mirror of
https://github.com/open-mmlab/mmdeploy.git
synced 2025-01-14 08:09:43 +08:00
5f59b4464f
* docs(docs/zh_cn): add doc and link checker * docs(REAME): update * docs(docs/zh_cn): update * docs(benchmark): update table * docs(zh_cn/benchmark): update link * CI(docs): update link check * ci(doc): update checker * docs(zh_cn): update * style(ci): remove useless para * style(ci): update * docs(zh_cn): update * docs(benchmark.md): fix mobilnet link error * docs(docs/zh_cn): add doc and link checker * docs(REAME): update * docs(docs/zh_cn): update * docs(benchmark): update table * docs(zh_cn/benchmark): update link * CI(docs): update link check * ci(doc): update checker * docs(zh_cn): update * style(ci): remove useless para * style(ci): update * docs(zh_cn): update * docs(benchmark.md): fix mobilnet link error * docs(zh_cn/do_regression_test.md): rebase * docs(docs/zh_cn): add doc and link checker * Update README_zh-CN.md * Update README_zh-CN.md * Update index.rst * Update check-doc-link.yml * [Fix] Fix ci (#426) * fix ci * add nvidia key * remote torch * recover pytorch * ci(codecov): ignore ci * docs(zh_cn): add get_started.md * docs(zh_cn): fix review advice * docs(readthedocs): update * docs(zh_CN): update * docs(zh_CN): revert * fix(docs): review advices * fix(docs): review advices * fix(docs): review Co-authored-by: q.yao <streetyao@live.com>
99 lines
3.0 KiB
Markdown
99 lines
3.0 KiB
Markdown
## ncnn Support
|
|
|
|
MMDeploy now supports ncnn version == 1.0.20211208
|
|
|
|
### Installation
|
|
|
|
#### Install ncnn
|
|
|
|
- Download VulkanTools for the compilation of ncnn.
|
|
|
|
```bash
|
|
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=$(pwd)/1.2.176.1/x86_64
|
|
export LD_LIBRARY_PATH=$VULKAN_SDK/lib:$LD_LIBRARY_PATH
|
|
```
|
|
|
|
- Check your gcc version.
|
|
You should ensure your gcc satisfies `gcc >= 6`.
|
|
|
|
- Install Protocol Buffers through:
|
|
```bash
|
|
apt-get install libprotobuf-dev protobuf-compiler
|
|
```
|
|
|
|
- Prepare ncnn Framework
|
|
|
|
- Download ncnn source code
|
|
```bash
|
|
git clone -b 20211208 git@github.com:Tencent/ncnn.git
|
|
```
|
|
|
|
- <font color=red>Make install</font> ncnn library
|
|
```bash
|
|
cd ncnn
|
|
export NCNN_DIR=$(pwd)
|
|
git submodule update --init
|
|
mkdir -p build && cd 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
|
|
```bash
|
|
cd ${NCNN_DIR} # To ncnn root directory
|
|
cd python
|
|
pip install -e .
|
|
```
|
|
|
|
#### Build custom ops
|
|
|
|
Some custom ops are created to support models in OpenMMLab, the custom ops can be built as follows:
|
|
|
|
```bash
|
|
cd ${MMDEPLOY_DIR}
|
|
mkdir -p build && cd build
|
|
cmake -DMMDEPLOY_TARGET_BACKENDS=ncnn ..
|
|
make -j$(nproc)
|
|
```
|
|
|
|
If you haven't installed ncnn in the default path, please add `-Dncnn_DIR` flag in cmake.
|
|
|
|
```bash
|
|
cmake -DMMDEPLOY_TARGET_BACKENDS=ncnn -Dncnn_DIR=${NCNN_DIR}/build/install/lib/cmake/ncnn ..
|
|
make -j$(nproc)
|
|
```
|
|
|
|
### Convert model
|
|
|
|
- This follows the tutorial on [How to convert model](../tutorials/how_to_convert_model.md).
|
|
- The converted model has two files: `.param` and `.bin`, as model structure file and weight file respectively.
|
|
|
|
|
|
### List of supported custom ops
|
|
|
|
| Operator | CPU | MMDeploy Releases |
|
|
|:--------------------------------|:---:|:------------------|
|
|
| [Expand](../ops/ncnn.md#expand) | Y | master |
|
|
| [Gather](../ops/ncnn.md#gather) | Y | master |
|
|
| [Shape](../ops/ncnn.md#shape) | Y | master |
|
|
| [TopK](../ops/ncnn.md#topk) | Y | master |
|
|
|
|
#### Reminder
|
|
|
|
- In ncnn version >= 1.0.20201208, the dimension of ncnn.Mat should be no more than 4.
|
|
|
|
### FAQs
|
|
|
|
1. When running ncnn models for inference with custom ops, it fails and shows the error message like:
|
|
|
|
```bash
|
|
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`
|