1、https://github.com/fo40225/tensorflow-windows-wheel/tree/master/1.14.0/py37/GPU/cuda101cudnn76sse2
ZC: thlflow's whl download address
2、CUDA Toolkit 10.1 original Archive _ NVIDIA Developer.html(https://developer.nvidia.com/cuda-10.1-download-archive-base)
ZC: CUDA 10.1 download address
3、cuDNN Archive _ NVIDIA Developer.html(https://developer.nvidia.com/rdp/cudnn-archive)
ZC: cuDNN download address (requires account login to Nvidia)
4. Build from source code _ TensorFlow.html ( https://tensorflow.google.cn/install/source#tested_build_configurations )
ZC: There are some versions of the corresponding information, such as:
version | Python version | translater | Build tools | cuDNN | CUDA |
---|---|---|---|---|---|
tensorflow_gpu-1.14.0 | 2.7、3.3-3.7 | GCC 4.8 | Basel 0.24.1 | 7.4 | 10.0 |
ZC: But it seems not suitable for " tensorflow_gpu-1.14.0-cp37-cp37m-win_amd64.whl " ( https://github.com/fo40225/tensorflow-windows-wheel/tree/master/1.14.0/py37/GPU/ cuda101cudnn76sse2 ), from the name of the github folder, it should correspond to "cuda10.1.?" and "cuDNN7.6.?". Actually operated a bit, during the training phase tensorflow will report an error saying that cuda10.1 is needed but it cannot be found ...
5、