|
- Version: Ryzen AI Software Platform v0.7
- Support: AMD Ryzen 7040U, 7040HS series mobile processors with Windows 11 OS.
- Last update: 7 Sep. 2023
Ryzen™ AI is a dedicated AI accelerator integrated on-chip with the CPU cores. The AMD Ryzen™ AI SDK enables developers to take machine learning models trained in PyTorch or TensorFlow and run them on laptops powered by Ryzen AI which can intelligently optimizes tasks and workloads, freeing-up CPU and GPU resources, and ensuring optimal performance at lower power.
In this Deep Learning(DL) tutorial, you will see how to deploy the Yolov8 detection model with ONNX framework on Ryzen AI laptop.
- Linux server (GPU is preferred)
- AMD Ryzen AI Laptop with Windows 11 OS
- Visual Studio 2019 (with Desktop dev c++ & MSVC v142-vs2019 x64/x86 Spectre-mitigated libs)
- Anaconda or Miniconda
- Git
- openCV (version = 4.6.0)
- glog
- gflags
- cmake (version >= 3.26)
- python (version >= 3.9) (Recommended for python 3.9.13 64bit)
- IPU driver & IPU xclbin release >= 20230726
- voe package >= (jenkins-nightly-build-id==205)
The Visual Studio is required to compile the related source code.
❗ The Yolov8 demo in this tutorial may not be compatible with other Visual Studio version.
Please download the Adrenalin IPU driver from the Link below and install it on your laptop.
Ensure that the IPU driver is installed from Device Manager -> System Devices -> AMD IPU Device as shown in the following image. The version is expected to be 10.106.6.52
The Ryzen AI Software Platform requires using a Conda environment for the installation process.
Start a Conda Prompt. In the Conda Prompt, create and activate an environment for the rest of the installation process. We will use ryzen_test here as an example.
# conda create --name ryzen_test python=3.9
# conda activate ryzen_test
Output:
## Package Plan ##
environment location: C:\Users\AMD\anaconda3\envs\yolov8
added / updated specs:
- python=3.9
The following NEW packages will be INSTALLED:
ca-certificates pkgs/main/win-64::ca-certificates-2023.05.30-haa95532_0
openssl pkgs/main/win-64::openssl-3.0.10-h2bbff1b_2
pip pkgs/main/win-64::pip-23.2.1-py39haa95532_0
python pkgs/main/win-64::python-3.9.17-h1aa4202_0
setuptools pkgs/main/win-64::setuptools-68.0.0-py39haa95532_0
sqlite pkgs/main/win-64::sqlite-3.41.2-h2bbff1b_0
tzdata pkgs/main/noarch::tzdata-2023c-h04d1e81_0
vc pkgs/main/win-64::vc-14.2-h21ff451_1
vs2015_runtime pkgs/main/win-64::vs2015_runtime-14.27.29016-h5e58377_2
wheel pkgs/main/win-64::wheel-0.38.4-py39haa95532_0
Proceed ([y]/n)? y
Downloading and Extracting Packages
Preparing transaction: done
Verifying transaction: done
Executing transaction: done
#
# To activate this environment, use
#
# $ conda activate yolov8
#
# To deactivate an active environment, use
#
# $ conda deactivate
C:\Users\AMD>conda activate yolov8
(yolov8) C:\Users\AMD>
# pip install onnxruntime
Output:
Collecting onnxruntime
Obtaining dependency information for onnxruntime from https://files.pythonhosted.org/packages/6a/fb/99bc0e75f3d23eab0dda640acaf23d0a3a68c3949a56ac5c25698eab4958/onnxruntime-1.15.1-cp39-cp39-win_amd64.whl.metadata
Using cached onnxruntime-1.15.1-cp39-cp39-win_amd64.whl.metadata (4.1 kB)
Collecting coloredlogs (from onnxruntime)
Using cached coloredlogs-15.0.1-py2.py3-none-any.whl (46 kB)
Collecting flatbuffers (from onnxruntime)
Obtaining dependency information for flatbuffers from https://files.pythonhosted.org/packages/6f/12/d5c79ee252793ffe845d58a913197bfa02ae9a0b5c9bc3dc4b58d477b9e7/flatbuffers-23.5.26-py2.py3-none-any.whl.metadata
Using cached flatbuffers-23.5.26-py2.py3-none-any.whl.metadata (850 bytes)
Collecting numpy>=1.21.6 (from onnxruntime)
