深度学习框架OneFlow是如何和ONNX交互的？

阳光穿透心脏的1/2处 2022-11-15 11:48 224阅读 0赞

> oneflow-onnx 工具开发者：daquexian， bbuf

# 0x0. 介绍 #

在开始阅读本篇文章之前，如果你对ONNX不是很了解介意先阅读我之前写的这几篇介绍ONNX文章：

*  [ONNX初探][ONNX]
 *  [ONNX 再探][ONNX 1]
 *  [onnx2pytorch和onnx-simplifier新版介绍][onnx2pytorch_onnx-simplifier]

以及大老师的：

[onnx simplifier 和 optimizer][onnx simplifier _ optimizer]

然后，这篇文章不会继续探索ONNX本身的东西，而是聊聊另外一个有趣的话题，即深度学习框架是如何和ONNX进行交互的？我最近配合大老师基于OneFlow深度学习框架做了一些和ONNX有关的工作，感觉自己对OneFlow和ONNX的交互过程也算熟悉一些了。因此，在这篇文章我将分享OneFlow和ONNX交互的具体实现思路以及介绍oneflow-onnx这个开源工具的一些特性。让读者了解OneFlow的模型是如何转换为ONNX模型，以及ONNX模型是如何转回OneFlow的模型（X2OneFlow）的。个人认为OneFlow目前和ONNX交互的做法是比较优雅并且具有较好扩展性的，因此我们将这项工作转换成了开源成果并分享实现思路，github地址为：`https://github.com/Oneflow-Inc/oneflow_convert_tools`。这个工具作为OneFlow 生态系统的一部分会被我们持续维护，同时这个工具也被我们制作成了一个wheel包，感兴趣的用户只需要pip安装oneflow-onnx即可快速体验。在下面的第二节以及工程的README也有详细的安装步骤。

oneflow-onnx工具包含两个功能，一个是将OneFlow导出ONNX，另外一个是将各个训练框架导出的ONNX模型转换为OneFlow的模型。本工程已经适配了TensorFlow/Pytorch/PaddlePaddle框架的预训练模型通过导出ONNX转换为OneFlow（我们将这一功能叫作X2OneFlow）。更多使用示例以及相关文档和源码均可以在开源`https://github.com/Oneflow-Inc/oneflow_convert_tools`工程中获得。

# 0x1. 算子支持和模型支持 #

## OneFlow2ONNX ##

#### OneFlow2ONNX 支持的OP列表 ####

> 目前OneFlow2ONNX 支持60+的ONNX OP，我们在下面的列表中列出了目前OneFlow支持导出的全部ONNX OP

<table> 
 <thead> 
  <tr> 
   <th>序号</th> 
   <th>OP</th> 
   <th>序号</th> 
   <th>OP</th> 
   <th>序号</th> 
   <th>OP</th> 
   <th>序号</th> 
   <th>OP</th> 
  </tr> 
 </thead> 
 <tbody> 
  <tr> 
   <td>1</td> 
   <td>GatherND</td> 
   <td>2</td> 
   <td>Transpose</td> 
   <td>3</td> 
   <td>Add</td> 
   <td>4</td> 
   <td>Sub</td> 
  </tr> 
  <tr> 
   <td>5</td> 
   <td>Mul</td> 
   <td>6</td> 
   <td>Div</td> 
   <td>7</td> 
   <td>Sum</td> 
   <td>8</td> 
   <td>LeakyRelu</td> 
  </tr> 
  <tr> 
   <td>9</td> 
   <td>Softplus</td> 
   <td>10</td> 
   <td>Softplus</td> 
   <td>11</td> 
   <td>Abs</td> 
   <td>12</td> 
   <td>Ceil</td> 
  </tr> 
  <tr> 
   <td>13</td> 
   <td>Elu</td> 
   <td>14</td> 
   <td>Exp</td> 
   <td>15</td> 
   <td>Floor</td> 
   <td>16</td> 
   <td>Log</td> 
  </tr> 
  <tr> 
   <td>17</td> 
   <td>Neg</td> 
   <td>18</td> 
   <td>Sigmoid</td> 
   <td>19</td> 
   <td>Sqrt</td> 
   <td>20</td> 
   <td>Tanh</td> 
  </tr> 
  <tr> 
   <td>21</td> 
   <td>Reciprocal</td> 
   <td>22</td> 
   <td>Relu</td> 
   <td>23</td> 
   <td>Acos</td> 
   <td>24</td> 
   <td>Asin</td> 
  </tr> 
  <tr> 
   <td>25</td> 
   <td>Atan</td> 
   <td>26</td> 
   <td>Cos</td> 
   <td>27</td> 
   <td>Sin</td> 
   <td>28</td> 
   <td>Tan</td> 
  </tr> 
  <tr> 
   <td>29</td> 
   <td>Acosh</td> 
   <td>30</td> 
   <td>Asinh</td> 
   <td>31</td> 
   <td>Atanh</td> 
   <td>32</td> 
   <td>Cosh</td> 
  </tr> 
  <tr> 
   <td>33</td> 
   <td>Sinh</td> 
   <td>34</td> 
   <td>Min</td> 
   <td>35</td> 
   <td>Max</td> 
   <td>36</td> 
   <td>Clip</td> 
  </tr> 
  <tr> 
   <td>37</td> 
   <td>Softmax</td> 
   <td>38</td> 
   <td>Sign</td> 
   <td>39</td> 
   <td>MatMul</td> 
   <td>40</td> 
   <td>Erf</td> 
  </tr> 
  <tr> 
   <td>41</td> 
   <td>FloorMod</td> 
   <td>42</td> 
   <td>Round</td> 
   <td>43</td> 
   <td>Not</td> 
   <td>44</td> 
   <td>And</td> 
  </tr> 
  <tr> 
   <td>45</td> 
   <td>Or</td> 
   <td>46</td> 
   <td>Equal</td> 
   <td>47</td> 
   <td>NotEqual</td> 
   <td>48</td> 
   <td>Greater</td> 
  </tr> 
  <tr> 
   <td>49</td> 
   <td>Less</td> 
   <td>50</td> 
   <td>Pad</td> 
   <td>51</td> 
   <td>AveragePool</td> 
   <td>52</td> 
   <td>MaxPool</td> 
  </tr> 
  <tr> 
   <td>53</td> 
   <td>Conv</td> 
   <td>54</td> 
   <td>QuantizeLinear</td> 
   <td>56</td> 
   <td>ReduceMin</td> 
   <td>57</td> 
   <td>BatchNormalization</td> 
  </tr> 
  <tr> 
   <td>58</td> 
   <td>ReduceSum</td> 
   <td>59</td> 
   <td>ReduceProd</td> 
   <td>60</td> 
   <td>ArgMax</td> 
   <td>61</td> 
   <td>ArgMin</td> 
  </tr> 
  <tr> 
   <td>62</td> 
   <td>Reshape</td> 
   <td>63</td> 
   <td>Squeeze</td> 
   <td>64</td> 
   <td>Transpose</td> 
   <td>65</td> 
   <td>Concat</td> 
  </tr> 
  <tr> 
   <td>66</td> 
   <td>Cast</td> 
   <td>67</td> 
   <td>Identity</td> 
   <td>68</td> 
   <td>Mul</td> 
   <td></td> 
   <td></td> 
  </tr> 
 </tbody> 
</table>

