PaddlePaddle · ooooo-create · Oct 29, 2025 · Oct 29, 2025
diff --git a/_typos.toml b/_typos.toml
@@ -37,9 +37,6 @@ Moible = "Moible"
 Operaton = "Operaton"
 Optimizaing = "Optimizaing"
 Optimzier = "Optimzier"
-REGISTE = "REGISTE"
-Reivew = "Reivew"
-Reuqest = "Reuqest"
 Setment = "Setment"
 Similarily = "Similarily"
 Simle = "Simle"
@@ -99,6 +96,7 @@ fundemental = "fundemental"
 funtion = "funtion"
 ilter = "ilter"
 inferface = "inferface"
+infor = "infor"
 instert = "instert"
 instrinsics = "instrinsics"
 interchangable = "interchangable"
@@ -135,15 +133,6 @@ overrided = "overrided"
 overwrited = "overwrited"
 porcess = "porcess"
 processer = "processer"
-recevied = "recevied"
-recomment = "recomment"
-registerd = "registerd"
-registring = "registring"
-reivew = "reivew"
-relase = "relase"
-requirments = "requirments"
-resut = "resut"
-resuts = "resuts"
 sacle = "sacle"
 samle = "samle"
 satifies = "satifies"

diff --git a/docs/api/gen_alias_api.py b/docs/api/gen_alias_api.py
@@ -121,7 +121,7 @@ def transform_file_to_api(self, file_path, api_last_name):
         return api
 
     # find the shortest path api which not starts with paddle.fluid
-    def _choose_recomment_api(self, api_list):
+    def _choose_recommend_api(self, api_list):
         min_len = len(api_list[0].split("."))
         rec_api = api_list[0]
         for api in api_list:
@@ -135,7 +135,7 @@ def sort_alias_name(self, api_list):
         real_api = self._choose_real_api(api_list)
         api_list.remove(real_api)
 
-        rec_api = self._choose_recomment_api(api_list)
+        rec_api = self._choose_recommend_api(api_list)
         api_list.remove(rec_api)
 
         # sort others api by path length

diff --git a/docs/design/mkldnn/acquire_api/acquire_api.md b/docs/design/mkldnn/acquire_api/acquire_api.md
@@ -70,7 +70,7 @@ Having memory objects and computational primitive we may trigger its execution .
 #### 5. Registering MKL-DNN memory format in corresponding Tensor
 Last step is to register MKL-DNN output memory object format inside of Output tensor eg. set Tensor::format_ to MKL-DNN enum that corresponds the way Tensor data is arranged (NCHW, NCHW16C etc.) This enum can be taken from dst memory object (wrapper to Output tensor) in Forward pass or from diff_src memory object (wrapper to X_grad Tensor).
 
-Example of registring MKL-DNN format in output tensor:
+Example of registering MKL-DNN format in output tensor:
 
     out->set_layout(framework::DataLayout::kMKLDNN);
     out->set_format(platform::GetMKLDNNFormat(*dst_memory));
diff --git a/docs/design/mkldnn/gru/gru.md b/docs/design/mkldnn/gru/gru.md
@@ -95,7 +95,7 @@ oneDNN `GRU` operator is based on Paddle Paddle `fusion_gru` operator. It uses p
 * PaddlePaddle Input LoD -> oneDNN TNC/NTC\
 Every time before executing `GRU`, each batch represented by PP tensor has to be converted into oneDNN tensor representation. It is done first by calculating length of the longest sentence in a batch to get `T` dimension and then creating oneDNN memory (or getting it from oneDNN cache). After that, based on the memory format chosen by oneDNN GRU primitive: `TNC` or `NTC` correct custom reorder is called.\
 \
-Because oneDNN assumes that all sentences are of equal length, before reorder, whole memory is set to 0 to add padding. Placement of this padding depends also on computation direction that is defined by `is_reverse` attribute. To get correct resuts, if computation is performed from left to right, the padding has to be on the right side of words. Otherwise, for right to left computation, it has to be on the left side.
+Because oneDNN assumes that all sentences are of equal length, before reorder, whole memory is set to 0 to add padding. Placement of this padding depends also on computation direction that is defined by `is_reverse` attribute. To get correct results, if computation is performed from left to right, the padding has to be on the right side of words. Otherwise, for right to left computation, it has to be on the left side.
 ![](images/input_is_reverse.svg)
 
 * PaddlePaddle WeightX -> oneDNN WeightX\

diff --git a/docs/design/others/releasing_process.md b/docs/design/others/releasing_process.md
@@ -43,7 +43,7 @@ PaddlePaddle 开发过程使用[git-flow](http://nvie.com/posts/a-successful-git
 * 其他用户的 fork 版本库并不需要严格遵守[git-flow](http://nvie.com/posts/a-successful-git-branching-model/)分支规范，但所有 fork 的版本库的所有分支都相当于特性分支。
     * 建议，开发者 fork 的版本库使用`develop`分支同步主版本库的`develop`分支
     * 建议，开发者 fork 的版本库中，再基于`develop`版本 fork 出自己的功能分支。
-    * 当功能分支开发完毕后，向 PaddlePaddle 的主版本库提交`Pull Reuqest`，进而进行代码评审。
+    * 当功能分支开发完毕后，向 PaddlePaddle 的主版本库提交`Pull Request`，进而进行代码评审。
         * 在评审过程中，开发者修改自己的代码，可以继续在自己的功能分支提交代码。
 
