diff --git a/.github/workflows/prebuild.yml b/.github/workflows/prebuild.yml
index b85fbc786e..bd113d0e6f 100644
--- a/.github/workflows/prebuild.yml
+++ b/.github/workflows/prebuild.yml
@@ -29,6 +29,7 @@ jobs:
           export MMDEPLOY_VERSION=$(python3 -c "import sys; sys.path.append('mmdeploy');from version import __version__;print(__version__)")
           echo $MMDEPLOY_VERSION
           echo "MMDEPLOY_VERSION=$MMDEPLOY_VERSION"  >> $GITHUB_ENV
+          echo "OUTPUT_DIR=$MMDEPLOY_VERSION-$GITHUB_RUN_ID" >> $GITHUB_ENV
       - name: Build MMDeploy
         run: |
           source activate mmdeploy-3.6
@@ -51,17 +52,55 @@ jobs:
           cd pack
           python ../tools/package_tools/generate_build_config.py --backend 'ort;trt' \
             --system linux --output config.yml --device cuda --build-sdk --build-sdk-monolithic \
-            --build-sdk-python --sdk-dynamic-net
+            --build-sdk-python --sdk-dynamic-net --onnxruntime-dir=$ONNXRUNTIME_GPU_DIR
+          python ../tools/package_tools/mmdeploy_builder.py --config config.yml
+      - name: Move artifact
+        run: |
+          mkdir -p /__w/mmdeploy/prebuild/$OUTPUT_DIR
+          cp -r pack/* /__w/mmdeploy/prebuild/$OUTPUT_DIR
+
+  linux_build_cxx11abi:
+    runs-on: [self-hosted, linux-3090]
+    container:
+      image: openmmlab/mmdeploy:build-ubuntu16.04-cuda11.3
+      options: "--gpus=all --ipc=host"
+      volumes:
+        - /data2/actions-runner/prebuild:/__w/mmdeploy/prebuild
+    steps:
+      - name: Checkout repository
+        uses: actions/checkout@v3
+        with:
+          submodules: recursive
+      - name: Get mmdeploy version
+        run: |
+          export MMDEPLOY_VERSION=$(python3 -c "import sys; sys.path.append('mmdeploy');from version import __version__;print(__version__)")
+          echo $MMDEPLOY_VERSION
+          echo "MMDEPLOY_VERSION=$MMDEPLOY_VERSION"  >> $GITHUB_ENV
+          echo "OUTPUT_DIR=$MMDEPLOY_VERSION-$GITHUB_RUN_ID" >> $GITHUB_ENV
+      - name: Build sdk cpu backend
+        run: |
+          mkdir pack; cd pack
+          python ../tools/package_tools/generate_build_config.py --backend 'ort' \
+            --system linux --output config.yml --device cpu --build-sdk --build-sdk-monolithic \
+            --sdk-dynamic-net --cxx11abi
+          python ../tools/package_tools/mmdeploy_builder.py --config config.yml
+      - name: Build sdk cuda backend
+        run: |
+          cd pack
+          python ../tools/package_tools/generate_build_config.py --backend 'ort;trt' \
+            --system linux --output config.yml --device cuda --build-sdk --build-sdk-monolithic \
+            --sdk-dynamic-net --cxx11abi --onnxruntime-dir=$ONNXRUNTIME_GPU_DIR --cudnn-dir /usr
           python ../tools/package_tools/mmdeploy_builder.py --config config.yml
       - name: Move artifact
         run: |
-          mkdir -p /__w/mmdeploy/prebuild/$MMDEPLOY_VERSION
-          rm -rf /__w/mmdeploy/prebuild/$MMDEPLOY_VERSION/*
-          mv pack/* /__w/mmdeploy/prebuild/$MMDEPLOY_VERSION
+          mkdir -p /__w/mmdeploy/prebuild/$OUTPUT_DIR
+          cp -r pack/* /__w/mmdeploy/prebuild/$OUTPUT_DIR
 
   linux_test:
     runs-on: [self-hosted, linux-3090]
-    needs: linux_build
+    needs:
+      - linux_build
+      - linux_build_cxx11abi
     container:
       image: openmmlab/mmdeploy:ubuntu20.04-cuda11.3
       options: "--gpus=all --ipc=host"
@@ -76,13 +115,14 @@ jobs:
           export MMDEPLOY_VERSION=$(python3 -c "import sys; sys.path.append('mmdeploy');from version import __version__;print(__version__)")
           echo $MMDEPLOY_VERSION
           echo "MMDEPLOY_VERSION=$MMDEPLOY_VERSION"  >> $GITHUB_ENV
+          echo "OUTPUT_DIR=$MMDEPLOY_VERSION-$GITHUB_RUN_ID" >> $GITHUB_ENV
       - name: Test python
         run: |
-          cd /__w/mmdeploy/prebuild/$MMDEPLOY_VERSION
+          cd /__w/mmdeploy/prebuild/$OUTPUT_DIR
           bash $GITHUB_WORKSPACE/tools/package_tools/test/test_sdk_python.sh
       - name: Test c/cpp
         run: |
-          cd /__w/mmdeploy/prebuild/$MMDEPLOY_VERSION
+          cd /__w/mmdeploy/prebuild/$OUTPUT_DIR
           bash $GITHUB_WORKSPACE/tools/package_tools/test/test_sdk.sh
 
   linux_upload:
@@ -100,20 +140,21 @@ jobs:
           export MMDEPLOY_VERSION=$(python3 -c "import sys; sys.path.append('mmdeploy');from version import __version__;print(__version__)")
           echo $MMDEPLOY_VERSION
           echo "MMDEPLOY_VERSION=$MMDEPLOY_VERSION"  >> $GITHUB_ENV
+          echo "OUTPUT_DIR=$MMDEPLOY_VERSION-$GITHUB_RUN_ID" >> $GITHUB_ENV
       - name: Upload mmdeploy
         run: |
-          cd $PREBUILD_DIR/$MMDEPLOY_VERSION/mmdeploy
+          cd $PREBUILD_DIR/$OUTPUT_DIR/mmdeploy
           pip install twine
           # twine upload * --repository testpypi -u __token__ -p ${{ secrets.test_pypi_password }}
           twine upload * -u __token__ -p ${{ secrets.pypi_password }}
       - name: Upload mmdeploy_runtime
         run: |
-          cd $PREBUILD_DIR/$MMDEPLOY_VERSION/mmdeploy_runtime
+          cd $PREBUILD_DIR/$OUTPUT_DIR/mmdeploy_runtime
           # twine upload * --repository testpypi -u __token__ -p ${{ secrets.test_pypi_password }}
           twine upload * -u __token__ -p ${{ secrets.pypi_password }}
       - name: Zip mmdeploy sdk
         run: |
-          cd $PREBUILD_DIR/$MMDEPLOY_VERSION/sdk
+          cd $PREBUILD_DIR/$OUTPUT_DIR/sdk
           for folder in *
           do
             tar czf $folder.tar.gz $folder
@@ -122,7 +163,7 @@ jobs:
         uses: softprops/action-gh-release@v1
         with:
           files: |
-              $PREBUILD_DIR/$MMDEPLOY_VERSION/sdk/*.tar.gz
+              $PREBUILD_DIR/$OUTPUT_DIR/sdk/*.tar.gz
 
