeOpMapper(OpMapper):
         data = node.data
         params = node.layer.inner_product_param
         if data is None:
-            print('The parameter of {} (type is {}) is not set. So we set the parameters as 0.'.format(
-            node.layer_name, node.layer_type))
+            print(
+                'The parameter of {} (type is {}) is not set. So we set the parameters as 0.'
+                .format(node.layer_name, node.layer_type))
             input_c = node.input_shape[0][1]
             output_c = params.num_output
             data = []
-            data.append(np.zeros([input_c, output_c]).astype('float32').astype('float32'))
-            data.append(np.zeros([output_c]).astype('float32').astype('float32'))
+            data.append(
+                np.zeros([input_c,
+                          output_c]).astype('float32').astype('float32'))
+            data.append(
+                np.zeros([output_c]).astype('float32').astype('float32'))
         else:
             data = self.adjust_parameters(node)
             # Reshape the parameters to Paddle's ordering
@@ -467,7 +482,7 @@ class CaffeOpMapper(OpMapper):
                 node.layer_name, node.layer_name + '_' + str(i)))
             if i == len(points) - 2:
                 break
-                
+
     def Concat(self, node):
         assert len(
             node.inputs
@@ -616,7 +631,8 @@ class CaffeOpMapper(OpMapper):
                                       param_attr=attr)
 
     def BatchNorm(self, node):
-        assert len(node.inputs) == 1, 'The count of BatchNorm node\'s input is not 1.'
+        assert len(
+            node.inputs) == 1, 'The count of BatchNorm node\'s input is not 1.'
         input = self.graph.get_bottom_node(node, idx=0, copy=True)
         params = node.layer.batch_norm_param
         if hasattr(params, 'eps'):
@@ -624,11 +640,16 @@ class CaffeOpMapper(OpMapper):
         else:
             eps = 1e-5
         if node.data is None or len(node.data) != 3:
-            print('The parameter of {} (type is {}) is not set. So we set the parameters as 0'.format(
-            node.layer_name, node.layer_type))
+            print(
+                'The parameter of {} (type is {}) is not set. So we set the parameters as 0'
+                .format(node.layer_name, node.layer_type))
             input_c = node.input_shape[0][1]
-            mean = np.zeros([input_c,]).astype('float32')
-            variance = np.zeros([input_c,]).astype('float32')
+            mean = np.zeros([
+                input_c,
+            ]).astype('float32')
+            variance = np.zeros([
+                input_c,
+            ]).astype('float32')
             scale = 0
         else:
             node.data = [np.squeeze(i) for i in node.data]
@@ -655,11 +676,16 @@ class CaffeOpMapper(OpMapper):
 
     def Scale(self, node):
         if node.data is None:
-            print('The parameter of {} (type is {}) is not set. So we set the parameters as 0'.format(
-            node.layer_name, node.layer_type))
+            print(
+                'The parameter of {} (type is {}) is not set. So we set the parameters as 0'
+                .format(node.layer_name, node.layer_type))
             input_c = node.input_shape[0][1]
-            self.weights[node.layer_name + '_scale'] = np.zeros([input_c,]).astype('float32')
-            self.weights[node.layer_name + '_offset'] = np.zeros([input_c,]).astype('float32')
+            self.weights[node.layer_name + '_scale'] = np.zeros([
+                input_c,
+            ]).astype('float32')
+            self.weights[node.layer_name + '_offset'] = np.zeros([
+                input_c,
+            ]).astype('float32')
         else:
             self.weights[node.layer_name + '_scale'] = np.squeeze(node.data[0])
             self.weights[node.layer_name + '_offset'] = np.squeeze(node.data[1])

x2paddle/op_mapper/caffe_shape.py

@@ -43,7 +43,8 @@ def get_kernel_parameters(params):
         [p_h, p_w] = [params.pad] * 2
     elif len(params.pad) > 0:
         p_h = params.pad_h if params.pad_h > 0 else params.pad[0]
-        p_w = params.pad_w if params.pad_w > 0 else params.pad[len(params.pad) - 1]
+        p_w = params.pad_w if params.pad_w > 0 else params.pad[len(params.pad) -
+                                                               1]
     elif params.pad_h > 0 or params.pad_w > 0:
         p_h = params.pad_h
         p_w = params.pad_w

x2paddle/op_mapper/onnx_op_mapper.py

@@ -94,7 +94,7 @@ class ONNXOpMapper(OpMapper):
                 print(op)
             return False
 
-    def directly_map(self, node, *args, name='', **kwargs):
+    def directly_map(self, node, name='', *args, **kwargs):
         inputs = node.layer.input
         outputs = node.layer.output
         op_type = node.layer_type

x2paddle/op_mapper/tf_op_mapper.py

@@ -24,6 +24,8 @@ import sys
 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
+    if pad_size < 0:
+        pad_size = 0
     pad0 = int(pad_size / 2)
     pad1 = pad_size - pad0
     return [pad0, pad1]
@@ -369,12 +371,13 @@ class TFOpMapper(OpMapper):
             pad_w = get_same_padding(in_shape[3], k_size[3], strides[3])
             pad_h = pad_h[0] + pad_h[1]
             pad_w = pad_w[0] + pad_w[1]
-            attr = {"paddings": [0, pad_h, 0, pad_w], "pad_value": -10000.0}
-            node.fluid_code.add_layer("pad2d",
-                                      inputs=input,
-                                      output=node,
-                                      param_attr=attr)
-            input = node
+            if pad_h != 0 or pad_w != 0:
+                attr = {"paddings": [0, pad_h, 0, pad_w], "pad_value": -10000.0}
+                node.fluid_code.add_layer("pad2d",
+                                          inputs=input,
+                                          output=node,
+                                          param_attr=attr)
+                input = node
         attr = {
             "pool_size": k_size[2:4],
             "pool_type": string("max"),
@@ -551,6 +554,7 @@ class TFOpMapper(OpMapper):
     def Reshape(self, node):
         input = self.graph.get_node(node.layer.input[0], copy=True)
         param = self.graph.get_node(node.layer.input[1], copy=True)
+        is_variable = False
         if param.layer_type == "Const":
             attr = {"shape": param.value.tolist()}
             self.add_omit_nodes(param.layer_name, node.layer_name)
@@ -582,6 +586,24 @@ class TFOpMapper(OpMapper):
                     new_param += (node.layer_name + "[{}]".format(i) + ", ")
                 new_param = new_param.strip(", ") + "]"
                 attr = {"shape": new_param}
+                is_variable = True
+
+        # to change [192, -1]->[-1, 192], allways put -1 in the first dimension
+        # optimization for Paddle-Lite
+        in_shape = input.out_shapes[0]
+        if is_variable and in_shape.count(-1) < 1:
+            total_size = 1
+            for i in range(len(in_shape)):
+                total_size *= in_shape[i]
+            for i in range(len(attr["shape"])):
+                if attr["shape"][i] == 0:
+                    attr["shape"][i] = in_shape[i]
+                if attr["shape"][i] != -1:
+                    total_size /= attr["shape"][i]
+            if attr["shape"].count(-1) > 0:
+                index = attr["shape"].index(-1)
+                attr["shape"][index] = int(total_size)
+                attr["shape"][0] = -1
 
         if len(input.out_shapes[0]) == 4 and node.tf_data_format == "NHWC":
             if len(attr["shape"]) < 3:

x2paddle/op_mapper/tf_op_mapper_nhwc.py

@@ -24,6 +24,8 @@ import sys
 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
+    if pad_size < 0:
+        pad_size = 0
     pad0 = int(pad_size / 2)
     pad1 = pad_size - pad0
     return [pad0, pad1]
@@ -500,6 +502,7 @@ class TFOpMapperNHWC(OpMapper):
     def Reshape(self, node):
         input = self.graph.get_node(node.layer.input[0], copy=True)
         param = self.graph.get_node(node.layer.input[1], copy=True)
+        is_variable = False
         if param.layer_type == "Const":
             attr = {"shape": param.value.tolist()}
             self.add_omit_nodes(param.layer_name, node.layer_name)
@@ -527,6 +530,24 @@ class TFOpMapperNHWC(OpMapper):
                     new_param += (node.layer_name + "[{}]".format(i) + ", ")
                 new_param = new_param.strip(", ") + "]"
                 attr = {"shape": new_param}
+                is_variable = True
+        # to change [192, -1]->[-1, 192], allways put -1 in the first dimension
+        # optimization for Paddle-Lite
+        in_shape = input.out_shapes[0]
+        if not is_variable and in_shape.count(-1) < 1:
+            total_size = 1
+            for i in range(len(in_shape)):
+                total_size *= in_shape[i]
+            for i in range(len(attr["shape"])):
+                if attr["shape"][i] == 0:
+                    attr["shape"][i] = in_shape[i]
+                if attr["shape"][i] != -1:
+                    total_size /= attr["shape"][i]
+            if attr["shape"].count(-1) > 0:
+                index = attr["shape"].index(-1)
+                attr["shape"][index] = int(total_size)
+                attr["shape"][0] = -1
+
         node.fluid_code.add_layer("reshape",
                                   inputs=input,
                                   output=node,

x2paddle/optimizer/tf_optimizer.py

@@ -16,6 +16,7 @@
 from x2paddle.op_mapper.tf_op_mapper import TFOpMapper
 from x2paddle.core.fluid_code import Layer
 from x2paddle.core.util import *
+import numpy
 import copy as cp
 