 * BugFix 分支也是在开发者自己的 fork 版本库维护，与功能分支不同的是，BugFix 分支需要分别给主版本库的`master`、`develop`与可能有的`release/版本号`分支，同时提起`Pull Request`。

diff --git a/docs/design/phi/design_cn.md b/docs/design/phi/design_cn.md
@@ -1670,7 +1670,7 @@ Fc(const Context& dev_ctx, const Tensor& x, const Tensor& w, const Tensor& b, Te
     phi::add<T, Context>(phi::mul<T，Context>(x, w), b, out);
 }
 
-PT_REGISTE_KERNEL("fc", Fc, ...)
+PT_REGISTER_KERNEL("fc", Fc, ...)
 ```
 
 mul 和 add 操作的拼接，代码量很少，再加一个注册声明。

diff --git a/docs/design/phi/kernel_migrate_en.md b/docs/design/phi/kernel_migrate_en.md
@@ -304,7 +304,7 @@ Differences include:
 1. The name of the registered macro is different, here is `PD_REGISTER_PLUGIN_KERNEL`.
 2. The name of the Backend is the name of the CustomDevice registered by the user, here is `ascend`.
 
-The kernel that registerd by REGISTER_OP_KERNEL_WITH_CUSTOM_TYPE in Fluid is also registered using `PD_REGISTER_KERNEL` or `PD_REGISTER_PLUGIN_KERNEL` in PHI. It should be noted that if there are two types of kernel template parameters registered in fluid, since the kernel registered in PHI only supports one type, the first type is used in PHI registration. For example, the example under fluid:
+The kernel that registered by REGISTER_OP_KERNEL_WITH_CUSTOM_TYPE in Fluid is also registered using `PD_REGISTER_KERNEL` or `PD_REGISTER_PLUGIN_KERNEL` in PHI. It should be noted that if there are two types of kernel template parameters registered in fluid, since the kernel registered in PHI only supports one type, the first type is used in PHI registration. For example, the example under fluid:
 
 ```c++
 

diff --git a/docs/dev_guides/git_guides/code_review_en.md b/docs/dev_guides/git_guides/code_review_en.md
@@ -1,4 +1,4 @@
-# Code Reivew promise
+# Code Review promise
 
 In order that reviewers focus on code in the code review,please follow these rules every time you submit your code:
 
@@ -8,11 +8,11 @@ In order that reviewers focus on code in the code review,please follow these rul
 
 In addition,please follow the following regulations in response to the suggestion of reviewers:
 
-1）A reply to every comment of reviewers（It's a fundamental complimentary conduct in open source community.An expression of appreciation is a need for help from others):
+1）A reply to every comment of reviewers(It's a fundamental complimentary conduct in open source community.An expression of appreciation is a need for help from others):
 
    - If you adopt the suggestion of reviewer and make a modification accordingly, it's courteous to reply with a simple `Done` .
 
-   - Please clarify your reason to the disagreenment
+   - Please clarify your reason to the disagreement
 
 2）If there are many suggestions
 

diff --git a/docs/dev_guides/op_optimization/kernel_primitive_api/add_example_cn.md b/docs/dev_guides/op_optimization/kernel_primitive_api/add_example_cn.md
@@ -39,7 +39,7 @@ __device__ void ElementwiseAddImpl(InT *in0, InT * in1, OutT * out, OpFunc func,
   ReadData<InT, InT, VecSize, 1, 1, IsBoundary>(arg0, in0, num);
   ReadData<InT, InT, VecSize, 1, 1, IsBoundary>(arg1, in1, num);
 
-  // compute resut[i] = args[i] + arg1[i]
+  // compute result[i] = args[i] + arg1[i]
   ElementwiseBinary<InT, OutT, VecSize, 1, 1, OpFunc>(result, arg0, arg1, func);
 
   // write data

diff --git a/docs/dev_guides/op_optimization/kernel_primitive_api/add_example_en.md b/docs/dev_guides/op_optimization/kernel_primitive_api/add_example_en.md
@@ -38,7 +38,7 @@ __device__ void ElementwiseAddImpl(InT *in0, InT * in1, OutT * out, OpFunc func,
   ReadData<InT, InT, VecSize, 1, 1, IsBoundary>(arg0, in0, num);
   ReadData<InT, InT, VecSize, 1, 1, IsBoundary>(arg1, in1, num);
 
-  // compute resut[i] = args[i] + arg1[i]
+  // compute result[i] = args[i] + arg1[i]
   ElementwiseBinary<InT, OutT, VecSize, 1, 1, OpFunc>(result, arg0, arg1, func);
 
   // write data

diff --git a/docs/eval/evaluation_of_docs_system.md b/docs/eval/evaluation_of_docs_system.md
@@ -779,7 +779,7 @@ Paddle 当前文档中的 Tutorial，特别是初级内容部分，都很好的
 - 模型保存
 - 推理
 