 
   windows_build:
@@ -138,6 +179,7 @@ jobs:
           $env:MMDEPLOY_VERSION=(python -c "import sys; sys.path.append('mmdeploy');from version import __version__;print(__version__)")
           echo $env:MMDEPLOY_VERSION
           echo "MMDEPLOY_VERSION=$env:MMDEPLOY_VERSION"  >> $env:GITHUB_ENV
+          echo "OUTPUT_DIR=$env:MMDEPLOY_VERSION-$env:GITHUB_RUN_ID" >> $env:GITHUB_ENV
       - name: Build MMDeploy
         run: |
           . D:\DEPS\cienv\prebuild_gpu_env.ps1
@@ -166,9 +208,8 @@ jobs:
           python ../tools/package_tools/mmdeploy_builder.py --config config.yml
       - name: Move artifact
         run: |
-          New-Item "D:/DEPS/ciartifact/$env:MMDEPLOY_VERSION" -ItemType Directory -Force
-          Remove-Item "D:/DEPS/ciartifact/$env:MMDEPLOY_VERSION/*" -Force -Recurse
-          Move-Item pack/* "D:/DEPS/ciartifact/$env:MMDEPLOY_VERSION"
+          New-Item "D:/DEPS/ciartifact/$env:OUTPUT_DIR" -ItemType Directory -Force
+          Move-Item pack/* "D:/DEPS/ciartifact/$env:OUTPUT_DIR"
 
   windows_test:
     runs-on: [self-hosted, win10-3080]
@@ -182,15 +223,16 @@ jobs:
           $env:MMDEPLOY_VERSION=(python -c "import sys; sys.path.append('mmdeploy');from version import __version__;print(__version__)")
           echo $env:MMDEPLOY_VERSION
           echo "MMDEPLOY_VERSION=$env:MMDEPLOY_VERSION"  >> $env:GITHUB_ENV
+          echo "OUTPUT_DIR=$env:MMDEPLOY_VERSION-$env:GITHUB_RUN_ID" >> $env:GITHUB_ENV
       - name: Test python
         run: |
-          cd "D:/DEPS/ciartifact/$env:MMDEPLOY_VERSION"
+          cd "D:/DEPS/ciartifact/$env:OUTPUT_DIR"
           . D:\DEPS\cienv\prebuild_cpu_env.ps1
           conda activate ci-test
           & "$env:GITHUB_WORKSPACE/tools/package_tools/test/test_sdk_python.ps1"
       - name: Test c/cpp
         run: |
-          cd "D:/DEPS/ciartifact/$env:MMDEPLOY_VERSION"
+          cd "D:/DEPS/ciartifact/$env:OUTPUT_DIR"
           . D:\DEPS\cienv\prebuild_cpu_env.ps1
           & "$env:GITHUB_WORKSPACE/tools/package_tools/test/test_sdk.ps1"
 
@@ -208,21 +250,22 @@ jobs:
           $env:MMDEPLOY_VERSION=(python -c "import sys; sys.path.append('mmdeploy');from version import __version__;print(__version__)")
           echo $env:MMDEPLOY_VERSION
           echo "MMDEPLOY_VERSION=$env:MMDEPLOY_VERSION"  >> $env:GITHUB_ENV
+          echo "OUTPUT_DIR=$env:MMDEPLOY_VERSION-$env:GITHUB_RUN_ID" >> $env:GITHUB_ENV
       - name: Upload mmdeploy
         run: |
-          cd "D:/DEPS/ciartifact/$env:MMDEPLOY_VERSION/mmdeploy"
+          cd "D:/DEPS/ciartifact/$env:OUTPUT_DIR/mmdeploy"
           conda activate mmdeploy-3.8
           # twine upload * --repository testpypi -u __token__ -p ${{ secrets.test_pypi_password }}
           twine upload * -u __token__ -p ${{ secrets.pypi_password }}
       - name: Upload mmdeploy_runtime
         run: |
-          cd "D:/DEPS/ciartifact/$env:MMDEPLOY_VERSION/mmdeploy_runtime"
+          cd "D:/DEPS/ciartifact/$env:OUTPUT_DIR/mmdeploy_runtime"
           conda activate mmdeploy-3.8
           # twine upload * --repository testpypi -u __token__ -p ${{ secrets.test_pypi_password }}
           twine upload * -u __token__ -p ${{ secrets.pypi_password }}
       - name: Zip mmdeploy sdk
         run: |
-          cd "D:/DEPS/ciartifact/$env:MMDEPLOY_VERSION/sdk"
+          cd "D:/DEPS/ciartifact/$env:OUTPUT_DIR/sdk"
           $folders = $(ls).Name
           foreach ($folder in $folders) {
             Compress-Archive -Path $folder -DestinationPath "$folder.zip"
@@ -231,4 +274,4 @@ jobs:
         uses: softprops/action-gh-release@v1
         with:
           files: |
-              D:/DEPS/ciartifact/$env:MMDEPLOY_VERSION/sdk/*.zip
+              D:/DEPS/ciartifact/$env:OUTPUT_DIR/sdk/*.zip
diff --git a/docs/en/02-how-to-run/prebuilt_package_windows.md b/docs/en/02-how-to-run/prebuilt_package_windows.md
index c7913fa217..143232fb65 100644
--- a/docs/en/02-how-to-run/prebuilt_package_windows.md
+++ b/docs/en/02-how-to-run/prebuilt_package_windows.md
@@ -21,26 +21,27 @@
 
 ______________________________________________________________________
 
-This tutorial takes `mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1.zip` and `mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8.2.3.0.zip` as examples to show how to use the prebuilt packages.
+This tutorial takes `mmdeploy-0.13.0-windows-amd64.zip` and `mmdeploy-0.13.0-windows-amd64-cuda11.3.zip` as examples to show how to use the prebuilt packages. The former supports onnxruntime cpu inference, the latter supports onnxruntime-gpu and tensorrt inference.
 