 
@@ -351,3 +352,311 @@ class TFOptimizer(object):
                 if node.fluid_code.layers[-1].op == "transpose":
                     node.fluid_code.layers[-2].output = name
                     del node.fluid_code.layers[-1]
+
+    def merge_batch_norm(self):
+        for i, name in enumerate(self.graph.topo_sort):
+            node = self.graph.get_node(name)
+            if node is None:
+                continue
+            is_batch_norm = True
+            if node.layer_type == "Add":
+                in_nodes0 = [
+                    self.graph.get_node(in_name) for in_name in node.inputs
+                ]
+                if in_nodes0[0].layer_type != "Mul" or in_nodes0[
+                        1].layer_type != "Sub":
+                    is_batch_norm = False
+                    continue
+
+                in_nodes1 = [
+                    self.graph.get_node(in_name)
+                    for in_name in in_nodes0[0].inputs
+                ]
+                in_nodes2 = [
+                    self.graph.get_node(in_name)
+                    for in_name in in_nodes0[1].inputs
+                ]
+                if len(in_nodes1[0].out_shapes[0]) != 4:
+                    is_batch_norm = False
+                    continue
+                if in_nodes1[1].layer_type != "Mul":
+                    is_batch_norm = False
+                    continue
+
+                if in_nodes2[0].layer_type != "Const" or in_nodes2[
+                        1].layer_type != "Mul":
+                    is_batch_norm = False
+                    continue
+
+                in_nodes3 = [
+                    self.graph.get_node(in_name)
+                    for in_name in in_nodes1[1].inputs
+                ]
+                if in_nodes3[0].layer_type != "Rsqrt" or in_nodes3[
+                        1].layer_type != "Const":
+                    is_batch_norm = False
+                    continue
+
+                in_nodes4 = [
+                    self.graph.get_node(in_name)
+                    for in_name in in_nodes2[1].inputs
+                ]
+                if in_nodes4[0].layer_type != "Const" or in_nodes4[
+                        1].layer_name != in_nodes1[1].layer_name:
+                    is_batch_norm = False
+                    continue
+
+                in_nodes5 = self.graph.get_node(in_nodes3[0].inputs[0])
+                if in_nodes5.layer_type != "Add":
+                    is_batch_norm = False
+                    continue
+
+                in_nodes6 = [
+                    self.graph.get_node(in_name) for in_name in in_nodes5.inputs
+                ]
+                if in_nodes6[0].layer_type != "Const" or in_nodes6[
+                        1].layer_type != "Const":
+                    is_batch_norm = False
+                    continue
+
+                if len(in_nodes0[0].outputs) != 1:
+                    is_batch_norm = False
+                    continue
+                if len(in_nodes0[1].outputs) != 1:
+                    is_batch_norm = False
+                    continue
+                if len(in_nodes1[1].outputs) != 2:
+                    is_batch_norm = False
+                    continue
+                if len(in_nodes2[0].outputs) != 1:
+                    is_batch_norm = False
+                    continue
+                if len(in_nodes2[1].outputs) != 1:
+                    is_batch_norm = False
+                    continue
+                if len(in_nodes3[0].outputs) != 1:
+                    is_batch_norm = False
+                    continue
+                if len(in_nodes3[1].outputs) != 1:
+                    is_batch_norm = False
+                    continue
+                if len(in_nodes4[0].outputs) != 1:
+                    is_batch_norm = False
+                    continue
+                if len(in_nodes5.outputs) != 1:
+                    is_batch_norm = False
+                    continue
+                if len(in_nodes6[0].outputs) != 1:
+                    is_batch_norm = False
+                    continue
+                if len(in_nodes6[1].outputs) != 1:
+                    is_batch_norm = False
+                    continue
+
+                conv_shape = in_nodes1[0].out_shapes[0]
+                if conv_shape[3] < 0:
+                    is_batch_norm = False
+                    continue
+
+                # moving_variance
+                if in_nodes6[0].value.size != conv_shape[3]:
+                    is_batch_norm = False
+                    continue
+
+                # epsilon
+                if in_nodes6[1].value.size != 1:
+                    is_batch_norm = False
+                    continue