-这里面包含了很多“干货”的内容，但是我觉得放在“升级指南”中有“明珠暗投”的风险。因为一般人对“升级指南”的刻板印象，应该类似是“relase log”级别的东西，可能会不会优先点开。
+这里面包含了很多“干货”的内容，但是我觉得放在“升级指南”中有“明珠暗投”的风险。因为一般人对“升级指南”的刻板印象，应该类似是“release log”级别的东西，可能会不会优先点开。
 
 据我体验，升级指南中的 [使用高层 API](https://www.paddlepaddle.org.cn/documentation/docs/zh/guides/01_paddle2.0_introduction/update_cn.html#id4) 和 [使用基础 API](https://www.paddlepaddle.org.cn/documentation/docs/zh/guides/01_paddle2.0_introduction/update_cn.html#id5) 包含了其它地方没有的内容，这些内容被用户错过的话，是可惜的。
 

diff --git a/docs/faq/2.0.md b/docs/faq/2.0.md
@@ -35,7 +35,7 @@
 
 1. 删除依赖这些库的 API，原来 Paddle 基于这些第三方库提供了一些 API，在删除这些依赖库的同时也删除了这些 API，如：移除 matplotlib 的同时，paddle.utils.plot 也被删除了。
 
-2. 删除依赖第三方库的数据集，如：移除 nltk 的同时，依赖 nltk 的数据集 sentiment(nltk.movie_reivew)也被删除了。
+2. 删除依赖第三方库的数据集，如：移除 nltk 的同时，依赖 nltk 的数据集 sentiment(nltk.movie_review)也被删除了。
 
 3. 删除依赖的第三方库后，如果需要使用这些库，需要重新安装`pip install objgraph`，直接 import objgraph 可能会报错。
 

diff --git a/docs/guides/jit/limitations_cn.md b/docs/guides/jit/limitations_cn.md
@@ -351,7 +351,7 @@ TypeError: In transformed code:
             out.append(i)
         return out
 
-TypeError: All values in `initialized_list` should be Variable, but recevied <class 'list'>.
+TypeError: All values in `initialized_list` should be Variable, but received <class 'list'>.
 ```
 
 **修改建议**：推荐将 [1,2] 和 [2,3] 转为 tensor，变为单层 list。如下面写法：

diff --git a/docs/install/compile/arm-compile.md b/docs/install/compile/arm-compile.md
@@ -52,7 +52,7 @@
     sudo make install
     ```
 
-3. 根据[requirments.txt](https://github.com/PaddlePaddle/Paddle/blob/develop/python/requirements.txt)安装 Python 依赖库，在飞腾加国产化操作系统环境中，pip 安装可能失败或不能正常工作，主要依赖通过源或源码安装的方式安装依赖库，建议使用系统提供源的方式安装依赖库。
+3. 根据[requirements.txt](https://github.com/PaddlePaddle/Paddle/blob/develop/python/requirements.txt)安装 Python 依赖库，在飞腾加国产化操作系统环境中，pip 安装可能失败或不能正常工作，主要依赖通过源或源码安装的方式安装依赖库，建议使用系统提供源的方式安装依赖库。
 
 4. 将 Paddle 的源代码克隆到当下目录下的 Paddle 文件夹中，并进入 Paddle 目录
 

diff --git a/docs/install/compile/mips-compile.md b/docs/install/compile/mips-compile.md
@@ -76,7 +76,7 @@
     cd Paddle
     ```
 
-6. 根据[requirments.txt](https://github.com/PaddlePaddle/Paddle/blob/develop/python/requirements.txt)安装 Python 依赖库。
+6. 根据[requirements.txt](https://github.com/PaddlePaddle/Paddle/blob/develop/python/requirements.txt)安装 Python 依赖库。
 
 
 7. 切换到`develop`分支下进行编译：

diff --git a/docs/install/compile/sw-compile.md b/docs/install/compile/sw-compile.md
@@ -48,7 +48,7 @@
    -- Found lapack in OpenBLAS (include: /usr/include)
    ```
 
-5. 根据[requirments.txt](https://github.com/PaddlePaddle/Paddle/blob/develop/python/requirements.txt)安装 Python 依赖库，注意在申威系统中一般无法直接使用 pip 或源码编译安装 python 依赖包，建议使用源的方式安装，如果遇到部分依赖包无法安装的情况，请联系操作系统服务商提供支持。此外也可以通过 pip 安装的时候加--no-deps 的方式来避免依赖包的安装，但该种方式可能导致包由于缺少依赖不可用。
+5. 根据[requirements.txt](https://github.com/PaddlePaddle/Paddle/blob/develop/python/requirements.txt)安装 Python 依赖库，注意在申威系统中一般无法直接使用 pip 或源码编译安装 python 依赖包，建议使用源的方式安装，如果遇到部分依赖包无法安装的情况，请联系操作系统服务商提供支持。此外也可以通过 pip 安装的时候加--no-deps 的方式来避免依赖包的安装，但该种方式可能导致包由于缺少依赖不可用。
 
 6. 请创建并进入一个叫 build 的目录下：