Check failed: error == cudaSuccess (8 vs. 0) invalid device function

[Lee Yoongu]

Hello all, I am asking the question because I did not find any solution for my issue. I am using Ubuntu 14.04, Cuda 7.5 and Cudnn 7.0, GTX 750 Ti and newest caffe. I installed correctly GPU by checking status "PASS" in the ./deviceQuery. I set USE_CUDNN:=1 and run make runtest and it has eror as follows. What is happen in my setting.  Could you help me to solve my issue? Thank all
       
       

[  FAILED  ] MathFunctionsTest/1.TestSgnbitGPU, where TypeParam = double (11776 ms)
[----------] 14 tests from MathFunctionsTest/1 (75631 ms total)

[----------] 1 test from HDF5OutputLayerTest/3, where TypeParam = caffe::DoubleGPU
[ RUN      ] HDF5OutputLayerTest/3.TestForward
[       OK ] HDF5OutputLayerTest/3.TestForward (2 ms)
[----------] 1 test from HDF5OutputLayerTest/3 (2 ms total)

[----------] 2 tests from SigmoidCrossEntropyLossLayerTest/3, where TypeParam = caffe::DoubleGPU
[ RUN      ] SigmoidCrossEntropyLossLayerTest/3.TestSigmoidCrossEntropyLoss
F0503 22:04:32.147617 27525 sigmoid_layer.cu:26] Check failed: error == cudaSuccess (8 vs. 0)  invalid device function
*** Check failure stack trace: ***
    @     0x2b55a3c9cdaa  (unknown)
    @     0x2b55a3c9cce4  (unknown)
    @     0x2b55a3c9c6e6  (unknown)
    @     0x2b55a3c9f687  (unknown)
    @           0xcfbeee  caffe::SigmoidLayer<>::Forward_gpu()
    @           0x8d28d1  caffe::Layer<>::Forward()
    @           0xceb4d4  caffe::SigmoidCrossEntropyLossLayer<>::Forward_gpu()
    @           0x8d28d1  caffe::Layer<>::Forward()
    @           0xa10b29  caffe::SigmoidCrossEntropyLossLayerTest<>::TestForward()
    @           0xa102d8  caffe::SigmoidCrossEntropyLossLayerTest_TestSigmoidCrossEntropyLoss_Test<>::TestBody()
    @           0xbb39a5  testing::internal::HandleSehExceptionsInMethodIfSupported<>()
    @           0xbaee36  testing::internal::HandleExceptionsInMethodIfSupported<>()
    @           0xb9c20b  testing::Test::Run()
    @           0xb9c9a4  testing::TestInfo::Run()
    @           0xb9cf92  testing::TestCase::Run()
    @           0xba1e1c  testing::internal::UnitTestImpl::RunAllTests()
    @           0xbb4d3a  testing::internal::HandleSehExceptionsInMethodIfSupported<>()
    @           0xbafb75  testing::internal::HandleExceptionsInMethodIfSupported<>()
    @           0xba0bb0  testing::UnitTest::Run()
    @           0xb972cf  main
    @     0x2b55a5d42ec5  (unknown)
    @           0x8be3b9  (unknown)
    @              (nil)  (unknown)
Aborted (core dumped)
make[3]: *** [src/caffe/test/CMakeFiles/runtest] Error 134
make[2]: *** [src/caffe/test/CMakeFiles/runtest.dir/all] Error 2
make[1]: *** [src/caffe/test/CMakeFiles/runtest.dir/rule] Error 2
make: *** [runtest] Error 2

[Jan]

I am not sure what you mean by cuDNN 7.0, the current (and only caffe-supported) version of cuDNN at the moment is cuDNN 4.

The error message means that some device function was not compiled for your architecture (you have compute capability 5.0). Make sure sm_50 and compute_50 is contained in the CUDA_ARCH variable in your Makefile.config.

Jan

[Lee Yoongu]

@Jan: Sorry. I was typo. It is Cudnn 5.0. I solved the issue. One thing is from your suggest, and other thing is from cudnn version. I need to use lower version of Cudnn as R1 which depends on the caffe version

Vào 23:25:41 UTC+9 Thứ Ba, ngày 03 tháng 5 năm 2016, Lee Yoongu đã viết:

Hello all, I am asking the question because I did not find any solution for my issue. I am using Ubuntu 14.04, Cuda 7.5 and Cudnn 7.0, GTX 750 Ti and newest caffe. I installed correctly GPU by checking status "PASS" in the ./deviceQuery. I set USE_CUDNN:=1 and run make runtest and it has eror as follows. What is happen in my setting.  Could you help me to solve my issue? Thank all
       
       

...