update final code

PaddlePaddle · Oct 18, 2023 · 7eb4aab · 7eb4aab
1 parent a59bc28
commit 7eb4aab
Show file tree

Hide file tree

Showing 7 changed files with 418 additions and 137 deletions.
diff --git a/docs/index.md b/docs/index.md
@@ -10,29 +10,29 @@
 
 ![panorama](https://paddle-org.bj.bcebos.com/paddlescience/docs/overview/panorama.png)
 
-!!! tip "快速安装"
+--8<--
+./README.md:update
+--8<--
 
-    === "方式1: 源码安装[推荐]"
+## 快速安装
 
-        --8<--
-        ./README.md:git_install
-        --8<--
+=== "方式1: 源码安装[推荐]"
 
-    === "方式2: pip安装"
+    --8<--
+    ./README.md:git_install
+    --8<--
 
-        ``` shell
-        pip install paddlesci
-        ```
+=== "方式2: pip安装"
 
-    === "[完整安装流程](./zh/install_setup.md)"
+    ``` shell
+    pip install paddlesci
+    ```
 
-        ``` shell
-        pip install paddlesci
-        ```
+=== "[完整安装流程](./zh/install_setup.md)"
 
---8<--
-./README.md:update
---8<--
+    ``` shell
+    pip install paddlesci
+    ```
 
 --8<--
 ./README.md:feature

diff --git a/docs/zh/examples/amgnet.md b/docs/zh/examples/amgnet.md
@@ -1,7 +1,62 @@
-# 2D-Darcy
+# AMGNet
 
 <!-- <a href="https://aistudio.baidu.com/aistudio/projectdetail/6184070?contributionType=1&sUid=438690&shared=1&ts=1684239806160" class="md-button md-button--primary" style>AI Studio快速体验</a> -->
 
+=== "模型训练命令"
+
+    === "amgnet_airfoil"
+
+        ``` sh
+        # linux
+        wget https://paddle-org.bj.bcebos.com/paddlescience/datasets/AMGNet/data.zip
+        # windows
+        # curl https://paddle-org.bj.bcebos.com/paddlescience/datasets/AMGNet/data.zip --output data.zip
+        # unzip it
+        unzip data.zip
+        python amgnet_airfoil.py
+        ```
+    === "amgnet_cylinder"
+
+        ``` sh
+        # linux
+        wget https://paddle-org.bj.bcebos.com/paddlescience/datasets/AMGNet/data.zip
+        # windows
+        # curl https://paddle-org.bj.bcebos.com/paddlescience/datasets/AMGNet/data.zip --output data.zip
+        # unzip it
+        unzip data.zip
+        python amgnet_cylinder.py
+        ```
+
+=== "模型评估命令"
+
+    === "amgnet_airfoil"
+
+        ``` sh
+        # linux
+        wget https://paddle-org.bj.bcebos.com/paddlescience/datasets/AMGNet/data.zip
+        # windows
+        # curl https://paddle-org.bj.bcebos.com/paddlescience/datasets/AMGNet/data.zip --output data.zip
+        # unzip it
+        unzip data.zip
+        python amgnet_airfoil.py mode=eval EVAL.pretained_model_path=https://paddle-org.bj.bcebos.com/paddlescience/models/amgnet/amgnet_airfoil_pretrained.pdparams
+        ```
+    === "amgnet_cylinder"
+
+        ``` sh
+        # linux
+        wget https://paddle-org.bj.bcebos.com/paddlescience/datasets/AMGNet/data.zip
+        # windows
+        # curl https://paddle-org.bj.bcebos.com/paddlescience/datasets/AMGNet/data.zip --output data.zip
+        # unzip it
+        unzip data.zip
+        python amgnet_cylinder.py mode=eval EVAL.pretrained_model_path=https://paddle-org.bj.bcebos.com/paddlescience/models/amgnet/amgnet_cylinder_pretrained.pdparams
+        ```
+
+| 预训练模型  | 指标 |
+|:--| :--|
+| [amgnet_airfoil_pretrained.pdparams](https://paddle-org.bj.bcebos.com/paddlescience/models/amgnet/amgnet_airfoil_pretrained.pdparams) | loss(RMSE_validator): 0.0001 <br> RMSE.RMSE(RMSE_validator): 0.01315 |
+| [amgnet_cylinder_pretrained.pdparams](https://paddle-org.bj.bcebos.com/paddlescience/models/amgnet/amgnet_cylinder_pretrained.pdparams) | loss(RMSE_validator): 0.00048 <br> RMSE.RMSE(RMSE_validator): 0.02197 |
+
 ## 1. 背景简介
 
