update faq (#1139)

* update faq * update volkan version * fix vulkan version * update benckmark model url
kendryte · Jan 3, 2024 · cb3fc2d · cb3fc2d
1 parent 583b64b
commit cb3fc2d
Show file tree

Hide file tree

Showing 6 changed files with 35 additions and 13 deletions.
diff --git a/.github/workflows/compiler-build.yml b/.github/workflows/compiler-build.yml
@@ -77,7 +77,7 @@ jobs:
           - {name: x86_64-windows, os: windows-latest, shell: bash}
 
     env:
-      VULKANSDK_VER: 1.2.182.0
+      VULKANSDK_VER: 1.3.268.0
 
     steps:
     - uses: actions/checkout@v3
@@ -99,8 +99,8 @@ jobs:
 
     - name: Set up test environment (Linux)
       run: |
-        wget https://sdk.lunarg.com/sdk/download/${VULKANSDK_VER}/linux/vulkansdk-linux-x86_64-${VULKANSDK_VER}.tar.gz -O vulkansdk.tar.gz
-        tar xf vulkansdk.tar.gz
+        wget https://sdk.lunarg.com/sdk/download/${VULKANSDK_VER}/linux/vulkansdk-linux-x86_64-${VULKANSDK_VER}.tar.xz -O vulkansdk.tar.xz
+        tar xf vulkansdk.tar.xz
         sudo cp -P ${VULKANSDK_VER}/x86_64/lib/libvulkan.so* /usr/local/lib/
         wget https://github.com/sunnycase/swiftshader/releases/download/v1.0/swiftshader-ubuntu-18.04-x86_64.zip -O swiftshader.zip
         unzip swiftshader.zip
@@ -112,8 +112,8 @@ jobs:
     - name: Set up test environment (Windows)
       shell: pwsh
       run: |
-        Invoke-WebRequest -Uri https://sdk.lunarg.com/sdk/download/${env:VULKANSDK_VER}/windows/VulkanSDK-${env:VULKANSDK_VER}-Installer.exe -O VulkanSDK-Installer.exe
-        .\VulkanSDK-Installer.exe /S
+        # Invoke-WebRequest -Uri https://sdk.lunarg.com/sdk/download/${env:VULKANSDK_VER}/windows/VulkanSDK-${env:VULKANSDK_VER}-Installer.exe -O VulkanSDK-Installer.exe
+        # .\VulkanSDK-Installer.exe /S
         Invoke-WebRequest -Uri https://github.com/sunnycase/swiftshader/releases/download/v1.0/swiftshader-windows-2019-x86_64.zip -OutFile swiftshader.zip
         Expand-Archive swiftshader.zip
         Copy-Item swiftshader\lib\vk_swiftshader_icd.json swiftshader\bin\

diff --git a/.github/workflows/compiler-python-release.yml b/.github/workflows/compiler-python-release.yml
@@ -17,7 +17,7 @@ jobs:
           - {name: x86_64-windows, os: windows-latest, arch: x64}
 
     env:
-      VULKANSDK_VER: 1.2.182.0
+      VULKANSDK_VER: 1.3.268.0
 
     steps:
     - uses: actions/checkout@v3

diff --git a/benchmark/gen_kmodel.py b/benchmark/gen_kmodel.py
@@ -31,11 +31,11 @@
 
 MODELS = {
     "mnist": {
-        "url": "https://github.com/onnx/models/raw/main/vision/classification/mnist/model/mnist-8.onnx",
+        "url": "https://github.com/onnx/models/raw/main/validated/vision/classification/mnist/model/mnist-8.onnx",
         "in_shapes": {"Input3": [1, 1, 28, 28]}
     },
     "mobilenet_v2": {
-        "url": "https://github.com/onnx/models/raw/main/vision/classification/mobilenet/model/mobilenetv2-7.onnx",
+        "url": "https://github.com/onnx/models/raw/main/validated/vision/classification/mobilenet/model/mobilenetv2-7.onnx",
         "in_shapes": {"data": [1, 3, 224, 224]}
     }
 }

diff --git a/docs/FAQ_EN.md b/docs/FAQ_EN.md
@@ -8,7 +8,7 @@
 2. What kinds of ops can be fully accelerated by KPU?
 
     All the following constraints must be met.
-    - Feature map shape: Input feature maps smaller than or equal to 320x240(WxH) and output features map larger than or equal to 4x4(WxH) and channels are between 1 to 1024 are supported.
+    - Feature map shape: Input feature maps smaller than or equal to *320x240(WxH)* and output features map larger than or equal to *4x4(WxH)* and channels are between 1 to 1024 are supported.
     - Same symmetric paddings (TensorFlow use asymmetric paddings when stride=2 and size is even).
     - Normal Conv2D and DepthwiseConv2D of 1x1 or 3x3 filters and stride is 1 or 2.
     - MaxPool(2x2 or 4x4) and AveragePool(2x2 or 4x4).
@@ -22,11 +22,22 @@
     - DilatedConv2D, nncase will replace it with a SpaceToBatch + KPUConv2D + BatchToSpace.
     - TransposeConv2D, nncase will replace it with a Pad + KPUConv2D.
 
+4. Attention point about K210。
+
+    - Use official SDK repo：[kendryte-standalone-sdk](https://github.com/kendryte/kendryte-standalone-sdk).
+    - Use a pure English path and keep the path as short as possible.
+    - Projects that need to be compiled are placed in the src directory, see other examples.
+    - Use administrator privileges in Windows.
+
 ### Compile models
 1. Fatal: Not supported tflite opcode: DEQUANTIZE
 
     Use float tflite models, nncase will take care of quantization.
 
+2. Fatal: inputs are not compatible to xxx.
+
+    Please check if the input of model is valid. Not support dynamic input shape.
+
 ### Deploy models
 1. Should I normalize inputs when running the model?
 

diff --git a/docs/FAQ_ZH.md b/docs/FAQ_ZH.md
@@ -8,7 +8,7 @@
 2. 哪些算子可以被 KPU 完全加速？
 
     下面的约束需要全部满足。
-    - 特征图尺寸：输入特征图小于等于 320x240(WxH) 同时输出特征图大于等于 4x4(WxH)，通道数在 1 到 1024。
+    - 特征图尺寸：输入特征图小于等于 *320x240(WxH)* 同时输出特征图大于等于 *4x4(WxH)*，通道数在 1 到 1024。
     - Same 对称 paddings (TensorFlow 在 stride=2 同时尺寸为偶数时使用非对称 paddings)。
     - 普通 Conv2D 和 DepthwiseConv2D，卷积核为 1x1 或 3x3，stride 为 1 或 2。
     - MaxPool(2x2 或 4x4) 和 AveragePool(2x2 或 4x4)。
@@ -22,11 +22,22 @@
     - DilatedConv2D, nncase 会把它替换为一个 SpaceToBatch + KPUConv2D + BatchToSpace。
     - TransposeConv2D, nncase 会把它替换为一个 Pad + KPUConv2D。
 
+4. 使用K210注意事项。
+
+    - 请使用官方SDK仓库：[kendryte-standalone-sdk](https://github.com/kendryte/kendryte-standalone-sdk).
+    - 使用纯英文路径，路径尽量短。
+    - 需要编译的项目放在src目录下，参考其他example。
+    - Windows下使用管理员权限。
+
 ### 编译模型
 1. Fatal: Not supported tflite opcode: DEQUANTIZE
 
     使用浮点 tflite 模型，nncase 会做量化。
 
+2. Fatal: inputs are not compatible to xxx.
+
+    请检查模型输入是否合法，不支持动态shape。
+
 ### 部署模型
 1. 运行模型是我需要归一化输入吗？
 

diff --git a/pyproject.toml b/pyproject.toml
@@ -47,9 +47,9 @@ before-all = [
   "pip install conan==1.59",
   "conan profile new default --detect",
   "conan profile update settings.compiler.libcxx=libstdc++11 default",
-  "wget https://sdk.lunarg.com/sdk/download/1.2.182.0/linux/vulkansdk-linux-x86_64-1.2.182.0.tar.gz -O vulkansdk.tar.gz",
-  "tar xf vulkansdk.tar.gz",
-  "cp -P 1.2.182.0/x86_64/lib/libvulkan.so* /usr/local/lib/"
+  "wget https://sdk.lunarg.com/sdk/download/1.3.268.0/linux/vulkansdk-linux-x86_64-1.3.268.0.tar.xz -O vulkansdk.tar.xz",
+  "tar xf vulkansdk.tar.xz",
+  "cp -P 1.3.268.0/x86_64/lib/libvulkan.so* /usr/local/lib/"
 ]
 before-build = "pip install https://github.com/sunnycase/auditwheel/releases/download/4.0.0-nncase-2/auditwheel-0.0.0-py3-none-any.whl"
 repair-wheel-command = "LD_LIBRARY_PATH=/usr/lib64 auditwheel repair -w {dest_dir} {wheel} --exclude libvulkan.so.1,libgomp.so.1"