Merge branch 'develop' into update-cI

Author: Ricardo-shuo-liu

_typos.toml

@@ -35,11 +35,6 @@ Optimzier = "Optimzier"
 Setment = "Setment"
 Simle = "Simle"
 Sovler = "Sovler"
-desgin = "desgin"
-desginated = "desginated"
-desigin = "desigin"
-determinated = "determinated"
-diffcult = "diffcult"
 dimention = "dimention"
 dimentions = "dimentions"
 dirrectories = "dirrectories"
@@ -90,8 +85,6 @@ outpu = "outpu"
 outpus = "outpus"
 overrided = "overrided"
 overwrited = "overwrited"
-porcess = "porcess"
-processer = "processer"
 samle = "samle"
 schedual = "schedual"
 secenarios = "secenarios"

docs/design/data_type/float16.md

@@ -130,7 +130,7 @@ fp16_tensor.set(tensor.astype(numpy.float16).view(numpy.uint16), GPUPlace)
 ```
 
 ### Consistent API requirement
-The basic inference in float16 mode requires users to feed input and obtain output both of float16 data type. However, in this way, the inference APIs are not consistent between float16 mode and float mode, and users may find it confusing and diffcult to use float16 inference since they need to do extra steps to provide float16 input data and convert float16 output data back to float. To have consistent API for different inference modes, we need to transpile the program desc in some way so that we can run float16 inference by feeding and fetching variables of float data type.
+The basic inference in float16 mode requires users to feed input and obtain output both of float16 data type. However, in this way, the inference APIs are not consistent between float16 mode and float mode, and users may find it confusing and difficult to use float16 inference since they need to do extra steps to provide float16 input data and convert float16 output data back to float. To have consistent API for different inference modes, we need to transpile the program desc in some way so that we can run float16 inference by feeding and fetching variables of float data type.
 
 This problem can be solved by introducing a type-casting operator which takes an input variable of certain data type, cast it to another specified data type, and put the casted data into the output variable. Insert cast operator where needed can make a program internally run in float16 mode.
 

docs/design/others/graph.md

@@ -56,7 +56,7 @@ For each parameter, like W and b created by `layer.fc`, marked as double circles
 
 ## Block and Graph
 
-The word block and graph are interchangable in the desgin of PaddlePaddle.  A [Block](https://github.com/PaddlePaddle/Paddle/pull/3708) is a metaphore of the code and local variables in a pair of curly braces in programming languages, where operators are like statements or instructions.  A graph of operators and variables is a representation of the block.
+The word block and graph are interchangable in the design of PaddlePaddle.  A [Block](https://github.com/PaddlePaddle/Paddle/pull/3708) is a metaphore of the code and local variables in a pair of curly braces in programming languages, where operators are like statements or instructions.  A graph of operators and variables is a representation of the block.
 
 A Block keeps operators in an array `BlockDesc::ops`
 

docs/design/quantization/fixed_point_quantization.md

@@ -79,7 +79,7 @@ From these formulas, dequantization also can be moved before GEMM, do dequantiza
 Figure 2. Equivalent forward in training with simulated quantization.
 </p>
 
-We use this equivalent workflow in the training. In our desigin, there is a quantization transpiler to insert the quantization operator and the de-quantization operator in the Fluid `ProgramDesc`. Since the outputs of quantization and de-quantization operator are still in floating point, they are called faked quantization and de-quantization operator. And the training framework is called simulated quantization.
