nervanasystems/cuda-convnet2

Custom fork containing our own python backend for integration into neon

cuda-convnet2's Issues

Policy "CMP0054" is not known to this version of CMake.

This error occurs when running a build on Ubuntu 14.04, which has the latest version of cmake set at 2.8, whilst Policy CMP0054 was introduced in 3.1. Is it possible to check the cmake version and change the build steps appropriately?

can't run example using cudanet

I try to run the example from neon page using cudanet:
neon --gpu cudanet examples/convnet/i1k-alexnet-fp32.yaml
I have no problems with cuda in other frameworks (e.g., Torch) but here I get a following error :

WARNING:neon.util.persist:deserializing object from: examples/convnet/i1k-alexnet-fp32.yaml
WARNING:neon.datasets.imageset:Imageset initialized with dtype <type 'numpy.float32'>
2015-08-17 13:42:37,419 WARNING:neon - setting log level to: 20
Traceback (most recent call last):
File "/opt/anaconda/bin/neon", line 235, in
experiment, result, status = main()
File "/opt/anaconda/bin/neon", line 197, in main
device_id=args.device_id)
File "/opt/anaconda/lib/python2.7/site-packages/neon/backends/init.py", line 99, in gen_backend
raise RuntimeError("cudanet not found, can't run via GPU")
RuntimeError: cudanet not found, can't run via GPU

Could you be so kind to help me using your framework?

CMake Error at cconv2_ConvNet_generated_weights.cu.o.cmake

I have an odd issue, whereby my automated builds on the Docker Hub have been failing for a while. The first issue with the CMake policy (#2) was solved. I now get the following error:

CMake Error at cconv2_ConvNet_generated_weights.cu.o.cmake:269 (message):
Error generating file
/tmp/pip_build_root/cudanet/cudaconvnet/CMakeFiles/cconv2_ConvNet.dir/src/./cconv2_ConvNet_generated_weights.cu.o

However when building locally (all available computers have CUDA-capable GPUs) with my Dockerfile works fine.