 The directory structure of the prebuilt package is as follows, where the `dist` folder is about model converter, and the `sdk` folder is related to model inference.
 
 ```
 .
-|-- dist
-`-- sdk
-    |-- bin
-    |-- example
-    |-- include
-    |-- lib
-    `-- python
+├── build_sdk.ps1
+├── example
+├── include
+├── install_opencv.ps1
+├── lib
+├── README.md
+├── set_env.ps1
+└── thirdparty
 ```
 
 ## Prerequisite
 
 In order to use the prebuilt package, you need to install some third-party dependent libraries.
 
-1. Follow the [get_started](../get_started.md) documentation to create a virtual python environment and install pytorch, torchvision and mmcv-full. To use the C interface of the SDK, you need to install [vs2019+](https://visualstudio.microsoft.com/), [OpenCV](https://github.com/opencv/opencv/releases).
+1. Follow the [get_started](../get_started.md) documentation to create a virtual python environment and install pytorch, torchvision and mmcv. To use the C interface of the SDK, you need to install [vs2019+](https://visualstudio.microsoft.com/), [OpenCV](https://github.com/opencv/opencv/releases).
 
    :point_right: It is recommended to use `pip` instead of `conda` to install pytorch and torchvision
 
@@ -80,9 +81,8 @@ In order to use `ONNX Runtime` backend, you should also do the following steps.
 5. Install `mmdeploy` (Model Converter) and `mmdeploy_runtime` (SDK Python API).
 
    ```bash
-   # download mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1.zip
-   pip install .\mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1\dist\mmdeploy-0.13.0-py38-none-win_amd64.whl
-   pip install .\mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1\sdk\python\mmdeploy_runtime-0.13.0-cp38-none-win_amd64.whl
+   pip install mmdeploy==0.13.0
+   pip install mmdeploy-runtime==0.13.0
    ```
 
    :point_right: If you have installed it before, please uninstall it first.
@@ -100,6 +100,8 @@ In order to use `ONNX Runtime` backend, you should also do the following steps.
    ![sys-path](https://user-images.githubusercontent.com/16019484/181463801-1d7814a8-b256-46e9-86f2-c08de0bc150b.png)
    :exclamation: Restart powershell to make the environment variables setting take effect. You can check whether the settings are in effect by `echo $env:PATH`.
 
+8. Download SDK C/cpp Library mmdeploy-0.13.0-windows-amd64.zip
+
 ### TensorRT
 
 In order to use `TensorRT` backend, you should also do the following steps.
@@ -107,9 +109,8 @@ In order to use `TensorRT` backend, you should also do the following steps.
 5. Install `mmdeploy` (Model Converter) and `mmdeploy_runtime` (SDK Python API).
 
    ```bash
-   # download mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8.2.3.0.zip
-   pip install .\mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8.2.3.0\dist\mmdeploy-0.13.0-py38-none-win_amd64.whl
-   pip install .\mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8.2.3.0\sdk\python\mmdeploy_runtime-0.13.0-cp38-none-win_amd64.whl
+   pip install mmdeploy==0.13.0
+   pip install mmdeploy-runtime-gpu==0.13.0
    ```
 
    :point_right: If you have installed it before, please uninstall it first.
@@ -128,6 +129,8 @@ In order to use `TensorRT` backend, you should also do the following steps.
 
 7. Install pycuda by `pip install pycuda`
 
+8. Download SDK C/cpp Library mmdeploy-0.13.0-windows-amd64-cuda11.3.zip
+
 ## Model Convert
 
 ### ONNX Runtime Example
@@ -138,7 +141,7 @@ After preparation work, the structure of the current working directory should be
 
 ```
 ..
-|-- mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1
+|-- mmdeploy-0.13.0-windows-amd64
 |-- mmclassification
 |-- mmdeploy
 `-- resnet18_8xb32_in1k_20210831-fbbb1da6.pth
@@ -186,7 +189,7 @@ After installation of mmdeploy-tensorrt prebuilt package, the structure of the c
 
 ```
 ..
-|-- mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8.2.3.0
+|-- mmdeploy-0.13.0-windows-amd64-cuda11.3
 |-- mmclassification
 |-- mmdeploy
 `-- resnet18_8xb32_in1k_20210831-fbbb1da6.pth
@@ -299,7 +302,7 @@ python .\mmdeploy\demo\python\image_classification.py cpu .\work_dir\onnx\resnet
 
