Windows 10 + NVIDIA GPU + TensorFlow

1. 參考 Install TensorFlow with GPU for Windows 10
2. 強烈建議：
• 確實檢查自己的NVIDIA顯示卡是否支援？官網
• Windows設定 %CUDA_HOME% 和添加CUDA相關目錄至%PATH%環境變數
• Anaconda 同時安裝 Python2 和 Python3 兩版本：下載連結
• 建立 Conda tensorflow-gpu 虛擬環境時，記得 Python3版本為 3.5
• conda create -n tensorflow-gpu python=3.5 (暫時別選3.6)
• activate tensorflow-gpu 後，安裝 numbapro 套件
• 開啟 jupyter notebook 後，執行下列 python3 code：參考網站

from numbapro import cuda, vectorize, guvectorize, check_cuda
from numbapro import void, uint8 , uint32, uint64, int32, int64, float32, float64, f8
import numpy as np
check_cuda()

• 確認執行結果是否類似下列結果，最後出現的是 True !：

------------------------------libraries detection-------------------------------
Finding cublas
 located at d:\Anaconda3\envs\tensorflow-gpu\DLLs\cublas64_70.dll
 trying to open library... ok
Finding cusparse
 located at d:\Anaconda3\envs\tensorflow-gpu\DLLs\cusparse64_70.dll
 trying to open library... ok
Finding cufft
 located at d:\Anaconda3\envs\tensorflow-gpu\DLLs\cufft64_70.dll
 trying to open library... ok
Finding curand
 located at d:\Anaconda3\envs\tensorflow-gpu\DLLs\curand64_70.dll
 trying to open library... ok
Finding nvvm
 located at d:\Anaconda3\envs\tensorflow-gpu\DLLs\nvvm64_30_0.dll
 trying to open library... ok
 finding libdevice for compute_20... ok
 finding libdevice for compute_30... ok
 finding libdevice for compute_35... ok
-------------------------------hardware detection-------------------------------
Found 2 CUDA devices
id 0    b'GeForce GTX 1060 6GB'                              [SUPPORTED]
                      compute capability: 6.1
                           pci device id: 0
                              pci bus id: 2
id 1      b'GeForce GTX 650'                              [SUPPORTED]
                      compute capability: 3.0
                           pci device id: 0
                              pci bus id: 1
Summary:
 2/2 devices are supported
PASSED
Out[1]:
True

3. 兩張顯示卡的GPU數量不同，另外在Windows環境設定TF_MIN_GPU_MULTIPROCESSOR_COUNT=2 錯誤訊息才消失

希望大家能少走點冤枉路...:P #