HDF5DataLayerTest: Failed to open source file: src/caffe/test/test_data/sample_data_list.txt

[BeamOfLight]

Issue summary

Hello!

I have problems with tests:

#./test_all.testbin
...
[ RUN      ] DeconvolutionLayerTest/3.TestNDAgainst2D
src/caffe/test/test_deconvolution_layer.cpp:276: Failure
Value of: 1
Expected: 0
[  FAILED  ] DeconvolutionLayerTest/3.TestNDAgainst2D, where TypeParam = caffe::GPUDevice<double> (1 ms)
[ RUN      ] DeconvolutionLayerTest/3.TestGradient3D
src/caffe/test/test_deconvolution_layer.cpp:309: Failure
Value of: 1
Expected: 0
[  FAILED  ] DeconvolutionLayerTest/3.TestGradient3D, where TypeParam = caffe::GPUDevice<double> (0 ms)
[----------] 5 tests from DeconvolutionLayerTest/3 (24602 ms total)

[----------] 1 test from HDF5DataLayerTest/0, where TypeParam = caffe::CPUDevice<float>
[ RUN      ] HDF5DataLayerTest/0.TestRead
F0818 17:11:29.112534 13231 hdf5_data_layer.cpp:88] Failed to open source file: src/caffe/test/test_data/sample_data_list.txt
*** Check failure stack trace: ***
    @     0x7f4602bbe5cd  google::LogMessage::Fail()
    @     0x7f4602bc0433  google::LogMessage::SendToLog()
    @     0x7f4602bbe15b  google::LogMessage::Flush()
    @     0x7f4602bc0e1e  google::LogMessageFatal::~LogMessageFatal()
    @           0xd1c885  caffe::HDF5DataLayer<>::LayerSetUp()
    @           0x5e955a  caffe::HDF5DataLayerTest_TestRead_Test<>::TestBody()
    @           0xb43964  testing::internal::HandleExceptionsInMethodIfSupported<>()
    @           0xb43825  testing::Test::Run()
    @           0xb449d0  testing::TestInfo::Run()
    @           0xb45217  testing::TestCase::Run()
    @           0xb4b5f7  testing::internal::UnitTestImpl::RunAllTests()
    @           0xb4b044  testing::internal::HandleExceptionsInMethodIfSupported<>()
    @           0xb4aff9  testing::UnitTest::Run()
    @           0xf8063a  main
    @     0x7f45fca8f830  __libc_start_main
    @           0xf73579  _start
    @              (nil)  (unknown)
Aborted

How can i fix it?

Steps to reproduce

Freshly compiled hipCaffe with Makefile.config parameters: USE_MIOPEN := 1 USE_ROCBLAS := 0 OPENCV_VERSION := 3 I am using the latest ROCm 1.6.127 from debian packages. Running the test_all.testbin show fails.

Your system configuration

Operating system: Ubuntu 16.04.2 LTS with 4.11.0-kfd-compute-rocm-rel-1.6-127 Compiler: GCC v5.4.0, HCC clang 5.0 CUDA version (if applicable): not applicable CUDNN version (if applicable): not applicable BLAS: ATLAS Python or MATLAB version (for pycaffe and matcaffe respectively): not applicable

[parallelo]

Hello @BeamOfLight,

Thanks for the heads-up. Sorry we haven't replied more quickly -- we've been working on some interesting framework updates.

Regarding the issue you reported, we'll take a look at our internal tests and see what we can find out.

Best,

Jeff

[parallelo]

Hello @BeamOfLight,

Some more details: We are aware of some failures in the unit tests, so this should be expected to some extent while we continue to enhance the ML frameworks.

However, in the immediate-term, we try to ensure two key outcomes for these unit tests:

1. Avoid regressions
2. Avoid all fatal aborts

Can you please re-test the unit test that caused the abort? (It works on my set up)

# ./build/test/test_all.testbin --gtest_filter=HDF5DataLayerTest/0.TestRead
...
[==========] Running 1 test from 1 test case.
[----------] Global test environment set-up.
[----------] 1 test from HDF5DataLayerTest/0, where TypeParam = caffe::CPUDevice<float>
[ RUN      ] HDF5DataLayerTest/0.TestRead
[       OK ] HDF5DataLayerTest/0.TestRead (6 ms)
[----------] 1 test from HDF5DataLayerTest/0 (6 ms total)

[----------] Global test environment tear-down
[==========] 1 test from 1 test case ran. (6 ms total)
[  PASSED  ] 1 test.

[tassilostefan]

I was having the same problem on a Win7Prof 64 machine. After searching in vain for a long time, I just tried whether it is a path problem...and it seemingly is: The sample_data_list.txt is under C:\Users\User\Downloads\Caffe\src\caffe\test\test_data The test exe (build with VS2013) is under C:\Users\User\build\test\Debug\

I was using, as usual, the C:\Users\User\build dir to start testing. With the result above mentioned. After working my way up from the C:\Users\User\Downloads\Caffe\src\caffe\test\ dir, I found that it works, when calling from the C:\Users\User\Downloads\Caffe dir, like this:

C:\Users\User\Downloads\Caffe>C:\Users\User\build\test\Debug\test.testbin-d.exe --gtest_filter=HDF5DataLayerTest/0.TestR
ead
Note: Google Test filter = HDF5DataLayerTest/0.TestRead
[==========] Running 1 test from 1 test case.
[----------] Global test environment set-up.
[----------] 1 test from HDF5DataLayerTest/0, where TypeParam = struct caffe::CPUDevice<float>
[ RUN      ] HDF5DataLayerTest/0.TestRead
[       OK ] HDF5DataLayerTest/0.TestRead (801 ms)
[----------] 1 test from HDF5DataLayerTest/0 (801 ms total)

[----------] Global test environment tear-down
[==========] 1 test from 1 test case ran. (817 ms total)
[  PASSED  ] 1 test.