+
+                # gamma
+                if in_nodes3[1].value.size != conv_shape[3]:
+                    is_batch_norm = False
+                    continue
+
+                # moving_mean
+                if in_nodes4[0].value.size != conv_shape[3]:
+                    is_batch_norm = False
+                    continue
+
+                # beta
+                if in_nodes2[0].value.size != conv_shape[3]:
+                    is_batch_norm = False
+                    continue
+
+                if is_batch_norm:
+                    index = in_nodes1[0].outputs.index(in_nodes0[0].layer_name)
+                    del in_nodes1[0].outputs[index]
+                    node.layer_type = "FusedBatchNorm"
+                    node.inputs = [in_nodes1[0].layer_name]
+                    node.outputs = node.outputs
+                    act = node.fluid_code.layers[-1].param_attr.get("act", None)
+                    node.fluid_code.clear()
+                    attr = {
+                        "epsilon": in_nodes6[1].value,
+                        "param_attr": string(in_nodes3[1].layer_name),
+                        "bias_attr": string(in_nodes2[0].layer_name),
+                        "moving_mean_name": string(in_nodes4[0].layer_name),
+                        "moving_variance_name": string(in_nodes6[0].layer_name),
+                        "is_test": True,
+                        "act": act
+                    }
+
+                    node.fluid_code.add_layer(
+                        "batch_norm",
+                        inputs=in_nodes1[0].fluid_code.layers[-1].output,
+                        output=node,
+                        param_attr=attr)
+
+                del self.graph.node_map[in_nodes0[0].layer_name]
+                del self.graph.node_map[in_nodes0[1].layer_name]
+                del self.graph.node_map[in_nodes1[1].layer_name]
+                del self.graph.node_map[in_nodes2[1].layer_name]
+                del self.graph.node_map[in_nodes3[0].layer_name]
+                del self.graph.node_map[in_nodes4[0].layer_name]
+                del self.graph.node_map[in_nodes5.layer_name]
+
+    def merge_prelu(self):
+        for i, name in enumerate(self.graph.topo_sort):
+            node = self.graph.get_node(name)
+            if node is None:
+                continue
+            is_prelu = True
+            if node.layer_type == "Add":
+                in_nodes0 = [
+                    self.graph.get_node(in_name) for in_name in node.inputs
+                ]
+                if in_nodes0[0].layer_type != "Relu" or in_nodes0[
+                        1].layer_type != "Mul":
+                    is_prelu = False
+                    continue
+                if len(in_nodes0[0].outputs) != 1 or len(
+                        in_nodes0[1].outputs) != 1:
+                    is_prelu = False
+                    continue
+
+                in_nodes1 = self.graph.get_node(in_nodes0[0].inputs[0])
+                in_nodes2 = [
+                    self.graph.get_node(in_name)
+                    for in_name in in_nodes0[1].inputs
+                ]
+                if in_nodes2[1].layer_type != "Const" or numpy.fabs(
+                        in_nodes2[1].value - 0.5) > 1e-06:
+                    is_prelu = False
+                    continue
+                if in_nodes2[0].layer_type != "Mul":
+                    is_prelu = False
+                    continue
+                if len(in_nodes2[1].outputs) != 1 or len(
+                        in_nodes2[0].outputs) != 1:
+                    is_prelu = False
+                    continue
+
+                in_nodes3 = [
+                    self.graph.get_node(in_name)
+                    for in_name in in_nodes2[0].inputs
+                ]
+                if in_nodes3[0].layer_type != "Const" or in_nodes3[
+                        1].layer_type != "Sub":
+                    is_prelu = False
+                    continue
+                if len(in_nodes3[0].outputs) != 1 or len(
+                        in_nodes3[1].outputs) != 1:
+                    is_prelu = False
+                    continue
+
+                in_nodes4 = [
+                    self.graph.get_node(in_name)
+                    for in_name in in_nodes3[1].inputs
+                ]
+                if in_nodes4[0].layer_name != in_nodes1.layer_name or in_nodes4[
+                        1].layer_type != "Abs":
+                    is_prelu = False
+                    continue
+                if len(in_nodes4[1].outputs) != 1:
+                    is_prelu = False
+                    continue
+
+                in_nodes5 = self.graph.get_node(in_nodes4[1].inputs[0])
