Hi Daniel, We are planning to upload CUDA 7.5 recently, and this would fix the compatibility of nvcc and gcc-5 [1].
The hot-spot deeplearning frameworks: caffe, torch, mxnet, tensorflow, theano caffe is blocked by CUDA and the nvidia team is working on CUDA curretly, https://bugs.debian.org/788539 torch is blocked by luarocks issue and we are discussing the solution, https://bugs.debian.org/794634 tensorflow is blocked by protobuf and bazel, https://bugs.debian.org/804612 mxnet is likely not blocked. https://bugs.debian.org/808235 theano is maintained by you, thanks! [1] http://bugs.debian.org/807579 -- .''`. : :' : `. `' `-