 近年来，深度学习在计算机视觉和自然语言处理方面的成功应用，促使人们探索人工智能在科学计算领域的应用，尤其是在计算流体力学(CFD)领域的应用。
@@ -29,7 +84,7 @@
 
 !!! info "注意事项"
 
-    本案例运行前需通过 `pip install pgl` 命令，安装 [**P**addle **G**raph **L**earning](https://github.com/PaddlePaddle/PGL) 图学习工具
+    本案例运行前需通过 `pip install pgl pyamg` 命令，安装 [**P**addle **G**raph **L**earning](https://github.com/PaddlePaddle/PGL) 图学习工具和 [PyAMG](https://github.com/pyamg/pyamg) 代数多重网格工具。
 
 ### 3.1 数据集下载
 
@@ -48,17 +103,17 @@ unzip data.zip
 
 === "airfoil"
 
-    ``` py linenums="60"
+    ``` py linenums="61"
     --8<--
-    examples/amgnet/amgnet_airfoil.py:60:71
+    examples/amgnet/amgnet_airfoil.py:61:62
     --8<--
     ```
 
 === "cylinder"
 
-    ``` py linenums="60"
+    ``` py linenums="61"
     --8<--
-    examples/amgnet/amgnet_cylinder.py:60:71
+    examples/amgnet/amgnet_cylinder.py:61:62
     --8<--
     ```
 
@@ -68,77 +123,77 @@ unzip data.zip
 
 在本案例中，我们使用监督数据集对模型进行训练，因此需要构建监督约束。
 
-在定义约束之前，我们需要指定数据集的路径、`batch_size` 等相关配置，将这些信息封装到字典中，如下所示。
+在定义约束之前，我们需要指定数据集的路径等相关配置，将这些信息存放到对应的 YAML 文件中，如下所示。
 
 === "airfoil"
 
-    ``` py linenums="73"
+    ``` py linenums="21"
     --8<--
-    examples/amgnet/amgnet_airfoil.py:73:90
+    examples/amgnet/conf/amgnet_airfoil.yaml:21:27
     --8<--
     ```
 
 === "cylinder"
 
-    ``` py linenums="73"
+    ``` py linenums="21"
     --8<--
-    examples/amgnet/amgnet_cylinder.py:73:90
+    examples/amgnet/conf/amgnet_cylinder.yaml:21:27
     --8<--
     ```
 
 接着定义训练损失函数的计算过程，如下所示。
 
 === "airfoil"
 
-    ``` py linenums="33"
+    ``` py linenums="35"
     --8<--
-    examples/amgnet/amgnet_airfoil.py:33:36
+    examples/amgnet/amgnet_airfoil.py:35:40
     --8<--
     ```
 
 === "cylinder"
 
-    ``` py linenums="33"
+    ``` py linenums="35"
     --8<--
-    examples/amgnet/amgnet_cylinder.py:33:36
+    examples/amgnet/amgnet_cylinder.py:35:40
     --8<--
     ```
 
 最后构建监督约束，如下所示。
 
 === "airfoil"
 
-    ``` py linenums="92"
+    ``` py linenums="82"
     --8<--
-    examples/amgnet/amgnet_airfoil.py:92:100
+    examples/amgnet/amgnet_airfoil.py:82:90
     --8<--
     ```
 