+We use this equivalent workflow in the training. In our design, there is a quantization transpiler to insert the quantization operator and the de-quantization operator in the Fluid `ProgramDesc`. Since the outputs of quantization and de-quantization operator are still in floating point, they are called faked quantization and de-quantization operator. And the training framework is called simulated quantization.
 
 #### Backward pass
 

docs/dev_guides/custom_device_docs/custom_runtime_en.rst

@@ -29,7 +29,7 @@ Device APIs
 +---------------------------+----------------------------------------------------------+----------+
 | get_device                | To get the current device                                | Y        |
 +---------------------------+----------------------------------------------------------+----------+
-| synchronize_device        | To synchronize the desginated device                     | Y        |
+| synchronize_device        | To synchronize the designated device                     | Y        |
 +---------------------------+----------------------------------------------------------+----------+
 | get_device_count          | To count available devices                               | Y        |
 +---------------------------+----------------------------------------------------------+----------+

docs/guides/model_convert/convert_from_pytorch/deprecated/apply_references_deprecated.py

@@ -265,7 +265,7 @@ def record_api(api):
         return [line]
 
 
-def reference_mapping_item_processer(line, line_idx, state, output, context):
+def reference_mapping_item_processor(line, line_idx, state, output, context):
     if not line.startswith("|"):
         output.append(line)
         return True
@@ -365,7 +365,7 @@ def get_c2a_dict(conditions, meta_dict):
 
     # 第二遍正式读，读并处理
     ret_code = reference_mapping_item(
-        mapping_index_file, reference_mapping_item_processer, reference_context
+        mapping_index_file, reference_mapping_item_processor, reference_context
     )
 
     # 检查是否重复出现

docs/guides/model_convert/convert_from_pytorch/deprecated/validate_mapping_files_deprecated.py

@@ -507,14 +507,14 @@ def get_meta_from_diff_file(
     return meta_data
 
 
-def process_mapping_index(index_path, item_processer, context={}):
+def process_mapping_index(index_path, item_processor, context={}):
     """
     线性处理 `pytorch_api_mapping_cn.md` 文件
     - index_path: 该 md 文件路径
-    - item_processer: 对文件每行的处理方式，输入参数 (line, line_idx, state, output, context)。
+    - item_processor: 对文件每行的处理方式，输入参数 (line, line_idx, state, output, context)。
                       如果处理出错则返回 False，否则返回 True。
     - context: 用于存储处理过程中的上下文信息
-               - output: 使用 context["output"] 初始化，如果不调用 item_processer，直接加入原文件对应行，否则 item_processer 处理 output 逻辑。
+               - output: 使用 context["output"] 初始化，如果不调用 item_processor，直接加入原文件对应行，否则 item_processor 处理 output 逻辑。
     - 返回值：是否成功处理，成功返回 0。
     """
     if not os.path.exists(index_path):
@@ -558,7 +558,7 @@ def process_mapping_index(index_path, item_processer, context={}):
             column_names.extend([c.strip() for c in columns])
             column_count = len(column_names)
 
-            if not item_processer(line, i, state, output, context):
+            if not item_processor(line, i, state, output, context):
                 break
 
             if column_names == expect_column_names:
@@ -577,7 +577,7 @@ def process_mapping_index(index_path, item_processer, context={}):
                 raise Exception(
                     f"Table seperator not match at line {i + 1}: {line}"
                 )
-            if not item_processer(line, i, state, output, context):
+            if not item_processor(line, i, state, output, context):
                 break
             state = IndexParserState.table_row_ignore
         elif state == IndexParserState.table_sep:
@@ -588,16 +588,16 @@ def process_mapping_index(index_path, item_processer, context={}):
                 raise Exception(
                     f"Table seperator not match at line {i + 1}: {line}"
                 )
-            if not item_processer(line, i, state, output, context):
+            if not item_processor(line, i, state, output, context):
