Merge branch 'develop' into update-cI

Author: Ricardo-shuo-liu

_typos.toml

@@ -28,27 +28,6 @@ datas = "datas"
 feeded = "feeded"
 
 # These words need to be fixed
-Learing = "Learing"
-Operaton = "Operaton"
-Optimizaing = "Optimizaing"
-Optimzier = "Optimzier"
-Setment = "Setment"
-Simle = "Simle"
-Sovler = "Sovler"
-libary = "libary"
-matrics = "matrics"
-metrices = "metrices"
-mutbale = "mutbale"
-occurence = "occurence"
-opeartor = "opeartor"
-opeartors = "opeartors"
-operaters = "operaters"
-optmization = "optmization"
-outpu = "outpu"
-outpus = "outpus"
-overrided = "overrided"
-overwrited = "overwrited"
-samle = "samle"
 schedual = "schedual"
 secenarios = "secenarios"
 sematic = "sematic"

docs/api/gen_doc.py

@@ -35,7 +35,7 @@
 #   "short_name":"",  # without module name
 #   "module_name":"",  # the module of the real api belongs to
 #   "display":True/Flase, # consider the not_display_doc_list and the display_doc_list
-#   "has_overwrited_doc":True/False  #
+#   "has_overwritten_doc":True/False  #
 #   "doc_filename"  # document filename without suffix
 #   "suggested_name":"",  # the shortest name in all_names
 # }

docs/api/paddle/linalg/svd_cn.rst

@@ -26,9 +26,9 @@ svd
 返回
 ::::::::::::
 
-    - Tensor U，奇异值分解的 U 矩阵。如果 full_matrics 设置为 False，则 Shape 为 ``[*, M, K]``，如果 full_metrices 设置为 True，那么 Shape 为 ``[*, M, M]``。其中 K 为 M 和 N 的最小值。
+    - Tensor U，奇异值分解的 U 矩阵。如果 full_matrices 设置为 False，则 Shape 为 ``[*, M, K]``，如果 full_matrices 设置为 True，那么 Shape 为 ``[*, M, M]``。其中 K 为 M 和 N 的最小值。
     - Tensor S，奇异值向量，Shape 为 ``[*, K]`` 。
-    - Tensor VH，奇异值分解的 VH 矩阵。如果 full_matrics 设置为 False，则 Shape 为 ``[*, K, N]``，如果 full_metrices 设置为 True，那么 Shape 为 ``[*, N, N]``。其中 K 为 M 和 N 的最小值。
+    - Tensor VH，奇异值分解的 VH 矩阵。如果 full_matrices 设置为 False，则 Shape 为 ``[*, K, N]``，如果 full_matrices 设置为 True，那么 Shape 为 ``[*, N, N]``。其中 K 为 M 和 N 的最小值。
 
 代码示例
 ::::::::::

docs/api/paddle/utils/cpp_extension/load_cn.rst

@@ -29,7 +29,7 @@ load
    from paddle.utils.cpp_extension import load
 
    custom_op_module = load(
-       name="op_shared_libary_name",                # 生成动态链接库的名称
+       name="op_shared_library_name",                # 生成动态链接库的名称
        sources=['relu_op.cc', 'relu_op.cu'],        # 自定义 OP 的源码文件列表
        extra_cxx_cflags=['-g', '-w'],               # 可选，指定编译。cc/.cpp 文件时额外的编译选项
        extra_cuda_cflags=['-O2'],                   # 可选，指定编译。cu 文件时额外的编译选项

docs/design/concurrent/parallel_do.md

@@ -15,7 +15,7 @@ AddOutput(kOutputs, "Outputs needed to be merged from different devices").AsDupl
 AddOutput(kParallelScopes,
           "Scopes for all local variables in forward pass. One scope for each device");
 AddAttr<framework::BlockDesc *>(kParallelBlock,
-                                "List of operaters to be executed in parallel");
+                                "List of operators to be executed in parallel");
 ```
 
 A vanilla implementation of parallel_do can be shown as the following (`|` means single thread and
@@ -94,7 +94,7 @@ There are serial places we can make this parallel_do faster.
 
 ### forward: split input onto different devices
 
-If the input of the parallel_do is independent from any prior opeartors, we can avoid this step by
+If the input of the parallel_do is independent from any prior operators, we can avoid this step by
 prefetching the input onto different devices in a separate background thread. And the python code
 looks like this.
 ```python

docs/design/data_type/float16.md

@@ -101,7 +101,7 @@ In Fluid, a neural network is represented as a protobuf message called [ProgramD
 ### Operator level requirement
 Each operator has many kernels for different data types, devices, and library types. The operator will select the appropriate kernel to run based on, among other things, the data type of the input variables. By default, every Fluid operator has a float data type kernel that takes float variables as input and generates float output.
 
