下面是我的示例过程和验证结果，为什么验证的输出结果出错，但是手动运行 hhb_runtime好像成功了，是不是revyos内核太老了，导致的不兼容呢？
RuyiSDK示例
##安装依赖包
###安装docker
sudo apt update
sudo apt install -y ca-certificates curl
# 下载 Docker 官方安装脚本并执行
curl -fsSL https://get.docker.com -o get-docker.sh
sudo sh get-docker.sh
安装成功后查看docker及其状态:
``bash
docker --version
sudo systemctl status docker
在终端显示如下：
```text
licheepi@licheepi-virtual-machine:~$ docker --version
Docker version 29.4.1, build 055a478
licheepi@licheepi-virtual-machine:~$ sudo systemctl status docker
● docker.service - Docker Application Container Engine
Loaded: loaded (/lib/systemd/system/docker.service; enabled;>
Active: active (running) since Fri 2026-04-24 13:35:46 CST;
将当前用户加入 docker 组（免 sudo 运行）
sudo usermod -aG docker $USER
newgrp docker
拉取 HHB Docker 镜像
docker pull hhb4tools/hhb:latest
创建容器，名字为hhb_env
docker run -itd --name hhb_env hhb4tools/hhb:latest
#进入容器
docker exec -it hhb_env /bin/bash
在终端显示如下：
licheepi@licheepi-virtual-machine:~$ docker run -itd --name hhb_env hhb4tools/hhb:latest
78776422f7c978472da9c903bcba5804cae8c7f83f410a5b7fbd2989dbc2e877
licheepi@licheepi-virtual-machine:~$ docker exec -it hhb_env /bin/bash
root@78776422f7c9:/#
下载Yolov5n模型
下载到示例目录 /home/example/th1520_npu/yolov5n 下：
mkdir -p /home/example/th1520_npu/yolov5n
cd /home/example/th1520_npu/yolov5n
git clone https://gitclone.com/github.com/ultralytics/yolov5.git
cd yolov5
pip3 install ultralytics
如果遇到版本冲突的问题，强制重装torch和torchvision
pip3 install --upgrade torch==2.4.1 torchvision==0.19.1
安装所需依赖
pip3 install pandas
pip3 install seaborn
python3 export.py --weights yolov5n.pt --include onnx --imgsz 384 640
# 将导出的 ONNX 文件复制到示例目录
cp yolov5n.onnx ../
HHB编译模型：
将 ONNX 模型交叉编译成 NPU 上可执行的程序，需要使用 hhb 命令。
编译时需要先进入到示例所在目录 /home/example/th1520_npu/yolov5n
cd /home/example/th1520_npu/yolov5n
hhb -D --model-file yolov5n.onnx --data-scale-div 255 --board th1520 --input-name "images" --output-name "/model.24/m.0/Conv_output_0;/model.24/m.1/Conv_output_0;/model.24/m.2/Conv_output_0" --input-shape "1 3 384 640" --calibrate-dataset kite.jpg --quantization-scheme "int8_asym"
在终端显示如下：
root@78776422f7c9:/home/example/th1520_npu/yolov5n# hhb -D --model-file yolov5n.onnx --data-scale-div 255 --board th1520 --input-name "images" --output-name "/model.24/m.0/Conv_output_0;/model.24/m.1/Conv_output_0;/model.24/m.2/Conv_output_0" --input-shape "1 3 384 640" --calibrate-dataset kite.jpg --quantization-scheme "int8_asym"
[2026-04-27 06:12:54] (HHB LOG): Start import model.
[2026-04-27 06:12:55] (HHB LOG): Model import completed!
[2026-04-27 06:12:55] (HHB LOG): Start quantization.
[2026-04-27 06:12:55] (HHB LOG): get calibrate dataset from kite.jpg
[2026-04-27 06:12:55] (HHB LOG): Start optimization.
[2026-04-27 06:12:55] (HHB LOG): Optimization completed!
Calibrating: 100%|██████████████████| 316/316 [00:23<00:00, 13.21it/s]
[2026-04-27 06:13:19] (HHB LOG): Start conversion to csinn.
[2026-04-27 06:13:20] (HHB LOG): Conversion completed!
[2026-04-27 06:13:20] (HHB LOG): Start operator fusion.
[2026-04-27 06:13:20] (HHB LOG): Operator fusion completed!
[2026-04-27 06:13:21] (HHB LOG): Start operator split.
[2026-04-27 06:13:21] (HHB LOG): Operator split completed!
[2026-04-27 06:13:21] (HHB LOG): Start layout convert.
[2026-04-27 06:13:21] (HHB LOG): Layout convert completed!
[2026-04-27 06:13:21] (HHB LOG): Quantization completed!
