[Windows] Tensorflow & Keras 사전 준비

1. CUDA Toolkit 설치 (Base: 10.1)

2. cuDNN 설치 (Base: 10.1)

---

1. CUDA Toolkit 설치(10.1)

https://developer.nvidia.com/cuda-10.1-download-archive-update2?target_os=Windows&target_arch=x86_64&target_version=10&target_type=exelocal

CUDA Toolkit 10.1 update2 Archive

해당 URL에 들어간다.

아래와 같이 선택하고 Legacy Releases를 클릭한다.

Category	선택
Operating System	Windows
Architecture	x86_64
Version	10
Installer Type	exe(local)

 

CUDA Toolkit 10.1을 클릭 후 download 진행한다.(앞에 선택 참고)

 

추가로 CUDA Toolkit 10.0 도 받는다. (밑에 설명할 오류가 발생하기 때문이다.)

---

2. cuDNN (cudnn-10.1)설치하기

https://developer.nvidia.com/rdp/cudnn-archive

cuDNN Archive

 

로그인을 진행하면 download 진행된다.

---

3. 설치하기

1) cuda 설치하기

cuda_10.1.105_418.96_win10를 먼저 설치하고 cuda_10.0.130_411.31_win10.exe를 설치한다.

설치가 완료되면 다음 파일을 copy한다. (from v10.0 to v10.1)

v10.0	v10.1
cudart64_100.dll	copy 위치

2) cudnn 설치하기

cudnn-10.1-windows10-x64-v7.6.5.32.zip의 압축을 풀다.

bin, include, lib를 모두 선택한다.

cuda가 설치된 Floder에 붙여넣는다.(C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1)

---

4. 환경변수 설정

path

C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\bin
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\lib
C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.1\include

 

 

참고

https://www.tensorflow.org/install/gpu#software_requirements

GPU support  |  TensorFlow

 

 

 

---

 

5. GPU 사용 확인

 

설치 전

import tensorflow as tf

hello = tf.constant('hello')
sess = tf.Session()
print(sess.run(hello))

2020-02-01 16:16:40.688461: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
b'hello'

 

설치 후

import tensorflow as tf

hello = tf.constant('hello')
sess = tf.Session()
print(sess.run(hello))

 

---

※ 오류 1

C:\Users\mellowlee\python_project\venv\Scripts\python.exe C:/Users/mellowlee/Project_Python_Pycharm01/sample/util.py
2020-02-01 16:53:32.153697: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'cudart64_101.dll'; dlerror: cudart64_101.dll not found

 

solution

cudart64_101.dll이 없다는 것이고 cuda10.1 버전을 받으면 될것 같다.

1) 받아서 설치(cuda10.1 설치)

 

https://developer.nvidia.com/cuda-10.1-download-archive-base?target_os=Windows&target_arch=x86_64&target_version=10&target_type=exelocal

CUDA Toolkit 10.1 original Archive

2) cudart64_101.dll copy

 

 

참고

https://www.joe0.com/2019/10/19/how-resolve-tensorflow-2-0-error-could-not-load-dynamic-library-cudart64_100-dll-dlerror-cudart64_100-dll-not-found/

How to resolve TensorFlow 2.0 Error – Could not load dynamic library ‘cudart64_100.dll’ – Jozef Jarosciak Blog

 

오류 2

2020-02-01 18:14:31.298078: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1651] Ignoring visible gpu device (device: 0, name: GeForce GTX 760, pci bus id: 0000:01:00.0, compute capability: 3.0) with Cuda compute capability 3.0. The minimum required Cuda capability is 3.5.
2020-02-01 18:14:31.298599: I tensorflow/core/platform/cpu_feature_guard.cc:142] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX2
2020-02-01 18:14:31.299583: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1096] Device interconnect StreamExecutor with strength 1 edge matrix:
2020-02-01 18:14:31.299806: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1102]    

 

그래픽 카드의 성능이 떨어져서 tenforflow를 사용할 수 가 없다는 내용이다.

3.5 이상은되어야 사용가능하다.

그래픽 성능표

https://developer.nvidia.com/cuda-gpus

CUDA GPUs