#### OneFlow2ONNX 模型测试库 ####

> 目前OneFlow2ONNX 支持60+的ONNX OP，我们在下面的模型列表中测试了OneFlow2ONNX的转换。

<table> 
 <thead> 
  <tr> 
   <th>模型</th> 
   <th>来源</th> 
   <th>operator version</th> 
  </tr> 
 </thead> 
 <tbody> 
  <tr> 
   <td>AlexNet</td> 
   <td><a href="https://github.com/Oneflow-Inc/OneFlow-Benchmark/tree/master/Classification/cnns" rel="nofollow">OneFlow-AlexNet</a></td> 
   <td>10</td> 
  </tr> 
  <tr> 
   <td>MobileNetV2</td> 
   <td><a href="https://github.com/Oneflow-Inc/OneFlow-Benchmark/tree/master/Classification/cnns" rel="nofollow">Oneflow-MobileNetV2</a></td> 
   <td>10</td> 
  </tr> 
  <tr> 
   <td>ResNet50</td> 
   <td><a href="https://github.com/Oneflow-Inc/OneFlow-Benchmark/tree/master/Classification/cnns" rel="nofollow">OneFlow-ResNet50</a></td> 
   <td>10</td> 
  </tr> 
 </tbody> 
</table>

## X2OneFlow ##

### X2OneFlow 支持的OP列表 ###

> 目前X2OneFlow 支持40+的ONNX OP，30+的Tensorflow/Pytorch/PaddlePaddle OP，覆盖了大部分CV分类模型常用的操作。OP的单元测试代码会逐渐移步到工程的examples目录下，并支持更多的OP。

#### ONNX ####

<table> 
 <thead> 
  <tr> 
   <th>序号</th> 
   <th>OP</th> 
   <th>序号</th> 
   <th>OP</th> 
   <th>序号</th> 
   <th>OP</th> 
   <th>序号</th> 
   <th>OP</th> 
  </tr> 
 </thead> 
 <tbody> 
  <tr> 
   <td>1</td> 
   <td>Conv</td> 
   <td>2</td> 
   <td>BatchNormalization</td> 
   <td>3</td> 
   <td>MaxPool</td> 
   <td>4</td> 
   <td>AveragePool</td> 
  </tr> 
  <tr> 
   <td>5</td> 
   <td>Concat</td> 
   <td>6</td> 
   <td>ReLU</td> 
   <td>7</td> 
   <td>AdaptiveMaxPool</td> 
   <td>8</td> 
   <td>Softmax</td> 
  </tr> 
  <tr> 
   <td>9</td> 
   <td>Unsqueeze</td> 
   <td>10</td> 
   <td>Transpose</td> 
   <td>11</td> 
   <td>Clip</td> 
   <td>12</td> 
   <td>Gather</td> 
  </tr> 
  <tr> 
   <td>13</td> 
   <td>Slice</td> 
   <td>14</td> 
   <td>Split</td> 
   <td>15</td> 
   <td>Flatten</td> 
   <td>16</td> 
   <td>Add</td> 
  </tr> 
  <tr> 
   <td>17</td> 
   <td>Sub</td> 
   <td>18</td> 
   <td>Mul</td> 
   <td>19</td> 
   <td>Div</td> 
   <td>20</td> 
   <td>Sqrt</td> 
  </tr> 
  <tr> 
   <td>21</td> 
   <td>Pow</td> 
   <td>22</td> 
   <td>Tanh</td> 
   <td>23</td> 
   <td>Sigmoid</td> 
   <td>24</td> 
   <td>Cast</td> 
  </tr> 
  <tr> 
   <td>25</td> 
   <td>Pad</td> 
   <td>26</td> 
   <td>ReduceMean</td> 
   <td>27</td> 
   <td>Reshape</td> 
   <td>28</td> 
   <td>AdaptiveAvgPool</td> 
  </tr> 
  <tr> 
   <td>29</td> 
   <td>Squeeze</td> 
   <td>30</td> 
   <td>Expand</td> 
   <td>31</td> 
   <td>Gather</td> 
   <td>32</td> 
   <td>Slice</td> 
  </tr> 
  <tr> 
   <td>33</td> 
   <td>Split</td> 
   <td>34</td> 
   <td>Min</td> 
   <td>35</td> 
   <td>Max</td> 
   <td>36</td> 
   <td>Constant</td> 
  </tr> 
  <tr> 
   <td>37</td> 
   <td>HardSigmoid</td> 
   <td>38</td> 
   <td>Gemm</td> 
   <td>39</td> 
   <td>MatMul</td> 
   <td>40</td> 
   <td>Erf</td> 
  </tr> 
  <tr> 
   <td>41</td> 
   <td>Cast</td> 
   <td>42</td> 
   <td>GlobalMaxPool</td> 
   <td>43</td> 
   <td>GlobalAveragePool</td> 
   <td>44</td> 
   <td>ReduceMax</td> 
  </tr> 
  <tr> 
   <td>45</td> 
   <td>Identity</td> 
   <td></td> 
   <td></td> 
   <td></td> 
   <td></td> 
   <td></td> 
   <td></td> 
  </tr> 
 </tbody> 
</table>

#### TensorFlow ####

<table> 
 <thead> 
  <tr> 
   <th>序号</th> 
   <th>OP</th> 
   <th>序号</th> 
   <th>OP</th> 
   <th>序号</th> 
   <th>OP</th> 
   <th>序号</th> 
   <th>OP</th> 
  </tr> 
 </thead> 
 <tbody> 
  <tr> 
   <td>1</td> 
   <td>relu</td> 
   <td>2</td> 
   <td>concatenate</td> 
   <td>3</td> 
   <td>expand_dims</td> 
   <td>4</td> 
   <td>transpose</td> 
  </tr> 
  <tr> 
   <td>5</td> 
   <td>batchnorm</td> 
   <td>6</td> 
   <td>slice</td> 
   <td>7</td> 
   <td>gather</td> 
   <td>8</td> 
   <td>clip_by_value</td> 
  </tr> 
  <tr> 
   <td>9</td> 
   <td>conv2d</td> 
   <td>10</td> 
   <td>depthwiseconv2d</td> 
   <td>11</td> 
   <td>flatten</td> 
   <td>12</td> 
   <td>add</td> 
  </tr> 
  <tr> 
   <td>13</td> 
   <td>sub</td> 
   <td>14</td> 
   <td>mul</td> 
   <td>15</td> 
   <td>div</td> 
   <td>16</td> 
   <td>pow</td> 
  </tr> 
  <tr> 
   <td>17</td> 
   <td>sqrt</td> 
   <td>18</td> 
   <td>tanh</td> 
   <td>19</td> 
   <td>sigmoid</td> 
   <td>20</td> 
   <td>erf</td> 
  </tr> 
  <tr> 
   <td>21</td> 
   <td>cast</td> 
   <td>22</td> 
   <td>pad</td> 
   <td>23</td> 
   <td>maxpool</td> 
   <td>24</td> 
   <td>avgpool</td> 
  </tr> 
  <tr> 
   <td>25</td> 
   <td>globalavgpool</td> 
   <td>26</td> 
   <td>globalmaxpool</td> 
   <td>27</td> 
   <td>reduce_mean</td> 
   <td>28</td> 
   <td>reshape</td> 
  </tr> 
  <tr> 
   <td>29</td> 
   <td>softmax</td> 
   <td>30</td> 
   <td>relu6</td> 
   <td></td> 
   <td></td> 
   <td></td> 
   <td></td> 
  </tr> 
 </tbody> 
</table>

