Getting "invalid device ordinal" when running `make runtest`

[Matthew Drill]

Hi, I just downloaded caffe, and was able to run

make; make run

without any problems, but now when I test it with

make runtest

I get this error:

F0328 17:54:31.815686 31058 pooling_layer.cu:212] Check failed: error == cudaSuccess (10 vs. 0)  invalid device ordinal
*** Check failure stack trace: ***
    @     0x7f1a1c09a26d  google::LogMessage::Fail()
    @     0x7f1a1c09c083  google::LogMessage::SendToLog()
    @     0x7f1a1c099dfb  google::LogMessage::Flush()
    @     0x7f1a1c09ca6e  google::LogMessageFatal::~LogMessageFatal()
    @     0x7f1a19810bd1  caffe::PoolingLayer<>::Forward_gpu()
    @           0x479c56  caffe::Layer<>::Forward()
    @           0x6a5bb5  caffe::PoolingLayerTest_TestForwardAve_Test<>::TestBody()
    @           0x91a693  testing::internal::HandleExceptionsInMethodIfSupported<>()
    @           0x913caa  testing::Test::Run()
    @           0x913df8  testing::TestInfo::Run()
    @           0x913ed5  testing::TestCase::Run()
    @           0x9151af  testing::internal::UnitTestImpl::RunAllTests()
    @           0x9154d3  testing::UnitTest::Run()
    @           0x46dd6d  main
    @     0x7f1a1896e830  __libc_start_main
    @           0x4757d9  _start
Makefile:523: recipe for target 'runtest' failed
make: *** [runtest] Aborted (core dumped)

The only thing I could find about this error said that it means that CUDA doesn't recognize your GPU, or you are using the wrong GPU index, but when I run

deviceQuery

in my cuda folder, It returns all the information about my gpu, so cuda is installed and recognizing my gpu

I'm running Ubuntu 16.04, CUDA 8.0, and cudNN 5.1. Does anyone have any idea how to fix this?

Thank you!

[Antonimus]

I have the same problem, where you able to solbe it ?

[Przemek D]

Is your Caffe the latest version? Get the most recent one by cloning the repository, then post the build log (better: attach it to the post). A full runtest log would be useful too.