Obtaining dependency information for numpy>=1.21.6 from https://files.pythonhosted.org/packages/df/18/181fb40f03090c6fbd061bb8b1f4c32453f7c602b0dc7c08b307baca7cd7/numpy-1.25.2-cp39-cp39-win_amd64.whl.metadata
Using cached numpy-1.25.2-cp39-cp39-win_amd64.whl.metadata (5.7 kB)
Collecting packaging (from onnxruntime)
Using cached packaging-23.1-py3-none-any.whl (48 kB)
Collecting protobuf (from onnxruntime)
Obtaining dependency information for protobuf from https://files.pythonhosted.org/packages/20/26/343db129c96d93a9d0820319a6beba4497aced1b0cd4b6051b0e4e1fd100/protobuf-4.24.2-cp39-cp39-win_amd64.whl.metadata
Using cached protobuf-4.24.2-cp39-cp39-win_amd64.whl.metadata (540 bytes)
Collecting sympy (from onnxruntime)
Using cached sympy-1.12-py3-none-any.whl (5.7 MB)
Collecting humanfriendly>=9.1 (from coloredlogs->onnxruntime)
Using cached humanfriendly-10.0-py2.py3-none-any.whl (86 kB)
Collecting mpmath>=0.19 (from sympy->onnxruntime)
Using cached mpmath-1.3.0-py3-none-any.whl (536 kB)
Collecting pyreadline3 (from humanfriendly>=9.1->coloredlogs->onnxruntime)
Using cached pyreadline3-3.4.1-py3-none-any.whl (95 kB)
Using cached onnxruntime-1.15.1-cp39-cp39-win_amd64.whl (6.7 MB)
Using cached numpy-1.25.2-cp39-cp39-win_amd64.whl (15.6 MB)
Using cached flatbuffers-23.5.26-py2.py3-none-any.whl (26 kB)
Using cached protobuf-4.24.2-cp39-cp39-win_amd64.whl (430 kB)
Installing collected packages: pyreadline3, mpmath, flatbuffers, sympy, protobuf, packaging, numpy, humanfriendly, coloredlogs, onnxruntime
Successfully installed coloredlogs-15.0.1 flatbuffers-23.5.26 humanfriendly-10.0 mpmath-1.3.0 numpy-1.25.2 onnxruntime-1.15.1 packaging-23.1 protobuf-4.24.2 pyreadline3-3.4.1 sympy-1.12
# pip install cmake
Output:
Collecting cmake
Obtaining dependency information for cmake from https://files.pythonhosted.org/packages/e0/67/3cc8ccb0cebac463033e1f8588328de32f8f85cfd9d3150c05b57b827893/cmake-3.27.4.1-py2.py3-none-win_amd64.whl.metadata
Downloading cmake-3.27.4.1-py2.py3-none-win_amd64.whl.metadata (6.8 kB)
Downloading cmake-3.27.4.1-py2.py3-none-win_amd64.whl (34.6 MB)
━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 34.6/34.6 MB 147.5 kB/s eta 0:00:00
Installing collected packages: cmake
Successfully installed cmake-3.27.4.1
Download the Execution Provider setup package with the link below:
Change the directory to the extracted Execution Provider setup package directory and install the necessary packages:
# cd voe-3.5-win_amd64\voe-3.5-win_amd64
# python installer.py
# pip install voe-0.1.0-cp39-cp39-win_amd64.whl
# pip install onnxruntime_vitisai-1.15.1-cp39-cp39-win_amd64.whl
Output:
2023-09-06 00:22:16,344 - INFO - copying C:\Windows\System32\AMD\xrt_core.dll to C:\Users\AMD\anaconda3\envs\yolov8\lib\site-packages\onnxruntime\capi
2023-09-06 00:22:16,373 - INFO - copying C:\Windows\System32\AMD\xrt_coreutil.dll to C:\Users\AMD\anaconda3\envs\yolov8\lib\site-packages\onnxruntime\capi
2023-09-06 00:22:16,405 - INFO - copying C:\Windows\System32\AMD\xrt_phxcore.dll to C:\Users\AMD\anaconda3\envs\yolov8\lib\site-packages\onnxruntime\capi
2023-09-06 00:22:16,428 - INFO - copying C:\Users\AMD\Downloads\IPU_Dependency\voe-3.5-win_amd64\voe-0.1.0-cp39-cp39-win_amd64\onnxruntime.dll to C:\Users\AMD\anaconda3\envs\yolov8\lib\site-packages\onnxruntime\capi
Processing c:\users\amd\downloads\ipu_dependency\voe-3.5-win_amd64\voe-0.1.0-cp39-cp39-win_amd64.whl
Collecting glog==0.3.1 (from voe==0.1.0)
Using cached glog-0.3.1-py2.py3-none-any.whl (7.8 kB)
Collecting python-gflags>=3.1 (from glog==0.3.1->voe==0.1.0)
Using cached python_gflags-3.1.2-py3-none-any.whl
Collecting six (from glog==0.3.1->voe==0.1.0)
Using cached six-1.16.0-py2.py3-none-any.whl (11 kB)
Installing collected packages: python-gflags, six, glog, voe
Successfully installed glog-0.3.1 python-gflags-3.1.2 six-1.16.0 voe-0.1.0
