WIN10+RTX2070 tensorflow-gpu環境搭建

參考

tensorflow-windows-wheel 我選擇了：1.12.0\py37\GPU\cuda100cudnn73sse2

前置環境

• Visual Studio 2017
• Anaconda3

這兩個我都安裝的最新版。
Visual Studio 2017不裝會致使安裝CUDA時"Visual Studio Integration"組件安裝失敗，我在此處卡了幾天，查了好多資料都無效。

GPU相關軟件下載

• Nvidai驅動
• CUDA10.0： cuda-tookit下載， 選擇CUDA Toolkit 10.0
• cuDNN：cuDNN下載， 選擇Download cuDNN v7.3.1 (Sept 28, 2018), for CUDA 10.0
• tensorflow安裝包：tensorflow-windows-wheel

安裝

• 顯卡驅動安裝
• 安裝CUDA

安裝好以後的環境變量默認包含：
CUDA_PATH：C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0
CUDA_PATH_V10_0：C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0
PATH:
	C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\bin
	C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\libnvvp
	C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\extras\CUPTI\libx64(這個須要本身添加)

• 安裝cuDNN

解壓cuDNN後獲得bin、include、lib三個文件夾
將bin下的cudnn64_7.dll拷貝到：C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\bin
將include下的cudnn.h拷貝到：C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\include
將lib\x64下的cudnn.lib拷貝到：C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v10.0\lib\x64

tensorflow-gpu環境配置

• 打開
• 建立環境

  conda create -n tensorflow_gpu_1.12_py37 python=3.7

• 激活環境

  conda activate tensorflow_gpu_1.12_py37

• 安裝tensorflow-gpu

  pip install E:\Downloads\tensorflow_gpu-1.12.0-cp37-cp37m-win_amd64.whl

• 測試

  進入python
  	>>> import tensorflow as tf
  	>>> hello = tf.constant('Hello, tensorflow!')
  	>>> sess = tf.Session()

  輸出：

  2019-03-02 09:37:52.875467: I tensorflow/core/platform/cpu_feature_guard.cc:141] Your CPU supports instructions that this TensorFlow binary was not compiled to use: AVX AVX2
  	2019-03-02 09:37:53.200917: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1432] Found device 0 with properties:name: GeForce RTX 2070 major: 7 minor: 5 memoryClockRate(GHz): 1.62
  	pciBusID: 0000:01:00.0
  	totalMemory: 8.00GiB freeMemory: 6.59GiB
  	2019-03-02 09:37:53.207587: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1511] Adding visible gpu devices: 0
  	2019-03-02 09:40:52.845289: I tensorflow/core/common_runtime/gpu/gpu_device.cc:982] Device interconnect StreamExecutor with strength 1 edge matrix:
  	2019-03-02 09:40:52.848702: I tensorflow/core/common_runtime/gpu/gpu_device.cc:988]      0
  	2019-03-02 09:40:52.850105: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1001] 0:   N
  	2019-03-02 09:40:52.854219: I tensorflow/core/common_runtime/gpu/gpu_device.cc:1115] Created TensorFlow 
  	device (/job:localhost/replica:0/task:0/device:GPU:0 with 6331 MB memory) -> physical GPU (device: 0, 
  	name: GeForce RTX 2070, pci bus id: 0000:01:00.0, compute capability: 7.5)

  接着執行

  print(sess.run(hello))

  輸出：

  b'Hello, tensorflow!'

  關閉session

  sess.close()

至此結束，第一次感受開發環境搭建不容易。