From fe20084c02480eb21d848831b1d0ffa5e282f019 Mon Sep 17 00:00:00 2001
From: teng <19603468+cmdbug@users.noreply.github.com>
Date: Wed, 28 Apr 2021 22:23:05 +0800
Subject: [PATCH] Format the faq.md (#2881)
---
docs/faq.md | 478 +++++++++++++++++++++++++++-------------------------
1 file changed, 247 insertions(+), 231 deletions(-)
diff --git a/docs/faq.md b/docs/faq.md
index a4ba8ed9f..1bdc5f018 100644
--- a/docs/faq.md
+++ b/docs/faq.md
@@ -3,290 +3,296 @@
# 如何加入技术交流QQ群？
# 如何看作者b站直播？
-nihui的bilibili直播间：https://live.bilibili.com/1264617
-# 编译
+
+- nihui的bilibili直播间：[水竹院落](https://live.bilibili.com/1264617)
-## 怎样下载完整源码？
+# 编译
-git clone --recursive https://github.com/Tencent/ncnn/
-或者
-https://github.com/Tencent/ncnn/releases 下载 ncnn-xxxxx-full-source.zip
+- ## 怎样下载完整源码？
-## 怎么交叉编译？
+ git clone --recursive https://github.com/Tencent/ncnn/
+
+ 或者
+
+ 下载 [ncnn-xxxxx-full-source.zip](https://github.com/Tencent/ncnn/releases)
-cmake 工具链怎么设置啊？
+- ## 怎么交叉编译？cmake 工具链怎么设置啊？
+
+ 参见 https://github.com/Tencent/ncnn/wiki/how-to-build
-参见 https://github.com/Tencent/ncnn/wiki/how-to-build
+- ## The submodules were not downloaded! Please update submodules with "git submodule update --init" and try again
-## The submodules were not downloaded! Please update submodules with "git submodule update --init" and try again
+ 如上，下载完整源码。或者按提示执行: git submodule update --init
-如上，下载完整源码。或者按提示执行 git submodule update --init。
+- ## Could NOT find Protobuf (missing: Protobuf_INCLUDE_DIR)
+
+ sudo apt-get install libprotobuf-dev protobuf-compiler
-## Could NOT find Protobuf (missing: Protobuf_INCLUDE_DIR)
+- ## Could NOT find CUDA (missing: CUDA_TOOLKIT_ROOT_DIR CUDA_INCLUDE_DIRS CUDA_CUDART_LIBRARY)
-sudo apt-get install libprotobuf-dev protobuf-compiler
+ https://github.com/Tencent/ncnn/issues/1873
-## Could NOT find CUDA (missing: CUDA_TOOLKIT_ROOT_DIR CUDA_INCLUDE_DIRS CUDA_CUDART_LIBRARY)
+- ## Could not find a package configuration file provided by "OpenCV" with any of the following names: OpenCVConfig.cmake opencv-config.cmake
-https://github.com/Tencent/ncnn/issues/1873
+ sudo apt-get install libopencv-dev
-## Could not find a package configuration file provided by "OpenCV" with any of the following names: OpenCVConfig.cmake opencv-config.cmake
+ 或者自行编译安装，set(OpenCV_DIR {OpenCVConfig.cmake所在目录})
-sudo apt-get install libopencv-dev
+- ## Could not find a package configuration file provided by "ncnn" with any of the following names: ncnnConfig.cmake ncnn-config.cmake
-或者自行编译安装，set(OpenCV_DIR {OpenCVConfig.cmake所在目录})
+ set(ncnn_DIR {ncnnConfig.cmake所在目录})
-## Could not find a package configuration file provided by "ncnn" with any of the following names: ncnnConfig.cmake ncnn-config.cmake
+- ## 找不到 Vulkan,
-set(ncnn_DIR {ncnnConfig.cmake所在目录})
+ cmake版本 3.10，否则没有带 FindVulkan.cmake
-## 找不到 Vulkan
+ android-api >= 24
-如何安装 vulkan sdk
-cmake版本 3.10，否则没有带 FindVulkan.cmake
-android-api >= 24
-macos 要先执行安装脚本
+ macos 要先执行安装脚本
-## 找不到库（需要根据系统/编译器指定）
+- ## 如何安装 vulkan sdk
-undefined reference to __kmpc_for_static_init_4 __kmpc_for_static_fini __kmpc_fork_call ...
