Example Docker does not work (crashes)

[michaelschwier]

After I couldn't get the notebook to work on my own caffe installation (see #29) I tried running your prepared docker. However this line

# Load network
net1 = caffe.Net(STEP1_DEPLOY_PROTOTXT, STEP1_MODEL_WEIGHTS, caffe.TEST)

throws the following error

...
F0509 19:58:49.805807    24 cudnn_relu_layer.cpp:15] Check failed: status == CUDNN_STATUS_SUCCESS (1 vs. 0)  CUDNN_STATUS_NOT_INITIALIZED

My host system is ubuntu 18.04, NVIDIA Driver Version 390.48, CUDA Version 9.1.85 (if that is any help) caffe is set to CPU mode in your example jupyter notebook

Any help would be appreciated.

[michaelschwier]

However, tha main issue is actually that the model does not seem to work on the current official Caffe version (see #29).

[mohamed-ezz]

Hi @michaelschwier , I think we used a cropping layer which is different from what original Caffe currently has, and it has a slightly different behavior.

The error you get when using the Docker image is most probably some Nvidia Driver and CUDA compatibility issue. The docker image uses an older CUDA, like 6.5 or 7.

Have you tried to install an older Nvidia driver ? for example Nvidia driver 367.57, it is compatible with Titan X, Titan X Pascal and Titan Xp and various other Nvidia GPU models but you have to check if your GPU is supported.

[mohamed-ezz]

@michaelschwier this comment highlights the difference between Jonlong's caffe version (the one we used) and the original caffe version : https://github.com/IBBM/Cascaded-FCN/issues/3#issuecomment-272343778

[michaelschwier]

Hi @mohamed-ezz, thank you for your answer. I gave up on the Nvidia Docker, but managed to get a CPU-only version running buy building Jonlong's caffe version. Thank you for the hint to #3.

[mohamed-ezz]

@michaelschwier So is the problem solved yet ?

[michaelschwier]

@mohamed-ezz I'd consider it solved for me, the CPU version is enough for my purposes. Thanks!