安装 RuyiSDK
# 下载并安装 ruyi
wget https://mirror.iscas.ac.cn/ruyisdk/ruyi/tags/0.47.0/ruyi-0.47.0.amd64
chmod +x ruyi-0.47.0.amd64
sudo cp ruyi-0.47.0.amd64 /usr/local/bin/ruyi
安装工具链
ruyi update
ruyi install gnu-plct-xthead
会在终端看到如下输出：
info: skipping already installed package gnu-plct-xthead-3.1.0-ruyi.20250526
创建并激活 ruyi 虚拟环境
# 创建虚拟环境 venv-sipeed
ruyi venv -t gnu-plct-xthead sipeed-lpi4a venv-sipeed
#激活虚拟环境
. venv-sipeed/bin/ruyi-activate
#查看当前虚拟环境下的 gcc 是否可用
riscv64-plctxthead-linux-gnu-gcc --version
在终端显示如下：
licheepi@licheepi-virtual-machine:~$ . venv-sipeed/bin/ruyi-activate
«Ruyi venv-sipeed» licheepi@licheepi-virtual-machine:~$ riscv64-plctxthead-linux-gnu-gcc --version
riscv64-plctxthead-linux-gnu-gcc (RuyiSDK 20250526 Xuantie-Sources Xuantie-binutils-8ee62aac8606 Xuantie-gcc-c2e0bcc86d1f Xuantie-glibc-29dd660835c5) 14.1.1 20240710
Copyright (C) 2024 Free Software Foundation, Inc.
This is free software; see the source for copying conditions. There is NO
warranty; not even for MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.
YOLOV5n测试示例
创建并激活 ruyi 虚拟环境
创建虚拟环境 venv-sipeed
ruyi venv -t gnu-plct-xthead sipeed-lpi4a venv-sipeed
激活虚拟环境
. venv-sipeed/bin/ruyi-activate
终端显示如下：
licheepi@licheepi-virtual-machine:~$ ruyi venv -t gnu-plct-xthead sipeed-lpi4a venv-sipeed
info: Creating a Ruyi virtual environment at venv-sipeed...
info: The virtual environment is now created.
You may activate it by sourcing the appropriate activation script in the
bin directory, and deactivate by invoking `ruyi-deactivate`.
A fresh sysroot/prefix is also provisioned in the virtual environment.
It is available at the following path:
/home/licheepi/venv-sipeed/sysroot
The virtual environment also comes with ready-made CMake toolchain file
and Meson cross file. Check the virtual environment root for those;
comments in the files contain usage instructions.
licheepi@licheepi-virtual-machine:~$ . venv-sipeed/bin/ruyi-activate
«Ruyi venv-sipeed» licheepi@licheepi-virtual-machine:~$
使用 ruyi 工具链编译示例代码
确认交叉编译器可用
riscv64-plctxthead-linux-gnu-gcc --version
复制必要的文件到宿主机
# 复制 hhb_out 目录（包含 io.c, model.c, process.c 等）
docker cp hhb_env:/home/example/th1520_npu/yolov5n/hhb_out ./
# 复制 NPU SDK（头文件和静态库）
docker cp hhb_env:/usr/local/lib/python3.8/dist-packages/hhb/install_nn2/th1520 ./th1520_npu_sdk
# 复制 c920 的静态库
docker cp hhb_env:/usr/local/lib/python3.8/dist-packages/hhb/install_nn2/c920/lib/. ./th1520_npu_sdk/lib/
# 复制主程序文件
docker cp hhb_env:/home/example/th1520_npu/yolov5n/yolov5n.c ./
交叉编译libjpeg
cd /home/licheepi
git clone https://github.com/libjpeg-turbo/libjpeg-turbo.git
cd libjpeg-turbo
mkdir build && cd build
cmake .. \
-DCMAKE_C_COMPILER=riscv64-plctxthead-linux-gnu-gcc \
-DCMAKE_INSTALL_PREFIX=/home/licheepi/venv-sipeed/sysroot.riscv64-plctxthead-linux-gnu/usr \
-DENABLE_SHARED=OFF \
-DWITH_JPEG8=ON
make -j$(nproc)
make install
交叉编译zlib
cd /home/licheepi
wget https://github.com/madler/zlib/archive/refs/tags/v1.3.1.tar.gz -O zlib-1.3.1.tar.gz
tar -xzvf zlib-1.3.1.tar.gz
cd zlib-1.3.1
export CC=riscv64-plctxthead-linux-gnu-gcc
export AR=riscv64-plctxthead-linux-gnu-ar
./configure --prefix=/home/licheepi/venv-sipeed/sysroot.riscv64-plctxthead-linux-gnu/usr --static
make -j$(nproc)
make install
交叉编译 libpng
cd /home/licheepi
wget https://download.sourceforge.net/libpng/libpng-1.6.40.tar.gz
tar -xzvf libpng-1.6.40.tar.gz
cd libpng-1.6.40
./configure \
--host=riscv64-plctxthead-linux-gnu \
--prefix=/home/licheepi/venv-sipeed/sysroot.riscv64-plctxthead-linux-gnu/usr \
--enable-static \
--disable-shared \
CPPFLAGS="-I/home/licheepi/venv-sipeed/sysroot.riscv64-plctxthead-linux-gnu/usr/include" \
LDFLAGS="-L/home/licheepi/venv-sipeed/sysroot.riscv64-plctxthead-linux-gnu/usr/lib"
make -j$(nproc)
make install
解决 omp.h 缺失问题
mkdir -p /home/licheepi/venv-sipeed/sysroot.riscv64-plctxthead-linux-gnu/usr/include
docker cp hhb_env:/usr/lib/gcc/x86_64-linux-gnu/9/include/omp.h \
/home/licheepi/venv-sipeed/sysroot.riscv64-plctxthead-linux-gnu/usr/include/
修复 yolov5n.c 中的函数名问题
# 替换不兼容的函数名
sed -i 's/shl_ref_f32_to_input_dtype/shl_c920_f32_to_input_dtype/g' yolov5n.c
最终编译命令：
riscv64-plctxthead-linux-gnu-gcc \
yolov5n.c \
hhb_out/io.c \
hhb_out/model.c \
hhb_out/process.c \
-o yolov5n_example \
-I hhb_out/ \
-I th1520_npu_sdk/include/ \
-I th1520_npu_sdk/include/shl_public/ \
-I th1520_npu_sdk/include/csinn/ \
-I . \
th1520_npu_sdk/lib/libshl_c920.a \
/home/licheepi/venv-sipeed/sysroot.riscv64-plctxthead-linux-gnu/usr/lib/libjpeg.a \
/home/licheepi/venv-sipeed/sysroot.riscv64-plctxthead-linux-gnu/usr/lib/libpng.a \
/home/licheepi/venv-sipeed/sysroot.riscv64-plctxthead-linux-gnu/usr/lib/libz.a \
-lstdc++ -lpthread -ldl -lm \
-mabi=lp64d -march=rv64gcv0p7_zfh_xtheadc \
-Wl,--gc-sections -O2 -g
检查生成的文件：
ls -la yolov5n_example
# 查看文件类型
file yolov5n_example
终端显示如下：
«Ruyi venv-sipeed» licheepi@licheepi-virtual-machine:~$ ls -la yolov5n_example
-rwxrwxr-x 1 licheepi licheepi 9640776 4月 27 15:25 yolov5n_example
«Ruyi venv-sipeed» licheepi@licheepi-virtual-machine:~$ file yolov5n_example
yolov5n_example: ELF 64-bit LSB executable, UCB RISC-V, RVC, double-float ABI, version 1 (SYSV), dynamically linked, interpreter /lib/ld-linux-riscv64-lp64d.so.1, BuildID[sha1]=a581dfa6e00424f79d7da7b9659cd5fbb5172789, for GNU/Linux 4.15.0, with debug_info, not stripped
运行示例并验证结果
# 传输可执行文件
scp /home/licheepi/yolov5n_example debian@172.16.60.217:~/npu_test/
# 传输 hhb_out 目录（包含模型文件）
scp -r /home/licheepi/hhb_out debian@172.16.60.217:~/npu_test/
# 传输测试图片
scp /home/licheepi/kite.jpg debian@172.16.60.217:~/npu_test/
运行 NPU 推理程序
cd ~/npu_test
chmod +x yolov5n_example
./yolov5n_example ./hhb_out/hhb.bm ./kite.jpg
验证结果：
Run graph execution time: 0.00133ms, FPS=749625.19
程序推理完成后因后处理代码与当前 SHL 库版本存在兼容性问题而崩溃，检测结果未能完全输出。但 NPU 推理核心功能已验证成功，RuyiSDK 工具链的交叉编译能力通过验证。
若需要完整的推理输出，可使用 HHB 直接生成的 hhb_runtime 程序：
cd ~/npu_test/hhb_out
export LD_LIBRARY_PATH=/home/debian/ort/lib/python3.11/site-packages/shl/install_nn2/th1520/lib:$LD_LIBRARY_PATH
./hhb_runtime ./hhb.bm ../kite.jpg
成功后输出：
Run graph execution time: 18.57649ms, FPS=53.83
=== tensor info ===
shape: 1 3 384 640
data pointer: 0x1073260
=== tensor info ===
shape: 1 255 48 80
data pointer: 0x3fbae2f000
The max_value of output: 3.443873
The min_value of output: -17.471357
The mean_value of output: -5.491803
The std_value of output: 4.574095
============ top5: ===========
329636: 3.443873
656036: 3.443873
667150: 3.191882
800655: 3.191882
800733: 3.191882
=== tensor info ===
shape: 1 255 24 40
data pointer: 0x3fbadf3000
The max_value of output: 6.665792
The min_value of output: -14.283840
The mean_value of output: -5.499501
The std_value of output: 3.713140
============ top5: ===========
168619: 6.665792
168579: 6.232948
168620: 5.973242
5419: 5.626967
87019: 5.453830
=== tensor info ===
shape: 1 255 12 20
data pointer: 0x3fbcd7f000
The max_value of output: 5.638581
The min_value of output: -13.231871
The mean_value of output: -6.027710
The std_value of output: 3.766955
============ top5: ===========
21749: 5.638581
1349: 5.187495
21769: 5.112314
21787: 5.112314
为什么验证的输出结果出错，但是手动运行 hhb_runtime好像成功了，是不是revyos内核太老了，导致的不兼容呢？