 #### TensorRT
 
-```
+```bash
  python .\mmdeploy\demo\python\image_classification.py cuda .\work_dir\trt\resnet\ .\mmclassification\demo\demo.JPEG
 ```
 
@@ -309,35 +312,19 @@ The following describes how to use the SDK's C API for inference
 
 #### ONNXRuntime
 
-1. Build examples
-
-   Under `mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1\sdk\example` directory
-
-   ```
-   // Path should be modified according to the actual location
-   mkdir build
-   cd build
-   cmake ..\cpp -A x64 -T v142 `
-     -DOpenCV_DIR=C:\Deps\opencv\build\x64\vc15\lib `
-     -DMMDeploy_DIR=C:\workspace\mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1\sdk\lib\cmake\MMDeploy `
-     -DONNXRUNTIME_DIR=C:\Deps\onnxruntime\onnxruntime-win-gpu-x64-1.8.1
-
-   cmake --build . --config Release
-   ```
+1. Add environment variables
 
-2. Add environment variables or copy the runtime libraries to the same level directory of exe
+   Refer to the README.md in sdk folder
 
-   :point_right: The purpose is to make the exe find the relevant dll
+2. Build examples
 
-   If choose to add environment variables, add the runtime libraries path of `mmdeploy` (`mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1\sdk\bin`) to the `PATH`.
-
-   If choose to copy the dynamic libraries, copy the dll in the bin directory to the same level directory of the just compiled exe (build/Release).
+   Refer to the README.md in sdk folder
 
 3. Inference：
 
    It is recommended to use `CMD` here.
 
-   Under `mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1\\sdk\\example\\build\\Release` directory：
+   Under `mmdeploy-0.13.0-windows-amd64\\example\\cpp\\build\\Release` directory：
 
    ```
    .\image_classification.exe cpu C:\workspace\work_dir\onnx\resnet\ C:\workspace\mmclassification\demo\demo.JPEG
@@ -345,35 +332,19 @@ The following describes how to use the SDK's C API for inference
 
 #### TensorRT
 
-1. Build examples
-
-   Under `mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8.2.3.0\\sdk\\example` directory
-
-   ```
-   // Path should be modified according to the actual location
-   mkdir build
-   cd build
-   cmake ..\cpp -A x64 -T v142 `
-     -DOpenCV_DIR=C:\Deps\opencv\build\x64\vc15\lib `
-     -DMMDeploy_DIR=C:\workspace\mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8 2.3.0\sdk\lib\cmake\MMDeploy `
-     -DTENSORRT_DIR=C:\Deps\tensorrt\TensorRT-8.2.3.0 `
-     -DCUDNN_DIR=C:\Deps\cudnn\8.2.1
-   cmake --build . --config Release
-   ```
-
-2. Add environment variables or copy the runtime libraries to the same level directory of exe
+1. Add environment variables
 
-   :point_right: The purpose is to make the exe find the relevant dll
+   Refer to the README.md in sdk folder
 
-   If choose to add environment variables, add the runtime libraries path of `mmdeploy` (`mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8.2.3.0\sdk\bin`) to the `PATH`.
+2. Build examples
 
-   If choose to copy the dynamic libraries, copy the dll in the bin directory to the same level directory of the just compiled exe (build/Release).
+   Refer to the README.md in sdk folder
 
 3. Inference
 
    It is recommended to use `CMD` here.
 
-   Under `mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8.2.3.0\\sdk\\example\\build\\Release` directory
+   Under `mmdeploy-0.13.0-windows-amd64-cuda11.3\\example\\cpp\\build\\Release` directory
 
    ```
    .\image_classification.exe cuda C:\workspace\work_dir\trt\resnet C:\workspace\mmclassification\demo\demo.JPEG
diff --git a/docs/en/get_started.md b/docs/en/get_started.md
index efdec195e6..a8b3601eea 100644
--- a/docs/en/get_started.md
+++ b/docs/en/get_started.md
@@ -69,8 +69,7 @@ mim install mmcv-full
 
 **Step 1.** Install MMDeploy and inference engine
 
-We recommend using MMDeploy precompiled package as our best practice.
-You can download them from [here](https://github.com/open-mmlab/mmdeploy/releases) according to your target platform and device.
+We recommend using MMDeploy precompiled package as our best practice. Currently, we support model converter and sdk inference pypi package, and the sdk c/cpp library is provided [here](https://github.com/open-mmlab/mmdeploy/releases). You can download them according to your target platform and device.
 
 The supported platform and device matrix is presented as following:
 
@@ -92,7 +91,7 @@ The supported platform and device matrix is presented as following:
   </tr>
   <tr>
     <td>CUDA</td>
-    <td>N</td>
+    <td>Y</td>
     <td>Y</td>
   </tr>
   <tr>
@@ -103,7 +102,7 @@ The supported platform and device matrix is presented as following:
   </tr>
   <tr>
     <td>CUDA</td>
-    <td>N</td>
+    <td>Y</td>
     <td>Y</td>
   </tr>
 </tbody>
@@ -114,46 +113,44 @@ The supported platform and device matrix is presented as following:
 Take the latest precompiled package as example, you can install it as follows:
 
 <details open>
-<summary><b>Linux-x86_64, CPU, ONNX Runtime 1.8.1</b></summary>
+<summary><b>Linux-x86_64</b></summary>
 
 ```shell
-# install MMDeploy
-wget https://github.com/open-mmlab/mmdeploy/releases/download/v0.13.0/mmdeploy-0.13.0-linux-x86_64-onnxruntime1.8.1.tar.gz
-tar -zxvf mmdeploy-0.13.0-linux-x86_64-onnxruntime1.8.1.tar.gz
-cd mmdeploy-0.13.0-linux-x86_64-onnxruntime1.8.1
-pip install dist/mmdeploy-0.13.0-py3-none-linux_x86_64.whl
-pip install sdk/python/mmdeploy_runtime-0.13.0-cp38-none-linux_x86_64.whl
-cd ..
-# install inference engine: ONNX Runtime
-pip install onnxruntime==1.8.1
-wget https://github.com/microsoft/onnxruntime/releases/download/v1.8.1/onnxruntime-linux-x64-1.8.1.tgz
-tar -zxvf onnxruntime-linux-x64-1.8.1.tgz
-export ONNXRUNTIME_DIR=$(pwd)/onnxruntime-linux-x64-1.8.1
-export LD_LIBRARY_PATH=$ONNXRUNTIME_DIR/lib:$LD_LIBRARY_PATH
-```
-
-</details>
-
-<details open>
-<summary><b>Linux-x86_64, CUDA 11.x, TensorRT 8.2.3.0</b></summary>
-
-```shell
-# install MMDeploy
-wget https://github.com/open-mmlab/mmdeploy/releases/download/v0.13.0/mmdeploy-0.13.0-linux-x86_64-cuda11.1-tensorrt8.2.3.0.tar.gz
-tar -zxvf mmdeploy-0.13.0-linux-x86_64-cuda11.1-tensorrt8.2.3.0.tar.gz
-cd mmdeploy-0.13.0-linux-x86_64-cuda11.1-tensorrt8.2.3.0
-pip install dist/mmdeploy-0.13.0-py3-none-linux_x86_64.whl
-pip install sdk/python/mmdeploy_runtime-0.13.0-cp38-none-linux_x86_64.whl
-cd ..
-# install inference engine: TensorRT
-# !!! Download TensorRT-8.2.3.0 CUDA 11.x tar package from NVIDIA, and extract it to the current directory
+# 1. install MMDeploy model converter
+pip install mmdeploy==0.13.0
+
+# 2. install MMDeploy sdk inference
+# you can install one to install according whether you need gpu inference
+# 2.1 support onnxruntime
+pip install mmdeploy-runtime==0.13.0
+# 2.2 support onnxruntime-gpu, tensorrt
+pip install mmdeploy-runtime-gpu==0.13.0
+
+# 3. install inference engine
+# 3.1 install TensorRT
+# !!! If you want to convert a tensorrt model or inference with tensorrt,
+# download TensorRT-8.2.3.0 CUDA 11.x tar package from NVIDIA, and extract it to the current directory
 pip install TensorRT-8.2.3.0/python/tensorrt-8.2.3.0-cp38-none-linux_x86_64.whl
 pip install pycuda
 export TENSORRT_DIR=$(pwd)/TensorRT-8.2.3.0
 export LD_LIBRARY_PATH=${TENSORRT_DIR}/lib:$LD_LIBRARY_PATH
-# !!! Download cuDNN 8.2.1 CUDA 11.x tar package from NVIDIA, and extract it to the current directory
+# !!! Moreover, download cuDNN 8.2.1 CUDA 11.x tar package from NVIDIA, and extract it to the current directory
 export CUDNN_DIR=$(pwd)/cuda
 export LD_LIBRARY_PATH=$CUDNN_DIR/lib64:$LD_LIBRARY_PATH
+
+# 3.2 install ONNX Runtime
+# you can install one to install according whether you need gpu inference
+# 3.2.1 onnxruntime
+wget https://github.com/microsoft/onnxruntime/releases/download/v1.8.1/onnxruntime-linux-x64-1.8.1.tgz
+tar -zxvf onnxruntime-linux-x64-1.8.1.tgz
+export ONNXRUNTIME_DIR=$(pwd)/onnxruntime-linux-x64-1.8.1
+export LD_LIBRARY_PATH=$ONNXRUNTIME_DIR/lib:$LD_LIBRARY_PATH
+# 3.2.2 onnxruntime-gpu
+pip install onnxruntime-gpu==1.8.1
+wget https://github.com/microsoft/onnxruntime/releases/download/v1.8.1/onnxruntime-linux-x64-gpu-1.8.1.tgz
+tar -zxvf onnxruntime-linux-x64-gpu-1.8.1.tgz
+export ONNXRUNTIME_DIR=$(pwd)/onnxruntime-linux-x64-gpu-1.8.1
+export LD_LIBRARY_PATH=$ONNXRUNTIME_DIR/lib:$LD_LIBRARY_PATH
 ```
 
 </details>
@@ -200,8 +197,8 @@ And they make up of MMDeploy Model that can be fed to MMDeploy SDK to do model i
 For more details about model conversion, you can read [how_to_convert_model](02-how-to-run/convert_model.md). If you want to customize the conversion pipeline, you can edit the config file by following [this](02-how-to-run/write_config.md) tutorial.
 