Processing c:\users\amd\downloads\ipu_dependency\voe-3.5-win_amd64\onnxruntime_vitisai-1.15.1-cp39-cp39-win_amd64.whl
Requirement already satisfied: coloredlogs in c:\users\amd\anaconda3\envs\yolov8\lib\site-packages (from onnxruntime-vitisai==1.15.1) (15.0.1)
Requirement already satisfied: flatbuffers in c:\users\amd\anaconda3\envs\yolov8\lib\site-packages (from onnxruntime-vitisai==1.15.1) (23.5.26)
Requirement already satisfied: numpy>=1.25.1 in c:\users\amd\anaconda3\envs\yolov8\lib\site-packages (from onnxruntime-vitisai==1.15.1) (1.25.2)
Requirement already satisfied: packaging in c:\users\amd\anaconda3\envs\yolov8\lib\site-packages (from onnxruntime-vitisai==1.15.1) (23.1)
Requirement already satisfied: protobuf in c:\users\amd\anaconda3\envs\yolov8\lib\site-packages (from onnxruntime-vitisai==1.15.1) (4.24.2)
Requirement already satisfied: sympy in c:\users\amd\anaconda3\envs\yolov8\lib\site-packages (from onnxruntime-vitisai==1.15.1) (1.12)
Requirement already satisfied: humanfriendly>=9.1 in c:\users\amd\anaconda3\envs\yolov8\lib\site-packages (from coloredlogs->onnxruntime-vitisai==1.15.1) (10.0)
Requirement already satisfied: mpmath>=0.19 in c:\users\amd\anaconda3\envs\yolov8\lib\site-packages (from sympy->onnxruntime-vitisai==1.15.1) (1.3.0)
Requirement already satisfied: pyreadline3 in c:\users\amd\anaconda3\envs\yolov8\lib\site-packages (from humanfriendly>=9.1->coloredlogs->onnxruntime-vitisai==1.15.1) (3.4.1)
Installing collected packages: onnxruntime-vitisai
Successfully installed onnxruntime-vitisai-1.15.1
It is recommended to build OpenCV form source code and use static build. Git is required to clone the repository.
Start a Git Bash. In the Git Bash, clone the repository
# git clone https://github.com/opencv/opencv.git -b 4.6.0
Switch back to the Conda Prompt, and compile the OpenCV source code with cmake.
# cmake -DCMAKE_EXPORT_COMPILE_COMMANDS=ON -DBUILD_SHARED_LIBS=OFF -DCMAKE_POSITION_INDEPENDENT_CODE=ON -DCMAKE_CONFIGURATION_TYPES=Release -A x64 -T host=x64 -G "Visual Studio 16 2019" '-DCMAKE_INSTALL_PREFIX=C:\Program Files\opencv' '-DCMAKE_PREFIX_PATH=.\opencv' -DCMAKE_BUILD_TYPE=Release -DBUILD_opencv_python2=OFF -DBUILD_opencv_python3=OFF -DBUILD_WITH_STATIC_CRT=OFF -B build -S opencv
# cmake --build build --config Release
# cmake --install build --config Release
In the Git Bash, clone the repository
# git clone https://github.com/gflags/gflags.git
Switch back to the Conda Prompt, and compile the gflags source code with cmake.
# cd gflags
# mkdir mybuild
# cd mybuild
# cmake .. -DCMAKE_EXPORT_COMPILE_COMMANDS=ON -DBUILD_SHARED_LIBS=ON -DCMAKE_POSITION_INDEPENDENT_CODE=ON -DCMAKE_CONFIGURATION_TYPES=Release -A x64 -T host=x64 -G "Visual Studio 16 2019" '-DCMAKE_INSTALL_PREFIX=C:\Program Files\gflag' -B build -S ../
# cmake --build build --config Release
# cmake --install build --config Release
# cd ../..
In the Git Bash, clone the repository
# git clone https://github.com/google/glog.git
Switch back to the Conda Prompt, and compile the glog source code with cmake.
# cd glog
# mkdir mybuild
# cd mybuild
# cmake .. -DCMAKE_EXPORT_COMPILE_COMMANDS=ON -DBUILD_SHARED_LIBS=ON -DCMAKE_POSITION_INDEPENDENT_CODE=ON -DCMAKE_CONFIGURATION_TYPES=Release -A x64 -T host=x64 -G "Visual Studio 16 2019" '-DCMAKE_INSTALL_PREFIX=C:\Program Files\glog' -B build -S ../
# cmake --build build --config Release
# cmake --install build --config Release
# cd ../..
All the dependencies on the Ryzen AI laptop are installed completely. User could run a end to end Yolov8 deplomyment progress with the following Section 4, which will start from the FP32 Yolov8 model. The whole progress will last for several hours or one day depending on the hardware computing ability.
