Merge pull request #342 from zhiminzhang0830/fix_doc

fix docs in transformer examples

Author: HydrogenSulfate

docs/zh/examples/cylinder2d_unsteady_transformer_physx.md

@@ -112,19 +112,19 @@ $$Re \sim(100, 750)$$
 
 首先展示代码中定义的各个参数变量，每个参数的具体含义会在下面使用到时进行解释。
 
-``` py linenums="51" title="examples/cylinder/2d_unsteady/transformer_physx/train_enn.py"
+``` py linenums="50" title="examples/cylinder/2d_unsteady/transformer_physx/train_enn.py"
 --8<--
-examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:51:62
+examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:50:65
 --8<--
 ```
 
 #### 3.2.1 约束构建
 
 本案例基于数据驱动的方法求解问题，因此需要使用 PaddleScience 内置的 `SupervisedConstraint` 构建监督约束。在定义约束之前，需要首先指定监督约束中用于数据加载的各个参数，代码如下：
 
-``` py linenums="67" title="examples/cylinder/2d_unsteady/transformer_physx/train_enn.py"
+``` py linenums="70" title="examples/cylinder/2d_unsteady/transformer_physx/train_enn.py"
 --8<--
-examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:67:84
+examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:70:87
 --8<--
 ```
 
@@ -143,9 +143,9 @@ examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:67:84
 
 定义监督约束的代码如下：
 
-``` py linenums="86" title="examples/cylinder/2d_unsteady/transformer_physx/train_enn.py"
+``` py linenums="89" title="examples/cylinder/2d_unsteady/transformer_physx/train_enn.py"
 --8<--
-examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:86:94
+examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:89:97
 --8<--
 ```
 
@@ -168,37 +168,37 @@ examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:86:94
 
 用 PaddleScience 代码表示如下：
 
-``` py linenums="99" title="examples/cylinder/2d_unsteady/transformer_physx/train_enn.py"
+``` py linenums="102" title="examples/cylinder/2d_unsteady/transformer_physx/train_enn.py"
 --8<--
-examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:99:105
+examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:102:108
 --8<--
 ```
 
 其中，`CylinderEmbedding` 的前两个参数在前文中已有描述，这里不再赘述，网络模型的第三、四个参数是训练数据集的均值和方差，用于归一化输入数据。计算均值、方差的的代码表示如下：
 
-``` py linenums="28" title="examples/cylinder/2d_unsteady/transformer_physx/train_enn.py"
+``` py linenums="29" title="examples/cylinder/2d_unsteady/transformer_physx/train_enn.py"
 --8<--
-examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:28:45
+examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:29:46
 --8<--
 ```
 
 #### 3.2.3 学习率与优化器构建
 
 本案例中使用的学习率方法为 `ExponentialDecay`，学习率大小设置为0.001。优化器使用 `Adam`，梯度裁剪使用了 Paddle 内置的 `ClipGradByGlobalNorm` 方法。用 PaddleScience 代码表示如下：
 
-``` py linenums="107" title="examples/cylinder/2d_unsteady/transformer_physx/train_enn.py"
+``` py linenums="110" title="examples/cylinder/2d_unsteady/transformer_physx/train_enn.py"
 --8<--
-examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:107:121
+examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:110:124
 --8<--
 ```
 
 #### 3.2.4 评估器构建
 
 本案例训练过程中会按照一定的训练轮数间隔，使用验证集评估当前模型的训练情况，需要使用 `SupervisedValidator` 构建评估器。代码如下：
 
-``` py linenums="123" title="examples/cylinder/2d_unsteady/transformer_physx/train_enn.py"
+``` py linenums="126" title="examples/cylinder/2d_unsteady/transformer_physx/train_enn.py"
 --8<--
-examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:123:150
+examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:126:153
 --8<--
 ```
 
@@ -208,39 +208,39 @@ examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:123:150
 
 完成上述设置之后，只需要将上述实例化的对象按顺序传递给 `ppsci.solver.Solver`，然后启动训练、评估。
 
-``` py linenums="153" title="examples/cylinder/2d_unsteady/transformer_physx/train_enn.py"
+``` py linenums="156" title="examples/cylinder/2d_unsteady/transformer_physx/train_enn.py"
 --8<--
-examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:153:
+examples/cylinder/2d_unsteady/transformer_physx/train_enn.py:156:
 --8<--
 ```
 