 === "cylinder"
 
-    ``` py linenums="92"
+    ``` py linenums="82"
     --8<--
-    examples/amgnet/amgnet_cylinder.py:92:100
+    examples/amgnet/amgnet_cylinder.py:82:90
     --8<--
     ```
 
 ### 3.4 超参数设定
 
-接下来我们需要指定训练轮数，此处使用 500 轮训练轮数。
+设置训练轮数等参数，如下所示。
 
 === "airfoil"
 
-    ``` py linenums="102"
+    ``` py linenums="41"
     --8<--
-    examples/amgnet/amgnet_airfoil.py:102:103
+    examples/amgnet/conf/amgnet_airfoil.yaml:41:51
     --8<--
     ```
 
 === "cylinder"
 
-    ``` py linenums="102"
+    ``` py linenums="41"
     --8<--
-    examples/amgnet/amgnet_cylinder.py:102:103
+    examples/amgnet/conf/amgnet_cylinder.yaml:41:51
     --8<--
     ```
 
@@ -148,53 +203,53 @@ unzip data.zip
 
 === "airfoil"
 
-    ``` py linenums="105"
+    ``` py linenums="92"
     --8<--
-    examples/amgnet/amgnet_airfoil.py:105:106
+    examples/amgnet/amgnet_airfoil.py:92:93
     --8<--
     ```
 
 === "cylinder"
 
-    ``` py linenums="105"
+    ``` py linenums="92"
     --8<--
-    examples/amgnet/amgnet_cylinder.py:105:106
+    examples/amgnet/amgnet_cylinder.py:92:93
     --8<--
     ```
 
 ### 3.6 评估器构建
 
-在训练过程中通常会按一定轮数间隔，用验证集(测试集)评估当前模型的训练情况，因此使用 `ppsci.validate.SupervisedValidator` 构建评估器，构建过程与 [约束构建](#34) 类似，只需把数据目录改为测试集的目录，并设置 `batch_size=1` 即可。
+在训练过程中通常会按一定轮数间隔，用验证集(测试集)评估当前模型的训练情况，因此使用 `ppsci.validate.SupervisedValidator` 构建评估器，构建过程与 [约束构建](#34) 类似，只需把数据目录改为测试集的目录，并在配置文件中设置 `EVAL.batch_size=1` 即可。
 
 === "airfoil"
 
-    ``` py linenums="108"
+    ``` py linenums="95"
     --8<--
-    examples/amgnet/amgnet_airfoil.py:108:131
+    examples/amgnet/amgnet_airfoil.py:95:118
     --8<--
     ```
 === "cylinder"
 
-    ``` py linenums="108"
+    ``` py linenums="95"
     --8<--
-    examples/amgnet/amgnet_cylinder.py:108:131
+    examples/amgnet/amgnet_cylinder.py:95:118
     --8<--
     ```
 
 评估指标为预测结果和真实结果的 RMSE 值，因此需自定义指标计算函数，如下所示。
 
 === "airfoil"
 
-    ``` py linenums="39"
+    ``` py linenums="43"
     --8<--
-    examples/amgnet/amgnet_airfoil.py:39:48
+    examples/amgnet/amgnet_airfoil.py:43:52
     --8<--
     ```
 === "cylinder"
 
-    ``` py linenums="39"
+    ``` py linenums="43"
     --8<--
-    examples/amgnet/amgnet_cylinder.py:39:48
+    examples/amgnet/amgnet_cylinder.py:43:52
     --8<--
     ```
 
@@ -204,16 +259,16 @@ unzip data.zip
 
 === "airfoil"
 
-    ``` py linenums="133"
+    ``` py linenums="120"
     --8<--
-    examples/amgnet/amgnet_airfoil.py:133:149
+    examples/amgnet/amgnet_airfoil.py:120:136
     --8<--
     ```
 === "cylinder"
 
-    ``` py linenums="133"
+    ``` py linenums="120"
     --8<--
-    examples/amgnet/amgnet_cylinder.py:133:149
+    examples/amgnet/amgnet_cylinder.py:120:136
     --8<--
     ```
 