*  分组卷积存在问题，已给TensorFlow2ONNX团队PR。

#### Pytorch ####

<table> 
 <thead> 
  <tr> 
   <th>序号</th> 
   <th>OP</th> 
   <th>序号</th> 
   <th>OP</th> 
   <th>序号</th> 
   <th>OP</th> 
   <th>序号</th> 
   <th>OP</th> 
  </tr> 
 </thead> 
 <tbody> 
  <tr> 
   <td>1</td> 
   <td>relu</td> 
   <td>2</td> 
   <td>cat</td> 
   <td>3</td> 
   <td>unsqueeze</td> 
   <td>4</td> 
   <td>transpose</td> 
  </tr> 
  <tr> 
   <td>5</td> 
   <td>batchnorm</td> 
   <td>6</td> 
   <td>slice</td> 
   <td>7</td> 
   <td>gather</td> 
   <td>8</td> 
   <td>clamp</td> 
  </tr> 
  <tr> 
   <td>9</td> 
   <td>conv2d</td> 
   <td>10</td> 
   <td>depthwiseconv2d</td> 
   <td>11</td> 
   <td>flatten</td> 
   <td>12</td> 
   <td>add</td> 
  </tr> 
  <tr> 
   <td>13</td> 
   <td>sub</td> 
   <td>14</td> 
   <td>mul</td> 
   <td>15</td> 
   <td>div</td> 
   <td>16</td> 
   <td>pow</td> 
  </tr> 
  <tr> 
   <td>17</td> 
   <td>sqrt</td> 
   <td>18</td> 
   <td>tanh</td> 
   <td>19</td> 
   <td>sigmoid</td> 
   <td>20</td> 
   <td>erf</td> 
  </tr> 
  <tr> 
   <td>21</td> 
   <td>cast</td> 
   <td>22</td> 
   <td>pad</td> 
   <td>23</td> 
   <td>maxpool</td> 
   <td>24</td> 
   <td>avgpool</td> 
  </tr> 
  <tr> 
   <td>25</td> 
   <td>globalavgpool</td> 
   <td>26</td> 
   <td>globalmaxpool</td> 
   <td>27</td> 
   <td>reduce_mean</td> 
   <td>28</td> 
   <td>reshape</td> 
  </tr> 
  <tr> 
   <td>29</td> 
   <td>softmax</td> 
   <td>30</td> 
   <td>relu6</td> 
   <td>31</td> 
   <td>CrossEntropyLoss</td> 
   <td></td> 
   <td></td> 
  </tr> 
 </tbody> 
</table>

#### PaddlePaddle ####

<table> 
 <thead> 
  <tr> 
   <th>序号</th> 
   <th>OP</th> 
   <th>序号</th> 
   <th>OP</th> 
   <th>序号</th> 
   <th>OP</th> 
   <th>序号</th> 
   <th>OP</th> 
  </tr> 
 </thead> 
 <tbody> 
  <tr> 
   <td>1</td> 
   <td>relu</td> 
   <td>2</td> 
   <td>concatenate</td> 
   <td>3</td> 
   <td>expand_dims</td> 
   <td>4</td> 
   <td>transpose</td> 
  </tr> 
  <tr> 
   <td>5</td> 
   <td>batchnorm</td> 
   <td>6</td> 
   <td>slice</td> 
   <td>7</td> 
   <td>gather</td> 
   <td>8</td> 
   <td>clip_by_value</td> 
  </tr> 
  <tr> 
   <td>9</td> 
   <td>conv2d</td> 
   <td>10</td> 
   <td>depthwiseconv2d</td> 
   <td>11</td> 
   <td>flatten</td> 
   <td>12</td> 
   <td>add</td> 
  </tr> 
  <tr> 
   <td>13</td> 
   <td>sub</td> 
   <td>14</td> 
   <td>mul</td> 
   <td>15</td> 
   <td>div</td> 
   <td>16</td> 
   <td>pow</td> 
  </tr> 
  <tr> 
   <td>17</td> 
   <td>sqrt</td> 
   <td>18</td> 
   <td>tanh</td> 
   <td>19</td> 
   <td>sigmoid</td> 
   <td>20</td> 
   <td>erf</td> 
  </tr> 
  <tr> 
   <td>21</td> 
   <td>cast</td> 
   <td>22</td> 
   <td>pad</td> 
   <td>23</td> 
   <td>maxpool</td> 
   <td>24</td> 
   <td>avgpool</td> 
  </tr> 
  <tr> 
   <td>25</td> 
   <td>adaptiveavgpool</td> 
   <td>26</td> 
   <td>adptivemaxpool</td> 
   <td>27</td> 
   <td>reduce_mean</td> 
   <td>28</td> 
   <td>reshape</td> 
  </tr> 
  <tr> 
   <td>29</td> 
   <td>softmax</td> 
   <td>30</td> 
   <td>relu6</td> 
   <td></td> 
   <td></td> 
   <td></td> 
   <td></td> 
  </tr> 
 </tbody> 
</table>

相关issue：

*  https://github.com/PaddlePaddle/Paddle2ONNX/issues/221
 *  https://github.com/PaddlePaddle/Paddle2ONNX/issues/220

### X2OneFlow模型测试库 ###

目前X2OneFlow 支持40+的ONNX OP，30+的Tensorflow/Pytorch/PaddlePaddle OP，覆盖了大部分CV分类模型常用的操作。我们在如下模型列表中测试了X2OneFlow的转换。

#### Pytorch ####

#### TensorFlow ####

<table> 
 <thead> 
  <tr> 
   <th>模型</th> 
   <th>是否支持</th> 
  </tr> 
 </thead> 
 <tbody> 
  <tr> 
   <td>VGGNet</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>ResNet</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>ResNetV2</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>XceptionNet</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>MobileNetV1</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>MobileNetV2</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>MobileNetV3</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>DenseNet</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>EfficientNet</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>InceptionNet</td> 
   <td>Yes</td> 
  </tr> 
 </tbody> 
</table>