 ```{tip}
-If MMDeploy-ONNXRuntime prebuilt package is installed, you can convert the above model to onnx model and perform ONNX Runtime inference
-just by 'changing detection_tensorrt_dynamic-320x320-1344x1344.py' to 'detection_onnxruntime_dynamic.py' and making '--device' as 'cpu'.
+You can convert the above model to onnx model and perform ONNX Runtime inference
+just by changing 'detection_tensorrt_dynamic-320x320-1344x1344.py' to 'detection_onnxruntime_dynamic.py' and making '--device' as 'cpu'.
 ```
 
 ## Inference Model
@@ -232,12 +229,14 @@ result = inference_model(
 You can directly run MMDeploy demo programs in the precompiled package to get inference results.
 
 ```shell
-cd mmdeploy-0.13.0-linux-x86_64-cuda11.1-tensorrt8.2.3.0
+wget https://github.com/open-mmlab/mmdeploy/releases/download/v0.13.0/mmdeploy-0.13.0-linux-x86_64-cuda11.3.tar.gz
+tar xf mmdeploy-0.13.0-linux-x86_64-cuda11.3
+cd mmdeploy-0.13.0-linux-x86_64-cuda11.3
 # run python demo
-python sdk/example/python/object_detection.py cuda ../mmdeploy_model/faster-rcnn ../mmdetection/demo/demo.jpg
+python example/python/object_detection.py cuda ../mmdeploy_model/faster-rcnn ../mmdetection/demo/demo.jpg
 # run C/C++ demo
-export LD_LIBRARY_PATH=$(pwd)/sdk/lib:$LD_LIBRARY_PATH
-./sdk/bin/object_detection cuda ../mmdeploy_model/faster-rcnn ../mmdetection/demo/demo.jpg
+# build the demo according to the README.md in the folder.
+./bin/object_detection cuda ../mmdeploy_model/faster-rcnn ../mmdetection/demo/demo.jpg
 ```
 
 ```{note}
diff --git a/docs/zh_cn/02-how-to-run/prebuilt_package_windows.md b/docs/zh_cn/02-how-to-run/prebuilt_package_windows.md
index 735a19ad97..c01f412300 100644
--- a/docs/zh_cn/02-how-to-run/prebuilt_package_windows.md
+++ b/docs/zh_cn/02-how-to-run/prebuilt_package_windows.md
@@ -21,23 +21,24 @@
 
 ______________________________________________________________________
 
-目前，`MMDeploy`在`Windows`平台下提供`TensorRT`以及`ONNX Runtime`两种预编译包，可以从[Releases](https://github.com/open-mmlab/mmdeploy/releases)获取。
+目前，`MMDeploy`在`Windows`平台下提供`cpu`以及`cuda`两种Device的预编译包，其中`cpu`版支持使用onnxruntime cpu进行推理，`cuda`版支持使用onnxruntime-gpu以及tensorrt进行推理，可以从[Releases](https://github.com/open-mmlab/mmdeploy/releases)获取。。
 
-本篇教程以`mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1.zip`和`mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8.2.3.0.zip`为例，展示预编译包的使用方法。
+本篇教程以`mmdeploy-0.13.0-windows-amd64.zip`和`mmdeploy-0.13.0-windows-amd64-cuda11.3.zip`为例，展示预编译包的使用方法。
 
 为了方便使用者快速上手，本教程以分类模型(mmclassification)为例，展示两种预编译包的使用方法。
 
-预编译包的目录结构如下，其中`dist`文件夹为模型转换相关内容，`sdk`文件夹为模型推理相关内容。
+预编译包的目录结构如下。
 
 ```
 .
-|-- dist
-`-- sdk
-    |-- bin
-    |-- example
-    |-- include
-    |-- lib
-    `-- python
+├── build_sdk.ps1
+├── example
+├── include
+├── install_opencv.ps1
+├── lib
+├── README.md
+├── set_env.ps1
+└── thirdparty
 ```
 
 ## 准备工作
@@ -48,7 +49,7 @@ ______________________________________________________________________
 
 首先新建一个工作目录workspace
 
-1. 请按照[get_started](../get_started.md)文档，准备虚拟环境，安装pytorch、torchvision、mmcv-full。若要使用SDK的C接口，需要安装vs2019+, OpenCV。
+1. 请按照[get_started](../get_started.md)文档，准备虚拟环境，安装pytorch、torchvision、mmcv。若要使用SDK的C接口，需要安装vs2019+, OpenCV。
 
    :point_right: 这里建议使用`pip`而不是`conda`安装pytorch、torchvision
 
@@ -88,9 +89,8 @@ ______________________________________________________________________
 5. 安装`mmdeploy`（模型转换）以及`mmdeploy_runtime`（模型推理Python API）的预编译包
 
