2. The first thing to determine annaconda3 5.2.0 64 - bit (Python 3.6.5) corresponding tensorflow - the gpu version of the relationship, make sure you want to install annaconda relation with tensorflow - the gpu version, otherwise it will cause cannot successful use problems,
? Install TensorFlow - GPU 2.0.0 alpha version
PIP install -i https://pypi.tuna.tsinghua.edu.cn/simple tensorflow - gpu==2.0.0 a0
3. Determine TensorFlow - cuda gpu corresponding version
Through the baidu search found installation TensorFlow - 2.0 corresponding cuda10.0 gpu is more, temporarily not check the TensorFlow - version 2.1 of the cuda gpu,,,
? Cuda10.0 install
Emmm, cuda installation after installation cudnn even, at the same time to find corresponding cudnn cuda version oh
https://developer.nvidia.com/rdp/cudnn-archive
4. Check the whole environment can use
Use tensorflow==version 2.0.0 and TensorFlow1.14.0 are the following problems
When the import, the following problems
Question 1 - TensorFlow with numpy incompatibilities
Solution:
Question 2
Solution:
Tf. Should be after two '_', and he has been writing a '_' [face] qq: 67. GIF [/face] vomiting blood
Question 3 -- PIP upgrade
Solution:
- the user can add after each time the PIP install
Brief: PIP install - user
Or use the administrator run CMD also ok, this didn't try
Question 4 - numpy and h5py incompatibilities
Solution 4:
Run successfully
