Ubuntu 18.04 설치 및 환경 셋팅 (RTX 3090, 2080) 및 기타 Tips

Ubuntu 18.04 설치

1, 부팅시 Delete 연타

2. advanced mode(F6) -> 컴퓨터마다 다름

3. Boot

4. Boot Option -> Usb 설정

5. Exit

6. 가만히 냅두기

7. install ubuntu

8. continue

9. Installation type에서 erase... 선택

10. 한글 설정

---

chrome 설치하기: https://webnautes.tistory.com/1184

---

환경 셋팅 (RTX 3090) - 465버전

3090 Driver - https://velog.io/@cychoi74/%EC%9A%B0%EB%B6%84%ED%88%AC-18.04-NVIDIA-%EB%93%9C%EB%9D%BC%EC%9D%B4%EB%B2%84-%EC%84%A4%EC%B9%98

 

Cuda 11.1

cuDNN 8.0.4

TensorRT 7.2.2.3

 

CUDA 11.0, cudnn 11.x: https://cafepurple.tistory.com/39

 

tensorRT: https://eehoeskrap.tistory.com/414

TAR File

tensorRT를 /usr/local/cuda-xxxx/ 에 옮겨주기

 

---

.bashrc 맨 마지막 경로 추가:

export CUDA_HOME=/usr/local/cuda

export TRT_HOME=${CUDA_HOME}/TensorRT-7.2.3.4

export LD_LIBRARY_PATH=${CUDA_HOME}/lib64:${TRT_HOME}/lib

PATH=${CUDA_HOME}/bin:${PATH}

export PATH

 

or

 

export CUDA_HOME=/usr/local/cuda
export TRT_HOME=${CUDA_HOME}/TensorRT-7.2.2.3
export LD_LIBRARY_PATH=${CUDA_HOME}/lib64:${TRT_HOME}/lib:/home/hiw/DEV4/lib
export PATH=${CUDA_HOME}/bin:${PATH}

---

opencv: https://webnautes.tistory.com/1186

우리가 쓰는 opencv는 3.4.5

opencv build(3.4.5)

• wget https://raw.githubusercontent.com/milq/milq/master/scripts/bash/install-opencv.sh

• wget https://soynet.io/demo/opencv_345_install.sh

• sudo bash ./opencv_345_install.sh

 

cmake-gui:

where is the source code: /home/hiw/opencv/opencv-3.4.5

where to build the binaries: /home/hiw/opencv/opencv-3.4.5/build

1. configure

2. world 검색, 체크

3. modul 검색, 경로 opencv-contrib-3.4.5/modules 설정

4. generate

5. 명령창에 'make -j(숫자)'

 

Opencv Package

sudo apt-get install build-essential cmake git unzip pkg-config libjpeg-dev libpng-dev libtiff-dev libavcodec-dev libavformat-dev libswscale-dev libgtk2.0-dev libcanberra-gtk* python3-dev python3-numpy python3-pip libxvidcore-dev libx264-dev libgtk-3-dev libtbb2 libtbb-dev libdc1394-22-dev gstreamer1.0-tools libv4l-dev v4l-utils libgstreamer1.0-dev libgstreamer-plugins-base1.0-dev libavresample-dev libvorbis-dev libxine2-dev libfaac-dev libmp3lame-dev libtheora-dev libopencore-amrnb-dev libopencore-amrwb-dev libopenblas-dev libatlas-base-dev libblas-dev liblapack-dev libeigen3-dev  libhdf5-dev protobuf-compiler libprotobuf-dev libgoogle-glog-dev libgflags-dev gfortran libtiff5-dev mesa-utils libgl1-mesa-dri libgtkgl2.0-dev libgtkglext1-dev python2.7-dev python-numpy -y

 

 

 

 

https://www.notion.so/SoyNet-Demo-1434f814c340484192f30f1f742035a9

 

 

---

RTX-2080

• 그래픽 드라이버: 460.89
• Cuda: 10.1
• Cudnn v7.6.5 for cuda 10.1
• Tensorflow-gpu 2.3.1
• Keras 2.4.3

 

 

RTX-3060

• 그래픽 드라이버: 516.40
• Cuda: 11.2
• Cudnn v8.1.0 for cuda 11.2
• Tensorflow-gpu 2.9.1
• Keras 2.4.3

 

GPU별 지원 CUDA 버전 확인하기

https://mickael-k.tistory.com/18

 

※ tensorflow gpu 사용가능 확인

from tensorflow.python.client import device_lib

device_lib.list_local_devices()

---

 

※ IDE 주석 관련 설정 ※

Visual Studio - 편집,선택영역 확장    편집.

Pycharm - Clone Caret Above, Below      Comment with Line Comment

 

 

※ Windows 같은 경우는 Visual Studio 먼저 설치하고 CUDA를 설치해줘야 Visual Studio에서 CUDA를 잡는다. ※

 

※ Visual Studio Linker -> Input ※

 - cudart.lib

 - cublas.lib

 - cuda.lib

 - cudnn.lib