When I use GPU, I get "Check failed: Check failed: status == CUDNN_STATUS_SUCCESS(1 vs. 0)". When I use CPU, I get :" Check failed: status == CUBLAS_STATUS_SUCCESS(1 vs. 0)"

[yuyadanyadan]

Hi, I use caffe to train for a regression issue. The input are images without labels, while the output are images. Where is the problem? and I don't know whether my network is OK!

[yuyadanyadan]

And my graphic card is GTX 960. It can support compute capability 3.0 and higher. My cudnn is 5.0, while my cuda is 8.0

[naibaf7]

Did you try with normal Caffe? This is OpenCL Caffe, and if you only use CUDA you won't get that many benefits from it anyways.

[yuyadanyadan]

Hi, I use normal caffe! Because I think you are familiar with caffe, I ask you! No one else can discuss it with me. I use GPU(include cuda and cudnn). I think that it may be my net! can you help me to watch my net? Thank you !

在 2016-09-12 20:26:18，"Fabian Tschopp" notifications@github.com 写道：

Did you try with normal Caffe? This is OpenCL Caffe, and if you only use CUDA you won't get that many benefits from it anyways.

— You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub, or mute the thread.

[naibaf7]

Sorry, I don't have time for doing non-bug user questions, and I am only responsible for the OpenCL code.

[yuyadanyadan]

Ok, thank you so much!