mmdeploy

2.1 KiB

Raw Blame History

ncnn Support

Installation

Install ncnn

Download VulkanTools for the compilation of ncnn.

wget https://sdk.lunarg.com/sdk/download/1.2.176.1/linux/vulkansdk-linux-x86_64-1.2.176.1.tar.gz?Human=true -O vulkansdk-linux-x86_64-1.2.176.1.tar.gz
tar -xf vulkansdk-linux-x86_64-1.2.176.1.tar.gz
export VULKAN_SDK=$(pwd)/1.2.176.1/x86_64

Check your gcc version. You should ensure your gcc satisfies gcc >= 6.

Prepare ncnn Framework

Download ncnn source code

git clone git@github.com:Tencent/ncnn.git

Make install ncnn library

cd ncnn
mkdir build
cmake -DNCNN_VULKAN=ON -DNCNN_SYSTEM_GLSLANG=ON -DNCNN_BUILD_EXAMPLES=ON -DNCNN_PYTHON=ON -DNCNN_BUILD_TOOLS=ON -DNCNN_BUILD_BENCHMARK=ON -DNCNN_BUILD_TESTS=ON ..
make install

Install pyncnn module
```
cd ncnn/python
pip install .
```

Build custom ops

Some custom ops are created to support models in OpenMMLab, the custom ops can be built as follows:

cd ${MMDEPLOY_DIR}
mkdir build
cd build
cmake -DBUILD_NCNN_OPS=ON ..
make -j$(nproc)

If you haven't installed NCNN in the default path, please add -DNCNN_DIR flag in cmake.

 cmake -DBUILD_NCNN_OPS=ON -DNCNN_DIR=${NCNN_DIR} ..
 make -j$(nproc)

Convert model

This follows the tutorial on How to convert model.
The converted model has two files: .param and .bin, as model structure file and weight file respectively.

FAQs

When running ncnn models for inference with custom ops, it fails and shows the error message like:
```
TypeError: register mm custom layers(): incompatible function arguments. The following argument types are supported:
    1.(ar0: ncnn:Net) -> int

Invoked with: <ncnn.ncnn.Net object at 0x7f7fc4038bb0>
```
This is because of the failure to bind ncnn C++ library to pyncnn. You should build pyncnn from C++ ncnn source code, but not by pip install

2.1 KiB Raw Blame History