#### PaddlePaddle ####

<table> 
 <thead> 
  <tr> 
   <th>模型</th> 
   <th>是否支持</th> 
  </tr> 
 </thead> 
 <tbody> 
  <tr> 
   <td>AlexNet</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>VGGNet</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>GoogleNet</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>ResNet</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>ResNext</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>SE_ResNext</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>SENet</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>MobileNetV1</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>MobileNetV2</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>MobileNetV3</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>RegNet</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>DenseNet</td> 
   <td>No（msg: “op_name: Concat_58 already exist in job: job_eval”）</td> 
  </tr> 
  <tr> 
   <td>EfficientNet</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>InceptionNet</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>ShuffleNetV2</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>SqueezeNet</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>DPNNet</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>DarkNet</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>GhostNet</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>RepVGG</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>XceptionNet</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>Xception_DeepLab</td> 
   <td>Yes</td> 
  </tr> 
  <tr> 
   <td>Vision_Transformer</td> 
   <td>No（“op_name: Constant_20 already exist in job: job_eval”）</td> 
  </tr> 
  <tr> 
   <td>Res2Net</td> 
   <td>No（split op bug，working）</td> 
  </tr> 
  <tr> 
   <td>Unet</td> 
   <td>No（OneFlow的上采样OP和Paddle未对齐）</td> 
  </tr> 
 </tbody> 
</table>

*  模型的测试代码均可以在工程的examples中找到。

# 0x2. 快速体验 #

##### 用户环境配置 #####

python>=3.5
    onnx>=1.8.0
    onnx-simplifier>=0.3.3
    onnxoptimizer>=0.2.5
    onnxruntime>=1.6.0
    oneflow>=0.3.4

如果你想使用X2OneFlow（X代表TensorFlow/Pytorch/PaddlePaddle）需要安装对应的深度学习框架，需要安装对应的深度学习框架，依赖如下：

pytorch>=1.7.0
    paddlepaddle>=2.0.0
    tensorflow>=2.0.0

#### 安装 ####

##### 安装方式1 #####

pip install oneflow_onnx

#### 安装方式2 ####

git clone https://github.com/Oneflow-Inc/oneflow_convert_tools
    cd oneflow_onnx
    python3 setup.py install

使用方法见工程的samples下的示例。

# 0x3. OneFlow-ONNX思路分享 #

我们将在这一节分享一下OneFlow的模型是如何被转换为ONNX的，这里我们以将OneFlow定义的AlexNet导出ONNX模型为例来分析源码。首先我们`https://github.com/Oneflow-Inc/oneflow_convert_tools/blob/main/examples/oneflow2onnx/test_alexnet.py#L133`进到这里，可以看到下面调用代码：

def test_alexnet():
        @flow.global_function()
        def alexnet_eval_job(x: tp.Numpy.Placeholder((1, 227, 227, 3))):
            return alexnet(x, None, False)
    
        convert_to_onnx_and_check(alexnet_eval_job, flow_weight_dir=None, onnx_model_path="/tmp")

这里通过`flow.global_function()`定义了一个预测用于eval的AlexNet `job`，网络的完整定义可以通过上面的链接访问，可以看到这里通过`convert_to_onnx_and_check`函数将OneFlow定义的AlexNet转换为了ONNX模型，我们跟进这个函数，就来到了这里：`https://github.com/Oneflow-Inc/oneflow_convert_tools/blob/main/oneflow_onnx/oneflow2onnx/util.py#L65-L73`，代码为：

while not os.path.exists(os.path.join(flow_weight_dir, "snapshot_done")):
         pass
     onnx_model_dir = onnx_model_path
     onnx_model_path = os.path.join(onnx_model_dir, "model.onnx")
     flow.onnx.export(
         job_func,
         flow_weight_dir,
         onnx_model_path,
         opset=opset,
         external_data=external_data,
     )

可以看到完成ONNX模型转换的核心函数就是这个`flow.onnx.export`函数，我们继续跳转到这个函数`https://github.com/Oneflow-Inc/oneflow_convert_tools/blob/main/oneflow_onnx/oneflow2onnx/flow2onnx.py#L229-L281`，代码如下：

def Export(
        job_func: Callable,
        model_save_dir: Text,
        onnx_filename: Text,
        continue_on_error: bool = False,
        opset: Optional[int] = None,
        extra_opset: Optional[int] = None,
        shape_override: Optional[Dict[Text, List[int]]] = None,
        external_data: bool = False,
    ):
        r"""Export a oneflow model into ONNX format.
        Args:
            job_func: OneFlow的作业函数
            model_save_dir: 包含OneFlow定义的模型权重的文件夹. 这个模型权重是用oneflow的check_point.save接口保存的。
            onnx_filename: 输出ONNX模型文件名，字符串类型
            continue_on_error: 如果某个OP无法处理（即没有映射），是否继续
            opset: ONNX Opset版本号，默认为10
            extra_opset: 额外Opset的列表，例如自定义操作使用的Opset
            shape_override: 带有输入信息的字典，覆盖OneFlow给定的输入形状
            external_data: 将权重另存为ONNX外部数据，通常是为了绕过protobuf的2GB文件大小限制。
        """
        assert os.getenv("ENABLE_USER_OP") != "False"
        # 确定模型的路径是存在的
        assert os.path.isdir(model_save_dir)
        # 通过c_api_util.GetJobSet()方法获取当前的所有job
        job_set = c_api_util.GetJobSet()
        # 我们要转的模型被定义在job_func中，所以我们先记录下它的名字
        job_name = job_func.__name__
        # 编译job_set，找到定义模型的job
        for job in job_set.job:
            # TODO(OYY) Modify the interface before modifying it
            if job.job_conf.job_name == job_name:
                # job找到了，可以开始进行下面的步骤，我们在外面详细解释
                onnx_graph = ProcessFlowGraph(
                    job,
                    model_save_dir,
                    continue_on_error=continue_on_error,
                    opset=opset,
                    extra_opset=extra_opset,
                    shape_override=shape_override,
                )
                onnx_graph = optimizer.OptimizeGraph(onnx_graph)
                model_proto = onnx_graph.MakeModel(
                    job_name, onnx_filename, external_data=external_data
                )
                with open(onnx_filename, "wb") as f:
                    try:
                        f.write(model_proto.SerializeToString())
                    except ValueError as e:
                        raise ValueError(
                            "Error occured when running model_proto.SerializeToString(). If the model is larger than 2GB, please specify external_data=True when calling flow.onnx.export. Original error message:\n{}".format(
                                e
                            )
                        )
                return
        raise ValueError('Cannot find job "{}" in jobset'.format(job_name))