 ### 3.3 Transformer 模型
 
 上文介绍了如何构建 Embedding 模型的训练、评估，在本节中将介绍如何使用训练好的 Embedding 模型训练 Transformer 模型。因为训练 Transformer 模型的步骤与训练 Embedding 模型的步骤基本相似，因此本节在两者的重复部分的各个参数不再详细介绍。首先将代码中定义的各个参数变量展示如下，每个参数的具体含义会在下面使用到时进行解释。
 
-``` py linenums="58" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
+``` py linenums="57" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
 --8<--
-examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:58:74
+examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:57:79
 --8<--
 ```
 
 #### 3.3.1 约束构建
 
 Transformer 模型同样基于数据驱动的方法求解问题，因此需要使用 PaddleScience 内置的 `SupervisedConstraint` 构建监督约束。在定义约束之前，需要首先指定监督约束中用于数据加载的各个参数，代码如下：
 
-``` py linenums="82" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
+``` py linenums="87" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
 --8<--
-examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:82:99
+examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:87:104
 --8<--
 ```
 
 数据加载的各个参数与 Embedding 模型中的基本一致，不再赘述。需要说明的是由于 Transformer 模型训练的输入数据是 Embedding 模型 Encoder 模块的输出数据，因此我们将训练好的 Embedding 模型作为 `CylinderDataset` 的一个参数，在初始化时首先将训练数据映射到编码空间。
 
 定义监督约束的代码如下：
 
-``` py linenums="101" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
+``` py linenums="106" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
 --8<--
-examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:101:106
+examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:106:111
 --8<--
 ```
 
@@ -255,9 +255,9 @@ examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:101:106
 
 用 PaddleScience 代码表示如下：
 
-``` py linenums="111" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
+``` py linenums="116" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
 --8<--
-examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:111:119
+examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:116:124
 --8<--
 ```
 
@@ -267,19 +267,19 @@ examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:111:119
 
 本案例中使用的学习率方法为 `CosineWarmRestarts`，学习率大小设置为0.001。优化器使用 `Adam`，梯度裁剪使用了 Paddle 内置的 `ClipGradByGlobalNorm` 方法。用 PaddleScience 代码表示如下：
 
-``` py linenums="121" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
+``` py linenums="126" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
 --8<--
-examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:121:135
+examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:126:140
 --8<--
 ```
 
 #### 3.3.4 评估器构建
 
 训练过程中会按照一定的训练轮数间隔，使用验证集评估当前模型的训练情况，需要使用 `SupervisedValidator` 构建评估器。用 PaddleScience 代码表示如下：
 
-``` py linenums="137" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
+``` py linenums="142" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
 --8<--
-examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:137:163
+examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:142:168
 --8<--
 ```
 
@@ -289,25 +289,25 @@ examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:137:163
 
 在本文中首先定义了对 Transformer 模型输出数据变换到物理状态空间的代码：
 
-``` py linenums="32" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
+``` py linenums="33" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
 --8<--
-examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:32:52
+examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:33:53
 --8<--
 ```
 
-``` py linenums="78" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
+``` py linenums="83" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
 --8<--
-examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:78:79
+examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:83:84
 --8<--
 ```
 
 可以看到，程序首先载入了训练好的 Embedding 模型，然后在 `OutputTransform` 的 `__call__` 函数内实现了编码向量到物理状态空间的变换。
 
 在定义好了以上代码之后，就可以实现可视化器代码的构建了：
 
-``` py linenums="165" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
+``` py linenums="170" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
 --8<--
-examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:165:193
+examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:170:197
 --8<--
 ```
 
@@ -317,9 +317,9 @@ examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:165:193
 
 完成上述设置之后，只需要将上述实例化的对象按顺序传递给 `ppsci.solver.Solver`，然后启动训练、评估。
 
-``` py linenums="193" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
+``` py linenums="199" title="examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py"
 --8<--
-examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:193:
+examples/cylinder/2d_unsteady/transformer_physx/train_transformer.py:199:
 --8<--
 ```