error == cudaSucess(30 vs. 0)

65 views
Skip to first unread message

Teddy Ma

unread,
Apr 1, 2016, 4:11:40 AM4/1/16
to Caffe Users
I meet this error when I use python-caffe to extract feature from images.
I use Tesla M40,with the env :CUDA7.5 ,nvidia driver: 352.79.I don't think these will cause this error.And I also add "-gencode -arch=compute_52,code=sm_52"to my Makefile.config. Then "make clean","make all ","make test", "make runtest " ,"make pycaffe"all go well.
This error occured when my program process 300s images or more, and then I check my gpu by using "nvidia-smi", shows my "GPU is lost".
I think it might be the driver's problem.So I update my driver from 352.39 to 352.79 ,but  it still occured  
So ,what should I do to solve this problem
Reply all
Reply to author
Forward
0 new messages