可以看到这个函数首先编译了OneFlow中的job\_set，然后找到了我们最开始定义AlexNet模型的那个job，然后就进入了`ProcessFlowGraph`函数，这个函数主要做了三件事情并最终获得了**初版的合法ONNX模型**（初版的意思是还没有经过优化以及填ONNX节点的权重），我们跟进这个函数，代码如下。

def ProcessFlowGraph(
        flow_graph,
        model_save_dir,
        continue_on_error=False,
        opset=None,
        extra_opset=None,
        shape_override=None,
    ):
    	# 这个函数用来获取导出的ONNX的Opset Version，OneFlow里面最高为10
        opset = util.FindOpset(opset)
        logger.info("Using opset <onnx, %s>", opset)
        # 判断当前的ONNX版本是否支持上面的Opset Version
        if opset > schemas.get_max_supported_opset_version():
            logger.warning(
                "Currently installed onnx package %s is too low to support opset %s, "
                "please upgrade onnx package to avoid potential conversion issue.",
                util.get_onnx_version(),
                opset,
            )
    
        if shape_override is None:
            shape_override = {
        }
    	# 用于将oneflow 的各个 node 转换为 onnx node 的格式，保持 op 类型、输入输出和属性值不变，这一步产生的还不是合法的 onnx 模型
        (onnx_nodes, op_cnt, attr_cnt, dtypes, output_shapes,) = FlowToOnnxNaive(
            flow_graph, shape_override
        )
    	# 构造一个 Graph 类，用于后续方便的修改 onnx 网络结构
        g = Graph(onnx_nodes, model_save_dir, output_shapes, dtypes, opset, extra_opset,)
    
        # create ops mapping for the desired opsets
        ops_mapping = handler.flow_op.CreateMapping(g.opset, g.extra_opset)
    
        # some nodes may already copied into inner Graph, so remove them from main Graph.
        TopologicalSort(g, continue_on_error)
    	# FlowOnnxMapping 函数调用各个转换函数（通过 @flow_op 注册）逐个转换 op，转换后产生的是合法的 onnx 模型
        mapped_op, unmapped_op, exceptions = FlowOnnxMapping(g, ops_mapping)
        if unmapped_op:
            logger.error("Unsupported ops: %s", unmapped_op)
        if exceptions and not continue_on_error:
            raise exceptions[0]
    
        # onnx requires topological sorting
        TopologicalSort(g, continue_on_error)
    
        g.UpdateProto()
    
        logger.debug(
            "Summay Stats:\n"
            "\toneflow ops: {}\n"
            "\toneflow attr: {}\n"
            "\tonnx mapped: {}\n"
            "\tonnx unmapped: {}".format(op_cnt, attr_cnt, mapped_op, unmapped_op)
        )
    
        return g

`FlowToOnnxNaive`这个函数用于将oneflow 的各个 node 转换为 onnx node 的格式，保持 op 类型、输入输出和属性值不变，最后将转换后的ONNX节点（这个地方这些ONNX节点还不是真正的合法ONNX节点，要后面执行一对一转换之后才是合法的ONNX节点）全部返回。接下来利用这些ONNX节点来构造Graph类，方便后续对ONNX模型进行修改。Graph类的实现在`https://github.com/Oneflow-Inc/oneflow_convert_tools/blob/18e041d92654cfc8b03e16c906c451a405c99fd2/oneflow_onnx/onnx_wrapper.py`，这个文件主要是定义了onnx graph和node的wrapper，包含各种修改 onnx 图结构的 api，这里复用了tensorflow-onnx项目的相关代码。注意构造Graph类之后还并没有构造ONNX模型，因为OneFlow的OP还没有一对一的转换为ONNX的OP。

接下来，我们调用`handler.flow_op.CreateMapping(g.opset, g.extra_opset)`这个函数，代码实现如下：

def CreateMapping(max_onnx_opset_version, extra_opsets):
            """Create the final mapping dictionary by stacking domains and opset versions.
    
            :param max_onnx_opset_version: The highest onnx opset the resulting graph may use.
            :param extra_opsets: Extra opsets the resulting graph may use.
            """
            mapping = {
        constants.ONNX_DOMAIN: max_onnx_opset_version}
            if extra_opsets:
                for extra_opset in extra_opsets:
                    mapping[extra_opset.domain] = extra_opset.version
            ops_mapping = {
        }
            for domain, opsets in flow_op.get_opsets().items():
                for target_opset, op_map in enumerate(opsets):
                    print('='*100)
                    print(target_opset)
                    print(op_map)
                    m = mapping.get(domain)
                    if m:
                        if target_opset <= m and op_map:
                            ops_mapping.update(op_map)
    
            flow_op._MAPPING = ops_mapping
            return ops_mapping

这个函数做的事情就是将每个ONNX Opset版本号（也就是for循环中的`domain`）和(OneFlow OP和ONNX OP的mapper，这个mapper是如何获得的请看后文)关联起来并返回，我们打印一下`target_opset`和`op_map`就可以理解了。以AlexNet为例打印如下：