+- ## 找不到库（需要根据系统/编译器指定）
-需要链接openmp库
+ undefined reference to __kmpc_for_static_init_4 __kmpc_for_static_fini __kmpc_fork_call ...
-undefined reference to vkEnumerateInstanceExtensionProperties vkGetInstanceProcAddr vkQueueSubmit ...
+ 需要链接openmp库
-需要 vulkan-1.lib
+ undefined reference to vkEnumerateInstanceExtensionProperties vkGetInstanceProcAddr vkQueueSubmit ...
-undefined reference to glslang::InitializeProcess() glslang::TShader::TShader(EShLanguage) ...
+ 需要 vulkan-1.lib
-需要 glslang.lib OGLCompiler.lib SPIRV.lib OSDependent.lib
+ undefined reference to glslang::InitializeProcess() glslang::TShader::TShader(EShLanguage) ...
-undefined reference to AAssetManager_fromJava AAssetManager_open AAsset_seek ...
+ 需要 glslang.lib OGLCompiler.lib SPIRV.lib OSDependent.lib
-find_library和target_like_libraries中增加 android
+ undefined reference to AAssetManager_fromJava AAssetManager_open AAsset_seek ...
-find_package(ncnn)
+ find_library和target_like_libraries中增加 android
-## undefined reference to typeinfo for ncnn::Layer
+ find_package(ncnn)
-opencv rtti -> opencv-mobile
+- ## undefined reference to typeinfo for ncnn::Layer
-## undefined reference to __cpu_model
+ opencv rtti -> opencv-mobile
-升级编译器 / libgcc_s libgcc
+- ## undefined reference to __cpu_model
-## unrecognized command line option "-mavx2"
+ 升级编译器 / libgcc_s libgcc
-升级 gcc
+- ## unrecognized command line option "-mavx2"
+ 升级 gcc
-## 为啥自己编译的ncnn android库特别大？
+- ## 为啥自己编译的ncnn android库特别大？
-https://github.com/Tencent/ncnn/wiki/build-for-android.zh 以及见 如何裁剪更小的 ncnn 库
+ https://github.com/Tencent/ncnn/wiki/build-for-android.zh 以及见 如何裁剪更小的 ncnn 库
-## ncnnoptimize和自定义层
+- ## ncnnoptimize和自定义层
-先ncnnoptimize再增加自定义层，避免ncnnoptimize不能处理自定义层保存。
+ 先ncnnoptimize再增加自定义层，避免ncnnoptimize不能处理自定义层保存。
-## rtti/exceptions冲突
+- ## rtti/exceptions冲突
-产生原因是项目工程中使用的库配置不一样导致冲突，根据自己的实际情况分析是需要开启还是关闭。ncnn默认是ON，在重新编译ncnn时增加以下2个参数即可：
- - 开启：-DNCNN_DISABLE_RTTI=OFF -DNCNN_DISABLE_EXCEPTION=OFF
- - 关闭：-DNCNN_DISABLE_RTTI=ON -DNCNN_DISABLE_EXCEPTION=ON
+ 产生原因是项目工程中使用的库配置不一样导致冲突，根据自己的实际情况分析是需要开启还是关闭。ncnn默认是ON，在重新编译ncnn时增加以下2个参数即可：
+ - 开启：-DNCNN_DISABLE_RTTI=OFF -DNCNN_DISABLE_EXCEPTION=OFF
+ - 关闭：-DNCNN_DISABLE_RTTI=ON -DNCNN_DISABLE_EXCEPTION=ON
-## error: undefined symbol: ncnn::Extractor::extract(char const*, ncnn::Mat&)
+- ## error: undefined symbol: ncnn::Extractor::extract(char const*, ncnn::Mat&)
-可能的情况：
- - 尝试升级 Android Studio 的 NDK 版本
+ 可能的情况：
+ - 尝试升级 Android Studio 的 NDK 版本
# 怎样添加ncnn库到项目中？cmake方式怎么用？
编译ncnn，make install。linux/windows set/export ncnn_DIR 指向 isntall目录下下包含ncnnConfig.cmake 的目录
-## android
+- ## android
-## ios
+- ## ios
-## linux
+- ## linux
-## windows
+- ## windows
-## macos
+- ## macos
-## arm linux
+- ## arm linux
# 转模型问题
-## caffe
+- ## caffe
+
+ `./caffe2ncnn caffe.prototxt caffe.caffemodel ncnn.param ncnn.bin`
-`./caffe2ncnn caffe.prototxt caffe.caffemodel ncnn.param ncnn.bin`
+- ## mxnet
-## mxnet
+ ` ./mxnet2ncnn mxnet-symbol.json mxnet.params ncnn.param ncnn.bin`
-` ./mxnet2ncnn mxnet-symbol.json mxnet.params ncnn.param ncnn.bin`
+- ## darknet
-## darknet
+ [https://github.com/xiangweizeng/darknet2ncnn](https://github.com/xiangweizeng/darknet2ncnn)
-[https://github.com/xiangweizeng/darknet2ncnn](https://github.com/xiangweizeng/darknet2ncnn)
+- ## pytorch - onnx
-## pytorch - onnx
+ [use ncnn with pytorch or onnx](https://github.com/Tencent/ncnn/wiki/use-ncnn-with-pytorch-or-onnx)
-[use ncnn with pytorch or onnx](https://github.com/Tencent/ncnn/wiki/use-ncnn-with-pytorch-or-onnx)
+- ## tensorflow 1.x/2.x - keras
-## tensorflow 1.x/2.x - keras
+ [https://github.com/MarsTechHAN/keras2ncnn](https://github.com/MarsTechHAN/keras2ncnn) **[@MarsTechHAN](https://github.com/MarsTechHAN)**
-[https://github.com/MarsTechHAN/keras2ncnn](https://github.com/MarsTechHAN/keras2ncnn) **[@MarsTechHAN](https://github.com/MarsTechHAN)**
+- ## tensorflow 2.x - mlir
-## tensorflow 2.x - mlir
+ [通过MLIR将tensorflow2模型转换到ncnn](https://zhuanlan.zhihu.com/p/152535430) **@[nihui](https://www.zhihu.com/people/nihui-2)**
-[通过MLIR将tensorflow2模型转换到ncnn](https://zhuanlan.zhihu.com/p/152535430) **@[nihui](https://www.zhihu.com/people/nihui-2)**
+- ## Shape not supported yet! Gather not supported yet! Cast not supported yet!
-## Shape not supported yet! Gather not supported yet! Cast not supported yet!
+ onnx-simplifier 静态shape
-onnx-simplifier 静态shape
+- ## convertmodel
-## convertmodel
+ [https://convertmodel.com/](https://convertmodel.com/) **[@大老师](https://github.com/daquexian)**
-[https://convertmodel.com/](https://convertmodel.com/) **[@大老师](https://github.com/daquexian)**
+- ## netron
-## netron
+ [https://github.com/lutzroeder/netron](https://github.com/lutzroeder/netron)
-[https://github.com/lutzroeder/netron](https://github.com/lutzroeder/netron)
+- ## 怎么生成有固定 shape 信息的模型？
-## 怎么生成有固定 shape 信息的模型？
+ Input 0=w 1=h 2=c
-Input 0=w 1=h 2=c
+- ## why gpu能更快
-why gpu能更快
+- ## ncnnoptimize 怎么转成 fp16 模型
-## ncnnoptimize 怎么转成 fp16 模型
+ `ncnnoptimize model.param model.bin yolov5s-opt.param yolov5s-opt.bin 65536`
-`ncnnoptimize model.param model.bin yolov5s-opt.param yolov5s-opt.bin 65536`
+- ## ncnnoptimize 怎样查看模型的 FLOPS / 内存占用情况
-## ncnnoptimize 怎样查看模型的 FLOPS / 内存占用情况
+- ## 怎么修改模型支持动态 shape？
-## 怎么修改模型支持动态 shape？
+ Interp Reshape
-Interp Reshape
+- ## 如何将模型转换为代码内嵌到程序里？
-## 如何将模型转换为代码内嵌到程序里？
+ ncnn2mem
-ncnn2mem
+- ## 如何加密模型？
-## 如何加密模型？
+ https://zhuanlan.zhihu.com/p/268327784
-https://zhuanlan.zhihu.com/p/268327784
+- ## Linux下转的ncnn模型，Windows/MacOS/Android/.. 也能直接用吗？
-## Linux下转的ncnn模型，Windows/MacOS/Android/.. 也能直接用吗？
+ Yes，全平台通用
-Yes，全平台通用
+- ## 如何去掉后处理，再导出 onnx？
-## 如何去掉后处理，再导出 onnx？
+ 检测：
-检测
-参考up的一篇文章(https://zhuanlan.zhihu.com/p/128974102)，步骤三就是去掉后处理,再导出onnx,其中去掉后处理可以是项目内测试时去掉后续步骤的结果。
+ 参考up的一篇文章(https://zhuanlan.zhihu.com/p/128974102)，步骤三就是去掉后处理,再导出onnx,其中去掉后处理可以是项目内测试时去掉后续步骤的结果。
-## pytorch 有的层导不出 onnx 怎么办？
+- ## pytorch 有的层导不出 onnx 怎么办？
-ONNX_ATEN_FALLBACK
+ ONNX_ATEN_FALLBACK
完全自定义的op，先改成能导出的（如 concat slice），转到 ncnn 后再修改 param
# 使用
-## vkEnumeratePhysicalDevices failed -3
+- ## vkEnumeratePhysicalDevices failed -3
-## vkCreateInstance failed -9
+- ## vkCreateInstance failed -9
-驱动
+ 驱动
-## ModuleNotFoundError: No module named 'ncnn.ncnn'
+- ## ModuleNotFoundError: No module named 'ncnn.ncnn'
-python setup.py develop
+ python setup.py develop
-## fopen nanodet-m.param failed
+- ## fopen nanodet-m.param failed
-文件路径
-working dir
+ 文件路径 working dir
-File not found or not readable. Make sure that XYZ.param/XYZ.bin is accessible.
+ File not found or not readable. Make sure that XYZ.param/XYZ.bin is accessible.
-## find_blob_index_by_name data / output / ... failed
+- ## find_blob_index_by_name data / output / ... failed
-layer name vs blob name
-param.bin 应该用 xxx.id.h 的枚举
+ layer name vs blob name
+
+ param.bin 应该用 xxx.id.h 的枚举
-## parse magic failed
+- ## parse magic failed
-## param is too old, please regenerate
+- ## param is too old, please regenerate
-模型本身有问题
+ 模型本身有问题
-Your model file is being the old format converted by an old caffe2ncnn tool.
+ Your model file is being the old format converted by an old caffe2ncnn tool.
-Checkout the latest ncnn code, build it and regenerate param and model binary files, and that should work.
+ Checkout the latest ncnn code, build it and regenerate param and model binary files, and that should work.
-Make sure that your param file starts with the magic number 7767517.
+ Make sure that your param file starts with the magic number 7767517.
-you may find more info on use-ncnn-with-alexnet
+ you may find more info on use-ncnn-with-alexnet
-## set_vulkan_compute failed, network use_vulkan_compute disabled
+- ## set_vulkan_compute failed, network use_vulkan_compute disabled
-你应该在 load_param / load_model 之前设置 net.opt.use_vulkan_compute = true;
+ 你应该在 load_param / load_model 之前设置 net.opt.use_vulkan_compute = true;
-## 多个blob输入，多个blob输出，怎么做？
-多次执行`ex.input()` 和 `ex.extract()`
+- ## 多个blob输入，多个blob输出，怎么做？
+ 多次执行`ex.input()` 和 `ex.extract()`
```
ex.input("data1", in);
ex.input("data2", in);
ex.extract("output1", out);
ex.extract("output2", out);
```
-## Extractor extract 多次会重复计算吗？
+- ## Extractor extract 多次会重复计算吗？
-不会
+ 不会
-## 如何看每一层的耗时？
+- ## 如何看每一层的耗时？
-cmake -DNCNN_BENCHMARK=ON ..
+ cmake -DNCNN_BENCHMARK=ON ..
-## 如何转换 cv::Mat CV_8UC3 BGR 图片
+- ## 如何转换 cv::Mat CV_8UC3 BGR 图片
-from_pixels to_pixels
+ from_pixels to_pixels
-## 如何转换 float 数据为 ncnn::Mat
+- ## 如何转换 float 数据为 ncnn::Mat
-https://github.com/Tencent/ncnn/wiki/use-ncnn-with-pytorch-or-onnx
+ [use-ncnn-with-pytorch-or-onnx](https://github.com/Tencent/ncnn/wiki/use-ncnn-with-pytorch-or-onnx)
-## 如何初始化 ncnn::Mat 为全 0
+- ## 如何初始化 ncnn::Mat 为全 0
-`mat.fill(0.f);`
+ `mat.fill(0.f);`
-## 如何查看／获取版本号
+- ## 如何查看／获取版本号
-cmake时会打印
+ cmake时会打印
-c_api.h ncnn_version()
+ c_api.h ncnn_version()
-（自己拼 1.0+yyyymmdd
+ 自己拼 1.0+yyyymmdd
-## 如何转换 yuv 数据
+- ## 如何转换 yuv 数据
-yuv420sp2rgb yuv420sp2rgb_nv12
+ yuv420sp2rgb yuv420sp2rgb_nv12
-**[@zz大佬](https://github.com/zchrissirhcz/xxYUV)**
+ **[@zz大佬](https://github.com/zchrissirhcz/xxYUV)**
-## 如何 resize crop rotate 图片
+- ## 如何 resize crop rotate 图片
-[efficient roi resize rotate](https://github.com/Tencent/ncnn/wiki/efficient-roi-resize-rotate)
+ [efficient roi resize rotate](https://github.com/Tencent/ncnn/wiki/efficient-roi-resize-rotate)
-## 如何人脸5点对齐
+- ## 如何人脸5点对齐
-get_affine_transform
-warpaffine_bilinear_c3
+ get_affine_transform
+
+ warpaffine_bilinear_c3
```c
// 计算变换矩阵 并且求逆变换
@@ -329,23 +335,25 @@ for(int c = 0; c < 3; c++)
ncnn::warpaffine_bilinear_c3(pSrc, SrcWidth, SrcHeight, SrcStride, pDst, DstWidth, DstHeight, DstStride, tm_inv, type, v);
```
-## 如何获得中间层的blob输出
+- ## 如何获得中间层的blob输出
+
+ ncnn::Mat output;
+
+ ex.extract("your_blob_name", output);
- ncnn::Mat output;
- ex.extract("your_blob_name", output);
+- ## 为什么我使用GPU，但是GPU占用为0
-## 为什么我使用GPU，但是GPU占用为0
+ windows 10 任务管理器 - 性能选项卡 - GPU - 选择其中一个视图左上角的下拉箭头切换到 Compute_0 / Compute_1 / Cuda
-windows 10 任务管理器 - 性能选项卡 - GPU - 选择其中一个视图左上角的下拉箭头切换到 Compute_0 / Compute_1 / Cuda
-你还可以安装软件：GPU-Z
+ 你还可以安装软件：GPU-Z
-## layer XYZ not exists or registered
+- ## layer XYZ not exists or registered
-Your network contains some operations that are not implemented in ncnn.
+ Your network contains some operations that are not implemented in ncnn.
-You may implement them as custom layer followed in how-to-implement-custom-layer-step-by-step.
+ You may implement them as custom layer followed in how-to-implement-custom-layer-step-by-step.
-Or you could simply register them as no-op if you are sure those operations make no sense.
+ Or you could simply register them as no-op if you are sure those operations make no sense.
```
class Noop : public ncnn::Layer {};
@@ -355,68 +363,68 @@ net.register_custom_layer("LinearRegressionOutput", Noop_layer_creator);
net.register_custom_layer("MAERegressionOutput", Noop_layer_creator);
```
-## network graph not ready
+- ## network graph not ready
-You shall call Net::load_param() first, then Net::load_model().
+ You shall call Net::load_param() first, then Net::load_model().
-This error may also happens when Net::load_param() failed, but not properly handled.
+ This error may also happens when Net::load_param() failed, but not properly handled.
-For more information about the ncnn model load api, see ncnn-load-model
+ For more information about the ncnn model load api, see ncnn-load-model
-## memory not 32-bit aligned at XYZ
+- ## memory not 32-bit aligned at XYZ
-The pointer passed to Net::load_param() or Net::load_model() is not 32bit aligned.
+ The pointer passed to Net::load_param() or Net::load_model() is not 32bit aligned.
-In practice, the head pointer of std::vector is not guaranteed to be 32bit aligned.
+ In practice, the head pointer of std::vector is not guaranteed to be 32bit aligned.
-you can store your binary buffer in ncnn::Mat structure, its internal memory is aligned.
+ you can store your binary buffer in ncnn::Mat structure, its internal memory is aligned.
-## crash on android with '__kmp_abort_process'
+- ## crash on android with '__kmp_abort_process'
-This usually happens if you bundle multiple shared library with openmp linked
+ This usually happens if you bundle multiple shared library with openmp linked
-It is actually an issue of the android ndk https://github.com/android/ndk/issues/1028
+ It is actually an issue of the android ndk https://github.com/android/ndk/issues/1028
-On old android ndk, modify the link flags as
+ On old android ndk, modify the link flags as
--Wl,-Bstatic -lomp -Wl,-Bdynamic
+ -Wl,-Bstatic -lomp -Wl,-Bdynamic
-For recent ndk >= 21
+ For recent ndk >= 21
--fstatic-openmp
+ -fstatic-openmp
-## dlopen failed: library "libomp.so" not found
-Newer android ndk defaults to dynamic openmp runtime
+- ## dlopen failed: library "libomp.so" not found
+ Newer android ndk defaults to dynamic openmp runtime
-modify the link flags as
+ modify the link flags as
--fstatic-openmp -fopenmp
+ -fstatic-openmp -fopenmp
-## crash when freeing a ncnn dynamic library(.dll/.so) built with openMP
+- ## crash when freeing a ncnn dynamic library(.dll/.so) built with openMP
-for optimal performance, the openmp threadpool spin waits for about a second prior to shutting down in case more work becomes available.
+ for optimal performance, the openmp threadpool spin waits for about a second prior to shutting down in case more work becomes available.
-If you unload a dynamic library that's in the process of spin-waiting, it will crash in the manner you see (most of the time).
+ If you unload a dynamic library that's in the process of spin-waiting, it will crash in the manner you see (most of the time).
-Just set OMP_WAIT_POLICY=passive in your environment, before calling loadlibrary. or Just wait a few seconds before calling freelibrary.
+ Just set OMP_WAIT_POLICY=passive in your environment, before calling loadlibrary. or Just wait a few seconds before calling freelibrary.
-You can also use the following method to set environment variables in your code:
+ You can also use the following method to set environment variables in your code:
-for msvc++:
+ for msvc++:
- SetEnvironmentVariable(_T("OMP_WAIT_POLICY"), _T("passive"));
+ SetEnvironmentVariable(_T("OMP_WAIT_POLICY"), _T("passive"));
-for g++:
+ for g++:
- setenv("OMP_WAIT_POLICY", "passive", 1)
+ setenv("OMP_WAIT_POLICY", "passive", 1)
- reference: https://stackoverflow.com/questions/34439956/vc-crash-when-freeing-a-dll-built-with-openmp
+ reference: https://stackoverflow.com/questions/34439956/vc-crash-when-freeing-a-dll-built-with-openmp
# 跑出来的结果对不上
[ncnn-produce-wrong-result](https://github.com/Tencent/ncnn/wiki/FAQ-ncnn-produce-wrong-result)
-## 如何打印 ncnn::Mat 的值？
+- ## 如何打印 ncnn::Mat 的值？
```C++
void pretty_print(const ncnn::Mat& m)
@@ -456,7 +464,7 @@ void pretty_print(const ncnn::Mat& m)
}
```
-## 如何可视化 ncnn::Mat 的值？
+- ## 如何可视化 ncnn::Mat 的值？
```
void visualize(const char* title, const ncnn::Mat& m)
@@ -511,30 +519,33 @@ void visualize(const char* title, const ncnn::Mat& m)
}
```
+- ## 总是输出第一张图的结果
+ 复用 Extractor？！
+- ## 启用fp16时的精度有差异
-## 总是输出第一张图的结果
-复用 Extractor？！
+ net.opt.use_fp16_packed = false;
-## 启用fp16时的精度差异
+ net.opt.use_fp16_storage = false;
-net.opt.use_fp16_packed = false;
-net.opt.use_fp16_storage = false;
-net.opt.use_fp16_arithmetic = false;
+ net.opt.use_fp16_arithmetic = false;
-https://github.com/Tencent/ncnn/wiki/FAQ-ncnn-produce-wrong-result
+ [ncnn-produce-wrong-result](https://github.com/Tencent/ncnn/wiki/FAQ-ncnn-produce-wrong-result)
# 如何跑得更快？内存占用更少？库体积更小？
-## fp32 fp16
+- ## fp32 fp16
-## 大小核绑定
-ncnn::set_cpu_powersave(int)绑定大核或小核
-注意windows系统不支持绑核。
-ncnn支持不同的模型运行在不同的核心。假设硬件平台有2个大核，4个小核，你想把netA运行在大核，netB运行在小核。
-可以通过std::thread or pthread创建两个线程，运行如下代码：
+- ## 大小核绑定
+ ncnn::set_cpu_powersave(int)绑定大核或小核
+ 注意windows系统不支持绑核。
+ ncnn支持不同的模型运行在不同的核心。假设硬件平台有2个大核，4个小核，你想把netA运行在大核，netB运行在小核。
+ 可以通过std::thread or pthread创建两个线程，运行如下代码：
+ 0:全部
+ 1:小核
+ 2:大核
```
void thread_1()
{
@@ -548,73 +559,78 @@ ncnn支持不同的模型运行在不同的核心。假设硬件平台有2个大
netB.opt.num_threads = 4;
}
```
-https://github.com/Tencent/ncnn/blob/master/docs/how-to-use-and-FAQ/openmp-best-practice.zh.md
