Win10 with CMake 3.5.2 and vs update1 compiling GPU version (Cuda 8.0, CUDNN v5 for Cuda 8.0) Open compile release and debug version with VS 2015 See the example on the net there are three inside the project Folders include (Include directories containing Mxnet,dmlc,mshadow)
Lib (contains Libmxnet.dll, libmxnet.lib, put it in vs. compiled)
Python (contains a mxnet,setup.py, and build, but the build contains the lib/mxnet, which is the same as the Python mxnet library, do not know is not redundant, oh, this may be installed after the)
The official release also adds a: 3rdparty (master above there is a 3rdparty, but for example CUDNN to update itself) Cudart also need to copy the past from the installation of Cuda/v8.0/bin, the following several files: ====== ============== and the root directory have a setupenv.cmd that sets the environment variable and license (copied from the original project) README.txt (some outdated)
Execute the environment variable settings at the end of the Python directory execute Python setup.py install.
But there is a problem, running on my Computer mlp_numpy example, GPU slower than CPU ...
From for notes (Wiz)
Win10 with CMake 3.5.2 and vs update1 compiling GPU version (Cuda 8.0, CUDNN v5 for Cuda 8.0)