====================================================================================================
    0
    {
        }
    ====================================================================================================
    1
    {
        'add_n': (<bound method AddN.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.AddN'>>, 'Sum', {
        }), 'leaky_relu': (<bound method DirectOpSinceOpset1.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOpSinceOpset1'>>, 'LeakyRelu', {
        }), 'softplus': (<bound method DirectOpSinceOpset1.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOpSinceOpset1'>>, 'Softplus', {
        }), 'abs': (<bound method DirectOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Abs', {
        }), 'ceil': (<bound method DirectOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Ceil', {
        }), 'elu': (<bound method DirectOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Elu', {
        }), 'exp': (<bound method DirectOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Exp', {
        }), 'floor': (<bound method DirectOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Floor', {
        }), 'log': (<bound method DirectOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Log', {
        }), 'neg': (<bound method DirectOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Neg', {
        }), 'sigmoid': (<bound method DirectOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Sigmoid', {
        }), 'sigmoid_v2': (<bound method DirectOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Sigmoid', {
        }), 'sqrt': (<bound method DirectOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Sqrt', {
        }), 'tanh': (<bound method DirectOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Tanh', {
        }), 'reciprocal': (<bound method DirectOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Reciprocal', {
        }), 'relu': (<bound method DirectOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Relu', {
        }), 'broadcast_maximum': (<bound method MinMaxOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.MinMaxOp'>>, 'Max', {
        }), 'broadcast_minimum': (<bound method MinMaxOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.MinMaxOp'>>, 'Min', {
        }), 'clip_by_scalar': (<bound method ClipByValueOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.ClipByValueOp'>>, 'Clip', {
        }), 'clip_by_scalar_min': (<bound method ClipByValueOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.ClipByValueOp'>>, 'Clip', {
        }), 'clip_by_scalar_max': (<bound method ClipByValueOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.ClipByValueOp'>>, 'Clip', {
        }), 'softmax': (<bound method Softmax.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.Softmax'>>, 'Softmax', {
        }), 'square': (<bound method Square.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.Square'>>, None, {
        }), 'rsqrt': (<bound method Rsqrt.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.Rsqrt'>>, None, {
        }), 'squared_difference': (<bound method SquaredDifference.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.SquaredDifference'>>, None, {
        }), 'sign': (<bound method Sign.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.Sign'>>, 'Sign', {
        }), 'matmul': (<bound method MatMul.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.MatMul'>>, 'MatMul', {
        }), 'batch_matmul': (<bound method MatMul.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.MatMul'>>, 'MatMul', {
        }), 'erf': (<bound method Erf.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.Erf'>>, 'Erf', {
        }), 'logical_not': (<bound method DirectOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Not', {
        }), 'broadcast_logical_or': (<bound method BroadcastOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.BroadcastOp'>>, 'Or', {
        }), 'broadcast_logical_and': (<bound method BroadcastOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.BroadcastOp'>>, 'And', {
        }), 'input': (<bound method DirectOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.misc.DirectOp'>>, None, {
        }), 'return': (<bound method DirectOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.misc.DirectOp'>>, None, {
        }), 'variable': (<bound method DirectOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.misc.DirectOp'>>, None, {
        }), 'distribute_split': (<bound method BoxingOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.misc.BoxingOp'>>, 'Identity', {
        }), 'distribute_concat': (<bound method BoxingOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.misc.BoxingOp'>>, 'Identity', {
        }), 'distribute_clone': (<bound method BoxingOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.misc.BoxingOp'>>, 'Identity', {
        }), 'distribute_add': (<bound method BoxingOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.misc.BoxingOp'>>, 'Identity', {
        }), 'conv2d': (<bound method ConvOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.nn.ConvOp'>>, None, {
        }), 'max_pool_2d': (<bound method PoolOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.nn.PoolOp'>>, 'MaxPool', {
        }), 'avg_pool_2d': (<bound method PoolOp.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.nn.PoolOp'>>, 'AveragePool', {
        }), 'reduce_prod': (<bound method ReduceOpBase.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.reduce.ReduceOpBase'>>, 'ReduceProd', {
        }), 'reduce_sum': (<bound method ReduceOpBase.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.reduce.ReduceOpBase'>>, 'ReduceSum', {
        }), 'reduce_min': (<bound method ReduceOpBase.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.reduce.ReduceOpBase'>>, 'ReduceMin', {
        }), 'argmax': (<bound method ArgMax.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.reduce.ArgMax'>>, 'ArgMax', {
        }), 'argmin': (<bound method ArgMax.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.reduce.ArgMax'>>, 'ArgMin', {
        }), 'squeeze': (<bound method Squeeze.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.array.Squeeze'>>, 'Squeeze', {
        }), 'transpose': (<bound method Transpose.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.array.Transpose'>>, 'Transpose', {
        }), 'concat': (<bound method Concat.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.array.Concat'>>, 'Concat', {
        }), 'identity': (<bound method Identity.Version_1 of <class 'oneflow_onnx.oneflow2onnx.handlers.array.Identity'>>, 'Identity', {
        })}
    ====================================================================================================
    2
    {
        'pad': (<bound method Pad.Version_2 of <class 'oneflow_onnx.oneflow2onnx.handlers.nn.Pad'>>, 'Pad', {
        })}
    ====================================================================================================
    3
    {
        }
    ====================================================================================================
    4
    {
        }
    ====================================================================================================
    5
    {
        'reshape': (<bound method Reshape.Version_5 of <class 'oneflow_onnx.oneflow2onnx.handlers.array.Reshape'>>, 'Reshape', {
        })}
    ====================================================================================================
    6
    {
        'broadcast_div': (<bound method BroadcastOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.BroadcastOp'>>, 'Div', {
        }), 'scalar_div_by_tensor': (<bound method BroadcastOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.BroadcastOp'>>, 'Div', {
        }), 'multiply': (<bound method BroadcastOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.BroadcastOp'>>, 'Mul', {
        }), 'broadcast_mul': (<bound method BroadcastOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.BroadcastOp'>>, 'Mul', {
        }), 'scalar_mul_by_tensor': (<bound method BroadcastOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.BroadcastOp'>>, 'Mul', {
        }), 'broadcast_sub': (<bound method BroadcastOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.BroadcastOp'>>, 'Sub', {
        }), 'scalar_sub_by_tensor': (<bound method BroadcastOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.BroadcastOp'>>, 'Sub', {
        }), 'broadcast_add': (<bound method BroadcastOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.BroadcastOp'>>, 'Add', {
        }), 'scalar_add_by_tensor': (<bound method BroadcastOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.BroadcastOp'>>, 'Add', {
        }), 'scalar_add': (<bound method ScalarBinaryOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.ScalarBinaryOp'>>, 'Add', {
        }), 'scalar_mul': (<bound method ScalarBinaryOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.ScalarBinaryOp'>>, 'Mul', {
        }), 'bias_add': (<bound method BiasAdd.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.BiasAdd'>>, 'Add', {
        }), 'abs': (<bound method DirectOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Abs', {
        }), 'ceil': (<bound method DirectOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Ceil', {
        }), 'elu': (<bound method DirectOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Elu', {
