ubuntu18.04系统安装mmdeploy_v1.0.0

一、先安装基础环境 :

Step 0. 从官网下载安装conda环境(具体步骤略.)

自行安装,   此处不做重点(略过)

Step 1. 创建一个conda虚拟环境,并进入该环境下.

conda create --name mmdeploy_v1.0.0 python=3.8 -y

source activate mmdeploy_v1.0.0 或者 conda  activate mmdeploy_v1.0.0

Step 2. 安装pytorch(官网地址: Previous PyTorch Versions | PyTorch).

具体版本依据自己的cuda版本, 本人cuda11.3, 或者自行输入 nvcc -V 命令查看自己的cuda

版本.

pip install torch==1.11.0+cu113 torchvision==0.12.0+cu113 torchaudio==0.11.0 --extra-index-url https://download.pytorch.org/whl/cu113

 至此基础环境安装完毕.

二、安装MMDeploy和MMDet环境 :

Step 0. 安装mmcv:

pip install -U openmim
mim install mmengine
mim install "mmcv>=2.0.0rc2"

Step 1.  安装 MMDeploy 推理引擎(Linux-x86_64架构下的):

# 1. install MMDeploy model converter
pip install mmdeploy==1.0.0

# 2. install MMDeploy sdk inference, support onnxruntime-gpu, tensorrt
pip install mmdeploy-runtime-gpu==1.0.0

# 3. 安装tensorrt的python接口
pip install TensorRT-8.2.3.0/python/tensorrt-8.2.3.0-cp38-none-linux_x86_64.whl
pip install pycuda

# 配置环境变量
vim ~/.bashrc
# 下载tensorrt, cudnn 解压后添加下面内容
# tensorrt的lib库
export TENSORRT_DIR=$(pwd)/TensorRT-8.2.3.0
export LD_LIBRARY_PATH=${TENSORRT_DIR}/lib:$LD_LIBRARY_PATH
# cudnn的lib库
export CUDNN_DIR=$(pwd)/cuda
export LD_LIBRARY_PATH=$CUDNN_DIR/lib64:$LD_LIBRARY_PATH
#添加完成后执行
source  ~/.bashrc

# 3.2 install ONNX Runtime, onnxruntime-gpu
pip install onnxruntime-gpu==1.8.1
wget https://github.com/microsoft/onnxruntime/releases/download/v1.8.1/onnxruntime-linux-x64-gpu-1.8.1.tgz
tar -zxvf onnxruntime-linux-x64-gpu-1.8.1.tgz
# 配置环境变量
export ONNXRUNTIME_DIR=$(pwd)/onnxruntime-linux-x64-gpu-1.8.1
export LD_LIBRARY_PATH=$ONNXRUNTIME_DIR/lib:$LD_LIBRARY_PATH
#添加完成后执行
source  ~/.bashrc

二、下载MMDeploy和MMDet代码 :

① 下载mmdeploy v1.0.0 版本的代码

git clone -b v1.0.0 https://github.com/open-mmlab/mmdeploy.git

② 下载mmdet v3.0的代码

git clone -b v3.0 https://github.com/open-mmlab/mmdetection.git

cd mmdetection-3.0.0
mim install -v -e .
cd ..

三、下载模型测试 :

① 转换  faster-rcnn 模型到tensorrt加速测试:

wget -P checkpoints https://download.openmmlab.com/mmdetection/v2.0/faster_rcnn/faster_rcnn_r50_fpn_1x_coco/faster_rcnn_r50_fpn_1x_coco_20200130-047c8118.pth

python mmdeploy-1.0.0/tools/deploy.py \
    mmdeploy-1.0.0/configs/mmdet/detection/detection_tensorrt_dynamic-320x320-1344x1344.py \
    mmdetection-3.0.0/configs/faster_rcnn/faster-rcnn_r50_fpn_1x_coco.py \
    mmdetection-3.0.0/checkpoints/faster_rcnn/faster_rcnn_r50_fpn_1x_coco_20200130-047c8118.pth \
    mmdetection-3.0.0/demo/demo.jpg \
    --work-dir mmdeploy_model/faster-rcnn \
    --device cuda \
    --dump-info

②转换RTMDet 到tensorrt测试:

python mmdeploy-1.0.0/tools/deploy.py \
  mmdeploy-1.0.0/configs/mmdet/detection/detection_tensorrt_static-640x640.py \
  mmdetection-3.0.0/configs/rtmdet/rtmdet_s_8xb32-300e_coco.py \
  mmdetection-3.0.0/checkpoints/rtmdet_s_8xb32-300e_coco_20220905_161602-387a891e.pth \
  mmdetection-3.0.0/demo/demo.jpg \
  --work-dir ./work_dirs/rtmdet \
  --device cuda:0 \
  --show

等待转换成功后的结果!