Alternatively, user who wants a quick benchmark could skip Section 4 and start from Section 5 with pre-quantized model.
In this section, we will leverage the Vitis AI docker container on Linux GPU server for a quantized awared training(QAT).
Please follow the instrucion here to build your docker container or pull prebuild docker from docker hub.
This tutorial will take GPU docker as a reference.
$ git clone https://github.com/Xilinx/Vitis-AI.git
$ cd <Vitis-AI-Home>/docker
$ ./docker_build.sh -t gpu -f pytorchDownload the COCO dataset from https://cocodataset.org/#download following the instruction and make sure the dataset structure is restored as below. Please also update variable "DATA_PATH" in "coco.yaml" to point to the correct location.
+ datasets/
+ coco/
+ labels/
+ annotations/
+ images/
+ test-dev2017.txt
+ train2017.txt
+ val2017.txtEnvironment setup
$ cd <Vitis-AI-Home>
$ git clone https://github.com/fanz-xlnx/Yolov8-on-Ryzen-AI.git
$ ./docker_run.sh xilinx/vitis-ai-pytorch-gpu:<Your-Image-Tag>
$ cd Yolov8-on-Ryzen-AI
$ sudo bash env_setup.shUser could use the run_test.sh script to validate the float point model first before the quantization.
$ bash run_test.shThen Quantize the model with following script.
$ bash run_ptq.shThen quantize the model with QAT technique.
$ bash run_qat.shCopy the quantized model to Ryzen AI laptop for the following deployment.
If the section 4 is skiped, please start a Git Bash. In the Git Bash, clone the repository
# git clone https://github.com/fanz-xlnx/Yolov8-on-Ryzen-AI.git
Switch back to the Conda Prompt, and compile the Yolov8 source code.
# cd YOLOv8_RyzenAI_demo
# build.bat
The output will be generated as below.
......
-- Installing: C:/Users/ibane/Desktop/voe-win_amd64-with_xcompiler_on-c07e419-latest/bin/camera_yolov8.exe
-- Installing: C:/Users/ibane/Desktop/voe-win_amd64-with_xcompiler_on-c07e419-latest/bin/camera_yolov8_nx1x4.exe
-- Installing: C:/Users/ibane/Desktop/voe-win_amd64-with_xcompiler_on-c07e419-latest/bin/test_jpeg_yolov8.exe
To validate your setup, the following command will do the inference with single image.
# run_jpeg.bat DetectionModel_int.onnx sample_yolov5.jpg
The output will be generated as below.
result: 0 person 490.38498 85.79535 640.00488 475.18262 0.932453
result: 0 person 65.96048 97.76373 320.66068 473.83783 0.924142
result: 0 person 182.15485 306.91266 445.14795 475.26132 0.893309
result: 27 tie 584.48022 221.15732 632.27008 244.21243 0.851953
result: 27 tie 175.62622 224.15210 235.84900 248.83557 0.651355
To run with live camera, user needs to change the display and camera settings manually as below.
- Go to
Display settings, change Scale to 100% in theScale & layoutsection. - Go to
Bluetooth & devices->Cameras->USB2.0 FHD UVC WebCam, turn off the Background effects in theWindows Studio Effectssection.
camera_nx.bat
Possible options to run the yolov8 demo.
# camera_nx.bat -h
Options:
-c [parallel runs]: Specifies the (max) number of runs to invoke simultaneously. Default:1.
-s [input_stream] set input stream, E.g. set 0 to use default camera.
-x [intra_op_num_threads]: Sets the number of threads used to parallelize the execution within nodes, A value of 0 means ORT will pick a default. Must >=0.
-y [inter_op_num_threads]: Sets the number of threads used to parallelize the execution of the graph (across nodes), A value of 0 means ORT will pick a default. Must >=0.
-D [Disable thread spinning]: disable spinning entirely for thread owned by onnxruntime intra-op thread pool.
-Z [Force thread to stop spinning between runs]: disallow thread from spinning during runs to reduce cpu usage.
-T [Set intra op thread affinities]: Specify intra op thread affinity string.
[Example]: -T 1,2;3,4;5,6 or -T 1-2;3-4;5-6
Use semicolon to separate configuration between threads.
E.g. 1,2;3,4;5,6 specifies affinities for three threads, the first thread will be attached to the first and second logical processor.
-R [Set camera resolution]: Specify the camera resolution by string.
[Example]: -R 1280x720
Default:1920x1080.
-r [Set Display resolution]: Specify the display resolution by string.
[Example]: -r 1280x720
Default:1920x1080.
-L Print detection log when turning on.
-h: help
The MIT License (MIT)
Copyright (c) 2022 Advanced Micro Devices, Inc.
Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:
The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.
THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.