fix-t11-t15

Author: Ricardo-shuo-liu

_typos.toml

@@ -169,9 +169,3 @@ sucess = "sucess"
 sucessor = "sucessor"
 sucessors = "sucessors"
 szie = "szie"
-traing = "traing"
-trainning = "trainning"
-transfered = "transfered"
-trasformed = "trasformed"
-treshold = "treshold"
-trian = "trian"

docs/api/paddle/vision/ops/generate_proposals_cn.rst

@@ -35,7 +35,7 @@ generate_proposals
         - **post_nms_top_n** (int，可选) - 每个图在 NMS 后要保留的总框数。默认值为 1000。
         - **nms_thresh** (float，可选) - NMS 中的阈值。默认值为 0.5。
         - **min_size** (float，可选) - 根据宽和高过滤候选框的阈值，宽或高小于该阈值的候选框将被过滤掉。默认值为 0.1。
-        - **eta** (float，可选) - 自适应阈值的衰减系数。仅在自适应 NMS 中且自适应阈值大于 0.5 时生效，在每次迭代中 ``adaptive_threshold = adaptive_treshold * eta`` 。默认值为 1.0。
+        - **eta** (float，可选) - 自适应阈值的衰减系数。仅在自适应 NMS 中且自适应阈值大于 0.5 时生效，在每次迭代中 ``adaptive_threshold = adaptive_threshold * eta`` 。默认值为 1.0。
         - **pixel_offset** (bool, 可选）- 是否有像素偏移。如果是 True, ``img_size`` 在计算时会偏移 1。默认值为 False。
         - **return_rois_num** (bool，可选) - 是否返回 ``rpn_rois_num`` 。当设定为 True 时会返回一个形状为[N,]的 1-D 的 Tensor，包含该 Batch 中每一张图片包含的 RoI 的数目。 N 是批大小和图片数量。默认值为 False。
         - **name** (str，可选) - 具体用法请参见 :ref:`api_guide_Name`，一般无需设置，默认值为 None。

docs/design/concepts/tensor.md

@@ -186,4 +186,4 @@ ViewIterator<ArrayView<T, D>> make_iterator(const Array<T, D>& in, Dim<D> idx) {
 
 The operations that manipulate DArray are defined as global functions, such as `ones`, `zeros`, `reshape`, `gemm` and so on.
 
-An array will be trasformed into an arrayview and then passed to the operation launching on a specific device(CPU/GPU).
+An array will be transformed into an arrayview and then passed to the operation launching on a specific device(CPU/GPU).

docs/design/mkldnn/caching/caching.md

@@ -83,7 +83,7 @@ Also MKL-DNN caching mechanism is used for transferring whole MKL-DNN memory all
 
 ##### Note on multi threading execution scenario
 PaddlePaddle's program(models) can be executed by single thread as well as multiple threads. Graph's operators can be executed by different threads . In particular forward kernel of given operator may be executed by different thread
-than backward kernel of the same instance of operator. In that situation hashing keys used for transferring MKL-DNN primitive descriptors cannot contain Thread ID specific values as key has to be shared among forward and backward kernel. Hence primitive descriptors (to be transfered to backward kernel) are using hashing key that does not contain any thread ID based value. As a result thread safety mechanics (critical section etc.) was employed to
+than backward kernel of the same instance of operator. In that situation hashing keys used for transferring MKL-DNN primitive descriptors cannot contain Thread ID specific values as key has to be shared among forward and backward kernel. Hence primitive descriptors (to be transferred to backward kernel) are using hashing key that does not contain any thread ID based value. As a result thread safety mechanics (critical section etc.) was employed to
 ensure MKL-DNN cache consistency.
 
 ### 4. Store once created MKL-DNN objects in order To avoid MKL-DNN recreation

docs/design/modules/batch_norm_op.md

@@ -58,7 +58,7 @@ As most C++ operators do, `batch_norm_op` is defined by inputs, outputs, attribu
 #### Attributes
 
 - `is_infer`: *bool*. If true, run `batch_norm_op` in inferencing mode.
-- `use_global_est`: *bool*. If true, use `saved_mean` and `saved_var` instead of `E[x]` and `STD[x]` in trainning.
+- `use_global_est`: *bool*. If true, use `saved_mean` and `saved_var` instead of `E[x]` and `STD[x]` in training.
 - `epsilon`: *float*. The epsilon value to avoid division by zero.
 - `momentum`: *float*. Factor used in `estimated_mean` and `estimated_var` updating. The usage is shown above.