        }), 'exp': (<bound method DirectOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Exp', {
        }), 'floor': (<bound method DirectOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Floor', {
        }), 'log': (<bound method DirectOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Log', {
        }), 'neg': (<bound method DirectOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Neg', {
        }), 'sigmoid': (<bound method DirectOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Sigmoid', {
        }), 'sigmoid_v2': (<bound method DirectOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Sigmoid', {
        }), 'sqrt': (<bound method DirectOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Sqrt', {
        }), 'tanh': (<bound method DirectOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Tanh', {
        }), 'reciprocal': (<bound method DirectOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Reciprocal', {
        }), 'relu': (<bound method DirectOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.DirectOp'>>, 'Relu', {
        }), 'broadcast_logical_or': (<bound method BroadcastOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.BroadcastOp'>>, 'Or', {
        }), 'broadcast_logical_and': (<bound method BroadcastOp.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.BroadcastOp'>>, 'And', {
        }), 'normalization': (<bound method BatchNorm.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.nn.BatchNorm'>>, None, {
        }), 'cast': (<bound method Cast.Version_6 of <class 'oneflow_onnx.oneflow2onnx.handlers.array.Cast'>>, 'Cast', {
        })}
    ====================================================================================================
    7
    {
        'acos': (<bound method TrigOpSinceOpset7.Version_7 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.TrigOpSinceOpset7'>>, 'Acos', {
        }), 'asin': (<bound method TrigOpSinceOpset7.Version_7 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.TrigOpSinceOpset7'>>, 'Asin', {
        }), 'atan': (<bound method TrigOpSinceOpset7.Version_7 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.TrigOpSinceOpset7'>>, 'Atan', {
        }), 'cos': (<bound method TrigOpSinceOpset7.Version_7 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.TrigOpSinceOpset7'>>, 'Cos', {
        }), 'sin': (<bound method TrigOpSinceOpset7.Version_7 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.TrigOpSinceOpset7'>>, 'Sin', {
        }), 'tan': (<bound method TrigOpSinceOpset7.Version_7 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.TrigOpSinceOpset7'>>, 'Tan', {
        }), 'broadcast_floor_mod': (<bound method FloorMod.Version_7 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.FloorMod'>>, 'FloorMod', {
        }), 'broadcast_equal': (<bound method Equal.Version_7 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.Equal'>>, 'Equal', {
        }), 'broadcast_not_equal': (<bound method Equal.Version_7 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.Equal'>>, 'NotEqual', {
        }), 'broadcast_greater': (<bound method GreaterLess.Version_7 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.GreaterLess'>>, 'Greater', {
        }), 'broadcast_less': (<bound method GreaterLess.Version_7 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.GreaterLess'>>, 'Less', {
        }), 'broadcast_less_equal': (<bound method GreaterLessEqual.Version_7 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.GreaterLessEqual'>>, 'Greater', {
        }), 'broadcast_greater_equal': (<bound method GreaterLessEqual.Version_7 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.GreaterLessEqual'>>, 'Less', {
        })}
    ====================================================================================================
    8
    {
        }
    ====================================================================================================
    9
    {
        'acosh': (<bound method TrigOpSinceOpset9.Version_9 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.TrigOpSinceOpset9'>>, 'Acosh', {
        }), 'asinh': (<bound method TrigOpSinceOpset9.Version_9 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.TrigOpSinceOpset9'>>, 'Asinh', {
        }), 'atanh': (<bound method TrigOpSinceOpset9.Version_9 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.TrigOpSinceOpset9'>>, 'Atanh', {
        }), 'cosh': (<bound method TrigOpSinceOpset9.Version_9 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.TrigOpSinceOpset9'>>, 'Cosh', {
        }), 'sinh': (<bound method TrigOpSinceOpset9.Version_9 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.TrigOpSinceOpset9'>>, 'Sinh', {
        }), 'sign': (<bound method Sign.Version_9 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.Sign'>>, 'Sign', {
        }), 'erf': (<bound method Erf.Version_9 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.Erf'>>, 'Erf', {
        }), 'normalization': (<bound method BatchNorm.Version_9 of <class 'oneflow_onnx.oneflow2onnx.handlers.nn.BatchNorm'>>, None, {
        }), 'cast': (<bound method Cast.Version_9 of <class 'oneflow_onnx.oneflow2onnx.handlers.array.Cast'>>, 'Cast', {
        })}
    ====================================================================================================
    10
    {
        'max_pool_2d': (<bound method PoolOp.Version_10 of <class 'oneflow_onnx.oneflow2onnx.handlers.nn.PoolOp'>>, 'MaxPool', {
        }), 'avg_pool_2d': (<bound method PoolOp.Version_10 of <class 'oneflow_onnx.oneflow2onnx.handlers.nn.PoolOp'>>, 'AveragePool', {
        }), 'min_max_observer': (<bound method MinMaxObserver.Version_10 of <class 'oneflow_onnx.oneflow2onnx.handlers.quantize.MinMaxObserver'>>, None, {
        }), 'moving_average_min_max_observer': (<bound method MovingAverageMinMaxObserver.Version_10 of <class 'oneflow_onnx.oneflow2onnx.handlers.quantize.MovingAverageMinMaxObserver'>>, None, {
        }), 'fake_quantization': (<bound method FakeQuantization.Version_10 of <class 'oneflow_onnx.oneflow2onnx.handlers.quantize.FakeQuantization'>>, 'QuantizeLinear', {
        })}
    ====================================================================================================
    11
    {
        'clip_by_scalar': (<bound method ClipByValueOp.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.ClipByValueOp'>>, 'Clip', {
        }), 'clip_by_scalar_min': (<bound method ClipByValueOp.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.ClipByValueOp'>>, 'Clip', {
        }), 'clip_by_scalar_max': (<bound method ClipByValueOp.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.ClipByValueOp'>>, 'Clip', {
        }), 'softmax': (<bound method Softmax.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.Softmax'>>, 'Softmax', {
        }), 'round': (<bound method Round.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.Round'>>, 'Round', {
        }), 'broadcast_equal': (<bound method Equal.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.Equal'>>, 'Equal', {
        }), 'broadcast_not_equal': (<bound method Equal.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.math.Equal'>>, 'NotEqual', {
        }), 'conv2d': (<bound method ConvOp.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.nn.ConvOp'>>, None, {
        }), 'max_pool_2d': (<bound method PoolOp.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.nn.PoolOp'>>, 'MaxPool', {
        }), 'avg_pool_2d': (<bound method PoolOp.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.nn.PoolOp'>>, 'AveragePool', {
        }), 'pad': (<bound method Pad.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.nn.Pad'>>, 'Pad', {
        }), 'reduce_prod': (<bound method ReduceOpBase.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.reduce.ReduceOpBase'>>, 'ReduceProd', {
        }), 'reduce_sum': (<bound method ReduceOpBase.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.reduce.ReduceOpBase'>>, 'ReduceSum', {
        }), 'reduce_min': (<bound method ReduceOpBase.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.reduce.ReduceOpBase'>>, 'ReduceMin', {
        }), 'argmax': (<bound method ArgMax.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.reduce.ArgMax'>>, 'ArgMax', {
        }), 'argmin': (<bound method ArgMax.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.reduce.ArgMax'>>, 'ArgMin', {
        }), 'squeeze': (<bound method Squeeze.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.array.Squeeze'>>, 'Squeeze', {
        }), 'concat': (<bound method Concat.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.array.Concat'>>, 'Concat', {
        }), 'gather_nd': (<bound method GatherND.Version_11 of <class 'oneflow_onnx.oneflow2onnx.handlers.array.GatherND'>>, 'GatherND', {
        })}
    ====================================================================================================
    12
    {
        }
    ====================================================================================================
    13
    {
        'min_max_observer': (<bound method MinMaxObserver.Version_13 of <class 'oneflow_onnx.oneflow2onnx.handlers.quantize.MinMaxObserver'>>, None, {
        }), 'fake_quantization': (<bound method FakeQuantization.Version_13 of <class 'oneflow_onnx.oneflow2onnx.handlers.quantize.FakeQuantization'>>, 'QuantizeLinear', {
        })}