                 break
             state = IndexParserState.table_row
         elif state == IndexParserState.table_row_ignore:
-            if not item_processer(line, i, state, output, context):
+            if not item_processor(line, i, state, output, context):
                 break
         elif state == IndexParserState.table_row:
             try:
                 context["columns"] = columns
-                if not item_processer(line, i, state, output, context):
+                if not item_processor(line, i, state, output, context):
                     break
                 context["table_row_idx"] += 1
             except Exception as e:

docs/guides/model_convert/convert_from_pytorch/nlp_fast_explore_cn.md

@@ -280,29 +280,29 @@ PyTorch 模块通常继承`torch.nn.Module`，飞桨模块通常继承`paddle.nn
 
 
 <p align="center">
-  <img src="https://raw.githubusercontent.com/ymyjl/docs/torch_migrate/docs/guides/model_convert/pictures/embedding.png"   align="middle"  width="500" />
+  <img src="../pictures/embedding.png"   align="middle"  width="500" />
 </p>
 
 
 - [EncoderLayer](https://github.com/PaddlePaddle/Paddle/blob/develop/python/paddle/nn/layer/transformer.py#:~:text=class%20TransformerEncoderLayer)：继承自 `torch.nn.Layer`，是 Bert 网络中基本模块，由 MultiHeadAttention、FeedForward 组成。后者由 LayerNorm，Dropout，Linear 层和激活函数构成。
 
 <p align="center">
-  <img src="https://raw.githubusercontent.com/ymyjl/docs/torch_migrate/docs/guides/model_convert/pictures/encoder.png" align="middle"  width="500" />
+  <img src="../pictures/encoder.png" align="middle"  width="500" />
 </p>
 
 
 - SelfAttention 层的 K,Q,V 矩阵用于计算单词之间的相关性分数，他们由 Linear 层组成。
 
 
 <p align="center">
-  <img src="https://raw.githubusercontent.com/ymyjl/docs/torch_migrate/docs/guides/model_convert/pictures/kqv.png" align="middle"  width="500" />
+  <img src="../pictures/kqv.png" align="middle"  width="500" />
 </p>
 
 
 - [MultiHeadAttention](https://github.com/PaddlePaddle/Paddle/blob/develop/python/paddle/nn/layer/transformer.py#:~:text=class%20MultiHeadAttention)：由 SelfAttention 层和 Softmax 函数构成。
 
 <p align="center">
-  <img src="https://raw.githubusercontent.com/ymyjl/docs/torch_migrate/docs/guides/model_convert/pictures/malti-head.png" align="middle"  width="500" />
+  <img src="../pictures/malti-head.png" align="middle"  width="500" />
 </p>
 
 

docs/guides/model_convert/convert_from_pytorch/nlp_migration_experiences_cn.md

@@ -16,7 +16,7 @@
 
 
 <p align="center">
-  <img src="https://raw.githubusercontent.com/ymyjl/docs/torch_migrate/docs/guides/model_convert/pictures/porcess.png" align="middle"  width="500" />
+  <img src="../pictures/process.png" align="middle"  width="500" />
 </p>
 
 
@@ -1137,7 +1137,7 @@ paddle.where(b, paddle.zeros(c.shape), c)
 
 
 <p align="center">
-  <img src="https://raw.githubusercontent.com/ymyjl/docs/torch_migrate/docs/guides/model_convert/pictures/information.png" align="middle"  width="500" />
+  <img src="../pictures/information.png" align="middle"  width="500" />
 </p>
 
 **【可能原因】**

docs/guides/model_convert/convert_from_pytorch/tools/README.md

@@ -119,13 +119,13 @@ API 别名表的生成逻辑与单个 API 项映射类似，实现于 `apply_ref
 
 生成工具读取时，当遇到符合预期的表格表头即进入准备读取的状态，随后跳过表格的分隔线，开始对预处理命令的读取状态，直到所在行不是预处理命令时回到普通状态。
 
-由于该读取逻辑可复用，因此将这部分逻辑实现在验证工具的 `process_mapping_index` 方法，通过传入 `item_processer` 回调和 `context` 上下文来控制行为，使用 `IndexParserState` 状态集来控制读取状态。
+由于该读取逻辑可复用，因此将这部分逻辑实现在验证工具的 `process_mapping_index` 方法，通过传入 `item_processor` 回调和 `context` 上下文来控制行为，使用 `IndexParserState` 状态集来控制读取状态。
 
 两次读取中，第一次读取用于分析表格匹配条件，第二次读取进行实际的预处理命令替换。
 
 第一次读取时使用 `reference_table_scanner` 方法作为回调，收集所有的 API 表引用项，记录其参数作为 API 分类的条件。随后在生成工具的 `get_c2a_dict` 方法中对所有条件按照优先 `prefix` 长度降序，次优 `max_depth` 升序的顺序进行排序，并对所有映射文件元数据按照条件进行匹配。
 
-第二次读取时使用 `reference_mapping_item_processer` 方法作为回调，对于所有需要处理的表格行进行转换，将转换结果写回 `context` 的 `output` 项中。
+第二次读取时使用 `reference_mapping_item_processor` 方法作为回调，对于所有需要处理的表格行进行转换，将转换结果写回 `context` 的 `output` 项中。
 
 完成读取后，检查是否有 API 重复出现，如果重复出现则输出重复出现的 API 名称和所在行，不写回源文件并进行 CI 报错。