+                if in_nodes5.layer_name != in_nodes1.layer_name:
+                    is_prelu = False
+                    continue
+
+                if len(in_nodes0[0].outputs) != 1:
+                    is_prelu = false
+                    continue
+                if len(in_nodes0[1].outputs) != 1:
+                    is_prelu = False
+                    continue
+                if len(in_nodes1.outputs) < 3:
+                    is_prelu = False
+                    continue
+                if len(in_nodes2[0].outputs) != 1:
+                    is_prelu = false
+                    continue
+                if len(in_nodes2[1].outputs) != 1:
+                    is_prelu = False
+                    continue
+                if len(in_nodes3[0].outputs) != 1:
+                    is_prelu = False
+                    continue
+                if len(in_nodes3[1].outputs) != 1:
+                    is_prelu = false
+                    continue
+                if len(in_nodes4[1].outputs) != 1:
+                    is_prelu = False
+                    continue
+
+                mode = None
+                in_shape = in_nodes1.out_shapes[0]
+                if in_shape == list(in_nodes3[0].value.shape):
+                    mode = "element"
+                elif len(in_nodes3[0].value.shape) == 0:
+                    mode = "all"
+                elif len(in_nodes3[0].value.shape
+                         ) == 1 and in_nodes3[0].value.shape[0] == 1:
+                    mode = "all"
+                elif len(in_shape) == 4 and len(
+                        in_nodes3[0].value.shape
+                ) == 1 and in_nodes3[0].value.shape[0] == in_shape[-1]:
+                    mode = "channel"
+                    weight = self.op_mapper.weights[in_nodes3[0].layer_name]
+                    weight = numpy.expand_dims(weight, 0)
+                    weight = numpy.expand_dims(weight, 2)
+                    weight = numpy.expand_dims(weight, 3)
+                    self.op_mapper.weights[in_nodes3[0].layer_name] = weight
+                    in_nodes3[0].fluid_code.layers[0].param_attr["shape"] = [
+                        1, in_shape[-1], 1, 1
+                    ]
+                else:
+                    is_prelu = False
+                    continue
+
+                if is_prelu:
+                    index = in_nodes1.outputs.index(in_nodes0[0].layer_name)
+                    del in_nodes1.outputs[index]
+                    index = in_nodes1.outputs.index(in_nodes3[1].layer_name)
+                    del in_nodes1.outputs[index]
+                    index = in_nodes1.outputs.index(in_nodes4[1].layer_name)
+                    del in_nodes1.outputs[index]
+
+                    node.layer_type = "Prelu"
+                    node.inputs = [in_nodes1.layer_name]
+                    node.outputs = node.outputs
+                    act = node.fluid_code.layers[-1].param_attr.get("act", None)
+                    node.fluid_code.clear()
+                    attr = {
+                        "mode": string(mode),
+                        "param_attr": string(in_nodes3[0].layer_name)
+                    }
+
+                    node.fluid_code.add_layer(
+                        "prelu",
+                        inputs=in_nodes1.fluid_code.layers[-1].output,
+                        output=node,
+                        param_attr=attr)
+                del self.graph.node_map[in_nodes0[0].layer_name]
+                del self.graph.node_map[in_nodes0[1].layer_name]
+                del self.graph.node_map[in_nodes2[0].layer_name]
+                del self.graph.node_map[in_nodes2[1].layer_name]
+                del self.graph.node_map[in_nodes3[1].layer_name]
+                del self.graph.node_map[in_nodes4[1].layer_name]

x2paddle_model_zoo.md

@@ -39,7 +39,7 @@
 | ResNet50 | [code](https://github.com/soeaver/caffe-model/blob/master/cls/resnet/deploy_resnet50.prototxt) |
 | Unet | [code](https://github.com/jolibrain/deepdetect/blob/master/templates/caffe/unet/deploy.prototxt) |
 | VGGNet | [code](https://gist.github.com/ksimonyan/211839e770f7b538e2d8#file-vgg_ilsvrc_16_layers_deploy-prototxt) |
-
+| FaceDetection | [code](https://github.com/ShiqiYu/libfacedetection/blob/master/models/caffe/yufacedetectnet-open-v1.prototxt) |
 
 
 
@@ -65,4 +65,3 @@
 | mNASNet | [pytorch(personal practice)](https://github.com/rwightman/gen-efficientnet-pytorch) |9|
 | EfficientNet | [pytorch(personal practice)](https://github.com/rwightman/gen-efficientnet-pytorch) |9|
 | SqueezeNet | [onnx official](https://s3.amazonaws.com/download.onnx/models/opset_9/squeezenet.tar.gz) |9|
-