    ```bash
-   # 先下载 mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1.zip
-   pip install .\mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1\dist\mmdeploy-0.13.0-py38-none-win_amd64.whl
-   pip install .\mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1\sdk\python\mmdeploy_runtime-0.13.0-cp38-none-win_amd64.whl
+   pip install mmdeploy==0.13.0
+   pip install mmdeploy-runtime==0.13.0
    ```
 
    :point_right: 如果之前安装过，需要先卸载后再安装。
@@ -108,6 +108,8 @@ ______________________________________________________________________
    ![sys-path](https://user-images.githubusercontent.com/16019484/181463801-1d7814a8-b256-46e9-86f2-c08de0bc150b.png)
    :exclamation: 重启powershell让环境变量生效，可以通过 echo $env:PATH 来检查是否设置成功。
 
+8. 下载 SDK C/cpp Library mmdeploy-0.13.0-windows-amd64.zip
+
 ### TensorRT
 
 本节介绍`mmdeploy`使用`TensorRT`推理所特有的环境准备工作
@@ -115,9 +117,8 @@ ______________________________________________________________________
 5. 安装`mmdeploy`（模型转换）以及`mmdeploy_runtime`（模型推理Python API）的预编译包
 
    ```bash
-   # 先下载 mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8.2.3.0.zip
-   pip install .\mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8.2.3.0\dist\mmdeploy-0.13.0-py38-none-win_amd64.whl
-   pip install .\mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8.2.3.0\sdk\python\mmdeploy_runtime-0.13.0-cp38-none-win_amd64.whl
+   pip install mmdeploy==0.13.0
+   pip install mmdeploy-runtime-gpu==0.13.0
    ```
 
    :point_right: 如果之前安装过，需要先卸载后再安装
@@ -136,6 +137,8 @@ ______________________________________________________________________
 
 7. 安装pycuda `pip install pycuda`
 
+8. 下载 SDK C/cpp Library mmdeploy-0.13.0-windows-amd64-cuda11.3.zip
+
 ## 模型转换
 
 ### ONNX Runtime Example
@@ -146,7 +149,7 @@ ______________________________________________________________________
 
 ```
 ..
-|-- mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1
+|-- mmdeploy-0.13.0-windows-amd64
 |-- mmclassification
 |-- mmdeploy
 `-- resnet18_8xb32_in1k_20210831-fbbb1da6.pth
@@ -194,7 +197,7 @@ export2SDK(deploy_cfg, model_cfg, work_dir, pth=model_checkpoint, device=device)
 
 ```
 ..
-|-- mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8.2.3.0
+|-- mmdeploy-0.13.0-windows-amd64-cuda11.3
 |-- mmclassification
 |-- mmdeploy
 `-- resnet18_8xb32_in1k_20210831-fbbb1da6.pth
@@ -315,7 +318,7 @@ python .\mmdeploy\demo\python\image_classification.py cpu .\work_dir\onnx\resnet
 