可以看到对于ONNX的每一个Opset Version的OP都对应了OneFlow实现的OP，需要特别注意的是这个OP Mapper过程在是在`https://github.com/Oneflow-Inc/oneflow_convert_tools/tree/main/oneflow_onnx/oneflow2onnx/handlers`这里完成的，只要安装了oneflow-onnx这个包或者编译了oneflow-onnx工程源码，Python就会自动将OneFlow的OP和ONNX的OP进行映射，这是通过`@flow_op(["avg_pool_2d"], onnx_op="AveragePool")`装饰器来实现的，`flow_op`装饰器的具体实现在`https://github.com/Oneflow-Inc/oneflow_convert_tools/blob/main/oneflow_onnx/oneflow2onnx/handler.py#L34`这里。

完成了ONNX每个Opset版本的OP和OneFlow OP的mapper之后，我们需要对`Graph`里面的ONNX节点（注意现在的ONNX节点并不是合法的ONNX节点，因为还没有执行一对一的转换，只是复制了OneFlow OP的类型、输入输出和属性值）先执行拓扑排序，然后再一对一的转换。这个地方很有意思，为什么要进行拓扑排序呢？

我们首先需要了解一下拓扑序算法，拓扑排序要解决的问题是给一个图的所有节点排序。

以下对拓扑排序的解释引自oi.wiki。

我们可以拿大学选课的例子来描述这个过程，比如学习大学课程中有：单变量微积分，线性代数，离散数学概述，概率论与统计学概述，语言基础，算法导论，机器学习。当我们想要学习 算法导论 的时候，就必须先学会 离散数学概述 和 概率论与统计学概述，不然在课堂就会听的一脸懵逼。当然还有一个更加前的课程 单变量微积分。这些课程就相当于几个顶点  u u u, 顶点之间的有向边  ( u , v ) (u,v) (u,v) 就相当于学习课程的顺序。显然拓扑排序不是那么的麻烦，不然你是如何选出合适的学习顺序。下面将介绍如何将这个过程抽象出来，用算法来实现。

但是如果某一天排课的老师打瞌睡了，说想要学习 算法导论，还得先学 机器学习，而 机器学习 的前置课程又是 算法导论，然后你就一万脸懵逼了，我到底应该先学哪一个？当然我们在这里不考虑什么同时学几个课程的情况。在这里，算法导论 和 机器学习 间就出现了一个环，显然你现在没办法弄清楚你需要学什么了，于是你也没办法进行拓扑排序了。因而如果有向图中存在环路，那么我们就没办法进行 拓扑排序 了。

因此我们可以说 在一个 DAG（有向无环图），我们将图中的顶点以线性方式进行排序，使得对于任何的顶点  u u u 到  v v v 的有向边  ( u , v ) (u,v) (u,v), 都可以有  u u u 在  v v v 的前面。

还有给定一个 DAG，如果从  i i i 到  j j j 有边，则认为  j j j 依赖于  i i i。如果  i i i 到  j j j 有路径（ i i i 可达  j j j），则称  j j j 间接依赖于  i i i。

**拓扑排序的目标是将所有节点排序，使得排在前面的节点不能依赖于排在后面的节点。** 伪代码实现如下：

void TopologicalSort(Graph G){
        
    	InitStack(S);
    	for(i = 0;i < G.vexnum; i++){
        
    		if(indegrdd[i]==0)
    			Push(S, i);
    	}
     
    	int count =0;
    	while(!Empty(S)){
        
    		Pop(S,i);
    		print[count++] = i;
    		for(p = G.vertices[i].firstarc; p; p = p->nextarc){
        
    			v = p->adjvex;
    			if(!(--indegree[v]))
    				Push(S, v);
    		}
    	}
     
    	if(count < G.vexnum)
    		return false;
    	else
    		return true;
    }

上面加粗的这句话即是拓扑排序的核心。一般深度学习模型也是一个DAG（有向无环图），我们这里同样使用了拓扑排序算法使得我们在一对一转换OP时和真实的网络结构是完全一致的。另外考虑到这里可能插入了一些新的节点如Identity可能会破坏原Graph的拓扑序，以及时刻需要判断计算图是否是一个完整合法的DAG，使用拓扑排序都是没有坏处的。

完成拓扑排序之后我们就可以执行`FlowOnnxMapping`完成OneFlow OP和ONNX OP的一对一转换了，代码如下：

def FlowOnnxMapping(g, ops_mapping):
        logger.debug("Mapping Oneflow node to ONNX node(s)")
        mapped_op = collections.Counter()
        unmapped_op = collections.Counter()
        exceptions = []
    
        ops = list(g.get_nodes())
        for node in ops:
            logger.debug("Process node: %s\n%s", node.name, node.summary)
    
            if node.skip_conversion:
                logger.debug("explicitly skip node " + node.name)
                continue
    
            op = node.op_type
            map_info = ops_mapping.get(op)
            if map_info is None:
                unmapped_op[op] += 1
                logger.error("oneflow op [%s: %s] is not supported", node.name, op)
                continue
            mapped_op[op] += 1
    
            func, onnx_op, kwargs = map_info
            if onnx_op is not None:
                node.op_type = onnx_op
            try:
                func(g, node, **kwargs)
                node.skip_conversion = True
            except Exception as ex:
                logger.error(
                    "Failed to convert node %s\n%s", node.name, node.summary, exc_info=1
                )
                exceptions.append(ex)
    
        return mapped_op, unmapped_op, exceptions

执行完这个函数会返回map上的OP容器，以及没有map上的OP容器，当然如果`Graph`中有OP没有map上也就是转换失败会抛出错误信息给用户。在转换完成之后，我们调用`Graph`中的每个`Node`的`UpdateProto()`构造函数将之前的假ONNX节点信息更新成真的ONNX节点信息。

接下来，我们调用各种 optimizer 优化网络结构，例如尽可能消除 nhwc->nchw 带来的 transpose op（Export 函数内的 optimizer.OptimizeGraph），即`https://github.com/Oneflow-Inc/oneflow_convert_tools/blob/main/oneflow_onnx/oneflow2onnx/flow2onnx.py#L264`。在oneflow-onnx里面主要有以下几种optimizer：

![oneflow-onnx里面的optimizer，获得更优的ONNX模型][oneflow-onnx_optimizer_ONNX]

这些 optimizer 继承自 tensorflow-onnx，我们后续会将其中的一部分用 onnx 原生的 optimizer 替代。

在优化了ONNX模型之后，最后调用下面的函数取磁盘中保存的 oneflow 权重，赋给 onnx 模型对象，并返回 protobuf 格式的 onnx 模型对象。至此就完成了创建合法的ONNX模型。

model_proto = onnx_graph.MakeModel(
                    job_name, onnx_filename, external_data=external_data
                )

我们的X2OneFlow分为X2ONNX和ONNX2Oneflow两个步骤，其中ONNX2OneFlow和OneFlow2ONNX共用了一套基础代码，所以需要修改的地方仅仅是将`handles`里面的注册OP转换的装饰器改个方向即可，这里不再赘述。

想了解更多细节可以看我们的源码`https://github.com/Oneflow-Inc/oneflow_convert_tools/tree/main/oneflow_onnx`。

# 0x4. 总结 #

在这篇文章中，我们分享了最近维护OneFlow和ONNX做的一系列工作，这项工作目前已经开源并在持续维护中，欢迎对OneFlow框架以及ONNX感兴趣的小伙伴体验和PR。点击阅读原文快速关注和体验oneflow-onnx工具。

# 0x5. 相关链接 #

*  OneFlow：https://github.com/Oneflow-Inc/oneflow
 *  https://github.com/onnx/tensorflow-onnx
 *  https://github.com/OI-wiki/OI-wiki

[ONNX]: https://mp.weixin.qq.com/s/FWEmk3PmaMhIG0GwaAOcjA
[ONNX 1]: https://mp.weixin.qq.com/s/_iNhfZNR5-swXLhHKjYRkQ
[onnx2pytorch_onnx-simplifier]: https://mp.weixin.qq.com/s/NDv-quXeBrPeDcCbg97FHA
[onnx simplifier _ optimizer]: https://mp.weixin.qq.com/s/q0Aa2LRpeCPCnIzRJbMmaQ
[oneflow-onnx_optimizer_ONNX]: /images/20221022/8377c7e2a30c432b80153a5224b35a42.png