Running DeepDream, causes to the error error == cudaSuccess (4 vs. 0) unspecified launch failure

291 views
Skip to first unread message

Hossein Hasanpour

unread,
Nov 27, 2016, 5:05:27 AM11/27/16
to Caffe Users

I'm trying to run Google's DeepDream on my CIFAR10 pretrained models.
The script can be found here .
The issue is recently I tried to run the script, to my surprise, halfway the operation, the whole thing crashes with this error message :

Check failed: error == cudaSuccess (4 vs. 0)  unspecified launch failure

I used to run the scrip without any issues, with different input image sizes. but now, only the CPU mode runs well, if I switch to GPU, it crashes!

I even re-installed CUDA, used newer version of cudnn (from 4 to cudnn5.1) and even re-installed the whole Win10 Os to no luck.

I tried using exceptions, but I get nothing, ! I dont know where this stems from. 

I run the scripts with iter_n=500 and octave_n=4 , and input images have varting sizes, from 32x32 to 1024x1024, they all fail halfway (some at the very begining! and some others at the very end!)

Any help is greatly appreciated. 

Hossein Hasanpour

unread,
Nov 27, 2016, 7:53:03 AM11/27/16
to Caffe Users
by the way The issue happens at math_functions.cu:79 which is here :
void caffe_gpu_memcpy(const size_t N, const void* X, void* Y) {
 
if (X != Y) {
 CUDA_CHECK
(cudaMemcpy(Y, X, N, cudaMemcpyDefault)); // NOLINT(caffe/alt_fn)
 
}
 
}
what could be the cause here? and How can I avoid it? any idea?

sayad yaghoobi

unread,
Apr 24, 2017, 8:35:29 AM4/24/17
to Caffe Users
i have this issue with caffe if you could solve the problem please share your way :

I0424 17:01:31.953949  4402 caffe.cpp:218] Using GPUs 0
I0424 17:01:31.956274  4402 caffe.cpp:223] GPU 0: ����
F0424 17:01:31.956310  4402 common.cpp:152] Check failed: error == cudaSuccess (30 vs. 0)  unknown error
*** Check failure stack trace: ***
    @     0x7f1e796995cd  google::LogMessage::Fail()
    @     0x7f1e7969b433  google::LogMessage::SendToLog()
    @     0x7f1e7969915b  google::LogMessage::Flush()
    @     0x7f1e7969be1e  google::LogMessageFatal::~LogMessageFatal()
    @     0x7f1e79e62012  caffe::Caffe::SetDevice()
    @           0x40b018  train()
    @           0x4072f0  main
    @     0x7f1e7860a830  __libc_start_main
    @           0x407b19  _start
    @              (nil)  (unknown)
Aborted (core dumped)

Reply all
Reply to author
Forward
0 new messages