 推理代码
 
-```
+```bash
  python .\mmdeploy\demo\python\image_classification.py cuda .\work_dir\trt\resnet\ .\mmclassification\demo\demo.JPEG
 ```
 
@@ -325,35 +328,19 @@ python .\mmdeploy\demo\python\image_classification.py cpu .\work_dir\onnx\resnet
 
 #### ONNXRuntime
 
-1. 编译 examples
-
-   在`mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1\sdk\example`目录下
-
-   ```
-   // 部分路径根据实际位置进行修改
-   mkdir build
-   cd build
-   cmake ..\cpp -A x64 -T v142 `
-     -DOpenCV_DIR=C:\Deps\opencv\build\x64\vc15\lib `
-     -DMMDeploy_DIR=C:\workspace\mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1\sdk\lib\cmake\MMDeploy `
-     -DONNXRUNTIME_DIR=C:\Deps\onnxruntime\onnxruntime-win-gpu-x64-1.8.1
-
-   cmake --build . --config Release
-   ```
+1. 添加环境变量
 
-2. 添加环境变量或拷贝动态库到exe同级目录
+   可根据 SDK 文件夹内的 README.md 使用脚本添加环境变量
 
-   :point_right: 目的是使exe运行时可以正确找到相关dll
+2. 编译 examples
 
-   若选择添加环境变量，则将`mmdeploy`的运行时库路径（`mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1\sdk\bin`）添加到PATH，可参考onnxruntime的添加过程。
-
-   若选择拷贝动态库，而将bin目录中的dll拷贝到刚才编译出的exe(build/Release)的同级目录下。
+   可根据 SDK 文件夹内的 README.md 使用脚本进行编译
 
 3. 推理：
 
    这里建议使用cmd，这样如果exe运行时如果找不到相关的dll的话会有弹窗
 
-   在mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1\\sdk\\example\\build\\Release目录下：
+   在mmdeploy-0.13.0-windows-amd64-onnxruntime1.8.1\\example\\cpp\\build\\Release目录下：
 
    ```
    .\image_classification.exe cpu C:\workspace\work_dir\onnx\resnet\ C:\workspace\mmclassification\demo\demo.JPEG
@@ -361,35 +348,19 @@ python .\mmdeploy\demo\python\image_classification.py cpu .\work_dir\onnx\resnet
 
 #### TensorRT
 
-1. 编译 examples
-
-   在mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8.2.3.0\\sdk\\example目录下
-
-   ```
-   // 部分路径根据所在硬盘的位置进行修改
-   mkdir build
-   cd build
-   cmake ..\cpp -A x64 -T v142 `
-     -DOpenCV_DIR=C:\Deps\opencv\build\x64\vc15\lib `
-     -DMMDeploy_DIR=C:\workspace\mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8 2.3.0\sdk\lib\cmake\MMDeploy `
-     -DTENSORRT_DIR=C:\Deps\tensorrt\TensorRT-8.2.3.0 `
-     -DCUDNN_DIR=C:\Deps\cudnn\8.2.1
-   cmake --build . --config Release
-   ```
-
-2. 添加环境变量或拷贝动态库到exe同级目录
+1. 添加环境变量
 
-   :point_right: 目的是使exe运行时可以正确找到相关dll
+   可根据 SDK 文件夹内的 README.md 使用脚本添加环境变量
 
-   若选择添加环境变量，则将`mmdeploy`的运行时库路径（`mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8.2.3.0\sdk\bin`）添加到PATH，可参考onnxruntime的添加过程。
+2. 编译 examples
 
-   若选择拷贝动态库，而将bin目录中的dll拷贝到刚才编译出的exe(build/Release)的同级目录下。
+   可根据 SDK 文件夹内的 README.md 使用脚本进行编译
 
 3. 推理
 
    这里建议使用cmd，这样如果exe运行时如果找不到相关的dll的话会有弹窗
 
-   在mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8.2.3.0\\sdk\\example\\build\\Release目录下：
+   在mmdeploy-0.13.0-windows-amd64-cuda11.1-tensorrt8.2.3.0\\example\\cpp\\build\\Release目录下：
 
    ```
    .\image_classification.exe cuda C:\workspace\work_dir\trt\resnet C:\workspace\mmclassification\demo\demo.JPEG
diff --git a/docs/zh_cn/get_started.md b/docs/zh_cn/get_started.md
index 8a02f6c66f..aa3151cf34 100644
--- a/docs/zh_cn/get_started.md
+++ b/docs/zh_cn/get_started.md
@@ -65,7 +65,7 @@ mim install mmcv-full
 
 **第二步**: 安装 MMDeploy 和 推理引擎
 
-我们推荐用户使用预编译包安装和体验 MMDeploy 功能。请根据目标软硬件平台，从[这里](https://github.com/open-mmlab/mmdeploy/releases) 选择最新版本下载并安装。
+我们推荐用户使用预编译包安装和体验 MMDeploy 功能。目前提供模型转换(trt/ort)以及 SDK 推理的 pypi 预编译包，SDK 的 c/cpp 库可从[这里](https://github.com/open-mmlab/mmdeploy/releases) 选择最新版本下载并安装。
 
 目前，MMDeploy 的预编译包支持的平台和设备矩阵如下：
 
@@ -87,7 +87,7 @@ mim install mmcv-full
   </tr>
   <tr>
     <td>CUDA</td>
-    <td>N</td>
+    <td>Y</td>
     <td>Y</td>
   </tr>
   <tr>
@@ -98,7 +98,7 @@ mim install mmcv-full
   </tr>
   <tr>
     <td>CUDA</td>
-    <td>N</td>
+    <td>Y</td>
     <td>Y</td>
   </tr>
 </tbody>
@@ -109,46 +109,43 @@ mim install mmcv-full
 以最新的预编译包为例，你可以参考以下命令安装：
 
 <details open>
-<summary><b>Linux-x86_64, CPU, ONNX Runtime 1.8.1</b></summary>
-
-```shell
-# 安装 MMDeploy ONNX Runtime 自定义算子库和推理 SDK
-wget https://github.com/open-mmlab/mmdeploy/releases/download/v0.13.0/mmdeploy-0.13.0-linux-x86_64-onnxruntime1.8.1.tar.gz
-tar -zxvf mmdeploy-0.13.0-linux-x86_64-onnxruntime1.8.1.tar.gz
-cd mmdeploy-0.13.0-linux-x86_64-onnxruntime1.8.1
-pip install dist/mmdeploy-0.13.0-py3-none-linux_x86_64.whl
-pip install sdk/python/mmdeploy_runtime-0.13.0-cp38-none-linux_x86_64.whl
-cd ..
-# 安装推理引擎 ONNX Runtime
-pip install onnxruntime==1.8.1
-wget https://github.com/microsoft/onnxruntime/releases/download/v1.8.1/onnxruntime-linux-x64-1.8.1.tgz
-tar -zxvf onnxruntime-linux-x64-1.8.1.tgz
-export ONNXRUNTIME_DIR=$(pwd)/onnxruntime-linux-x64-1.8.1
-export LD_LIBRARY_PATH=$ONNXRUNTIME_DIR/lib:$LD_LIBRARY_PATH
-```
-
-</details>
-
-<details open>
-<summary><b>Linux-x86_64, CUDA 11.x, TensorRT 8.2.3.0</b></summary>
+<summary><b>Linux-x86_64</b></summary>
 
 ```shell
-# 安装 MMDeploy TensorRT 自定义算子库和推理 SDK
-wget https://github.com/open-mmlab/mmdeploy/releases/download/v0.13.0/mmdeploy-0.13.0-linux-x86_64-cuda11.1-tensorrt8.2.3.0.tar.gz
-tar -zxvf mmdeploy-0.13.0-linux-x86_64-cuda11.1-tensorrt8.2.3.0.tar.gz
-cd mmdeploy-0.13.0-linux-x86_64-cuda11.1-tensorrt8.2.3.0
-pip install dist/mmdeploy-0.13.0-py3-none-linux_x86_64.whl
-pip install sdk/python/mmdeploy_runtime-0.13.0-cp38-none-linux_x86_64.whl
-cd ..
-# 安装推理引擎 TensorRT
-# !!! 从 NVIDIA 官网下载 TensorRT-8.2.3.0 CUDA 11.x 安装包并解压到当前目录
+# 1. 安装 MMDeploy 模型转换工具（含trt/ort自定义算子）
+pip install mmdeploy==0.13.0
+
+# 2. 安装 MMDeploy SDK推理工具
+# 根据是否需要GPU推理可任选其一进行下载安装
+# 2.1 支持 onnxruntime 推理
+pip install mmdeploy-runtime==0.13.0
+# 2.2 支持 onnxruntime-gpu tensorrt 推理
+pip install mmdeploy-runtime-gpu==0.13.0
+
+# 3. 安装推理引擎
+# 3.1 安装推理引擎 TensorRT
+# !!! 若要进行 TensorRT 模型的转换以及推理，从 NVIDIA 官网下载 TensorRT-8.2.3.0 CUDA 11.x 安装包并解压到当前目录。
 pip install TensorRT-8.2.3.0/python/tensorrt-8.2.3.0-cp38-none-linux_x86_64.whl
 pip install pycuda
 export TENSORRT_DIR=$(pwd)/TensorRT-8.2.3.0
 export LD_LIBRARY_PATH=${TENSORRT_DIR}/lib:$LD_LIBRARY_PATH
-# !!! 从 NVIDIA 官网下载 cuDNN 8.2.1 CUDA 11.x 安装包并解压到当前目录
+# !!! 另外还需要从 NVIDIA 官网下载 cuDNN 8.2.1 CUDA 11.x 安装包并解压到当前目录
 export CUDNN_DIR=$(pwd)/cuda
 export LD_LIBRARY_PATH=$CUDNN_DIR/lib64:$LD_LIBRARY_PATH
+
+# 3.2 安装推理引擎 ONNX Runtime
+# 根据是否需要GPU推理可任选其一进行下载安装
+# 3.2.1 onnxruntime
+wget https://github.com/microsoft/onnxruntime/releases/download/v1.8.1/onnxruntime-linux-x64-1.8.1.tgz
+tar -zxvf onnxruntime-linux-x64-1.8.1.tgz
+export ONNXRUNTIME_DIR=$(pwd)/onnxruntime-linux-x64-1.8.1
+export LD_LIBRARY_PATH=$ONNXRUNTIME_DIR/lib:$LD_LIBRARY_PATH
+# 3.2.2 onnxruntime-gpu
+pip install onnxruntime-gpu==1.8.1
+wget https://github.com/microsoft/onnxruntime/releases/download/v1.8.1/onnxruntime-linux-x64-gpu-1.8.1.tgz
+tar -zxvf onnxruntime-linux-x64-gpu-1.8.1.tgz
+export ONNXRUNTIME_DIR=$(pwd)/onnxruntime-linux-x64-gpu-1.8.1
+export LD_LIBRARY_PATH=$ONNXRUNTIME_DIR/lib:$LD_LIBRARY_PATH
 ```
 
 </details>
@@ -193,7 +190,7 @@ python mmdeploy/tools/deploy.py \
 转换结果被保存在 `--work-dir` 指向的文件夹中。**该文件夹中不仅包含推理后端模型，还包括推理元信息。这些内容的整体被定义为 SDK Model。推理 SDK 将用它进行模型推理。**
 
