Okay. I found a “permanent” solution - feel free to pinch your nose now - by hardcoding two sysenv vars:
INCLUDE="C:\Program Files (x86)\Windows Kits0\Include0.0.10240.0\ucrt"
LIB="C:\Program Files (x86)\Windows Kits0\Lib0.0.10240.0\um\x64;C:\Program Files (x86)\Windows Kits0\Lib0.0.10240.0\ucrt\x64"
Yes, I know…
Gory details for my Keras1.1.0 + Theano0.8.2 + VS2015 + CUDA8.0 + cuDNN5.1 setup for Windows 10, here:
Tested on the following hardware:
- Dell Precision T7900, 64GB RAM [Intel Xeon E5-2630 v4 @ 2.20 GHz (1 processor, 10 cores total, 20 logical processors)]
- NVIDIA GeForce Titan X, 12GB RAM [Driver version: 372.90 / Win 10 64]
Uses the following tools/libraries:
- Visual Studio 2015 Community Edition Update 3 w. Windows Kit 10.0.10240.0 [Used for its C/C++ compiler (not its IDE) and SDK]
- CUDA 8.0.44 (64-bit) [Used for its GPU math libraries, card driver, and CUDA compiler]
- MinGW-w64 (5.4.0) [Used for its Unix-like compiler and build tools (g++/gcc, make…) for Windows]
- Anaconda (64-bit) w. Python 2.7 (Anaconda2-4.2.0) [A Python distro that gives us NumPy, SciPy, and other scientific libraries]
- Theano 0.8.2 [Used to evaluate mathematical expressions on multi-dimensional arrays]
- Keras 1.1.0 [Used for deep learning on top of Theano]
- OpenBLAS 0.2.14 (Optional) [Used for its CPU-optimized implementation of many linear algebra operations]
- cuDNN v5.1 (August 10, 2016) for CUDA 8.0 (Recommended) [Used to run vastly faster convolution neural networks]
Hope this helps!