-## get_cpu_count get_gpu_count
+ [openmp-best-practice.zh.md](https://github.com/Tencent/ncnn/blob/master/docs/how-to-use-and-FAQ/openmp-best-practice.zh.md)
+
+- ## 查看 CPU 或 GPU 数量
+ get_cpu_count
+
+ get_gpu_count
-## ncnnoptimize
+- ## ncnnoptimize
-使用方式一：
- - ./ncnnoptimize ncnn.param ncnn.bin new.param new.bin flag
-
注意这里的flag指的是fp32和fp16，其中0指的是fp32，1指的是fp16
+ 使用方式一：
+ - ./ncnnoptimize ncnn.param ncnn.bin new.param new.bin flag
+
注意这里的flag指的是fp32和fp16，其中0指的是fp32，1指的是fp16
-使用方式二：
- - ./ncnnoptimize ncnn.param ncnn.bin new.param new.bin flag cutstartname cutendname
-
cutstartname: 模型截取的起点
-
cutendname: 模型截取的终点
+ 使用方式二：
+ - ./ncnnoptimize ncnn.param ncnn.bin new.param new.bin flag cutstartname cutendname
+
cutstartname：模型截取的起点
+
cutendname：模型截取的终点
-## 如何使用量化工具？
+- ## 如何使用量化工具？
-[Post Training Quantization Tools](https://github.com/Tencent/ncnn/tree/master/tools/quantize)
+ [Post Training Quantization Tools](https://github.com/Tencent/ncnn/tree/master/tools/quantize)
-## 如何设置线程数？
+- ## 如何设置线程数？
-opt.num_threads
+ opt.num_threads
-## 如何降低CPU占用率？
+- ## 如何降低CPU占用率？
-net.opt.openmp_blocktime = 0;
-OMP_WAIT_POLICY=passive
+ net.opt.openmp_blocktime = 0;
+
+ OMP_WAIT_POLICY=passive
-## 如何 batch inference？
+- ## 如何 batch inference？
-## partial graph inference
+- ## partial graph inference
-先 extract 分类，判断后，再 extract bbox
+ 先 extract 分类，判断后，再 extract bbox
-## 如何启用 bf16s 加速？
+- ## 如何启用 bf16s 加速？
```
net.opt.use_packing_layout = true;
net.opt.use_bf16_storage = true;
```
-[用bf16加速ncnn](https://zhuanlan.zhihu.com/p/112564372) **@[nihui](https://www.zhihu.com/people/nihui-2)**
+ [用bf16加速ncnn](https://zhuanlan.zhihu.com/p/112564372) **@[nihui](https://www.zhihu.com/people/nihui-2)**
-A53
+ A53
-## 如何裁剪更小的 ncnn 库？
+- ## 如何裁剪更小的 ncnn 库？
-https://github.com/Tencent/ncnn/wiki/build-minimal-library
+ [build-minimal-library](https://github.com/Tencent/ncnn/wiki/build-minimal-library)
-## net.opt sgemm winograd fp16_storage 各是有什么作用？
+- ## net.opt sgemm winograd fp16_storage 各是有什么作用？
-对内存消耗的影响
+ 对内存消耗的影响
# 白嫖项目
-## nanodet
+- ## nanodet
# 其他
-## up主用的什么系统/编辑器/开发环境？
+- ## up主用的什么系统/编辑器/开发环境？
-| 软件类型 | 软件名称 |
-| ------------| ----------- |
-| 系统 | Fedora |
-| 桌面环境 | KDE |
-| 编辑器 | Kate |
-| 画草图 | kolourpaint |
-| 画函数图像 | kmplot |
-| bilibili直播 | OBS |
+ | 软件类型 | 软件名称 |
+ | ------------| ----------- |
+ | 系统 | Fedora |
+ | 桌面环境 | KDE |
+ | 编辑器 | Kate |
+ | 画草图 | kolourpaint |
+ | 画函数图像 | kmplot |
+ | bilibili直播 | OBS |