PaddlePaddle · zhiminzhang0830 · Jun 24, 2024 · Apr 25, 2024 · Apr 25, 2024 · Apr 26, 2024
diff --git a/docs/zh/examples/earthformer.md b/docs/zh/examples/earthformer.md
@@ -10,9 +10,9 @@
 === "模型训练命令"
 
     ``` sh
-    # ICAR-ENSO 数据预训练模型
+    # ICAR-ENSO 数据模型训练
     python examples/earthformer/earthformer_enso_train.py
-    # SEVIR 数据集预训练模型
+    # SEVIR 数据模型训练
     python examples/earthformer/earthformer_sevir_train.py
 
     ```
@@ -74,7 +74,7 @@ EarthFormer 原代码中训练了 ICAR-ENSO 数据集中海面温度 (sst) 和 S
 
 ### 2.1 ICAR-ENSO 和 SEVIR 模型的训练、推理过程
 
-模型预训练阶段是基于随机初始化的网络权重对模型进行训练，如下图所示，其中 $[x_{i}]_{i=1}^{T}$ 表示长度为 $T$ 时空序列的输入气象数据，$[y_{i}]_{i=1}^{K}$ 表示预测未来 $K$ 步的气象数据，$[y_{i_True}]_{i=1}^{K}$ 表示未来 $K$ 步的真实数据，如海面温度数据和云总降水量数据。最后网络模型预测的输出和真值计算 mse 损失函数。
+模型预训练阶段是基于随机初始化的网络权重对模型进行训练，如下图所示，其中 $[x_{i}]_{i=1}^{T}$ 表示长度为 $T$ 时空序列的输入气象数据，$[y_{T+i}]_{i=1}^{K}$ 表示预测未来 $K$ 步的气象数据，$[y_{T+i_true}]_{i=1}^{K}$ 表示未来 $K$ 步的真实数据，如海面温度数据和云总降水量数据。最后网络模型预测的输出和真值计算 mse 损失函数。
 
 <figure markdown>
   ![earthformer-pretraining](https://paddle-org.bj.bcebos.com/paddlescience/docs/earthformer/earthformer-pretrain.png){ loading=lazy style="margin:0 auto;height:70%;width:70%"}

diff --git a/docs/zh/examples/neuraloperator.md b/docs/zh/examples/neuraloperator.md
diff --git a/mkdocs.yml b/mkdocs.yml
@@ -49,6 +49,7 @@ nav:
         - Volterra_IDE: zh/examples/volterra_ide.md
         - NLSMB: zh/examples/nlsmb.md
         - XPINN: zh/examples/xpinns.md
+        - NeuralOperator: zh/examples/neuraloperator.md
       - 技术科学(AI for Technology):
         - 流体:
           - AMGNet: zh/examples/amgnet.md

diff --git a/ppsci/paddle_harmonics/legendre.py → ppsci/arch/paddle_harmonics/legendre.py b/ppsci/paddle_harmonics/legendre.py → ppsci/arch/paddle_harmonics/legendre.py
diff --git a/ppsci/paddle_harmonics/quadrature.py → ppsci/arch/paddle_harmonics/quadrature.py b/ppsci/paddle_harmonics/quadrature.py → ppsci/arch/paddle_harmonics/quadrature.py
diff --git a/ppsci/paddle_harmonics/random_fields.py → ppsci/arch/paddle_harmonics/random_fields.py b/ppsci/paddle_harmonics/random_fields.py → ppsci/arch/paddle_harmonics/random_fields.py
@@ -35,7 +35,7 @@
 import paddle
 from paddle import nn
 
-from ppsci.paddle_harmonics.sht import InverseRealSHT
+from ppsci.arch.paddle_harmonics.sht import InverseRealSHT
 
 
 class GaussianRandomFieldS2(nn.Layer):

diff --git a/ppsci/paddle_harmonics/sht.py → ppsci/arch/paddle_harmonics/sht.py b/ppsci/paddle_harmonics/sht.py → ppsci/arch/paddle_harmonics/sht.py
@@ -39,11 +39,11 @@
 import paddle.fft
 import paddle.nn as nn
 
-from ppsci.paddle_harmonics.legendre import _precompute_dlegpoly
-from ppsci.paddle_harmonics.legendre import _precompute_legpoly
-from ppsci.paddle_harmonics.quadrature import clenshaw_curtiss_weights
-from ppsci.paddle_harmonics.quadrature import legendre_gauss_weights
-from ppsci.paddle_harmonics.quadrature import lobatto_weights
+from ppsci.arch.paddle_harmonics.legendre import _precompute_dlegpoly
+from ppsci.arch.paddle_harmonics.legendre import _precompute_legpoly
+from ppsci.arch.paddle_harmonics.quadrature import clenshaw_curtiss_weights
+from ppsci.arch.paddle_harmonics.quadrature import legendre_gauss_weights
+from ppsci.arch.paddle_harmonics.quadrature import lobatto_weights
 
 
 class RealSHT(nn.Layer):

diff --git a/ppsci/arch/sfnonet.py b/ppsci/arch/sfnonet.py
@@ -10,7 +10,7 @@
 
 from ppsci.arch import base
 from ppsci.arch import fno_block
-from ppsci.paddle_harmonics import sht as paddle_sht
+from ppsci.arch.paddle_harmonics import sht as paddle_sht
 from ppsci.utils import initializer
 
 einsum_symbols = "abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ"