diff --git a/deployment/libtorch/README.md b/deployment/libtorch/README.md
index 3f683535..1c0cf2b6 100644
--- a/deployment/libtorch/README.md
+++ b/deployment/libtorch/README.md
@@ -8,7 +8,7 @@ The LibTorch inference for `yolort`, both GPU and CPU are supported.
 
 - LibTorch 1.8.0+ together with corresponding TorchVision 0.9.0+
 - OpenCV
-- CUDA 10.2+ \[Optional\]
+- CUDA 10.2+ [Optional]
 
 *We didn't impose too strong restrictions on the version of CUDA.*
 
diff --git a/deployment/onnxruntime/README.md b/deployment/onnxruntime/README.md
index 87707040..5b64f6a1 100644
--- a/deployment/onnxruntime/README.md
+++ b/deployment/onnxruntime/README.md
@@ -8,7 +8,7 @@ The ONNX Runtime inference for `yolort`, both CPU and GPU are supported.
 
 - ONNX Runtime 1.7+
 - OpenCV
-- CUDA \[Optional\]
+- CUDA [Optional]
 
 *We didn't impose too strong restrictions on the versions of dependencies.*
 
@@ -30,7 +30,7 @@ The ONNX model exported by yolort differs from other pipeline in the following t
 
    And then, you can find that a ONNX model ("best.onnx") have been generated in the directory of "best.pt". Set the `size_divisible` here according to your model, 32 for P5 ("yolov5s.pt" for instance) and 64 for P6 ("yolov5s6.pt" for instance).
 
-1. \[Optional\] Quick test with the ONNX Runtime Python interface.
+1. [Optional] Quick test with the ONNX Runtime Python interface.
 
    ```python
    from yolort.runtime import PredictorORT
diff --git a/deployment/ppq/README.md b/deployment/ppq/README.md
index 03639753..011e3022 100644
--- a/deployment/ppq/README.md
+++ b/deployment/ppq/README.md
@@ -13,7 +13,7 @@ The ppq int8 ptq example of `yolort`.
 
 ## Usage
 
-Here we will mainly discuss how to use the ppq interface, we recommend that you check out  [tutorial](https://github.com/openppl-public/ppq/tree/master/ppq/samples) first. This code can be used to do the following stuff:
+Here we will mainly discuss how to use the ppq interface, we recommend that you check out [tutorial](https://github.com/openppl-public/ppq/tree/master/ppq/samples) first. This code can be used to do the following stuff:
 
 1. Distill your calibration data (Optional: If you don't have images for calibration and bn is in your model, you can use this)
 
diff --git a/deployment/tensorrt/README.md b/deployment/tensorrt/README.md
index 228ee88a..8996956f 100644
--- a/deployment/tensorrt/README.md
+++ b/deployment/tensorrt/README.md
@@ -27,7 +27,7 @@ Here we will mainly discuss how to use the C++ interface, we recommend that you
    trtexec --onnx=best.trt.onnx --saveEngine=best.engine --workspace=8192
    ```
 
-1. \[Optional\] Quick test with the TensorRT Python interface.
+1. [Optional] Quick test with the TensorRT Python interface.
 
    ```python
    import torch
@@ -58,7 +58,7 @@ Here we will mainly discuss how to use the C++ interface, we recommend that you
      cmake --build .  # Can also use the yolort_trt.sln to build on Windows System
      ```
 
-   - \[Windows System Only\] Copy following dependent dynamic link libraries (xxx.dll) to Release/Debug directory
+   - [Windows System Only] Copy following dependent dynamic link libraries (xxx.dll) to Release/Debug directory
 
      - cudnn_cnn_infer64_8.dll, cudnn_ops_infer64_8.dll, cudnn64_8.dll, nvinfer.dll, nvinfer_plugin.dll, nvonnxparser.dll, zlibwapi.dll (On which CUDA and cudnn depend)
      - opencv_corexxx.dll opencv_imgcodecsxxx.dll opencv_imgprocxxx.dll (Subsequent dependencies by OpenCV or you can also use Static OpenCV Library)