 ```{tip}
-在安装了 MMDeploy-ONNXRuntime 预编译包后，把上述转换命令中的detection_tensorrt_dynamic-320x320-1344x1344.py 换成 detection_onnxruntime_dynamic.py，并修改 --device 为 cpu，
+把上述转换命令中的detection_tensorrt_dynamic-320x320-1344x1344.py 换成 detection_onnxruntime_dynamic.py，并修改 --device 为 cpu，
 即可以转出 onnx 模型，并用 ONNXRuntime 进行推理
 ```
 
@@ -226,12 +223,15 @@ result = inference_model(
 你可以直接运行预编译包中的 demo 程序，输入 SDK Model 和图像，进行推理，并查看推理结果。
 
 ```shell
-cd mmdeploy-0.13.0-linux-x86_64-cuda11.1-tensorrt8.2.3.0
+wget https://github.com/open-mmlab/mmdeploy/releases/download/v0.13.0/mmdeploy-0.13.0-linux-x86_64-cuda11.3.tar.gz
+tar xf mmdeploy-0.13.0-linux-x86_64-cuda11.3
+
+cd mmdeploy-0.13.0-linux-x86_64-cuda11.3
 # 运行 python demo
-python sdk/example/python/object_detection.py cuda ../mmdeploy_model/faster-rcnn ../mmdetection/demo/demo.jpg
+python example/python/object_detection.py cuda ../mmdeploy_model/faster-rcnn ../mmdetection/demo/demo.jpg
 # 运行 C/C++ demo
-export LD_LIBRARY_PATH=$(pwd)/sdk/lib:$LD_LIBRARY_PATH
-./sdk/bin/object_detection cuda ../mmdeploy_model/faster-rcnn ../mmdetection/demo/demo.jpg
+# 根据文件夹内的 README.md 进行编译
+./bin/object_detection cuda ../mmdeploy_model/faster-rcnn ../mmdetection/demo/demo.jpg
 ```
 
 ```{note}
diff --git a/tools/package_tools/generate_build_config.py b/tools/package_tools/generate_build_config.py
index a7c7139568..1040bb73fd 100644
--- a/tools/package_tools/generate_build_config.py
+++ b/tools/package_tools/generate_build_config.py
@@ -63,6 +63,7 @@ def parse_arguments():
         '--cudnn-dir',
         type=str,
         help='cudnn root dir, default use $ENV{CUDNN_DIR}')
+    parser.add_argument('--cxx11abi', action='store_true', help='new cxxabi')
     parser.add_argument(
         '--output', required=True, type=str, help='output config file path')
 
@@ -139,8 +140,10 @@ def generate_config(args):
         # sdk package template
         if args.system in ['windows', 'linux']:
             name = 'mmdeploy-{mmdeploy_v}-{system}-{machine}'
+            if args.cxx11abi:
+                name = name + '-cxx11abi'
             if args.device == 'cpu':
-                name = '{}-cpu'.format(name)
+                pass
             elif args.device == 'cuda':
                 name = '{}-cuda'.format(name) + '{cuda_v}'
             else:
diff --git a/tools/package_tools/scripts/linux/README.md b/tools/package_tools/scripts/linux/README.md
index 85d65b3cd8..4e44b47cf9 100644
--- a/tools/package_tools/scripts/linux/README.md
+++ b/tools/package_tools/scripts/linux/README.md
@@ -8,8 +8,11 @@
 2. set environment variable and path
    in sdk folder:
 
-   `source ./set_env.sh` \
-   (**you have to additionally install cuda and cudnn if you use sdk cuda version**)
+   `source ./set_env.sh`
+
+   (**you have to additionally install cuda and cudnn if use sdk cuda version**)
+
+   (**may need to set CUDNN environment variable point to cudnn root folder if use sdk cuda version**)
 
 3. build sdk
    in sdk folder:
diff --git a/tools/package_tools/scripts/windows/README.md b/tools/package_tools/scripts/windows/README.md
index e11d269a9b..7d6cdf05f0 100644
--- a/tools/package_tools/scripts/windows/README.md
+++ b/tools/package_tools/scripts/windows/README.md
@@ -12,7 +12,12 @@
    in sdk folder:
 
    `. .\set_env.ps1`
-   (you have to additionally install cuda and cudnn if you use sdk cuda version)
+
+   if you use sdk cuda version:
+
+   (**you have to additionally install cuda and cudnn if use sdk cuda version**)
+
+   (**may need to set CUDNN environment variable point to cudnn root folder if use sdk cuda version**)
 
 4. build sdk
    in sdk folder:
diff --git a/tools/package_tools/test/test_sdk.sh b/tools/package_tools/test/test_sdk.sh
index ec87596ffa..472a8d5b4c 100755
--- a/tools/package_tools/test/test_sdk.sh
+++ b/tools/package_tools/test/test_sdk.sh
@@ -11,14 +11,14 @@ fi
 pushd $WORKSPACE
 pushd $SDK_DIR
 
-test_pkg=$(ls | grep *cpu*)
+test_pkg=$(find "." -type d -iname "*-x86_64")
 work_dir=/tmp/_test
 cp -r $test_pkg $work_dir
 
 pushd $work_dir
 
 # opencv
-if [[ ! -d $OpenCV_DIR ]]; then
+if [ ! -d "$OpenCV_DIR" ]; then
     ./install_opencv.sh
 fi