hipCaffe
hipCaffe copied to clipboard
miopenStatusUnknownError
On a fresh install with Ubuntu 16.04 and rocm 1.7 Haswell 5820k, Vega FE. Similar to #26 during test_all.testbin Up until the error, many tests pass and many tests fail with "which exceeds threshold_" etc For some reason, the train_lenet works fine, and the accuracy is between 98 - 99 % Interestingly his error is in softmaxocl.cpp:59 and mine is /data/repo/MIOpen/src/ocl/activ_ocl.cpp:47
MIOpen Error: /data/repo/MIOpen/src/ocl/activ_ocl.cpp:47: Only alpha=1 and beta=0 is supported F0204 17:32:58.538528 20810 cudnn_tanh_layer_hip.cpp:23] Check failed: status == miopenStatusSuccess (7 vs. 0) miopenStatusUnknownError *** Check failure stack trace: *** @ 0x7f55c3fd95cd google::LogMessage::Fail() @ 0x7f55c3fdb433 google::LogMessage::SendToLog() @ 0x7f55c3fd915b google::LogMessage::Flush() @ 0x7f55c3fdbe1e google::LogMessageFatal::~LogMessageFatal() @ 0x154768e caffe::CuDNNTanHLayer<>::Forward_gpu() @ 0x4f7787 caffe::Layer<>::Forward() @ 0x1b39bb7 caffe::Net<>::ForwardFromTo() @ 0x1c3a5ba caffe::RecurrentLayer<>::Forward_gpu() @ 0x4f7787 caffe::Layer<>::Forward() @ 0x69a982 caffe::RNNLayerTest_TestForward_Test<>::TestBody() @ 0x108f334 testing::internal::HandleExceptionsInMethodIfSupported<>() @ 0x108f1f6 testing::Test::Run() @ 0x1090341 testing::TestInfo::Run() @ 0x1090b97 testing::TestCase::Run() @ 0x1097277 testing::internal::UnitTestImpl::RunAllTests() @ 0x1096cb4 testing::internal::HandleExceptionsInMethodIfSupported<>() @ 0x1096c69 testing::UnitTest::Run() @ 0x2006a7a main @ 0x7f55bf494830 __libc_start_main @ 0x2005f19 _start @ (nil) (unknown) Aborted (core dumped)
I see this same error mentioning activ_ocl.cpp
, but train_lenet
does not work for me (segfault) on either a from-scratch ROCm setup or a ROCm docker container.
@acowley - Thanks for the heads-up regarding the segfault. Please open a new ticket, as that appears to be a separate issue.