Use NVIDIA Docker:
nvidia-docker run -it kaixhin/cuda-caffe.
For more information on CUDA on Docker, see the repo readme.
If you find this useful in research please consider citing this work.
So cool, but I found that in 'kaixhin/cuda-caffe:8.0', the result of 'cat /etc/issue' is 'Ubuntu 16.04.2 LTS', actually not ubuntu-14.04 as 'kaixhin/cuda-caffe:7.5'.
Thanks for your excellent work. When i run this image(cuda-caffe:8.0) in the docker, I find that there is no space left in /.
root@0e0658ba9fe1:~/caffe# df -h
Filesystem Size Used Avail Use% Mounted on
none 48G 45G 0 100% /
tmpfs 16G 0 16G 0% /dev
tmpfs 16G 0 16G 0% /sys/fs/cgroup
/dev/sda1 48G 45G 0 100% /etc/hosts
shm 64M 0 64M 0% /dev/shm
tmpfs 16G 0 16G 0% /sys/firmware
Can you give me some advice? Thanks a lot.
@yanleirex You will need to make sure that you have the right driver version for the tagged image - more details can be found in the NVIDIA Docker project.
Check failed: error == cudaSuccess (35 vs. 0) CUDA driver version is insufficient for CUDA runtime version
@ezamir Builds were failing around a month ago, so the issue might be fixed with one of the latest versions. Some people have had issues with the CUDA 7.5 driver version and EC2, so this might also be your problem.
@chanmees2k14 You shouldn't need a GPU to run this image, but you should use the CPU-only version instead (kaixhin/caffe).
Good Work. Thanks. My petty doubt is :
Am using intel i5 core CPU using docker.Do I need GPU(Physical Hardware) to run this CUDA-CAFFE? please update me earliest.
Can't for the life of me get this to work on EC2. I've tried so many different "recipes" for installing CUDA too. When I run caffe.set_mode_gpu() I just get errors telling me CUBLAS and CUBLAND(?) don't work.