End solve a problem on test import
>>> import tensorflow
2020-04-05 17:15:13.738389: W tensorflow/stream_executor/platform/default/dso_loader.cc:55] Could not load dynamic library 'cudart64_101.dll'; dlerror: cudart64_101.dll not found
2020-04-05 17:15:13.775424: I tensorflow/stream_executor/cuda/cudart_stub.cc:29] Ignore above cudart dlerror if you do not have a GPU set up on your machine.Internet Solutions https://blog.csdn.net/qq_41999081/article/details/104515513
It was found that I did not have C: \ Program Files \ NVIDIA GPU Computing Toolkit folder
Finally found the premise of the installation can be installed GPU are:
1, video card supports CUDA
(1) Right-click My Computer - Properties
(2) open Device Manager
https://developer.nvidia.com/cuda-gpus
Reference Links: https://www.cnblogs.com/jfdwd/p/11178704.html
I do not seem to support
Support it
2, pip version> = 8.1
View pip version: pip -V
update pip: python -m pip install -U pip
Anaconda installation
Installation anaconda, this in the previous blog had already been introduced in detail
(https://blog.csdn.net/qq_39295665/article/details/88206251)
Installation CUDA
(1) Enter the official website, https: //developer.nvidia.com/cuda-toolkit-archive
download the corresponding version of cuda, here to download the 8.0 version
(2) opening the installation package, store the path selection
(3) start extracting
(4) Compatibility test
(5) Installation Protocol
(6) select a custom installation location
Install CUDA-related items
(8) the installation is complete
(9) Testing
At the command line, type nvcc -V
Installation cudnn
(1) Enter the official website corresponding to the selected version download
https://developer.nvidia.com/cudnn
Select download
(2) register
(3) corresponding to the selected version
(4) decompress
Copy the .dll files are extracted to CUDA folder corresponding address