@@ -223,16 +278,16 @@ unzip data.zip
 
 === "airfoil"
 
-    ``` py linenums="151"
+    ``` py linenums="138"
     --8<--
-    examples/amgnet/amgnet_airfoil.py:151:
+    examples/amgnet/amgnet_airfoil.py:138:
     --8<--
     ```
 === "cylinder"
 
-    ``` py linenums="151"
+    ``` py linenums="138"
     --8<--
-    examples/amgnet/amgnet_cylinder.py:151:
+    examples/amgnet/amgnet_cylinder.py:138:
     --8<--
     ```
 
@@ -276,7 +331,7 @@ unzip data.zip
         ![Cylinder_0_p](https://paddle-org.bj.bcebos.com/paddlescience/docs/AMGNet/cylinder_0field.png1_field.png){ loading=lazy }
         <figcaption>左：预测压力 p，右：实际压力 p</figcaption>
         ![Cylinder_0_vec_y](https://paddle-org.bj.bcebos.com/paddlescience/docs/AMGNet/cylinder_0field.png2_field.png){ loading=lazy }
-        <figcaption>左：预测y方向流速 p，右：实际 y 方向流速</figcaption>
+        <figcaption>左：预测 y 方向流速 p，右：实际 y 方向流速</figcaption>
     </figure>
 
 可以看到模型预测结果与真实结果基本一致。

diff --git a/docs/zh/examples/bracket.md b/docs/zh/examples/bracket.md
@@ -28,7 +28,7 @@
 
 | 预训练模型  | 指标 |
 |:--| :--|
-| [bracket_pretrained.pdparams](https://paddle-org.bj.bcebos.com/paddlescience/models/bracket/bracket_pretrained.pdparams) | loss(commercial_ref_u_v_w_sigmas): 32.28704, MSE.u(commercial_ref_u_v_w_sigmas): 0.00005, MSE.v(commercial_ref_u_v_w_sigmas): 0.00000, MSE.w(commercial_ref_u_v_w_sigmas): 0.00734, MSE.sigma_xx(commercial_ref_u_v_w_sigmas): 27.64751, MSE.sigma_yy(commercial_ref_u_v_w_sigmas): 1.23101, MSE.sigma_zz(commercial_ref_u_v_w_sigmas): 0.89106, MSE.sigma_xy(commercial_ref_u_v_w_sigmas): 0.84370, MSE.sigma_xz(commercial_ref_u_v_w_sigmas): 1.42126, MSE.sigma_yz(commercial_ref_u_v_w_sigmas): 0.24510 |
+| [bracket_pretrained.pdparams](https://paddle-org.bj.bcebos.com/paddlescience/models/bracket/bracket_pretrained.pdparams) | loss(commercial_ref_u_v_w_sigmas): 32.28704 </br> MSE.u(commercial_ref_u_v_w_sigmas): 0.00005 </br> MSE.v(commercial_ref_u_v_w_sigmas): 0.00000 </br> MSE.w(commercial_ref_u_v_w_sigmas): 0.00734 </br> MSE.sigma_xx(commercial_ref_u_v_w_sigmas): 27.64751 </br> MSE.sigma_yy(commercial_ref_u_v_w_sigmas): 1.23101 </br> MSE.sigma_zz(commercial_ref_u_v_w_sigmas): 0.89106 </br> MSE.sigma_xy(commercial_ref_u_v_w_sigmas): 0.84370 </br> MSE.sigma_xz(commercial_ref_u_v_w_sigmas): 1.42126 </br> MSE.sigma_yz(commercial_ref_u_v_w_sigmas): 0.24510 |
 
 ## 1. 背景简介