-This means that if we provide float input to the first operator in a program, then each opeartor will use float kernel to compute float output and send it as input to the next operator to trigger the float kernel. Overall, the program will run in float mode and give us a final output of float data type.
+This means that if we provide float input to the first operator in a program, then each operator will use float kernel to compute float output and send it as input to the next operator to trigger the float kernel. Overall, the program will run in float mode and give us a final output of float data type.
 
 The same principle applies if we want a program to run in float16 mode. We provide input variable of float16 data type to the first operator, and then one by one, each operator in the program will run the float16 kernel (provided that each operator in this program has float16 kernels registered) until we finally obtain a float16 output variable.
 

docs/design/dynamic_rnn/rnn_design.md

@@ -198,7 +198,7 @@ std::vector<SortedSeqItem> SortBySeqLen(const LODTensor& tensor);
 由于输入序列的顺序变化，以下现有的接口需要针对性地修改：
 
 - InitMemories, memory 需要根据 `sorted_seqs` 重新排列
-- SetmentInputs
+- SegmentInputs
 - ConcatOutputs
 
 此外，由于 `sorted_seqs` 需要被 `RecurrentGradientOp` 复用，因此会变成 `RecurrentOp` 一个新的 output 输出，

docs/design/dynamic_rnn/rnn_design_en.md

@@ -136,7 +136,7 @@ std::vector<SortedSeqItem> SortBySeqLen(const LODTensor& tensor);
 Due to the sequence of input sequences, the following existing interfaces need to be modified:
 
 - InitMemories, memory needs to be rearranged according to `sorted_seqs`
-- SetmentInputs
+- SegmentInputs
 - ConcatOutputs
 
 In addition, because `sorted_seqs` needs to be multiplexed with `RecurrentGradientOp`, it will become a new output of `RecurrentOp`.

docs/design/memory/memory_optimization.md

@@ -79,7 +79,7 @@ In former control flow graph, the out-edges of node 5 are 5 --> 6 and 5 --> 2, a
 
 - Uses and Defs
 
-An assignmemt to a variable or temporary defines that variable. An occurence of a variable on the right-hand side of an assignment(or in other expressions) uses the variable. We can define the *def* of a variable as the set of graph nodes that define it; or the *def* of a graph node as the set of variables that it defines; and the similarly for the *use* of a variable or graph node. In former control flow graph, *def(3)* = {c}, *use(3)* = {b, c}.
+An assignmemt to a variable or temporary defines that variable. An occurrence of a variable on the right-hand side of an assignment(or in other expressions) uses the variable. We can define the *def* of a variable as the set of graph nodes that define it; or the *def* of a graph node as the set of variables that it defines; and the similarly for the *use* of a variable or graph node. In former control flow graph, *def(3)* = {c}, *use(3)* = {b, c}.
 
 - Liveness
 

docs/design/mkldnn/int8/QAT/README.md

@@ -62,7 +62,7 @@ Notes:
    ```... → input1 → conv2d → output1 → batch_norm → output2 → relu → output3 → ...```
    and we want to quantize the `conv2d` op, then after applying FP32 optimizations the sequence will become
    ```... → input1 → conv2d → output3 → ...```
-   and the quantization scales have to be collected for the `input1` and `outpu3` tensors in the Quant model.
+   and the quantization scales have to be collected for the `input1` and `output3` tensors in the Quant model.
 2. Quantization of the following operators is supported: `conv2d`, `depthwise_conv2d`, `mul`, `fc`, `matmul`, `pool2d`, `reshape2`, `transpose2`, `concat`.
 3. The longest sequence of consecutive quantizable operators in the model, the biggest performance boost can be achieved through quantization:
    ```... → conv2d → conv2d → pool2d → conv2d → conv2d → ...```