Marvin is a GPU-only neural network framework made with simplicity, hackability, speed, memory consumption, and high dimensional data in mind.
This fork was modified to be compilable on Windows + Visual Studio 2013, and enable integration inside other projects (latter still WIP)
Download CUDA 7.5 and cuDNN 5.
To enable CUDA:
- Open VS2013 Express
- Click on New Project...
- Select Visual C++, then General, then Empty Project
- At this point you should have a new project open. Right click on the project (name) and select Build Dependencies... Build Customizations...
- A new dialog opens, select the CUDA 7.5 (.targets,.props) checkbox, and click OK
- In your project explorer window, right click on marvin.cu file, and select CUDA C++, then make any changes (for example, you may want to change the build target from sm_20 to something else...)
- Change project properties C++/Additional Include Directories and add cuDNN include dir
- Change project properties Linker/Additional Library Directories and add cuDNN lib dir and CUDA 7.5 lib dir
- Change project properties Linker/Input/Additional Dependencies and add cudart_static.lib;cudnn.lib;cublas.lib;curand.lib;
- Add appropriate source code to your Source.cu file
- Build...
Refer to the original base repo at https://github.com/PrincetonVision/marvin for everything else :D